Retention Secured Nonlinear and Self-Rectifying Analog Charge Trap Memristor for Energy-Efficient Neuromorphic Hardware
Open accessThe DOI matches its Crossref record; 44 of the 45 references it cites were found in a registry or an index; 3 of the 34 sources its claims cite were checked for support.
Abstract
A memristive crossbar array (MCA) is an ideal platform for emerging memory and neuromorphic hardware due to its high bitwise density capability. A charge trap memristor (CTM) is an attractive candidate for the memristor cell of the MCA, because the embodied rectifying characteristic frees it from the sneak current issue. Although the potential of the CTM devices has been suggested, their practical viability needs to be further proved. Here, a Pt/Ta 2 O 5 /Nb 2 O 5‐ x /Al 2 O 3‐ y /Ti CTM stack exhibiting high retention and array‐level uniformity is proposed, allowing a highly reliable selector‐less MCA. It shows high self‐rectifying and nonlinear current‐voltage characteristics below 1 µA of programming current with a continuous analog switching behavior. Also, its retention is longer than 10 5 s at 150 °C, suggesting the device is highly stable for non‐volatile analog applications. A plausible band diagram model is proposed based on the electronic spectroscopy results and conduction mechanism analysis. The self‐rectifying and nonlinear characteristics allow reducing the on‐chip training energy consumption by 71% for the MNIST dataset training task with an optimized programming scheme.
Overview
CoChat Summary
## TL;DR
The paper proposes a Pt/Ta2O5/Nb2O5-x/Al2O3-y/Ti charge trap memristor that combines self-rectification, nonlinear analog switching, and high retention for selector-less crossbar arrays. In a 32×32 array and MNIST simulation, the device enabled an estimated 71% energy reduction under an optimized programming scheme and about 91% recognition accuracy.
## Key findings
- The PTNAT device shows strong rectifying and nonlinear behavior, with a maximum on/off ratio of ≈8 × 10^3 at 4.5 V and a maximum rectifying current ratio of ≈5.7 × 10^4 at 7.5 V.
- The device operates with analog switching below 1 µA programming current, is electroforming-free, and shows endurance up to 10^5 cycles.
- Retention is reported as longer than 10^5 s at 150 °C in the abstract, and the conclusion states >2 × 10^5 s at 150 °C; the body also reports stable LRS and HRS at 150 °C (read at 2 V).
- In a 32 × 32 crossbar array, 80 randomly selected cells showed high array-level uniformity, and the proposed biasing scheme reduced on-chip training energy from 0.66 J to 0.19 J, a 71% reduction; compared with the conventional half-voltage/serial scheme, energy fell from 0.51 J to 0.31 J (39% reduction) with parallel programming.
- MNIST training on a fully connected network (784-256-10), trained for 10 epochs on 60,000 samples, achieved an estimated recognition accuracy of ≈91%.
## Methods
- Device design and fabrication: a Pt/Ta2O5/Nb2O5-x/Al2O3-y/Ti stack was fabricated on SiO2/Si; the device area was 25 µm² and the crossbar line width was 5 µm.
- Characterization: single-device and array electrical measurements were performed with a probe station/Keithley 4200A-SCS and an ArC ONE platform; spectroscopy used XPS, TEM, REELS, and UPS to analyze composition and band structure.
- Analysis and simulation: conduction was fit with temperature-dependent and state-dependent Schottky emission analysis; neuromorphic energy and MNIST accuracy were evaluated using a software-based emulator of a fully connected DLP network with bi-directional weight updates.
## Limitations
- The paper notes the device is “not ideal yet”; the estimated ≈91% MNIST accuracy is lower than the ideal ≈97% possible in the same network, due to non-ideal linearity and conductance variation.
- The array testing platform’s current sensing limit was about ≈10^-8 A, so lower current levels were not detectable.
- Evident from the text: some retention results are reported at different temperatures and times (125 °C for 1000 s in one measurement, 150 °C for >10^5 s or >2 × 10^5 s elsewhere), so the exact headline retention figure is not fully consistent across sections.
Model-written summary. Check it against the paper before relying on it.
Integrity signals
Registry
Cites
Statistics
Figures





Verification
Claims the paper makes
2 Valid · 1 Not valid · 18 Not checked · 90 Own results and interpretations
Every source the paper cites for this claim was checked and supports it.
In the paper
“Both plots confirmed that the Schottky emission dominates the conduction of the device.[ 24 ]”
§ Retention Secured Nonlinear and Self‐Rectifying Analog Charge Trap Memristor for Energy‐Efficient Neuromorphic Hardware > Results and Discussion > Conduction Mechanism Analysis and Band Diagram Modeling of Pt/Ta2O5/Nb2O5‐ x /Al2O3‐ y /Ti Device Conduction mechanisms of the PTNAT device were examined to confirm the proposed band diagram model and understand the detailed switching mechanism. Figure 3a shows the LRS I–V curves measured at temperatures ranging from 40 to 70 °C. The results revealed that the LRS is highly temperature‐dependent, suggesting that the conduction is associated with the thermionic conduction mechanism. For more detailed analysis, the temperature‐dependent I–V curves at a high bias region (3.5–6.5 V) were plotted to the Schottky emission equation (ln(J/T 2) vs E 1/2), as shown in Figure 3b. In addition, multiple LRS curves programmed by different voltages from +6 to +9 V at fixed room temperature were plotted with the Schottky emission equation, as shown in Figure 3c. Both plots confirmed that the Schottky emission dominates the conduction of the device.[ 24 ] From the fitting, the optical dielectric constant (ε op) and SBH (Φ B) of the conduction‐limiting interface could be extracted at each temperature and each conductance state, as shown in Figure 3d and Figure 3e, respectively. For the fitting, the local electric field on each layer was calculated using the thickness and dielectric constant of the oxide layers. The thicknesses were 9.7 nm of Ta2O5, 27.7 nm of Nb2O5‐ x , and 8.3 nm of Al2O3‐ y from the TEM image in Figure 1a. The dielectric constants were 24 for Ta2O5,[ 25 ] 45 for Nb2O5‐ x ,[ 26 ] and 9 for Al2O3‐ y ,[ 27 ] which were taken from literature. After investigating all possible cases, we could conclude that the Schottky barrier at Ta2O5/Nb2O5‐ x interface limited the overall conduction (see Figure S4, Supporting Information, for the conduction mechanism fitting results limited by other layers.) The estimated ε op values ranged from 4.4 to 3.4 as the temperature increased from 40 to 70 °C, which coincided with the reference value (ε op ≈ 4) of Ta2O5 calculated by the square of its refractive index (n ≈ 2). In addition, the estimated Φ B ranged from 0.588 to 0.591 eV at 40–70 °C, suggesting the band offset at Ta2O5/Nb2O5‐ x interface. (The effective electron mass condition (m e ≈ 0.3 m 0) of Ta2O5 was used for the Φ B calculation.[ 28 ]) This band offset also coincides with the χ difference of the Ta2O5/Nb2O5‐ x interface measured from the UPS analysis, 4.03 eV (Nb2O5‐ x ), and 3.44 eV (Ta2O5). Similarly, the ε op and Φ B were also calculated from multiple conductance states. The estimated ε op was 3.82–4.08, consistent with the temperature‐dependent fitting results in Figure 3d. Interestingly, the Φ B were decreased from 0.72 eV (lower conductance state) to 0.60 eV (higher conductance state) as the set voltage increased, suggesting the Φ B decreased by further charge trapping. Such Φ B modulation behavior with respect to the charge trapping process can be plausibly explained by the following switching model. Figure 4a–c shows the XPS depth profile results of the Ta 4f, Nb 3d, and Al 2p core levels at each dielectric layer of the PTNAT device. The left panels show the raw XPS depth profile data (etch level from 1 to 57) to navigate the etch levels of interest (right panels) for each figure. In Figure 4a, the binding energy of two peaks have coincided with the reference Ta2O5 phase peaks (black dashed lines at 26.8 eV for Ta 4f 7/2 and 28.6 eV for Ta 4f 5/2).[ 29 ] The core levels showed no specific changes with the etch level increase, meaning a stoichiometric Ta2O5 layer was formed by the plasma‐enhanced atomic layer deposition (PEALD) process. Figure 4b shows Nb 3d core levels with the reference peak positions are as follows: Nb5+ (Nb2O5) peaks at 210.0 and 207.3 eV (blue dashed line), Nb4+ (NbO2) peaks at 208.8 and 206.0 eV (red dashed line), and Nb2+ (NbO) peaks at 206.8 and 204 eV (black dashed line).[ 30 ] The XPS results of the Nb2O5‐ x layer suggested complicated composition of various sub‐oxide phases, meaning a significant amount of oxygen vacancies exist in the Nb2O5‐ x charge trap layer. (see Figure S5, Supporting Information, for the Nb 3d XPS spectra deconvolution results of the PTNAT device.) At the Nb2O5‐ x /Al2O3‐ y interface, the peaks shifted toward the lower binding energy slightly, meaning the higher oxygen vacancies at the interface. Moreover, Al 2p peaks also showed lower binding energy shifting (orange dashed line) characteristics at the Nb2O5‐ x /Al2O3‐ y interface compared with the bulk region (black dashed line),[ 31 ] as shown in Figure 4c. Both indicate that a high concentration of oxygen vacancies is formed at the Nb2O5‐ x /Al2O3‐ y interface. This may be originated from the sputtering damage during the Nb2O5‐ x deposition kicking off the oxygens from the underlying Al2O3‐ y layer. Eventually, high‐density trap levels were formed at the interface helping the trap‐assisted tunneling process. Based on the conduction mechanism study and REELS, UPS, and XPS results, we could establish a detailed switching model of the PTNAT device. Figure 4d shows an energy band diagram of the PTNAT device. The work functions (Φ) of the electrodes were assumed to 5.6 eV for Pt top electrode and 4.3 eV for Ti bottom electrode. The bandgap (E g) and the electron affinity (χ) values were estimated from REELS and UPS results in Figure S3, Supporting Information, which are 4.5 and 3.44 eV for Ta2O5; 4.2 and 4.03 eV for Nb2O5‐ x ; and 6.6 and 2.29 eV for Al2O3‐ y , respectively. As mentioned above, the Nb2O5‐ x charge trap layer contains two trap levels (≈0.22 eV and ≈1.1 eV below the conduction band) associated with the oxygen vacancies. The deep trap level is responsible for the charge trapping. The shallow trap levels assist fluent carrier injection via Poole‐Frenkel (P‐F) emission.[ 32 ] Also, in the Al2O3‐ y layer, oxygen vacancies were observed at the Nb2O5‐ x /Al2O3‐ y interface. These defects could enhance the band‐to‐trap tunneling through the Al2O3‐ y layer, allowing the Al2O3‐ y to act as a tunneling layer despite the high band offset at the Al2O3‐ y /Ti interface. With the proposed band diagram, the analog charge trapping associated resistance switching characteristics can be well understood. Figure 4e shows schematic band diagrams during the set and reset process. In the pristine HRS state (panel i), the deep traps are empty, and the conduction is not fluent due to the high SBH at the Ta2O5/Nb2O5‐ x interface. When a positive set bias (>6 V) is applied to the Pt top electrode, the injected electrons start to transit to the deep traps and act as negative space charges. They form an internal electric field and pull up the conduction band, resulting in two changes to the programming process; first, it lowers the SBH at the Ta2O5/Nb2O5‐ x interface, making the conduction more fluent. Second, it also pulls up the shallow trap levels relative to the Fermi level of the Ti electrode and suppresses additional charge trapping. Therefore, the set switching process stops at the given set switching voltage, making an analog conductance change possible. Once the space charges are stored, the SBH at the Ta2O5/Nb2O5‐ x interface can be decreased by an image force‐associated Schottky barrier lowering effect.[ 33 ] The amount of barrier lowering depends on the applied partial electric field across the dielectric film, which can be given as (1) ΔϕE=e3E/4πεs+eαE2 where α is a material constant given by (2) α=eℏ2/24m∗(kBT)2 ε s is a permittivity, h is the Plank constant, and m* is the effective electron mass. Thus, as more charges are trapped, the lower SBH is formed. These trapped charges are stable due to the high band offset at the Ta2O5/Nb2O5‐ x and Nb2O5‐ x /Al2O3‐ y interfaces, as shown in Figure 2f. At a sufficient negative bias for reset, the trapped charges are de‐trapped by the reverse sequence of the trapping process from deep traps via the shallow traps to the electrode.
What the sources say
- Supportedcites [24] source
Our copy of Chiu (2014) treats Schottky emission as the prevailing mechanism in dielectric films: the section on Schottky/thermionic emission states 'Schottky emission the most often observed conduction mechanism in dielectric films, especially at relatively high temperature,' and §2.1 defines it as electron emission over the metal-dielectric barrier with image-force barrier lowering. The same section presents the Schottky plot (J/T^2 vs E^1/2, eqs. (1)-(2), with the Al/ZrO2/p-Si and related MIS data) as the means by which this mechanism is identified, which is what the citing 'both plots confirmed' analysis rests on. The device-specific conclusion belongs to the citing authors' own measurement; the review supplies the general/identification backing, with the mild qualifier 'especially at relatively high temperature' rather than unconditional dominance. (LLM, conf=0.82)
boostedstates a cause- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Supported
The cited sources support only part of this claim — read each source’s verdict below.
In the paper
“The shallow trap levels assist fluent carrier injection via Poole‐Frenkel (P‐F) emission.[ 32 ]”
§ Retention Secured Nonlinear and Self‐Rectifying Analog Charge Trap Memristor for Energy‐Efficient Neuromorphic Hardware > Results and Discussion > Conduction Mechanism Analysis and Band Diagram Modeling of Pt/Ta2O5/Nb2O5‐ x /Al2O3‐ y /Ti Device Conduction mechanisms of the PTNAT device were examined to confirm the proposed band diagram model and understand the detailed switching mechanism. Figure 3a shows the LRS I–V curves measured at temperatures ranging from 40 to 70 °C. The results revealed that the LRS is highly temperature‐dependent, suggesting that the conduction is associated with the thermionic conduction mechanism. For more detailed analysis, the temperature‐dependent I–V curves at a high bias region (3.5–6.5 V) were plotted to the Schottky emission equation (ln(J/T 2) vs E 1/2), as shown in Figure 3b. In addition, multiple LRS curves programmed by different voltages from +6 to +9 V at fixed room temperature were plotted with the Schottky emission equation, as shown in Figure 3c. Both plots confirmed that the Schottky emission dominates the conduction of the device.[ 24 ] From the fitting, the optical dielectric constant (ε op) and SBH (Φ B) of the conduction‐limiting interface could be extracted at each temperature and each conductance state, as shown in Figure 3d and Figure 3e, respectively. For the fitting, the local electric field on each layer was calculated using the thickness and dielectric constant of the oxide layers. The thicknesses were 9.7 nm of Ta2O5, 27.7 nm of Nb2O5‐ x , and 8.3 nm of Al2O3‐ y from the TEM image in Figure 1a. The dielectric constants were 24 for Ta2O5,[ 25 ] 45 for Nb2O5‐ x ,[ 26 ] and 9 for Al2O3‐ y ,[ 27 ] which were taken from literature. After investigating all possible cases, we could conclude that the Schottky barrier at Ta2O5/Nb2O5‐ x interface limited the overall conduction (see Figure S4, Supporting Information, for the conduction mechanism fitting results limited by other layers.) The estimated ε op values ranged from 4.4 to 3.4 as the temperature increased from 40 to 70 °C, which coincided with the reference value (ε op ≈ 4) of Ta2O5 calculated by the square of its refractive index (n ≈ 2). In addition, the estimated Φ B ranged from 0.588 to 0.591 eV at 40–70 °C, suggesting the band offset at Ta2O5/Nb2O5‐ x interface. (The effective electron mass condition (m e ≈ 0.3 m 0) of Ta2O5 was used for the Φ B calculation.[ 28 ]) This band offset also coincides with the χ difference of the Ta2O5/Nb2O5‐ x interface measured from the UPS analysis, 4.03 eV (Nb2O5‐ x ), and 3.44 eV (Ta2O5). Similarly, the ε op and Φ B were also calculated from multiple conductance states. The estimated ε op was 3.82–4.08, consistent with the temperature‐dependent fitting results in Figure 3d. Interestingly, the Φ B were decreased from 0.72 eV (lower conductance state) to 0.60 eV (higher conductance state) as the set voltage increased, suggesting the Φ B decreased by further charge trapping. Such Φ B modulation behavior with respect to the charge trapping process can be plausibly explained by the following switching model. Figure 4a–c shows the XPS depth profile results of the Ta 4f, Nb 3d, and Al 2p core levels at each dielectric layer of the PTNAT device. The left panels show the raw XPS depth profile data (etch level from 1 to 57) to navigate the etch levels of interest (right panels) for each figure. In Figure 4a, the binding energy of two peaks have coincided with the reference Ta2O5 phase peaks (black dashed lines at 26.8 eV for Ta 4f 7/2 and 28.6 eV for Ta 4f 5/2).[ 29 ] The core levels showed no specific changes with the etch level increase, meaning a stoichiometric Ta2O5 layer was formed by the plasma‐enhanced atomic layer deposition (PEALD) process. Figure 4b shows Nb 3d core levels with the reference peak positions are as follows: Nb5+ (Nb2O5) peaks at 210.0 and 207.3 eV (blue dashed line), Nb4+ (NbO2) peaks at 208.8 and 206.0 eV (red dashed line), and Nb2+ (NbO) peaks at 206.8 and 204 eV (black dashed line).[ 30 ] The XPS results of the Nb2O5‐ x layer suggested complicated composition of various sub‐oxide phases, meaning a significant amount of oxygen vacancies exist in the Nb2O5‐ x charge trap layer. (see Figure S5, Supporting Information, for the Nb 3d XPS spectra deconvolution results of the PTNAT device.) At the Nb2O5‐ x /Al2O3‐ y interface, the peaks shifted toward the lower binding energy slightly, meaning the higher oxygen vacancies at the interface. Moreover, Al 2p peaks also showed lower binding energy shifting (orange dashed line) characteristics at the Nb2O5‐ x /Al2O3‐ y interface compared with the bulk region (black dashed line),[ 31 ] as shown in Figure 4c. Both indicate that a high concentration of oxygen vacancies is formed at the Nb2O5‐ x /Al2O3‐ y interface. This may be originated from the sputtering damage during the Nb2O5‐ x deposition kicking off the oxygens from the underlying Al2O3‐ y layer. Eventually, high‐density trap levels were formed at the interface helping the trap‐assisted tunneling process. Based on the conduction mechanism study and REELS, UPS, and XPS results, we could establish a detailed switching model of the PTNAT device. Figure 4d shows an energy band diagram of the PTNAT device. The work functions (Φ) of the electrodes were assumed to 5.6 eV for Pt top electrode and 4.3 eV for Ti bottom electrode. The bandgap (E g) and the electron affinity (χ) values were estimated from REELS and UPS results in Figure S3, Supporting Information, which are 4.5 and 3.44 eV for Ta2O5; 4.2 and 4.03 eV for Nb2O5‐ x ; and 6.6 and 2.29 eV for Al2O3‐ y , respectively. As mentioned above, the Nb2O5‐ x charge trap layer contains two trap levels (≈0.22 eV and ≈1.1 eV below the conduction band) associated with the oxygen vacancies. The deep trap level is responsible for the charge trapping. The shallow trap levels assist fluent carrier injection via Poole‐Frenkel (P‐F) emission.[ 32 ] Also, in the Al2O3‐ y layer, oxygen vacancies were observed at the Nb2O5‐ x /Al2O3‐ y interface. These defects could enhance the band‐to‐trap tunneling through the Al2O3‐ y layer, allowing the Al2O3‐ y to act as a tunneling layer despite the high band offset at the Al2O3‐ y /Ti interface. With the proposed band diagram, the analog charge trapping associated resistance switching characteristics can be well understood. Figure 4e shows schematic band diagrams during the set and reset process. In the pristine HRS state (panel i), the deep traps are empty, and the conduction is not fluent due to the high SBH at the Ta2O5/Nb2O5‐ x interface. When a positive set bias (>6 V) is applied to the Pt top electrode, the injected electrons start to transit to the deep traps and act as negative space charges. They form an internal electric field and pull up the conduction band, resulting in two changes to the programming process; first, it lowers the SBH at the Ta2O5/Nb2O5‐ x interface, making the conduction more fluent. Second, it also pulls up the shallow trap levels relative to the Fermi level of the Ti electrode and suppresses additional charge trapping. Therefore, the set switching process stops at the given set switching voltage, making an analog conductance change possible. Once the space charges are stored, the SBH at the Ta2O5/Nb2O5‐ x interface can be decreased by an image force‐associated Schottky barrier lowering effect.[ 33 ] The amount of barrier lowering depends on the applied partial electric field across the dielectric film, which can be given as (1) ΔϕE=e3E/4πεs+eαE2 where α is a material constant given by (2) α=eℏ2/24m∗(kBT)2 ε s is a permittivity, h is the Plank constant, and m* is the effective electron mass. Thus, as more charges are trapped, the lower SBH is formed. These trapped charges are stable due to the high band offset at the Ta2O5/Nb2O5‐ x and Nb2O5‐ x /Al2O3‐ y interfaces, as shown in Figure 2f. At a sufficient negative bias for reset, the trapped charges are de‐trapped by the reverse sequence of the trapping process from deep traps via the shallow traps to the electrode.
What the sources say
- Partly supportedcites [32] sourceExperimental — accuracy not yet measured
The Discussion describes a shallow-defect Poole–Frenkel current and says it controls the injected tunneling current above turn-on. It distinguishes that current from carrier injection: electrons enter by tunneling above turn-on or thermionic emission below it, rather than by Poole–Frenkel emission. (LLM, conf=0.85)
plainstates a cause- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Partly supported
The cited sources that could be checked couldn’t confirm or dispute this claim.
In the paper
“Although our device is not ideal yet, the estimated accuracy is comparable to other hardware‐based simulation performances ranging from 70% to 94%.[ 41 , 42 , 43 , 44 , 45 ]”
§ Retention Secured Nonlinear and Self‐Rectifying Analog Charge Trap Memristor for Energy‐Efficient Neuromorphic Hardware > Results and Discussion > Demonstration of Energy‐Efficient Neuromorphic Computing
What the sources say
- Not checked — no full text of the sourcecites [41] source
- Not checked — no full text of the sourcecites [42] source
- Could not verifycites [43] sourceExperimental — accuracy not yet measured
The Results section on device fabrication and switching performance reports resistance ratios, retention, and switching times, not simulation accuracy. Searches for “70%,” “94%,” and hardware-based classification accuracy found no passage in our copy reporting the claimed performance range. (LLM, conf=0.76)
hedged- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: unclear — It is unclear whether the paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“Also, its retention is longer than 105 s at 150 °C, suggesting the device is highly stable for non‐volatile analog applications.”
§ Retention Secured Nonlinear and Self‐Rectifying Analog Charge Trap Memristor for Energy‐Efficient Neuromorphic Hardware > Abstract
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
A result the paper states — it cites no source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“The self‐rectifying and nonlinear characteristics allow reducing the on‐chip training energy consumption by 71% for the MNIST dataset training task with an optimized programming scheme.”
§ Retention Secured Nonlinear and Self‐Rectifying Analog Charge Trap Memristor for Energy‐Efficient Neuromorphic Hardware > Abstract
plainstates a cause- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
61 more claims
Create a free account to see all 111 claims and every source verdict.
- It may state two separate claims as one. 40
- The quote only partly states this claim. 2
- Probably not a substantial claim. 7
Claims other papers make about this one
Each is checked against this paper’s own text: does it say what the citing paper says it does?
The cited source says what the claim attributes to it.
In the citing paper
“We have previously proposed this structure, but it was on a SiO2/Si substrate22, so the device itself may not be novel.”
From Flexible self-rectifying synapse array for energy-efficient edge multiplication in electrocardiogram diagnosis (2025) · cites this paper as [22]
The cited source says what the claim attributes to it.
In the citing paper
“It has a low current analog operation capability and electroforming-free characteristics, allowing high reliability and uniformity.”
From Flexible self-rectifying synapse array for energy-efficient edge multiplication in electrocardiogram diagnosis (2025) · cites this paper as [22]
The cited source says what the claim attributes to it.
In the citing paper
“the self-rectifying characteristics of the device enable large-array operation without complex transistor or selector processes”
From Flexible self-rectifying synapse array for energy-efficient edge multiplication in electrocardiogram diagnosis (2025) · cites this paper as [22]
The cited source says what the claim attributes to it.
In the citing paper
“The NbOx layer is non-stoichiometric and amorphous, which is a key factor that acts as a charge trap layer22.”
From Flexible self-rectifying synapse array for energy-efficient edge multiplication in electrocardiogram diagnosis (2025) · cites this paper as [22]
The cited source says what the claim attributes to it.
In the citing paper
“The analog switching characteristics are attributed to the Schottky barrier height modulation accompanied by electron trapping and detrapping in the NbOx layer22.”
From Flexible self-rectifying synapse array for energy-efficient edge multiplication in electrocardiogram diagnosis (2025) · cites this paper as [22]
1 more claim about this paper
Create a free account to see every claim other papers make about this one.
Citations
Cites
Gradient-based learning applied to document recognition
Lecun, Y., Bottou, L., Bengio, Y., & Haffner, P. (1998) · Proceedings of the IEEE · doi:10.1109/5.726791 · Cited by 58,805
Redox‐Based Resistive Switching Memories – Nanoionic Mechanisms, Prospects, and Challenges
Waser, R., Dittmann, R., Staikov, G., & Szot, K. (2009) · Advanced Materials · doi:10.1002/adma.200900375 · Cited by 5,080
The future of electronics based on memristive systems
Zidan, M. A., Strachan, J. P., & Lu, W. D. (2018) · Nature Electronics · doi:10.1038/s41928-017-0006-8 · Cited by 2,138
A Review on Conduction Mechanisms in Dielectric Films
Chiu, F.-C. (2014) · Advances in Materials Science and Engineering · doi:10.1155/2014/578168 · Cited by 1,574
Efficient and self-adaptive in-situ learning in multilayer memristor neural networks
Li, C., Belkin, D., Li, Y., Yan, P., Hu, M., Ge, N., Jiang, H., Montgomery, E., Lin, P., Wang, Z., et al. (2018) · Nature Communications · doi:10.1038/s41467-018-04484-2 · Cited by 931
Yu, S., Gao, B., Fang, Z., Yu, H., Kang, J., & Wong, H. ‐. P. (2013) · Advanced Materials · doi:10.1002/adma.201203680 · Cited by 527
Woo, J., Moon, K., Song, J., Lee, S., Kwak, M., Park, J., & Hwang, H. (2016) · IEEE Electron Device Letters · doi:10.1109/led.2016.2582859 · Cited by 512
Crossbar RRAM Arrays: Selector Device Requirements During Read Operation
Zhou, J., Kim, K.-H., & Lu, W. (2014) · IEEE Transactions on Electron Devices · doi:10.1109/ted.2014.2310200 · Cited by 310
Self-citationIn situ training of feed-forward and recurrent convolutional memristor networks
Wang, Z., Li, C., Lin, P., Rao, M., Nie, Y., Song, W., Qiu, Q., Li, Y., Yan, P., Strachan, J. P., et al. (2019) · Nature Machine Intelligence · doi:10.1038/s42256-019-0089-1 · Cited by 307
Zhang, Y., Mao, G.-Q., Zhao, X., Li, Y., Zhang, M., Wu, Z., Wu, W., Sun, H., Guo, Y., Wang, L., et al. (2021) · Nature Communications · doi:10.1038/s41467-021-27575-z · Cited by 300
Research works printed but not linked: 1
- Modelli A., 2004, 2004 International Conference on Integrated Circuit and Design Technology, 211
Cited by
Memristor-based neural networks: a bridge from device to artificial intelligence
Cao, Z., Sun, B. T., Zhou, G., Mao, S., Zhu, S., Zhang, J., Ke, C., Zhao, Y. X., & Shao, J. (2023) · Nanoscale Horizons · doi:10.1039/d2nh00536k · Cited by 141
Xu, Z. G., Li, Y., Xia, Y., Shi, C., Chen, S., Ma, C., Zhang, C., & Li, Y. (2024) · Advanced Functional Materials · doi:10.1002/adfm.202312658 · Cited by 103
Self‐Rectifying Memristors for Three‐Dimensional In‐Memory Computing
Ren, S., Dong, A., Yang, L., Xue, Y., Li, J., Yu, Y., Zhou, H., Zuo, W., Li, Y., Cheng, W., et al. (2024) · Advanced Materials · doi:10.1002/adma.202307218 · Cited by 96
From fundamentals to frontiers: a review of memristor mechanisms, modeling and emerging applications
Thakkar, P., Gosai, J., Gogoi, H. J., & Solanki, A. (2024) · Journal of Materials Chemistry C · doi:10.1039/d3tc03692h · Cited by 100
Kim, Y. S., Baek, J. H., Im, I. H., Lee, D. H., Park, M. H., & Jang, H. W. (2024) · ACS Nano · doi:10.1021/acsnano.4c12884 · Cited by 75
Zhang, C., Chen, M., Pan, Y., Li, Y., Wang, K., Yuan, J., Sun, Y., & Zhang, Q. (2023) · Advanced Science · doi:10.1002/advs.202207229 · Cited by 66
Memristors with Tunable Volatility for Reconfigurable Neuromorphic Computing
Woo, K. S., Park, H., Ghenzi, N., Talin, A. A., Jeong, T.-Y., Choi, J., Oh, S., Jang, Y. H., Han, J., Williams, R. S., et al. (2024) · ACS Nano · doi:10.1021/acsnano.4c03238 · Cited by 67
Ran, H., Ren, Z., Li, J., Sun, B. T., Wang, T., Gu, D., Wang, W., Hu, X., Dong, Z., Song, Q., et al. (2024) · Advanced Functional Materials · doi:10.1002/adfm.202418113 · Cited by 63
Emerging memristive artificial neuron and synapse devices for the neuromorphic electronics era
Li, J., Abbas, H., Ang, D. S., Ali, A., & Ju, X. (2023) · Nanoscale Horizons · doi:10.1039/d3nh00180f · Cited by 56
Sahu, D. P., Park, K., Chung, P. H., Han, J., & Yoon, T. (2023) · Scientific Reports · doi:10.1038/s41598-023-36784-z · Cited by 68
Memcapacitor Crossbar Array with Charge Trap NAND Flash Structure for Neuromorphic Computing
Hwang, S., Yu, J., Song, M. S., Hwang, H., & Kim, H. (2023) · Advanced Science · doi:10.1002/advs.202303817 · Cited by 45
Lian, M., Gao, C., Lin, Z., Shan, L., Chen, C., Zou, Y., Cheng, E., Liu, C., Guo, T., Chen, W., et al. (2024) · Light Science & Applications · doi:10.1038/s41377-024-01516-z · Cited by 55
“This comprehensive analysis encompassed the examination of ion dynamics 34,35 , nonlinear attenuation 36 – 38 , short-term memory 39 – 41 , separability 42 – 44 , and echo state characteristics 45,46 of the device.”
BackgroundLu, C., Meng, J., Song, J., Xu, K., Wang, T., Zhu, H., Sun, Q., Zhang, D. W., & Chen, L. (2024) · ACS Nano · doi:10.1021/acsnano.4c09199 · Cited by 39
Filament-free memristors for computing
Choi, S., Moon, T., Wang, G., & Yang, J. J. (2023) · Nano Convergence · doi:10.1186/s40580-023-00407-0 · Cited by 36
Self-citationMultielement Filament Memristor Enabling Multifunctional Neuromorphic Device
Jang, M., Song, H., Kim, H. J., Yoon, J. H., & Kim, K. M. (2025) · Advanced Functional Materials · doi:10.1002/adfm.202423273 · Cited by 35
Self-citationTwo-dimensional material-based memristive devices for alternative computing
Panisilvam, J., Lee, H. Y., Byun, S., Fan, D., & Kim, S. (2024) · Nano Convergence · doi:10.1186/s40580-024-00432-7 · Cited by 38
“Charge trapping [115, 116], magneto resistive effects, and phase transition effects [117] are mechanisms that change a device between several levels of resistance states without forming a conductive filament.”
BackgroundThreshold Modulative Artificial GABAergic Nociceptor
Kim, G., Lee, Y., Jeon, J. B., Cheong, W. H., Park, W., Song, H., & Kim, K. M. (2023) · Advanced Materials · doi:10.1002/adma.202304148 · Cited by 29
Self-citationLocalized Conduction Channels in Memristors
Woo, K. S., Williams, R. S., & Kumar, S. (2024) · Chemical Reviews · doi:10.1021/acs.chemrev.4c00454 · Cited by 33
Lu, C., Meng, J., Yu, J., Song, J., Wang, T., Zhu, H., Sun, Q., Zhang, D. W., & Chen, L. (2024) · Nano Letters · doi:10.1021/acs.nanolett.3c04577 · Cited by 27
Physical Unclonable Function with 3D Stacked Memristor Crossbar Array Using Self-Differential Pair
Park, J., & Kim, H. (2025) · ACS Nano · doi:10.1021/acsnano.4c18621 · Cited by 30
Kim, D., Chabungbam, A. S., Kim, G., Son, J., Lim, B. M., Lee, Y. M., Kim, S. J., Kim, M., Lee, H., & Park, H. (2025) · Advanced Functional Materials · doi:10.1002/adfm.202503883 · Cited by 31
Self-citationLee, Y., Rhee, H., Kim, G., Cheong, W. H., Kim, D. H., Song, H., Kay, S. N., Lee, J., & Kim, K. M. (2025) · Nature Communications · doi:10.1038/s41467-025-59589-2 · Cited by 19
Self-citationZhang, G., Wang, Z., Fan, X., Wang, Z., Li, P., Luo, Q., Gao, D., Wan, Q., & Zhang, Y. (2024) · Applied Physics Letters · doi:10.1063/5.0225833 · Cited by 26
Park, T. W., Moon, J., Shin, D. H., Kim, H. J., Kim, S. S., Cho, J. M., Park, H., Woo, K. S., Kim, D. Y., Cheong, S., et al. (2024) · ACS Applied Materials & Interfaces · doi:10.1021/acsami.4c15598 · Cited by 18
Self-citationMemristive Explainable Artificial Intelligence Hardware
Song, H., Park, W., Kim, G., Choi, M. G., In, J. H., Rhee, H., & Kim, K. M. (2024) · Advanced Materials · doi:10.1002/adma.202400977 · Cited by 18
Self-citationZhao, J., Zhu, Y., Yan, S., Li, G., Liu, R., Zhong, Q., Bian, J., Peng, M., Li, Q., Li, Y., et al. (2025) · Small · doi:10.1002/smll.202500062 · Cited by 15
Zhang, G., Fan, X., Wang, Z., Li, P., Zhang, Z., Yu, B., Gao, D., Loke, D. K., Zhong, S., Wan, Q., et al. (2026) · Nano-Micro Letters · doi:10.1007/s40820-025-02035-1 · Cited by 19
“This work confirms the great potential of SRMs in developing neuromorphic computing, provides a promising synaptic device platform for building high-density, low-power ANNs with high learning capabilities and provides initial guidance for related work in the following years [63, 93, 119].”
Recent Advancements in 2D Material-Based Memristor Technology Toward Neuromorphic Computing
Park, S., Naqi, M., Lee, N., Park, S., Hong, S., & Lee, B. H. (2024) · Micromachines · doi:10.3390/mi15121451 · Cited by 17
Analog Ion‐Slicing LiNbO3 Memristor Based on Hopping Transport for Neuromorphic Computing
Wang, J., Zeng, H., Xie, Y., Zhao, Z., Pan, X., Luo, W., Shuai, Y., Tang, L., Zhu, D., Xie, Q., et al. (2023) · Advanced Intelligent Systems · doi:10.1002/aisy.202300155 · Cited by 15
Ren, S., Xue, Y.-B., Zhang, Y., Zuo, W., Li, Y., & Miao, X. (2024) · Applied Physics Letters · doi:10.1063/5.0190308 · Cited by 11
Martinez, A., Lee, Y., Kim, G., Park, W., Yu, J., Kay, S. N., Kim, E., Rhee, H., Kim, D. H., & Kim, K. M. (2025) · Advanced Functional Materials · doi:10.1002/adfm.202511469 · Cited by 12
Self-citationRen, S., Mao, G., Xue, Y.-B., Zhang, Y., Sun, J., Zuo, W., Li, Y., Xue, K., & Miao, X. (2024) · Applied Physics Letters · doi:10.1063/5.0227603 · Cited by 9
Sun, Y., Yuan, Q., Wang, Y., Li, B., & Wen, D. (2023) · ACS Sustainable Chemistry & Engineering · doi:10.1021/acssuschemeng.3c05027 · Cited by 8
Park, W., Song, H., Kim, E., Choi, M. G., Lee, M. G., Rhee, H., Kim, G., Go, T., Martinez, A., Kim, D., et al. (2025) · Advanced Materials · doi:10.1002/adma.202502255 · Cited by 8
Self-citationHe, S., Ye, X., Zhu, X., Zhong, Q., Liu, Y., Li, G., Liu, R., Meng, X., Xiao, Y., Yan, S., et al. (2024) · Nano Research · doi:10.26599/nr.2025.94907085 · Cited by 12
Liu, C., Jin, J., Xiao, Y., Wang, X., Yan, P., Cao, Y., Jiang, L., Sheng, C., & Yu, Y. (2024) · Rare Metals · doi:10.1007/s12598-024-02640-w · Cited by 10
Jang, J., & Gi, S.-G. (2025) · Advanced Materials Technologies · doi:10.1002/admt.202500588 · Cited by 13
Park, T., Kim, J., Kwon, Y. J., Kim, H., Yim, S., Shin, D. H., Kim, Y. R., Kim, H. J., & Hwang, C. S. (2024) · Advanced Materials Technologies · doi:10.1002/admt.202400965 · Cited by 6
Chanchal, Jindal, K., Tomar, M., & Jha, P. K. (2024) · ACS Applied Electronic Materials · doi:10.1021/acsaelm.4c00386 · Cited by 6
Coupling-Free Readout Scheme for Memcapacitors With NAND Flash Structure
Ahn, S., Yu, J., Hwang, H., Song, M. S., Yu, D., Hwang, S., Chung, I.-Y., Choi, W. Y., & Kim, H. (2024) · IEEE Transactions on Electron Devices · doi:10.1109/ted.2024.3418289 · Cited by 6
“In particular, resistance-based memristors have been a popular choice due to their simple two-terminal structure, fast processing speed, and high cell density [12], [13], [14], [15], [16], [17], [18], [19], [20], [21], [22], [23], [24], [25], [26].”
BackgroundDevice‐Level Implementation of Reservoir Computing With Memristors
Park, S., Kim, H., & Jang, H. W. (2026) · Advanced Intelligent Systems · doi:10.1002/aisy.202500843 · Cited by 9
“As illustrated in Figure 4, the interface-type memristors have a similar structure to filament-based memristors, but switching involves trapping of charges in interfacial defect sites instead of conducting filaments [98 – 100].”
Pham, P., Tran, T., Anh, T. V. V., Nguyen, T. D., Brügger, J., Kawazoe, Y., Phan, T. B., Nguyen-Tran, T., Vu, N. H., & Pham, N. K. (2025) · Advanced Functional Materials · doi:10.1002/adfm.202516695 · Cited by 7
A Neuransistor with Excitatory and Inhibitory Neuronal Behaviors for Liquid State Machine
Cheong, W. H., Kim, G., Lee, Y., Kim, E. Y., Jeon, J. B., Kim, D. H., & Kim, K. M. (2025) · Advanced Materials · doi:10.1002/adma.202419122 · Cited by 5
Self-citationBiswas, S., Paul, A. D., Ghosh, C. K., et al. (2024) · IEEE Transactions on Electron Devices · doi:10.1109/ted.2024.3482258 · Cited by 5
“It also reduces training energy consumption, though programming complexity increases [47].”
Optimization of self-rectifying analog memristors by insertion of an interfacial layer
Zhao, X., Li, K., Chen, Z., Dellith, A., Dellith, J., Hübner, U., Bengel, C., Liu, F., Menzel, S., Schmidt, H., et al. (2024) · Applied Physics Letters · doi:10.1063/5.0213396 · Cited by 5
Modeling of Bilayer Modulated RRAM and Its Array Performance for Compute-in-Memory Applications
Lee, J.-W., Chou, T. H., Chen, P., et al. (2023) · IEEE Journal on Exploratory Solid-State Computational Devices and Circuits · doi:10.1109/jxcdc.2023.3311899 · Cited by 5
“Many studies have shown that the bilayer structure that creates high-nonlinearity RRAM demonstrates FNT behavior [6], [7], [8].”
BackgroundSelf-citationUpdate Disturbance‐Resilient Analog ReRAM Crossbar Arrays for In‐Memory Deep Learning Accelerators
Choi, W., Stecconi, T., Falcone, D. F., Galetta, M., Clerico, V., Zaccaria, E., Ram, M. S., Porta, A. L., Horst, F., Jubin, D., et al. (2025) · Advanced Science · doi:10.1002/advs.202504578 · Cited by 9
Zhang, Y., Ren, S., Zuo, W., Xue, Y.-B., Sun, J., Xue, K., Li, Y., & Miao, X. (2025) · IEEE Electron Device Letters · doi:10.1109/led.2025.3578071 · Cited by 6
Atomic Layer Deposition in Transistors and Monolithic 3D Integration
Liu, Y., Zou, T., Xu, C., Zhen, L., Li, Y., & Noh, Y. (2026) · Advanced Functional Materials · doi:10.1002/adfm.74729 · Cited by 5
Yang, B., Cao, Z., Sun, B. T., Gao, K., Wang, M., Zhou, G., Lv, J., Luo, S., Chen, X., Li, X., et al. (2025) · Applied Materials Today · doi:10.1016/j.apmt.2025.102771 · Cited by 7
Electric Field‐Driven Conformational Changes in Molecular Memristor and Synaptic Behavior
Lim, C., Kim, T., Park, Y. J., Kim, D., Shin, C., Ha, S., Lin, J., Li, Y., & Park, J. (2025) · Advanced Science · doi:10.1002/advs.202505016 · Cited by 3
Acevedo, W. R., Aguirre, M. H., Noheda, B., & Rubi, D. (2024) · Physical Review Materials · doi:10.1103/physrevmaterials.8.075003 · Cited by 3
Neuromorphic Near‐Sensor and In‐Sensor Computing Enabled by Next‐Generation Material‐Based Sensors
Jung, S. Y., Kim, G. Y., Kim, S., Seo, H., Lee, S. G., Won, S. M., Rhee, D., & Kang, J. Y. (2026) · Advanced Science · doi:10.1002/advs.75742 · Cited by 5
“When metal oxide is used as the switching layer, charge carriers injected from the electrode are trapped in deep-level defects such as oxygen vacancies within the oxide or near the interface [ 114, 117 ].”
Self-citationKim, H. J., Kim, S. S., Park, S. J., Oh, Y., & Han, S. (2025) · Journal of the Korean Ceramic Society · doi:10.1007/s43207-025-00491-4 · Cited by 2
Song, Y., Yoon, J. H., Lee, S., Ham, W., Park, J.-M., Lee, J., & Kwon, J. (2023) · Applied Physics Letters · doi:10.1063/5.0158280 · Cited by 2
Self-citationEvolution of Neural Network Models and Computing-in-Memory Architectures
Kumar, A., Devi, M., Singh, S. K., Singh, K., & Mishra, D. (2025) · Archives of Computational Methods in Engineering · doi:10.1007/s11831-025-10481-8 · Cited by 3
Bian, J., Zhu, Y., Yan, S., Tang, Y., Guo, J., Li, G., Zhao, J., Zhong, Q., Li, Q., Liu, S., et al. (2025) · Nano Research · doi:10.26599/nr.2025.94907803 · Cited by 1
Guo, W., Wu, Z., Wang, H., Zou, J., Hao, Y., Wang, W., Zhao, X., Cao, R., Feng, Z., Xu, L., et al. (2026) · Advanced Materials · doi:10.1002/adma.75096
Park, H., Jeong, H. Y., Woo, K. S., Kim, H. W., Han, J., Shin, J. H., & Hwang, C. S. (2026) · ACS Applied Materials & Interfaces · doi:10.1021/acsami.6c14083
Lee, J. W., Shin, D. H., Myoung, S. J., Eo, S., Kim, W. C., Lee, S., Kim, C., Lee, J., Lee, Y. J., Choi, S., et al. (2026) · Advanced Electronic Materials · doi:10.1002/aelm.70579
Self-citationWafer-Scale All-Silicon Self-Rectifying Memristor for Synaptic Response and Reservoir Computing
Yoon, J. H., Cho, S. H., Moroshkin, P., Kim, S., Bo, M. D., Kwon, J., Nam, K. T., & Xu, J. (2026) · Nano Letters · doi:10.1021/acs.nanolett.6c00994
Multimode Oxide‐Based Optoelectronic Memtransistor for In‐Sensor Vision Processing
Lee, M., Kim, G. T., Kim, H. S., Kim, D. K., Song, H., Noh, H., Noh, J., Lee, S., Park, S. K., Cho, H., et al. (2026) · Advanced Functional Materials · doi:10.1002/adfm.77205
Self-citationLee, S., Ju, S., Lee, W. J., Kang, M., Kim, S., & Kim, Y. (2026) · Advanced Science · doi:10.1002/advs.76640
Raj, J. V., Archana, T., & Jinesh, K. B. (2026) · Journal of Vacuum Science & Technology B Nanotechnology and Microelectronics Materials Processing Measurement and Phenomena · doi:10.1116/6.0005462
Emerging Memory and Device Technologies for Hardware‐Accelerated Model Training and Inference
Cho, Y., Kim, D., Kimh, Y., Na, J., Jeong, Y., Lee, D., & Choi, S. (2026) · Advanced Electronic Materials · doi:10.1002/aelm.202500796
Self-citationArtificial nociceptor with Pt/NbO x /Graphene memristive device
Dai, Y., Lin, Z., Wang, H., Wang, Z., Zou, J., Guo, W., Feng, Z., Xu, Z., Zhu, Y., & Wu, Z. (2025) · Nanotechnology · doi:10.1088/1361-6528/ae1dfe
Kim, G., Park, J., Park, W., Ko, M., Lee, M. G., Cho, H., & Kim, K. M. (2025) · Small Science · doi:10.1002/smsc.202500309 · Cited by 1
Self-citationElectrical Behavior of Combinatorial Thin-Film ZrxTa1−xOy
Flynn‐Hepford, M., Emery, R., Randolph, S., Retterer, S. T., Eres, G., Sumpter, B. G., Ievlev, A. V., Ovchinnikova, O. S., & Rack, P. D. (2025) · Nanomaterials · doi:10.3390/nano15100732
“Kim et al. also produced charge trapping-based Nb 2 O 5-x memristors, but in order to stabilize the high and low-resistance states, Ta 2 O 5 and Al 2 O 3-y capping layers were inserted between the Nb 2 O 5-x active layer and Pt and Ti electrodes, respectively [16].”
Half-Oxidized MoS2-Based Memristor by UV-Ozone Treatment
Shin, B., Kim, J. Y., Gwon, O. H., et al. (2025) · ACS Applied Electronic Materials · doi:10.1021/acsaelm.5c00031
Self-citationKim, H. Y., Ghenzi, N., Park, H., Shin, D. H., Kim, D. Y., Park, T. W., Cho, J. M., Park, T., & Hwang, C. S. (2025) · Nanoscale Horizons · doi:10.1039/d5nh00256g · Cited by 1
(2024) · doi:10.21203/rs.3.rs-4575664/v1 · Cited by 19
Self-citation
Metadata
- Authors
- Kim, G., Son, S., Song, H., Jeon, J. B., Lee, J., Cheong, W. H., Choi, S., & Kim, K. M.
- Affiliations
- Korea Advanced Institute of Science and Technology (KR)
- Samsung (South Korea) (KR)
- Corresponding author
- Kim, Kyung Min
- Venue
- Advanced Science
- Publisher
- Wiley
- Published
- 2022-11-27
- Volume
- 10
- Issue
- 3
- Pages
- e2205654-e2205654
- ISSN
- 2198-3844 (ISSN-L)
- PMC
- PMC9875615
- OpenAlex
- W4310277646
- Cited by
- 72
- References
- 45
- Citations by year
- 2026: 13 · 2025: 22 · 2024: 24 · 2023: 12 · 2022: 1
- FWCI
- 4.04
- Type
- journal article
- Language
- en
- Open access
- gold
- doi.org
- Licence
- Locations
- Advanced Science (publishedVersion, cc-by)
- PubMed Central (publishedVersion, cc-by)
- PubMed (publishedVersion)
- DOAJ (DOAJ: Directory of Open Access Journals) (submittedVersion, cc-by-sa)
- Funders
- National Research Foundation of Korea (2020M3F3A2A01081585, 2020M3F3A2A01081916, 2022M3I7A4085484, 2022M3F3A2A01085466)
- Topics
- Physical Sciences › Engineering › Electrical and Electronic Engineering › Advanced Memory and Neural Computing
- Physical Sciences › Engineering › Electrical and Electronic Engineering › Ferroelectric and Negative Capacitance Devices
- Physical Sciences › Engineering › Electrical and Electronic Engineering › Semiconductor materials and devices
- Keywords
- memristor crossbar, nonlinear current-voltage characteristics, sneak current suppression, retention stability, conduction mechanism, neuromorphic hardware, energy-efficient training, MNIST classification
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
The claim verdicts this report counts are experimental: their accuracy is not yet measured.
A language model drew the claims from the paper on 2026-10-10, under version 3 of the claim definition (model deepseek-v4.1-flash, prompt extract-v3).
A second model labelled each claim on 2026-10-10 (model jev-1.13-20260917, definition revision gold-guideline-v0.2).
Cited sources are checked with a language model’s help (qwen3.8-omni-flash and gpt-6-sol).
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
Data sources
Bibliographic data: Semantic Scholar, OpenAlex, Crossref.
Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).
Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.