Compliance-Free, Digital SET and Analog RESET Synaptic Characteristics of Sub-Tantalum Oxide Based Neuromorphic Device

Abbas, Yawar; Jeon, Yu‐Rim; Соколов, А.С.; Kim, Sohyeon; Ku, Bon‐Cheol; Choi, Changhwan

doi:10.1038/s41598-018-19575-9

Cited by 110 publications

(85 citation statements)

References 68 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Resistive random access memory (RRAM), which achieves data storage through reversible resistance changes in certain dielectrics, is one of the leading candidates for this purpose. Reproducible resistive switching phenomena have been widely observed in oxides [6][7][8][9], chalcogenides [10] and even organic materials [11]. Due to their simple device structure (metal/insulator/metal), high compatibility to standard CMOS flow line and the prospect of 3D integration, certain binary oxides such as Al 2 O 3 [12], HfO 2 [13,14], TiO2 [15,16], and ZrO 2 [17] have attracted more attention.…”

Section: Introductionmentioning

confidence: 99%

GGA-1/2 self-energy correction for accurate band structure calculations: the case of resistive switching oxides

et al. 2018

View full text Add to dashboard Cite

First-principles calculation has become an indispensable methodology in revealing the working principles of nanoscale electronic devices, but ultra-large supercells are usually required in modeling the devices with critical metal/dielectric interfaces. Traditional density functional theory within the generalized gradient approximation (GGA) suffers from the inaccurate band gap problem when metal oxides are present, but they serve as the core component in resistive random access memory (RRAM), which is a promising path for novel high speed non-volatile memories. To obtain improved oxide band gaps, we applied the efficient GGA-1/2 method for self-energy correction, whose computational load is at the same level as standard GGA. In particular, we have investigated the influence of exchange-correlation functional flavors on the GGA-1/2 band structures, taking four important binary oxide RRAM materials (α-Al 2 O 3 , r-TiO 2 , m-ZrO 2 and m-HfO 2 ) as benchmark examples. Five GGA functionals (PBE, PBEsol, PW91, revPBE and AM05) were considered and their band structures were compared in detail. We have found that the performance of GGA-1/2 is comparable to state-ofthe-art GW and generally superior to the HSE06 hybrid functional. Among the five GGA functionals, PBEsol yields the best results in general. In addition, the applicability of a single self-energy potential for various GGA-1/2 flavors is discussed. Our work provides a guide to the GGA flavor selection, when applying the GGA-1/2 method to metal oxides. manifest a transition from a high resistance state (HRS) to a low resistance state (LRS) upon the formation of conductive filaments (CFs) connecting the two electrodes. Yet, at present there are still controversies on the exact composition and geometry of the CFs [19][20][21][22]. Since it is difficult to directly observe and determine experimentally the exact composition of the CFs within the dielectric media, theoretical approaches can be particularly helpful.Among atomistic simulation techniques, ab initio calculations based on density functional theory (DFT) [23] are highly successful in the prediction of total energies, bond lengths, and vibration frequencies. However, DFT within local density approximation (LDA) or generalized gradient approximation (GGA) severely underestimates the band gaps of semiconductors and insulators, which limits its capability of predicting the band alignment in metal/semiconductor contacts, a significant task in understanding charge transmission through CFs [24]. Various theoretical approaches have been proposed to overcome this shortcoming, thus providing more reliable predictions of physical properties that depend on excited states. Hedin's GW approximation is considered to be the state-of-the-art, which is still the most accurate method for electronic structure calculation. It calculates the quasiparticle energies in terms of the perturbation theory [25], and is a beyond-DFT technique. Hybrid functionals [26], which combine standard DFT and Hartree-Fock exchange functionals thro...

show abstract

Section: Introductionmentioning

confidence: 99%

GGA-1/2 self-energy correction for accurate band structure calculations: the case of resistive switching oxides

et al. 2018

View full text Add to dashboard Cite

show abstract

“…2020, 6,1901012 It is speculated that the BFO layer contains much non-lattice oxygen and influences the Schottky barrier height between HfO 2 and TiN electrode, resulting in a large ON/OFF ratio.…”

Section: Discussionmentioning

confidence: 99%

“…[3] Recently, neuromorphic computing system has been extremely considered as a promising candidate as the next generation computing scheme for an efficient information processor, which is inspired by the biologic brain. It is widely accepted that the brain contains two kinds of basic units, neurons and connected synapses, [6][7][8] which serve as critical parts in memory and learning. It is widely accepted that the brain contains two kinds of basic units, neurons and connected synapses, [6][7][8] which serve as critical parts in memory and learning.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Designing High‐Performance Storage in HfO₂/BiFeO₃ Memristor for Artificial Synapse Applications

Liu

Xiong

Liu

et al. 2020

Adv Elect Materials

View full text Add to dashboard Cite

using artificial synapses is an essential step to accomplish the neuromorphic computing system. [9,10] Formerly, artificial synapses were realized by complementary metal-oxide-semiconductor (CMOS) circuitry containing dozens of electronic components. [11] However, many electronic components result in complicated architecture and high energy consumption. As comparison, two-terminal memristors, especially resistive random access memory and phase change random access memory, that recently entered our field of vision have been widely discussed as artificial synapses owing to their structures which is similar to that of synapses and the reproducible tuning of resistance. [12][13][14][15][16] Also, for an ideal synapse device it is better to meet these requirements, such as symmetric potentiation-depression characteristics, 5-bit/cell analog levels, and high non-volatility with ≈100 conductance ON/OFF ratio. They are the key points we need to take into account. [17,18] In particular, HfO 2 -based memristors have been demonstrated as a leading alternative as synapse in virtue of its distinctive superiority, such as simple structure, <10 ns switching speed, <10 pJ power consumption, multilevel ability, and compatibility with CMOS fabrication process. [19][20][21] However, the resistance contrast (ON/OFF ratio) of HfO 2 -based memristors ranges from ≈40 to ≈150. [22] Considering the resistance fluctuation of these memristors across a silicon wafer, larger ON/OFF ratio is needed to guarantee high recognition accuracy (>97%). [23] Recently, several works were reported on improvement of the memristors' performance with thin interfacial layer. [24][25][26] As one of the most important multiferroic materials, bismuth iron oxide with perovskite structure has come into notice for its potential in multifunctional device applications. [27] Moreover, BiFeO 3 (BFO) has attracted much attention because it possesses superior characteristics of resistance switching (RS) such as large ON/OFF ratio in some researches. [28] It is also confirmed that BFO can be applied in bi-layer design memristor to significantly improve RS characteristics, [29] which gives inspiration to insert BFO thin film as the goal of high device performances.In this work, ultrathin BFO film was inserted to fabricate Pt/BFO/HfO 2 /TiN memristor to improve the RS characteristic of HfO 2 -based memristor. The material characterization and RS behavior were systemically analyzed. The role of the inserting BFO layer on the RS behavior was evaluated and the RS mechanism triggered by inserting BFO film was explored.HfO 2 -based memristors that remembers the history of the current that has passed through them have attracted great interest for use as artificial synapses in neuromorphic systems. However, the low resistance contrast exhibited by HfO 2 -based memristors seriously decreases their recognition accuracy. By inserting a 2 nm BiFeO 3 layer a large memory window of 10 4 and remarkable pulse endurance of 10 8 cycles are achieved. Multilevel storage capability is also d...

show abstract

“…The exact required device performance specifications, including switching speed and retention, strongly depend on the biological functionality that is required from the neural network [2,3,5,6]. For example, the processes of learning and forgetting in the human brain are governed by the modulation of the strengths of paths connecting neurons, known as synaptic weights, as a result of action potentials [4,7,10,11]. This synaptic plasticity behavior can be emulated by memristors [12]: two-terminal devices of which the microscopic internal state can be modified and measured as a change in the conductance.…”

Section: Introductionmentioning

confidence: 99%

Electric field driven memristive behavior at the Schottky interface of Nb-doped SrTiO3

Goossens

Das

Banerjee

2018

Journal of Applied Physics

View full text Add to dashboard Cite

Computing inspired by the human brain requires a massive parallel architecture of low-power consuming elements of which the internal state can be changed. SrTiO3 is a complex oxide that offers rich electronic properties; here Schottky contacts on Nb-doped SrTiO3 are demonstrated as memristive elements for neuromorphic computing. The electric field at the Schottky interface alters the conductivity of these devices in an analog fashion, which is important for mimicking synaptic plasticity. Promising power consumption and endurance characteristics are observed. The resistance states are shown to emulate the forgetting process of the brain. A charge trapping model is proposed to explain the switching behavior.

show abstract

Compliance-Free, Digital SET and Analog RESET Synaptic Characteristics of Sub-Tantalum Oxide Based Neuromorphic Device

Cited by 110 publications

References 68 publications

GGA-1/2 self-energy correction for accurate band structure calculations: the case of resistive switching oxides

GGA-1/2 self-energy correction for accurate band structure calculations: the case of resistive switching oxides

Designing High‐Performance Storage in HfO₂/BiFeO₃ Memristor for Artificial Synapse Applications

Electric field driven memristive behavior at the Schottky interface of Nb-doped SrTiO3

Contact Info

Product

Resources

About

Compliance-Free, Digital SET and Analog RESET Synaptic Characteristics of Sub-Tantalum Oxide Based Neuromorphic Device

Cited by 110 publications

References 68 publications

GGA-1/2 self-energy correction for accurate band structure calculations: the case of resistive switching oxides

GGA-1/2 self-energy correction for accurate band structure calculations: the case of resistive switching oxides

Designing High‐Performance Storage in HfO2/BiFeO3 Memristor for Artificial Synapse Applications

Electric field driven memristive behavior at the Schottky interface of Nb-doped SrTiO3

Contact Info

Product

Resources

About

Designing High‐Performance Storage in HfO₂/BiFeO₃ Memristor for Artificial Synapse Applications