Vertical Metal‐Oxide Electrochemical Memory for High‐Density Synaptic Array Based High‐Performance Neuromorphic Computing

Lee, Hyunjoon; Ryu, Da Gil; Lee, Giho; Song, Moon Hee; Moon, Hyungjin; Lee, Jaehyeong; Ryu, Jongchan; Kang, Ji‐Hoon; Suh, Junmin; Kim, Sang-Bum; Lim, Jongwoo; Jeon, D.; Kim, Seyoung; Kim, J.; Lee, Yun Seog

doi:10.1002/aelm.202200378

Cited by 14 publications

(7 citation statements)

References 25 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…Switching behavior at room temperature. The switching mechanism of the ECRAM device is articulated through three steps: 1) ion migration within the electrolyte, 2) ion diffusion into the channel, and 3) electrochemical reactions 32 . Upon application of a gate voltage, oxygen ions (or oxygen vacancies) migrate from the electrolyte toward the channel, accumulating at the interface before diffusing into the channel.…”

Section: Resultsmentioning

confidence: 99%

Unveiling ECRAM Switching Mechanism Using Variable Temperature Hall Measurements for Accelerated AI Computation

Kim,

Kwak,

Choi

et al. 2024

Preprint

Self Cite

View full text Add to dashboard Cite

Electrochemical random-access memory (ECRAM) devices are promising cross-point elements for implementing analog cross-point array-based AI computation accelerators due to their remarkable programmability and high-stability driven by ion movement. However, efforts to investigate the fundamental physical parameters governing switching have been limited by the complexity of ECRAM multilayer film structure and the low mobility and high-resistivity of the oxide material. Additionally, tracking the movement of oxygen vacancies within metal oxides has been challenging compared to other active ions. Herein, we fabricate ECRAM devices with multi-terminal Hall-bar structures and conduct AC magnetic parallel dipole line (PDL) Hall measurements. Using two rotating DC magnets and lock-in technique, we extract key transport parameters such as mobility and carrier density of high-resistivity WO3-x channel layer. Through variable-temperature Hall measurements at 50 - 300 K, we experimentally measure, for the first time, the oxygen donor level at approximately 0.1 eV in WO3-x thin film. In addition, we characterize the resistive switching of ECRAM at low temperatures, for the first time, and reveal that the conductance potentiation is caused by increase in both mobility and carrier density. This phenomenon can be attributed to the reversible changes in electronic and atomic structure supported by density functional theory calculation result. Overall, our experimental study on ECRAM switching mechanism contributes to a deep understanding of the operational principles of ECRAM and provides valuable insights for enabling high-performance and energy-efficient AI computation in analog hardware.

show abstract

Section: Resultsmentioning

confidence: 99%

Unveiling ECRAM Switching Mechanism Using Variable Temperature Hall Measurements for Accelerated AI Computation

Kim,

Kwak,

Choi

et al. 2024

Preprint

Self Cite

View full text Add to dashboard Cite

show abstract

“…About 50% of multi-level RS demonstrations based on hafnium oxide were implemented with pure hafnium oxide [30, 37, 38, 42, 44-47, 52, 54, 56, 57, 59, 71-81], another about 20% used hafnium oxide and aluminum oxide in various combinations such as bi-/multi-layers or aluminum hafnium oxide [31,33,47,48,64,[82][83][84][85], and the rest combined hafnium oxide with various other materials such as tantalum oxide [34,55,58,86,87], titanium oxide [35,40,51,88,89], tungsten oxide [53,63,90], molybdenum oxide [90], zirconium oxide [91], hafnium oxynitride [39], titanium oxynitride [32], or Pt nanoparticles [43], Ti [56,92], Ni [41], or Ba [36] interspersed in hafnium oxide or aluminum oxide.…”

Section: Methodsmentioning

confidence: 99%

Multi-level resistive switching in hafnium-oxide-based devices for neuromorphic computing

Hellenbrand,

MacManus-Driscoll

2023

Nano Convergence

View full text Add to dashboard Cite

In the growing area of neuromorphic and in-memory computing, there are multiple reviews available. Most of them cover a broad range of topics, which naturally comes at the cost of details in specific areas. Here, we address the specific area of multi-level resistive switching in hafnium-oxide-based devices for neuromorphic applications and summarize the progress of the most recent years. While the general approach of resistive switching based on hafnium oxide thin films has been very busy over the last decade or so, the development of hafnium oxide with a continuous range of programmable states per device is still at a very early stage and demonstrations are mostly at the level of individual devices with limited data provided. On the other hand, it is positive that there are a few demonstrations of full network implementations. We summarize the general status of the field, point out open questions, and provide recommendations for future work.

show abstract

“…8 c. Lee et al . successfully implemented vertical ECRAM cells with a minimum size (F) of four times the minimum feature size (4F 2 ) [ 67 ]. The manufacturing process involved sequential deposition of the WO 3 channel, HfO 2 electrolyte, and MoO y reservoir in a planar structure through sputtering.…”

Section: Device-level Analysismentioning

confidence: 99%

Electrochemical random-access memory: recent advances in materials, devices, and systems towards neuromorphic computing

Kwak,

Kim,

Jeon

et al. 2024

Nano Convergence

Self Cite

View full text Add to dashboard Cite

Artificial neural networks (ANNs), inspired by the human brain's network of neurons and synapses, enable computing machines and systems to execute cognitive tasks, thus embodying artificial intelligence (AI). Since the performance of ANNs generally improves with the expansion of the network size, and also most of the computation time is spent for matrix operations, AI computation have been performed not only using the general-purpose central processing unit (CPU) but also architectures that facilitate parallel computation, such as graphic processing units (GPUs) and custom-designed application-specific integrated circuits (ASICs). Nevertheless, the substantial energy consumption stemming from frequent data transfers between processing units and memory has remained a persistent challenge. In response, a novel approach has emerged: an in-memory computing architecture harnessing analog memory elements. This innovation promises a notable advancement in energy efficiency. The core of this analog AI hardware accelerator lies in expansive arrays of non-volatile memory devices, known as resistive processing units (RPUs). These RPUs facilitate massively parallel matrix operations, leading to significant enhancements in both performance and energy efficiency. Electrochemical random-access memory (ECRAM), leveraging ion dynamics in secondary-ion battery materials, has emerged as a promising candidate for RPUs. ECRAM achieves over 1000 memory states through precise ion movement control, prompting early-stage research into material stacks such as mobile ion species and electrolyte materials. Crucially, the analog states in ECRAMs update symmetrically with pulse number (or voltage polarity), contributing to high network performance. Recent strides in device engineering in planar and three-dimensional structures and the understanding of ECRAM operation physics have marked significant progress in a short research period. This paper aims to review ECRAM material advancements through literature surveys, offering a systematic discussion on engineering assessments for ion control and a physical understanding of array-level demonstrations. Finally, the review outlines future directions for improvements, co-optimization, and multidisciplinary collaboration in circuits, algorithms, and applications to develop energy-efficient, next-generation AI hardware systems.

show abstract

Vertical Metal‐Oxide Electrochemical Memory for High‐Density Synaptic Array Based High‐Performance Neuromorphic Computing

Cited by 14 publications

References 25 publications

Unveiling ECRAM Switching Mechanism Using Variable Temperature Hall Measurements for Accelerated AI Computation

Unveiling ECRAM Switching Mechanism Using Variable Temperature Hall Measurements for Accelerated AI Computation

Multi-level resistive switching in hafnium-oxide-based devices for neuromorphic computing

Electrochemical random-access memory: recent advances in materials, devices, and systems towards neuromorphic computing

Contact Info

Product

Resources

About