Impact of Switching Layer Architecture on Power Consumption in RRAM Journal: Journal of Electronic Materials
John F. Hardy II, Jack Garrard, Guilherme S.Y. Giradini, Carlo R. daCunha · 2025
Read Impact of Switching Layer Architecture on Power Consumption in RRAM Journal: Journal of Electronic Materials by John F. Hardy II, Jack Garrard, Guilherme S.Y. Giradini, Carlo R. daCunha (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2510-17831.txt.
Introduction
John F. Hardy and three co-authors show that porous helical layers cut memory chip power sharply. Their devices are resistive random-access memory, a switchable resistance memory type. The test stack pairs indium tin oxide with tungsten oxide layers. Helical and flat film devices behaved alike at high current limits.
Only helical devices kept working reliably at a much lower current limit. There, reset voltages fell by about sixty percent. Switching currents dropped by roughly seventy percent as well. The readable memory window grew several-fold at that low setting. Geometry-driven field concentration explains the low-power, high-margin behavior.
Catalog record
| CREATOR | John F. Hardy II, Jack Garrard, Guilherme S.Y. Giradini, Carlo R. daCunha |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 29 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2510-17831
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Impact of Switching Layer Architecture on Power Consumption in RRAM Journal: Journal of Electronic Materials John F. Hardy II, Jack Garrard, Guilherme S.Y. Giradini, Carlo R. daCunha Abstract This work demonstrates that porous helical WO x architectures enable a distinct low-power regime for planar ITO/WO x /ITO resistive random-access devices. While thin film and helical devices behave similarly at a 5 mA compliance, only helical devices sustain reproducible operation at 500 500 μ \mu A, where RESET voltages reduce by ∼ 60 % \mathord{\mathord{\sim}}60\% , switching currents decrease by 68 − 75 % 68-75\% , and SET/RESET power drops by ∼ 89 % \mathord{\mathord{\sim}}89\% and ∼ 83 % \mathord{\mathord{\sim}}83\% . With helical devices operating at 500 500 μ \mu A, the memory window expands 400 400 – 600 % 600\% due to selective suppression of high-resistive-state leakage, yielding both lower-power and improved read margin in a regime inaccessible to thin film devices. These results highlight geometry-driven field enhancement and confinement as practical design principles for low-power, high-margin resistive memories and point toward opportunities in transparent, flexible, and high-surface-area material systems. 1 Introduction Traditionally, silicon-based Flash memories are the current prevailing non-volatile storage devices, yet their fundamental operating principles present increasing limitations as device dimensions demand improved scalability and applications demand intensified operations [ 1 ] . Flash memory encodes information by storing electrons in a floating gate within a metal-oxide-semiconductor field-effect transistor (MOSFET) [ 2 ] . While Flash remains widely adopted due to its non-volatility, data retention capabilities, and relatively high storage density, it suffers from several critical limitations. These limitations consist of high program and erase voltages, long write/erase times, and limited endurance. Down-scaling below the 20 nm node presents more challenges, such as charge leakage, increased variability, and degradation in reliability and yield [ 3 , 4 , 5 ] . More broadly, conventional memory technologies are constrained by the Von Neumann architecture, where physically separated memory and processing units lead to the well-known memory bottleneck [ 6 ] . The limitations of conventional memory technologies are particularly detrimental in emerging applications such as edge computing, artificial intelligence, and real-time data processing, which necessitate high-speed and low-power memory solutions [ 7 ] . To overcome these constraints, extensive research has been directed to the development of emerging non-volatile memory technologies that simultaneously address performance and scalability requirements [ 8 ] . Among these, resistive random-access memory (RRAM) has received considerable attention for applications in both high-density memory and neuromorphic computing hardware [ 9 ] . RRAM is a two-terminal, non-volatile device in which data storage is achieved by modulating the resistance state through the application of an electric field [ 10 ] . Compared to conventional Flash memory, RRAM exhibits distinct advantages, including low operating voltage, fast switching speed, scalability to sub- 10 nm technology nodes, and compatibility with three-dimensional integration [ 11 , 12 , 13 ] . A typical RRAM device’s architecture consists of a metal-insulator-metal thin-film (TF) stack, which offers both design simplicity and scalability [ 14 ] . In a vertical stack, the bottom electrode (BE) is typically grounded while a bias is applied to the top electrode (TE) to program the device by altering the resistance state of the insulating-layer, also referred to as the active-layer [ 15 ] . In addition to the conventional vertical configuration, planar RRAM architectures present an alternative geometry that can simplify fabrication and may increase compatibility with flexible substrates [ 16 , 17 , 18 ] . In the planar configuration, the two electrodes are patterned laterally on the same substrate surface, with the active-layer bridging the gap between them. In this geometry, the voltage is applied to one of the electrodes to establish an electric field across the lateral channel, and resistive switching occurs within the region of the active-layer between the electrodes. Unlike charge-based memories, RRAM does not rely on charge storage but instead on the formation and rupture of conductive filaments (CFs) [ 19 , 20 , 21 ] . In filamentary switching, CFs originate from the field-driven migration of oxygen ions or metal cations within the active-layer, producing localized conductive pathways that can be repeatedly formed and ruptured [ 22 , 23 ] . When a metal oxide serves as the active-layer, a positive bias applied to the TE drives oxygen anions toward the electrode, leaving behind a chain of oxygen vacancies that coalesce into a CF [ 24 ] . This oxygen-deficient region places the device in a low-resistive state (LRS) [ 25 ] . A subsequent negative bias reverses this process, driving oxygen anions back into the filament region, where recombination with vacancies ruptures the CF and restores insulating behavior, thereby switching the device into a high-resistive state (HRS) [ 26 , 27 , 28 ] . The initial creation of a CF is referred to as the forming process, while subsequent rupture and reformation correspond to RESET and SET processes, respectively. To prevent permanent breakdown and to control filament dimensions, a compliance current limit (CCL) is applied during forming and SET operations [ 29 ] . The CCL directly determines filament diameter and also the resulting LRS value [ 30 ] . While RRAM has been extensively investigated across various materials and device geometries, comparatively few studies have examined how active-layer architecture influences switching power and memory window characteristics in planar RRAM devices. In this work, planar ITO/WO x /ITO RRAM structures were fabricated with either compact TF active-layers or porous helical active-layers deposited by glancing angle deposition (GLAD). The helical morphology was selected because its porous, interconnected geometry increases surface area relative to TF layers and provides reproducible, tunable parameters for systematic study [ 31 ] . Accordingly, this study aims to (i) determine whether planar RRAM devices exhibit dependence on active-layer thickness across the 50 – 200 nm range, (ii) evaluate the ability of porous helical architectures to sustain operation at reduced CCLs, and (iii) investigate how current scaling influences the memory window in planar devices. Through this comparison of TF and helical architectures, the work seeks to clarify how geometry, rather than thickness, governs the low-power and memory characteristics of planar RRAM, and to assess the potential of helical designs for enabling energy-efficient and flexible memory technologies. 2 Materials and Methods In this study, two devices with differing active-layer architectures were investigated. The first active-layer architecture was a traditional TF layer, while the second active-layer utilized a helical geometry fabricated by a GLAD method. Both device types were fabricated with active-layer thicknesses of 50 , 100 , and 200 nm to confirm that the observed switching behavior was not dominated by thickness effects. Both device architectures used indium tin oxide (ITO) as electrodes with a tungsten oxide (WO x ) active-layer. The ITO/WO x /ITO metal-insulator-metal (MIM) planar configuration was fabricated on a manufactured ITO-coated glass slide substrate (SPI Supplies). While devices in this study were tested in a planar configuration, an ITO-coated glass slide substrate was chosen to facilitate potential future vertical studies and transparent device testing. Before deposition, the substrates were cleaned with deionized (DI) water and acetone, followed by an oxygen plasma treatment in a Harrik Plasma PDC- 32C system.