Mechanisms of Resistive Switching in 2D Monolayer and Multilayer Materials

TL;DR

This paper reviews resistive switching in two-dimensional layered materials (2DLM), focusing on monolayer and defect-localized mechanisms and how electrode contacts shape operation. It synthesizes experimental observations and ab initio simulations to map energy landscapes for switching pathways in hBN and TMDCs, revealing atomistic processes such as metal-atom adsorption and vacancy-mediated transitions that enable ultra-low power, highly scalable memory devices. The authors highlight remarkable HRS/LRS ratios (up to ~$10^{11}$) and switching energies in the fJ–aJ range, discuss crossbar and 3D-stack demonstrations, and emphasize defect engineering, contact control, and monolayer advantages as key enablers. They argue that 2DLM RS devices hold promise for ultra-dense memory, neuromorphic computing, and RF switching, but require advances in reproducibility, defect control, and integration strategies to compete with mature oxide-based approaches.

Abstract

The power and energy consumption of resistive switching devices can be lowered by reducing their active layer dimensions. Efforts to push this low-energy switching property to its limits have led to the investigation of active regions made with two-dimensional layered materials (2DLM). Despite their small dimensions, 2DLM exhibit a rich variety of switching mechanisms, each involving different types of atomic structure reconfigurations. In this review, we highlight and classify the mechanisms of resistive switching in mono and bulk 2DLM, with a subsequent focus on those occurring in a monolayer and/or localized to point defects in the crystalline sheet. We discuss the complex energetics involved in these fundamentally defect-assisted processes, including the co-existence of multiple mechanisms and influence of the contacts used. Examining the highly localized 'atomristor'-type switching, we provide insights into the atomic motions and electronic transport across the metal-2D interfaces underlying their operation. Finally, we present the progress and our perspective on the challenges associated with the development of 2D resistive switching devices. Promising application areas and material systems are identified and suggested for further research.

Figures (4)

• Figure 1: Operation, structure, and mechanisms of 2DLM RS devices. A schematic pinched-hysteresis current vs. voltage (I-V) curve resulting from resistive switching (RS) device measurement is shown in (a)/(b) for bi/nonpolar switching characteristics, indicating the non-volatile transitions from the High Resistance State (HRS) to the Low Resistance State (LRS) and back. The device structure usually consists of a switching layer sandwiched between two contacts. The switching layer can be realized with 2DLM in (c) multilayer or (d-e) monolayer forms, and in (d) vertical or (e) lateral configurations. The black arrows in each case indicate the direction of current flow across the 2DLM. Schematics of the switching mechanisms are shown in (f) lateral and (g-h) vertical 2DLM devices. The mechanisms are sorted by their scaling potential, both vertical (limited to multilayer, or capable of existing in monolayers) and lateral (requires extended area for switching, or can be localized to a point). The light shaded areas highlight the conductive areas once the device is switched to its LRS. In all cases, the 2DLM has the structure of a transition metal dichalcogenide or hexagonal boron nitride, and is representative of diverse 2DLM. The mechanisms shown have been reported experimentally, for example in Refs (f - grain boundary reconfiguration) Sangwan2018, (f - metal filament) Farronato2022, (f - schottky barrier modulation) Huang2021, (g - metal filament) Chen2020, (g - vacancy filament) Yan2019, (g - defect bridging) Ducry2022Mao2022, (g - Schottky-barrier modulation) Pam2022, (h - conductive metal substitution) Hus2020, (h - ferroelectric polarization) Li2020, (h - phase change) Hou2019 ).
• Figure 2: Structural transitions leading to resistive switching in hBN and TMDC devices. (a) Cross-sectional TEM image of a Ti/hBN/Cu device, indicating the location of structural distortions in the hBN multilayer. The scale bar is 1.5 nm. A conductive AFM map of the surface of this structure is shown in (b) (scale bar is 250 nm). (a)-(b) are adapted from Ref Chen2020. (c) Molecular structures of different interlayer bridges formed across hBN layers, taken from Ref Strand2019. Similar bridge structures have been observed in (d) TEM images of hBN RS devices with atomically flat contacts, as highlighted by the yellow dashed lines in (e) after switching to the LRS (adapted from Ref Mao2022). (f)-(g) Scanning tunnelling microscope (STM) images of a RS mechanism in which a single Au atom from the STM tip substitutes into a sulfur divacancy in an MoS$_2$ monolayer, leading to (f) HRS and (g) LRS states, adapted from Ref Hus2020. (h)-(i) shows the existence of Ag contact migration into the active layer of a lateral MoS$_2$ device (h) before and (i) after switching, adapted from Ref Farronato2022. (j) TEM imaging of tungsten and sulfur vacancies in WS$_2$, adapted from Ref Yan2019.
• Figure 3: Resistive switching mechanisms at the electrode/monolayer 2H-MoS$_2$ interface. (a) TEM image of the Au-2H MoS$_2$ interface in which the Au (111) surface termination is clearly visible, adapted from Ref Ge2017. Schematics representing (b) Au metal interstitial creation, (c) Au adsorption into a Sulfur vacancy site, and (d) adsorption of an Au adatom at the interface into a Sulfur vacancy site are shown on top of plots representing the energy surfaces along the minimal-energy pathways for each of these processes. An ab initio Nudged Elastic Band (NEB) approach was used for that purpose JNSSON1998, considering the (111) crystallographic plane of Au. In (e-f) the mechanism of metal adsorption from (c) is further explored for (e) the (100) crystallographic plane of Au and (f) the (111) plane of Ag. In each case, the reaction coordinate enumerates seven states found during the simulation of the minimum-energy trajectory. All calculations are performed with initial and final images corresponding to the shortest path available for the mobile species to complete such a transition. (g-i) Electronic transport mechanisms across the vertical Au/monolayer-MoS$_2$/Au stack in case of (g) a pristine monolayer, (h) a monolayer with intrinsic defects, such as Sulfur vacancies, and (i) a monolayer with an adsorbed Au atom. The wavy arrow indicates a phonon-assisted transition to a higher energy level, leading to enhanced Schottky emission.
• Figure 4: Device performance and scaling trends for 2DLM RS vertical device stacks made with mono- or multilayer active areas. (a)-(c) show existing device metrics for HRS/LRS ratio, operating voltage (SET Voltage), and HRS current (leakage current) for (a)-(b) monolayer and (c) multilayer 2DLM devices. In each case, the OFF-current is read at a voltage of V$_{SET}$/2, and the ON current corresponds to the compliance current used in the measurements. The size of each marker in (a-b) refers to the area of the corresponding device, which ranges from 0.03 $\mu$m$^2$ to 0.20 mm$^2$. The legend (electrode metal) are the same for (a)-(b). The data is colored according to the electrode metal. It is tabulated from Refs Wu2020Ge2017Ge2020Zhao2018-1Wu2019Kim2018Ge2018Kim2020Mao2022Yang2023-2Yang2024-2 (monolayer) and Refs He2012Cheng2015Das2019Feng2019Li2018-8Kim2018Qian2016Zhang2016-5Zhou2016-7Han2017Das2019-2Rehman2017Zhang2018Wang2018Chen2023-2Pam2022Aggarwal2023Zhuang2023-2Puglisi2016Shi2018Vlkel2023Lu2021Li2021-3Yin2022Li2021Hou2023Zhang_2018Yang2023 (multilayer). (d) Variation in the RESET and SET Voltages for 48 Au/multilayer hBN/Au device from a single crossbar array, using a compliance current of 1 mA Chen2020. (e) Thickness vs. operating voltage for MoTe$_2$ and several Mo$_{1-x}$W$_x$Te devices (x = 0.03 - 0.09), adapted from Ref Zhang2018. The black dashed line serves as a guide to the eye. (f) Change in HRS and LRS resistance with device area for Au/monolayer MoS$_2$/Au devices Ge2017. The inset shows the effect of area scaling on the SET Voltage for the multilayer hBN stack in Ref Shi2017.