December 1 - 6, 2024
Boston, Massachusetts
Symposium Supporters
2024 MRS Fall Meeting & Exhibit
EL03.09.01

2D Materials for Neuromorphic Computing Devices

When and Where

Dec 4, 2024
9:15am - 9:45am
Sheraton, Second Floor, Back Bay C

Presenter(s)

Co-Author(s)

Max Lemme1,2

RWTH Aachen University1,AMO GmbH2

Abstract

Max Lemme1,2

RWTH Aachen University1,AMO GmbH2
2D Materials are one of several material classes considered for resistive switching (RS) devices, potential game-changers for neuromorphic computing hardware [1]. Such ”memristive” devices can be switched between two or more resistive states in a volatile or non-volatile way [2], and may be used in computing-in-memory architectures [3], cross-bar arrays [4], or as electronic synapses [5].

Semiconducting molybdenum disulfide (MoS2) has been investigated intensively as a RS material. The physical origin of the switching has been attributed to different mechanisms: bias-induced migration of sulfur vacancies and grain boundaries, lattice distortion, reversible modulation of MoS2 phases, and ion migration. Here, I will discuss different MoS2 device configurations that depend on the material growth conditions and lead to different RS behavior and mechanisms.

Cross-point devices, where two metal electrodes are separated by one or several MoS2 layers oriented perpendicular to the electrodes, are probably the best-studied devices. We show that the choice of electrodes and even the choice of their deposition method can influence the switching behavior and reliability. In particular, metal-organic chemical vapor deposited (MOCVD) MoS2 with sputtered Al top-electrodes show significantly reduced variability, with high yields of up to 95% and competitive endurance, retention, and switching speeds compared to the state-of-the-art [6].

Vertical MoS2 films can be grown under certain conditions by thermal conversion of thin molybdenum films. Here, the van der Waals gaps between the vertical layers facilitate electric-field-driven ion movement [7], leading to non-volatile RS. I will show volatile RS in a device combining a SiOx layer with vertically aligned MoS2. The devices exhibit repeatable threshold switching at low switching voltages below 1 V, switching times below 400 ns, and endurance of close to 6000 cycles [8].
Similar ion transport in van der Waals gaps can be observed in lateral two-terminal devices based on MOCVD MoS2. I will show experiments on silver (Ag) ion migration in lateral MoS2 memristors that do not require higher “forming” voltages for their initial RS event. These Pd/MoS2/Ag devices show volatile switching with ultra-low threshold voltages of 0.3 V and 100-600 ns switching [9].

Furthermore, I will present a study on the current conduction mechanism in memristors with MOCVD-grown hexagonal boron nitride (h-BN) as the active material. The devices were electrically measured in their low and high resistance states in a wide range of temperatures. We propose the formation and retraction of nickel filaments along boron defects as the RS mechanism, with defect-mediated transport in the high resistive state. The devices exhibit low cycle-to-cycle variability of 5% and a large On/Off current ratio of 107 [10].

Finally, I will discuss the parasitic effect of resist residues, which could potentially lead to misinterpretations of RS in 2D materials [11].

All experiments are corroborated with transmission electron microscopy, Raman spectroscopy, and through other analytical means.

This work has received funding from the German BMBF through grants 03ZU1106xx (NeuroSys) and 16ME0399/16ME0400 (NEUROTEC II).

[1] M.C. Lemme et al., Nat Comm 2022, 13, 1392.
[2] D.B. Strukov et al., Nat Electron 2018, 1, 333.
[4] S. Chen et al., Nat Electron 2020, 3, 638.
[5] G. Zhou et al., Adv Electron Mat 2022, 8, 2101127.
[6] D. Braun et al., in Device Research Conference (DRC), College Park, Maryland, USA, 2024.
[7] M. Belete et al., Adv Electro Mat 2020, 1900892.
[8] J. Lee et al., Device Research Conference (DRC), College Park, Maryland, USA, 2024.
[9] S. Cruzes et al., Device Research Conference (DRC), College Park, Maryland, USA, 2024.
[10] L. Völkel et al., Adv Funct Mat 2024, 34, 2300428.
[11] D. Braun et al., 2023, DOI 10.48550/arXiv.2309.13900.

Symposium Organizers

Deji Akinwande, The University of Texas at Austin
Cinzia Casiraghi, University of Manchester
Carlo Grazianetti, CNR-IMM
Li Tao, Southeast University

Session Chairs

Cinzia Casiraghi
Cecilia Mattevi

In this Session