Programmable Multilevel Memtransistors Based on van der Waals Heterostructures
Abstract
Neuromorphic computing that mimics the energy‐efficient cortical neural network in the human brain is attractive because of its possibility to process complex and massive data sets and achieve fast computing capability. Herein, a heterosynaptic and programmable memtransistor architecture with high computing functionality is reported by monolithically integrating a hexagonal boron nitride (h‐BN) memristor with a molybdenum disulfide (MoS2) transistor. Memristors consisting of a vertically stacked van der Waals materials (multilayer graphene (MLG) and h‐BN) exhibit a stable bipolar resistive switching behavior with a memory window more than three orders of magnitude due to the formation and rupture of the metallic filament within the h‐BN layer. By controlling the resistance state of the h‐BN memristor, the behaviors of the memtransistor can be programmed with a high switching ratio of ≈104, showing ≈16 pW standby power consumption. A multistate computing window and tunable current on/off ratio can be achieved by controlling the synaptic weight of the memristor, demonstrating that the presented 2D architecture can be exploited as a logic inverter device. The results pave the way toward the development of highly functional neuromorphic systems for the next‐generation in‐memory computing.
Document Details
- Document Type
- Pub Defense Publication
- Publication Date
- Jul 24, 2019
- Source ID
- 10.1002/aelm.201900333
Entities
People
- Hyunik Park
- Jihyun Kim
- Marko J. Tadjer
- Michael A. Mastro
Organizations
- Korea Institute of Energy Technology Evaluation and Planning
- Korea University
- National Research Foundation of Korea
- Office of Naval Research
- United States Naval Research Laboratory