Full metadata record
| DC Field | Value | Language |
|---|---|---|
| dc.contributor | Department of Applied Physics | en_US |
| dc.contributor.advisor | Chai, Yang (AP) | en_US |
| dc.creator | Chen, Hongye | - |
| dc.identifier.uri | https://theses.lib.polyu.edu.hk/handle/200/13776 | - |
| dc.language | English | en_US |
| dc.publisher | Hong Kong Polytechnic University | en_US |
| dc.rights | All rights reserved | en_US |
| dc.title | Two-dimensional material-based energy-efficient electronic devices towards neuromorphic computing | en_US |
| dcterms.abstract | The human brain is considered a highly parallel and energy-efficient computing system, able to process complex cognitive tasks. Compared to the human brain, artificial neural networks (ANN) based on transistors and random access-resistive memories (RRAMs) face critical challenges with energy efficiency in the way of further scaling of neural network. Two-dimensional (2D) materials, including transition metal dichalcogenides (TMDs), hexagonal boron nitride (h-BN), and CuInP2S6, have been regarded as one of the potential candidates for low-power and high-performance electronic devices in neuromorphic computing. Differing from bulk materials, the atomically thin 2D materials could achieve low-temperature fabrication, low-power switching, and electrostatic gate tunability. This thesis focuses on enhancing energy efficiency through the design and demonstration of devices utilizing 2D materials, including (1) selectors to suppress leakage current in RRAM array; (2) dual-gate transistor based on high-k dielectric as dendristor to calculate the synaptic input and achieve a complete dendritic artificial neuron network. | en_US |
| dcterms.abstract | The first work focuses on the suppression of leakage current in RRAM array. The RRAM array exhibits great potential for information storage and computing due to its small cell area (~ 4F2). Nevertheless, in the RRAM array, there exists the sneak current through unselected low resistance state cells when the unselected word lines and bit lines are biased with half of the operation voltage in the V/2 scheme, which will contribute to energy consumption and useless data. To address this issue, we introduce the two-terminal highly nonlinear MoS2/WSe2/MoS2 n-p-n selector to suppress the sneak current, which performs a high current density of 2 × 103 A cm-2 due to the low Schottky barrier height of Au/MoS2 while maintaining a high nonlinearity of above 200 based on the punch-through mechanism. Furthermore, we successfully integrate this n-p-n selector with bipolar h-BN memory and demonstrate a two-terminal all 2D material-based 1S1R architecture, whose maximum crossbar size was estimated to be 6.5 Kbit. This work contributes to a framework for advancing 3D crossbar array memory devices. | en_US |
| dcterms.abstract | Secondly, the artificial dendrite device was proposed to complete dendritic artificial neuron model through the nonlinear computation ability of dendrite. Neuromorphic computation in ANN typically uses a point neuron model, which only consists of synapses and soma and ignores the computational function of dendrites. In biology, dendrites nonlinearly integrate the synaptic inputs and control the somatic membrane potential, which plays a key role in the accuracy and energy-efficiency improvement of neuron network. Therefore, a more complete neural network with artificial dendrite has been proposed. The Cu0.67Ag0.33InP2S6 (CAIPS) as 2D high-k dielectric and MoS2 as channel was used in the gate-all-around transistor, termed as dendristor. The adaptation to van der Waals layered materials and dual-gate design enables the dendristor with high performance, resulting from the clean interface between CAIPS and MoS2. Meanwhile, this device performs superlinear/sublinear integration of synaptic input and synaptic plasticity with low energy consumption. We further introduce the dendristor device into the neuron system and achieve multisensory perception. | en_US |
| dcterms.abstract | In conclusion, we investigate the 2D material-based energy-efficient electronic devices towards neuromorphic computing. With the development of modern electronics with high performance, our work will provide new insight into the development of energy-efficient electronic devices for neuromorphic computing. | en_US |
| dcterms.extent | xvi, 91 pages : color illustrations | en_US |
| dcterms.isPartOf | PolyU Electronic Theses | en_US |
| dcterms.issued | 2025 | en_US |
| dcterms.educationalLevel | Ph.D. | en_US |
| dcterms.educationalLevel | All Doctorate | en_US |
| dcterms.accessRights | open access | en_US |
Copyright Undertaking
As a bona fide Library user, I declare that:
- I will abide by the rules and legal ordinances governing copyright regarding the use of the Database.
- I will use the Database for the purpose of my research or private study only and not for circulation or further reproduction or any other purpose.
- I agree to indemnify and hold the University harmless from and against any loss, damage, cost, liability or expenses arising from copyright infringement or unauthorized usage.
By downloading any item(s) listed above, you acknowledge that you have read and understood the copyright undertaking as stated above, and agree to be bound by all of its terms.
Please use this identifier to cite or link to this item:
https://theses.lib.polyu.edu.hk/handle/200/13776