Sharbati, Mohammad Taghi
(2022)
Two Dimensional Dynamic Synapse With Programmable Spatio-Temporal Dynamics For Neuromorphic Computing.
Doctoral Dissertation, University of Pittsburgh.
(Unpublished)
Abstract
In today’s era of big-data, a new computing paradigm beyond today’s von-Neumann architecture is needed to process large-scale datasets efficiently. In response to this need, the field of neuromorphic computing has recently emerged. Inspired by the brain, neuromorphic approaches are better at complex tasks than even supercomputers and show much better efficiency. This is because, unlike modern computers that use digital ‘0’ and ‘1’ for computation, biological neural networks exhibit analog changes in synaptic connections during the decision-making and learning processes. However, the existing approaches of using digital complementary metal-oxide-semiconductor (CMOS) devices to emulate gradual/analog behaviors in the neural network are energy intensive and unsustainable; furthermore, emerging memristor devices still face challenges such as non-linearities and large write noise. Here, we propose a novel artificial synaptic device use of an electrochemical dynamic synapse based on two-dimensional (2D) materials. The synaptic weight (channel conductance) of these dynamic synapses can be tuned via both a long-term doping effect from electrochemical intercalation and a short-term doping effect from ionic gating, thereby demonstrating programmable spatio-temporal dynamics, an essential feature for implementing spiking neural networks (SNNs). The electrical conductance of the channel is reversibly modulated by a concentration of Li ions between the layers of the 2D materials. This fundamentally different mechanism allows us to achieve a good energy efficiency (< 700 aJ per switching event), analog tunability (>5000 non-volatile states), good endurance and retention performances, and a linear and symmetric resistance response. We demonstrate essential neuronal functions such as excitatory and inhibitory synapses, short term and long term plasticity, paired pulse facilitation (PPF), spike timing dependent plasticity (STDP), and spike rating dependent plasticity (SRDP), with good repeatability. Our scaling study suggests that this simple, two-dimensional (2D) synapse is scalable in terms of switching energy and speed. This work can lead to the low-power hardware implementation of neural networks for neuromorphic computing.
Share
| Citation/Export: |
|
| Social Networking: |
|
Details
| Item Type: |
University of Pittsburgh ETD
|
| Status: |
Unpublished |
| Creators/Authors: |
|
| ETD Committee: |
|
| Date: |
16 January 2022 |
| Date Type: |
Publication |
| Defense Date: |
9 September 2021 |
| Approval Date: |
16 January 2022 |
| Submission Date: |
26 October 2021 |
| Access Restriction: |
No restriction; Release the ETD for access worldwide immediately. |
| Number of Pages: |
157 |
| Institution: |
University of Pittsburgh |
| Schools and Programs: |
Swanson School of Engineering > Electrical and Computer Engineering |
| Degree: |
PhD - Doctor of Philosophy |
| Thesis Type: |
Doctoral Dissertation |
| Refereed: |
Yes |
| Uncontrolled Keywords: |
Neuromorphic Computing, Artificial Synapse, Two Dimensional Dynamic Synapse, Graphene Synapse, MoS2 Synapse, Electrochemical Intercalation, Li Intercalation, Electric Double Layer Gating Effect, Ionic Gating, Synaptic Plasticity, Short Term Potentiation (STP), Long Term Potentiation (LTP), Paired Pulse Facilitation (PPF), Spike Timing Dependent Plasticity (STDP), Spike Rating Dependent Plasticity (SRDP), Spiking Neural Network (SNN), 2D material, ECRAM, Electrochemical Intercalation, Li Intercalation, Ionic Gating, Electric Double Layer Gating, Fast Speed Artificial Synapse, Low Energy Artificial Synapse, Synaptic Device, Event-Based Spatio Temporal Pattern Recognition, Programmable Spatio-Temporal Dynamics. |
| Date Deposited: |
16 Jan 2022 17:29 |
| Last Modified: |
16 Jan 2022 17:29 |
| URI: |
http://d-scholarship.pitt.edu/id/eprint/41885 |
Metrics
Monthly Views for the past 3 years
Plum Analytics
Actions (login required)
 |
View Item |
![[feed]](/images/feed-icon-32x32.png){kind=icon}