Three-Terminal Synaptic Device for Neural Network Accuracy
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional resistance-variable synaptic devices with a two-terminal structure face challenges in stable operation when applied to three-terminal electrochemical memory-based cross-point arrays due to voltage drop issues, limiting the accuracy of inference and learning operations in artificial neural networks.
Innovation Solution
A three-terminal synaptic device comprising an electrochemical memory (ECRAM) and two transistors, configured in a parallel and series connection, allows for improved accuracy in inference and learning operations by overcoming voltage drop issues through a parallel operation method.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If a half-bias scheme is applied to a three-terminal electrochemical memory-based cross-point array, then voltage pulses can be applied to update synaptic devices, but voltage drop in the channel layer prevents sufficient update
Solution Approach 1:
The synaptic device is segmented into three separate terminals (gate, source, drain) instead of using a conventional two-terminal structure. This segmentation allows independent control of voltage application at each terminal, enabling the gate terminal to receive full voltage pulses for sufficient updates while the source and drain terminals handle current measurement, thereby eliminating the voltage drop problem that plagues two-terminal devices.
Solution Approach 2:
The gate terminal acts as an intermediary that receives the full voltage pulse without suffering from voltage drop. By applying voltage through the gate terminal rather than through the channel current path, the system can update synaptic weights sufficiently without energy loss. The gate terminal mediates between the voltage source and the electrochemical memory, enabling full-voltage operation.
2Adaptability or versatility
If conventional two-terminal synaptic devices are used in a three-terminal cross-point array, then the array structure can be formed, but stable operation is compromised due to voltage drop
Solution Approach 1:
The invention changes the fundamental operating parameters of the synaptic device by transitioning from a two-terminal to a three-terminal configuration. This parameter change allows the device to operate with separate voltage and current paths, eliminating the voltage drop issue while maintaining compatibility with cross-point array structures. The electrochemical memory device operates with gate voltage control rather than direct current through the memory layer.
3Reliability
If three-terminal electrochemical memory is used as synaptic device, then voltage drop issues can be overcome, but device complexity increases
Solution Approach 1:
The three-terminal electrochemical memory device serves multiple functions simultaneously: the gate terminal controls the electrochemical reactions, the source and drain terminals measure current, and the device integrates both memory and transistor functionality in a single structure. This multi-functionality justifies the increased terminal complexity by providing superior voltage stability and eliminating the need for separate control circuitry.
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
The proposed device enhances the accuracy of inference and learning operations in artificial neural networks by enabling stable parallel operations and overcoming voltage drop degradation, particularly when using oxide semiconductor-based transistors.
Implementation Method 1
an operation method of a three-terminal synaptic device for artificial neural network learning... applying stochastic voltage pulses of opposite polarities to a gate electrode and a source electrode of the electrochemical memory... to change conductance of a channel region
Data Source
Figure 1
Figure 2(a)~2(c)
Figure 3
AI summary
The present disclosure relates to a three-terminal synaptic device for artificial neural network learning, a synaptic array using the same, and a method of operating the same. The three-terminal synaptic device includes a first transistor; an electrochemical memory (ECRAM) connected in parallel to the first transistor; and a second transistor connected in series to the parallel structure. Accordingly, the present disclosure can achieve an accuracy improvement of inference and learning operations in an artificial neural network through parallel operation by configuring a cross-point array based on the synaptic device with the three-terminal structure comprised of the electrochemical memory and two transistors.