Complementary ferroelectric logic

DWPI Title: Method for solving discrete optimization problem i.e. MAXCUT problem, in neuromorphic device, involves outputting approximate solution to discrete optimization problem according to correlated activity variables by output leaky integrate and fire neuron
Abstract: A new form of logic and processing functionality, complementary ferroelectric logic (CFL), which employs ferroelectric tunnel junction (FTJ) devices as the underlying building blocks, is disclosed. By interconnecting the FTJ devices, one can implement various logical functions and memory devices. Of particular note is that the memory devices retain their written states even when power is removed, i.e., CFL memory is nonvolatile, due to the polarization retention properties of the ferroelectric material in the FTJ devices. Exemplary CFL logic functions include inverter (NOR), NAND, and multi-input comparison circuits. The CFL circuits include a writing pulse step to store either a logic “0” or “1” state, followed by a low voltage read pulse to output the implemented logic function. Due to the CMOS-compatible processes required to fabricate CFL circuits, CFL may be heterogeneously integrated with standard CMOS circuitry.
Use: Method for solving discrete optimization problem i.e. MAXCUT problem, in a neuromorphic device (claimed).
Advantage: The method enables a circuit to solve the MAXCUT problem entirely within the circuit without requiring any external preprocessing, and demonstrating capacity of neuromorphic circuits driven by intrinsic randomness to solve parallel computationally relevant problems.
Novelty: The method involves assigning weights (116) to connections (114) between random noise generators and leaky integrate and fire (LIF) neurons, where the weights are chosen to produce a correlation matrix (130) determined by discrete optimization problem. Random bits (106) generated by the random noise generators are integrated by the LIF neurons according to the assigned weights to produce specified correlated activity variables (112). The correlated variables are fed to an output LIF neuron (118) that operates according to plasticity rule. An approximate solution (120) is output to the discrete optimization problem by the output LIF neuron according to the correlated activity variables.
Filed: 12/21/2023
Application Number: US18391785A
Tech ID: SD 16326.1
This invention was made with Government support under Contract No. DE-NA0003525 awarded by the United States Department of Energy/National Nuclear Security Administration. The Government has certain rights in the invention.
Data from Derwent World Patents Index, provided by Clarivate
All rights reserved. Republication or redistribution of Clarivate content, including by framing or similar means, is prohibited without the prior written consent of Clarivate. Clarivate and its logo, as well as all other trademarks used herein are trademarks of their respective owners and used under license.