Back to search

Democritus University of Thrace (DUTH)

Design, modelling and simulation of memristor circuits towards conventional and unconventional computing

Abstract

dc:description

The increase in big data, in part due to the Internet of Things, the rise of quantum computing and the flourishing of machine learning are driving the development of increasingly data-centric computing, which poses a challenge to traditional digital computers based on the von Neumann architecture. In these computers, the frequent movement of data between separate processing and memory units results in significant drawbacks in terms of energy efficiency and data bandwidth. This limitation is known as the von Neumann bottleneck. Furthermore, the disparity in speed between the memory unit and the central processing unit, referred to as the memory wall, further exacerbates this bottleneck. As a result, there is a need for entirely different computing paradigms that can enhance the area and energy efficiency to meet the growing demands for computing power in the era of big data. The aim of this doctoral thesis is the development of hybrid unconventional computational circuits using nanoelectronic non-volatile memristor-based memory arrays. The proposed circuits aim to perform storage and processing of information in the same medium, performing unconventional computations and addressing the known problems of current computational architecture, specifically the memory and energy wall. Conductive bridge random access memory (CBRAM) type memristors, as emergent nanoelectronic devices, have been successfully fabricated and used in non-conventional and neuromorphic computing systems in the last years. Several behavioral or physical based models have been developed to explain their operation and optimize their fabrication process parameters. Among them, the resistance switching of the insulating layer due to the formation of conductive filaments is the most well respected and experimentally proven. All existing memristor models offer trade-offs between accuracy, universality and realism, but, none of them is purely characterized as quantum mechanical, despite the fact that quantum mechanical processes constitute a major part of the memristor operation. Consequently, in the first phase of this PhD thesis, quantum mechanical methods has been employed to develop a complete and accurate filamentary model for the resistance variation that governs the memristor's operating cycle. The model has been fitted to fabricated devices of the Institute of Microelectronics/NCSR Demokritos. Following the microscopic model, a macroscopic compact memristor model that utilizes the drift, diffusion and thermo-diffusion effects has been developed. The objective of this model is to achieve a realistic implementation of the switching mechanisms of a memristor device, while minimizing the overhead on computing resources during simulation and most significantly, being compatible with the already available and well established circuit design tools. The model incorporates the effect of thermo-diffusion on the switching pattern, providing a different perspective to the ionic transport processes, what enable the unipolar type of switching. The model has been fitted to fabricated devices of the National Technical University of Athens. Utilizing this compact memristor model, the objective was to design memristive circuits capable of performing quantum computations. Quantum computing is a compelling new technology that is becoming increasingly practical as challenges regarding the construction of suck quantum systems are being solved in an unprecedented pace. Quantum computers have the potential to solve problems of great complexity and magnitude across many different industries by utilizing quantum specific developed algorithms. However, creating a working quantum computer is still no easy task. This hinders the further development, implementation and testing of quantum algorithms. Initially, the representation of qubits using memristors was achieved. Subsequently, a circuit based on memristor crossbar was designed, capable of reproducing the action of quantum gates on qubits, resulting in the execution of quantum computations. This circuit was further enhanced to incorporate a closed-loop CBRAM crossbar system for the execution and acceleration of quantum algorithms. This circuit supports a universal set of quantum gates representation, and by exploiting its reprogramming capabilities and feedback loop, can compute any quantum algorithm. A comparative analysis of power and speed between the proposed circuit and conventional hardware such as CPU, GPU, and FPGA has been conducted, demonstrating the high efficiency and performance gains. Furthermore, a framework has been developed that supports the design of custom closed-loop memristive crossbars through a graphical user interface (GUI), providing the user with the capability to execute quantum algorithms while having a complete overview of the programming and computations of the circuit, assisting with the realization of a hardware prototype.

Degree

thesis:*
Grantor dc:publisher
Democritus University of Thrace (DUTH)
Year dc:date
2023

Author and committee

dc:creator, dc:contributor.*
Authors dc:creator
  • Fyrigos, Iosif-Angelos
  • Φυρίγος, Ιωσήφ-Άγγελος

Subjects

dc:subject × 16

Rights

Language dc:language
eng

Identifiers

dc:identifier.*
Identifier
10.12681/eadd/58037

Chain of custody

source
Harvested from
Greek National Archive of PhD Theses
Base URL
phdtheses.ekt.gr/eadd_oai/request
Last updated
2026-11-18
Source record
OAI-PMH GetRecord
citation

Fyrigos, Iosif-Angelos; Φυρίγος, Ιωσήφ-Άγγελος. Design, modelling and simulation of memristor circuits towards conventional and unconventional computing. Democritus University of Thrace (DUTH), 2023. http://hdl.handle.net/10442/hedi/58037