Files in this item

FilesDescriptionFormat

application/pdf

application/pdfJO-DISSERTATION-2018.pdf (14MB)Restricted to U of Illinois
(no description provided)PDF

Description

Title:Analysis and simulation of carbon-based thermal interface material
Author(s):Jo, Michael Kim
Director of Research:Ravaioli, Umberto
Doctoral Committee Chair(s):Ravaioli, Umberto
Doctoral Committee Member(s):Lyding, Joseph W; Aluru, Narayana R; Schutt-Aine, Jose E
Department / Program:Electrical & Computer Eng
Discipline:Electrical & Computer Engr
Degree Granting Institution:University of Illinois at Urbana-Champaign
Degree:Ph.D.
Genre:Dissertation
Subject(s):Dynamic Thermal Interface Material
Percolation Theory
Finite Difference
Machine Learning
Genetic Algorithm
Carbon Nanotube
Graphite
Abstract:Dynamic-Thermal Interface Material (D-TIM) is an adaptive thermal interface material with thermal conductivity proportionally increasing with system temperature. This thesis introduces a percolation theory-based simulation to investigate this unique trend. Brief reviews of the important aspects of carbon-based thermal interface material and the concept of percolation theory are discussed. In order to capture the thermal conductivity enhancement with temperature, we employ thermoelectric effect and variable-range hopping as the governing equations of the simulation. The complete framework of D-TIM percolation simulation and simulation results are presented. On the other hand, machine learning has been successfully adopted by the industry since current computation techniques can manage massive calculations for such algorithms. However, most of these successes are based on software, computer architecture and circuit design level. On the contrary, machine learning for analysis of data in micro-nanotechnology has not been widely applied, resulting in ample space for researchers to extract inference not only from experimental data but also from simulation data. Here, we present examples of machine learning algorithms applied to experiment data. After a brief review of machine learning algorithms, we present three complete projects as examples. The first example demonstrates a technique to extract geometric properties from Raman spectra of few-layer graphenes. Secondly, we present how to detect Chevron Graphene Nanoribbon (CGNR) from scanning tunneling spectroscopy data and calculate the orientation autonomously. Lastly, the third project is to remove silicon substrate effect from the measured current imaging tunneling spectroscopy in order to extract the pure local density of states of the CGNR. Then, we use machine learning, genetic algorithm in this problem, for parameter optimization of D-TIM simulation. We discuss the optimized parameters of D-TIM percolation simulation and in turn analyze the mass ratio of nano-graphite and multi-wall carbon nanotube to improve the effective thermal conductivity of D-TIM. Finally, we analyze the mass ratio and surface roughness effect on D-TIM performance.
Issue Date:2018-07-11
Type:Thesis
URI:http://hdl.handle.net/2142/101705
Rights Information:Copyright 2018 Michael Jo
Date Available in IDEALS:2018-09-27
Date Deposited:2018-08


This item appears in the following Collection(s)

Item Statistics