National and Kapodistrian University of Athens
Optimization of nanomaterials for energy applications with simulation and machine learning
Abstract
dc:descriptionThis thesis addresses the scientific question of the optimization of width-modulated nanowaveguides (nWVGs) for thermoelectric (TE) applications. Among energy applications, TE applications, which are based on the direct conversion of heat into electricity, occupy a special position because they could serve the current needs of our society for green energy, power the Internet of Things, cool nanoelectronic circuits, and improve the performance of quantum applications. Width-modulated nWVGs is a class of nanomaterials promising for micro-TE generators and micro-TE coolers because their transport properties can be geometrically tuned for high TE efficiency. In the last decade, they have attracted increasing research interest stimulated by theoretical predictions for exceptional performance. However, there are many open questions related to their properties and their operation that need to be addressed before they can find applications. Stimulated by this need, we performed systematic research with calculations and machine learning to understand and optimize electrons and phonons transport properties in width-modulated nWVGs. To validate the potential of modulated nWVGs, a better understanding of the effect of width modulation on the TE efficiency is needed. So far, it is considered that (a) the TE efficiency increases monotonically with increasing width-mismatch due to decreasing phonon thermal conduction taking place without significantly affecting electron transport, (b) width-mismatch dominates the effect of width-modulation in transport, and (c) phonons play the main role in increasing the TE efficiency. Here, we demonstrate counterevidence based on an investigation of the effect of width-modulation on electrons so far overlooked. We reveal that (a) the TE efficiency varies non-monotonically with the modulation mismatch due to quantum effects on electron transport, (b) the modulation mismatch is quantified by the size mismatch of the modulation rather than by the width mismatch, and (c) it is electrons rather than phonons that play the main role in optimizing width modulation for maximum TE efficiency when quantum effects dominate. Our findings indicate that research should reorient from large width mismatch towards optimal modulation mismatch for width-modulated nanostructures with enhanced TE efficiency in the quantum confinement regime. We reach these conclusions with simulations of electrons and phonons in nWVGs and extensive calculations on their transport properties varying the characteristic confinement dimensions of the quantum dot (QD) modulation unit. Electron and phonon transport properties were calculated using Landauer formalism in terms of the transmission coefficient determined through the calculation of the scattering matrix within a continuum theory for the description of the energy states. We analyze the calculations, identify dependencies, provide physical interpretation, and optimize width modulation for maximum TE efficiency. We first consider a single QD modulation unit and vary systematically all characteristic confinement dimensions to investigate the effects of the width mismatch, the length mismatch, and the overall modulation mismatch on the TE efficiency. Optimization of the single-modulation-unit nWVG is achieved for minimum phonon conduction, maximum TE power factor (TPF), and maximum TE figure of merit ZT. Next, we investigate the effect of increasing modulation degree considering the cases of (a) a two-QD modulation, and (b) multiple-QD modulation in periodic and aperiodic sequences. We show that the thermoelectric efficiency depends on the coupling between the modulation units and the interplay between periodicity and aperiodicity in the modulation profile. It becomes evident that maximization of the TPF is for periodic width modulation whereas maximization of the TE efficiency is for aperiodic width modulation profiles. Adding more modulation units in the modulation profile increases electron transport properties and the TPF. The increase is the maximum for the periodic superlattice (p-SL) modulation profile. The phonon thermal conductance decreases with increasing modulation extent reaching the p-SL value for periodic modulation. The ZT increases significantly with increasing modulation up to the p-SL value. A further increase is possible at quasi-localized states in aperiodic modulation profiles where phonon thermal conductance decreases below the p-SL value and the TPF is comparable to or higher than the p-SL value. The results reveal that maximization of the TPF is for periodic width-modulation whereas maximization of the TE efficiency is for aperiodic width-modulation profiles that create quasi-localized states for electrons. We proceed with the optimization of aperiodicity in the modulation profile with calculations and machine learning, more specifically Bayesian optimization. We point out the non-trivial role of disorder in decreasing thermal conduction in the presence of non-uniformity. Machine learning optimization of hetero-SLs for minimum coherent phonon conduction showed optimal aperiodicity for moderate disorder against physics intuition. Importantly, we show optimal aperiodicity in width-modulated nWVGS for maximum disorder according to a physics rule. Optimizing aperiodic geometry modulation is particularly challenging due to the enormous possible configurations. We set up a feasible optimization problem removing unnecessary complexity and demonstrated efficient Bayesian optimization. Our results confirm that minimum thermal conductance occurs for the most disordered arrays of modulation units; the degree of disorder being quantified by the number of non-identical modulation units. We further explore the mechanisms underlying the optimization of width-modulated nWVGs and demonstrate generic behaviour. The results support that the distribution of the thermal conductance among the aperiodic width-modulation configurations is controlled by the modulation degree irrespective of choices of constituent material, width-modulation geometry, and composition constraints. The efficiency of Bayesian optimization is evaluated against increasing temperature and sample size. It decreases with increasing temperature due to the thermal broadening of the thermal conductance distribution and shows weak dependence on temperature in samples with high discreteness in the distribution spectrum. Finally, a two-step methodology is applied to design optimized width-modulated nWVGs for minimum phonon conduction combining optimization of modulation mismatch with simulations and optimization of modulation profile with Bayesian optimization. The two-step methodology is shown suitable to deal with counter-dependencies and capable of leading to optimization of the nWVG-modulation for optimal electron and phonon transport properties and maximum TE efficiency. In this thesis, optimization of the width-modulation profile of nWVGs is accomplished with systematic research, extensive simulations, and machine learning techniques. The thesis provides new physics insight and reveals novel effects, non-trivial evidence, validation of a physics optimization rule, and design guidelines for efficient energy conversion at the nanoscale due to quantum confinement effects. The new knowledge should be a useful contribution to the currently evolving research on thermoelectric metamaterials for efficient energy applications.
Degree
thesis:*- Grantor dc:publisher
- National and Kapodistrian University of Athens
- Year dc:date
- 2024
Author and committee
dc:creator, dc:contributor.*- Authors dc:creator
-
- Stefanou, Antonios-Dimitrios
- Στεφάνου, Αντώνιος-Δημήτριος
Subjects
dc:subject × 18- Θερμοηλεκτρικά
- Θερμική αγωγή
- Νανοδομές
- Κυματοδηγοί
- Μηχανική μάθηση
- Thermoelectric
- Thermal conduction
- Nanostructures
- Waveguides
- Machine learning
- Νανοτεχνολογία
- Επιστήμες Μηχανικού και Τεχνολογία
- Νανοεπιστήμη και Νανοτεχνολογία
- Νανοϋλικά
- Nano-Technology
- Engineering and Technology
- Nanoscience and Nanotechnology
- Nano-materials
Rights
- Language dc:language
- eng
Identifiers
dc:identifier.*- Identifier
- 10.12681/eadd/56609