Design and Optimization of Stable Alloy Anodes Using Experimental Analysis and Artificial Neural Network Models for Thermodynamic Property Prediction

Abstract

Sodium-ion batteries (SIBs) are emerging as a sustainable and cost-effective alternative to lithium-ion batteries (LIBs) due to the abundant availability and low cost of sodium resources. Although LIBs currently dominate the market, concerns about lithium scarcity and increasing production costs have motivated the search for alternatives like SIBs, particularly for large-scale energy storage. Alloy-based anode materials offer promising potential with their high theoretical capacities and favorable redox potentials, but they encounter challenges such as significant volume expansion during sodiation and desodiation cycles, which can limit cycling stability and long-term performance. In this study, several synthesis strategies were developed to address these challenges, focusing on surface engineering, binder optimization, and structural/morphological design to enhance the performance of alloy-based anodes. After extensive evaluation, antimony phosphate (SbPO₄) was selected as the most promising material for this thesis. SbPO₄ demonstrated a capacity of 337 mAh/g at a current density of 1 A/g, and after 100 cycles, retained 295 mAh/g, achieving 87.5% capacity retention with a Coulombic efficiency exceeding 99%. Further optimization showed that SbPO₄/1% MXene achieved the highest specific capacity of 451.7 mAh/g, while Fe₀.₂Sb₀.₈PO₄ recorded 398.3 mAh/g, both exhibiting excellent stability with minimal volume expansion over cycling. These findings highlight the potential of MXene additives and iron substitution in enhancing the capacity and stability of SbPO₄ based anodes. To complement experimental work, machine learning techniques were employed to predict key performance metrics, including specific capacity, volume change, and voltage. Among the developed models, the Deep Neural Network (DNN) achieved the highest predictive accuracy, ranging from 84.5% to 95.4%. The DNN was integrated with the NSGA2 optimization algorithm to identify optimal solutions that balance capacity and volume change, aiding in precise material design. This combination of experimental strategies and predictive modeling contributes to advancing SIB technology, offering practical solutions for large-scale, sustainable energy storage.

Date of Award	12 Dec 2024
Original language	American English
Supervisor	Ali Alhammadi (Supervisor)