# Predictive Modeling of Core–Shell Magnetoplasmonic Nanoparticles: GPR-Based Optimization for Enhanced Photothermal Heating

## TL;DR

This paper studies how core-shell nanoparticles can be optimized for better photothermal therapy using simulations and machine learning.

## Contribution

The study introduces a GPR-based model to optimize core-shell nanoparticle design for enhanced photothermal heating.

## Key findings

- Fe3O4@Au core-shell nanoparticles outperform Au@Fe3O4 in photothermal efficiency.
- Optimal shell thickness ranges from 10 to 30 nm with a core radius of around 80 nm.
- Absorption cross-section (C_abs) most strongly influences photothermal performance.

## Abstract

This study presents
a comprehensive analysis of the photothermal
therapy (PTT) performances of single gold (Au), single magnetite (Fe3O4), and Au@Fe3O4 and Fe3O4@Au core–shell nanoparticles using PyMieLab
and COMSOL Multiphysics simulations. Through a detailed parametric
study, we elucidate how core–shell configurations influence
photothermal conversion efficiency, heat generation, and spatial temperature
distribution. Our results reveal that Fe3O4 nanoparticles
exhibit a higher intrinsic photothermal efficiency than Au; however,
integrating both materials in core–shell structures markedly
enhances thermal responses, with Fe3O4@Au consistently
outperforming Au@Fe3O4. This enhancement arises
from synergistic core–shell interactions that optimize light
absorption and thermal conversion. A Gaussian process regression (GPR)-based
efficiency parameter was introduced to predict optimal geometric combinations
and temperature profiles. The analysis revealed clear design rules;
the effective shell thicknesses ranged from 10 to 30 nm, while the
most efficient core radius was around 80 nm. The highest predicted
temperature elevations occurred at a 4:1 core-to-shell ratio, although
the largest J
o values were observed at
a ratio of 5:1. Among the optical parameters, the absorption cross-section
(C
abs) had the strongest influence on
the photothermal performance, while J
o, despite its limited predictive reliability in the current data
set, remains a promising descriptor for future studies with broader
material combinations. These findings underscore the pivotal role
of nanoparticle design in PTT optimization.