Synergistic effects of copper nanoparticles and multi-walled carbon nanotubes on dielectric behaviour of poly methyl methacrylate-based hybrid nanocomposites with machine learning insights

Abstract

Polymer-based dielectric nanocomposites have attracted considerable attention for electronic applications; however, achieving enhanced dielectric performance while maintaining electrical insulation remains a significant challenge due to inadequate filler dispersion and limited understanding of interfacial polarization mechanisms. To address this issue, the present study investigates the dielectric behaviour and machine learning-assisted prediction of poly (methyl methacrylate) (PMMA)/multi-walled carbon nanotube (MWCNT)/copper (Cu) hybrid nanocomposites fabricated using a solution mixing and solvent evaporation approach. The synergistic interaction between conductive MWCNTs and Cu nanoparticles was explored to enhance interfacial polarization, conductive pathway formation, and dielectric performance while preserving the insulating characteristics of the PMMA matrix. The nanocomposites were fabricated using a solution mixing and solvent evaporation method with a fixed 5 wt% MWCNT loading and varying Cu nanoparticle concentrations (1, 1.5, and 2 wt%). Structural characterization using X-ray diffraction confirmed the coexistence of amorphous PMMA, graphitic MWCNTs, and crystalline Cu phases without significant lattice distortion. Dielectric spectroscopy revealed strong frequency-dependent behavior, where dielectric constant and dielectric loss decreased with increasing frequency due to reduced interfacial polarization and dipolar relaxation effects. Among all compositions, the composite containing 1 wt% Cu exhibited the highest dielectric response and AC conductivity, indicating optimized interfacial charge transport and conductive pathway formation. Higher Cu loadings resulted in reduced dielectric performance because of charge carrier screening and partial disruption of conductive networks. Despite the incorporation of conductive fillers, all composites maintained low AC conductivity, preserving their insulating characteristics. To further validate the experimental observations, Random Forest (RF) and Extra Trees (ET) machine learning models were employed to predict dielectric properties. Both models demonstrated excellent predictive capability, particularly for AC conductivity and loss tangent, with the ET model showing slightly superior generalization performance. The combined experimental and machine learning-assisted analysis demonstrates that optimizing the synergistic interaction between MWCNTs and Cu nanoparticles enables improved dielectric performance while maintaining the insulating characteristics of the PMMA matrix. These findings provide valuable guidelines for the design and optimization of polymer-based dielectric nanocomposites for applications in flexible electronics, electromagnetic interference (EMI) shielding, sensors and advanced biomedical electronic components.