Advanced control methodologies for atmospheric pressure plasmas used for materials synthesis

Noble metal nanoparticles exhibit pronounced plasmonic features that enable a wide
range of applications in medicine, catalysis, nanoelectronics, imaging and sensor
technology. However, the commercial adoption of these nanoparticles is hampered by
difficulties in reproducibility, stability and precise control of parti cle size, shape and
dispersity. Cold atmospheric pressure plasma (CAP) has emerged as a promising
synthesis method that offers unique ways to generate nanoparticles due to its ability to
initiate complex chemical reactions without conventional reducing agents or stabilisers.
Despite considerable advancements, CAP synthesis of noble metal nanoparticles
continues to experience reproducibility issues due to the inherent non-linearity and
sensitivity of plasma processes. Slight variations in voltage, gas flow or electrode
positioning can significantly alter the discharge dynamics, leading to considerable
variations in nanoparticle properties. In addition, the volatile and highly reactive species
formed at the interfaces between plasma and liquid make precise control and scale-up
difficult.
In this work, it is hypothesised that the integration of machine learning (ML) methods
with real-time diagnostics can effectively model, predict and control the stochastic
dynamics of CAP processes, enabling reproducible, tailored nanoparticle synthesis. T o
prove this suitable CAP electrode geometries were evaluated, synthesis parameters to
nanoparticle outputs were mapped, ML-based control strategies tailored to CAP
optimisation were evaluated and ML-driven CAP synthesis with control of nanoparticle
properties was illustrated.