Description
Low-temperature plasmas enable electron-driven plasma NH3 synthesis, yet a quantitative understanding of microphysical characteristics relative to reactor-scale performance remains elusive. Building on a systematic experimental N₂–H₂ dataset spanning varied reactor configurations and operating conditions, we developed a machine-learning-based framework that links plasma microphysics to macroscopic ammonia synthesis, using electrical diagnostics (Lissajous analysis, micro-discharge statistics), optical emission spectroscopy, and colorimetric NH3 quantification. These charge and excitation descriptors, along with varied operational conditions, were used to train a Random Forest model. The model reveals that total charge per pulse and median micro-discharge energy are the most informative descriptors of the plasma characteristics, which are governed by barrier thickness, gap geometry, voltage, and plasma on time, altering the reactor capacitance and electric field. Further, it is observed that when the discharge gap is fixed, increasing barrier thickness reduces the total micro discharge charge but increases the median energy. On the contrary, fixing the electrode gap and increasing the barrier thickness decreases both total charge and median energy, producing fewer, less energetic filaments. Further, the model showed that the resulting total charge per pulse is the dominant predictor of NH(A–X) and H Balmer intensities intern related to ammonia synthesis. Thus, these insights translate into mechanistic design rules: reducing the discharge gap or increasing barrier thickness increases reactor capacitance and micro discharge energy, which boosts NH(A–X)/H emission and NH₃ concentration but at the expense of energy efficiency, whereas higher total flow may produce less power-dense micro discharges, improving energy efficiency. Overall, this data-driven framework shows how machine learning can deconvolute complex multidimensional plasma behaviour into physically interpretable guidelines for optimising non-thermal plasma reactors for sustainable, decentralised ammonia synthesis.
This work builds on Ref. [1].
Reference:
1. Dange, R., Sinausia, D., Bekkerman, A., Leybo, D. & Vogt, C. Systematic evaluation of plasma and reactor parameters in non-thermal dielectric barrier discharge plasma ammonia synthesis. Green Chem. 28, 1234–1256 (2026).