Machine Learning for Chemistry Reduction in N$_2$-H$_2$ Low-Temperature Plasmas

TL;DR

This paper introduces a machine learning model that simplifies complex chemical reaction schemes in low-temperature N$_2$-H$_2$ plasmas, aiding in understanding key pathways for ammonia production.

Contribution

The study develops a novel machine learning approach to identify and reduce critical reactions in plasma chemistry, enhancing interpretability of complex reaction networks.

Findings

The model successfully identifies key reactions in N$_2$-H$_2$ plasmas.

Reduction of chemical schemes improves understanding of ammonia formation pathways.

The approach can be applied to other plasma chemistries.

Abstract

Low-temperature plasmas are partially ionized gases, where ions and neutrals coexist in a highly reactive environment. This creates a rich chemistry, which is often difficult to understand in its full complexity. In this work, we develop a machine learning model to identify the most important reactions in a given chemical scheme. The training data are an initial distribution of species and a final distribution of species, which can be obtained from either experiments or simulations. The model is trained to provide a set of reaction weights, which become the basis for reducing the chemical scheme. The approach is applied to N$_2$-H$_2$ plasmas, created by an electric discharge at low pressure, where the main goal is to produce NH$_3$. The interplay of multiple species, as well as of volume and surface reactions, make this chemistry especially challenging to understand. Reducing the chemical scheme via the proposed model helps identify the main chemical pathways.