Chemical space mapping for multicomponent gas mixtures

Airat Kotliar-Shapirov, Fedor S. Fedorov, Henni Ouerdane, Stanislav Evlashin, Albert G. Nasibulin, Keith J. Stevenson

Research output: Contribution to journalArticlepeer-review

Abstract

Analysis of gas mixtures using multivariable sensors or multisensor gas analytical systems requires adapted protocols of data processing. Notably, interference of analytes yields fingerprints that differ from a combination of the projected patterns of the single-components in the chemical space. Here, employing cyclic voltammetry only, we analyze the classification of single components, NO2 and CO2, and their mixtures based on the response of electrochemical sensors we specifically designed. The sensor sensitivity towards NO2 is 15.7 nA/ppm, and the sensitivity towards CO2 is 1.56 nA/ppm with the limits of detection of 1 ppm and 11 ppm, respectively. The overlap of the analytes’ voltammetry profiles makes it difficult to extract their exact concentrations, so data-driven approaches were used, specifically principal component analysis (PCA). After data classification, inverse PCA was used to determine the characteristics of the gas mixture response kinetics. Partial dependence analysis helps to identify which measured potential contributes the most to the selective recognition of analytes, both single-components and mixtures. As an outcome, accurate gas mixture concentration profiles were obtained which allows deconvolution of the gas mixture in the chemical space.

Original languageEnglish
Article number115472
JournalJournal of Electroanalytical Chemistry
Volume895
DOIs
Publication statusPublished - 15 Aug 2021

Keywords

  • Carbon dioxide
  • Cyclic voltammetry
  • Feature importance
  • Gas analytical system
  • Machine learning
  • Nitrogen dioxide

Fingerprint

Dive into the research topics of 'Chemical space mapping for multicomponent gas mixtures'. Together they form a unique fingerprint.

Cite this