Mechanistic insights into the removal of As(III) and As(V) by iron modified carbon based materials with the aid of machine learning. (April 2023)

• Record Type: Journal Article
• Authors: Yan, Changchun; Wang, Xuejiang; Xia, Siqing; Zhao, Jianfu
• Is Part Of: Chemosphere. Volume 321(2023)
• Publication Date: 2023-04
• Subjects: Iron-carbon based materials -- Arsenic -- Surface oxygen-containing functional groups -- Machine learning -- Environmental remediation; Pollution -- Periodicals; Pollution -- Physiological effect -- Periodicals; Environmental sciences -- Periodicals; Atmospheric chemistry -- Periodicals; 551.511
• DOI: 10.1016/j.chemosphere.2023.138125
• Languages: English
• ISSNs: 0045-6535

Abstract

Abstract: The machine learning (ML) technique was used to examine the effects of different microscopic material features on the ability of iron modified carbon-based materials (Fe-CBMs) to remove As(V) and As(III). The findings showed that specific CBMs and Fe-CBMs features (such as surface functionality) from sophisticated microscopic and spectroscopic techniques led to models that were more accurate than those constructed using more basic information, such as bulk elemental composition and surface area (the root-mean-square error fell by 44.7% for As(V) and 56.9% for As(III), respectively). The high non-polar carbon (NPC) content of CBMs and Fe-CBMs had a detrimental influence on As(V) and As(III) removal capability, whereas surface oxygen-containing functional groups (SOFGs) contents on CBMs and Fe-CBMs played an essential role in arsenic removal based on ML approaches. The relative importance of CO was greater by 77.8% and 40.6% than that of C–O on the elimination of As(V) and As(III), respectively. The accurate ML models are helpful for the future design of Fe-CBMs and the relative importance and partial dependence plot analysis can direct the use of Fe-CBMs for arsenic removal in a sensible manner under different application situations. Graphical abstract: Image 1 Highlights: Arsenic removal capacity on Fe-CBMs was modeled by machine learning. Surface functional groups provided high prediction accuracy of RF models. Higher content of NPC showed negative effects on the removal of As(V) and As(III). CO played significant role on the removal of As(V) and As(III).