Fluoride adsorption from aqueous solution using a protonated clinoptilolite and its modeling with artificial neural network-based equations

B. G. Saucedo-Delgado, D. A. De Haro-Del Rio, L. M. González-Rodríguez, H. E. Reynel-Ávila, D. I. Mendoza-Castillo, A. Bonilla-Petriciolet, J. Rivera de la Rosa

Research output: Contribution to journalArticlepeer-review

32 Scopus citations

Abstract

Water defluoridation properties of a protonated clinoptilolite has been studied and analyzed. This adsorbent has been obtained by a thermochemical treatment with NH4Cl to protonate the zeolite surface and to increase its specific surface area. Results of adsorption kinetics and isotherms showed that the defluoridation properties of this protonated clinoptilolite were better than those reported for raw and modified zeolites with multivalent cations such as aluminum or iron. Defluoridation performance of this protonated clinoptilolite was endothermic and increased at acidic conditions in contrast to other zeolites modified with multivalent cations that should operate at pH ≥ 7 to maintain the adsorbent chemical stability. In addition, new models have been also developed to fit the fluoride adsorption on this protonated zeolite. These models were based on a hybridization of artificial neural networks and Langmuir and Pseudo-second order equations. Results showed that these hybrid models satisfactorily fitted the kinetics and isotherms of the fluoride adsorption on protonated clinoptilolite. These new models are promising to correlate and predict the fluoride adsorption with this zeolite or other types of adsorbents.

Original languageEnglish
Pages (from-to)98-106
Number of pages9
JournalJournal of Fluorine Chemistry
Volume204
DOIs
StatePublished - Dec 2017

Keywords

  • Adsorption modeling
  • Artificial neural network
  • Clinoptilolite
  • Fluoride adsorption
  • Zeolite protonation procedure

Fingerprint

Dive into the research topics of 'Fluoride adsorption from aqueous solution using a protonated clinoptilolite and its modeling with artificial neural network-based equations'. Together they form a unique fingerprint.

Cite this