Selection of Protic Ionic Liquids for the Pretreatment of Lignocellulosic Biomasses Using the COSMO-SAC Model and Machine Learning Techniques

- 338038
Resumo
Favoritar este trabalho
Como citar esse trabalho?
Resumo

Pretreatment of lignocellulosic biomass is a crucial step for 2G ethanol production. Protic ionic liquids (PILs) are presented in literature as sustainable alternatives in this case. However, experimental evaluation of PIL precursors is time-cost consuming. In this work, the solubility of guaiacol-syringol β-O-4 dimer from lignin structure (solute) in PILs was calculated with COSMO-SAC model through activity coefficient at infinite dilution (γi∞) values. PILs were obtained by combining precursors used in literature, with 31 anions and 10 cations, totaling 310 PILs. Systems with different water fractions (0 to 50%) and temperatures were considered. Machine Learning (ML) models (Random Forest, Gradient Boost, SVR and linear regression) and Explainable AI (XAI) tools were used to describe/explain responses based on molecular descriptors: dissociation constant (pKa), dielectric energy, octanol-water partition coefficient (Kow), O/C ratio, OH groups, and surface charge distribution (hydrogen bond donor and acceptor fractions) of the precursors. LIPs formulated with ethylenediamine and carboxylic acids (formic, trans-aconitic, and malonic), although not commonly used in the literature, showed the lowest γi∞ values, indicating greater potential for lignin solubilization. Results show linear regression (R² = 0.7342) with lower predictive capacity, while non-linear models (Gradient Boost, R² of 0.9982) could correlate descriptors with γi∞ values. According to feature importance analysis and SHAP values, solubility shows higher correlation with descriptors related to surface charge distribution and Kow. Water fractions have intermediate relevance, tending to decrease solubility. Temperature exerts a smaller effect, depending on LIP. COSMO-SAC model is a potential tool for screening LIPs for biomass pretreatment for E2G. Associated with ML, it was possible to correlate main significant factors, facilitating phenomenon comprehension and choice of best solvents in this context. 

Compartilhe suas ideias ou dúvidas com os autores!

Sabia que o maior estímulo no desenvolvimento científico e cultural é a curiosidade? Deixe seus questionamentos ou sugestões para o autor!

Faça login para interagir

Tem uma dúvida ou sugestão? Compartilhe seu feedback com os autores!

Instituições
  • 1 Universidade Estadual de Campinas (UNICAMP)
  • 2 FEA/UNICAMP
Eixo Temático
  • Simulação molecular
Palavras-chave
Ethanol 2G
Ionic Liquids
Biomass
COSMO-SAC