Quantitative structure-property relationship models for gas adsorption and toxicity prediction of ionic liquids

Mgr. Xuejing Kang

Doctoral thesis

Abstract:

Ionic liquids (ILs) have gained extensive attention in both industrial and academic areas in recent three decades. Due to their remarkable properties, ILs have been applied in diverse fields, such as chemical engineering, pharmaceuticals, biotechnology, etc. ILs have been regarded as novel potential solvents and alternative media for gas separation and adsorption in environmental and ecological protection fields due to their extraordinary non-volatility. Moreover, the toxicity assessment of ILs is required before they are widely applied and released into the environment. However, because of the possible combinations of cations and anions of ILs, it is impractical to rely entirely on experimental methods to acquire the property data for all ILs. Exploring efficient computer-aided methods for property prediction of ILs and screening suitable ILs for a specific target is desired and significant. Quantitative structure-property relationship (QSPR) models are commonly considered as a promising way to diminish and substitute experimental processes involving measurements for the properties of compounds. This dissertation developed serval novel methods based on the QSPR methodology to achieve better model performance to predict gas absorption and toxicity of ILs. The main research contents and results of this thesis include: (1) a dataset with 502 data points of ammonia solubility in 17 ILs and another dataset consisting of 122 henry’s law constant (HLC) data of hydrogen sulfide (H2S) for 22 ILs were gathered; the relationships between the electrostatic potential surface areas (SEP) and the gas absorption capability of ILs were studied; the extreme learning machine (ELM) method was applied to develop predictive models based on the SEP descriptors of ILs. Results show that the proposed models were efficient and reliable, and can be potentially used for screening reasonable ILs to absorb NH3/H2S from chemical industry processes. (2) A QSPR model was developed to assess the ecotoxicity of ILs towards Vibrio fischeri using the ELM method; A comprehensive toxicity dataset for Vibrio fischeri involving 142 ILs were collected, while the parameters of cations and anions in ILs within various SEP ranges were calculated. The developed model with the R2 (0.9297) and the root-mean-square error (RMSE=0.3262) exhibited good reliability and predictive power for predicting ILs ecotoxicity on Vibrio fischeri. (3) Two new types of group/fragment contribution methods, atom surface fragment contribution (ASFC) method and non-integer group contribution (NGC) method, were first proposed to build predictive models for the ecotoxicity of ILs towards leukemia rat cell line (IPC-81); A dataset of 140 ILs toxicity towards IPC-81 was built; the group parameters used for modeling were based on the sigma surface area (Sσ-surface) or the surface areas of electrostatic potential (SEP) of atoms in ILs. The developed ASFC/NGC models showed better performance than the traditional group contribution method. Thus, the proposed techniques have novelty and superiority in assessing the properties of ionic liquids and other compounds. What’s more, they can be applied in the screening and designing of green and functional ILs for special applications in various fields.

Keywords

Ionic liquids, QSPR models, gas absorption, toxicity, exteme learning machine algorithm, group contribution method, quantum chemical parameters

Language used: English

Date on which the thesis was submitted / produced: 14. 10. 2022

Thesis defence

• Supervisor: doc. Zhongbing Chen

Czech University of Life Sciences Prague

Faculty of Environmental Sciences

Doctoral programme:
Applied and Landscape Ecology