2018
DOI: 10.1016/j.molliq.2018.03.090
|View full text |Cite|
|
Sign up to set email alerts
|

Predicting ionic liquid melting points using machine learning

Abstract: The melting point (T m ) of an ionic liquid (IL) is of crucial importance in many applications. The T m can vary considerably depending on the choice of the anion and cation. This study explores the use of various machine learning (ML) methods to predict the melting points (−96 • C -359 • C range) of structurally diverse 2212 ILs based on a combination of 1369 cations and 141 anions. Among the ML models applied to independent training and test sets, tree-based ensemble methods (Cubist, random forest and gradie… Show more

Help me understand this report

Search citation statements

Order By: Relevance

Paper Sections

Select...
2
2
1

Citation Types

2
47
0

Year Published

2019
2019
2023
2023

Publication Types

Select...
4
2
1

Relationship

1
6

Authors

Journals

citations
Cited by 80 publications
(49 citation statements)
references
References 79 publications
2
47
0
Order By: Relevance
“…In QSPR methods, several simple descriptors are correlated to the melting point. Numerous QSPR [46][47][48][49][50][51][52] and other statistical models [53,54] were assembled, with perhaps the most impressive being a work by Torrecilla et al, who report a mean prediction error of only 1.30% [46]. However, the QSPR and QSAR-based models have an intrinsic shortcoming-they are reliable as far as the characterized substances are similar to those of the training set that were used to build the model.…”
Section: Discussionmentioning
confidence: 99%
“…In QSPR methods, several simple descriptors are correlated to the melting point. Numerous QSPR [46][47][48][49][50][51][52] and other statistical models [53,54] were assembled, with perhaps the most impressive being a work by Torrecilla et al, who report a mean prediction error of only 1.30% [46]. However, the QSPR and QSAR-based models have an intrinsic shortcoming-they are reliable as far as the characterized substances are similar to those of the training set that were used to build the model.…”
Section: Discussionmentioning
confidence: 99%
“…Although fragment/group contribution descriptors have been popular, models trained on such variables often fail when presented with new fragments for which they were not trained. We have therefore chosen molecular descriptors that focus on charge distributions and geometrical indices which have been shown to yield good predictive performances for IL properties such as melting points [17], thermal decomposition temperatures [15], refractive indices [18] and CO 2 solubilities [52]. The variables were calculated independently for each cation and anion using the software KrakenX [53,54].…”
Section: Machine Learningmentioning
confidence: 99%
“…The top ranked variables (ranked according to the contribution of the variable to the response) in the models included the charged partial surface area descriptors (summarize the charge distribution in the ion), chemical reactivity parameters such as the HOMO/LUMO energies that are closely related to electrophilic/nucleophilic attack and the charge distribution in the ion, and softness (inverse of the HOMO-LUMO gap) which are indicative of the cation-anion electrostatic (nucleophilic-electrophilic) interactions. Experimental data for the properties was taken from various literature sources [13,17,18,52,55]. Machine learning models were evaluated for 12 different properties: melting points (T m ), glass transition temperatures (T g ), thermal decomposition temperatures (T d ), viscosities (η), densities (ρ), heat capacities (C p ), CO 2 capacity (x CO 2 ), electrical (κ) and thermal conductivities (λ), cytotoxicities towards the leukemia rat cell line IPC-81 (log 10 (EC 50 ), surface tension (σ) and refractive indices (n D ).…”
Section: Machine Learningmentioning
confidence: 99%
See 2 more Smart Citations