Predicting the Enthalpy and Gibbs Energy of Sublimation by QSPR Modeling

Area of Science:

• Computational chemistry
• Physical chemistry

Background:

• Quantitative structure-property relationship (QSPR) models are valuable tools for predicting chemical properties.
• Predicting the enthalpy of sublimation has been explored with various QSPR models.
• Fewer QSPR models exist for predicting the Gibbs energy of sublimation.

Purpose of the Study:

• To compare existing QSPR models for predicting the enthalpy of sublimation.
• To develop and evaluate novel QSPR models for predicting the Gibbs energy of sublimation.
• To assess the performance of different modeling techniques, including multiple linear regression and neural networks.

Main Methods:

• Reproduced and validated previously reported QSPR models for enthalpy of sublimation.
• Developed new QSPR models for Gibbs energy of sublimation, building upon existing enthalpy models.
• Employed multiple linear regression (MLR) and neural network (NN) training methods.

Main Results:

• Successfully reproduced existing QSPR models for enthalpy of sublimation with high correlation coefficients (0.82–0.97).
• Developed an MLR model for Gibbs energy of sublimation with R² training=0.71, R² test=0.62, and SD=9.1 kJ mol^-1.
• Improved predictive performance using a neural network for Gibbs energy of sublimation, achieving R² training=0.80, R² test=0.63, and SD=8.9 kJ mol^-1.

Conclusions:

• Established reliable QSPR models for predicting enthalpy of sublimation.
• Demonstrated the feasibility and improved accuracy of using QSPR, particularly neural networks, for predicting Gibbs energy of sublimation.
• The developed neural network model offers a robust approach for estimating Gibbs energy of sublimation.