有机化学品不同温度下(过冷)液体蒸气压预测模型的建立与评价

(过冷)液体蒸气压(PL)是评价化学品在环境中分配、迁移和归趋行为的重要参数。PL具有较强的温度依附性。发展一种能够精确预测不同环境温度下化学品PL的方法,有助于填补化学品生态风险评估的大量数据缺失。本研究收集整理了661种有机化合物在不同温度下(200~830 K)共计10 478个log PL值。在此基础上,采用偏最小二乘(PLS)回归和支持向量机(SVM)方法,构建了PL的线性和非线性预测模型。结果表明:2种模型均具有良好的拟合度、稳健性及预测能力,SVM模型的预测性能略高于PLS模型(PLS:R2adj.tra=0.912,RMSEtra=0.477,Q2ext=0.910;SVM:R2adj.tra=0.997,RMSEtra=0.092,Q2ext=0.967)。机理分析表明,温度是影响PL的主要因素,温度越高,蒸气压越大;其次,X1sol也影响PL大小,X1sol用来描述分子间的色散作用,分子间色散力越小,蒸气压越大;此外,化合物的氢键个数、极性和分子构型等因素也影响PL大小。采用Wiliams plot方法表征了PLS模型应用域。所建立的模型可用来预测烷烃、烯烃、醇、酮、羧酸、苯、酚、联苯、卤代芳香烃、含N化合物及含S化合物在不同温度下的PL数据。

Development and Evaluation for a Predictive Model of (Subcooled) Vapor Pressure of Organic Chemicals at Different Temperatures

The (subcooled) liquid vapor pressure (PL) is an important parameter to assess the distribution, transport and fate of organic chemicals in the environment. PL possesses strong temperature-dependence. In order to fill the gap of the data for the ecological risk assessment, it is of significance to develop an effective tool for predicting PL at different temperatures. In this study, overall 10 478 logPL values of 661 compounds at different temperatures are collected and used to develop two models, a partial least square (PLS) regression linear model and a support vector machine (SVM) nonlinear model. Results reveal that both models have satisfactory goodness-of-fit, robustness and external predictive performance. The predictive performance of SVM model is slightly better than that of PLS model (PLS: R²_adj.tra = 0.912, RMSE_tra = 0.477, Q²_ext = 0.910; SVM: R²_adj.tra = 0.997, RMSE_tra = 0.092, Q²_ext = 0.967). Mechanistic explanation shows that the PL of organic chemicals is mainly influenced by the temperature and X1sol (intermolecular dispersive forces). Increasing temperature and decreasing intermolecular dispersive forces lead to the increase of logPL values. Moreover, the number of hydrogen bond, dipole and the molecular structure can also exert influence on PL values. The applicability domain of the PLS model was characterized by the Williams plot. The developed models can be used to predict logPL values of alkanes, alkenes, ketones, carboxylic acids, benzene, phenol, biphenyl, halogenatearomatic hydrocarbonsas well as N- and S-containing chemicals.