(429a) The Properties of Gases and Liquids: 2020

For decades, The Properties of Gases and Liquids (PGL) has been a standard reference for professionals in the field, advanced undergraduate, and graduate students. With 27,000 google scholar citations and counting, this book serves a similar role to "Consumer Reports" in objectively evaluating the range of estimation methods available in the literature. Nevertheless, this important resource has not been updated since 2001. The current authors have undertaken to provide this update, with careful attention to the many changes in technology and methods that have occurred over the years. Our mission is to survey the entire field of physical property estimation in order to promote the best approaches and objectively characterize the quality of the available data and methods, such that current applications are reliable and future progress of the field is advanced.

A lot has happened since the fifth edition of PGL was published. Broadly speaking, advanced computer methods have become much more feasible for practical engineering calculations. These include quantum mechanical (QM) methods like Moller-Plesset and quantum density functional theory (DFT), quantitative structure property relations (QSPR; both closed-form and machine-learning methods), and molecular simulations like Monte Carlo (MC) and molecular dynamics (MD) methods. Additionally, group contribution (GC) methods have become more sophisticated, with second and third order groups in many cases. Despite the increased sophistication, these approaches are not guaranteed to be more accurate than traditional methods, so rigorous evaluation of all available methods, both old and new, is essential for practical application. Doing so requires the specification of a common set of metrics against which the performance of all methods is compared. This set of metrics cannot simply be obtained from the original literature as the range of compounds, conditions, and comparative statistics varies from study to study. As in previous versions of PGL, care is taken in this edition to define an objective and comprehensive set of metrics for proper evaluation of existing methods so that proper recommendations may be made about predictive methods.

Modern chemical databases and computer automation make evaluation possible in a manner not available previously, but this increased capacity for objective comparisons is consistent with PGL editions of the past. For many years, the spirit of PGL has reflected the working hypothesis that predictive methods are most reliable when they combine the fundamental principles of physics as a foundation with an appropriate degree of empiricism to ensure accurate reflection of experimental measurements. In this sense, experimental measurement comprises the final arbiter of any engineering model. Therefore, the first order of business in evaluating physical property methods is the compilation of standard databases for comparing the methods on a consistent basis. Despite best intentions, it would not be feasible for these authors to code and test every method that has been published since the fifth edition. The advancing sophistication of predictive models has made it more difficult to reproduce and evaluate many methods. Therefore, we envision a two stage process for making comparisons. In the first stage, method developers will submit working executables of their method in accordance with working protocols that we define. These submissions will be evaluated against the databases and the top few methods will be selected for the sample calculation stage. In the second stage, sample calculations will be performed by the authors. Final recommendations will include the sample calculations in textbook form along with a possibly subjective consideration of how easily reproducible the calculations would be for our readers to perform.

In this presentation, we outline our expectations for database development and method comparisons in greater detail.