•¨«„ŽZ–@‚̏«—ˆ“W–]

“Œ‹ž—‰È‘åŠw‹³Žö

‘å]C‘¢

”M—ÍŠw•¨«‚̍‘Û‰ï‹ciPPEPPD2001j‚ōu‰‰ 2001”N5ŒŽ22“ú ‘q•~

PPEPPD 2001 Conference, Kurashiki, Japan, May 22, 2001

(9th International Conference on Properties and Phase Equilibria for Product and Process Design)

Workshop

Future perspectives in computer-aided prediction of physical properties

of chemicals and computer-aided molecular and materials design in chemical engineering

Future of Prediction Method for Physical Properties

@

Shuzo Ohe

Graduate School of Chemical Engineering

Science University of Tokyo

@

Future prediction methods of physical properties will be changed to the methods based on computational chemistry instead of current empirically derived methods from principles of physical chemistry. However, it in the states of the arts is not enough for prediction from the point of practical engineering accuracy, while the computational chemistry well represents behavior of physical properties by its theoretical summation of individual molecule's energy. The accuracy of prediction by molecular dynamics depends on the parameters used for potential function. For the substances of which parameters are well investigated, the accuracy is satisfactory, whereas in the case of the parameters are not well studied, it is not satisfactory. The computational chemistry is quite effective to express of physical properties qualitatively, but is not still insufficient to predict them quantitatively.

Most of current prediction methods of physical properties except a few cases for ideal gas and ideal solution are empirically derived based on physical chemistry. Most of the future prediction of them, however, will be theoretically determined by molecular dynamics or molecular Monte Carlo simulation. With the present prediction method, there are some gaps between the practical prediction method and the method based on computational chemistry in accuracy. As adjusting technology of these gaps between the current practical prediction method and the method based on theoretically attractive computational chemistry, artificial intelligence, especially computational neural network methodologies, which have capability of association of non-linearity, will be effective technology for predicting of non-linear and complex properties, especially for the properties of non-ideal solution.

All rights reserved Shuzo Ohe 2014