The refractive index is an important parameter to study the physical properties of materials. This paper takes supercritical carbon dioxide(S-CO_2) as the object to integrate the measured data of S-CO_2 refractive index of various literatures at different temperatures and pressures. The BP neural network optimized by genetic algorithm(GA) is used to establish a model for predicting S-CO_2 refractive index

and the density of S-CO_2 is inverted based on the Lorentz-Lorenz relationship between S-CO_2 density and wavelength and refractive index. The results show that

the maximum error of the model is 0.844%. The average error of inversion of S-CO_2 density value is no more than 3.65% compared with that of REFPROP. It is feasible to study the physical property of CO_2 by experimental measurement of refractive index in subcritical and supercritical regions. In the near critical region

due to the drastic change of the physical property of CO_2

the measurement of the law of refractive index change and the relationship between the refractive index and the physical property of CO_2 still need to be further studied.