| Supercritical fluid extraction and supercritical fluid chromatography have been widely used in the separation and purification of various natural products with high added value such as coumarins and flavonoids. As an important parameter for process design, optimization and amplification, diffusion coefficient (D12) is significant to the development and application of supercritical fluid technology. Chromatographic impulse response (CIR) method can determine the infinite dilute diffusion coefficients of various solutes in scCO2 quickly and accurately, especially for solutes with high polarity or viscosity, large molecular weight and solid matter.We firstly developed an improved CIR method. One of the major error sources of CIR is the measurement of velocity in diffusion column. Especially for ternary system containing modifier, the phase separation caused by decompression makes it difficult to obtain accurate velocity value. Therefore, we proposed the improved CIR method. The hold-up time of system measured by n-hexane was used to calculate the solute retention factor. And then combined with parameters obtained from curve fitting, we could get D12 without the direct measurement of velocity. The accuracy and reliability of the improved CIR method for binary and ternary diffusion coefficients determination were verified by comparing the experimental and literature data.The diffusion coefficients D12 and retention factors k of coumarins in scCO2 were determined by the improved CIR method. Moreover, we applied CIR method to scCO2 system containing modifier for the first time, and measured the D12 and k values of flavanones in scCO2 containing 15 mol% ethanol accurately. It was found that both the binary and ternary D12 values decrease with increasing pressure, decrease linearly with the increase of density, and can be well represented by the equation based on hydrodynamic approach. Meanwhile, the D12 data can be well correlated with temperature using an Arrhenius type equation, which also provide the diffusion activation energy. The accuracy of several prediction models for diffusion coefficient was tested. For binary system, the rough hard sphere and free volume models are superior to the hydrodynamic equation with the average AARD lower than 6.00%. However, all these models show considerable deviation for the prediction of D12 of flavanones in ethanol modified scCO2. The Schmidt number correlation and DHB free volume model with adjustable parameters make a good prediction for both binary and ternary D12 with all AARD less than 4.00%.The influence of solute molecular configuration and intermolecular interaction on diffusion coefficient was investigated by analyzing various organic compounds’ D12 data in this work and literature. It was found that the binary diffusion coefficient simply decreases with the increase of solute molecular weight, and the relationship can be quantitatively represented by D12=aMM-0.5. It’s mainly due to the slower molecular motion rate and stronger interaction with SCCO2 of large molecule. While in addition to the molecular weight and volume, D12 of flavanones in ethanol modified SCCO2 are remarkably influenced by intermolecular hydrogen-bonding between solutes and ethanol. The stronger the interaction of solute with ethanol, the larger should be the solute cluster as true diffusion unit and smaller the corresponding diffusion coefficient.We employed quantum chemistry method to delve into the type and strength of interaction between solutes and ethanol and made a further study on qualitative and quantitative relationship between D12 and interaction energy AE. The result indicated that intermolecular hydrogen bonding with different intensity is formed, proved that diffusion coefficient is negatively related to interaction energy and the relationship can be well correlated by simple empirical model. Based on the quantitative calculation, we put forward modified prediction models for ternary D12 for the first time. The predictive accuracy for D12 of flavanones in ethanol modified SCCO2 is significantly improved. All the average AARD are lower than 6.00%, while the AARD of unrevised model can be as high as 100%. |
PDF Full Text Request