| Abstract: | The three‐dimensional solubility parameter model was applied to analyze solution thermodynamic data of 27 solutes in poly(ε‐caprolactone) (PCL) between 70 and 110 °C. A linear regression method was compared with a nonlinear least square regression method, which searches solubility parameter components by minimization of the sum of error squares. The parameters of polymers were the same by both methods. When compared with the error in predicting χRT/V, the data showed a different slope from the simple three‐dimensional model. These deviations were reduced by a different model using a smaller weight on the polar and hydrogen bonding components. In the new model, the solubility parameter components were closer to the value of a structure analogue of PCL. The confidence intervals for the parameters were estimated from a linearized equation based on the sum of error squares. The solubility parameter components obtained were different from the average values of the five solutes with the smallest χ. The inclusion of solutes with high hydrogen bonding components contributed to the increase of the component in the nonlinear regression method. © 2006 Wiley Periodicals, Inc. J Appl Polym Sci 100: 2002–2009, 2006 |
