A noise elimination technique was applied to the determination of binary diffusion coefficients D-12 from the response curves having extremely low absorbance intensities in impulse response methods under supercritical conditions of carbon dioxide. The effectiveness of this technique was experimentally examined for the analyses of response curves through both the curve-fitting and the moment methods in two cases: the chromatographic impulse response method for phenol and beta-carotene with a polymer-coated capillary column, and the Taylor dispersion method for acetone with an uncoated capillary column. Unreliable D-12 values were obtained from the moment method of the response curves at lower absorbance intensities, even treated with noise elimination. The curve-fitting method with the noise elimination treatment was quite effective for determining the D-12 values accurately, and was valid at the lowest absorbance intensities, on the order of 10(-4) absorbance unit of UV-Vis multi-detector, corresponding to the smallest quantity of the solute, i.e. 6 x 10(-5), 6 x 10(-6), and 5 x 10(-2) mu mol for phenol, beta-carotene, and acetone, respectively, under conditions studied. Infinite dilution regions for binary diffusion coefficients were obtained by injecting various amounts: the binary diffusion coefficients showed constant values at concentrations less than 0.6, 0.004, and 0.08 mol m(-3) for phenol, beta-carotene, and acetone, respectively, in supercritical carbon dioxide at 313.2 K and 16-18 MPa. (C) 2004 Elsevier B.V. All rights reserved.