## Abstract

Thermal diffusion measurements made by the two-bulb method are reported for the system CO<latex>$_{2}$</latex> + N <latex>$_{2}$</latex> these cover the composition range 10 to 90% CO<latex>$_{2}$</latex> and the approximate temperature range 300 to 800 K. By using the results in conjunction with an extension upwards in temperature of a previously described technique, values of the binary (concentration) diffusion coefficient D<latex>$_{12}$</latex> are determined for the temperature range 300 to 1800 K. The present work is not only novel in using the technique to probe concentration diffusion at elevated temperatures, but also in testing the technique's viability for use with relatively complex gas mixtures. The upper temperature limit of this extended technique is determined by the onset of invalidity of assumptions implicit in classical kinetic theory, but is thought not to be reached in this work. Comparison of the values obtained for D<latex>$_{12}$</latex> with the direct measurements of Walker & Westenberg, which cover the range 300 to 1150 K, shows a mean absolute deviation of 1.3% and a mean algebraic deviation of +0.5%; a similar comparison for the range 1100 to 1800 K with the data of Pakurar & Ferron, individual values of which scatter considerably and have uncertainties of about 10%, shows a mean absolute deviation of 3.7% and a mean algebraic deviation of -2.7%. The overall concordance of the present calculations with the experimental data seems good enough to establish both the utility of the calculational technique and the freedom from systematic error of the point-source method of measuring high-temperature diffusion coefficients. The experimental thermal diffusion data are also compared with the predictions of the commoner intermolecular potential models; of these, given a reasonable value of the well-depth parameter, the exp: 6(<latex>$\alpha $</latex> = 13) and Lennard-Jones (9:6) perform well, especially at the higher temperatures, and the addition of a quadrupole interaction term to the L.-J. (12:6) model effects a marked improvement in the latter's performance.