A Study on the differential diffusion caused by the difference in the molecular diffusivities in turbulent H-Air diffusion flame in coaxial air jet is investigated. The approach is to use the perturbation method based on the conventional conserved scalar approach. Turbulent momentum fluxes are represented by turbulence model. The turbulence models used are:Algebraic stress model, standard kε model and Rodi's corrected k-ε model. The scalar fluctuations are represented the clipped Gaussian P.D.F with mixture fraction. The species equations are closed by the combustion model. The combustion models used are:chemical equilibrium model, eddy break-up model and Arrhenius model. The computational results considering the differential diffusion for Re=4200 are better predicted than those which don't consider it. Because of the differential diffusion of hydrogen, the hydrogen concentration decays more rapidly than that of equal diffusivity theory, so temperatures rise more rapidly and formation rates of H 0 therefore increase a long axis between 0-150 mm. Turbulence models are tested at Re=4200 and Re=11000. As a result, at Re=4200, the algebraic stress model is superior near the initial region, but at downstream, the Rodi's k-ε model predicts the experiments well, but there is little difference between Rodi's model and standard k-ε model. As the Reynolds number increases to Re=11000, the standard model is better than the other models. This is thought to be in local equilibrium of turbulence at high Reynolds number. Arrhenius model predicts greatly the temperature, and $H_2O$ concentrations than Magnussen's eddy break model. The effects of nozzle wall thickness and axial diffusion become apparent, and they are dicussed, The two methods of averaging are discussed, and additionally, discussions are made about the validity of p.d.f. model.
