Planar isotropy of passive scalar turbulent mixing with a mean perpendicular gradient L. Danaila, 1 J. Dusek, 2 P. Le Gal, 1 F. Anselmet, 1 C. Brun, 3 and A. Pumir 3 1 IRPHE, 12 Avenue du Ge ´ne ´ral Leclerc, 13003 Marseille, France 2 IMF, 2 rue Boussingault, 67000 Strasbourg, France 3 INLN, 1361 route des Lucioles, 06560 Valbonne, France Received 16 October 1998; revised manuscript received 23 March 1999 A recently proposed evolution equation Vaienti et al., Physica D 85, 405 1994 for the probability density functions PDF’sof turbulent passive scalar increments obtained under the assumptions of fully three- dimensional homogeneity and isotropy is submitted to validation using direct numerical simulation DNS results of the mixing of a passive scalar with a nonzero mean gradient by a homogeneous and isotropic turbulent velocity field. It is shown that this approach leads to a quantitatively correct balance between the different terms of the equation, in a plane perpendicular to the mean gradient, at small scales and at large Pe ´clet number. A weaker assumption of homogeneity and isotropy restricted to the plane normal to the mean gradient is then considered to derive an equation describing the evolution of the PDF’s as a function of the spatial scale and the scalar increments. A very good agreement between the theory and the DNS data is obtained at all scales. As a particular case of the theory, we derive a generalized form for the well-known Yaglom equation the isotropic relation between the second-order moments for temperature increments and the third-order velocity-temperature mixed moments. This approach allows us to determine quantitatively how the integral scale properties influence the properties of mixing throughout the whole range of scales. In the simple con- figuration considered here, the PDF’s of the scalar increments perpendicular to the mean gradient can be theoretically described once the sources of inhomogeneity and anisotropy at large scales are correctly taken into account. S1063-651X9902108-X PACS numbers: 47.27.Gs, 47.27.Ak, 47.27.Eq I. INTRODUCTION Understanding the statistical and dynamical properties of high Reynolds number turbulent flows is currently the sub- ject of many investigations. Kolmogorov theory 1, hereaf- ter referred to as K41, has provided some major insight in this field. It assumes local isotropy in high Reynolds number turbulent flows. The K41 theory successfully predicts the structure of the two-point correlation function of the velocity field in the inertial range, intermediate between the large anisotropic scales, where energy injection takes place, and the small scales, where viscosity is important. Local isotropy for the inertial range signifies simply that all the statistical properties of the field are invariant under any rotation direc- tion. In particular, this property implies the independence of these scales with respect to the large scale forcing, in general anisotropic and very specific nonuniversal. The K41 theory leads to a scaling law for the velocity increments, S p =u ( r ) -u (0) p ¯ p /3 r p /3 , when the separation r is in the inertial range, and where ¯ is the mean dissipation rate. Ac- cordingly, these moments should generally determine self- similar probability density functions PDF’s. The K41 theory, originally derived for the velocity field was later naturally extended to the passive scalar field, by Oboukhov 2and Corrsin 3. Indeed, it was reasonable to argue that the passive scalar properties would be completely determined by the driving field behavior. Significant deviations of the high order velocity moments from the K41 predictions were found experimentally 4. Un- derstanding these deviations, attributed to the phenomenon called ‘‘intermittency,’’ and predicting the corresponding anomalous scaling laws, has become a major challenge. Sev- eral ‘‘phenomenological models,’’ from the Kolmogorov model K41to the refined similarity hypothesis RSH5, or the multifractal model 6, tried, with more or less suc- cess, to refine the predictions qualitatively and quantitatively for the velocity scaling law exponents see Ref. 7for a recent review. Recently, a general analysis of the turbulent cascade was proposed in Ref. 8. A canonical distribution of velocity differences at any scale was introduced, with the help of a conserved quantity throughout the whole scale range. Several of the previously mentioned models can be recovered as particular cases in this general formulation. A phenomenological description of the statistical properties of the cascade was also presented in Ref. 9, providing an evo- lution equation for the probability density function of the velocity increment. A method to extract the exponents cor- responding to the various irreducible representations of the rotation group was proposed in Ref. 10, therefore address- ing the issue of isotropy in turbulent flows. Alternatively, in 11, anomalous scaling exponents for the dynamic field were obtained, by taking into account the interaction be- tween random and large-scale coherent components of a tur- bulent field. Generally speaking, the passive scalar field was found to exhibit a more anomalous behavior than the velocity field 12, being in this sense ‘‘more intermittent than the velocity field.’’ Specifically, the probability density functions PDF’sassociated with the inertial and dissipative scales deviate from self-similar behavior, and present wider than exponential tails more pronounced than those for the velocity field. PHYSICAL REVIEW E AUGUST 1999 VOLUME 60, NUMBER 2 PRE 60 1063-651X/99/602/169117/$15.00 1691 © 1999 The American Physical Society