PHYSICS OF FLUIDS 20, 025102 共2008兲

Effects of molecular diffusion on the subgrid-scale modeling

of passive scalars
C. Brun,1,2 G. Balarac,1,a兲 C. B. da Silva,3 and O. Métais1
1
Equipe MoST/LEGI, BP. 53, 38041, Grenoble Cedex 09, France
2
Laboratoire de Mécanique et d’Energétique, 8 rue Léonard de Vinci, 45072 Orléans Cedex 2, France
3
IDMEC/IST, Technical University of Lisbon, Mecânica I, 1o andar/esquerdo (LASEF), Av. Rovisco Pais,
1049-001 Lisbon, Portugal
共Received 3 November 2007; accepted 15 January 2008; published online 25 February 2008兲
The spectral eddy-viscosity and eddy-diffusivity closures derived from the Eddy-Damped
Qusasi-Normal Markovian theory, and one of its physical space counterparts, i.e., the structure
function model 关Métais and Lesieur, J. Fluid Mech. 239, 157 共1992兲兴, are revisited to account for
molecular viscosity and diffusivity effects. The subgrid-scale Schmidt number 共usually set to Sct
⬇ 0.6兲 is analytically derived from the Eddy-Damped Qusasi-Normal Markovian theory and shown
to be Reynolds number dependent, a property of utmost importance for flows involving scalar
transport at moderate Reynolds numbers or during the transition to turbulence. A priori tests in
direct numerical simulation of homogeneous isotropic turbulence 关da Silva and Pereira, Phys. Fluids
19, 1 共2007兲兴 and in spatially evolving turbulent plane jets 关da Silva and Métais, J. Fluid Mech. 473,
103 共2002兲兴, as well as a posteriori 共large eddy simulation兲 tests in a round jet are carried out and
show that the present viscous structure function model improves the results from the classical
approaches and at a comparatively small computational cost. © 2008 American Institute of Physics.
关DOI: 10.1063/1.2844469兴

I. INTRODUCTION II. SGS MODELING IN SPECTRAL SPACE

The large eddy simulation 共LES兲 technique consists of In the Fourier space, the SGS transfers across the cutoff
describing the large-scale motions based on a spatial filtering wave number kc 共which represents the smallest resolved
operation while the effect of the subgrid-scales 共SGS兲 needs scale for LES兲 are modeled based on eddy-viscosity and
to be modeled.1 While many SGS models have been de- eddy-diffusivity concept defined as8

冕 冋 册
signed to close the resulting filtered Navier–Stokes equations ⬁
1 ⳵E
for incompressible flows,2–4 the corresponding problem ap- ␯⬁t = ␪0pp 5E共p兲 + p dp, 共1兲
plied to turbulent combustion, turbulent mixing, or com- 15 kc ⳵p
pressible flow has not yet been fully addressed.5,6 For these
situations, the filtered transport equation of a scalar variable
共e.g., mixture fraction, temperature兲 must be solved and a
subgrid-scale scalar flux has to be modeled. The simplest
␬⬁t =
2
3
冕kc
⬁
␪0pp
T
E共p兲dp, 共2兲

way to address this issue consists in considering a constant

where E共k兲 is the three-dimensional kinetic energy spectrum
SGS Schmidt number7 Sct ⬇ 0.6 and in determining an eddy-
and ␪kpq and ␪kpq
T
are the velocity and scalar-velocity triple
diffusivity term proportional to the eddy-viscosity model:
correlation relaxation times, respectively,
␬t = ␯t / Sct. The main limitation of such a model is that Sct is
set to a constant while it should be strongly affected by the 1
slope of the kinetic energy spectrum.8 Moin et al.9 have ␪kpq = , 共3兲
␮共k兲 + ␮共p兲 + ␮共q兲 + ␯共k2 + p2 + q2兲
shown indeed that a dynamically computed value of Sct
could greatly improve the results of the simulations, since ␬t
depends on the molecular Schmidt number Sc and on the 1
␪kpq
T
= . 共4兲
local turbulence level of the flow. In the present work we ␮⬘共k兲 + ␮⬘共p兲 + ␮⬙共q兲 + ␬共k2 + p2兲 + ␯q2
revisit the eddy-viscosity and eddy-diffusivity models in
They are introduced in the frame of the EDQNM theory,
spectral space8 in order to integrate the molecular viscosity/
to close the evolution equation for the third-order moments
diffusivity effects from the original Eddy-Damped Qusasi-
of the velocity and temperature components1,10,11 involving
Normal Markovian 共EDQNM兲 equations. The new
triadic interactions,
formulation/model is useful to address two related issues: 共i兲
The effects of low Reynolds number and 共ii兲 a variable SGS
Schmidt number for transitional flows. 再 ⳵
⳵t
+ ␯共k2 + p2 + q2兲 + 关␮共k兲 + ␮共p兲 + ␮共q兲兴 冎
a兲
Present address: CTR Stanford University, CA 94301, USA. ⫻具û共k兲û共p兲û共q兲典 = ⌺具ûû典具ûû典, 共5兲

1070-6631/2008/20共2兲/025102/6/$23.00 20, 025102-1 © 2008 American Institute of Physics

Downloaded 25 Feb 2008 to 193.136.128.19. Redistribution subject to AIP license or copyright; see http://pof.aip.org/pof/copyright.jsp
025102-2 Brun et al. Phys. Fluids 20, 025102 共2008兲

再 ⳵
⳵t
+ ␬共k2 + p2兲 + ␯q2 + 关␮⬘共k兲 + ␮⬘共p兲 + ␮⬙共q兲兴 冎 X=
15a21 m + 1 ␯ti
5−m3−m ␯
,
⬁

共11兲

⫻具T̂共k兲T̂共p兲û共q兲典 = ⌺具ûû典具T̂T̂典, 共6兲

⬁
3共a2 + a3兲2 m + 1 ␬ti
Y= . 共12兲
where ␮共k兲2 = a21兰k0 p2E共p兲dp, ␮⬘共k兲2 = a22兰k0 p2E共p兲dp and 2 3−m␯+␬
␮⬙共k兲2 = a23兰k0 p2E共p兲dp are linear damping functions that ap-
proximate the fourth-order cumulants with a1 = 0.218C3/2 The present formulation leads to a damping effect with re-
k ,
spect to the basic inviscid eddy-viscosity and eddy-
a2 = 0, and a3 = 0.7848C3/2k so that 共a2 + a3兲 / 6a1 = 0.6. In Eqs.
共3兲 and 共4兲, the role of the molecular viscosity/diffusivity diffusivity defined as8
terms ␯k2 and ␬k2 is to partially inhibit the linear terms as-
sociated to ␮共k兲, ␮⬘共k兲, and ␮⬙共k兲. This contribution had
been neglected in the original formulation of the spectral
␯t⬁i =
5 − m 冑3 − m
15a1 m + 1
冑 E共kc兲
kc
, 共13兲

eddy-viscosity model8 because the aim was to deal with high

Reynolds number flows. It is presently reintroduced here to
address the issue of modeling turbulent flows involving a
transition to turbulence or a moderate Reynolds number. In
order to evaluate the integrals defined in Eqs. 共1兲 and 共2兲, the
shape of the energy spectrum E共k兲 for k ⬎ kc must be known.
In the present work we assume that the kinetic energy spec-
trum is modeled as E共k兲 = Cmk−m for k ⬎ kc, where 5 / 3 ⱕ m
⬍ 3. The case m = 5 / 3 is considered to be a good approxima-
tion for high Reynolds number flows. In particular, the accu-
racy of the approximation increases with the Reynolds num-
ber. For low Reynolds number flows, the spectrum exhibits
only a dissipative range with a local slope that is always m
⬎ 5 / 3.8,12 Despite the limitations of this approximation, par-
ticularly in low Reynolds number flows, it has yielded good
results in numerous theoretical and numerical works.1 It is
interesting to note that the upper bound m = 3 value in the
present analytical developments corresponds to forced flow
situations such as two-dimensional turbulence in geophysical
flows.1
Thus, the relaxation times are8,12
1 1
␪0pp = , 共7兲
2 a 1 冑C m 共3−m兲/2
+ ␯p 2
冑3 − m p
␬t⬁i =
4 冑3 − m
3共a2 + a3兲 m + 1
冑 E共kc兲
kc
. 共14兲
The new model equations 共9兲–共14兲 are first assessed using
kinetic energy spectra E共k兲 from DNS of homogeneous iso-
tropic turbulence14 for Re␭ = 39− 96, Sc= 0.2− 0.7− 3. Ex-
plicit filtering is obtained with a sharp cutoff filter in Fourier
space applied with filter sizes ranging from ⌬ ¯ = 2⌬ to ⌬¯
= 10⌬, where ⌬ is the size of the grid from the DNS in each
direction. In practice, the m slope is obtained by interpolation
using a least-squares method12 applied on the three-
dimensional energy spectra in the wave number range kc / 2
ⱕ k ⱕ kc. In this sense, m is a function of the global slope of
the kinetic energy spectrum in the neighborhood of the cutoff
wave number. For low Reynolds number flows no inertial
range exists and the cutoff filter is located in the dissipative
region of the kinetic energy spectrum. Therefore, m becomes
steeper as the filter size decreases. Figures 1共a兲 and 1共b兲
show indeed that the spectral eddy-viscosity ␯⬁t and diffusiv-
ity ␬⬁t are strongly damped for low Reynolds numbers, e.g.,
for Re␭ = 39, in comparison with their inviscid counterparts
␯t⬁ and ␬t⬁, respectively. The inviscid SGS Schmidt number
i i
is defined8 as
␯t⬁i a2 + a3
Sc共m兲
ti = = 共5 − m兲 . 共15兲
␬t⬁i 20a1

1 Introducing Q as the ratio between Y and X,

␪0pp
T
= . 共8兲
共a2 + a3兲冑Cm
p共3−m兲/2 + 共␬ + ␯兲p2 Y a2 + a3 ␯ Sc
冑3 − m Q= =2 = 12Sc共m=5/3兲 , 共16兲
X a1 ␯ + ␬ ti
1 + Sc

Using Eqs. 共1兲, 共2兲, 共7兲, and 共8兲, the eddy-viscosity and eddy- we obtain an expression for the SGS Schmidt number,
diffusivity are now obtained as13
␯⬁t QX − Q ln共1 + X兲
Sc共m兲
t = = Sc共m兲 . 共17兲

冋 册
␬⬁t ti
QX − ln共1 + QX兲
ln共1 + X兲
␯⬁t = ␯t⬁i 1 − , 共9兲
X In the limit of small X; i.e., small Reynolds number, the SGS
Schmidt number is

冋
␬⬁t = ␬t⬁i 1 −
ln共1 + Y兲
Y
册 共10兲 Sc共m兲
t =
共5 − m兲
40
1+ 冉 冊
1
Sc

instead of its inviscid value defined in Eq. 共15兲. Thus, the

with new model provides a damping adjustment of the SGS

Downloaded 25 Feb 2008 to 193.136.128.19. Redistribution subject to AIP license or copyright; see http://pof.aip.org/pof/copyright.jsp
025102-3 Effects of molecular diffusion on the subgrid-scale modeling Phys. Fluids 20, 025102 共2008兲

κti, Reλ=39, Sc=3.0

2 2 κti, Reλ=96, Sc=0.7
10 10 κt, Reλ=39, Sc=3.0
νti, Reλ=39, Sc=3.0
νt∞ νti, Reλ=96, Sc=0.7 κ∞
t κt, Reλ=96, Sc=0.7
νt, Reλ=39, Sc=3.0 κti, Reλ=96, Sc=0.2
νt, Reλ=96, Sc=0.7 κt, Reλ=96, Sc=0.2
1 1
10 10

100 100

-1 -1
10 10
4 6 8 10 4 6 8 10

∆/∆ ∆/∆
(a) (b)

Scti, Reλ=39, Sc=3.0

1.4 Scti, Reλ=96, Sc=0.7
(m) Sct, Reλ=39, Sc=3.0
Sct Sct, Reλ=96, Sc=0.7
1.2 Scti, Reλ=96, Sc=0.2
Sct, Reλ=96, Sc=0.2
1

0.8

0.6

0.4

0.2

0
4 6 8 10

∆/∆
(c)

FIG. 1. A priori tests on isotropic turbulence with passive scalar for Re␭ = 39– 96 and Sc= 0.2– 0.7– 3. A sharp cutoff filter with filter widths equal to ⌬ ¯ /⌬
苸 关2 ; 10兴 was used. The results were obtained with Eqs. 共9兲, 共10兲, and 共17兲. 共a兲 Spectral eddy-viscosity ␯⬁t and ␯t⬁. 共b兲 Spectral eddy-diffusivity ␬⬁t and ␬t⬁. 共c兲
i i
SGS Schmidt number Sc共m兲 t and Scti共m兲.

Schmidt number, which includes the effects of the global III. SGS MODELING IN PHYSICAL SPACE
slope of the kinetic energy spectrum m around the cutoff A. Viscous structure function model
wave number and of the molecular diffusion 共accounted for
by the Schmidt number兲. These two effects are expected to In LES carried out in the physical space, the filtered
be of critical importance in the simulation of flows involving Navier–Stokes equations and a corresponding filtered scalar
scalar transport in low to moderate Reynolds numbers or transport equation have to be evaluated,
during the transition to turbulence. As an example of appli-
cation using Eqs. 共15兲 and 共17兲, consider the case m = 5 / 3
and Sc= 0.7. We get Scti = 0.6 and Sct = 0.2, with these equa- ⳵ ūi ⳵ 1 ⳵ P̄ ⳵ S̄ij ⳵ ␶ij
tions, respectively, while for m = 3 and Sc= 0.7, we obtain + 共ūiū j兲 = − + 2␯ − , 共18兲
⳵t ⳵xj ␳ ⳵ xi ⳵xj ⳵xj
Scti = 0.36 and Sct = 0.12; i.e., the damping functions lead to a
reduction in the SGS Schmidt number to about one third.
These effects are well illustrated in Fig. 1共c兲, where again the
shape of the energy spectra from isotropic turbulence14 were ⳵ ūi
= 0, 共19兲
used. ⳵ xi

Downloaded 25 Feb 2008 to 193.136.128.19. Redistribution subject to AIP license or copyright; see http://pof.aip.org/pof/copyright.jsp
025102-4 Brun et al. Phys. Fluids 20, 025102 共2008兲

10 0.8
νt /ν 8 κt /κ Sct VSF
REF
SF
8
VSF VSF
REF REF 0.6
SF SF
6
6

0.4
4
4

0.2
2 2

0 0 0
0 2 4 6 8 10 12 0 2 4 6 8 10 12 0 2 4 6 8 10 12

z/H z/H z/H

(a) (b) (c)
FIG. 2. Downstream evolution at y / H = 0.5 of the 共a兲 eddy-viscosity ␯t, 共b兲 eddy-diffusivity ␬t, and 共c兲 SGS Schmidt number Sct in a priori tests on a plane
jet for ReH = 3000 and Sc= 0.7. The reference values 共REF兲 were obtained by box filtering the DNS, and using Eqs. 共27兲, while results for the classical structure
function model 共SF兲, and viscous structure function model 共VSF兲 were obtained with Eqs. 共23兲–共26兲, respectively.

⳵ T̄ ⳵
+
⳵ t ⳵ xi
共T̄ūi兲 = ␬
⳵ Ḡi ⳵ qi
−
⳵ xi ⳵ xi
, 共20兲
␬VSF =
␯SF
t 1−冋 1
Qx
ln共1 + Qx兲 册 , 共25兲
Sct共m兲
t
i
where S̄ij = 21 共 ⳵ūi / ⳵x j + ⳵ū j / ⳵xi 兲 is the strain rate tensor and
respectively, where Q is given by Eq. 共16兲, and x is a func-
Ḡi = ⳵T̄ / ⳵xi is the scalar gradient. The subgrid-scale stresses
tion of the angle ␣ between the local vorticity ␻ and the
tensor ␶ij = uiu j − ūiū j and the subgrid-scale scalar flux qi
¯
mean vorticity ␻m averaged over a shell of radius equal to ⌬
= Tui − T̄ūi are unknown quantities requiring modeling. The
共see Sagaut4兲,
use of the eddy-viscosity and eddy-diffusivity concepts per-
mits one to relate these unknowns to a turbulent eddy-
viscosity ␯t共z , t兲 and a turbulent eddy-diffusivity ␬t共z , t兲, x = tan 冉冊␣ ␯SF
2 ␯
t
␣ = arccos 冉 ␻m · ␻
储␻m兩兩兩兩␻储
. 冊 共26兲

␶ij − 31 ␦ij␶kk = − 2␯tS̄ij , 共21兲 The resulting VSF model constitutes therefore a continu-
ous formulation of the selective structure function model.2,15

qi = − ␬tḠi . 共22兲 B. A priori tests

The structure function 共SF兲 model8 was developed as a
physical space representation of the spectral eddy-viscosity
model derived from the EDQNM closure. It seems, there-
fore, natural to assess the present model in the physical space
using the SF model as a starting point. The eddy-viscosity
for the SF model is given by
A priori tests were performed on a DNS of a turbulent
plane jet16 for ReH = 3000 and Sc= 0.7. Explicit filtering was
made with a top-hat filter for the evaluation of the SGS
¯ / ⌬ = 3, 5, and 7. The results
model in the physical space for ⌬
demonstrated that ␯t and ␬t are roughly proportional to
¯ / ⌬兲4/3 as expected from the literature1 共not shown兲. For
共⌬
comparison we computed reference values for the eddy-

␯sf −3/2 ¯ ¯
t = 0.105CK ⌬ F2共z,⌬,t兲,
冑 共23兲
viscosity and eddy-diffusivity from the DNS data through

具␶ijS̄ij典 具qiḠi典
¯ , t兲 is the local second-order velocity structure
where F2共z , ⌬ ␯REF
t =− , ␬REF
t =− , 共27兲
2具S̄klS̄kl典 具q jq j典
function computed on a shell of diameter equal to ⌬ ¯ . This
model showed good results in a number of turbulent shear where 具·典 means a temporal averaging. Figure 2共a兲 compares
flows.8 The extension to the physical space of the present the new viscous SF eddy-viscosity ␯VSF , the classical SF
t
low Reynolds number model consists in the viscous structure eddy-viscosity8 ␯SF
t , and the reference eddy-viscosity ␯t
REF
function model 共VSF兲 developed by Brun et al.13 In this obtained with Eq. 共27兲. As can be seen the VSF model has
model the 共physical space兲 eddy-viscosity13 and eddy- the expected behavior along the jet streamwise direction; i.e.,
diffusivity are given by it damps the influence arising from the presence of a strong

冋 册
mean velocity gradient during the transition stage. Similar
1 results are obtained for the eddy-diffusivity 关see Fig. 2共b兲兴.
␯VSF
t = ␯SF
t 1− ln共1 + x兲 , 共24兲
x Figure 2共c兲 displays the evolution of the SGS Schmidt num-

Downloaded 25 Feb 2008 to 193.136.128.19. Redistribution subject to AIP license or copyright; see http://pof.aip.org/pof/copyright.jsp
025102-5 Effects of molecular diffusion on the subgrid-scale modeling
3
FSF
VSF
2.5
2
1/K1
1.5
1/β
1
0.5
0 2 4 6 8 10
z/D
(a)
FIG. 3. Centerline evolution of the 共a兲 mean velocity and passive scalar field and 共b兲 RMS velocity and RMS passive scalar field for LES of spatially evolving
round jets at ReD = 25 000 and Sc= 0.2.
ber corresponding to Figs. 2共a兲 and 2共b兲. The constant value
of Sct = 0.6 usually set in LES of passive scalar is also shown
for comparison. The potential of the new model is clearly
illustrated by the damping effect present in the initial stage of
the plane jet development.
C. A posteriori tests: LES of round jet at ReD = 25 000
Finally, LES of spatially evolving turbulent round jet
were carried out with the new model 共VSF兲 using an accurate
Navier–Stokes solver with pseudospectral methods and
sixth-order compact schemes.16,17 The Reynolds number is
ReD = 25 000 based on the inflow velocity Uo of the jet and
on the diameter D, and the molecular Schmidt number is
Sc= 0.2. The flow configuration consists of a co-flowing jet
with about Uext / Uo = 7% co-flow. Hyperbolic tangent profiles
were applied as inflow conditions for both the velocity and
the scalar,16 with a momentum thickness ␪o = D / 40. Results
are compared with the filtered structure function model2
共FSF兲 with constant SGS Schmidt number Sct = 0.6. Figure
3共a兲 shows the mean velocity 具Uc典 and mean scalar 具Tc典 de-
cay rates along the centerline z, which are in good agreement
with the experimental results18,19 with ␤ = 5.8 and K1 = 4.48
for the velocity and passive scalar fields, respectively. A
good agreement is obtained also for the root-mean square
共RMS兲 of the velocity and passive scalar fields 关Fig. 3共b兲兴
with respect to experimental results19,20 including a linear
decay behavior in the fully turbulent region. Notice that the
transition to turbulence starts about 1D earlier for both the
velocity and the scalar fields 关see Fig. 3共b兲兴 due to the inhi-
bition of the SGS model caused by the damping functions.
The scalar RMS reaches a lower 共about 10%兲 saturation
value with the VSF model than with the FSF model, an effect
which is related to the variation in SGS Schmidt number in
the transition zone. In the experiments,19 the transition to
turbulence is strongly enhanced when the flow is issued from
a smooth contraction and yields TRMS / To ⬇ 3% at about z
Downloaded 25 Feb 2008 to 193.136.128.19. Redistribution subject to AIP license or copyright; see http://pof.aip.org/pof/copyright.jsp
Phys. Fluids 20, 025102 共2008兲
To

Tc  z −1
FSF
VSF
0.1
Uo −Uext

Uc −Uext
uRM S
Uo −Uext
0.01
TRM S
To
12 1 10
z/D
(b)
= 2D. The present LES results show a clear improvement
with a similar trend when the VSF model is used.
IV. CONCLUSION
In the present study, we revisited the spectral eddy-
viscosity and eddy-diffusivity closures derived from the
EDQNM theory and accounted for molecular diffusion ef-
fects which were neglected in the original formulation dedi-
cated to high Reynolds number flows.8 Two related issues
were addressed: 共i兲 The effects of low Reynolds number and
共ii兲 a variable SGS Schmidt number for transitional flows.
The analytical present spectral formulation involves a damp-
ing of the inviscid eddy-viscosity and inviscid eddy-
diffusivity which vanishes for increasing Reynolds number
flows. The spectral model has been transposed to the physi-
cal space SGS modeling and yields a so-called viscous struc-
ture function 共VSF兲 model, which consists of an improve-
ment of the original structure function 共SF兲 model8 for flows
involving transition to turbulence. The present VSF model is
of particular interest for transitional flows with scalar trans-
port since the resulting SGS Schmidt number is no longer
constant and set to Sct = 0.6 but is reduced in low Reynolds
number flow regions, as expected from analytical and nu-
merical analysis. The new model was validated based on
both a priori tests in DNS of homogeneous isotropic turbu-
lence and DNS of turbulent plane jet and a posteriori tests in
LES of round turbulent jet with scalar transport.
ACKNOWLEDGMENTS
The computing resources were provided by IDRIS-
France. The authors gratefully acknowledge the financial
support from the Centre National de la Recherche Scienti-
fique 共CNRS兲. The study was part of DFG-CNRS research
025102-6 Brun et al.
project FR 507, titled “LES of complex flows.” Carlos B. da
Silva is supported by the Portuguese Minister of Science and
Technology 共MCTES兲 under “Ciência 2007”.
1
M. Lesieur, Turbulence in Fluids, 4th ed. 共Springer, New York, 2008兲.
2
M. Lesieur and O. Métais, “New trends in large-eddy simulations of tur-
bulence,” Annu. Rev. Fluid Mech. 28, 45 共1996兲.
3
C. Meneveau and J. Katz, “Scale invariance and turbulence models for
large-eddy simulation,” Annu. Rev. Fluid Mech. 33, 1 共2001兲.
4
P. Sagaut, Large Eddy Simulation for Incompressible Flows – An Intro-
duction 共Springer-Verlag, Berlin, 1998兲.
5 15
B. Vreman, B. J. Geurts, and H. Kuerten, “Large-eddy simulation of the
turbulent mixing layer,” J. Fluid Mech. 339, 357 共1997兲.
6 16
M. Lesieur and P. Comte, “Favre filtering and macro-temperature in large-
eddy simulations of compressible turbulence,” C. R. Acad. Sci., Ser. IIb
Mec. 329, 363 共2001兲.
7 17
M. Lesieur and R. Rogallo, “Large eddy simulation of passive scalar dif-
fusion in isotropic turbulence,” Phys. Fluids A 1, 718 共1989兲.
8
O. Métais and M. Lesieur, “Spectral large-eddy simulation of isotropic and
stably stratified turbulence,” J. Fluid Mech. 239, 157 共1992兲.
9
P. Moin, K. Squires, W. H. Cabot, and S. Lee, “A dynamic subgrid-scale
model for compressible turbulence and scalar transport,” Phys. Fluids A 3,
2746 共1991兲.
10
J. C. André and M. Lesieur, “Influence of helicity on the evolution of
isotropic turbulence at high Reynolds number,” J. Fluid Mech. 81, 187
共1977兲.
11
J. R. Herring, D. Schertzer, M. Lesieur, G. R. Newman, J. P. Chollet, and
Downloaded 25 Feb 2008 to 193.136.128.19. Redistribution subject to AIP license or copyright; see http://pof.aip.org/pof/copyright.jsp
Phys. Fluids 20, 025102 共2008兲
M. Larcheveque, “A comparative assessment of spectral closures as ap-
plied to passive scalar diffusion,” J. Fluid Mech. 124, 411 共1982兲.
12
E. Lamballais, O. Métais, and M. Lesieur, “Spectral-dynamic model for
large-eddy simulations of turbulent rotating channel flow,” Theor. Com-
put. Fluid Dyn. 12, 149 共2005兲.
13
C. Brun, C. B. da Silva, O. Métais, and R. Friedrich, “A new mixed model
based on the velocity structure function,” in Advances in LES of Complex
Flows, edited by R. Friedrich and W. Rodi 共Kluwer, Dordrecht, 2002兲, pp.
50–64.
14
C. B. da Silva and J. C. F. Pereira, “Analysis of the gradient-diffusion
hypothesis in large-eddy simulations based on transport equations,” Phys.
Fluids 19, 035106 共2007兲.
C. Ackerman and O. Métais, “A modified selective structure function
subgrid-scale model,” J. Turbul. 2, 011 共2001兲.
C. B. da Silva and O. Métais, “On the influence of coherent structures
upon interscale interactions in turbulent plane jets,” J. Fluid Mech. 473,
103 共2002兲.
G. Balarac, O. Métais, and M. Lesieur, “Mixing enhancement of coaxial
jets through inflow forcing: A numerical study,” Phys. Fluids 19, 075102
共2007兲.
18
H. J. Hussein, S. P. Capp, and W. K. George, “Velocity measurements in a
high-Reynolds-number, momentum conserving, axisymmetric, turbulent
jet,” J. Fluid Mech. 258, 31 共1994兲.
19
J. Mi, D. S. Nobes, and G. J. Nathan, “Influence of jet exit conditions on
the passive scalar field of an axisymmetric free jet,” J. Fluid Mech. 432,
91 共2001兲.
20
S. C. Crow and F. H. Champagne, “Orderly structure in jet turbulence,” J.
Fluid Mech. 48, 547 共1971兲.