Wall-Bounded Turbulent Flows

May 5, 2010

Nearly a century ago, Prandtl (1904) presented the concept of “bound-

ary layer” which brought a revolution in fluid mechanics. The existence of
boundary layer is due to the no-slip condition at the solid boundary. Prandtl
argued that for high Reynolds number flows, viscous effects should be con-
fined in a thin layer along the solid surface. Such a layer must always exist,
since the velocity must rapidly change from the freestream value outside to
zero on the solid boundary. This variation inevitably produces large gradi-
ents and hence, significant shear forces. Figure 1 shows a schematic of the
boundary layer over a flat surface.

Figure 1: Schematic of a boundary layer

Although this region occupies a small portion of the total flow field, it
cannot be negelected. The study of boundary layers is important for the
calculation of skin friction drag, transition to turbulence and flow separa-
tion among others. Near-wall flows can be either laminar or turbulent. In
laminar wall-bounded flows, acceleration/deceleration takes place in layers
or laminates of fluid that interact with each other through molecular viscos-
ity. Most wall-bounded flows of practical concern are turbulent. Turbulent
flows are characterized by eddies or swirls of fluid that vary in space and
time. Figure 2 shows various configurations of wall-bounded flows, where d

1
 2

represents boundary layer “thickness”.

Outside the boundary layer, the flow behaves as if it is inviscid. Thus

the flow domain is usually divided into two regions with different governing
equations. For the outer region, viscous effects can be neglected and the
inviscid Euler equations can be used. For the boundary layer, a simplified
version of the Navier-Stokes equations, based on its small thickness, is used.

Figure 2: Configurations of wall-bounded flows from Pope (2000).

 3

1 Equations for the Turbulent Boundary Layer

with Zero Pressure Gradient
When a turbulent flow is decomposed according to Reynolds decomposition
into a fluctuating and mean part, the equations for an incompressible and
two-dimensional (in the mean) turbulent boundary layer are the continuity
equation,
∂U ∂V
+ =0 (1)
∂x ∂y
and the momentum equation,
 
∂U ∂U 1 dP ∂ ∂U 0 0
U +V =− + ν −uv . (2)
∂x ∂y ρ dx ∂y ∂y
Capital letters correspond to mean quantities and lower case letters with
a prime denote fluctuations. x is the streamwise coordinate and y is the wall-
normal coordinate. The overbar in −u0 v 0 denotes an average. The velocity
components U and V are in directions x and y respectively. Symbols ρ and
ν denote the fluid density and kinematic viscosity respectively.

1.1 The Inner Layer

The flow region very near the wall is critical because many flow parameters
which are important from engineering point of view are determined in this
layer. According to Pope (2000), the inner layer extends upto y/δ < 0.1 and
consists of a viscous sublayer (y + < 5), a buffer layer (5 < y + < 30) and the
overlap layer (y + > 50, y/δ < 0.1). In the case of a zero pressure gradient
flow, the term dP/dx is zero. The inner layer is dominated by viscous forces
and the convection terms are negligible, and thus equation (2) can be reduced
to,
 
∂ ∂u 0 0
0= ν −uv (3)
∂y ∂y
The right hand side of this equation is the gradient of the viscous and
Reynolds shear stresses. Very near the wall (viscous sublayer), ν ∂u
∂y
-u0 v 0 ,
and thus we can assume,
∂u τw
ν = = constant (4)
∂y ρ
integrating the above equation,
 4

Z u Z y
τw
du = dy (5)
0 ρν 0
from which we can get, q q
u = τρν
wy
= u2τ νy ⇒ uuτ = yuτ
ν
where uτ = ν( ∂u
∂y
)w = τw
ρ
and w stands for
wall or y=0.

Finally we can write,

u+ = y + (6)
This equation is the linear law of velocity and is applicable only in the
viscous sublayer (y + < 5). If equation (4) is scaled in wall units, i.e., u+ =
u/uτ , y + = yuτ /ν and τ + = τ /(ρu2τ ), then it can be written as:

du+ +
τ+ = +
− u0 v 0 = 1 (7)
dy
This implies that all dependence on x (streamwise coordinate) is included
in uτ .

1.2 Outer Layer: Velocity Defect Law

According to Pope (2000), the outer layer exists for y + > 50, which evidently
overlaps with the inner layer to some extent. The general characteristics of
the outer region are inviscid flow resembling that of wall-free turbulent flow.
The existence of the Reynolds stresses in the outer region results in a drag
on the flow and generates a velocity defect (Ue − u), which when normalized
by uτ is considered to be only a function of y/δ. Therefore, we can write,
Ue − u
= F (y/δ) (8)
uτ
This is called the velocity defect law. According to Kline et al. (1967),
flow region outside the inner layer is called the wake region. Empirical laws
have been proposed for F, but no theoretical solution is available even in the
ZPG case.

1.3 The Overlap Region: Log Law

This is a region of overlap between inner and outer layers at large Reynolds
numbers. In the inner layer, the velocity profile is described by the law of
the wall,
 5

u yuτ
= = f (y + ) (9)
uτ ν
Logically, the velocity profiles of inner and outer layers should match in
the overlap layer. But instead of matching the velocities of inner and outer
layers, it is more convenient to match their gradients. From equations (9)
and (8), the velocity gradients in the inner and outer layers are given as,

dU u2 df
= τ +, (10)
dy ν dy
dU uτ dF
= . (11)
dy δ dη
where η = y/δ. Equating equation(10) and equation(11), and multiplying
by y/uτ , we get,
dF df 1
η = y+ + = (12)
dη dy κ
The term on left hand side is a function of η only and the term on right
hand side is a function of y + only, both sides therefore must be equal to some
universal constant, 1/κ, where κ is the Von Karman constant. Experimen-
tally, k ∼ 0.41. Integrating equation (12) yields,

1
f (y + ) = ln y + + A, (13)
κ
1
F (η) = ln η + B. (14)
κ
Experiments show that A = 5.0 and B = -1.0 for a smooth flat plate,
hence we can write equations (13) and (14) as,

u 1 yuτ
= ln + 5.0, (15)
uτ κ ν
Ue − u 1 y
= ln − 1.0 (16)
uτ κ δ
These equations are valid only for large y + and small y/δ. Due to the
logarithmic form of the velocity profile, the overlap layer is also called the
logarithmic layer. The above-mentioned logarithmic layer equation was first
derived by Clark.B.Millikan in 1938.
 6

2 Scaling of Wall-Bounded Turbulent Flows

Self-similarity is an important concept in fluid dynamics in general and tur-
bulent flows in particular. Self-similarity is achieved, if for a given geometry,
various turbulence statistics (mean, rms, Reynolds stresses, etc...) measured
at different Reynolds numbers and in different facilities collapse to a univer-
sal profile when scaled by proper length and velocity scales. Self-similarity is
important because it allows to extrapolate the lower Reynolds number labo-
ratory results to real world high Reynolds number applications.

A flow usually needs two scales to be described universally ,i.e. velocity

and length (see figure 3). In unsteady flows, a time scale may also be nec-
essary. Apart from length and velocity, scalings have also been sought for
statistics like correlations, spectra etc.

Figure 3: General purpose of scaling

Earlier research into self-similarity and equilibrium boundary layers by

Clauser (1954) and Rotta (1962) assumed the friction velocity uτ as veloc-
ity scale for the whole boundary layer, which was later contested by many
researchers. They assumed a self-similarity of the following form:

Ue − U y −u0 v 0 y
= f ( ), = g 12 ( )
uτ ∆ u2τ ∆

u0 2 y v02 y
2
= g11 ( ), 2
= g22 ( )
uτ ∆ uτ ∆
 7

For velocity and Reynolds stresses, uτ has classically been used as internal
scale and Ue as external scale because all wall-bounded flows have distinct
p
inner and outer regions. Here uτ is the friction velocity defined as τw /ρ
and Ue is the local freestream flow velocity. Near separation, uτ becomes zero
and thus the wall-unit scaling becomes difficult to implement. In most cases,
the inner velocity scaling based on uτ was unable to collapse the velocity
and Reynolds stress profiles in the outer layer and thus an effect of Reynolds
number was evident in the normalised profiles.

For length, the classical inner scaling has been uτ /ν and the boundary
layer thickness δ as outer scale. Owing to the difficulty of precisely mea-
suring boundary layer thickness, especially over curved surfaces, many re-
searchers have instead preferred to use boundary layer momentum thickness
θ or dispalcement thickness δ ∗ as length scales because these are integral
quantites and relatively easy to compute. Clauser (1954) originally proposed
∆ = (δ ∗ Ue /uτ ) as the length scale but it was largely replaced by boundary
layer thickness δ. The usual length scales for fully-developped channel flow,
fully-developped pipe flows and bumps/humps/ramps are the half-channel
height H, diameter D and maximum height h respectively. According to
Pope (2000), a flow is fully-developped when the velocity statistics no longer
change with streamwise coordinate.

Castillo & George (2001) reconsidered the similarity analysis of the tur-
bulent boundary layer and suggested outer velocity scales as Ue , Ue2 and
Ue2 dδ/dx for the velocity defect, Reynolds normal stress and Reynolds shear
stress respectively. Thus their similarity analysis took the following form:

Ue − U y −u0 v 0 y
= f ( ), dδ
= g 12 ( ),
Ue δ Ue2 ( dx ) δ

u0 2 y v02 y
= g11 ( ), = g 22 ( )
Ue2 δ Ue2 δ
The validity of Ue , Ue2 and Ue2 dδ/dx as outer velocity scales was put in
question by Maciel et al. (2006a) for flows with strong streamwise pressure
gradients. They pointed out that Reynolds stresses scaled with Ue2 can grow
by an order of magnitude for a flow experiencing a strong favorable pressure
gradient and then a strong adverse pressure gradient like on the suction side
of an airfoil.
 8

Zagarola & Smits (1998) proposed Uc − Ub as the outer velocity scale for
turbulent pipe flow, where Uc is the centreline velocity and Ub is the bulk
mean velocity. They showed that Uc − Ub is better than uτ for collapsing
the velocity defect profiles. For turbulent boundary layers, they proposed
an equivalent scale Ue δ ∗ /δ, which proved to be a valid scale for the outer
layer. Their scale, Uzs , not only removed Reynolds number effects but also
the upstream condition effects were also noted to disappear. Later Maciel
2
et al. (2006b) showed that Uzs is a valid scale for Reynolds stresses. To com-
plement the Zagarola & Smits (1998) velocity scale Uzs , Maciel et al. (2006a)
replaced uτ by Uzs to get ∆zs = Ue δ ∗ /Uzs , which is the same as boundary
layer thickness δ.

DeGraaff & Eaton (2000) proposed a mixed scaling uτ Ue for the Reynolds
stresses but scaling was found not to be universal in the outer region of the
boundary layer. They did notice however the disappearance of Reynolds
number effects with the mixed scaling.

Stanislas et al. (2008) carried out stereo PIV experiments of flat plate
turbulent boundary layer over a range of Reynolds number Reθ from 7800
to 15000. They argued that if Kolmogorov scales determines the friction
at the wall and if they are more representative of dissipation than , then
they should replace the viscosity ν as a scale. The viscosity shall then be-
come a kind of ‘turbulence thermodynamic’ quantity by contributing towards
Kolmogorov scale (η). They found that very near to the wall, the ratio of
Kolmogorov scales to the integral scales is of order 1 and thus near the wall
the Kolmogorov scale η should be used as a length scale. They plotted the
mean velocity U/Ue as a function of y/η and obtained a good collapse near
the wall. However, a good collapse was not observed in the outer layer and
to use δ as the outer length scale, they defined a new length scale η ∗ in form:
 
∗ δ
η = min η, . (17)
A
Here A is of the order of 1000. The function was designed such that δ
shall take over as a length scale if Kolmogorov scale η is greater than δ/1000.
Therefore, as summarised above, the existence of a plethora of velocity,
length and stress scales proves that we are still far from having a single
universally accepted set of scales for turbulent wall-bounded flows.
 9

3 Vortical Structures of Wall-Bounded Flows

A turbulent flow consists of coherent and incoherent motions with varying
scales. These coherent structures were also termed as “organised motion” in
an effort to search for order in the apparent disorder of turbulence. These
structures have been observed in a variety of types according to size, shape,
orientation, aspect ratio, convection velocities, strength, etc. Various kinds of
vortical structures have been observed like vortex “worms” found in isotropic
turbulence (Jiménez et al., 1993), vortex “braids” in turbulent shear lay-
ers (Rogers & R.D., 1994), quasi-streamwise vortices (Robinson, 1991a) and
“hairpin” vortices found in wall turbulence (Adrian et al., 2000) etc. Figure
4 shows a model of hairpin vortex.

Figure 4: Model of a hairpin vortex.

Despite an extensive literature on turbulence structure, the turbulence

community is unable to provide a universally accepted definition for ‘vorti-
cal structure’ or ‘coherent structure’. There is also no concensus over the
definition of an ‘Eddy’ or ‘Vortex’, which is the basic element for coherent
structures.

Robinson (1991a) defined ‘coherent motion’ for turbulent flows as “three-

dimensional region of flow over which at least one fundamental flow variable
(velocity component, density, temperature, etc.) exhibits significant correla-
tion with itself or with another variable over a range of space and/or time
that is significantly larger than the smallest local scales of the flow”. In words
 10

of Fazle Hussain (1986), “A coherent structure is a connected turbulent fluid

mass with instantaneously phase-correlated vorticity over its spatial extent”,
thus making coherent vorticity as primary identifier of coherent structures.
Due to fluid continuity, all motions in fluid possess some degree of spatial
coherence, therefore only spatial coherence is not sufficient to define an or-
ganised motion and temporal coherence should also be taken into account.

Vortical structures play an important role in heat, mass and momentum

transport in turbulent flows and have been called as “the sinews and mus-
cles of turbulence”. The role that they play for heat, mass and momentum
transport makes them important for combustion, chemical reactions as well
as drag and aerodynamic noise. The early research into turbulence struc-
ture was largely conducted at low Reynolds numbers (Reθ < 5000) where
the range of turbulent scales was limited, direct numerical simulations were
possible and flow visualisation and experimentation were easy to conduct.

To fully understand the vortical structures and be able to model them,

we need not only the kinematic properties (size, scaling properties, shape,
vorticity, energy) but also the dynamic properties (origin, stability, growth,
evolution etc). In this thesis, we shall mostly focus on the kinematic prop-
erties of vortical structures.

The existence of hairpin vortices inclined to the streamwise direction, first

proposed by Theodorsen (1952), was confirmed through flow visualization by
Head & Bandyopadhyay (1981). They further proposed that this structure
is inclined at 45 ◦ to the streamwise direction. This three-dimensional vor-
tical structure is composed of a pair of counter-rotating legs that are joined
through a head segment. It has been observed by Adrian et al. (2000) that
these vortices travel together in groups called ‘hairpin packets’. However, a
detailed description of their size, shape and dynamics is still not available.
The aspect ratio of these vortices changes with Reynolds number and a hair-
pin vortex becomes a horseshoe vortex at lower Reynolds numbers with lower
aspect ratio.

Theodorsen (1952) was the first to propose the presence of hairpin vor-
tices attached to the wall in the incompressible turbulent boundary layers but
later in the twentieth century, advanced experimental and numerical tech-
niques like PIV (Particle Image Velocimetry) and DNS (Direct Numerical
Simulation) permitted a further insight into vortical structures of near-wall
flows.
 11

The kinematic properties of the streamwise vortical structures very near

to the wall like size, orientation, etc were first investigated by Blackwelder
& Eckelmann (1979), who experimentally detected counter-rotating vortex
pairs by means of conditional analysis. They found the expected vortex cen-
ter at y + ≈ 15, ro+ ≈ 15, streamwise length of around 200+ and spanwise
spacing of 50+ − 200+ .

Carlier & Stanislas (2005) recently used Stereoscopic PIV to investi-

gate the vortical structures in the log layer of a turbulent boundary layer
at Reynolds numbers Reθ between 7500 and 19000. They used a pattern-
recognition algorithm and an Oseen model was fitted to vortex core to get
vortex parameters like radius, circulation, vorticity etc. Their results showed
that vortical structures have a caned-shape tilted downstream at angle of
approximately 45 ◦ . The vortices had an origin at y + ≈ 25 and their radius
ro+ was in the range 18-25 for a range of Reynolds number Reθ from 7500 to
19000. The size of vortices was noted to increase with the distance from the
wall with an approximately constant circulation (Γ+ = (Γ/ν) ≈ 235 − 250).
Their results were in good agreement with the existing literature.

They also found that the mean advection velocity of spanwise vortices was
equal to the local mean flow velocity. Carlier & Stanislas (2005), using signed
swirling strength education criteria, detected dense populations of prograde
and retrograde vortices with prograde vortices mostly found near the wall
and decreasing with wall normal distance. Retrograde vortices, on the other
hand, were found to increase with wall-normal distance till the top of the log
layer and then found to decrease further upwards. Wu & Christensen (2006)
also performed a PIV study on vortical structures and obtained almost the
same results regarding effects of Reynolds number, prograde and retrograde
vortices and advection velocities of vortices.

Elsinga et al. (2007) conducted tomographic PIV of a turbulent boundary

layer at Reθ = 1900 and employed Q-criterion to educe the vortical struc-
tures.They observed mostly the asymmetric hairpin vortices in the lower part
of the boundary layer (y/δ < 0.5). Streamwise vortices were found to be lo-
cated near the sweep events and to have an approximate length of 0.2δ. With
increasing distance from the wall, the length was noted to increase and at
y/δ = 0.6, the largest structure was found to be approximately 0.5δ. They
also provided some details on hairpin packets. A hairpin packet consisting of
three vortices at y/δ = 0.35 was found to be approximately 0.3δ wide, 0.35δ
high and 0.8δ long.
 12

DNS (Direct Numerical Simulation) studies of near-wall flows remained

restricted to low-Reynolds number regime for a long time with Reτ = huτ /ν
from 180 to 600 but recently some direct simulations have been carried out at
Reynolds numbers upto Reτ = 2000. The majority of these simulations were
of fully-developped channel flow but recently Pirozolli et al. (2008) carried
out a DNS of a supersonic spatially-developping boundary layer at Mach 2
and Reθ = 950.

Using the DNS database of Spalart (1988), Robinson (1991b) found that
the number density of vortices is maximum at y + ≈ 10 − 50, the vortex
core radius ro+ ≈ 5 − 20 and the circulation Γ+ = (Γ/ν) ≈ 60 − 250. del
Álamo et al. (2006) performed a DNS of turbulent channel flow and classi-
fied the population of vortices into two groups of wall-detached eddies and
attached eddies, in terms of their dependance on the distance from the wall.
The typical size of the wall-detached eddy was found to be of order of the
Kolmogorov length scale (η), and not to depend on the distance from the
wall. The second group of the Townsend’s wall-attached eddies was found
to closely resemble the hairpins as described by Adrian et al. (2000). Das
et al. (2006) performed DNS of channel flow at Reτ upto 1270 and found out
that the size of vortices ranges from 6 and 90η. For validation, they fitted a
Burgers vortex model to the detected vortices.

Skote & Henningson (2002) performed DNS of a turbulent boundary layer

undergoing adverse pressure gradient with a displacement thickness Reynolds
number U δ ∗ /ν of 400 at the starting position of the simulation (x=0), where
U is the freestream velocity. They found that near-wall streaks become weak
due to adverse pressure gradient and the spacing in viscous units tend to
increase. The streaks were noted to vanish at separation. The length in the
streamwise direction was found to be 3400 wall units. APG was found to
have a damping effect on these structures. Spacing between structures was
noted to increase from 100 (as in Zero Pressure Gradient) to 130 viscous
units towards the end of the computational domain.

Stanislas et al. (2008) studied streamwise vortical structures in a turbu-

lent boundary layer with Reynolds number Reθ ranging from 7800 to 15000.
They obtained velocity data through stereoscopic PIV in the (y,z)-plane. The
schemes for vortex education and validation were the same as used by Car-
lier & Stanislas (2005) but they added Linear Stochastic Estimation (LSE)
and conditional analysis to further study the form and characteristics of the
vortices. In addition to fit to an Oseen vortex model, they also used linear
stochastic estimation of swirling events and correlation of swirling strength
 13

to estimate the radius of educed vortices and found that all three methods
were in good agreement. Through one- and two-point LSE, they proposed
that the most probable form of streamwise vortices is strongly asymmetri-
cal and resembles a cane or hook rather than an hairpin. The mean radius
of vortices was found to be around 6η with a range from 1 to 30η. The
mean vorticity was found to be about 1.6τ −1 with a range from 0 to 10τ −1 .
With Kolmogorov scaling, they found that distributions of vortex mean ra-
dius and vorticity are independent of the wall-normal location of the vortex
centre. For the origin of vortices, they found that most of the vortices are
created near the wall and are then lifted up away from it.

Wu & Moin (2009) performed a DNS of zero-pressure gradient flat-plate

boundary layer and took it from Blasius layer at Reθ = 80 through tran-
sition to 1000. They argued that none of the previous simulations were of
genuinly spatially-developping turbulent boundary layers. They found that
the instantaneous flow fields in both transitional and turbulent regions were
populated by hairpin vortices. In contrast to the earlier studies, they found
out that the hairpin vortices were mostly quasi-symmetric. A dense popu-
lation of hairpin vortices in near-wall region, referred to as “hairpin forest”
was found to exist close to the wall.

In summary, while there exists some concensus on the geometric features

of vortical structures in wall-bounded flows, still a concensus on the shape
and dynamical properties is far from being achieved. Experiments and simu-
lations have mostly been performed on the canonical flows like zero pressure
gradient boundary layers and fully-developped channels. Studies of vortical
structures with additional strain rates like adverse pressure gradient or sur-
face curvature are indeed rare.

References
Adrian, R., Meinhart, C. & Tomkins, C., 2000 Vortex organization in
the outer layer of the turbulent boundary layer. J. Fluid Mech. 422, 1–54.

Blackwelder, R. F. & Eckelmann, H., 1979 Streamwise vortices asso-

ciated with the bursting phenomenon. J. Fluid Mech. 94, 577–594.

Carlier, J. & Stanislas, M., 2005 Experimental study of eddy structures

in a turbulent boundary layer using particle image velocimetry. J. Fluid
Mech. 535, 143–188.
 14

Castillo, L. & George, W., 2001 Similarity analysis for turbulent bound-
ary layers with pressure gradient: Outer flow. AIAA J. 39, 41–47.

Clauser, F. H., 1954 The turbulent boundary layer. Advances Applied

Mechanics 4, 1–51.

Das, S. K., Tanahashi, M., Shoji, K. & Miyauchi, T., 2006 Statistical
properties of coherent fine eddies in wall-bounded turbulent flows by direct
numerical simulation. Theor. Comput. Fluid. Dyn. 20 (2), 55–71.

DeGraaff, D. B. & Eaton, J. K., 2000 Reynolds-number scaling of the

flat-plate turbulent boundary layer. J. Fluid Mech. 422, 319–346.

del Álamo, J. C., Jiménez, J., Zandonade, P. & Moser, R. D., 2006
Self-similar vortex clusters in the logarithmic region. J. Fluid Mech. 561,
329–358.

Elsinga, G. E., Kuik, D. J., van Oudheusden, B. W. & Scarano,

F., 2007 Investigation of the three-dimensional coherent structures in a
turbulent boundary layer with tomographic PIV. AIAA paper , 1–11.

Fazle Hussain, A., 1986 Coherent structures and turbulence. J. Fluid

Mech. 173, 303–356.

Head, M. R. & Bandyopadhyay, P., 1981 New aspects of turbulent

boundary layer structure. J. Fluid Mech. 107, 297–338.

Jiménez, J., WRAY, A., SAFFMAN, P. & ROGALLO, R., 1993 The
structure of intense vorticity in isotropic turbulence. J. Fluid Mech. 255,
65–90.

Kline, S. J., Reynolds, W. C., Schraub, W. C. & Runstadler,

P. W., 1967 The structure of turbulent boundary layers. J. Fluid Mech.
30, 741–773.

Maciel, Y., Rossignol, K.-S. & Lemay, J., 2006a Similarity in the
outer region of adverse-pressure-gradient turbulent boundary layers. AIAA
Journal 44 (11), 2450–2464.

Maciel, Y., Rossignol, K.-S. & Lemay, J., 2006b A study of a turbu-
lent boundary layer in stalled-airfoil-type flow conditions. Experiments in
Fluids 41, 573–590.
 15

Pirozolli, S., Bernardini, M. & Grasso, F., 2008 Characterization of

coherent vortical structures in a supersonic turbulent boundary layer. J.
Fluid Mech , 205–231.

Pope, S. B., 2000 Turbulent Flows. Cambridge Univ. Press.

Robinson, S., 1991a Coherent motions in the turbulent boundary layer.

Ann. Rev. Fluid Mech. 23, 601–639.

Robinson, S., 1991b The kinematics of turbulent boundary layer structure.

NASA TM 103859.

Rogers, M. M. & R.D., M., 1994 Direct numerical simulation of a self-

similar turbulent mixing layer. Phys. Fluids 6, 903–923.

Rotta, J., 1962 Turbulent boundary layers in incompressible flow. Progress

in Aerospace Sciences. 2, 1–95.

Skote, M. & Henningson, D. S., 2002 Direct numerical simulation of

separating turbulent boundary layers. J. Fluid Mech. 471, 107–136.

Spalart, P., 1988 Direct simulation of a turbulent boundary layer up to

rθ = 1410. J. Fluid Mech. 187, 61–99.

Stanislas, M., Perret, L. & Foucaut, J. M., 2008 Vortical structures

in the turbulent boundary layer: a possible route to a universal represen-
tation. J. Fluid Mech. 602, 327–382.

Theodorsen, T., 1952 Mecanism of turbulence. In Proc. Midwest Conf.

Fluid Mech. 2nd Edn. Columbus, Ohio, pages 1–18.

Wu, X. & Moin, P., 2009 Direct numerical simulation of turbulence in a

nominally zero-pressure-gradient flat-plate boundary layer. J.Fluid.Mech
630, 5–41.

Wu, Y. & Christensen, K. T., 2006 Population trends of spanwise vor-

tices in wall turbulence. J. Fluid Mech. 568, 55–76.

Zagarola, M. & Smits, A. J., 1998 Mean-flow scaling of turbulent pipe

flow. J. Fluid Mech 373, 33–79.