2.

External Flows

H.-H. FERNHOLZ
With 7 Figures

"External" turbulent flows are those in which the turbulent region is

significantly affected by at.most one wall. An example is the boundary
layer on an airfoil (Fig. 2.1) and an example without walls is the wake
behind the airfoil. Although the interaction between a turbulent shear
layer and the non-turbulent "inviscid" flow outside it is often important
even in external flows, it is convenient to defer the treatment of interac-
tions until Chapter 3 on internal flow (bounded by at least two walls)
in which they are always important. Conversely the treatment of
boundary layers in the present chapter is applicable to turbomachine
blades and other internal flows. Several topics such as streamline-
curvature effects and separation are introduced in the present chapter
and illustrated by simple external-flow examples, while the more
complicated flow situations are discussed in Chapter 3.

2.1 Introduction
We begin with a brief survey of some general turbulent shear flow
configurations which, merged or distorted, make up the more complex
flow patterns which are often found in practice but which are often not
accessible to experiment or calculation. The greater part of this chapter
will be concerned with a discussion of steady two- and three-dimensional
turbulent boundary layers at low Mach numbers. Boundary layers are
the commonest members of the class of "thin shear layers" (Chapt. 1)
and provide a convenient framework within which to discuss the effects
of boundary conditions on shear layers. If the wall or free stream
boundary conditions are changed, the shear layer will feel this change
as a more or less strong perturbation. Most shear layers in real life
undergo such perturbations, and the difference between this chapter
and some of the existing reviews of th~ subject is that we have tried to
give perturbed shear layers a share of space proportional to their im-
portance. Unperturbed free shear layers--the classical wake, jet and
mixing layer-- are rather rare in practice and the data on growth rate,
etc., which are normally needed can be summarized quite briefly (Sect.
5.9). Free shear layers near solid boundaries are usually perturbed and
 46 H.-H. FERNHOLZ

Seporotion region Seporotionbubble

-3
A_p
%0

T-1
1;o -011-~'5~x/~ ~ 10
0 ~.~
1~-"~-~" xll 1'.0
I ~ ," x/l IJ v h
Fig. 2.1. Attached and separating flow (after THWAITES[2.446], Fig. V.13)

often form part of a separation/reattachment flow. Of these flows

separation bubbles, in which the free shear layer remains very close to
the surface, are discussed in Subsection 2.3.6. The more extensive se-
parated regions, whose turbulence behavior is of most importance in
internal flows, are discussed in Section 3.4. Basic references on (un-
perturbed) free shear layers are given in [-1.7, 22, 2.1], and good entries
to the current research literature are provided by [2.2, 3-1. Ref. [2.2]
contains extensive data tabulations. See Section 1.10 for work on orderly
structure in jets. A very comprehensive survey of aircraft trailing
vortices, with particular reference to turbulence structure, has recently
appeared [2.4], and we therefore omit this important subject. Vortices
frequently appear as constituent parts of turbulent flow fields, but in
that case are better treated as examples of turbulence with superimposed
rotation (Sect. 3.1).
For most of the present chapter we concentrate on constant-
property ("incompressible") flows for simplicity, but the chapter is
concluded by a discussion of compressibility effects on turbulent bound-
ary layers. In all cases discussed here, the boundary conditions can be
defined clearly. This has been intentional. It should be mentioned,
however, that there exist many external flows, aeronautical or otherwise,
where the definition of the conditions at the interface of two merging
shear flows is very difficult or where there is no asymptotic transition
between the boundary layer and the outer flow if the latter is highly
disturbed. However such configurations are even more important in
internal flows, and are therefore discussed in Chapter 3; the results of
that chapter are at least qualitatively applicable to external flows.
 External Flows 47

2.2 Flow Configurations and Boundary Conditions

Turbulent shear flows can be divided into three groups according to
the number of fixed boundaries:
1) Free turbulent shear flows bounded by no wall, such as jets, wakes
and mixing layers (Fig. 2.2), and plumes (Chapt. 4).
Jets and wakes differ in principle from each other only in the sign of
the momentum creating them. Among them the turbulent jet is the most
common flow configuration, ranging in its application from propulsive
devices for rockets and aircraft to pneumatic control systems.
2) Turbulent shear flows bounded by one free and one fixed boundary,
such as boundary layers or wall-jets, where the wall may be straight
or curved, permeable or impermeable.
For both jet and wall-jet the velocity of the ambient flow can have the
same or the opposite sign as the shear flow.
3) Turbulent shear flows bounded by two or more fixed boundaries,
such as pipe, duct and channel flows which are dealt with in Chapter 3.
In this book a "pipe" is a closed duct of circular cross section, a "duct"
has a non-circular cross section and a "channel" has a free surface.
These basic flow phenomena do not always appear in their pure
state but often as a combination of two or more configurations, e.g.,
as a separated and reattaching boundary layer, a jet emanating into a
duct flow (ejector problems or a flap configuration with merging shear
layers (Fig. 2.3). In this chapter the discussion will be confined almost
entirely to simple thin or slender shear layers, having one sign of shear
throughout. It is obvious from Fig. 2.2 that it would be rather fortuitous
if the shear layers on either side of the velocity maximum in a wall jet
flow, say, behaved exactly like an isolated thin shear layer, however well
the flow obeyed the thin shear layer approximation; shear-layer inter-
actions require further discussion (Chapt. 3). Other cases where the
i

I ~' I/ I I II ~~ iI \\

Jet Wake Mixfng layer Boundary layer Wall jet

Fig. 2.2. Some basic turbulent shear flow configurations

48 H.-H. FERNHOLZ

Seporoted region Reottochment region

Slot woke Aerofoil B.L.

Fig. 2.3a--c. Some complex turbulent shear flow configurations. (a) boundary layer flow
over a step, (b) duct flow in an internal ejector, (c) flow over an aerofoil with a front slot
(from HORTON 1-2.267])

concept of simple shear layers is violated occur in flow regions in which

the thin shear layer is strongly perturbed. Such a perturbation may be
caused by a pressure gradient that greatly exceeds the Reynolds stress
gradient, as in an impinging jet or in the boundary layer separating in
front of a step (Fig. 2.3a). In such regions a quasi-inviscid flow model
suffices, few data on turbulence structure are available, and lengthy
discussion would be unprofitable.
We shall, however, deal with shear flows, initially in a state of equi-
librium 1, which are more or less strongly perturbed by a change of the
boundary conditions and which tend to a new state of equilibrium or
relax to the original equilibrium according to the boundary conditions
downstream from the perturbation. There exists a strong motivation
for investigating perturbed shear flows apart from their practical im-
portance. CLAUSER[2.5-1 suggested that useful basic information about
turbulent flows could be obtained by studying their response to per-
turbation, after the style of the classical "black box" problem. Many
experimenters have taken up this idea. A survey of turbulent boundary
layers subject to perturbations in general was provided by TANI [2.61,
accompanying the exhaustive review of boundary-layer data by COLES
i A concept to be defined in Subsection 2.3.2.
 External Flows 49

and HIRST [2.7]. In the following the effects of perturbations on simple

thin and slender shear layers, in the first instance boundary layers, will
be emphasized. These perturbations are caused by changes of the
boundary conditions at the outer edge of the boundary layer or at the
wall and may affect the outer or inner region of the boundary layer
only, or, as in many cases, the boundary layer as a whole.

Boundary and Initial Conditions

Once the flow configuration, i.e., the type of the shear layer, is established,
the (parabolic) partial differential equations describing it (Sect. 1.7)
must be supplemented by initial conditions and by boundary conditions
at the edges of the shear layer. For many problems in aeronautics these
conditions can fortunately be specified without too great difficulties,
so that a great number of important flow configurations can be classified
as well-posed problems. This has certainly favored the solution of
problems in aeronautics, and it is hoped that this experience can be
transfered to more complex flows, both in the same field and in other
closely related areas such as mechanical and chemical engineering
where the occurring problems are often ill defined.
Before we can solve the thin-shear-layer equations (1.38) to (1.41),
together with a "turbulence model" to close the equations, the following
initial and boundary conditions must be known (for steady flow); for
convenience we treat a boundary layer on the solid surface y = 0 .

Initial conditions (X=Xo where Xo denotes the starting position of the

investigation, see Fig. 2.2)

U(y),p(y), T(y) or 2 e(y).

According to the closure model chosen: Vg, ~'w, v-0,q2 etc. (see Chapt. 5).
For three-dimensional flow it is necessary to know a further function,
being a combination of 0, U, V, and W which is determined from a
compatibility condition (TIN(; [2.8]).

Boundary Conditions (y = 0 and y= 6) 3

y=O:
U = 0 (where U is the relative velocity between the wall and the fluid)
V= 0 (impermeable wall) or

2 e is the internal energyof the flow.

a t5 denotesthe free-streamboundary of the boundarylayer where the index used is e
(for external).
50 H.-H. FERNHOLZ

V<0 (suction) or V>0 (blowing)

IV=0 (for three-dimensional flow)
u=v=w=O (in the case of a permeable wall v is also assumed to be
zero, which is reasonable if the pressure drop across
the wall is large)
T= Tw (for flow with heat transfer)
y=6:
U= Uc
W=W~ and dpJeo+d(UZ~+W~Z)2=O
T=T~
~--~]~=b--~]e=v0[e=0 (for high free-stream turbulence these values
must be known).
For complex configurations, for example where there is strong inter-
action between the viscous and the inviscid flow (see e.g., GREEN [2.9])
or between two shear layers (Fig. 2.3 and Sect. 3.1), iterative matching
of boundary conditions in two regions, or use of the full elliptic equations
of motion, may become necessary. In other cases the boundary conditions
themselves may be given in indirect forms, perhaps by differential
equations as in two-phase flow problems where a vapor and a condensate
boundary layer interact.
It is this multiplicity of boundary conditions which creates such a
variety of thin shear layers with widely differing flow characteristics.
In the following subsections of this chapter we shall deal with some of
these cases by varying the boundary conditions, indicating their effects
on the boundary layer and giving a selection of references for a more
detailed study. Many of the remarks apply to other types of turbulent
wall flow such as wall jets or duct flows, while remarks about free-stream
boundaries apply to free shear layers also.

2.3 Two-Dimensional Boundary Layers at Small

Mach Numbers
2.3.1 The Multi-Layer Model of Turbulent Boundary Layers
As indicated in Chapter 1 (see also Chapt. 5), any solution of the tur-
bulence problem short of a complete solution of the time-dependent
Navier-Stokes equations relies on empirical information. It is therefore
desirable to have as detailed an account of the turbulence structure in the
boundary layer as possible. To put the observed phenomena into some
sort of order it has proven useful to split up the boundary layer into an
inner layer and an outer layer. The elementary dimensional analysis
for the former, which applies to all wall flows, was carried out in Section
 External Flows 51

Fig. 2.4a and b. Visualization of a boundary layer by means of smoke. Photograph a from
[2.282], photograph b by courtesy of P. BRAOSr~AW.(a) Streamwise view, (b) side view, flow
right to left

1.8. We now seek more details of the turbulent motion, and of the charac-
teristic velocity and length scales of the outer layer. Despite the
difficulties of the task, obvious from the smoke photographs of Fig. 2.4,
some progress can be made, though the outer-layer scales are much more
complicated than those of the inner layer.
In turbulent boundary layers one boundary is formed by the wall,
where the turbulent shear flow must accommodate itself to the wall
through a viscous sublayer and the large velocity gradients associated
52 H.-H. FERNHOLZ

with it. Very close to the wall the fluctuating velocities and therefore the
Reynolds shear stresses tend to zero, while the total shear stress remains
constant. Thus the viscous shear stress pOU/Oy, and the velocity gradient
OU/Sy, are much larger than elsewhere in the boundary layer. The other
boundary is the turbulent/non-turbulent interface where the shear
flow with high vorticity accommodates to an essentially irrotational
outer flow through the "viscous superlayer" (Sect. 1.9). This latter
irregular but very distinct boundary is common to all turbulent thin
shear layers with at least one free boundary, but its quantitative proper-
ties depend on the characteristic scales and the boundary conditions
of the specific shear layer. It has been found useful (see Fig. 1.6) to divide
the inner layer at the wall into the linear sublayer and the buffer layer
(together comprising the viscous sublayer) and the Jog-law region.
The outer layer which covers about 80 % of the boundary layer is bounded

Table 2.1. Multi-layer model of turbulent boundary layer: see text for symbols

Inner layer Outer layer

Viscous Log-law region Outer-law region Viscous

sublayer superlayer

Range within O< y< 4Ov/u~ 40v/u,< y<0.2~ 0.2 < y/fi Inhabits
boundary layer y>0.46
Characteristic (v/uOi kr (v/u), kr 6 or A (v3/%)~
length or v/Vo
Characteristic u, u, u,, Uo V~~ K(v~op
velocity

Typeof U=f(yu~ kru,I OU/Oy=(_~.~)/Ky y Ur

velocity law u, \ v ' v /
Flow Intermittent Fully turbulent Fully turbulent, but Laminar-
condition turbulent intermittent like
rotational/irrotational
for y > 0.46
Dominating Dissipation ~ production of Extraction of kinetic Transfer of
mechanism turbulent energy mean flow energy by vorticity to
the Reynolds stress the irrota-
gradient tional free
stream
Dependence Dependent on Independent of viscosity Dependent
on viscosity viscosity on viscosity
Dominating -ao<~aU/~y -~>t~Su/(~y
shear stress (for most of the
term time)
 External Flows 53

towards the free-stream by the viscous superlayer 4. Table 2.1 briefly

surveys this multi-layer model, specifying the dominating turbulence
mechanisms, the characteristic scales and the types of the mean velocity
distribution, mainly based on the monographs by TOWNSEND [1.22]
and ROTTA [1.5]. A similar survey could be made for other types of
turbulent shear flows, but the boundary layer is by far the best documen-
ted and most important.

The Inner Layer (0 < y/t~ < O.1-0.2 roughly; 0 < U/U~ < 0.7 very roughly)
Initially we consider two-dimensional constant-density flow over an
impermeable wall, as outlined in Section 1.8, in a region close enough
to the wall for the total shear stress to be nearly equal to the wall value
zw and for all derivatives with respect to x to be negligible. The behavior
of the mean velocity profile, and of the turbulence, depends on the
relative size of the terms on the right-hand side of

Zw=pdU/Oy_off~ (2.1)

Linear SubIayer (u~y/v < 3, roughly)

In the innermost part of the sublayer, ~--~can be neglected compared
with #dU/dy, and on integrating (2.1) one obtains

U/u,=u~y/v or U+=y + (2.2)

hence the name "linear sublayer". Here u~ is the friction velocity (Zw/Qw)~,
and we recall from Section 1.8 that u,y/v=y is a typical Reynolds
number of the energy-containing turbulence. Eq. (2.2) applies for y + < 3
approximately. Support for the neglect of the Reynolds stresses in the
viscous part of the inner layer, based on experiments, has been given
recently by ECKELMANN[2.17"]. Experimental evidence shows that the
pressure gradient has little effect on the sublayer and that (2.2) applies
for flows in ducts [2.18] and pipes [2.19] also.
If one measures the instantaneous velocity profile instead of the
time-mean profile the sublayer is revealed to have a longitudinally

4 This multi-layermodel for turbulent boundary layers has been developedover the
past fifty years. Referring to a few papers only one should mention especially:for the
sublayer PRANDTL[2.10], KLIN~et al. [2.111; for the logarithmiclaw PRAND'rL[2.10, 12],
LUDWIEOand TILLMANN[2.13]; for the outer layer VONKARt,IAN [2.14], CLAUSlm[2.15],
COL~s[2.16]; for the superlayerCORRSINand KISrLERi-1.30].
54 H.-H. FERNHOLZ

streaky, spatial and time-dependent flow structure (KLINE and Rtm-

s'rADLER [2.20]), continuously disturbed by small-scale velocity fluctua-
tions of low magnitude and periodically disturbed by fluid elements
which penetrate into this region from positions further removed from
the wall (COmNO and BRODK~ [2.21]). For further references the reader
is referred to papers by BAKEWELL[2.19], KUNE et al. [2.11], KIM et al.
[2.22], (}UPTA et al. [2.23], WALLACE et al. [2.24], OFFENand KLINE
[2.25] discussing the phenomenology and the triggering mechanisms
and to UEDA and HI~zE [2.26] and KtJo and CoRRsIN [2.285] for the
fine structure of the turbulence. The streaky flow structure seems to
have been observed earlier by HAMA as pointed out by CORRSlN[2.27].

Buffer Layer (3 < u,y/v < 40 roughly)

Since both viscous and Reynolds shear stresses in (2.1) are of the same
order of magnitude, no simple relationship for the mean velocity in the
buffer layer can be derived. According to the experiments of LAUFL~R
[2.28] and KLEBANOrr [2.29] the production and dissipation of turbulent
energy are larger, but not much larger, than the diffusion: this is not a
local-equilibrium region. The production of turbulence is apparently
strongly connected to the bursts observed in the sublayer (KrM et al.
[2.22]; the experiments were done at low bulk Reynolds numbers, but
this should not matter according to inner-layer similarity arguments).
For further details see also WILLMARTHand Lu [2.30] and BLACK-
WELDERand KAPL^N [2.31] who applied conditional sampling techni-
ques to investigate the link between the bursting phenomenon and the
production of shear stress.

Looarithmic Law Region (40v/u~ < y < 0.2~ roughly)

For large turbulent Reynolds numbers (u,y/v > 40) the viscous term in
(2.1) may be neglected compared with the Reynolds shear stress. Then
dimensional analysis as. in Chapter 1 or in RoTrA [2.32] yields the
"logarithmic law" or the "law of the wall" for the mean velocity distribu-
tion on an impermeable smooth wall with negligible shear stress gradient,

U /u, = (1/x) In (u~y/v) + C, (2.3)

where x=0.40 to 0.41 and C=5.0 to 5.2 are determined from experi-
ments s. This is one of the most important relationships for turbulent
boundary layers since, apart from the floating element, almost all
5 The possibilitythat x does in fact depend on the bulk Reynoldsnumberis discussed
in [1.15, 28].
 External Flows 55

measuring techniques for skin friction are based on its validity, as well
as most semi-empirical relationships for the wall shear stress (Sect. 5.7).
For further references see, for example, PRESTON [2.33], CLAUSER[2.5],
HEAD and RECHENBERG [2.34] and PATELand HEAD [2.35]. There is
good experimental evidence that the logarithmic law is valid also in wall
layers with a streamwise pressure gradient--though the assumptions
used for its derivation did not include pressure gradients--but that its
range of validity becomes considerably smaller in an adverse pressure
gradient compared with that in zero pressure gradient (cf. the Stanford
Data Catalogue by COLES and HIRST [2.7]). Deviations from the log-
arithmic law occur, however, in wall layers with strong favorable
pressure gradients (PATEL and HEAD [2.35]). A range of validity for the
application of Preston tubes [2.33] to measure skin friction was given
by PATEL [2.36].
Analyses of the effect of the pressure gradient on the velocity distribu-
tion near the wall were performed by TOWNSEND[1.26], MELLOR[2.37]
and MCDONALD [2.38] using (1.50) or refinements. For z-----w+ ~Y, inte-
grating (1.50) and requiring compatibility with (2.3) in the limit ~ 0
gives, with z = oty/-c w,

u ~ = rc v (l+z)+l +2((l+z)-1) ' (2.4)

where C may depend slightly on ~v/(Qu,3).

Corresponding to the result z = zw for the inner layer in flow over an
impermeable wall and negligible acceleration, one obtains, for an inner
layer with transpiration, (Vw= constant):

= Xw+ 0U Vw, (2.5)

where Vw is the normal velocity at the surface. Substituting (2.5) into

(1.50), integrating this equation and requiring compatibility with (2.3)
yields

2.,[([~1+ u2 ] ] .,y
1 = __lxIn --v + C (2.6)

where C is now a function of the dimensionless parameter V,,/u, and,

if the surface porosity is not of very small scale, of the properties of that
porosity too. For a more detailed discussion see for example COLIn
[2.39] and BAKERand LAUNDER [2.40]. Other integrals of (1.50), for
different shear-stress distributions, can be generated in the same way.
 56 H.-H. FERNHOLZ

Before turning to the outer region, attention should be drawn to

some flow visualization studies in the inner region. As can be seen for
instance from the data of KIM et al. [2.22] the flow structure close to the
wall is qualitatively very different at periods of oscillatory growth of the
"streaky" flow structure from the flow at "non-bursting" periods. The
former flow is strongly intermittent and displays features of an organized
oscillation or wave packet. This has led to investigations of the turbulence
structure where use was made of wave models, e.g., LANDAHL [2.41],
BARK [2.421, PHILLIPS [2.43], MOFFATT [2.44], LIGHTHILL[2.45] and
HUSSAINand REYNOLDS[2.46, 47]. So far no decision has been reached
whether a random array of deterministic eddies or a model based on
travelling waves is more appropriate to describe the structure of tur-
bulence (see also KOVASZNAY[2.48]).

The Outer Layer (0.2 < y/c5 < 1.0)

Casual observation of turbulent shear flows with a free-stream boundary
made visible by some contaminant (e.g., chimney plumes, clouds) shows
that the boundary between contaminated and uncontaminated fluid
is highly irregular, unsteady in any coordinate system and moving
outwards to entrain previously uncontaminated fluid. In the case of
irrotational outer flow, velocity fluctuations, with wavelengths of the
order of the largest turbulent eddies but containing negligible
vorticity, are found to extend into the irrotational flow region beyond
this boundary. Therefore it is advisable to divide the flow into a rotational
and an irrotational flow region (see Fig. 2.4). The demarcation line
between the two zones gives an intermittent character to the signal of a
fixed hot-wire probe (CoRRSnq [2.49]), and the fraction of time spent
by the probe in the rotational region is a characteristic feature of a
turbulent shear layer, called the intermittency factor V (TOWNSEND
[2.50]). V equals one near the wall and tends to zero as y goes to infinity.
Typical intermittency distributions, which seem to follow a Gaussian
(error function) cumulative probability distribution were measured in
various shear flows by CORRSINand KISTLER [1.30] and in boundary
layers with variable pressure gradient by FEDLER and HEAD [2.511 for
instance. Recent assessments of discrimination processes for the deter-
mination of the rotational/irrotational interface have been given by
KAPLAN and LAUFER [2.52], KOVASZNAY et at. [1.32], ANTONIA [2.53],
. VAN ATTA [2.541, LARUE [2.551, ANTONIAand ATKINSON [2.56] and
BRADSHAWand MtmLIS [2.571, the latter paper revealing some of the
difficulties one may encounter. The scales of the largest surface indenta-
tions are comparable with the total width of the shear layer, and the
average velocity difference between the rotational and irrotatfonal
 External Flows 57

flow is about 5% of the maximum velocity difference in wake flow

(TOWNSEND [1.22]) and about 5% of Ue in boundary layer flow
(KOVASZ)qAe et al. [1.32]), showing the rotational zone lagging slightly
behind. Velocity and shear stress profiles in the rotational and irrotational
zone were measured by KAPLANand LAUFER [2.52] and BLACKWELDER
and KOVASZNAV [2.58], respectively. KLEBANOFr [2.29] showed the
free-stream boundary to be at a mean position Y=0.786 with the standard
deviation a=0.146 in a zero-pressure gradient boundary layer. These
values vary with the state of the boundary layer [2.51]. Though the
spreading rates for distinct shear flows are well known (see, e.g.,
SCHLICHTING [Ref. 1.3, p. 678]) the entrainment process as such is not
yet fully understood (see also MOLLO-CrIPaSTENSE~q[2.286] and FAVRE
et al. [2.287]).

T h e 14scous Superlayer

The viscous superlayer, as described in Chapter 1, was postulated by

CORRSIN and KISTLER [1.30]. It is maintained thin by propagation
relative to the fluid with a propagation velocity Vo and by the random
stretching of vortex lines in its local velocity gradient. By dimensional
reasoning the thickness of the superlayer was estimated to be of the
same order as the Kolmogorov length rl=(v3/eo) ~, its propagation
velocity normal to itself being as proportional to the Kolmogorov
velocity v~=(Veo) ~' (see also PmLLIPS [2.59]) where e0 is the viscous
dissipation within the turbulent fluid. So the propagation of the inter-
face into the irrotational fluid is effected by viscous stresses but at a
rate set by the spatial transport of turbulent energy from below by the
large eddies. The mean velocity component normal to the mean position
of the interface, y = Y(x), is called the "entrainment velocity", a notation
which is used in practice for the velocity normal to the line y=6(x).
By calling this velocity VE, application of the principle of conservation
of mass to a control volume bounded by y=0, y = 6 ( x ) , x = x 0 and
x = x o + A x gives

d ~ o U d y = d [peUe(6_6.)]
#eVE= ~xxSO (2.7)

where the suffix e denotes conditions at the edge, y=fi(x), and where c5"
is the displacement thickness as defined by (2.22).
Note that VE is orders of magnitude more than the propagation
velocity of the interface because the latter is highly convoluted [1.33].
A semi-empirical relationship between the entrainment velocity VE
and the velocity profile has lead HEAD [2.60] to a simple closure for the
58 H.-H. FERNI-IOLZ

boundary layer equations (see also H ~ o and PATEL [2.61]); details

are given in Chapter 5. Theoretical approaches for a description of the
evolution of the superlayer were presented by TOWNSEND [2.62] and
PmLLmS [2.59], and further experiments on its structure were published
by KOVASZNAYet al. [1.32].

The Outer-Layer Eddy Structure

So far the entrainment process has been solely ascribed to the normal
propagation of the interface as a nonlinear diffusion of vorticity. This
is increased as compared with that of a laminar boundary layer by the
enlarged free surface of the turbulent boundary layer which is convoluted
by the large eddies in the outer layer. The large eddies contribute at
least 50% to the turbulent energy associated with the u- and v-fluctuations
and about 80% to the Reynolds stress (BLACKWELDERand KOVASZNAY
[2.58]; see also FALCO [2.63]). Whether the large eddies arise in the
outer layer because of some form of interfacial instability near the
superlayer (KovASZNAY et al. [1.32]), or result from the ejection of
bursts from the inner region (ANTONIA [2.531t or both, is still an open
question.
For a survey on the structure of turbulent shear flows the reader
is referred to KOVASZNAY[2.48], or WILLMARTH[2.447].

The Outer-Layer Velocity Profile

If the mean velocity is averaged over the rotational and irrotational
zones of the outer layer, then it is usual to express the velocity distribu-
tion in terms of the velocity defect U , - U (e.g., ROTTA [2.32]):

(Ue - U)/u, = f (y/A) (2.8)

where the function f depends on the pressure-gradient history and on

the bulk Reynolds number, although the direct effect of the latter is
small except in small-scale laboratory flows. Here A is a characteristic
length scale, introduced by ROTTA[2.64] and defined as

A j~ U e - U dY (2.9)

Equation (2.8) is a good approximation to the velocity distribution

(outside the viscous sublayer) in constant-pressure boundary layers,
where f is a universal function if u,6/v is greater than about 2000 (see
Sect. 1.9). The same holds for fully developed flow in parallel-wall
 External Flows 59

ducts. In this case ~ is replaced by the half-height of the duct and Uc

by the centerline velocity. A convenient way of rewriting (2.8) for
constant-pressure boundary layers (RorrA [2.65]) is

(U, - U)/u, = - (l/x) In (y/8) + (B/K)[2-- w(y/$)] (2.10)

where w(y/6), with w(1)= 2, is Coles' wake function, and the wake factor
B is an absolute constant for ufi/v > 2000.
Equation (2.10) applies only to constant-pressure flows (or, with a
change of B or w, to the self-preserving boundary layers discussed in the
next section). However COLIn [2.16] showed that the velocity profile
outside the viscous sublayer in almost any turbulent boundary layer
could be represented by (2.10) with w(y/6) taken as a universal function,
independent of pressure gradient or Reynolds number, and B (replaced
by H to avoid confusion) taken as a free parameter. It is important to
realize that this is merely a convenient data correlation, the effect of
pressure-gradient history being submerged in /I. Equation (2.10), on
the other hand, is based on the assumption that pressure-gradient
history is unimportant--as indeed it is in a constant-pressure flow.
COLES refers to the function w(y/8) as the "law of the wake", analogous
to the "law of the wall" function f~(u~y/v) in (1.47a), and writes, for the
velocity profile outside the viscous sublayer in any turbulent boundary
layer at low Mach number,

(O/Uc) = (uJU)[tc- 1 In (u,y/v) + C + (ll(x)/tc)w(y/6)] . (2.11)

The function ll(x) is determined by inserting the boundary conditions
at the outer edge of the boundary layer into (2.11), and the wake function
w is approximated by w(r/)=(1--cos(r~/)) according to COLES, or,
according to ROTTA [2.65] by

~q)=39~3--125q4+183qs--133q6+38~ 7 (2.12)
where in each case ~/= y/& For a different formulation of the wake law
giving dU/dy = 0 at y = ~ see Section 3.2, and for skin-friction laws derived
from the inner and outer laws see Chapter 5.
A family of velocity profiles covering the whole width of the boundary
layer has been given by THOMPSON [2.66], taking into account the
intermittency in the outer layer and the inner layer by two universal
functions. This approach is readily adaptable to account for the effects
of wall roughness, or of distributed suction or injection, by adapting
the velocity profile in the inner region. It may be regarded as a numerically
specified extension of the simple analytic data fit of COL~. The fact that
60 H.-H. FERNIIOLZ

these are merely data fits and not laws of nature is illustrated by the
quite large differences between Thompson's "intermittency function"
and the true behavior of the intermitteney. The main use of these profile
families is in checking experimental data and in "integral calculation
methods" (Chapt. 5).

The Turbulent Boundary Layer at Small Reynolds Numbers

Turbulent boundary layers in the range of 320<Reo<-5000 show a
behavior which deviates from that in higher Reynolds number boundary
layers with decreasing values of Ree. The following brief discussion is
therefore of some importance for incompressible boundary layers with
zero or negative pressure gradients and compressible boundary layers
at high Mach numbers.
If the ratio A U/u~ as given in Fig. 1.6, equal to 2H/x in present nota-
tion, is used as a measure for the strength of the wake component, it
can be shown from measurements (COLa [2.67]) that the wake compo-
nent disappears 6 at about a value of Reo ~ 500, which is not far from
that given as a lower limit for turbulent boundary layers (Reo= 320) by
PRESTON [2.68]. This leads to a breakdown of the outer law. With
decreasing Reynolds number both viscous sublayer and superlayer
thicken considerably, so that viscous effects begin to dominate over
effects due to turbulence, a process which finally can lead to relaminariza-
tion of the boundary layer (see Subsect. 2.3.11). The inner law on the
other hand seems to be very little, if at all, affected. Smoke photographs
in boundary layers at low Reynolds numbers indicate changes of the
shape of the interface and the structure of the large eddies (private
communication HEADandFALCO) but more detailed experimental
evidence must be awaited. It is this lack of reliable experiments which
makes it impossible at present to decide whether the constants in the
law of the wall (2.3) are dependent on the Reynolds number (SIMPSON
[2.69]) or not ( H ~ A N and BPO,DSHAW [2.70]). Further basic ex-
perimental evidence about turbulent boundary layers in the Reynolds
number range 800_~ Reo <6000 has just been reported by M u l l s [2.71-1
and appears to support Head and Falco's findings as well as those of
HUFFMAN and BRADSHAW.

2.3.2 Self-Preserving Shear Layers

In the case of a laminar boundary layer the equation of motion can be
reduced from a partial to an ordinary differential equation dependent

6 I.e., the velocitynever exceeds the logarithmic-lawvalue--not in itselfa critical

situation.
 External Flows 61

on q ~yg(x)/Re~ and a shape parameter m if the free-stream velocity

U~ is proportional to x" (FALKNERand SKAN [2.72]). For constant m
the mean velocity profile is similar at all distances downstream, say
from some leading edge. Since the various values of m cover a range
between highly accelerated and separated flows, the variation of the
velocity profile is described by these similar solutions which are exact
solutions of the boundary layer equations (see LOITSIANSKI [2.73]).
In order to obtain dynamic similarity in turbulent shear layers not
only the mean velocity profile, but also the profiles of Reynolds stresses
and other turbulence quantities, would have to be similar. Such a strict
similarity does not exist except in trivial cases. Some turbulent flows,
however, have regions where the action of viscosity on the mean flow
is negligible and where the motion of the energy containing components
of the turbulence is determined by the boundary conditions alone.
At sufficiently high turbulence Reynolds numbers the motions of
such flows differ, at given y/6 and outside viscous regions, only in their
scales of velocity and length and are then called self-preserving flows
(TowNsEnD [1.22]). Clearly, the condition for a flow to be self-preserving
is that the equations for the mean velocity and the turbulent energy
(see Chapt. 1) and the boundary conditions can be satisfied to an ade-
quate approximation by self-preserving distributions of mean velocity,
Reynolds stress, etc. (see, e.g., TOWNSEND [2.74]). For example, both
plane and circular jets and wakes are characteristic shear flows of this
type, if observed far enough downstream from their origin. It is in this
sense that the word "equilibrium" was used above.
The general form of the mean velocity and shear stress profiles in a
self-preserving flow is

(U,~r-- U)luo = F(y/lo) (2.13a)

--ff-ff/U2o = 9(y/lo) (2.13 b)

where u o and lo are the velocity and length scales of the flow, both
functions of x in general. U must be referred to Ur,f, usually the free-
stream velocity, because the self-preserving region excludes the viscous
sublayer and therefore the point where U=0. Outside the viscous sub-
layer, but still in the inner layer, there is a region where the local rates
of production and dissipation of turbulence energy are large compared
with the rates of advection and diffusion (1.36) so that the latter can
be neglected. A shear layer where these local rates of production and
dissipation balance each other is called a (local) equilibrium layer
(TowNSEND [1.26]). Local equilibrium can be thought of as a special
case of self preservation, with Uo= ( - u v ) and l o =y. However having
62 H.-H. FERNHOLZ

the scale and the coordinate the same makes this rather confusing
and we have therefore considered local equilibrium separately (Sect. 1.8).
Local equilibrium is the more important; it is believed that almost any
inner layer has a local-equilibrium region, whereas truly self-preserving
flows are rare.
TOWNSEND [1.26, 2.74, 76, 77] and Ro'rrA [1.5, 2.75], for example,
have investigated self-preserving flows. ROTTA determined the condi-
tions under which self-preserving flows can exist by substituting self-
similar distributions of mean velocity and Reynolds shear stress.

( U e - U)/u~ = F(tl, o2, fl) (2.14a)

- ~ / u ~ = a(n, ",', fl) (2.14b)

into the x-momentum equation (1.39) obtaining the following dif-

ferential equation:

fl(2F-oF2)+[(oo-ldA/dx)-fl](,-(-o S~, Fd,)F'=9' (2.15)

where A is defined by (2.9) and

fl = (A/zw)o~dp/dx- (6*/zw)dp/dx )
n =ylA (2.16)
= uJUe

Unfortunately, symbol H is often used for fl, as well as for the wake
parameter. Equation (2.15) is only independent of x if fl, o9 and A or
dA/dx are constant. The condition fl=constant is fulfilled by velocity
distributions of the type Ue~x" or Ue,-~expx and A---constant. ~o
equals constant can be satisfied only if the following relationship holds

tn = (1/~) In (U~6*/v) + C, + C(k*) (2.17)

where the variation of the displacement thickness 6* with x is balanced

by the variation of the characteristic roughness k * = (kru,/v), kr being
the height of a roughness element.
For a further discussion of the theoretical aspects of self-preserving
boundary layers the reader is referred to the above-mentioned investiga-
tions of ROTTAand TOWNSEND and a paper by MELLORand GmSON
[2.78]. The latter authors have supplemented Rotta's work by extending
the range of fl covering values -0.5 < f l < oo by using an eddy viscosity
relationship for the Reynolds shear stresses and keeping the parameters
 External Flows 63

fl and G constant. G is defined as

G = S~) [ ( U e - U)/u,]2d(y/6)/S~ [Ue- U)/u~]d(y/6). (2.18)

The "defect shape parameter" G is connected with fl and o9 via the

momentum integral equation (ROYrA [2.32]).
An approximate relation between G and fl based on the available,
rather scattered experimental data was given by NASH [2.79] (for self-
preserving flows only)

G-~6.10 (fl+ 1.81)~ - 1.70. (2.19)

This gives G = 6.5 at fl=0, while the actual value is about 6.8.
It is by no means easy to verify self-preserving boundary layers
experimentally. CLAUSER [2.15] kept the velocity profile in the outer
layer (2.8) independent of x by adjusting the adverse pressure gradient
distribution in the x-direction accordingly. In these experiments G
was nearly constant while fl varied, especially in the boundary layer
where the pressure gradient was kept high. Both HERRINGand NORBURY
[2.80] and LAUNDERand STINCHCOMBE[2.81] set up self-preserving
flows with favorable pressure gradients. But there are doubts about the
applicability of their method to determine skin friction in highly ac-
celerated flows. Therefore it is difficult to decide whether self-preserva-
tion was reached at all (cf. COLESand HmST [2.7]). BRADSHAWand
FERRISS [2.82] and BRAI)SI-tAW [2.83] have investigated two self-pre-
serving flows in adverse pressure gradients (Ue,,,x -'z55 and x -'is)
measuring mean velocity profiles and turbulence structure. They found
that Reynolds shear stress and fluctuating velocities show self-preserva-
tion in the outer layer on the average, without being able to specify
the degree of accuracy reached in their measurements of turbulence
quantities.
Similarity laws for turbulent boundary layers with suction or injec-
tion were given by COLES [2.39] and STEVENSON[2.84].

2.3.3 Upstream History, Relaxation Effects and Downstream Stability

A shortcoming of many of the older boundary layer calculation methods
consists in their inability to take into account the upstream history of
the boundary layer and the relaxation of these upstream effects down-
stream, since it is assumed that the boundary layer development, and
especially the shear stress distribution, is determined by local conditions
only, via an "eddy viscosity" relation (Sect. 1.7) or otherwise. Relaxation
is not used here in the strict sense of the word as it is defined for example
64 H.-H. FERNHOLZ

in thermodynamics, but is meant to describe the transition process

between two well-defined, possibly equilibrium states of a shear flow.
Upstream history and relaxation effects may influence the inner and the
outer region of a boundary layer more or less strongly or for a different
length of time. In Klebanoff and Diehl's experiment [2.851 where the
boundary layer had been disturbed by a spanwise rod, lying on the wall,
diameter d about equal to the undisturbed boundary layer thickness 60
at the trip position, a downstream distance of 26 60 was not sufficient
for the velocity profile to regain similarity. If a much smaller trip rod
with d/6 o = 0.055 (CLAUSERI-2.5]) is placed parallel to the wall and across
the boundary layer, but now at distances y/6 o =0.15 and 0.55 from the
wall, the inner layer returns much more quickly to the universal inner
law than the outer layer, confirming the strong influence of upstream
history and relaxation on the large eddies. One reason for this behavior
consists in the rather long lifetime of these eddies. If a typical eddy time
scale is defined as (typical wavelength)/(typical velocity fluctuation),
one finds a time scale of roughly 6/0.003 U~, during which time the
large eddies travel a distance of about 30 60. The response of a retarded
self-preserving turbulent boundary layer to the sudden removal of the
pressure gradient has been investigated by BRADSnAWand FERRlSS
[2.82] who simulated the pressure distribution near the leading edge
of a "peaky" aerofoil. It turns out from measurements of the production
and advection terms of (1.43) that the advection is an unusually large
fraction of the production and that the decrease of the advection after
the release of the pressure gradient is small, indicating a slow approach
to the new self-preserving state in zero pressure gradient. Not only is
the advection term large compared with that in a self-preserving bound-
ary layer, but also the turbulent energy integrated across the layer is
much higher than in the zero-pressure gradient case. This must be
attributed to the large-scale turbulence in the outer region of the bound-
ary layer, i.e., to the large eddies. A further flow dominated by its up-
stream history is a turbulent boundary layer responding to a step
change in surface roughness from rough to smooth (JACOBS [2.86] and
ANTONIA and LUXTON [2.87]). The more intense initial rough-wall
flow dictates the rate of diffusion of the disturbance for a considerable
distance, and the turbulent energy budget shows that the advection
term is comparable with the production or dissipation terms.
Relaxation phenomena in boundary layers due to the upstream
heat transfer history (GRAN et al. [2.88], GATES [2.89]) or to severe
pressure gradients in the vicinity of the throat of supersonic nozzles
(LEE et al. I-2.90]) have been observed to last far downstream into the
zero-pressure gradient region. In the same type of flow the laminar-
turbulent transition process seems to take much longer in the near
 External Flows 65

wall region than in the outer part of the boundary layer, evidently for
similar reasons. As far as downstream stability is concerned CLAUSER
[2.15] reported difficulties in establishing self-preserving boundary
layers in strong adverse pressure gradients where a small change in
pressure gradient upstream produced large changes in the flow down-
stream. Problems of this sort are often linked to secondary flows or to
the unsteadiness of a separation region, but Clauser's results cannot
be ruled out completely. A plausible argument for downstream stability
was put forward by BRAOSHAW [2.83] and a more extensive discussion
of the stability of self-preserving flow was given by TOWNSEND [1.26].

2.3.4 The Effect of Free-Stream Turbulence

From the aeronautical viewpoint a discussion of the effects of freestream
turbulence on a turbulent boundary layer is of little importance. For a
large subsonic aircraft at cruise the rms velocity fluctuation over an
octave bandwidth centered on a wavelength of 0.3 m (the order of t$ over
the rear upper surface of the wing) will be only 0.03 % of the free-stream
velocity in heavy turbulence (GREEN [2.91]). However in wind tunnel
tests of aircraft and other bodies, and in many internal-flow rigs, the
influence of free-stream turbulence on the boundary layer must be known
in order to interpret or improve the results. The rate of growth of the
boundary layer thickness increases by about 50% and that of the skin
friction by about 20 % (CHARNAY et al. [2.92]) for a change of the tur-
bulence level T~= (u2)~/Ue from 0.2 to 5 %. As expected from inner-layer
similarity arguments the log-law is hardly affected. It is difficult, however,
to reconcile the increase in skin friction with Kestin's [2.93] statement
that a measurable effect of the free-stream turbulence on heat transfer
could not be found. As is to be expected, the effects of free-stream tur-
bulence on the outer part of the boundary layer (here y/6 > 0.7) lead to
an increase in the level of u 2, ~ (by about 150%) and of the turbulent
kinetic energy, to a sharp decrease of the wake component A U/u~ for
T~~ 5 % and to a slightly fuller mean velocity profile (for the use of the
latter effect to increase the "effective" Reynolds number in windtunnels
see GREEN [2.91]). HUFFMAN et al. [2.94] state further that at the high
turbulence level a highly non-equilibrium boundary layer exists because
the advection term is no longer small and therefore production and
dissipation are no longer roughly equal. Apart from the measurements
by KLINE et al. [2.95], CHARNAYet al. [2.92] and HUFFMAN et al. [2.94]
from which the above conclusions were drawn, there are measurements
by PICHAL [2.96], KEARNEY et al. [2.97] and ROBERTSONand HOLT
[2.278]. Further references together with a discussion of semi-empirical
relationships for the outer law, the skin friction, the shape parameters,
66 "H.-H. FERNHOLZ

etc., under the effect of free-stream turbulence were given by BRADSnAW

[2.281].
In the above work, the free-stream turbulence was generated by an
upstream grid and was not too far from spatial homogeneity and iso-
tropy, having a scale (wavelength) comparable with the thickness of the
shear layer. The effects of much larger or much smaller scales are not
clear yet. This is especially true for the boundary layer in a turbomachine
where the so-called free-stream turbulence is a mixture of roughly
homogeneous turbulence from far upstream, highly inhomogeneous
turbulence from the wakes of the blades and miscellaneous large-scale
unsteadiness, presenting a good example of an ill-posed boundary
condition. The influence of free-stream turbulence on the structure of a
wake and on the "potential" core region of a jet has been investigated
by KOMODA [2.98] and VAor [2.99], respectively. A thorough survey
and experiments on the management of free-stream turbulence have been
carried out by LO~-mKE and N~,Gm [2.100].

2.3.5 The Effect of Streamwise Pressure Gradient

The pressure gradient, adverse (positive) or favorable, is the parameter
which has the most far-reaching consequences for a boundary layer,
possibly leading to separation in the one case and to supersonic flow
and/or reverse transition in the other. More detailed reviews of the effect
of the pressure gradient on incompressible turbulent boundary layers
were given by CLAUSER[2.5], ROTTA [2.32] and TowNsn~) [1.22], and
a data catalogue containing mean velocity profiles, skin friction data
and pressure distributions of turbulent boundary layers was published
by COL~S and HIRSX [2.7]. A critical compilation of measurements of
turbulence data, updated at the same time, is very much needed as an
addition to the Stanford catalogue [2.7]. The interaction between
pressure gradient, mean velocity distribution and skin friction can be
seen from the integral form of the momentum equation for two-dimen-
sional thin shear layers which is obtained by integrating (1.45) from y = 0
to y=~5:

d i QU(U~- U)dy- -~x 1 - -~ dy-zw=O, (2.2o)

dx o

or, in more common forms, valid also for compressible boundary layers,

d
(Qov~o) = ~w(1+P) (2.21a)
dr
 External Flows 67

or

dO 0 dUe(H+2_ 2 2 _ Cf (2.21b)
dx + U dx M~)=(Zw/&U,)= --~

where

6" or dil= ~ [1-(QU/qeUc)]dy (displacement thickness)

o (2.22)
t~

0 or 62= S (QU/QcUe)[1-(U/Ue)]dY
o
(momentum loss thickness) (2.23)
H or H12=61/62 or 6*/0 (shape parameter) (2.24)

and fl was defined in (2.16). Eq. (2.21a) shows that fl measures the relative
effects of pressure gradient and surface shear stress on the rate of loss
of momentum. Since the mean pressure gradient does not appear in the
transport equations for the fluctuating quantities (see Chapt. 1), the
changes in turbulent intensity, shear stress, etc., which are characteristic
of pressure gradient flow are caused by changes of the mean velocity
profile. This again is influenced by the pressure gradient and implicitly
by the skin friction. It is this rather complex relationship between mean
velocity and turbulence quantities which causes the difficulties inherent
in the closure problem of the boundary layer equations. The structure
of the turbulent boundary layer with longitudinal pressure gradients
was investigated by SCnRAtra and KLINE [2.279].
We have seen in Subsection 2.3.1 that the logarithmic law is valid
also in boundary layers with adverse pressure gradients (Ltroww.G and
TmLMANN [2.13] if the flow is not too close to separation, where its
range of validity vanishes. It does not hold, however, for highly ac-
celerated boundary layers (PAnt, [2.36]). The outer law (2.8) is strongly
influenced by the pressure gradient, exhibiting under certain conditions
similar (self-preserving) velocity profiles with uJU and fl as parameters
(see RoTrA or CLAUSER). Typical Reynolds stress ~ , kinetic energy
and mean velocity profiles are shown in Fig. 2.5. At the wall the gradient
of the shear stress profile is determined by the following relation

(&/~Y)w= dp/dx (2.25)

indicating a constant stress layer in zero pressure gradient flow and a

maximum of the shear stress in the boundary away from the wall if an
adverse pressure gradient is present. Unfortunately there still exist few
68 H.-H. FERNHOLZ

~u~
u
Ue 0,8 .... -~ ~-- - .... 002

0 0,2
T0,4- [ ...... ~
0.6 y/5 1,0 0
.... 0,4" -
0,2 O.B ~ y/6 1.0

0.012 t 1
.... ~=o
d
0.008
~>0
Ue dx
O.OOl,

Ue
0.2 0,4 0.6 b y/(5 1,0

Fig. 2.5. Reynoldsstress, kinetic energyand mean velocityprofilesin a turbulent boundary

layer with zero and mediumadverse pressure gradient (from I-2.29,83])

experimental investigations in which even basic turbulence quantities

have been measured so thoroughly that all terms in the transport equa-
tions can be determined for a wide enough range of pressure gradients,
thus allowing assessment of the existing turbulence models. For adverse
pressure gradient boundary layers see for example B~DSHAW [2.101]
(equilibrium flow), BRAOSHAWand FERPdSS [2.82] (relaxing to zero
pressure gradient), GOLDBERG [2.102], SAMUELand JOUBERT [2.103],
SPANGENBERG et al. [2.104], SANDBORNand L]u [2.105], SCHUBAUER
and KLEBANOFF [2.106], NEWMAN [2.107] and for favorable pressure
gradients LAUNDER[2.t08].

2.3.6 Separation, Separation Bubbles and Reattachment Flow

Separation
Separation was probably the phenomenon which inspired Prandtl's
thinking about those shear flows which are now called boundary layers.
There exists a vast body of references on the topic, covering almost
every conceivable kind of flow geometry (see, e.g., CHANG [2.109]).
Following ROSHKO [2.110] we are concerned here neither with separa-
tion at low Reynolds numbers nor with the difficult mathematical
questions relating to the singularity at the laminar separation point
 External Flows 69

(see, e.g., KAPLUN [2.111] and BROWNand STEWARTSON [2.112]).

Separation may occur at sharp edges, for instance at the shoulder of a
wedge, or in boundary layers on smoothly curved surfaces in adverse
streamwise pressure gradient. In the first case there is no question as to
the point or line of separation, in many other cases the determination
of the separation point is part of the problem.
In some flows the wall-streamline leaves the wall at the point where
the gradient of the mean velocity at the wall becomes zero, as in Prandtl's
[2.113] classical case, followed by a region of reverse flow (Fig. 2.1).
The angle of this separating streamline was determined by OSWATITSCH
[2.114]. ROSHKOthen names secondly the flow with a separation bubble
and thirdly the flow with breakaway separation (for the latter see also
STEWARTSON [2.115], [2.116]).
All these different types of separation have in common the interaction
between the boundary layer and the flow in the free stream. Even in
incompressible flow the pressure distribution upstream of separation
is influenced by the flow downstream of separation, thus converting
the pressure from an independent boundary condition to a variable.
This means that the pressure calculated from inviscid flow theory and
the measured pressure are no longer consistent. Furthermore the as-
sumptions made to derive the thin shear-layer equations, such as
d6/dx,,,O(~) where e,~ 1 and Op/t~y=O(e), are no longer valid due to the
sharp increase of the boundary layer thickness and the curvature of the
streamlines close to separation. Streamline curvature will again have a
strong influence on the turbulence structure in this region (see BRADSHAW
[2.117], and Sect. 3.1).
Most of the criteria for separation are based on determining the
position of zero skin friction (see also Chapt. 5), e.g., the "rapid separa-
tion criterion" of Stratford as modified by TOWNSEND [2.118], who
showed that the pressure distribution near separation depends only on
the pressure rise towards separation and on the characteristics of the
initial boundary layer and not on the flow geometry. Further investiga-
tions were performed by SANDBORNandLIU [2.105] based on the
separation model of KLrNE [2.119] and SAra)BORN and KLINE[2.120].
An important phenomenon related to separation is that the boundary
layer close to separation may show a steady, an unsteady or a transient
behavior, superimposed on the turbulent motion proper, exhibiting a
difficult feedback mechanism. For similar effects of unsteady flow
behavior in boundary layers on wings with transonic flow regions
where buffeting occurs due to shock-boundary layer interaction see,
e.g., MABEY [2.121]. Boundary layers which showed a steady behavior
up to separation were investigated by STRAaWORD [2.122], FERNHOLZ
and GmSON [2.123], GOLDBERG[2.102] and SPANGENBERGet al. [2.104]
70 H.-H. FERNHOLZ

among others. In the former two investigations the pressure distributions

were set in such a way that the boundary layer remained on the verge
of separation, i.e., the flow had nominally zero skin friction. Theoretical
discussions about the boundary conditions (pressure distribution) for,
and the development of turbulent boundary layers with negligible wall
shear stress were given by STRATFORD [2.122], TOWNSEND [2.124],
FERNHOLZ [2.125] and ALBER [2.126] for example. Hardly any informa-
tion is available, however, about the structure of the turbulence in the
two extreme cases when a boundary layer reaches separation gradually
or is caused to separate rapidly by an obstacle. In the latter case the
boundary layer is subject to a strong perturbation (e.g., BRADSHAW
and GALEA [2.127]), thus behaving in a distinctly different way from the
former case.
Since separation is a phenomenon which normally has detrimental
effects on the overall flow, aerodynamicists have developed many
ingenious techniques to influence separation, the most important one,
of course, being to design the streamwise pressure distribution in such
a way that separation is prevented as far downstream as possible. A
survey of the techniques to prevent separation by blowing, suction,
moving walls or propeller slipstreams was given by G~a~STEN [2.128],
updating the monograph edited by LACHMANN[2.129]. For discussions
of large-scale separated regions and reattachment in internal flows see
Sect. 3.4.

Separation Bubbles (Fig. 2.1c)

After the discovery of the phenomenon of separation bubbles by JONES
[2.130], fresh interest in this topic was aroused through the use of thin
aerofoil sections for reducing the effects of compressibility. The behavior
of separation bubbles is therefore of particular interest both for the
stalling characteristics of these aerofoils, showing high suction peaks
near the leading edge followed by a sharp pressure rise, and also for the
heat transfer on turbine blades or for flow in poorly designed nozzles at
low or medium Reynolds numbers. Separation bubbles are also observed
in the flow around circular cylinders and in connection with the inter-
action of shock waves and laminar boundary layers. There is a whole
series of earlier experimental investigations summarized by YOUNO
and HORTON [2.131] and followed by those of HORTON [2.132, 133] and
DELPAK [2.134]. A survey paper on the subject which is still an excellent
introduction was given by TANI [2.135].
At first sight the problem of separation bubbles seems to be one of
laminar flow only, for it is the laminar boundary layer which separates:
but, under certain conditions not yet clarified, it reattaches in the tur-
 External Flows 71

bulent state, having undergone transition somewhere between separa-

tion and reattachment. The shallow recirculation region shown in
Fig. 2.1c is called a separation bubble. The necessary conditions for the
formation of a separation bubble are that the pressure gradient shall
decrease soon after separation (so that the shear layer does not move
rapidly away from the surface) and, except in the rare cases of laminar
reattachment, that the Reynolds number and free-stream turbulence
level shall be high enough for transition to occur soon after separation
but not so high that transition occurs before the expected laminar
separation point. At the higher Reynolds numbers the extent of the
bubble is small, of the order of 1% chord, and the slight step in the
pressure distribution produced by the bubble has a negligible effect
on the flow around an aerofoil. It should be noted that the upper limit
of the Reynolds number at separation is about 350, where 0s is the
characteristic length. With an increase of incidence or a reduction of
speed (Reynolds number) the short bubble may burst to form either a
long bubble or a separated shear layer (GASHER [2.136]) which then
leads to a severe reduction of the lift (Fig. 2.1c). No conclusive criterion
is known so far, either for the formation or for the bursting of separation
bubbles. If the separation region is small, conventional iteration pro-
cedures should suffice to calculate the effect of the bubble on the flow
around an aerofoil. Promising steps to solve the separation bubble
problem via the Navier-Stokes equations for laminar flows have been
taken by BRILEYand MCDONALD [2.137], LUGT and HAUSSLING[2.138]
and MEHTAand LAVAN [2.139]. Some experiments on the turbulence
structure of the flow were performed by McGREGOR [2.140] and GASTm~
[2.136] without conclusive results, and separation bubbles in three-
dimensional boundary layers were investigated by EICHELBRENNER
[2.141, 142] and HORTON [2.133]. The latter author also presented
measurements of three-dimensional turbulent boundary layers re-
developing after reattachment behind short separation bubbles.

Reattaehment of Turbulent Shear Flow

The phenomenon of reattachment of a separated shear layer, which we
have just met downstream of a separation bubble, has been of special
interest in supersonic flows, mainly because of the need to predict base
drag and heat transfer for flight vehicles. Let it suffice here to mention
the names of the authors who presented the basic theoretical approaches:
CHAPMAN [2.143], KORST [2.144] and CRoccoand LEES [2.145]. An
eclectic merger of the Crocco-Lees and Chapman-Korst approach to
the near wake and reattachment region has been proposed recently by
SMITH and LAMB [2.146] including about 70 references. Reattachment
72 H.-H. FERNHOLZ

is also discussed in I-2.9]. Though there are numerous applications in

low speed flow such as spoilers, forward- and backward-facing steps,
windbreaks, etc., the problem has attracted less scholarly attention
than in supersonic flow (see, e.g., the survey paper by B~DSHXW and
WONG [2.147]).
Shear layers with reattachment are in most cases highly perturbed
flows, the more so if they have two regions of separated flows as in Fig. 2.3
or if in the case of a backward-facing step the thickness of the oncoming
boundary layer 30 is much smaller than the step height (h>>30). In the
first case the upstream history of the flow is less important due to the
severe perturbation by the first separation region, but it is surprising
to note that upstream conditions seem to play a minor role also in the
latter case (e.g., ABBOTTand KLIr~ [2.148]). These authors have applied
flow visualization techniques to the flow in the separated region down-
stream from a step, finding a rather complex vortex pattern and a two-
dimensional recirculation region. BRADSHAWand WONG [2.147] have
noticed large differences between the turbulence structure in the re-
attachment region and that in a conventional boundary layer; both a
marked "dip" below the log-law for the near wall region and persisting
characteristics of the mixing layer in the outer part of the boundary
layer far downstream.
There are further experimental investigations by ARm and Rouse
[2.149], MUELLERand ROBERTSON[2.150] and LE BALLEURand MIRANDE
[2.151] (with variable pressure gradients). MUELLERand ROBERTSONalso
observed a slow redevelopment from the mixing layer to the boundary
layer flow. Again there is a need for further measurements in shear
layers with reattachment, with more refined measurement techniques.
Calculation methods for reattached shear flows were developed by
MUELLERet al. [2.152] and LE BALLEURand MIRANDE[2.151].

2.3.7 The Effect of Changing Wall Geometry

In this subsection we shall discuss the effects of two kinds of wall boundary
conditions on the turbulent boundary layer, the effects of roughness
elements and those of waviness of the wall. Wall curvature will be dealt
with in Subsection 2.3.8.

Wall Roughness
The effect of wall roughness on a turbulent shear layer was first investi-
gated in pipe flow by NmtmADSE [2.153] in 1933. Since then a wide
variety of experiments has been performed in shear flows, ranging between
the extremes of flows in natural watercourses and over aircraft wings or
 External Flows 73

compressor blades. Due to the multiplicity of geometrical shapes and

distribution patterns of roughness elements not even a semi-empirical
correlation of the experimental data exists. There are, however, relation-
ships which cover certain types of roughness, such as those of PERRYet al.
[2.160; see 2.154] and SIMPSON[2.155]. Naturally it is the boundary layer
over the so-called sand-grain roughness covering a surface uniformly
which was investigated most thoroughly (PRANDTL[2.156], SCHUCHTmG
[2.157-1). The former showed that results obtained in pipe and duct flows
also hold for boundary layers (see, e.g., HAMA [2.158]). If the roughness
length scale for this type of roughness kr is made dimensionless by v/u,,
one can distinguish between three cases (for nominally zero pressure
gradient flow: if Or/t3y~O, a parameter (k~/zw)O'r/Oy must appear but its
effects are unknown):
1) Hydraulically smooth ( O < k r u J v < 5 ) . The roughness elements
are submerged in the viscous sublayer and do not affect the boundary
layer significantly.
2) Transient roughness (5 < k~uJv< 70). The velocity profile is af-
fected by the roughness of the wall, and this effect is taken into account
by a function f which is substituted into the law of the wall instead of the
constant C in (2.3).

f ( kru,/v) = C - d Uk/u ~ . (2.26)

The roughness function A Uk/u ~ can be expressed according to HAMA

[2.158] for natural roughnesses such as wrought or cast iron by

A Ut,/u~ = (l/x) In [(kruJV) + 3.30] - 2.92. (2.27)

The logarithmic law then reads:

( U /u~) = (1/~c) In (u~y /v) + f ( k,u,Jv) . (2.28)

An analytic approximation for the function f was given by IOSELEVICH

and PILIPENKO [2.159] and a skin friction law was given by CLAUSER
[2.15]. For flows other than with sand-grain roughness (2.28) can be used
also if an "equivalent sand-grain roughness" is determined. PERRY et al.
[2.160] have found that for a roughness geometry consisting of rectangular
elements of square cross section placed normal to the flow direction
d Uk/u ~ is of the form

A Uk/u ~= (l/x) ln(eu,/v) + C + , (2.29)

74 H.-H. FERNHOLZ

where ~ is called the error in origin and is the distance below the crest
of the roughness which locates the origin of y in (2.28). C + is a constant
for the specific roughness geometry.
3) Fully developed roughness (70 < kruJV). The velocity distribution
in the log-law region is now independent of viscosity because the Reynolds
number characteristic of the flow over the rough wall is large. It takes the
following form, of which (2.27) yields a special case for large k,u~/v,

U/u,=(1/x) In (y/Zo). (2.30)

This is a relationship often adopted for velocity profiles in atmospheric

boundary layers (see Chapt. 4) where Zo is chosen to absorb the constant
C in (2.3) altogether.
It has become usual to distinguish between two types of roughness
according to e in (2.29), or z 0 in (2.30): e proportional to kr, defining
"k-type" or sand-grain roughness and e proportional to (5, or a function
of the distance from the leading edge x, defining a "d-type" roughness.
The latter type is generated for example by the above-mentioned rec-
tangular elements, now spaced at distances such that quasi-stable vortices
are formed in the cavity in between (e.g., Lit) et al. [2.161], PF_a~Y et al.
[2.160] and WOOD and ANTONIA [2.154]). For a d-type roughness little
direct interaction seems to occur between the roughness and the mean
flow in the boundary layer (see also TOWNF_Sand SABERSKY[2.162]), SO
that there are features similar to those of a smooth-wall boundary layer.
A calculation method for the flow over a d-type roughness was published
by ANTONIAand WOOD [2.163]. Further investigations dealing with the
structural features of turbulent flow over rough boundaries were per-
formed by GRASS [2.164] and by ANa'ONIXand LtrXTON [2.280].
Since the pressure gradient does not influence the flow in the inner
layer directly, little effect on the rough wall boundary layer is to be
expected (e.g., PERRYand JOUBER'r [2.165]). It is important, however, to
distinguish secondary effects due to size or spacing of roughness elements,
such as displacement effects (kr>>6o)leading to a change in the overall
pressure distribution or to vortex shedding from the crest of the first
roughness element. In the experiments of ANTONIAand LUXTON[2.166]
the crests of the roughness elements were aligned with the smooth surface
upstream to avoid these detrimental effects. Of course upstanding rough-
ness elements, such as belts of trees, may be important in practice. The
same authors [2.167] then found an internal layer growing over the rough
surface like 6. ~ x 8, i.e., similar to the growth rate of a smooth boundary
layer which is primarily controlled by diffusion. This internal layer
denotes the outward limit of the influence of the roughness and has two
important consequences: the influence of the "old" upstream boundary
 External Flows 75

layer is confined to the outer part of the composite boundary layer, and,
due to the intermittent behavior of the interface between the two layers
of the composite boundary layer, the flow switches between regions of
the "new" and the "old" turbulence. A self-preserving state is reached
about 15 6 o downstream from the first roughness element (LUXTON
[2.168]). The fast growth together with too crude assumptions about
velocity and shear stress distributions may account for the unsatisfactory
theoretical results (e.g., TOWNSEND [2.169, 170] and TAYLOR [2.171]).
RAO et al. [2.172] obtained better results from a higher-order closure
method.
In a fully developed rough-wall boundary layer the turbulent energy
production is significantly higher than in a smooth-wall boundary layer
and the shear stress has a positive gradient at the wall (ANrO~A and
LUXTON [2.173]). The reason for the nonzero shear stress gradient in the
absence of acceleration is not wholly clear. The only plausible explanation
is that U and Vvary cyclically with x as the flow passes over the roughness
elements, leading to an apparent "Reynolds stress".
The response of a turbulent boundary layer to a step change in
surface roughness from rough to smooth was investigated experimentally
by ANTONIAand LUXTON [2.87]. Measurements of the drag of some
characteristic aircraft excrescences immersed in turbulent boundary
layers were performed by GAtJDETand WINTER [2.174]. For the discus-
sion of three-dimensional and of nonuniformly distributed roughness
the reader is referred to SCHUCHaaNG [2.157].

Wavy Walls
There are three main areas where one is interested in the flow over wavy
walls: aeronautical, civil, and chemical engineering. Wave generation by
wind and the interaction of wind and sea are outside our scope here. In
aerodynamics even very small ratios of amplitude to wavelength of a
wavy wall, say 0.005, are of interest, since a surface waviness of such a
magnitude can provide a significant contribution to the total drag of a
transonic aircraft (ROGERS[2.175]). INGER and WILLIAMS[2.176] showed
that large changes in the phase of the pressure and temperature per-
turbation occur across the highly nonuniform flow of a turbulent
boundary layer if the boundary layer thickness is comparable to the
surface wavelength. As a result, a wall pressure distribution typical of
subsonic flow can exist in the presence of supersonic external inviscid
flow (see also KENDALL[2.283] and DAVIS[2.284]). The reader is referred
to ASHTON [2.177] and OWEN and THOMSON [2.178] for heat transfer
problems along wavy walls. Civil engineers are interested in the formation
of bed forms in alluvial channels. For flow over rigid wavy boundaries
76 H.-H. FERNHOLZ

the objective has been to evaluate the interactions between the boundary
layer and the external flow as reflected in pressure distribution and flow
resistance (e.g., Ho and GELHAR [2.179]). Turbulent flow in wavy pipes
was investigated by Hso and KEr,rN~v [2.180] who measured both
mean flow and fluctuating quantities. Besides boundary layers along
wavy walls as such, there exists considerable interest for gas-liquid
boundary layer flow where the phase changing interface shows a wavy
form (e.g., KOTAKE [2.181]). In this area little research on turbulent
flow has become known.

2.3.8 The Effect of Wall Curvature

Dealing with the effect of wall curvature on turbulent shear layers one
must distinguish between curvature in the transverse (cross-sectional)
and in the longitudinal (camber) direction (the flow along a body of
revolution as investigated by WINTER et at. [2.182] contains both effects).
If for a body of constant radius the boundary layer thickness is small
compared with the body radius, then the flow can be treated as if it were
two dimensional. This assumption does not hold for long slender bodies
of revolution, and the boundary layer equations and turbulence models
must contain terms which take into account the transverse curvature
(e.g., CEBECIand SMITH [ 1.4]). Effects of transverse curvature on turbulent
boundary layers were considered by RICHMOND [2.183] and GINEVSKn
and SOLODKIN [2.184] who derived relationships for the mean velocity
and the Reynolds shear stress profile. For the flow in the vicinity of the
wall they obtained for the shear stress distribution

z/z w = [ 1 + (dp/dx)y/Zw]/[1 _ y/rw], (2.31)

where rw is the radius of the transverse curvature and where the positive
sign refers to a convex and the negative sign to a concave surface. The
reader is referred to discussions of the inner law by WILLMARTHand YANG
[2.185], BRADSHAW and PATEL [2.186] and of the eddy-viscosity distri-
bution by C~BECIand SMm-I [1.4]. The boundary layer on air-drawn
artificial fibers was investigated by WALZ and MAY~R [2.187] and that
near the tail of a body of revolution by PATEL et al. [2.188].
Effects on the turbulent boundary layer due to longitudinal curvature
are much more difficult to cope with since the structure of the turbulence
is strongly affected, in apparent contrast to the case of transverse
curvature. Details are given in Subsection 3.1.1; here we give a few basic
references to work on boundary layers. PRANOTL [2.189] explained why
the flow of a turbulent boundary layer along a convex surface is more
stable than that along a concave wall, a result consistent with the in-
 External Flows 77

stability mode in laminar boundary layers on concave walls represented

by Taylor-G6rtler vortices (TAYLOR [2.190] and GORTLER[2.191]). Such
longitudinal vortices in turbulent boundary layers were first found by
TANI [2.192]. Finally BRADSHAW [2.193] made quantitative use of the
analogy between streamline curvature and buoyancy in turbulent shear
flow.
On a convex wall both Reynolds shear stress and turbulent kinetic
energy across the boundary layer are strongly reduced as compared with
a boundary layer on a plane wall, whereas the inverse effect can be
observed in the flow along a concave wall ([So and MELLOR[2.194, 195]).
Tt-IOMPSON [2.196] provided evidence that even at ratios of b/R=0.003 7
changes in the entrainment rate could be attributed to curvature effects.
For b/R=0.01 MERONEY [2.197] noticed a steep decrease in Reynolds
shear stress outside the near wall region compared with a flat plate
boundary layer and a change of the integral properties by about 10%.
Ratios of fi/R=0.01 are often found with thick aerofoils or with turbo-
machine blades. In compressible turbulent boundary layers on curved
walls with zero longitudinal pressure gradient, changes in heat transfer
of _+20 % were measured for a concave and a convex wall, respectively
(THOMANN [2.198]).
In most of the cases discussed above the wall curvature was so small
that changes in static pressure across the boundary layer were still
negligibly small. In experiments in curved ducts (So and MELLOR, and
PATEL [2.1993) the normal pressure gradient due to the strong curvature
must be taken into account. The relevant equations and the importance
of these terms with regard to curvature effects were discussed by
BRADSHAW [2.117] who gives a large number of references for both
boundary layers and wall jets. The reader is referred to NEWMAN[2.200]
for a review on curved turbulent wall jets and to FF~NHOLZ[2.201, 202]
for the influence of upstream history, boundary conditions and effects
due to very high curvature.
The effect of longitudinal curvature on eddy viscosity was correlated
by CEBECIand SMITH [1.4]. Calculation methods which account for
curvature effects on turbulent boundary layers were presented by
DVORAK [2.203] and IRWIN and SMrrI-i [2.204].

2.3.9 The Effect of Heat Transfer at the Wall

Since Chapter 6 will deal extensively with heat transfer problems in

turbulent flows the discussion of "shear flow--heat transfer" interaction
will be confined to a few items in preparation for the discussion of

7 6 is the thickness of the boundary layer and R the radius ofthe longitudinal curvature.
78 H.-H. FERNHOLZ

compressible flow in Section 2.5. We shall refer to turbulent boundary

layers at low Mach numbers and at a Prandtl number of about one only,
first with small temperature differences and then with temperature
differences large enough to affect the velocity field. One of the technically
very important relationships is that between skin friction and heat
transfer at a fixed wall, the so-called Reynolds analogy which allows heat
transfer to be determined by a measurement of the skin friction

S t - Q,,/Eee Uccp( T, - Tw)]= F . cf/2 . (2.32)

The analogy in its simplest form applies only to boundary layers with
zero pressure gradient along a smooth isothermal wall. S t denotes the
Stanton number, cf the skin friction coefficient and F the Reynolds
analogy factor (see RoTTA [2.205]). Simple transfer problems could be
solved by boundary layer methods but most of the technical flow
configurations with heat transfer are extremely complex (e.g., heat ex-
changers) and many of them in addition ill-posed problems. The Reynolds
analogy factor is fairly well documented for constant-pressure boundary
layers, or duct flows, with small temperature differences. Data for flows
in pressure gradients, and plausible methods of correlating the data, are
rarer, and in the case of large temperature differences little information
is available about the turbulence structure, which is now influenced also
by the fluctuations of density and temperature. In the face of these prob-
lems, the usual approach is to transfer first-order closure concepts, such
as those of mixing length or eddy viscosity, from the momentum equation
to the energy equation and to introduce an eddy diffusivity Yx:

~ T = -o--'O/( OT/ c3y) (2.33)

where 0 is the fluctuating temperature and cp the specific heat at constant

pressure. In most cases the turbulent Prandtl number ~rt defined, analog-
ously to the molecular Prandtl number a, as the ratio of eddy viscosity to
diffusivity is assumed to be constant across the boundary layer (for
at 4=constant see MEIER et al. [2.206]). First-order closure schemes were
used in the calculation methods of LEONT'EV [2.207], PATANKARand
SPALDING [2.208], GOSMAN et al. [2.209], and CEBEO et al. [2.210]
FLAHERTY [2.211] and CRAWrORDand KAYS [2.212] and more modern
closure schemes are discussed in Chapter 6.
There exist few experimental investigations of turbulent boundary
layers with heat transfer and even fewer with high heat transfer rates. For
zero pressure gradient flow see KELNHOFER[2.213] (heated wall), BACK
et al. [2.214] (cooled wall) and for accelerated flows BOLl)MAN et al.
[2.215], BACKand CUFFEL [2.216], and KEARNEY et al. [2.217]. The
 External Flows 79

other experiments known to the author have only moderate to small

temperature differences across the boundary layer (e.g., HATTONand
EUSTACE [2.218], and TmV.LBARet al. [2.219]) and assume constant fluid
properties. Little information is available on boundary layers with heat
transfer in an adverse pressure gradient. As separation is approached,
the heat transfer at the wall falls off, but less rapidly than the skin friction
(DvoRAK and HEAD [2.220]). It is obvious that the heat transfer does not
vanish at separation or reattachment since dT/ay does not tend to zero
(e.g., CttILCOTT [2.221] and FLETCHERet al. [2.222]).
In most of the references quoted above the wall was kept more or less
isothermal. Under this condition there is good reason to assume that the
influence of the pressure gradient on the temperature profile is weak
(CottoN and R~HOTKO [2.223]). However, if there is a streamwise
gradient of wall temperature, a strong influence of the heat transfer on
the temperature profile must be expected (DIEN~tAr~N [2.224]). Another
important problem is the effect of the upstream thermal history of the
boundary layer which markedly influences the surface heat transfer
characteristics downstream (REYNOLDSet al. [2.2251 MOREa~n and KAYS
[2.226], and K~RNL~Y et al. [2.217]).
Finally we should mention relationships for the mean velocity and the
temperature in turbulent boundary layers. For the inner layer they have
the following form (DvaSSLEa~[2.227]):

U = u , f ( y u J v w, fl), (2.34)

T = T w f +(yuJv,,,, fl). (2.35)

Here fl = Qw/(ecpu, Tw) =- TJTw, where T~ is Squire's "friction temperature"

[1.2, p. 823] which according to [2.228] was introduced independently
by LANDAUand LIFSCHITZ in 1944. Relationships for the log-law region
were derived by B~oSrIAW [2.229] and the author [2.230], for the sub-
layer by M ~ K and BAER [2.231] and for the whole profile by Ror'rA
[2.205]. Equilibrium profiles were calculated by AtaER and COATS
[2.232]. Survey papers on heat transfer problems were published by
KESTIN and tOCHARX)SON[2.233] and SPALDING[2.234].

2.3.10 Transition from Laminar to Turbulent Flow

The phenomenon of laminar-turbulent transition was identified as long

ago as 1883, by REYNOLDS,and accordingly there exists a large body of
theoretical and experimental results, a description of which would go
well beyond the scope of this book. What can be done, however, is to
warn the reader to be especially critical before applying transition
80 H.-H. FERNHOLZ

criteria and other research results, usually obtained under quite specific
conditions. MORKOVlN[2.235] gives an admirable survey of the problems
and of current knowledge, while RESI4OTKO [2.236] outlines the most
urgent problems and suggests further work. It is generally accepted
now that transition is preceded by instability of the laminar shear layer.
Since the theory of hydrodynamic stability in shear layers, with boundary
layers as a special case, can be considered as a field of its own, it must
suffice here to refer to the monograph by LIN [1.10] and the papers by
MACK [2.237, 238], CRAm [2.239] and GASTER [2.240] to name but a
few. The more recent stability calculation schemes deal with non-
parallel flow models and take into account nonlinear effects.
In the "ideal case" of an incompressible two-dimensional laminar
boundary layer with zero pressure gradient which is free of "artifical"
disturbances due to boundary conditions, the process of instability and
"natural" transition was summarized by STUART [2.241] and MOLLO-
CHRISTENSEN [2.242]: "Beyond a certain Reynolds number travelling
waves of velocity fluctations develop and grow with downstream
distance. Next, spanwise disturbances cause local free shear layers to
form which again cause bursts of high-frequency instability fluctuations
with a very high growth rate. The flow then erupts locally into the
Emmon's spots (i.e., patches of turbulent flow in a laminar surrounding)
which further downstream cause such a confused flow that we call it
turbulent". It is obvious from this description, from the basic ex-
perimental investigations by SCHUBAUERand KLEaANOFF [2.243] and
KLEBANOFFet al. [2.244], and from the smoke photographs by KNAPP
and ROACHE [2.245] that transition does not occur instantaneously
and that both the location and the extend of the transition region in a
specific flow configuration are required. For this ideal case--even with
variable pressure gradient--the critical point (denoted by a critical
Reynolds number (Reo)crlt. as the upstream boundary) of the transition
region can be reasonably well predicted (e.g., SMITH and GAMBERONI
[2.246], JArFE et al. [2.247]). A general prediction of transition is
impossible at present as we have to cope with an ill-posed problem.
It is rarely possible to specify the details of the initial disturbances
that exist in a real flow because they arise from a number of boundary
conditions such as free-stream turbulence, surface roughness, surface
curvature, surface temperature, three-dimensional effects (secondary
flow), compressibility, noise and structural vibrations. This means
that besides the model and flow conditions, the disturbance environment,
i.e., the wind tunnel or flight path conditions, must also be known.
Furthermore, though it is generally known whether the above-mentioned
parameters have a stabilizing or a destabilizing effect on the boundary
layer, little information is available about the mutual interaction of
 External Flows 81

these effects. As for supersonic boundary layers, the reader is referred

to Morkovin's survey paper with its comprehensive study of the available
literature (345 references) up to 1969. One of Morkovin's conclusions
was that the empirical basis for the correlations and the prediction
methods of transition in high-speed boundary layers used in industry
was severely challenged, with little hope offered for more solid evidence
in the near future.
We close with a brief summary of the effects on transition of the
parameters and boundary conditions reviewed in previous sections (for
more details see MORKOV1N [2.235] and TANI [2.248]).
A favorable pressure gradient increases the stability of a flow in the
sense that the critical Reynolds number is higher than for the boundary
layer with zero pressure gradient. An unfavorable pressure gradient
destabilizes the flow due to the formation of velocity profiles with an
inflection point (for a criterion see SMn'rI and GAMBERONI [2.246]).
The transition region can be rather long in a zero pressure gradient with
low free-stream turbulence (ScI4UBAUER and KLEBANOFF[2.243]) and
rather short in the free shear layer of a reattachment region or on a
separation bubble, i.e., in case where the pressure gradient is unfavorable.
The transition process is accelerated by the presence of free-stream
turbulence which appears to control the rapidity with which the rest
of the spectrum feeds on the amplified component. A lower limit for the
significant influence of free-stream turbulence seems to be a turbulence
level of 0.08 % (TAN0. The effect of surface roughness on transition
has been investigated more than any other parameter. This is due to the
multiplicity of two- and three-dimensional roughness elements and
configurations and to the impossibility of determining the effect of transi-
tion devices by theoretical methods. For incompressible boundary
layers the reader is referred to SCHLICHTING [1.3] and TANI [2.248]
for further references.
It is known from theoretical studies by G6RTLER [2.191] and his
co-workers that the instability of a boundary layer along a concave
surface is enhanced due to the normal pressure gradient, resulting in a
configuration of longitudinal vortices. On a convex surface the vortex
mode is not excited, and transition proceeds as on a fiat surface.
So far as we know today, stabilization of the laminar boundary
layer can be achieved only by cooling of the wall, suction or a favorable
pressure gradient. But here again a warning is necessary since transition
on so-called "heat-sink" noses of ballistic missiles was not delayed by
the strong favorable influences of cooling and rapidly falling pressure
(MORKOVIN [2.235]).
The problem of the disturbance environment as summarized by
MORKOVIN was investigated further by DOVGHERTV [2.249]. For more
82 H.-H. FERNHOLZ

recent experimental investigations of transition the reader is referred

to the papers by PATE [2.250, 251], LA GRAFF [2.252], OWEN and
HORSTMAN [2.253] and STAINBACK and ANDERS [2.254]. Among the
attempts to predict the boundary layer behavior through the transition
region the investigations by HARRIS [2.255], MCDONALD and FISH
[2.256] and CESECl (Sect. 5.8) should be mentioned, where the eddy
viscosity is modified according to the statistical distribution of turbulent
spots. A paper using a more refined approach is that by BUSHNELLand
ALSTON [2.257].
Transition in pipe flow was investigated by SMITH [2.258],
WYGNANSKI and CHAMPAGNE [2.259] and WVGNANSKI et al. [2.260].
A detailed picture of the sequence of instabilities and interactions which
lead to transition in free shear layers was given by MIKSAD [2.261].

2.3.11 Relaminarization and Reverse Transition

Reverse transition and relaminarization have been observed in pipe-

and duct flows and in boundary layers. Though the terms are used
synonymously in most investigations, a distinction will be made in this
subsection. An initially turbulent shear flow is in a state of relaminariza-
tion if the laminar state of the flow is clearly established, such as by the
parabolic shape of the mean velocity profile in pipe flow (SmuLKIN
[2.262], LALrFER [2.263], MCELIGOT et al. [2.264] and BADm
NAaAYANAN [2.265] ( in duct flow)) or by a typical laminar profile in
boundary layers (LAUNDER[2.108] and BAom NARAYANANand RAMJEE
[2.266]). Relaminarization of pipe flow for example was achieved by
decreasing the pipe Reynolds number below its critical value of about
2100.
The process of reverse transition is much more difficult to define.
Some features exist, however, the occurrence of which indicates reverse
transition of an initially turbulent boundary layer. The mean velocity
profile departs from its characteristic log-law form, being different
from a laminar and from a highly accelerated velocity profile (PATEL
and HEAD [2.35]); the relevant Reynolds number is comparatively low
(Reo < 2500) and the sublayer thickness increases. There is some dis-
crepancy between the "reverse transition" mean velocity profiles
measured by BLACKWELDER and KOVASZNAY [2.268] and BADRI
NARAYANANand RAMJEE[2.266], showing a similar behavior to forward-
transition profiles, and those of PATEL and HEAD [2.35] which fall
below the log-law in the outer layer.
Several parameters controlling the onset of reverse transition have
been suggested (K = (v/U2)dUddx by LAUNDER[2.108], Zip-----(v/Qu~)dp/dx
and A,=(v~/~y)/(Oua~) by PATELand HEAD [2.35], and combinations
 External Flows 83

thereof (e.g., BRADSHAW[2.269]): A~ is the correct inner-layer parameter.

More reliable skin friction measurements are needed before distinct
values can be allocated to these parameters (NARASIMHAand SREBNI-
VASAN[2.270]).
As far as the turbulence structure in a boundary layer undergoing
reverse transition is concerned, the following was observed (BLACK-
WELDERand KOVASZNAY [2.268]). Measurements of the fluctuating
velocity components and the Reynolds shear stress indicated that the
absolute levels of the velocities and the shear stress were approximately
constant along a mean streamline except near the wall. However, the
relative levels were decreasing, as reported already by LAUNDER[2.108],
for example. KLn'~E et al. [2.11] found a decrease in the number of
turbulent bursts in the wall region if the pressure gradient parameter K
reached a critical value of 3.7 x 10 -6, i.e., a reduction in the generation
of turbulent energy which again leads to a decay of the turbulence in the
outer part of the boundary layer. This cessation of turbulent bursts
probably accounts for the departure of the velocity profiles in the log-law
region. It is not clear, however, why the large eddy structure was not
changed significantly after passing through the region where the favorable
pressure gradient was strongest (BLACKWELDERand KOVASZNAY[2.268]
from space-time autocorrelations). "Sink-flow" turbulent boundary
layers between converging planes were investigated theoretically by
LAUNDERand JONES [2.271], and the same authors presented a pre-
diction method of reverse transition with a two-equation model of
turbulence [2.272]. Reverse transition and relaminarization are phenome-
na which are of special practical importance for the heat transfer in
sub- and supersonic nozzle flows, and several papers deal with this
aspect, e.g., O'BmEN [2.273], MORETTI and KAYS [2.226], BOLDMANet al.
[2.274], BACK et al. [2.275], NASH-WEBBERandOATES [2.276] and
PERKINSand MCEUGOT [2.277].

2.4 Three-Dimensional Boundary Layers

2.4.1 Classification of Three-Dimensional Flows

PRANDTL [2.288] opened his lecture on "Boundary Layers in Three-
Dimensional flow" by introducing the two phenomena which distinguish
three-dimensional from two-dimensional boundary layers:
a) The lateral (spanwise, z-wise) convergence or divergence of the
velocity components of the potential flow parallel to the wall. In this
case only the thickness of the boundary layer is affected and consequently
the velocity profile and the wall shear stress. The mean velocity profiles
84 H.-H. FERNHOLZ

are collateral, i.e., the velocity vector points in one direction throughout
the boundary layer. The departure from two dimensionality can be ac-
counted for by a single term in the momentum integral equation, adding
O/(x o - x)to the right-hand side of (2.21b). Here x o is the virtual origin
of the streamlines as seen in plan view. For details, see PIERCE [2.289"1.
b) The secondary flow due to the lateral curvature of the potential
flow. Because of the turning of the external streamlines the velocity
vectors in the boundary layer, subjected to the same radial pressure
gradient but of smaller size than in the outer flow, follow a turning angle
which is a function of y and is usually largest at the surface, that is, the
velocity profiles are skewed.
Three-dimensional boundary layers belong to the class of three-
dimensional "thin shear layers" where 8/tgy~> ~/dx ,,~ 8/9z when operating
on any velocity component. The boundary layer equations are presented
here for reasons of simplicity in a cartesian coordinate system where the
x-direction can be either surface orientated or be given by the direction
of the external streamlines. Then the coordinates on the wall, x and z,
are the projections of the external streamlines and their orthogonal
trajectories. As for the choice of more general, i.e., curvilinear coordinate
systems for three-dimensional boundary layers and a detailed discussion
thereof, see for example HAYES [2.290], SQUIRE [2.291], RAKICH and
MATEER [2.292], CEBECI et al. [2.293], KRAUSE [2.294] and the review
papers by EICHELBRENNER [2.295] and BLOTTNER [2.296], the latter
especially on calculation methods.
The second kind of three-dimensional shear flow along a wall is
typified by the flow in the corner of a duct or of a wing-body junction
where t3/dy, ~/az >3>d/ax (slender shear flow). The shear stress gradients in
both the y- and z-directions are important in the Reynolds equations.
The flow field as mapped by GESSNERand JONES [2.297] or EICHEL-
BRENNER [2.298] for example shows a corner vortex in the case of
turbulent flow which is called a secondary flow of Prandtl's second kind.
With corner-flow configurations we encounter one of the rare cases where
laminar and turbulent flow show completely different tendencies in their
general behavior. F r o m the considerations and results of ZAMm [2.299],
ZAMIR and YOUNG [2.300], PERKINS [2.301] and BRAGG [2.302] it ap-
pears that the secondary flow along the bisector of the corner is directed
away from the corner in laminar flow s and towards the corner in turbu-
lent flow (e.g., EICBELBRENNER and NGUYEN [2.303], GESSNER and
JONES [2.297]). This latter effect has been found to be linked to the
turbulence characteristics of the flow and has been investigated by

s It is just possible that the secondary flow that appears in the laminar case results
from a localizedinstabilityin the corner, wherethe velocityprofileshave points of inflection.
 External Flows 85

EICHELBRENNER [2.304, 298"], EICHELBRENNERand PRF~TON [2.305],

EICHELB~NNER et al. [2.306] and MOJOLAand YOUNG [2.307]. ROTTA
[1.5] presents the Reynolds equations for the boundary layer in the corner
of a square duct and gives the conditions under which no secondary flow
would occur. Transitional corner flow was investigated by ZAMIR and
YOUNG and corner flow with an adverse pressure gradient by BRAGG,
NGUYEN [2.308] and MOJOLAand YOUNG in turbulent flow showing
similar characteristics as in two-dimensional retarded boundary layers.
For further discussions of secondary flow in ducts see Section 3.1.
The third kind of three-dimensional shear flow, in which ~/3x~
~/3y,~ (~/~z, needs either higher order boundary layer equations or the
full Navier-Stokes equations, as for example in some cases of the flow in
turbomachines (Chapt. 3).
Before we discuss some specific problems of turbulent three-di-
mensional boundary layers i t may be advantageous to mention some
further review papers; COOKEand HALL [2.309], JOUaERT et al. [2.310],
SHERMAN [2.311], WHEELER and JOHNSTON [2.312, 313], NASH and
PATEL [2.314] and BLOTTNER (especially on numerical calculations
[2.296]), and three progress reports, HORLOCKet al. [2.315], FERNHOLZ
[2.316] and EmHELBREh'NER[2.295].

2.4.2 Three-Dimensional Thin Shear Layers

Under the conditions stated below, the equations for three-dimensional

boundary layers as given in Chapter 1 can be simplified further. Since
the validity range of the underlying assumptions has often been extended
to flow configurations where these additional assumptions are not
justified, a brief discussion seems to be appropriate. One can distinguish
between three kinds of simplification:
a) the "independence principle" according to PRANDTL [2.2883,
b) the infinite swept wing or infinite yawed cylinder,
c) the "small crossflow assumption" or the "principle of prevalence"
according to EICHELBRENNERand OUDART [2.317"]; the "crossflow" is
the velocity component normal to the local free stream.
Since the "independence principle" applies to laminar flow only, the
reader is referred to COOKE [2.318J. In this case, however, the underlined
terms in (2.36) to (2.39), the equations of mean motion in rectangular
cartesian coordinates, vanish.

~(ffu)+~(Sv)+~(~w)=o, (2.36)
86 H.-H. FERNHOLZ

_ ou Off
(2.37)

( OW+vOW wOWI off o(ow _~_~) (2.38)

+ / = + Ty ,

y=0: U=V=W=O; u=v=w=O, (2.39)

y = 6: U = U e ; W= We . (2.40)

The reason why the independence principle cannot be applied to turbu-

lent boundary layers is that there is interaction between the spanwise
and chordwise components of the velocity fluctuations.
Though the turbulent flow on an infinite swept wing is not different
from a general three-dimensional boundary layer from a physical
point of view, it is easier to treat mathematically (derivatives in z-
direction must be identically zero and the surface must be developable).
There is a further advantage that under infinite swept-wing conditions
it is sufficient to perform measurements at one section only, thus reducing
the amount of experimental data considerably. These considerations
do not hold for finite swept wings or tapered wings where there exists
a boundary layer drift towards the wing tip. For calculation methods
dealing with the boundary layer on infinite swept wings see for example
CUMPSTY and HEAD [2.319], ADAMS [2.320], KRAUSE [2.294] and
CEBECi et al. [2.293].
The "small crossflow assumption" or the "principle of prevalence"
due to EICHELBRENNERand OUDART[2.317] is applicable to compressible
and incompressible laminar and turbulent boundary layers. It amounts
to assuming that W is so small that it may be ignored in the momentum
equation for U. The equation then becomes uncoupled from the equation
for W. EICHELBRENNER and PEUBE [2.321] have extended the small
crossflow assumption to an attenuated "principle of prevalence".
Originally the principle of prevalence was deduced for laminar boundary
layers and applied to the case of slender bodies with small angles of
incidence and consequently moderate crossflow. An extension to
turbulent flow is, however, confined to even smaller crossflow due to
the strong crosswise momentum and energy transport caused by the
transverse turbulent velocity fluctations. Furthermore Reynolds shear
stresses are influenced implicitly by the spanwise flow (see for example
the transport equation for ~ (Chapt. 1)). Lacking a better criterion
it may be advisable to follow CUMPSTY and HEAD [2.322] who restrict
the small crossflow assumption to angles between the free stream and the
limiting streamline smaller than 6 degrees in turbulent boundary layers.
 External Flows 87

_-~ P

Fig. 2.6. Domain of dependence of a point P in a three-dimensional boundary layer

According to the boundary layer equations, the solution at a given

point is affected by upstream conditions only within a wedge-shaped
"domain of dependence" whose apex is the point considered [-2.323, 324].
This is RaETZ'S [2.323] "influence principle". The wedge (Fig. 2.6) is
bounded by envelopes of the characteristics (normals to the surface
through the limiting streamline and the outer edge streamline) and the
subcharacteristics which are the streamlines themselves. Though this
model, like the boundary layer equations, neglects diffusion of vorticity
in the x, z plane and assumes instantaneous propagation of pressure
disturbances normal to the x, z plane, it is of great practical importance
since it allows one to determine which upstream conditions must be
known to calculate the boundary layer in a specific region (see WANG
[2.325] and WESSELING[2.324]).
There are two further flow configurations for which the equations
for three-dimensional boundary layers can be simplified (see PIERCE
[-2.326] and NASH and PAT~ [2.314]). In all flows which have a plane
of symmetry, the cross-flow components of the mean velocity ( W = 0 )
and of the shear stress (~-~= 0) vanish whereas the gradients in z-direction
must be retained. The equations are of course the same as derived via
the small crossflow approximation and hence may be expected to hold
for some short distance on either side of the plane of symmetry. For a
more detailed discussion of the turbulent flow at a plane of symmetry
the reader is referred to investigations by JOHYSTON [2.327], PIERCE
[2.326], MELLOR [-2.328] and HEAD and PRAHL~ [2.329]. The second
special flow is the flow along an attachment line, for instance along the
nose of a swept wing. In this ease the equations of the three-dimensional
boundary layer are even more simplified by setting all derivatives in the
spanwise direction zero (see for example CtnUPSTYand HEAD [2.330, 331]
and CEBECl [2.332]). However as the chordwise velocity is zero, chord-
wise differentiation is necessary to give equations which can be solved
by a numerical "marching" method.
88 H.-H. FERNHOLZ

Now, general three-dimensional thin shear layers may be divided

into two groups: firstly, those which are mainly pressure driven such as
on swept wings, in curved channels and on a flat plate around an obstacle;
and secondly, mainly shear-driven boundary layers, which are more
difficult to generate. Flows of both kinds can have very complex shear
and velocity distributions (see, e.g., EAST and HOXEV [2.333] and can
include separation regions. Experiments of the first type which have been
published after Nash and Patel's review [2.314] are those on swept
wings by ETHERIDGE [2.334], VAN DEN BERG et al. [2.335], ELSENAAR
and BOELSMA [-2.336] and EAST [2.337]; in curved channels by
KLINKSIEK and PIERCE [2.338], VERMEULEN [2.339] and BANSOD and
BRADSHAW [2.340]; and on ship hulls by LARSSON [2.341]. Only three
experiments of the second type are known so far, MOORE and
RICHARDSON [2.342], BRADSHAW and TERRELL [2.343] and CRABBE
[2.344]. In contrast to the case of two-dimensional boundary layers
little information about the turbulence structure in three-dimensional
thin shear layers and the flow structure in the immediate vicinity of the
wall is available. Therefore many investigations have assumed that most
of the properties of collateral boundary layers can be carried over to
skewed turbulent boundary layers. Prediction methods based on such
assumptions may give acceptable agreement with experiments where
the three-dimensional boundary layer extends only a few boundary
layer thicknesses downstream but may well fail in cases where a three-
dimensional boundary layer had time to develop properly. Due to the
limited amount of space we cannot discuss here the whole range of
seirfi-empirical relations, but some of the controversial assumptions
will be pointed out in the hope to make it clear that more experiments
are needed. For a more detailed discussion the reader is referred to
NASH and PATEL[2.314].
Semi-empirical relationships for the components of the mean
velocity vector as deduced from experiments have been given by many
authors, e.g., for the "crossflow" profile by MAGER [2.345] and the
polar representation which devides the velocity profile into an inner
and outer region by JOHNSTON [2.346]. Both models assume a collateral
velocity vector in the wall region, which is very unlikely according to
more recent measurements (see for instance PIERCE [2.347]). Neither
of these models can take into account velocity profiles with a change
of signs of the crossfiow component (see also PRAHLAD[2.348]). Down-
stream of inflection points of the outer flow streamlines, crossflow
velocity profiles of multiple sign have been observed by BOUSGARm~
[2.349], EICHELBRENNERand PEUBE [2.350], KLINKSIEKand PIERCE
[2.338], ETa4ERngGE [2.334] and others. EICHELBRENNER and PEUBE
[2.350] developed a relationship which can take account of S-shaped
crossflow profiles, but agreement with measurements is not satisfactory.
 External Flows 89

As with two-dimensional boundary layers the set of equations for the

three-dimensional case can only be closed if the Reynolds stresses can be
determined either from semi-empirical relations (first-order closure) or
from transport equations for turbulence properties (second-order
closure). For a first-order closure, mixing-length or eddy-viscosity
concepts (Sect. 1.8) are generally carried over from the two-dimensional
case, very often even assuming isotropy of the eddy viscosity vx, for
instance:

- u - ~ = (vT), 8c~x2
U, (2.41)

where the index 2 denotes the coordinate normal to the wall and where
may be either 1 or 3. There are several reasons for criticism here. It is
well known that first-order closures have been successful in two-
dimensional flow only if the boundary layer was always close to equili-
brium. Neither eddy viscosity nor mixing length correlations hold for
flows which are strongly influenced by their upstream history. Equa-
tion (2.41) implies that the direction of the Reynolds shear stress vector
is always the same as the direction of the velocity-gradient vector. This is
often expressed by another auxiliary equation

~ v / ~ = ( aU / ay/( aW / ay) . (2.42)

It has been shown, however (cf. JOHNSTON [2.351] and EAST [2.337]),
that the angle through which the shear stress has turned is generally
intermediate between the angle of the mean velocity and the mean
velocity gradient. Johnston found angular differences between the
direction of the turbulent shear stress and the direction of the mean
velocity gradient of about 20 . There is a further group of calculation
methods where an extended version of HeRd's entrainment concept
is used for three-dimensional turbulent boundary layers. A typical
method is described in Subsection 5.7.3. Suggestions for a second-
order closure procedure have been made by BRADSHAW [2.352] and
SPALDING [2.353]. A version of the law of the wall generalized to three-
dimensional boundary layers was derived by VAN DEN BERG [2.354]
and EAST [2.355].
There exist few measurements of turbulence quantities which is one
of the reason why many of the concepts mentioned above could not be
checked properly (see for example ASI4KENAS[2.356], BRADSHAWand
TERRELL[2.343], JOHNSTON[2.351] and ELSENAARand BOELSMA[2.336])
One of the unsolved questions remains whether and how much the local
flow angle varies down to the wall (PIERCE [2.347], PIERCE and
KROMMENHOEK [2.357] and PRAHLAD [2.358]). VERMEULEN [2.339]
90 H.-H. FERNHOLZ

reported rates of change of the crossflow angle as large as 0.2 deg per
unit of u~y/v in the immediate vicinity of the wall. For measurements
in compressible turbulent boundary layers in three dimensions the reader
is referred to HALL and DICKENS [2.359], RAINBIRD [2.360], FISHER
and WEINSTEIN [2.361, 362] and COUSTEIXand MICHEL [2.363].
Three-dimensional turbulent boundary layers on rotating bodies
have been dealt with for example by PARR [2.364], STEINHEUER[2.365]
and CHAM and HEAD [2.366].

2.4.3 Separationin Three-DimensionalFlow

In Subsection 2.3.6 the separation problem in two-dimensional boundary
layers has been found to be rather complex, for reasons connected with
the structure of the problem (lack of experimental evidence) and with
the validity of the boundary layer approximations, and also for ma-
thematical reasons. Beginning with three-dimensional laminar boundary
layers we find that our understanding of the separation phenomenon
is even more incomplete, as can be seen from the investigations by
BROWN and STEWARTSON [2.112] and by BUCKMASTER [2.367]. The
latter author states a basic difficulty as lying in the very definition of
separation in three-dimensional flow: "In general the skin friction only
vanishes at singular points--where both components of the wall shear
stress vector are zero--although naturally the boundary layer leaves
the body along a curve". EICnELBRENNER[2.295] discussed separation
in three-dimensional boundary layers in more detail arriving mainly
at the definition of HAVES [2.290] : "If a particular streamline is found
which separates streamlines coming from an unseparated part of the
surface from those coming from a clearly separated part of the surface,
this bounding line may be taken as the separation line". This separation
criterion is also applicable to turbulent boundary layer, but more
careful measurements in the separation region of skewed turbulent
boundary layer are necessary (TAYLOR [2.368]) and with special em-
phasis on aerodynamical problems (SMITH [2.369]).

2.5 Two-Dimensional Boundary Layers at High

Mach Numbers

Supersonic flight of aircraft and rockets has undoubtedly been the main
incentive for the research on compressible turbulent boundary layers,
and one of the most important aims has been to determine shear stress
and heat transfer at the wall. It is not surprising therefore, to find a large
 External Flows 91

body of investigations in the open literature, and the number of those

still classified can only be guessed. The momentum and energy equations
for steady two-dimensional compressible turbulent boundary layers are
as follows if effects due to dissociation or ionization are excluded. The
energy equation is given here in dimensionless form to show the explicit
influence of the Mach number Moo.

~U OU OU dP t3 (fi OU _~fff) (2.43)

- dx + T y \

~ 0 ~ +(~lT'+q v ) ~ = ( 7 - 1)M ooU ~x

+ a~oRe~ -ff~~ ~ ] /

Quantities denoted by ^ , unless characteristic parameters, were made

dimensionless by flow quantities of the undisturbed flow; overbars on
/~, 0, etc., denote mean values, and some small terms have been neglected.
For boundary layers with zero pressure gradient the only direct effect
of the Mach number occurs in the term where heat is added to the
boundary layer, by direct dissipation and indirectly via production and
dissipation of the turbulent fluctuations. The larger the Mach number,
the greater is the dissipation. Heat transfer at the wall and heat addition
by dissipation influence the velocity field by the distribution of the
density ff and the temperature T, the latter via the viscosity fi and the
heat conductivity k. Since dissipation occurs mainly in the vicinity of the
wall the maximum of the static temperature is found close to (cooled wall)
or at the wall (heated or adiabatic wall). Such temperature distributions
again lead to mass-flow profiles where the bulk of the mass flow is in the
outer layer of the boundary layer and to regions of low density at the wall
(a ,-~ T -I for Op/~y~O). Measurements performed close to the wall need
therefore rarefied gas corrections under certain conditions (see e.g.
BECKWITI-I et al. [2.371]). Typical velocity, temperature and mass-flow
distributions are given in Fig. 2.7. Other quantities being equal, the
increased temperature at the surface implies a larger viscosity and a
smaller density, which cause the ratio fi/ff to increase and the local
Reynolds number to decrease, leading to an increase of the viscous sub-
9 h=cpT; ?=Cp/Cv;Mo~=UI(TP/~; a~o=(IZcp/k)~o;,~=k/cp; Re~=(UoL/IO~; O=
temperature fluctuation. For a derivation of this equation for laminar boundary layers see
MOORE 12.370].
92 H.-H. FERNHOLZ

2,5 . . . .

yxlO2 [ml

S
T 2,0'

1,5
\ , u_.
U6

PaU6

' / \\
0.5

0.2 O,Z, 0.6 0.8 1.0

pu u
P6U6 U6

r
T,
Fig. 2.7. Mean velocity, temperature and mass-flow profile for a Mc=7.2 boundary layer
with cooling (Tw/T,=0.51) at Reo= 2074 (taken from [2.393-1. Suffix 6 denotes conditions
at y=6

thickness. It may be mentioned in passing that the characteristic Reynolds

number

Reo = Je UeO/,ffw (2.45)

the definition of which arises naturally from the momentum integral

relation (WALZ [2.372]), is a sensible compromise taking into account
the large variation of p and 0 across the compressible boundary layer.
Finally there is, of course, an indirect influence of the Mach number on
the boundary layer via the boundary conditions Ue, Pe, and ho.
A question which cannot be answered from inspection of (2.43) and
(2.44) only is whether and how strongly the terms ~ and v--0are influenced
by compressibility effects. Both flow visualization and hot wire measure-
ments have shown that the compressible boundary layer exhibits a large-
scale structure and an intermittent flow pattern in the outer layer. Due to
the relatively thick sublayer and the location of the maximum of the
turbulent fluctuations near the viscous sublayer a similar intermittent
behavior between high- and low-intensity fluid has been observed in the
inner layer (KOVASZNAY [2.373] and OWEN et al. [2.374]). The latter
 External Flows 93

layer authors showed that the intermittency at the outer edge is more
sharply defined in a Mach 7 boundary layer, but does not start until y/6
much larger than in the low-speed case.
The foundations for fluctuation measurements in compressible
boundary layers were laid by KOVASZNAV[2.373, 375] and MORKOVIN
I-2.376, 377], but there are still only a few data sets available, mainly
obtained with hot-wire anemometers (for a recent investigation see
DEMErRIADES and LADERMAN [2.378]). In these investigations it could
be shown that the three types of disturbance fields or modes (vorticity,
entropy and sound wave), all of which obey the Navier-Stokes equa-
tions, are only weakly dependent on each other when the fluctuation
intensities are small, but interact at larger intensities when linearization
is no longer permissible. As long as the temperature fluctuations can be
assumed isobaric, the sound wave mode (though important in itself) can be
neglected compared with the rms mass-flow fluctuations. It is a lucky
coincidence that the hot wire responds mainly to stagnation-temperature
fluctuations at low overheat ratios and to mass-flow fluctuations at high
overheat ratios, so that these two fluctuation components can be
measured almost independently.
It must not be forgotten, however, that--possibly at Mach numbers
around lO---the sound mode and its interaction with the other modes
mentioned above become important, forming a barrier against the
measuring techniques applied so far, so that electron beam or laser-
Doppler measuring techniques may have to supplement the hot-wire
technique (WALLACE[2.379], HARVEYand BUSHNELL [2.380], I~YNER
[2.381], and YANTAand LEE [2.382]).
KISTLER [2.383] found that mass-flow fluctuations and total tempera-
ture fluctuations (0t) increased with Mach number in adiabatic boundary
layers "dp/dx=0 and Me ~4.7), whereas OWEN et al. [2.374] state that for
moderately cooled boundary layers (dp/dx=O, M e ~ 7 ) the mass-flow
fluctuations appear to be independent of Mach number while the 0 t-
fluctuations decrease with rising Mach number. A relationship between
the fluctuating quantities was derived by MORKOVIN [2.384] if second-
order terms can be neglected

~ . 1 0 --M2)-lu+ (7-- 1)M2 (1 + ~ - ~ M2) -1

0, = _~(1+ (2.46)

or for adiabatic wall conditions (0t = 0)

0 u
= U ( 7 - 1)M2. (2.47)

Equation (2.47) was used extensively by BRAOSHAW[2.385] for example.

94 H.-H. FERNHOLZ

For a further discussion of the mode interaction and the limitations

of the mode-concept (e.g., one field-point measurement only) see the
excellent survey paper by MORKOVIN [2.384] which forms the base for
Morkovin's hypothesis [2.386] where it is stated that "the essential
dynamics of adiabatic compressible boundary layers follow the in-
compressible pattern for flows with M <5", a limit which may now be
extended to Mach numbers of about 8 and with cooling of the wall.
For further details of the turbulence structure of compressible turbu-
lent boundary layers the reader is referred to OWEN and HORSTMAN
[2.387] and ROSE [2.388]. Calculations of shear stress, eddy viscosity
and mixing length distributions were made by MAISE and MCDONALD
[2.389], MARTELLUCCIet al. [2.388] and BUSHNELLand MORRIS [2.391]
from mean velocity and temperature measurements. Prandtl number
distributions across the boundary layer have been investigated by
MEIER and ROTTA[2.392] and HORSTMANand OWEN [2.393] for example.
Since the measurements of fluctuating quantities in compressible
turbulent boundary layer (e.g., KISTLER [2.383], LADERMANand
DEMETmADES [2.394], OWEN et al. [2.387, 374] and ROSE [2.388, 395])
are not sufficient yet to develop adequate models of the turbulence
structure, information about the distribution of mean velocity and
temperature across the boundary layer is still very important. Beginning
with the mean temperature distribution, we know of the existence of two
energy integrals, i.e., a relation between the static enthalpy h and the
mean velocity U, which are valid throughout the boundary layer in-
dependent of x and y. For laminar boundary layers, Prandtl number one
and zero pressure gradient, such an energy integral exists if hw is constant
(see MOOl~ [Ref. 2.370, p. 211]). A second energy integral can be ob-
tained even if the pressure gradient is different from zero, provided that
the heat transfer is zero at the wall. In the first case, the existence of an
energy integral implies the so-called Reynolds analogy which means that
the heat transfer at the wall is proportional to the skin friction. For
turbulent boundary layers and Prandtl number close to one--as is the
case for the technically important gases--these energy integrals (first
derived by CROCCO and BUS~,~AtCN) have been modified by VAN DRIEST
[2.396] under the additional condition noted first by MoRIcOVlN [2.384]
that (b"~t)/(v-"0),~1. The energy integral for a boundary layer along an
adiabatic wall, cr ~ 1 and variable pressure gradient reads:

T I T e = 1 + (;~- 1 ) r M z [1 - ( u I u o ) 2] (2.48)

where r is the recovery factor, r is assumed here to be 0.89 for air as a

first approximation or can be expressed as a function of the Prandtl
number (MEIER et al. [2.397]).
 External Flows 95

For an isothermal wall, zero pressure gradient and cr--~1 VAN DRIEST
obtained the following relationship:

T / T e -- T w / T e + ( U / U e ) { 1 - (Tw/Te) + 0.5r(T - 1)M 2 [ 1 - (U/Ue)]}

(2.49)

and for the Reynolds analogy:

St = Fzw/Oe U 2 (2.50)

where St is the Stanton number and Fthe Reynolds analogy factor, which
unfortunately has been found to vary between 0.8 and 1.2 (CARY [2.398]
and COLEMANet al. [2.399]).
Equations (2.48) and (2.49) are often written for the total temperature
or the total enthalpy but it is little known that the form given above
permits a more realistic comparison of measured and calculated data.
Attempts have been made to extend these relations to flows with pressure
gradients and heat transfer but the results are not yet convincing (WALZ
[2.372] and KOSTER [2.400] for example). In order to gain more in-
formation about the mean velocity distribution, which must be known
to evaluate the energy integrals, it is natural to revert to the multi-layer
concept of the turbulent boundary layer, especially since the basic
features of the turbulence structure in compressible and incompressible
flow can be assumed to be approximately similar, by virtue of Morkovin's
hypothesis. Many attempts have been made to scale or transform both
abscissa and ordinate of the log law (2.3) or only one of them (COLES
[2.401], ROTTA [2.402], BARONTIand LIBBY [2.403] and TENNEKES
[2.404] to mention but a few). The best agreement with measurements
over a wide range of Mach numbers, heat transfer ratios and even
pressure gradients (GRAN et al. [2.405]) has been obtained if U*/u, is
plotted against y u J v w, where U* is the mean velocity transformed ac-
cording to VAN DRIEST [2.396]. It is, of course, important to heed the
validity range of (2.48) and (2.49) which are used in the transformation,
but the log-law is probably relatively insensitive to adverse pressure
gradients, as in incompressible flow. Deviations from the log-law must,
however, be expected if the upstream history of the boundary layer--be it
a temperature history (e.g., FELLER[2.445]) or the uncompleted transition
process (FERNHOLZ [2.406J)--still affects the boundary layer. These
deviations must not be counted as a failure of the transformation (RoTTA
[2.4071 FERNHOLZ [2.408] and KEENERand HOPKINS [2.409]).
As for the log-law it has been advantageous to use van Driest's
transformation for the outer law, too (FERNrlOLZ [-2.408]). This relation
96 H.-H. FERNHOLZ

is valid for zero pressure gradient and constant wall temperature:

U,*- U*
- - - - 4.70 ln(y/A*) - 6.74 (2.51)
U~r

where quantities denoted by an asterisk are transformed according to

VAN DRIEST. It will be noted that the Rotta-Clauser thickness A* (2.9)
was used instead of the boundary layer thickness 6. Due to the slower
asymptotic approach of the velocity distribution in the boundary layer
to the free-stream conditions, the evaluation of 6 is even more liable to
the personal touch than in incompressible boundary layers and therefore
not a useful length scale. Coles' wake-law was extended to compressible
boundary layers by MArHBWS et ai. [2.410] and GRAY et al. [2.405].
Since we have just advocated van Driest's transformation, it is ap-
propriate to mention a series of research papers which endeavor to find
a transformation allowing a point-to-point mapping between a com-
pressible and an incompressible turbulent boundary layer. This is the
transformation concept of COLES [2.67] for adiabatic boundary layers
which was extended by CROCCO [2.411] and LEWXSet al. [2.412] to
boundary layers with pressure gradient and heat transfer. The develop-
ment of this transformation theory has not yet been completed.
The majority of compressible two-dimensional boundary layers
that have been investigated consists of zero-pressure-gradient flows
along adiabatic or cooled walls (hydraulically smooth and rough)
generated on surfaces such as contoured plane and axisymmetric
nozzles, tunnel side walls, flat plates and axisymmetric bodies. A survey
of most of the unclassified experiments is due to be published by FERN-
HOLZ and FINLEY [2.413]. No experiments have become known so far
where the temperature gradient varied along the wall (there exists one
experiment with a step change of the wall temperature by GRAN et al.
[2.405]). Such a variable wall temperature T (x) is expected to exert a
strong influence on the temperature distribution across the boundary
layer.
The determination of the skin friction---even for zero pressure
gradient flow--still causes problems. The reliability and comparability
of floating-element balances and the choice of the correct calibration
curve for Preston tubes may serve as examples. These problems are
aggravated, of course, in boundary layers with pressure gradients,
and it is not Surprising to find only a few experiments where mean
velocity profiles and values of and skin friction have been measured
(e.g., PEAKEet al. [2.414], LEWtSet al. [2.415], VOISINEXand LeE [2.416],
WALa'RtJP and SCHETZ [2.417] and ZWARTS [2.418]). Though none of
 External Flows 97

the four adverse pressure gradient flows leads to separation, it appears

unusual at first sight that the skin friction can rise in an adverse pressure
gradient. One possible reason for this is that the Mach number influence
on the skin friction is stronger than that of the pressure gradient. For
zero-pressure-gradient boundary layers an increase in Mach number
reduces the skin friction considerably, especially at Reynolds numbers
Re62<5000 (see for example SPALDING and CHI [2.419] or FERNHOLZ
[2.420] and for an evaluation of several other theories HOPKINS and
INOUYE [2.421] and CARY and BERTRAM[2.422]).
Deviations between calculations of the characteristic flow param-
eters (especially for el) and measurements in boundary layers with
variable pressure gradient found by BRADSHAW[2.423] were attributed
to a direct influence of the mean dilatational rate of strain or compression
div U on the turbulence structure. This hypothesis is based on fluctation
measurements by BEH~NS [2.424] and ROSE [2.395] in oblique shock/
boundary layer interactions and by LEwis and BErt~NS [2.425] in the
shear layer emerging from a Prandtl-Meyer expansion. These measure-
ments have shown unexpectedly large increases in turbulence intensity
during compression and large decreases during expansion.
Another flow configuration especially important for air intakes of
jet engines is the boundary layer where the adverse pressure gradient
is generated by means of a ramp or compression corner (see for example
STROUD and MILLER [2.426] or STUREK and DANBERG [2.427]). Boun-
dary layers over curved surfaces are subject to pressure gradients normal
to the wall, the influence of which was investigated experimentally by
THOMAr,rN [2.198] and theoretically by ROTTA [2.428]. As for com-
pressible turbulent boundary layers with separation the reader is referred
to a few of the more recent experimental investigations (THOMKE and
ROSHKO [2.429], BATHAM [2.430], ELFSTROM [2.431], COLEMAN and
STOLLERY [2.432], ROSE et al. [2.433] and LAW [2.434]).
There exists a sequence of research papers on compressible flow
along rough walls which were performed at the University of Texas.
Since it is difficult to obtain these reports, details may be taken from
FERNHOLZ and FINLEY [2.413], and for another characteristic paper
see CHEN [2.435].
For a discussion of transition phenomena in compressible boundary
layers the reader is referred to the survey papers by MORKOVIN [2.235]
and KISTLER[2.436] and to some more recent experimental investigations
(MADDALOrq [2.437], LA GRAFT [2.438], OWEN [2.439], CAR'," [2.440],
BECKWITH and BERTRAM [2.441], WATSON et al. [2.442], BECKWITH
[2.443] and KENDALL [2.444]).
98 H.-H. FERNHOLZ

References
2.1 G.N.ABRAMOVICH: The Theory of Turbulent Jets. (MIT Press, Cambridge, Mass,
1963)
2.2 S.F.BIRCH, D.H.RuDv, D.M.BuSHNELL (Eds.): NASA SP 321 (1972) (2 vols.)
2.3 S.N.B. MuRTHY (Ed.): Turbulent Mixing in Non-Reactive and Reactive Flows.
(Plenum Press, New York 1975)
2.4 C. DuP. DoNALDSON, A.J. BILANIN: AGARDograph 204 (1975)
2.5 F.H. CLAUSER: The turbulent boundary layer, in: Advances in Applied Mechanics 1V.
(Academic Press, New York 1956)
2.6 I. TANl: Proceedings, Computation of Turbulent Boundary Layers--1968 AFOSR-
IFP-Stanford Conference, S.J.KLINE et al. (Eds.) (Thermosciences Div., Stanford
Univ. 1969), Vol. I, p. 483
2.7 D.E.CoLES, E.A.HIRST (Eds.): Proceedinys, Computation of Turbulent Boundary
Layers--1968 AFOSR-IFP-Stanford Conference, Vol. II. (Thermosciences Divn.,
Stanford Univ. 1969)
2.8 L.TING: J. Math. Phys. 44, 353 (1965)
2.9 J.E. GREEN : RAE TR 72050 (1972)
2.10 L. PRANDTL: Ergebnisse der Aerodynamischen Versuchsanstalt zu G6ttingen (Olden-
bourg, Miinchen-Berlin 1927), vol. III, p. 1
See also W.TOLLMIEN, H. SCHLICHTING, H. GORTLER (Eds.): Ludwig Prandtl Ge-
sammelte Abhandlunqen. (Springer, Berlin-Gfttingen-Heidelberg 1961). For all
Prandtl references
2.11 S.J. KLINE, W.C. REYNOLDS, F.A. SCHRAUB, P.W. RuNSTADLER: J. Fluid Mech. 30,
741 (1967)
2.12 L. PRANDTL: Ergebnisse der Aerodynamischen Versuchsanstalt zu G6ttingen (Olden-
bourg, Miinchen-Berlin 1932), vol. IV, p. 18
2.13 H.LuDWIEG, W.TILLMANN: Ingr.-Arch. 17, 288 (1949), translated as NACA TM
1285
2.14 TH. YON KARMAN: NACA TM 611 (1931)
2.15 F.H.CLAUSER: J. Aeronaut. Sci. 21, 91 (1954)
2.16 D. CoLES: J. Fluid Mech. 1,191 (1956)
2.17 H.ECKELMANN:J. Fluid Mech. 65, 439 (1974)
2.18 H.ECKELMANN: Mitt. No. 48, Max-Planck-lnst. f/ir Str6mungsforschung (1970)
2.19 H.P. BAKEWELL: Thesis, Dept. Aerospace Sci., Penn. State Univ. (1966); see also
Phys. Fluids 10, 1880 (1967)
2.20 S.J. KLINE, P.W. RUNSTADLER: Trans. ASME 81E, 166 (1959)
2.21 E.R. CORINO, R. S. BRODKEY: J. Fluid Mech. 37, 1 (1969)
2.22 H.Z. KIM, S. J. KLINE, W. C. REYNOLDS: J. Fluid Mech. 50, 133 (1971)
2.23 A.K. GUPTA, J. LAUEER, R. E. KAPLAN: J. Fluid Mech. 50, 493 (1971)
2.24 J.M.WALLACE, H.ECKELMANN, R.S. BRODKEY: J. Fluid Mech. 54, 39 (1972)
2.25 G.R. OFFEN, S.J. KLINE: J. Fluid Mech. 62~ 223 (1974)
2.26 H. UEDA, J.O. HINZE: J. Fluid Mech. 67, 125 (1975)
2.27 S. CORRSIN: Naval Hydrodynamics, Publication 515, National Academy of Sciences
- - National Research Council (1957), Chapt. 15
2.28 J. LAUEER: NACA Rept. 1174 (1955)
2.29 P.S. KLEBANOFF: NACA Rept. 1247 (1955)
2.30 W.W. WILLMARTH, S.S. LU: AGARD-CP-93, 3.1 (1972)
2.31 R.F. BEACKWELDER, R.E. KAPLAN: AGARD-CP-93, 5.1 (1972)
2.32 J.ROTTA: Prog. Aeronaut. Sci. 2, 1 (1962)
2.33 J.H.PR~STON: J. Roy. Aeronaut, Soc. 58, 109 (1954)
2.34 M.R. HEAD, I. RECHENBERG: J. Fluid Mech. 14, 1 (1962)
 External Flows 99

2.35 V.C. PATEL, M.R. HEAD: J. Fluid Mech. 34, 371 (1968)
2.36 V.C. PATEL: J. Fluid Mech. 23, 185 (1965)
2.37 G.L.MELLOR: J. Fluid Mech. 24, 255 (1966)
2.38 H. MCDONALD: J. Fluid Mech. 35, 311 (1969)
2.39 D.E.CoLES: AGARD-CP-93, 25.1 (1972)
2.40 R.J. BAKER, B.E. LAUNDER: Intern. J. Heat Mass Transfer 17, 275 and 293 (1974)
2.41 M.LANDAHL: J. Fluid Mech. 29, 441 (1967)
2.42 F.H. BARK: J. Fluid Mech. 70, 229 (1975)
2.43 O.M. PHILEIPS: J. Fluid Mech. 27, 131 (1967)
2.44 H.K. MOFFATT: Proc. AFOSR-IFP Stanford Conf. 1,495 (1969)
2.45 M.J.LIGHTHILL: Proc. AFOSR-IFP Stanford Conf. 1,511 (1969)
2.46 A . K . M . F . HUSSAIN, W.C. REYNOLDS: J. Fluid Mech. 41,241 (1970)
2.47 A . K . M . F . HUSSAIN,W. C. REYNOLDS: J. Fluid Mech. 54, 241 (1972)
2.48 L.S.G. KOVASZNAY: AGARD-CP-93, D1 (1972)
2.49 S.CORRSIN: NACA WR W-94 (1943)
2.50 A.A. ToWNSEND: Aust. J. Sci. Res. Set. A t , 161 (1948)
2.51 H. FIEDLER, M.R. HEAD: J. Fluid Mech. 25, 719 (1966)
2.52 R.E.KAPLAN, J. LAUFER: Proc. 12th Int. Congr. AppL Mech. (Springer, Berlin-
Heidelberg-New York 1969), p. 236
2.53 R.A.ANTONIA: J. Fluid Mech. 56, 1 (1972)
2.54 C.W. VANATTA: T.N. F-54, Mech. Engng. Dept., Sydney Univ. (1973)
2.55 J.C. LARuE: Phys. Fluids 17, 1513 (1974)
2.56 R.A. ANTONIA, J. D. ATgINSON : J. Fluid Mech. 64, 679 (1974)
2.57 P. BRADSHAW, J. MURLIS: Aero Rept. 74-04, Imperial College, London (1974)
2.58 R.F. BLACKWELDER, L.S.G. KOVASZNAY: Phys. Fluids 15, 1545 (1972)
2.59 O.M.PHILLIPS: J. Fluid Mech. 51, 97 (1972)
2.60 M.R. HEAD:ARC R. and M. 3152 (1960)
2.61 M.R. HEAD, V. C. PATEL: ARC R. and M. 3643 (1969)
2.62 A.A. TowNSEND: J. Fluid Mech. 26, 689 (1966)
2.63 R.E.FALco: AIAA Paper 74-99 (1974)
2.64 J. ROTTA: Ingr.-Arch. 19, 31 (1951)
2.65 J.C.RoTTA:FluidMechanicsoflnternalFlow(G.SOVRAN, Ed.)(Elsevier, Amsterdam
1967)
2.66 B.G.J.THOMPSON: ARC R. and M. 3463 (1967)
2.67 D.E.COLES: Rand Corp. R-403-PR (1962)
2.68 J.H.PRESTON: J. Fluid Mech. 3, 373 (1958)
2.69 R.L. SIMPSON: J. Fluid Mech. 42, 769 (1970)
2.70 G.D. HUFFMAN, P. BRADSI-IAW:J. Fluid Mech. 53, 45 (1972)
2.71 J.MuRus: PhD thesis, Imperial College, London (1975)
2.72 V.M.FALKNER, S.W. SKAN: Phil. Mag. 12, 865 (1931)
2.73 L.G.LoITSIANSKI: Laminare Grenzschichten (Akademie-Verlag, Berlin 1976)
2.74 A.A. ToWNSEND: J. Fluid Mech. 22, 773 (1965)
2.75 J.C. RoTTA: J. Aeronaut. Sci. 22, 215 (1955)
2.76 A.A.TowNSEND: J. Fluid Mech. 1,561 (1956)
2.77 A.A.ToWNSEND: J. Fluid Mech. 23, 767 (1965)
2.78 G.L. MELLOR, D. M. GmsoN: J. Fluid Mech. 24, 225 (1966)
2.79 J.F. NASH: AGARDogmph 97, 245 (1965)
2.80 H.J.HERRING, J.F. NORBURY: J. Fluid Mech. 27, 541 (1967)
2.81 B.E. LAUNDER, H.S. STINCHCOMBE: Rept. TWF/TN/21, Dept. Mech. Engng.,
Imperial College, London (1967)
2.82 P. BRADSHAW, D. H.FERRISS : NPL Aero Rept. 1145 (1965)
2.83 P. BRADSHAW: J. Fluid Mech. 29, 625 (1967)
100 H.-H. FERNHOLZ

2.84 T.N. STEVENSON: College of Aeronautics Rept. Aero 170 (1963)

2.85 P.S.KLEBANO~, Z.W. DIEHL: NACA Rept. 1110 (1952)
2.86 W. JACOBS: ZAMM 19, 87 (1939) and NACA TM 951 (1940)
2.87 R.A. ANTONIA, R. E. LUXTON: J. Fluid Mech. 53, 737 (1972)
2.88 R.L.GRAY, J.E. LEWIS, T. KUBOTA: J. Fluid Mech. 66, 507 (1974)
2.89 D.F.GATES: TR-73-152, U.S. Naval Ordnance Lab. (1973)
2.90 R.E.LEE, W.I. YANTA, A.C.LEONAS: Proc. 1968 Heat Transfer. (Fluid Mech.
Institute, Stanford : University Press 1968)
2.91 J.E. GREE~ : RAE TR 72201 (1972)
2.92 G. CrtARNAY, G.CoMTE BELLOT, J. MATHmU: AGARD-CP-93, 27.1 (1972) and
G. CHARNAY: P h D thesis, Ecole Centrale de Lyon (1974)
2.93 J. KESTIN: Advanc. Heat Transf. 3, 1 (1966)
2.94 G.D. HUrFMAN, D. R. ZIMMERMAN,W. A. BENNETT: AGARDograph 164 (1972)
2.95 S.J. KLINE, A.V. LISIN, B.A. WAITMAN: NASA TN D-368 (1960)
2.96 M.PJCHAL: Z A M M 52, T 407 (1972)
2.97 D.W. KEARNEY, W.M. KAYS, R.J. MOFFAT, R.J. LoYD: NASA CR 11 3590 (1969),
Rept. HMT-9 Dept. Mech. Engng., Stanford Univ. (t970)
2,98 H.KOMODA: J. Jap. Soc. Aeronaut. Eng. 1957, 274
2.99 J.D.VAGT: DFG-Bericht zu Schade 154/1 (1968/69)
2.100 R.J. LOEHRKE, FI. M. NAGIB: AGARD Rept. R-598 (1973)
2.101 P. BRADSHAW: ARC R. and M. 3575 (1969)
2.102 P.GoLDaEgG: Gas Turbine Lab. Rept. 85, MIT (1966)
2.103 A.E. SAMUEL, P. N. JOUBERT: J. Fluid Mech. 66, 481 (1974)
2.104 W.G. SPANGENBERG,W. R. ROWLAND, N. E. MEASE: In : Fluid Mechanics of Internal
Flow, G. SOVRAN(Ed.). (Elsevier, Amsterdam 1967), p. 110
2.105 V.A.SANDBORN, C.Y. LIu: J. Fluid Mech. 32, 293 (1968)
2.106 G. B. SCHUBAUER, P. S. KLEBANOFF" NACA Rept. 1030 (1951)
2.107 B.G.NEWMAN: Australian Dept. Supply Rept. ACA-53 (1951)
2.108 B.E. LAUNDER: Trans. ASME 31E, 707 (1964)
2.109 P.K. CHANG: Separation of Flow. (Pergamon Press, Oxford 1970)
2.110 A.ROSHKO: Canadian Congr. Appl. Mech. (Univ. Laval Press, Quebec 1967)
2.111 S. KAPLUN: Fluid Mechanics and Singular Perturbations. (Academic Press, New
York 1967)
2.112 S.N. BROWN, K.STEWARTSON: Ann. Rev. Fluid Mech. 1, 45 (1969)
2.113 L. PRANDTL: Verh. 111. Int. Math. Kongr., Heidelberg (Teubner, Leipzig 1905)p. 484;
(see also Gesammelte Abhandlungen [2.10])
2.114 K.OswATITSCH: Grenzschichtforschung, (Springer, Berlin-G6ttingen-Heidelberg
1958), p. 357
2.115 K.STEWARTSON: Quart. J. Mech. Appl. Math. 11,399 (1958)
2.116 K.STEWARTSON: Phys. Fluids 12, II 282 (1969)
2.117 P. BRADSHAW:AGARDograph 169 (1973)
2.118 A.A.TowNSEND: J. Fluid Mech. 12, 536 (1962)
2.119 S.J. KL1NE: Trans. ASME glD, 305 (1959)
2.120 V. A. SANDBORN, S.J. KLINE: Trans. ASME 83D, 317 (1961)
2.121 D.G. MABEV: Aeronaut. J. 77, 201 (1973)
2.122 B.S. STRATFORD: J. Fluid Mech. 5, 17 (1959)
2.123 H. FERNHOLZ, P. GIaSON: Gas Turbine Lab. Rept. No. 91, MIT (1967)
2.124 A.A.ToWNSEND: J. Fluid Mech. 8, 143 (1960)
2.125 H. FERNHOLZ: Ingr.-Arch. 35, 192 (1966)
2.126 I,E.ALBER: AIAA Paper, 71-203 (1971)
2.127 P. BRADSHAW,P. V. GALEA: J. Fluid Mech. 27, 111 (1967)
2.128 K.GERSTEN : AGARD-CP-4, 863 (1966)
 External Flows 101

2.129 G. LACHMANN (Ed.): Boundary Layer and Flow Control. London: Pergamon Press
1961
2.130 B. M. JONES: J. Roy. Aeronaut. Soc. 38, 753 (1934)
2.131 A.D. YOUNG, H. P. HORTON : AGARD-CP-4, 779 (1966)
2.132 H.P. HORTON: PhD thesis, Queen Mary College, London (1968)
2.133 H.P. HORTON : Aeronaut, Quart. 23, 211 (1972)
2.134 H.T. DELPAK: PhD thesis, Queen Mary College, London (1973)
2.135 I.TAut: Progr, Aeronaut. Sci. 5, 70. London: Pergamon Press 1964
2.136 M. GASTER: ARC R, and M. 3595 (1967)
2.137 W.R.BRILEY, H. McDoNALD: J. Fluid Mech. 69, 631 (1975)
2.138 H.J.LuGT, H.J.HAUSSL1NG: Rept. 3748, Naval Ship Res. Devel. Center (1972)
2.139 U.B. MEHTA, Z. LAVAN: J. Fluid Mech. 67,227 (1975)
2.140 I. McGREGOR: PhD thesis, Queen Mary College, London (1954)
2.141 E.A. EICHELaRENNER: Jahrbuch der WGL I959, (Vieweg, Braunschweig 1960),
p. 119
2.142 E.A. EICHELBRENNER: New York Acad, Sci. Ann. 154, 655 (1968)
2.143 D.R.CHAPMAN: NACA TN-2t37 (1950)
2.144 H.H. KoRsT: Trans. ASME 78, 593 (1956)
2,145 L.CRocCo, L. LEES: 3. Aeronaut. Sci. 19, 649 (1952)
2.146 J.H. SMITH, J. P. LAMB: Intern. J. Heat Mass Transl. 17, 1571 0974)
2.147 P. BRADSl-IAW,F. Y. F. WONG : J. Fluid Mech. 52, 113 (1972)
2.148 D.E.ABBOTT, S.J.KLINE: Trans. ASME 84D, 317 (1962)
2.149 M.ARIE, H. ROUSE: J. Fluid Mech. I, 129 (1956)
2.150 T.J. MOELLER, J. M. ROBERTSON: Modem Develop. Theor. Appl. Mech. 1,326 (1963)
2.151 J.C. LE BALLEUR, J. MIRANDE: ONERA T.P. No. 1975-16(1975)
2.152 T.J. MUELLER, H.H.KoRST, W.L. CHOW: Trans. ASME 86D, 221 (1964)
2.153 J. NIKURADSE: VDI Forschungshefl 361 (1933)
2.154 D.H.WooD, R.A.ANTON1A: Aeronaut. Quart. 26, 202 (1975)
2.155 R.L. SIMPSON: AIAA J. I1,242 (1973)
2.156 L.PRANDTL: Ergebnisse der AVA Gbttingen 4 (Oldenbourg, Miinchen 1932) (see also
Gesammelte Abhandlungen I-2.10])
2.157 H.SCHLICHT1NG: Ingr.-Arch. 7, 1 (1936) and NACA TM 823
2.158 F.R. HAMA: Trans. Soc. Nay. Architects Marine Engrs. 62, 333 (1954)
2.159 V.A. IOSELEVICH,V.I.PILIPENKO: Sov. Phys. Dokl. 213, 1266 (1973)
2.160 A.E. PERRY, W. H. SCHOF1ELD, P. N. JOUBERa': J. Fluid Mech. 37,383 (1969)
2.161 C.K.L1u, S.J. KLINE, J.P.JOHNSTON: Rept. MD-15, Mech. Dept. Engng., Stanford
Univ. (1966)
2.162 H. W. TOWNES, R.H. SABERSKY; Intern. J. Heat Mass Transf. 9, 729 (1966)
2.163 R.A.ANTONIA, D.H. WooD: TN F-71, Mech. Engng. Dept., Sydney Univ. (1974)
2.164 A.J.GRASS: J. Fluid Mech. 50, 233 (1971)
2.65 A.E.PERRY, P,N.JoUBERT: J. Fluid Mech, 17, 193 (1963)
2.166 R.A.ANTONIA, R. E. LUXTON: TN F-I, Mech. Engng. Dept., Sydney Univ. (1969)
2,167 R.A.ANTONIA, R.E. LUXTON: Advan. Geophysics 18A, 263 (1974)
2.168 R.E.LuxTON: TN F-12, Mech. Engng. Dept., Sydney Univ. (1970)
2.169 A.A. TowNsEND: J. Fluid Mech. 22, 799 (1965)
2.170 A.A.ToWNSEND: J. Fluid Mech. 26, 255 (1966)
2.171 R.J. TAYLOR: J. Fluid Mech. 13, 529 (1962)
2.172 K.S. RAO, J. C. WYNGAARD, O. R. COTr~: J. Atmosph. Sci. 31,738 (1974)
2.173 R.A. ANTONIA, R. E. LUXTON; Phys. Fluids 14, 1027 (1971)
2.174 L.GAUDET, K.G. WINTER: RAE TM Aero 1538 (1973)
2.175 K.H.ROGERS: J. Aircraft 11,382 (1974)
2.176 G.R. INGER, E.P. WILLIAMS: AIAA J. 10, 636 (1972)
102 H.-H. FERNHOLZ

2.177 G.D. ASHTON: Proc. 1972 Inst. Heat Mass Transfer (1972)
2.178 P.R. OWEN,W.R. THOMSON: J. Fluid Mech. 15, 321 (1963)
2.179 R.T.Ho, L.W.GELHAR; J. Fluid Mech. 58, 403 (1973)
2.180 S.T.Hsu, J.F. KENNEDY: J. Fluid Mech. 47, 481 (1971)
2,181 S. KOTAKE: Intern. J. Heat Mass Transl. 17, 885 (1974)
2.182 K.G. WINTER, J.C. ROTTA, K.G. SMXTH:ARC R. and M. 3633 (1970)
2.183 R. RICHMOND: PhD thesis, Caltech, Pasadena (1957)
2.184 A. S. GINEVSKII,E. E. SOLODKIN:J. Appl. Math. Mech. 22, 1169 (1958)
2.185 W.W. WILLMARTH,C. S. YANG: J. Fluid Mech. 41, 47 (1970)
2.186 P. BRADSHAW,V. C. PATEL: AIAA J. 11,893 (1973)
2.187 A. WALZ, M. MAYER: Glastechn. Ber. 39, 359 and 409 (1966)
2.188 V.C. PATEL, A. NAKAYAMA,R. DAMIAN: J. Fluid Mech. 63, 345 (1974)
2.189 L. PRANDTL: NACA TM-625 (1929)
2.190 G.I.TAYLOR: Phil. Trans. Roy. Soc. A 223, 289 (1923)
2.191 H.GtSRTLER: Nachr. dr. Wiss. Ges. G6ttingen, Math. Phys. 1, 1 (1940)
2.192 I.TANI: J. Geophys. Sci. 67, 3075 (1962)
2.193 P. BRADSHAW:J. Fluid Mech. 36, 177 (1969)
2.194 R.M.C. So, G. L. MELLOR: J. Fluid Mech. 60, 43 (1973)
2.195 R.M.C. So, G. L. MELLOR: Aeronaut. Quart. 26, 25 (1975)
2.196 B.G.J. THOMPSON: AGARDograph 97, 159 (1965)
2.197 R.N. MERONEY, P. BRADSHAW:AIAA J. 13, 1448 (1975)
2.198 H.THOMANN: J. Fluid Mech. 33, 283 (1968)
2.199 V.C. PATEL: ARC R. and M. 3599 (1969)
2.200 B.G. NEWMAN: Canadian Aero. Space J. 15, 288 (1969)
2.201 H.H.FERNHOLZ: DLR FB 66-21 (1966)
2.202 H.H.FERNHOLZ: Z. Flugwiss. 15, 136 (1967)
2.203 F.A. DvORAK: AIAA J. 11,517 (1973)
2.204 H.P.A.H. IRWIN, P. Arnot SMITH: Phys. Fluids !8, 624 (1975)
2.205 J.C. ROTTA: W~rrne Stofffibertragung 7, 133 (1974)
2.206 H.U. MEIER, D. F. GATES, R. L. P. VOISINET: AIAA Paper 74-596 (1974)
2.207 A.I. LEONT'EV: Advances in Heat Transfer (Academic Press, New York 1966),
vol. 3, p. 33
2.208 S.V. PATANKAR, D.B. SPALDING: Heat and Mass Transfer in Boundary Lasers
(Morgan-Grampian, London 1967)
2.209 A. D. GOSMAN,W.M. PUN, A. K. RUNCHAL,D. B. SPALDING,M. WOLFSHSTEIN: Heat
and Mass Transfer, in Recirculating Flows (Academic Press, New York 1968)
2.210 T. CEBECI, A. M. O. SMITH, G. MOSINSKIS:Trans. ASME 92C, 133 (1970)
2.211 R.J. FLAHERTY:J. Aircraft 11,293 (1974)
2.212 M.E. CRAWFORD,W. M. KAYS: Rept. HMT-23, Thermosciences Div., Mech. Engng.
Dept., Stanford Univ. (1975)
2.213 W.J. KELNHOFER:DFVLR-FB 70-66 (1970) (see also Trans. ASME 91A, 281 (1969)
2.214 L.H. BACK, F. CUFFEL, P. F. MASSIER: Intern. J. Heat Mass Transf. 13, 1029 (1970)
2.215 D.R. BOLDMAN,J. F. SCHMIDT, R. C. EHLERS: Trans. ASME 89C, 341 (1967)
2.216 L.H. BACK, R. F. CUEFEL:Trans. ASME 93C, 397 (1971)
2.217 D.W. KEARNEY,W. M. KAYS,R. J. MOFFAT:Int. J. Heat Mass Transl. 16, 1289 (1973)
2.218 A.P.HATTON, V.A.EuSTACE: Proc. 3rd Int. Heat Transfer Conf. Chicago 2, 34
(1966)
2.219 W.H.THIELBAHR, W.M.KAYS, R.J. MoFFAT: Rept. HMT-5, Dept. Mech. Engng.,
Stanford Univ. (1969)
2.220 F.A. DVORAK, M. R. HEAD: Intern. J. Heat Mass Transl. 10, 61 (1967)
2.221 R.E.CmLCOTT: Intern. J. Heat Mass Transf. 10, 783 (1967)
2.222 L. S. FLETCHER, D. G. BRIGGS, R. H. PAGE: AIAA Paper 70-767 (1970)
 External Flows 103

2.223 C. B. COHEN. E. RESHOTKO: N A C A R 1293 (1956)

2.224 W. DIENEMANN: Z A M M 33, 89 (1953)
2.225 W. C. REYNOLDS, W. M. KAYS, S. J. KLINE : NASA Memo 12-3-58 W (1958)
2.226 P. M. MO~TTI, W. M. KAYS: Intern. J. Heat Mass Transf. 8, 1187 (1965)
2.227 R. G. DEISSLER: Trans. ASME 76, 73 (1954)
2.228 B. A. KADER, A. M. YAGLOM: Intern. J. Heat Mass Transl. 15, 2329 (972)
2.229 P. BRADSHAW: AIAA J. 8, 1375 (1970)
2.230 H. FERNHOLZ : D L R - F B 72-26, 135 (1972)
2.231 R, L. MEEK, A. D, BAER: Intern. J. Heat Mass Transl. 16, 1385 (1973)
2.232 I. E. ALBER, D. E. COATS: AIAA J. 8, 791 (1971)
2.233 J. KESTIN, P.D. RICHARDSON: Forsch. Gebiete Ingenieurw. 29, 93 (1963)
2.234 D. B. SPALDING : Convective Mass Transfer. London : Edward Arnold 1963
2.235 M. V. MORKOVIN" TR-68-149, A F F D L , Wright-Patterson AFB (1969)
2.236 E. RESHOTKO: A I A A Paper 74-130 (1974)
2.237 L. M. MACK: A G A R D o g r a p h 97 (1965)
2.238 L. M. MACK : A I A A Paper 74-134 (1974)
2.239 A . D . D . C R A m : J. Fluid Mech. 50, 393 (1971)
2.240 M.GASrER: J. Fluid Mech. 66, 465 (1974)
2.241 J.T.STUART: NPL Aero Rept. 1147 (1965)
2.242 E. MOLLO-CHRISTENSEN: A I A A J. 9, 1217 (1971)
2.243 G. B. SCHUBAUER, P. S. KLEaANOFF: N A C A Rept. 1289 (1956)
2,244 P. S. KLEBANOEF, K. D. TIDSTROM, L. M. SARGENT: J. Fluid Mech. 12, 1 (1962)
2.245 C.F.KNAPP, P.J.ROACHE: A I A A J. 6, 29 (1968)
2.246 A. M.O. SMITH, N. GAMBERONI: Rept. ES 26388, Douglas Aircraft Co. (1956)
2.247 N.A.JAEEE, T.T.OKAMURA, A . M . O . SMITH: AIAA J. 8, 301 (1970)
2.248 I.TANI: Ann. Rev. Fluid Mech. 1,169 (1969)
2.249 N. S. DOUOHERTY : AIAA Paper 74-627 (1974)
2.250 S.R.PATE: A I A A J. 9, 1082 (1971)
2.251 S.R.PATE: A I A A J. 12, 1615 (1974)
2.252 J. E. LA GRAFF : AIAA J. 10, 762 (1972)
2.253 F. K. OWEN, C.C. HORSTMAN: A IAA J. 10, 769 (1972)
2.254 P. C. STAINBACK, J. B. ANDE~ : AIAA Paper 74-136 (1974)
2.255 J. E. HARRIS: N A S A TR-R-368 (1971)
2.256 H. McDONALD, R. W. FtSH: Intern. J. Heat Mass Transf. 16, 1729 (1973)
2.257 D. M. BUSHNELL, D.W. ALSTON : A I A A J. 11,554 (1973)
2.258 A . M . O . SMITH: J. Fluid Mech. 7, 565 (1960)
2.259 I. J. WYGNANSKI,F. H. CHAMPAGNE" J. Fluid Mech. 59, 281 (1973)
2.260 I. WYGNANSKI~ M. SOKOLOV, D. FRIEDMAN: J. Fluid Mech. 69, 283 (1975)
2.261 R. W. MIKSAD : J. Fluid Mech. 56, 695 (1972)
2.262 M. SIBULKIN: Phys. Fluids 5, 282 (1962)
2.263 J.LAUEER: Miszellaneen der Angewandten Mechanik (Akademie-Verlag, Berlin 1962),
p. 166
2.264 D . M . MCELIGOT, C. W. COON, H.C. PERKINS: Intern. J. Heat Mass Transf.13, 431
(1970)
2.265 M.A. BADRI NARAYANAN: J. Fluid Mech. 31,609 (1968)
2.266 M.A. BADRI NARAYANAN,V. RAMJEE: J. Fluid Mech. 35, 225 (1969)
2.267 H.P. HORTON : AGARD-LS-43-71 (1971)
2.268 R.F. BLACKWELDER,L. S. G. KOVASZNAY: J. Fluid Mech. 53, 61 (1972)
2.269 P. BRADSHAW: J. Fluid Mech. 35, 387 (1969)
2.270 R. NARASIMHA,K. R. Sr~ENIVASAN: J. Fluid Mech. 61,417 (1973)
2.271 B.E.LAUNDER, W . P . J o N ~ : J. Fluid Mech. 38, 817 (1969)
2.272 W.P.JoNES, B. E. LAUNDER: Intern. J. Heat Mass Transl. 15, 301 (1972)
104 H.-H. FERNHOLZ

2.273 R.L.O'BRIEN : United Aircraft Corp. Rept. (1964)

2.274 D.R. BOLDMAN, J. F. SCHMIDT, A. K. GALLAGHER: NASA TN D-4788 (1968)
2.275 L.H. BACK, R. F. CUEFEL, P. F. MASSlER: AIAA J. 7,730 (1969)
2.276 J.L. NASH-WEBBER, G. C. OATES: Trans. ASME 94D, 897 (1972)
2.277 K.R. PERKINS, D. M. MCELIGOT: Trans. ASME 97C, 589 (1975)
2.278 J.M. ROaERTSON, C. F. HOLT: ASCE J. Hydraulics Div., 98, H Y6, 1095 (1972)
2.279 F.A.SCHRAUB, S.J.KLINE: Rept. MD-12, Dept. Mech. Engng., Stanford Univ.
(1965)
2.280 R.A.ANTONIA, R.E. LuxToN: TN F-31, Mech. Engng. Dept., Sydney Univ. (1971)
2.281 P. BRADSHAW: Aero Rept. 74-10, Imperial College, London (1974)
2.282 H.FERNHOLZ: ARC R. and M. 3368 (1964)
2.283 J.M.KENDALL: J. Fluid Mech. 41,259 (1970)
2.284 R.E. DAvis: J.,Fluid Mech. 52, 287 (1972)
2.285 A. Y.-S. Kuo, S. CORRSJN: J. Fluid Mech. 50, 285 (1971)
2.286 E. MOLLO-CHRISTENSEN: Ann. Rev. Fluid Mech. 5, 101 (1973)
2.287 A. FAVRE, J.GAV1GLIO, R. DUMAS: Phys. Fluids 10, S 138 (1967)
2.288 L.PRANOTL: Festschrift zum 60. Geburtstage yon A. BETZ (G6ttingen 1945) p. 134
(see also Gesammelte Abhandlungen [2.10])
2.289 F.J.PIEgCE: Trans. ASME 88D, 101 (1966)
2.290 W.D.HAVES: NAVORD Rept. 1313 (1951)
2.291 L.C.SQUIRE: ARC R. and M. 3006 (1957)
2.292 J.V.RAKICH, G.G.MATEER: AIAA J. 10, 1538 (1972)
2.293 T. CEBECI, K. KAUPS, J. RAMSEY, A. MOSER: Proc. NASA Langley Conf. Aero
Analyses Requiring Advanced Computers (NASA SP-347, 1975) and Douglas
Aircraft Co. Rept. MDC J6866 (1975)
2.294 E. KRAUSE: Int. Congr. Aerospace Sci. Paper 74-20 (1974)
2.295 E.A. EICHELBRENNER: Ann. Rev. Fluid Mech. 5, 339 (1973)
2.296 F.G. BLOTTNER: AGARD-LS-73, 3-1 (1975)
2.297 F.B. GESSNER, J. B. JONES: Trans. ASME 83D, 657 (1961)
2.298 E.A. EICHELBRENNER: Rech. A6ro. 104, 3 (1965)
2.299 M. ZAMIR: Aeronaut. J, 74, 330 (1970)
2.300 M. ZAMIR, A. D. YOUNG: Aeronaut. Quart. 21,313 (1970)
2.301 H.J. PERKINS: J. Fluid Mech. 44, 721 (1970)
2.302 G. M. BRAGG: J. Fluid Mech. 36, 485 (1969)
2.303 E.A. EICHELBRENNER, K.T. NGUYEN: Compt. Rend. Acad. Sci. A 269, 869 (1969)
2.304 E.A. EICHELBRENNER: Rech. A6ro. 83 (1961)
2.305 E . A . EICHELBRENNER, J. H. PRESTON : J. M6canique 10, 91 (1971)
2.306 E.A.EIcrIELBRENNER, P. FLORENT, K.T. NGUVEN: Compt. Rend. Acad. Sci. Ser.
A Paris 274, 1063 (1972)
2.307 O.O. MoJOLA, A.D. YouNG: AGARD-CP-93, 12.1 (1972)
2.308 K.T. NGUYEN : MSc thesis, Universit6 Laval, Quebec (1968)
2.309 J.C.CoOKE, M.G. HALL: Prog. Aero. Sci. 2, 221 (1962)
2.310 P.N. JouBERT, A.E. PERRY, K.C. BRowN: In: Fluid Mechanics of Internal Flow,
SOVRAN,G. (Ed.) (Elsevier, Amsterdam 1967)
2.311 F.S.SHERMAN : Rept. RM-4843-PR, Rand Corp., Santa Monica (1968)
2.312 A.J. WHEELER, J. P. JOHNSTON: Trans. ASME 95I, 415 (1973)
2.313 A.J.WHEELER, J. P. JOHNSTON: Rept. MD-32, Dept. Mech. Engng., Stanford Univ.
(1972)
2.314 J.F. NASH, V.C. PATEL: Three-Dimensional Turbulent Boundary Layers (SBC
Technical Books, Atlanta 1972)
2.315 J.H.HoRLOCK, J.F.NoRBURY, J.C.CooKE: J. Fluid Mech. 27, 369 (1966)
2.316 H. FERNHOLZ: J. Fluid Mech. 58, 177 (1973)
 External Flows 105

2.317 E. A. EICHELBRENNER,A. OUDART: ONERA Pub. 76 (t955)

2.318 J.C.CoO~E: Aeronaut. Quart. 11,333 (1960)
2.319 N.A. CUMPST~, M. R. HEAD: Aeronaut. Quart. 21,121 (1970)
2.320 J.C.ADAMS: J. Spacecraft 12, 131 (1975)
2.321 E.A. EICHELBRENNER, J. L. PEUBE: Univ. Poitiers, Contract Number: N 62558-3863
(1966)
2.322 N.A. CUMPSTV, M. R. HEAD: Aeronaut. Quart. 18, 55 (1967)
2.323 G.S. RAETZ: Rept. NAI-58-73 (BLC-114), Northrop Aircraft, Inc. (1957) and ARC
paper 23634 (1962)
2.324 P. WESSELING: Rept. AT-69-01, NLR, Amsterdam (1969)
2.325 K.C. WANG: J. Fluid Mech. 48, 397 (1971)
2.326 F.J. PIERCE: Trans. ASME 86D, 227 (1963)
2.327 J.P. JOHNSTON:Trans. ASME 82D, 622 (1960)
2.328 G.L. MELLOR: AIAA J. 5, 1570 (1967)
2.329 M.R. HEAD, T. S. PRAHLAD: Aeronaut. Quart. 25, 293 (1974)
2.330 N.A. CUMPSTY,M. R. HEAD: Aeronaut. Quart. 18, 150 (1967)
2.331 N.A. CuMPSTV, M.R. HEAD: Aeronaut. Quart. 20, 99 (1969)
2.332 T. CEaECI: AIAA J. 12, 242 (1974)
2.333 L.F. EAST, R.P.HoxEY: ARC R. and M. 3653 (1971)
2.334 D.W. ETnER1DGE: PhD thesis, Queen Mary College, London (1971)
2.335 B. VANDEN BERG, A. ELSENAAR, J,P.F. LINDHOUT, P. WEssELING: J. Fluid Mech.
70, 127 (1975)
2.336 A. ELSENAAR, S.H. BOELSMA: TR-74095U, NLR, Amsterdam (1974)
2.337 L.F, EAsT: ARC R. and M. 3768 (1975)
2.338 W.F. KLINKSIEK, F. J. PIERCE:Trans. ASME 92D, 83 (1970)
2.339 A.J. VERMEULEN: PhD thesis, Univ. Cambridge (1971)
2.340 P. BANSOD, P. BRADSHAW: Aeronaut. Quart. 23, 131 (1972)
2.341 L. LARSSON: PhD thesis, Chalmers Univ., G6teborg (1975)
2.342 R.W. MOORE, D. L. RICHARDSON: Trans. ASME 79, 1789 (1957)
2.343 P. BRADSHAW, M.G.TERRELL: NPL Aero Rept. 1305 (1969)
2.344 R. CRAaBE: Rept. 71-2, Mech. Engng. Dept., McGill Univ. (1971)
2.345 A. MAGER: NACA R 1067 (1952)
2.346 J.P. JoHNSTON: Trans. ASME 82D, 233 (1960)
2.347 F.J.PIERCE: AIAA J. 10, 334 (1972)
2.348 T.S. PRAHLAD: AIAA J. 11,359 (1973)
2.349 J.L. BOUSGARBIES: thesis, Univ. Poitiers (1963)
2.350 E.A.EICHELBRENNER, J. L. PEUBE: Univ. Poitiers. Lab. Mech. Fluides (1966)
2.351 J.P. JOHNSTON: J.Fluid Mech. 42, 823 (1970)
2.352 P. BRADSHAW: J. Fluid Mech. 46, 417 (1971)
2.353 D.B. SPALDING: Paper presented at Euromech 60, Trondheim (1975)
2.354 B. VANDEN BERG: J. Fluid Mech. 70, 149 (1975)
2.355 L.F. EAST: RAE TR 72178 (1972)
2.356 H.ASHKENAS: NACA TN 4140 (1958)
2.357 F.J. PIERCE, D. KRoMMENHOEK: Interim Tech. Rept. 2, Dept. Mech. Engng.,
Virginia Poly. (1968)
2.358 T. PRAHLAD: AIAA J. 6, 1772 0968)
2.359 M.G. HALL, H. B. DICKENS: RAE TR 66214 (1966)
2.360 W.C. RAINB1RD: AIAA J. 6, 2410 (1968)
2.361 M.C. FISHER, L. M. WEINSTEIN: AIAA J. 12, 131 (1974)
2.362 M.C. FISHER, L. M. WEINSTE1N: AIAA Paper 73-635 (1973)
2.363 J. COUSTEIX, R. MICHEL: ONERA T.P. No. 1975-22 (1975)
2.364 O. PARR: Ingr.-Arch. 32, 393 (1963)
106 H.-H. FERNHOLZ

2.365 J. S~lNHEUER: AGARDograph 97, 567 (1965)

2.366 T.S. CHAM, M. R. HEAD: J. Fluid Mech. 42, 1 (1970)
2.367 J. BUCKMASTER:Phys. Fluids 15, 2106 (1972)
2.368 E.S.TAYLOR: In: SOVRAN,G. (Ed.): Fluid Mechanics of Internal Flow (Elsevier,
Amsterdam 1967), p. 320
2.369 J.H.B. SMITH : RAE Memo, 1620 (1975)
2.370 F.K. MOORE: Theory of Laminar Flows (Princeton Univ. Press, Princeton 1964)
2.371 I.E. BECIWITH, W.D. HARVEY, F. L. CLARK: NASA TN D-6192 ( 1971)
2.372 A. WALz: Str6mungs- und Temperatur#renzschichten.(Braun, Karlsruhe 1966)
2.373 L.S.G. KOVASZNAY:J. Aeronaut. Sci. 20, 657 (1953)
2.374 F.K.OwEN, C.C.HoRSTMAN, M.I.KussoY: J. Fluid Mech. 70, 393 (1975)
2.375 L.S.G. KOVASZNAV:J. Aeronaut. Sci. 17, 565 (1950)
2.376 M.V. MORKOVIN: AGARDograph 24 (1956)
2.377 M.V. MORKOVlN: Proc. Int. Syrup. Hot-Wire Anemometry, Univ. Maryland (1967)
pp. 38--51
2.378 A. DEMETmADES, A. J. LADERMAN: SAMSO Rept. 73-129 (1973)
2.379 J.E. WALLACE: AIAA J. 7, 757 (1969)
2.380 W. D. HARVEY, D. M. BUSHNELL: NASA SP 216, 11 AI--11 A10 (1968)
2.381 S.S. PENNER: AIAA Paper 71-283 (1971)
2.382 W.J. YANTA, R. E. LEE: AIAA Paper 74-575 (1974)
2.383 A.L.KIsTLER: Phys. Fluids 2, 296 (1959)
2.384 M.V. MoRKOVtN: A G A R D Wind Tunnel and Model Testing PANEL, London
(1960)
2.385 P. BRADSHAW, D. H. FERRISS: J. Fluid Mech. 46, 83 (1971)
2.386 M.V. MoRKovIN: In: The Mechanics of Turbulence, FAVRE,A., (ed.) (New York:
Gordon and Breach 1964)
2.387 F.K.OwEN, C.C. HoRsTMAN: J. Fluid Mech. 53, 611 (1972)
2.388 W.C.RosE: AIAA J. 12, 1060 (1974)
2.389 G. MAISE, H. McDoNALD: AIAA J. 6, 73 (1968)
2.390 A.MARTELLUCCI, H. RIE, J. F.SoNTowsKt: AIAA Paper 69-688 (1969)
2.391 D.M. BUSHNELL, D.J. MORRIS: AIAA J. 9, 764 (1971)
2.392 H. U. MEIER, J. C. ROTTA: AIAA J. 9, 2149 (1971)
2.393 C.C.HORSTMAN, F.K.OwEN: AIAA J. 10, 1418 (1972)
2.394 A.J. LADERMAN,A. DEMETRIADES"J. Fluid Mech. 63, 121 (1974)
2.395 W.C.RosE: PhD thesis, Univ. of Washington (1972). See also NASA TN-D 7092
(1973)
2.396 E.R. VANDRIEST: J. Aeronaut, Sci. lg, 145 (1951)
2.397 H.U. MEIER, R. L. P. VO~S1NET,D. F. GATES: AIAA Paper 74-596 (1974)
2.398 A.M. CARY: NASA TN D-5560 (1970)
2.399 G.T. COLEMAN, C. OSBORNE,J. L. STOLLERY: J. Fluid Mech. 60, 257 (1973)
2.400 H.J. Ki3STER: Dissertation, Technische Universit~t Berlin (1972)
2.401 D.CoLES: J. Aeronaut. Sci. 21,433 (1954)
2.402 J.RorrA: Z. Flugwiss. 7,264 (1959)
2.403 P.O. BARONTI, P.A. LmBV: AIAA J. 4, 193 (1966)
2.404 H.TENNEKES: AIAA J. 5, 489 (1967)
2.405 R.L. GRAN, J. E. LEWIS, T. I(UBOTA: J. Fluid Mech. 66, 507 (1974)
2.406 H.FERNHOLZ: Fluid Dynamic 7?ansactions (Polish Academy of Sciences, Warsaw
1971), vol. 6, pt. 2, p. 161
2.407 J.C. ROTTA: Rept. 64 A 10, AVA-DFVLR, G6ttingen (1964)
2.408 H. FERNHOLZ: lngr.-Arch. 38, 311 (1969)
2.409 E.R. KEENER, E.J.HOPKINS: AIAA J. 11, 1784 (1973)
2.410 D.C. MATr~EWS, M.E. CmLDS, G.C. PAYNTER: J. Aircraft 7, 137 (1970)
 External Flows 107

2.411 L. CRoCCO: AIAA J. 1, 2723 (1963)

2.412 J. E. LEWIS, T. KUBOTA, W. H. WEBB: AIAA J. 8, 1644 (1970)
2.413 H. H. FERNHOLZ, P. J. FINLEY: To be published by A G A R D o g r a p h 223 (1977)
2.414 D. J. PEAKE, G. BRAKMAN, J. M. ROMmKIE : AGARD-CP-93, 11-1 (1972)
2.415 J.E. LEwlS, R. L. GRANT, T. KUBOTA: J. Fluid Mech. 51,657 (1972)
2.416 R. L. VOISINET, R. E. LEE: TR-73-224, U.S. Naval Ordnance Lab. (1973)
2.417 P. J. WALTRUP, J. A. SCHETZ : AIAA J. 11, 50 (1973)
2.418 F.ZWARTS: PhD thesis, McGill Univ. (1970)
2.419 D. B. SPALDING,S. W. CHI" J. Fluid Mech. 18, 117 (1964)
2.420 H. FERNHOLZ: Z A M M 51, T 146 (1971)
2.421 E.J. HoPKINS, M. INouYE: AIAA J. 9, 993 (1971)
2.422 A. M. CARY, M. H. BERTRAM : NASA TN-D-7507 (1974)
2.423 P. BRADSHAW: J. Fluid Mech. 63, 449 (1974)
2.424 W. BEHRENS: AIAA Paper 71-127 (1971)
2.425 J. E. LEWIS, W. BEHRENS: AIAA J. 7, 664 (1969)
2.426 J. F. STROUD, L.D. MILLER; Tech. Rept. AFFDL-TR-65-123 and J. Aircraft 3, 548
(1966)
2.427 W. B. STUREK,J. E. DANBERG: AIAA J. 10, 475 and 630 (1972)
2.428 J.C. ROTTA: Phys. Fluids 10, S 174 (1967)
2.429 G. J. THOMKE, A. ROSHKO: NASA CR-73308 (1969)
2.430 J.P. BATHAM: J. Fluid Mech. 52, 425 (1972)
2.431 G. M. ELFSTROM: J. Fluid Mech. 53, 113 (1972)
2.432 G. T. COLEMAN, J. L. STOLLERY: J. Fluid Mech. 56, 741 (1972)
2.433 W. C. ROSE, R. J. PAGE, M. E. CHILDS; A I A A J. 11,761 (1973)
2.434 C . H . L A w : AIAA J. 12, 794(1974)
2.435 K. K. CHEN: A I A A Paper 71-166 (1971)
2.436 A. L. KISTLER : N A S A CR-128540 (1971)
2.437 D.V. MADDALON : AIAA J. 7, 2355 (1969)
2.438 J. E. LA GRAFF: PhD thesis, Oxford Univ. (1970)
2.439 F. K. OWEN : AIAA Paper 70-745 (1970)
2.440 A.M.CARY: NASA TN-D-5863 (1970)
2.441 I. E. BECKWITH, M. H. BERTRAM: NASA-TM-X-2566 (1972)
2.442 R. D. WATSON, J. E. HARRIS, J. B. ANDERS: AIAA Paper 73-165 (1973)
2.443 I. E. BECKWITH : A I A A Paper 74-135 (1974)
2.444 J.M. KENDALL: AIAA J. 13, 290 (1975)
2.445 W. V. FELLER: A I A A J. 11,556 (1973)
2.446 B.THWAITF.S (Ed.): Incompressible Aerodynamics (Clarendon Press, Oxford 1960)
2.447 W.W.WILLMARTH: Advances in Appl. Mech. 15, 159 (1975)