ANNUAL

REVIEWS Further
Quick links to online content

Copyright 1976. All rights reserved

BOUNDARY-LAYER �8091
STABILITY AND TRANSITION

Eli Reshotko
Department of Fluid, Thermal, and Aerospace Sciences,
Case Western Reserve University, Cleveland, Ohio 44106

I INTRODUCTION

The phenomenon of boundary-layer transition has so far eluded a fully rational

explanation from first principles. Periodically the state of our knowledge is reviewed,
most recently in these volumes by Tani (1969), where emphasis was given to the Iow­
speed regime. However baffling the low-speed regime, transition at supersonic and
hypersonic speed is even more so. An orderly approach having both theoretical and
experimental aspects is required. The elements of such an approach are described
here and the accomplishments to date reviewed.
It was early hypothesized by Reynolds that transition is a consequence of
instability of the laminar flow. This hypothesis was further developed by Rayleigh
and to this day remains most highly regarded. It has certainly stimulated much
theoretical and experimental work in boundary-layer stability_ The excellent agree­
ment between the boundary-layer-stability experiments of Schubauer & Skramstad
(1948), Liepmann (1943), Laufer & Vrebalovich (1960), Kendall (1967), and others
with appropriate theories has provided a basis for further development of the
consequences of the Reynolds-Rayleigh viewpoint.
Nevertheless, over the years transition data have been accumulated and correlated
by traditional aerodynamic-testing procedures and quite independently of stability
considerations. These tests have yielded much data on the effects of Mach number,
surface-temperature level, the mysterious "unit Reynolds number," surface rough­
ness, bluntness, pressure gradient, suction and blowing, angle of attack, sweep, etc.
These efforts, however, have yielded neither an acceptable transition theory nor any
even moderately reliable means of predicting transition Reynolds number.
Attention has again focused in recent years (Morkovin 1969, Reshotko 1969) on
the importance in the transition process of the response of the boundary layer to the
available disturbance environment. For example, it has been shown that the transition
behavior in supersonic wind tunnels above a Mach number M of 2.5 can be clearly
ascribed to the noise radiated from the turbulent boundary layers on the tunnel walls
(Pate 1969, Pate & Schueler 1969, Wagner et al 1970, Kendall 1967). In the JPL
(Jet PropUlsion Laboratory) 20" supersonic tunnel at M = 4.5, Kendall observed no

311
312 RESHOTKO

transition on a flat plate of length Reynolds number 3.3 x 106 when the tunnel-wall
boundary layer was laminar ("quiet" operation), whereas in the same tunnel at the
same Mach number but with turbulent side-wall boundary layers, Coles (1954)
observed transition on a plate at Reynolds numbers of the order of 1 x 106• Thus the
vast body of transition data obtained in supersonic wind tunnels is suspect. Nor are
ballistic ranges free of difficulties. In a series of experiments in an enclosed range
where the model precedes any disturbances resulting from sabot impact, Potter (1968,
1975) nevertheless obtains a variation of transition Reynolds number with unit
Reynolds number that has yet to be explained. These severe effects of facility
on transition are salient among many. It is clear that the interpretation and
utilization of experimental transition data will require resolution of these various
difficulties.
One may view the transition of the boundary layer to a turbulent state as the non­
linear response of a very complicated oscillator-the laminar boundary layer-to
a random forcing function whose spectrum is assumed to be of infinitesimal
amplitude compared with the appropriate laminar-flow quantities. The initial
response to this disturbance is covered by infinitesimal-disturbance considerations.
An infinitesimal disturbance is one whose amplitude is insufficient to alter the
basic flow whose stability is being studied. The response of the boundary layer to
infinitesimal disturbances is described by linear equations. Disturbances are referred
to as large or finite when they reach a sufficient amplitude for the time-independent
or time-averaged flow quantities to depart from their laminar values.
It would seem most rlesirable to formulate stability theory in a way that simulates
experiment-namely, to take a given initial disturbance spectrum (forcing function)
and follow it forward in time. The response would depend on the receptivity
(Morkovin 1969) of the boundary layer to the particular disturbances assumed and
the subsequent disturbance amplification.
Receptivity denotes the means by which a particular forced disturbance enters the
boundary layer and the nature of its signature in the disturbance flow. If the initial
disturbances are sufficiently large, they can grow by forcing mechanisms to nonlinear
levels and lead directly to turbulent flow. If they are small, they will tend to excite free
disturbances in the boundary layer. These free disturbances are the normal modes of
the boundary layer and are often referred to as Tollniien-Schlichting waves. The
nature of each of these normal modes is determined from the solution of the eigenvalue
problem arising from considerations of the linearized disturbance equations subject to
appropriate boundary conditions. Boundary-layer-stability analyses have been
generally restricted to studies of the normal modes. However, the normal-modes
representation of a disturbance spectrum does not extend conveniently to finite
amplitude, and so the nonlinear processes between initial instability and the com­
pletion of the transition process are to date poorly understood theoretically and only
slightly better experimentally.
Therefore, the relationship between transition Reynolds number and some
representative Reynolds number from infinitesimal-disturbance stability theory is
weak quantitatively and only moderately strong qualitatively. Conversely, when it
comes to evaluating experimental transition, the results of stability theory serve
primarily as a guide.
 BOUNDARY-LAYER STABILITY AND TRANSITION 313

This article briefly reviews normal-modes procedures and the receptivity

mechanisms studied to date. It then indicates how normal-modes results can be used
for guidance as regards the factors affecting transition. Finally some remarks are
made about the prediction of transition, the fixing of transition, and the directions of
current and future transition investigation.

2 NORMAL-MODES PROCEDURES
FOR BOUNDARY LAYERS

Formulation
The normal-modes methods can be described generally as follows: Let each flow
quantity be composed of its value for the specified basic flow plus a disturbance
component as follows:
Q = Q(x, t) + Q'(x, t). (1)

For most problems, Qis independent of time. The time variation is left in temporarily
in deference to those who study the stability of basic flow patterns that are time
dependent (e.g. Shen 1961, Yang & Kelleher 1964).
The total flow satisfies the time-dependent conservation laws of mass, momentum,
and energy, whereas the basic flow satisfies a more restricted set of equations. If one
is studying the stability of steady laminar boundary layers, then the basic-flow
equations are steady boundary-layer equations. Subtraction of the basic-flow
equations from the total-flow equations yields the set of conservation-law equations
satisfied by the disturbances. Since it is stipulated that the fluctuation amplitudes
be very small compared with the basic-flow quantities, products and squares of
fluctuation quantities are neglected. The resulting equations are then linear partial­
differential equations in the variables (x, t).

PARALLEL-FLOW ASSUMPTION The equations can be further simplified by treating

the boundary layer as a parallel flow. By a parallel flow, we mean one whose stream­
lines are everywhere parallel to each other and parallel to any bounding surface.!
Strictly speaking, growing boundary layers are not parallel flows. It has been shown
(Cheng 1953, Dunn 1953), however, that to leading asymptotic approximation, the
parallel-flow approximation is valid for boundary layers. It has 'been customary to
treat the boundary layer as a parallel flow even to higher approximation, as evidenced
by the extensive numerical solutions of Mack (1965, 1969).
The calculation of a proper nonparallel-flow solution and considerations of its
importance are matters currently being pursued by a number of investigators (Barry
& Ross 1970, Ling & Reynolds 1973, Gaster 1974, Nayfeh, Saric & Mook 1974,
Wazzan, Taghavi & Keltner 1974, Saric & Nayfeh 1975, among others). A critical
review of the prior work, together with a reasonably satisfactory treatment of
the problem, is given by Saric & Nayfeh (1975). They analyze the spatial stability of
two-dimensional incompressible boundary-layer flows using the method of mUltiple
scales and apply their analysis to the Blasius and Falkner�Skan flows. Their results
1 This definition is for a two-dimensional flow. For a three-dimensional parallel flow, the

streamlines all lie in parallel planes that are also parallel to any bounding planes.
314 RESHOTKO

show that nonparallel effects are insignificant for flows with favorable pressure
gradient, are quite noticeable for Blasius flow, and become more pronounced as the
pressure gradient becomes more adverse. Their results are discussed in greater detail
later in this chapter.

FORM OF DISTURBANCE Under the parallel-flow assumptions, terms involving the

mean normal velocity and longitudinal derivatives of mean quantities are omitted.
The stability of a local flow is calculated as if its profiles existed from 00 to + 00
-

in a boundary layer of constant thickness. Under these circumstances the profiles are
functions of the normal coordinate y only, and the equations admit of a disturbance
of the form

Q'(x, t) = q(y) exp [iex(xcost/t+zsin I/! c t)]

- . (2)
This is the equation of an oblique plane wave propagating at angle t/t with respect to
the x direction. The wave number of the disturbance is ex(ex = 2n/J.., where J.. is the
wavelength) and c is the phase velocity of the disturbance. The disturbance may be
assumed to grow spatially (S: ex complex and exc real) or temporally (T: ex real and
c complex). Disturbances that neither grow nor decay are referred to as neutral.
Despite the obvious compatibility of the spatial description with the growth of
disturbances in boundary layers, the temporal description was used almost
exclusively until about 1964 when Kaplan (1964) presented results based on his
method of exact numerical integration of the Orr-Sommerfeld equation. Of course
most interest until that time was in defining the boundaries of neutral stability, in
which limit the two alternatives degenerate to the same analytical problem. For
disturbances propagating in the flow direction of a two-dimensional boundary layer,
Gaster (1962) has shown that in the limit of small amplification the spatial and
temporal descriptions both yield c, as the phase velocity and that the growth rates in
the two descriptions are related through the group velocity as follows:
(X;(S) � _ exCt(T)
(3)
Cg ,
where the group velocity is given by
8 (exc,) OC,
.

Cg = a;- = c,+ ex (4)

oex'

In most boundary-layer problems, the amplification rates are sufficiently small for
Gaster's relation to hold. For more general situations such as two-dimensional flows
subject to oblique plane-wave disturbances and for three-dimensional flows subject
to arbitrary plane-wave disturbances, the relationship is not quite as simple. This is
because the phase and group velocities are vector quantities but are not necessarily
in the same direction.
For the nonparallel-flow treatment where the profiles in addition to being
functions of y are also weakly dependent on x, the wave number ex and phase velocity
C display some x dependence ( Saric & Nayfeh 1975).

In summary, the disturbances considered in normal-modes analyses are plane

 BOUNDARY-LAYER STABILITY AND TRANSITION 315

waves. For a two-dimensional boundary layer those disturbances that propagate in

the direction of boundary-layer development (local free-stream direction) are called
two-dimensional disturbances, whereas those that propagate at some angle to the
local free-stream direction are called three-dimensional disturbances. For a three­
dimensional boundary layer where the local free stream is not in the direction of the
pressure gradient-and, therefore, one might say that there is no single direction of
boundary-layer development-it is convenient to consider all disturbances as three­
dimensional and to identify them by the angle '" of the direction in which they
propagate relative to the reference, say x, direction.
For boundary layers in incompressible and subsonic flow, the phase velocities of
the normal modes are generally within the velocity spread of the basic flow. For
boundary layers in supersonic flow, one generally deals only with "subsonic"
disturbances, that is, disturbances that move subsonically with respect to the
component of the free stream in the direction of wave propagation. Such disturbances
have amplitudes that decay exponentially in the free stream. A disturbance that
propagates supersonically with respect to the free stream would be expected to have
a nonvanishing amplitude far from the wall. In the past, the supersonic disturbances
have been dismissed as being of doubtful significance. Mack (1971, 1975a) has shown,
however, that radiated sound from wall boundary layers of supersonic wind tunnels
can be identified with incoming supersonic disturbances to a test boundary layer.
This has been one ofthe important accomplishments to date in the area of receptivity.
A most important development in the last decade in the stability of supersonic
laminar boundary layers is the discovery of the higher modes. Following the
questioning of the uniqueness of subsonic disturbances (Lees & Reshotko 1962),
Mack (1965) encountered the higher modes in his numerical calculations, and Lees
(Lees & Gold 1966) confirmed the conditions for the existence of these additional
modes. We now know that if the wall is supersonic relative to the phase velocity of
infinitesimal disturbances (crlaw > 1), then the boundary layer is rich in unstable
normal modes, some of which are not damped by cooling (Mack 1965). For insulated
surfaces, higher modes appear for M > 2.2; however, it is not until the Mach number
is of the order of 4 or greater that the second mode is at low enough frequency to have
experimental consequences. With cooled walls, since c, > 1- 11M for subsonic
disturbances, the higher modes can be significant at Mach numbers as low as 1 if the
cooling is sufficient.

PROPERTIES OF DISTURBANCE EQUATIONS The disturbance equations derived by the

procedures indicated in this section2 have been shown to display the properties
outlined below.
With regard to the stability of two-dimensional parallel flows to three-dimensional
disturbances, Squire (1933) has shown that for an incompressible fluid the disturbance
equations can be transformed to the completely two-dimensional Orr-Sommerfeld
equation and that the two-dimensional disturbance is the least stable. Dunn & Lin
(1955) considered the stability of a two-dimensional compressible boundary layer to
2 The complete disturbance equations for a three-dimensional compressible parallel flow

subject to an arbitrary plane-wave disturbance are derived and stated in Reshotko (1962).
316 RESHOTKO

three-dimensional disturbances. They showed that when only the leading viscous­
conductive effects on the disturbances are considered the equations for three­
dimensional disturbances can be transformed to those for two-dimensional dis­
turbances. They carefully point out that for compressible flow these transformed
equations are not the equations of a proper two-dimensional disturbance, so that no
"families of solutions" are obtainable. Furthermore, while the transformation does
permit the use of solution procedures for two-dimensional disturbances in problems
of three-dimensional disturbances, with modern computing machinery the additional
time required for the proper three-dimensional solution is small enough that the two­
dimensional shortcut is not justified.
The stability of three-dimensional boundary layers to three-dimensional dis­
turbances is considered for incompressible flow by Owen & Randall (1953) and by
Gregory, Stuart & Walker (1955). Their results for a parallel flow have been
concisely summarized by Moore (1956) : " For a disturbance assumed to be moving
in a certain direction, the eigenvalue problem may be treated as a two-dimensional
one, governed by the boundary-layer velocity profile measured in that direction."
Of course, for incompressible flow the energy equation is irrelevant, and within the
framework of the paralIe1-flow assumption this statement is exact. It is shown for
compressible flows (Reshotko 1962) that the transformation implied by Moore's
statement applies exactly for the continuity and momentum equations but only for
the leading terms of the energy equation. As has already been pointed out by Dunn
& Lin (1955), the dissipation terms do not all transform. Mack (1967) compared
results for first-mode disturbances with (eighth-order system) and without (sixth­
order system) the nontransforming terms and found the differences in amplification
rate to be generally less than 10%. The differences are most pronounced at low
Reynolds numbers, as would be expected. Nevertheless, present-day calculations are
for the eighth-order system so that ambiguities in interpretation will not arise.

Results of Normal-Modes Calculations3

The results of normal-modes calculations are usually presented in diagrams of wave
number IX versus a thickness Reynolds number. Such diagrams for three different
Mach numbers are shown in Figure 1. Since the dimensionless frequency
F == wv/U2 can be written

F == �� = (:e) C'
and since for a given frequency c, varies very little, a line of constant frequency is
almost a straight line through the origin of the IX- Re diagram. Note that when higher
modes are present, a given frequency may correspond to progressively higher modes

3 All stability results presented and discussed herein are from exact numerical solutions of

appropriate sets of disturbance equations. Enough exact solutions exist to convince anyone
that results of asymptotic approximations are quantitatively unreliable. There are enough
successful numerical procedures available that obtaining other than exact solutions for
quantitative purposes is inexcusable.
 [M:fOI [M:4.5] 0(
[Moo: 5.81

STA6LE
STI\BLE.
0(
0(
§
z

i
�

STA8LE
�
Remi �
. Re. Re Re.
",.it
�z
1
�
Figure Stability diagrams for insulated boundary layers.

w
.­
-.I
 w
-
00

400 �O
ro ----.----,--.--

� �-o.-" - - - PARALLEl
•
I �:Ol
PARALLEL
NON-PARALLEl
I
- NON-PARALLEL
70- 300-

60-

-0
.00 o
-
50-
-
)( )( 200-
�
L&-

40-

30- 100-
1203 1336

ro ----�----���--�--
1000 1200 1600 1400 1600
Re.�.
o , I

400 600 800 1000 1200

I I

Re� ....
 W::O: I9SS I
---r--r-�--�-'�.'--'--.r--r--�-'--,-�

[[:=�Dl - - - PARALLEL PARALLEL

- NON-PARALLEl NON-PARALLEL

2000-

�
....
1000- -0
o
.....
10,000-
- �
)(
)(
�
j
LL
LL

500- 5000 - r' ''' , ,

-. "

\ ,

\ .....
......

200- 115 203

\
\
\
\
......
......... i
�
\ >
\ Z
o
\
WO 100°
o 100 200 500 600 700 0 40 80 120 160
Re.�..
200 240 280 �
Figure 2 Neutral-stability diagrams for Falkner-Skan flow. 0: data of Schubauer & Skramstad (1948); 0, . : lower- and upper-branch
�
data, respectively, of Ross et al (1970) (Saric & Nayfeh 1 975). \;J
......
'"
 320 RESHOTKO

as the Reynolds number is increased, or else may excite the higher modes without
exciting the first mode. The Reynolds number below which all wave numbers are
damped is termed the minimum critical Reynolds number. The stability diagram may
be alternatively plotted as dimensionless frequency F versus Reynolds number, and
it is in this form that the recent results for the nonparallel-flow treatment of
incompressible boundary layers are described and discussed.

NONPARALLEL-FWW EFFECTS Figure 2 shows neutral-stability curves of F vs Re6' for

four different values of the Falkner-Skan pressure-gradient parameter p. The non­
parallel effects as calculated by Saric & Nayfeh (1975) decrease the minimum critical
Reynolds number by about 100 in each of the cases. This is a much more significant
change for adverse pressure gradients than for favorable ones. Note the good agree­
ment between the nonparallel results and the flat-plate experimental data of
Schubauer & Skramstad (1948) and Ross et al (1970), particularly near the minimum
critical point of the curve. Similar results have been obtained recently in water­
tunnel stability experiments by Strazisar et al (1975). The growth rates near the
minimum critical point are very small; and so, while the neutral points can be
identified, there is noticeable scatter. From the viewpoint of instability as a
precursor of transition, a more important range of frequencies is that located in the
center of the unstable region where growth rates are large. In this latter region, there
is little difference between the results of parallel and nonparallel treatments; and

1.4

SfCONDMODE
1.2 0/ = Odeg
..
0

"

1.0

'-'
�

""'
z

::>
0.8
""'
0-

'"
"-

:::
V>

.J 0.6
:::
Z

V)
%
W 50
:E
0.4
0
FIRST MODE

0.2
= 45 deg

0 1 1 1 1
0 2 3 4 6 10
MACH NUMBER, M..

Figure 3 Effect of Mach number on first- and second-mode most unstable frequencies at
Re = 1500. Insulated wall; wind-tunnel temperatures (Mack 1969).
 BOUNDARY-LAYER STABILITY AND TRANSITION 321

hence, while the nonparallel calculations are impressive in their agreement with
experiment, the parallel-flow calculations are adequate for considerations relevant to
transition. The balance of the stability information to be presented is for the parallel­
flow treatment.

FLAT-PLATE BOUNDARY LAYERS IN AIR Stability diagrams of results available to

1966 can be found in Betchov & Criminale (1967). The reported results of Mack's
extensive calculations for flat-plate boundary layers are to be found in Mack (1969).

6.6

4.0
SECOND MODE
'" = 0 dog
0')

a 3.6
I

)(
x
.«
::E
3.2
"
.�

...; 2.8
I-
«
'"
z
8 2.4
I-
«
u

:J
u.

0- 2.0
::E
«
-r--6S_ _ 60
-r .......c
::;
...J

I-
« 1.6
0- 55
Vl
::E
::::>
:5 1.2
x fiRST MODE
«
::E

0.8

0.4

0 I. 1 I 1
0 4 5 6 8 10

MACH NUMBER, M

Figure 4 Effect of Mach number on spatial amplification rate of most unstable first- and
second-mode disturbances at Re = 1500. Insulated wall; wind-tunnel temperatures (Mack
1969).
322 RESHOTKO

0.36

TwiT, = 0.05 (T.. = 12S·K)

0.32

0.28

0.24
"

INSULATED WALL
r;;
.... 0.20
to
:>:
:=>
:z

0.16
w
'- TwiT, = 0.25
"'"
:>

'"

0.12

0.08

0.04

o I I .. I I I I I ! I
o 200 400 600 eoo 1000 1200 1400 1600 1600 2000

REYNOLDS NUMBER, Re

Figure 5 Effect of wall cooling on neutral-stability curve at Mach number 5.S. Two­
dimensional disturbances, T <Xl = 500K (Mack 1969).

A rather concise summary of stability characteristics for flat-plate boundary layers

as presently understood is given by means of the following figures taken from
Mack (1969).
Figures 3 and 4 describe the characteristics of the most unstable first- and second­
mode frequencies for insulated boundary layers. The data are for Re = 1 500, which
corresponds to a length Reynolds number of 2.25 x 106.4 This is sufficiently ahead of
observed transition Reynolds numbers that the stability results are relevant. In
Figure 3, the dimensionless frequencies are shown. The most unstable first-mode
frequencies at supersonic speeds occur for oblique waves with if! generally between
45° and 65°, while the most unstable second- (and higher-) mode frequencies occur
for if! 0°. The M 1 0 point is of course for if! 0° by virtue of Squire's theorem.
= = =

The associated spatial growth rates are shown in Figure 4. The second mode, once
activated, clearly displays higher growth rates than the first. The rapidity of decline

4 Mack's definition of Re corresponds to the square root of the length Reynolds number.
 BOUNDARY-LAYER STABILITY AND TRANsmON 323

of spatial growth rate with Mach number, particularly around Mach number zero,
should also be noted.
The effect of surface cooling on stability is of significance because of the great
variety of aerodynamic applications that require cooling. The first mode is generally
stabilized by cooling. In fact, two-dimensional first-mode disturbances can be
completely stabilized by cooling up to Mach numbers of the order of 9 (Lees 1947,
Dunn & Lin 1955, Reshotko 1963). While the oblique waves cannot all be completely
stabilized, it is expected that.cooling greatly increases minimum critical Reynolds
numbers and diminishes growth rates. On the other hand, the higher modes are not
stabilized by cooling. They tend toward higher frequency and higher growth rate as
the surface temperature is reduced. Stability diagrams for two-dimensional dis­
turbances at M 1 = 5.8 with different degrees of cooling are shown in Figure 5. It is
seen that when the temperature level has decreased to TwiT,. = 0.25, the first mode
has completely disappeared while the second-mode bulge has shifted to higher wave
numbers (higher frequencies). The effect of surface temperature on growth rate at
M 1 = 5.8 is shown in Figure 6 in the inviscid limit. The effect of cooling on growth
rate in the inviscid limit for the first four modes at M 1 8 and M 1 = 10 are shown
=

M
o

)(
>< ____ SECOND MODE, 2-D
� --- �
--

:--- ----
---
--
---
---
4

TH ST MODE, 3-D�

2
IRD MODE, 2-0"""",
T "'" 55d.. ...,

"t'a.
.--
-��
- i--
--
I -
�= 70d.. .--- -
� -
� r---
/"'"
- fiRST MODE, 2-D
0
0.1 0.2 0.3 0.4 0.5 0.6 0.7 O.B 0.9 1.0

RATIO OF WALL TO RECOVERY TEMPERATURE, TwIT r

Figure 6 Effect of wall cooling on maximum temporal amplification rate of first three
modes at Mach number 5.8. Too 500K (Mack 1969).
=
324 RESHOTKO

in Figures 7 and 8, respectively. The upward shift with cooling in both frequency and
growth rate is apparent.

FLAT-PLATE BOUNDARY LAYERS IN WATER It is curious that for boundary layers

in water, the effect ofcooling is destabilizing whereas the effect of heating is stabilizing
(Wazzan, Okamura & Smith 1968, 1970). Since water is essentially incompressible,
only the "first" or traditional mode is involved. Because the viscosity of water
decreases sharply with increase in temperature, heating yields a fuller velocity
profile whereas cooling tends to give an inflected velocity profile. The results of

S.O

I. INSULATED WALL

- - COOLED WAll
2
,..
I ' MODE NUM8ER

I \
n

•. 0
, \
, ,
, I
M I \
0

'" I I
U 3.0 , I
"
.,; , ,
!;;:
'"
I I
z 100 = SO".
0
I ,
;::
'-'
- ,
<C 3

"-
::l
('
"- I I \
�
2.0

-' , I \
�
"-
0 I I \
::E
LU
I- I \
I \
1.0
, \
, \
., I
n

I \
, \
I \
/ \
, "
0.1 0.2 0.3 0.'
WAVE NUMBER, (l

Figure 7 Temporal amplification rate vs wave number for first three modes for insulated
wall and cooled wall. M., 8 (Mack 1969).
=
 BOUNDARY-LAYER STABILITY AND TRANSITION 325

Wazzan et al were obtained by solving a fourth-order system of disturbance equations

that included mean-viscosity variations but ignored the disturbance energy equation
and the associated temperature and viscosity fluctuations. Because of the rather
unusual results obtained, LoweIl & Reshotko (1974) obtained solutions to the sixth­
order system of disturbance equations that included all mean- and disturbance­
property variations. The results for minimum critical Reynolds number of a heated­
flat-plate boundary layer are shown in Figure 9. Both systems predict significant
boundary-layer stabilization (increased Remin crib decreased disturbance amplification
rates, etc) with moderate heating, but display a maximum and subsequent decrease
as the wall-to-free-stream temperature difference is further increased. Over the
normal liquid range of water, the results of the sixth-order calculations show only
slight enhancement of stability over those obtained from the fourth-order system.
Since the amplitude of the temperature fluctuations is not small, the insensitivity of
the stability characteristics to their inclusion is attributed to the narrowness of the

4.0

- INSULATED WAll

- - COOLED WALL

i
'w ',
0,05

n MODE NUMBER

M
0 3.0

"

U
"
.
f-
LW

'"
«
('
z
0 I \
>=
« 7.0
I
- I
u
"-
-
-'
0- I
::E
«
-' I
�
"-
0 I
::E
LW
f-
I ,
1.0

\
\
\
0 ·- I

"
"
\
0.1 0.4

WAVE NUMBER, (l

Figure 8 Temporal amplification rate vs wave number for first four modes for insulated wall
and cooled wall. Moo 10, Too
= SOaK (Mack 1969).
=
326 RESHOTKO

Res *min. crit ..

X IO-�

-0- LOWELL � RESHOTKO

(1974-)
12 -0(;- WAZ.'ZAN,OI<AMUQA
• &MIT� (1970)
It
III II
10
� 10
:>
2 9
dI
0 e
.J
0 �
2 7 \
\
> \
III 6 \
U
.J .s
'" ,
,
� ,

t=
4 ,

ii 'Q."
u �
''''0
� '2
J
�
Z
� I , I
100 '200 &00

Figure 9 Effect of wall temperature on minimum critical Reynolds number (Lowell &
Reshotko 1974).

region in which such disturbances are important for fluids of large Prandtl number,
such as water.
The calculated sensitivity of the stability characteristics to small amounts of
heating has been confirmed experimentally by Strazisar et al ( 1975). Figures 10 and 1 1
show neutral-stability curves for the unheated flat plate and for the case where the
wall temperature is about 5°F above the stream temperature. The theoretical curves
(Figure 10) are the parallel-flow results of Lowell & Reshotko (1974). Note that the
experimental shifts in neutral curve due to 5° of wall heating are consistent with the
theoretical shifts. There is almost a 50% increase in (Re�')mincrit indicative of a
doubling in length to the minimum critical point, and the heated neutral curve falls.
generally within its unheated counterpart. 5 These results portend exciting possibilities
for drag reduction.

5 As indicated earlier, the experimental results of Figure 1 1 are quite consistent in behavior

near the minimum critical point with the results shown in Figure 2.
 BOUNDARY-LAYER STABILITY AND TRANsmON 327

EFFECT OF SURFACE CURVATURE This review is concerned by and large with

propagating waves of the form of equation (2), historically termed Tollmien­
Schlichting waves in the boundary-layer context. Over concave surfaces Gortler
( 1940a,b) showed that the boundary layer is unstable to longitudinal-vortex
disturbances (Taylor-Gortler vortices) very much akin to the vortices that appear
between two cyiinders, the inner one rotating and the outer at rest (Taylor 1923).
Liepmann (1943) has in fact observed for incompressible flow that transition on
convex surfaces occurs at about the same Reynolds numbers as on flat plates, whereas
on concave surfaces the transition Reynolds number decreases almost linearly with
8/R from the flat-plate result. In this expression the symbol R denotes the radius

Tw= 75°F

Tw= BO°F
300

200

...
o

100

o ' I I I I

�OO 600 800 1000 1200

- '

R. .
,
Figure 10 Theoretical neutral-stability diagrams for unheated and heated fiat-plate
boundary layers in water (Strazisar et al 1975).
328 RFSHOTKO

of curvature ofthe plate. A comparison ofthe calculations of Kaplan (1964) and Smith
( 1955) shows that for incompressible boundary layers over concave surfaces, the
minimum critical Reynolds number for Tollmien-Schlichting instability is lower than
that for Taylor-Gortler vortices when J/R < 1/40,000, and vice versa.

100

o , I I I I I

400 600 800 1000 1200

Re�t
Figure 1 1 Experimental neutral-stability results for unheated and heated fiat-plate
boundary layers in water (Strazisar et al 1975).
 BOUNDARY-LAYER STABILITY AND TRANSITION 329

3 RECEPTIVITY

The properties of the laminar boundary layer as a linear oscillator have been reviewed.
The normal modes have been identified and their propagation and growth
characteristics described as a function of wave-number and boundary-layer Reynolds
number. But how are these modes excited by the available disturbance environment?
What specifically are the means by which free-stream turbulence, entropy
disturbances, tunnel sound fields, surface roughness, etc, trigger the growing normal
modes? These are the questions addressed under the heading of "receptivity"
(Morkovin 1969). Although these questions had been articulated for some time, they

� �------T---�--�--�

RESPONSE TO EXTERNAL
.. 0
PRESSURE F1 ELO, .,. cleg
(e, •0.65, 0.01)
a ..

16�------�--��--+---�
MI· 4.5
R .. 800
0.08125 10"" F .. x

12�------�--��-+--+---�
E
......
E

8�--�----��---+--+---�

4 �----�--+-----4---tt---�--�
MOST UNSTABLE
EIGENSOlUTlON
t" 60 deg
(Cr 0.667,
.. a .. 0.0(98)
,
I
I rc,
O�______����__��______�________�
o 0.4 1.6

y/o
Figure 12 Distribution through boundary layer of mass-flow-fluctuation amplitude
(Mack 1970).
330 RESHOTKO

have been addressed only within the last six or so years, and the treatments are all
somewhat incomplete. Nevertheless, some intriguing and possibly important results
have been obtained.
The receptivity phenomenon differs from stability both physically and mathemati­
cally. Physically, it is the signature in the boundary layer of some externally imposed
disturbance. Mathematically, the problem is no longer one ofhomogeneous equations
with homogeneous boundary conditions, but one where either the equations and/or
the boundary conditions are nonhomogeneous. Hence in contrast to the normal­
modes stability calculations, the receptivity phenomenon is not an eigenvalue
problem. The boundary layer is driven by the external forced oscillations, and its
response is a neutral solution of the linearized disturbance equations having the same
frequency and phase speed as the particular forcing disturbance being considered.
Each class offorcing disturbance has its own particular inhomogeneity. Therefore the
study ofthis important initial element of the transition process is reviewed by focusing
on three receptivity phenomena for which there is some information.

18

F X 104 O. 040625
16 0.08125

12

-
E
CoRS --
E

o , I I I I
o 400 800 1200 1600 2000
Re
Figure 13 Effect of frequency on peak mass-flow fluctuation. Moo 4.5, insulated wall,
'"
=

= 0°, Cr =0.65 (Mack 1971).

 BOUNDARY-LAYER STABILITY AND TRANSITION 331

Response of Boundary Layer to aMoving Sound Wave

Measurements by Kendall ( 1970) on a flat plate at M", = 4.5 in the JPL 20" super­
sonic tunnel with turbulent boundary layers on the tunnel walls show all frequencies
to grow monotonically from the leading edge-well ahead of the minimum critical
point-and the source of these disturbances has been identified as the sound field
radiated from the tunnel-wall boundary layers. Thus the sound field interacts with the
boundary layer to provide growing disturbances in a region that is stable according
to stability theory. In an attempt to account for these observations, Mack ( 1970, 1971)
altered his stability theory to include the response of a layer to incoming sound waves.
He in fact identified the radiated sound from the tunnel walls with incoming super­
sonic disturbances to the test boundary layer and proceeded to calculate the response
of the boundary layer to this incoming forced disturbance. Satisfaction of the
impermeability condition at the plate (v' = 0) requires that there be a reflected sound
field from the plate.
Figure 12 compares mass-flow-fluctuation amplitudes due to an incoming sound
field with the result from the most unstable eigensolution at the same frequency and
Reynolds number. The ratio m'!m'r is of the rms mass-flow fluctuation to that of the
incoming wave in the free stream at y = O. For the forced disturbance the peak.
amplitude in the boundary layer is 16 times that of the incoming wave, while in the
free stream m'/m'r of the combined incoming and reflected waves is less than 2. The

200

M. = �.5
• -I
Re = 3.7·AND 7.2 x 10 em
100
oeF = 0.04 >110-4
6. 0.16
o. 0.65
Q 50
�
...

2 THEORY, COMBINED
Q FORCING AND STABILITY
...

�...
20
NEUTRAL POINT
::l
...
::E
-<

•
t 0

10
�
5

2 I I
o 400 800 1200

Re.
Figure 14 Comparison with combined theory. M", = 4.5 (Kenda1l 1975).
332 RESHOTKO

peak m' of the eigensolution is scaled to the same peak as the forced curve. Note how
close the two functions are over most of the boundary layer despite the mismatch in
phase speed and direction of propagation. Figure 13 gives the ratio of m�, the peak
rms mass-flow fluctuation in the boundary layer, to mI' Disturbances of all
frequencies are seen to grow rapidly with distance from the leading edge, reaching a
peak in the vicinity of the region at which amplification due to instability begins. The
magnitude of the peak is inverse to the frequency.
Mack then calculated the growth of disturbances at selected frequencies by using
the forcing theory up to the neutral-stability point and stability theory beyond the
neutral point. The calculated results compare very well with measurements of

1.0 r---�--��--�---

Moo 4.5:l

R
e
= 72,OOO/cm

Re= 425
600
'L
850
U 1040
1200
1470
0.4 EMPTY
TUNNEL
STAB ILiTY THEORY,
R = 1470, 'II = 60°

1 -1/Moo' SONIC

0.2 I I I
o 0.2 0.4
III

0.6 0.8

F X 104
Figure 15 Variation of wave speed with distance along the plate. Moo = 4.5 (Kendall 1971).
 BOUNDARY-LAYER STABILITY AND TRANsmON 333

disturbance amplitudes at three different frequencies as shown in Figure 14 (Kendall

1975), lending some credence to the hypothesis. However, the mechanism by which
the forced wave turns into the free wave is unknown and has been ignored by Mack,
although the process most assuredly does take place, as seen in Figure 15 (Kendall
1971). The measurements show the dimensionless wave speed, initially supersonic
relative to the stream, to accelerate during their streamwise travel, approaching a
speed approximately equal to that predicted by stability theory.
The results in Figure 14 are essentially the signature of disturbance growth
triggered by tunnel sound and are representative of additional measurements by
Kendall (1975) at Mach numbers 3 and 5.6. At Mach numbers 1.6, 2.2, and 8.5, the
observed frequency response in the boundary layer shows definite peaks at the most
unstable frequencies of stability theory and little if any effect of tunnel sound. Hence
the most pronounced contamination of the wind-tunnel environment by radiated
tunnel sound seems to be in the middle-Mach-number range that stretches from
2.5 to 7 or so.

!I

UfO) = - cos(A-f/+�) sin-f(x-t)

V(a) = sia(.hf/+�} cos Jr(x-t)
Figure 16 Convected array of counterrotating harmonic vortices (Rogier & Reshotko
1975).
334 RESHOTKO

Disturbances in a Boundary Layer Introduced by an Array of Vortices

To acquire insight into the role. of fre�stream turbulence on transition, RogIer &
Reshotko (1975) analytically and numerically studied the interaction between an
incompressible boundary layer and a low-intensity array of single-wave-number
vortices convected at the mean free-stream velocity (Figure 16). The flow field is taken
to be the sum of the steady laminar field (Blasius) plus a flow field ascribable to the
effects of the vortex array. This latter flow field is further subdivided into the portion
that exists in the absence of the plate (the vortex array itself) plus a flow field
representing the alteration to that array due to the shearing mean flow and no-slip
and impermeability conditions at the plate surface. This last portion of the flow field is
described by a nonhomogeneous Orr-Sommerfeld equation with dimensionless
phase-speed unity and real wave number. The forcing function depends on the mean
flow and on the free-stream disturbance array.
It was found (Figure 17) that amplitudes in the boundary layer grow in the down-

-1
10 -

Rt\ 5000

-,
1 2 4 5

LENGTH REYNOLDS NO. Uoox/v

Figure 17 Variation of V" with length Reynolds number along the trajectory � 4 =

(� is the Blasius similarity variable) for three Reynolds numbers based on vortex diameter
(Rogier & Reshotko 1975).
 BOUNDARY-LAYER STABILITY AND TRANsmON 335

stream direction, with the maximum amplitude arising near the leading edge for small
vortex diameters and further downstream for larger diameters. This is somewhat
related to the trend found by Mack (1971) for tunnel sound.
Rogier & Reshotko (1975) further speculate that if initial turbulence levels are
sufficiently large the disturbances can grow by forcing mechanisms to nonlinear levels
and lead to turbulent flows without resort to Tollmien-Schlichting amplification.6
This is an example of what Morkovin (1969) calls a "high-intensity bypass."

Surface Roughness
In an unpublished paper, R. E. Kaplan treats roughness as a disturbance forced by a
sinusoidal wall whose amplitude is the roughness height. The disturbance equation
is the homogeneous Orr-Sommerfeld equation with zero phase speed but with non­
homogeneous boundary conditions. The mechanism by which eigensolutions are
excited by this forced disturbance is again unknown.
The consideration of forced disturbances is an important new direction in studies
related to instability and transition. There is more to be done in understanding the
three "receptivity" phenomena described as well as others not yet considered.

4 FACTORS AFFECTING TRANSITION

Whether one proceeds from the discussion of the prior sections or goes through
similitude arguments (Reshotko 1969), it is abundantly clear that, in addition to
being a function of the mean-flow conditions, transition must in some way be related
to the wave-number and orientation spectra of the disturbance environment. This
was pointed out by Laufer (l�54) many years ago and again emphasiZed by
Morkovin (1969), as evidenced by Figure 18 taken from his work. The disturbances
are identified unequivocally at the top of the diagram as input. The vertical arrows
between input and linear amplification represent the receptivity phenomenon just
discussed. The linear amplification is due to growing normal modes. The traditional
"factors affecting transition" are identified in the diagram as operation modifiers, i.e.
factors modifying the amplification characteristics of the oscillator. This diagram is
well worth studying in that it summarizes in a very concise way the behavior in the
linear and early nonlinear regimes of instability that may eventually lead to
transition. Not shown in the diagram is what Morkovin (1969) calls the "high-

6 A possible case in point is Poiseuille pipe flow. Wygnanski &- Champagne (1973)
observed that when their smooth pipe was carefully aligned (with a very good entrance
section) turbulent slugs were observed at Re > 5 x 104, which they ascribed to the
consequences of entrance-flow instability. Transition could be initiated at lower Re
(2000 < Re < 27(0) when a large disturbance was introduced into the inlet. Furthermore,
their observed disturbance amplitude near the wall at Re = 2360 was greatly damped,
consistent with the aforementioned calculations of Rogier & Reshotko (1975) for a flat plate.
Hence the speCUlation that observed pipe-flow transitions at Re � 2500 are due to large
initial disturbances that grow to nonlinear levels by forcing mechanisms, since Poiseuille
pipe flow is stable to Tollmien-Schlichting waves and the entrance flow does not become
unstable until much larger values of Re are attained.
336 RESHOTKO

tf.
A.C . I N PUT == D 'ST UR BA NCE_
:::::;:::;;:::
free-stream ;;;:;,:-;;�_
vorticity
S_1-
If-
_
_

�
poor ob servat ion of disturbances
poor control of disturbances
j
" sound .
" entropy spots
high
frequency
vi brations mu ltiple c hannels

�
in paralle l :
RACE between
instability MODES
______ -..J

LINEAR AMPLIFICATION OPERATION MODIFIERS ==

== MEAN B. l. PROPERTIES
2 0 : D I R ECT 3 0 : IN DIRECT
of Xl Fourier
com p onents
of disturbances ­
p (x) 3D roughness
Tw I T, properties -

m
past Recr of functions of l
each mode -
curvature e.g. p ( l )
slow and waviness angle of yow
extended - 20 roughness leading - edge

CUMULATiVE EFFECTS
angle of sweep
attock 3D non ­
now directly observable low f homogeneity
vibrations
etc . etc.

= D.C. modifier
{
effect on mean fIOW 2 dim.
3 dim.
vorticity stretching
lateral energy transfer �
in overgrown wa ves
I
SECONDARY +
I iNSTABILITY

Figure 18 Laminar boundary layer as a linear and nonlinear operator (Mor�ovin 1969�.
intensity bypass"-the process by which large-amplitude input leads to turbulent­
spot formation without resort to Tollmien-Schlichting amplification. The implication
in the diagram is that once turbulent spots are formed, transition will generally
follow.
 BOUNDARY-LAYER STABILITY AND TRANsmON 337

Based on stability considerations, Reshotko (1969) has deduced the folIowing

plausible forms for the relation between the transition Reynolds number and the
characteristic dimensionless frequencies and/or wavelengths of the disturbance
spectra :

(Re)tr �
(U2)"
OJV (5)

(Re)'r �
( -)
VA. "
.
VCr

acterized by OJ, the transition Reynolds number will vary with U�/v.
Equation (5) indicates that for a given disturbance-frequency spectrum char­
The coefficient
and exponent wiII be functions of Mach number and surface-temperature level and
possibly also wvlU2 to allow for deviations from the power law of equation (5).

characterized by A, the transition Reynolds number will vary with Ulv,

Equivalently, equation (6) indicates that for a given wavelength spectrum
where again

and Ullvc,.
the coefficient and exponent will depend on Mach number, surface-temperature level,
The dimensionless phase velocity is a very slowly varying quantity
particularly at hypersonic Mach numbers.
In less abstract language, we are saying that the importance of a given physical
(dimensional) frequency or wavelength depends on the amplification associated with
that frequency or wavelength. But the amplification depends on the dimensionless
frequency or dimensionless wavelength, and so the importance of a given physical
spectrum (characterized by m or .-1.) in leading to transition depends on the associated
values of U2/v or U/v, respectively.
The phenomenon just described through equations (5) and (6) may well be what
has been traditionally referred to as the unit-Reynolds-number effect. From the
arguments presented, this effect is to be expected in any facility or test where the
spectrum of available disturbances is nonwhite in the bands that have relevance to
instability and transition. Accordingly, it may be encountered in any facility. If it is a
physicalfrequency spectrum that remains invariant from one test to the next in a given
U v.
facility, then (Re)'r will depend on 2/ On the other hand, if it is a physical wave­
length spectrum of disturbances that remains fairly constant over a range of facility
operating conditions, then (Re)" will vary with UIv. A combin ation of the two is also
possible.
The discussion so far has for simplicity ignored the orientation spectra of dis­
turbances. These can be readily accommodated. It is known that the growth rate of
disturbances is orientation dependent, and so it is quite possible that the transition
Reynolds number would also show some dependence on orientation spectrum. This
dependence has yet to be sought experimentally.
Reshotko (1969) points out another consequence of the dimensionless-frequency
and/or dimensionless-wavelength arguments, that is, the tendency offacilities or flight
altitudes to emphasize particular modes of instability of the supersonic and hyper­
sonic boundary layer. It is shown there that in the ballistic-range tests of Sheetz (1965)

7 Note that wv/U 2 = 2n/(U).jvc,).

338 RESHOTKO

at Mach number 5, second- and higher-mode excitation is highly improbable and

hence the observed transition behavior is dominated by first-mode considerations.
Shock-tunnel data reported by Stetson & Rushton (1967) at about the same Mach
number and surface-temperature level but at an order of magnitude lower value of
U2jv show a decrease of transition Reynolds number with cooling as might be
expected through the involvement of second and higher modes. It is also shown that
in data reported by Deem & Murphy (1965) and by Sanator et al (1965) at Mach
number 10 in the AEDC von Karman Facility Tunnel C, second- and higher-mode
excitation is quite likely and tends to explain the insensitivity of transition Reynolds
number to surface cooling.
The prospective involvement of higher modes in a given supersonic or hypersonic
situation is as follows : The lower the value of U2jv (therefore the higher that of
wv/U2), the greater the importance of the higher modes. It seems that it may be
difficult to escape the higher modes in steady-flow hypersonic wind tunnels, while on
the other hand they may have little relevance to transition in a ballistic range.
While the physical disturbance frequencies in flight are unknown, the correspond-

107

FACILITY
2

0 J PL 20 TU�EL
• J PL 20"
TUNNEL �
=>
IO
'

QUIET OPERATrON

£; VKF BALLISTIC
RANGE K

0 NOL BALLISTIC
RANGE

L1 AVCO RAD
SHOCK TUNNEL

[:,. NASA LANGLEY

18" VARIABLE
DENSITY TUNNEL

<> V K f TUNNEL C

5
5 15O,OOQ

2 2

161L---:!,--L...1+LJ...l. �_� 10
•
10
I 1 1 1 1 I
2 20
I I I
I 2 I 5
MACH NUMBER, M MACH NUM8ER, M

IL ) ;. = 10 kH'l b) ).= I INCH

Figure 19 Dimensionless disturbance parameters in flight and in various test facilities.

 BOUNDARY-LAYER STABILITY AND TRANSITION 339

ing dimensionless frequencies wv/U2 are strongly dependent on Mach number and
altitude. This is shown in Figure 19a for an assumed frequency of 10 kHz. The
dimensionless frequency changes by about an order of magnitude for each 50,000 feet
of altitude. Thus a 1OO-kHz disturbance at 100,000 feet has the same dimensionless
frequency as a 10-kHz disturbance at 1 50,000 feet. The dimensionless frequencies
corresponding to 10 kHz in each of a number of hypersonic facilities are super­
imposed. Because of the equality of frequencies, the altitudes indicated for each
facility are their U2/v altitudes.
If the disturbance spectrum of a given facility is known, then its corresponding
range of dimensionless frequencies would indicate the range of frequency-altitude
combinations simulated.
Again, equivalent arguments can be presented in terms of wavelength through the
parameter UNv. The values of the dimensionless wavelength UA/V for a disturbance
having a physical wavelength of one inch at various flight altitudes are shown in
Figure 19b. The values for a one-inch-wavelength disturbance in each of the facilities
of Figure 19a are superimposed. In comparing Figures 19a and 19b note that the
U/v altitude of a given facility is not necessarily equal to the U2/v altitude. For the
cited facilities the U/v altitude is slightly lower than the U2/v altitude.
The order-of-magnitude variations of U/v and U2/v with each 50,000 feet of altitude
indicate that significant attention must be given to the choice of laboratory test
conditions in order to ·simulate closely a particular dimensionless disturbance
environment.

5 PREDICTION OF TRANSITION

As has been indicated, the objective of the foregoing presentation is to lead to a

rational scheme of predicting transition behavior in wind tunnels as well as in flight.

Calculation Methods
AMPLITUDE-RATIO METHODS Significant attempts in accomplishing this objective
were made by Smith & Gamberoni (1956) and independently by Van Ingen (1956)
using a method based on linear stability theory. For low-speed flow, they correlated
transition Reynolds number over plates, wings, and bodies with the amplitude ratio
of the most unstable frequency from its neutral point to the transition point. Using
theoretical values of Ci from the temporally growing calculations of Pretsch (1942)
for the Falkner-Skan profiles, together with experimental data on transition Reynolds
number, Smith & Gamberoni (1956) found that the transition Reynolds number
Rex•tr as predicted by assuming an amplification factor of e9 was seldom in
error by more than 20%. Jaffe, Okamura & Smith (1970) updated the Smith­
Gamberoni method by using spatial growth rates calculated by exact solution of the
Orr-Sommerfeld equation for the locally observed (or calculated) profiles on various
shapes. They found good correlation with estimations based on an amplification
factor of elO• This procedure with exponents ranging anywhere from about 8 to 1 1
has found wide application in transition correlation and prediction.
Despite the apparent success of these procedures, they are defective in principle
340 RESHOTKO

and perhaps also in practice. From the beginning, Smith (Smith & Gamberoni 1956)
acknowledged that the boundary layer is "agitat!Xl by disturbances impressed upon
it by external turbulence, surface roughness, noise, and vibration," and that "the
true flow is similar to a forced vibration." Yet, the disturbance spectrum is in no way
involved in his method, and accordingly there is no way of introducing a unit­
Reynolds-number effect. A pointed example of the defectiveness of the method is that
it cannot explain why for a flat plate Schubauer & Skramstad (1948) obtain a
transition Reynolds number of 2.84 x 106 while Wells (1961) obtains 4.9 x 106. The
difference is no doubt due to the reduction in background noise in the Wells (1967)
experiment but there is no accommodation of this fact into the Jaffe et al (1970)
procedure. This points out the need for a criterion based on amplitude rather than
amplification.

METHODS INCORPORATING AN AMPLITUDE CRITERION There is nothing novel about

amplitude criteria. Liepmann (1945) suggested many years ago that transition is
initiated when the ratio of disturbance Reynolds stress to the mean viscous stress

4 -,

Re lln. X 10-5
3.0 0
3
1. 5 a

1'....
0
)( 2
or
IX

a
1

- CALCULAT ION
o.o COL£ S (195....)

0 , I I I
1 2 3 4 5

Moo
Figure 20 Calculated effect of Mach number on the transition of insulated-flat-plate
boundary layers at (Re per inch) x 10 - 5 = 1.5 and 3.0 and comparison with Coles's (1954)
experiment (Mack 1975b).
 BOUNDARY-LAYER STABILITY AND TRANsmON 341

reaches some critical value. This idea was developed further by Van Driest & Blumer
(1963) and by Benek & High ( 1973).
In a pair of recent papers, Mack (1975a,b) estimates the variation of transition
Reynolds number by appending amplitude criteria to his prior calculations of
disturbance growth according to linear stability theory and incorporating initial­
disturbance information. Two examples of his results are described.
The first example (Mack 1975b) is that of the variation of transition Reynolds
number with free-stream Mach number for an insulated-fiat-plate boundary layer.
The spectrum of free-stream rms pressure fluctuation and its dependence on unit
Reynolds number play an essential role in this variation and must be included in the
calculation. With Laufer's (1964) measurements of rms-pressure spectra taken in the
JPL 20" supersonic tunnel as input, the disturbance of maximum amplitude is
calculated. The amplitude transition criterion chosen is that the rms pressure
fluctuation be 1 % of the free-stream static pressure and an arbitrary coupling
constant is chosen to give the measured (Coles 1954) transition Reynolds number
at M 1 = 4.5, Re per inch = 3 x 105• On the basis of this same coupling constant, a
series of calculations was carried out for M 1 = 2.2, 3.0, and 4.5 with 1 < (Re per
inch) x 10-5 < 4. The results are shown in Figure 20, where they are compared with

3."
'OItCING + INSTABILITY, AlAI . 50

3.0

INSTABILITY . SPECTRUM �
AlA, • 50, I./in, • I .. I "
2.6

�
tIS FORCING + INSTABILITY
2.2 + SPECTRUM
R./in. - I x loS
......
<
�
'-" A/AI - 50
-
< 1.8
�

1." 25

- - - -REF "
1.0 I I
0.7 0.8 0.9 1 .0

Figure 21 Five theoretical calculations of the effect of cooling on transition at M <Xl = 3.0
(Mack 1975a).
342 RESHOTKO

the Coles (1954) measurements taken in the same JPL 20" tunnel. Not only is there
reasonable agreement between calculated and experimental values but also the unit­
Reynolds-number effect seems to be reasonably accounted for, as seen from this figure
and other supporting information (Mack 1975b).
The second example (Mack 1975a) deals with the effect of TwlT"won Ret,/(Re/,)aw for
a noisy wind-tunnel environment. This again was done by calculating the growth of
the most unstable disturbance from its initial amplitude in the tunnel. The results of
Mack's various calculations are shown in Figure 21. The ordinate is the ratio of the
length to attain amplitude A with cooling to that without cooling. The abscissa is of
course the wall-temperature ratio Tw/T"w- The dashed line in the figure is the
experimental value of Re",t,/(Re"" ,).w of Van Driest & Blumer (1968) at Moo = 2.7
in the JPL 20" supersonic tunnel. The first calculation is that based on stability alone
to give AlA, 50, where A, is a fixed reference amplitude. The first improvement is
=

to account for the fact that as the boundary layer is cooled, the most unstable
frequency at a fixed unit Reynolds number shifts to a lower value and hence to a
region of the spectrum with increased energy. The spectra as measured by Kendall
(1975) at Moo = 3 in the aforementioned JPL tunnel are used by Mack in this
calculation, This result (instability + spectrum) is a slight improvement over the initial
calculation. The third calculation uses the forcing theory in the sense illustrated by
Figure 14. The resulting curve (forcing + instability, A/Al 50) is no improvement
:=

over the earlier result. Here Al is the amplitude of the mass-flow fluctuations in the
incoming sound wave.
The final calculations combine alI the elements of the previous ones : the forcing
theory, the stability theory, and the spectrum of free-stream disturbances. As seen in
Figure 21, a much reduced effect of cooling is obtained, and for AIAl = 25 the trend
is in close agreement with the measurements of Van Driest & Blumer (1968).8
These calculations give a hint of what is possible and also show the importance of
knowing the disturbance environment in a test facility. The second one in particular
is an initial example of the theoretical consideration of forcing mechanisms in a
transition context.

METHOD BASED ON TURBULENCE-MODEL EQUAnONS The procedures based on linear

stability theory do not account directly for the nonlinear effects that are physically
significant in the transition process.9 This prompted an alternative approach
(Yates & Donaldson 1972, Wilcox 1975) of using turbulence model equations to
calculate the downstream consequences of a disturbance introduced into the
laminar boundary layer. The disturbances are described by their intensity and scale.
Such procedures can also yield plausible reproduction of observed transition trends
but require adjustment of a number of "constants" from their values for turbulent
flow. Such procedures, however, are, unable to deal with the details of initiation of

9
B Note that all of these amplitude ratios are well below the e or e
10 levels suggested by
Smith & Gamberoni (1956) and by Jaffe, Okamura & Smith (1970).
9 A brief summary of these nonlinear effects was given for low-speed flows by Tani (1969).
 BOUNDARY-LAYER STABILITY AND TRANSITION 343

disturbances in the boundary layer, and even though they are basically nonlinear,
they do not describe the evolution of nonlinearity and three-dimensionality in a
boundary layer as evidenced, for example, in the measurements of Klebanoff,
Tidstrom & Sargent (1962). Until the turbulence-model-equation procedures are
better developed, it seems preferable to deal correctly with receptivity and initial
growth and develop an amplitude criterion either by physical argument or by
correlation.

Prediction of Transition in Flight

Whatever the difficulties of transition prediction in wind tunnels where disturbance
spectra are readily measurable, the rational prediction of transition Reynolds
numbers in free flight borders on the impossible because of the lack of information
on the disturbance environment in flight. The traditional ways of extrapolating wind­
tunnel data to flight conditions fail to account for the differences in disturbance
environment. For example, the extrapolation of a wind-tunnel result to flight unit
�) Reynolds numbers tends, according to equation (6), to assume the constancy of a
characteristic disturbance wavelength. There is no basis for such an assumption.
Morkovin (1969) indicated some effort in determining the distribution, intensity, and
00,
scales of disturbances at altitudes up to 2 000 feet. He suggests as an interim
working hypothesis that the disturbances at high altitudes have characteristics that
are no worse than those at 20,()(}()-40,OOO feet, where considerable information is
available in connection with commercial airline operation.
Any rational procedure for the prediction of transition should follow the processes
of Figure 18 as far along as is possible and then try to correlate transition with an
amplitude level or a level of distortion of the basic flow. The work of Mack (197Sa,b)
has demonstrated that such calculations are possible provided that there is adequate
input information.

6 FIXING OF TRANSITION

The achievement of earlier transition through artificial tripping of the boundary

layer is often desired in order to simulate turbulent-boundary-Iayer behavior at full­
scale Reynolds numbers. The testing literature is replete with descriptions of the
response of the boundary layer to varieties of surface roughness-single roughness
elements, multiple elements, spherical bodies, distributed roughness, etc. These trips
have been developed more or less by trial and error.
Stability considerations offer us both an explanation for the observed behavior
due to various kinds of trips as well as suggestions for more effective tripping (e.g.
Klebanoff & Tidstrom 1972). The general decay in spatial amplification rate with
Mach number (Figure 4) is probably responsible for the increasing difficulty of
tripping with increase in Mach number. Also the art of tripping has not really had
the chance to benefit from the recent documentation of greater first-mode instability
to oblique waves than to two-dimensional waves.
It is suggested that tripping devices be designed so as to capitalize on the known
instability characteristics of laminar boundary layers. A trip should generate oblique
344 RESHOTKO

waves of appropriate wavelength to be most effective at Mach numbers up to 4

(see Figure 4). Note Hama's success with "triangular patch stimulators" at Mach
numbers up to about 5 (Hama 1964). Beyond Mach number 4 it seems desirable to
excite the second mode for most efficient tripping. Furthermore, the trips need not be
mechanical. It is apparent that radiated sound at appropriate frequencies (Figure 3)
can have a noticeable effectiveness in promoting transition.

7 DIRECTIONS OF CURRENT AND FUTURE

INVESTIGATION

While the various sections of this paper presented the current transition work and its
antecedents, it is helpful to summarize briefly and point to the future in the areas of
basic mechanisms, transition prediction, and transition testing.

Basic M echanisms
The study of basic mechanisms relating to boundary-layer transition can be divided
into three categories : linear stability, receptivity, and no�linear processes.

LINEAR STABILITY Within this category, it is necessary to expand the catalog of

normal-modes results. This means additional results should be obtained through
exact numerical solution of appropriate sets of disturbance equations with experi­
mental corroboration where deemed desirable. So far most attention has been
given to zero-pressure-gradient cases with heating and cooling. Since within linear
stability considerations one studies the stability of a profile, anything that modifies
the profile changes the stability characteristics of a boundary layer. Factors to be
considered further are : pressure gradient, suction and blowing, bluntness,
longitudinal temperature variations; and three-dimensional effects such as those
introduced by angle of attack and sweep.

RECEPTIVITY The understanding of the coupling of radiated sound to a boundary

layer through the supersonic disturbances is fairly well understood. More must be
done, however, to acquire understanding for free-stream turbulence, entropy
disturbances, roughness, model vibrations, etc. In all cases including that of tunnel
sound, understanding of the mechanisms by which a forcing disturbance of a given
frequency and prescribed phase velocity excites a free disturbance of the same or
related frequency but different phase velocity is of great importance.

NONLINEAR PROCESSES The term nonlinear processes is meant to include all aspects
of behavior subsequent to the growth of infinitesimal disturbances. Some of the
important features of nonlinear growth are the changes in the frequency and .
orientation spectra of the disturbances through distortion of the mean flow,
generation of harmonics, mode-mode coupling, secondary instabilities, etc, leading
perhaps to turbulent-spot formation and eventual transition. Nonlinear processes are
not treated in any detail in this review. There is a meager theoretical literature (see
Stuart 1971), some carefully conceived and executed experiments at low speeds, and
much study of the turbulent spot at low speeds and "observation" of bursts, etc,
 BOUNDARY-LAYER STABILtrY AND TRANSITION 345

at high speeds. There seems to be no evidence of finite-amplitude metastable

configurations of the boundary layer. Hence when significant nonlinearity develops
and certainly by the time turbulent spots are formed, transition seems inevitable.
One cannot even begin to list the myriad of nonlinear phenomena that are waiting
to be identified and explained. But from the viewpoint of transition prediction, the
important contribution from nonlinear studies might be in the development of
amplitude criteria.

Transition Prediction
All transition-prediction theories, whether based on linear stability considerations or
on turbulence model equations, should have ways of incorporating input-dis­
turbance information. It is also desirable that the amplitude criterion developed for
and compatible with a given method have some physical and/or correlation basis.

Transition Testing
Experimental studies of stability and transition have always been fraught with
difficulty and have often .yielded anomalous data. Some of these have now been
identified with the disturbance environment of the test facility, particularly where
tunnel sound is involved. Some of the anomalies can be resolved on the basis of the
mode-selection argument advanced by Reshotko (1969). But by and large what is
required is a general improvement in experimental technique displaying sensitivity to
potential difficulties and care in overcoming them.

GUIDELINES Toward these ends, the Boundary Layer Transition Study Group
(Reshotko 1974, Reshotko 1975) was founded in 1970 under NASA auspices
and continues to function now under AEDC auspices. Among the guidelines
developed by the Group for the conduct of its work are the following :
1. Any effects specifically and only associated with test-facility characteristics must
be identified and if possible avoided. This points to emphasizing studies in
ballistic ranges and "quiet" tunnels.
2. Attention must be given to disturbances introduced by model surface, model
material, and internal structure. Experimental studies should include documenta�
tion of these various factors.
3. Details of coupling of disturbances of various kinds to the boundary layer must
be understood theoretically and experimentally, so that the sensitivity of the
transition process to the flight environment might be determined.
4. Whenever possible, tests should involve more than one facility. Tests should have
ranges of overlapping parameters, and whenever possible, experiments should
have redundancy in transition measurements.

These guidelines should come as no surprise following the exposition in this article.
The work of Mack (1975a) and Kendall (1975) already described herein is part of the
program of the Transition Study Group.
Another of the Group's programs is that reported by Owen et al (1975) of an
investigation undertaken to resolve differences in transition data obtained on two
similar 50-half-angle cone models in the Ames 3.5-foot hypersonic wind tunnel and
346 RESHOTKO

in the Langley 18" variable-density tunnel. By careful attention to instrumentation

and measurement, the differences were resolved. The final results agreed with
neither of the original sets of data, pointing out the fallibility of even experienced
investigators.
In still another of the Group's programs, Potter (1975) very carefully investigated
the effects of roughness, model vibration, nonuniform surface temperature and other
,

range-peculiar factors on sharp slender cones in the AEDC von Karman Facility
Ballistic Range K at Mach numbers 2 and 4.5. Despite the quiescence of the air on the
centerline of the range prior to arrival of the model and negative indications of the
range-peculiar factors on transition, the transition-Reynolds-number data display a
decided variation with unit Reynolds number. The resolution of this matter remains
a challenge to the Transition Study Group and to the transition community at large.

DEVELOPMENT Of QUIET WIND TUNNELS An important undertaking ofthe Transition

Study Group is the development of a new experimental facility. The "quiet" tunnel­
having laminar rather than turbulent boundary layers on the nozzle walls-averts a
prime cause of facility-induced transition at Mach numbers above 2.5. It was felt

15 i n
RODDED SOUND S H I ELD MODEL

RAP I D EXPANS I ON NOZZLE

BOUNDARY LAYER REMOVAL SLOT

NOZZLE APPROACH

15. 5 i n

12 i n

Figure 22 Mach-S Pilot Quiet Tunnel�slotted nozzle with rodded-wall sound shield
installed in test section (Beckwith 1975).
 BOUNDARY-LAYER STABILITY AND TRANSITION 347

desirable to develop a high-unit-Reynolds-number quiet tunnel for transition

research and also for the study of turbulent-boundary-layer development in the
absence of tunnel noise. As with other facilities, the proper utilization of a quiet
tunnel depends on identification and documentation of its disturbance environment.
A conception of such a facility as described by Beckwith (1975) is shown in Figure
22. The nozzle consists of a convergent approach·section followed by an annular slot
whose function is to remove the turbulent boundary layer that forms in the settling
chamber and nozzle approach, including the Taylor-Gortler vortices that would be
present due to the concave curvature in the approach. The slot lip is placed just
downstream of the inflection point in the local streamlines. The new boundary layer
that forms is therefore not subjected to concave curvature until far downstream of
the throat. The rapid expansion section expands the flow to the desired test-section
Mach number and a rodded sound shield enclosing the test section is installed at the
nozzle exit. The progress to date in the development of the throat section and sound
shield is described by Beckwith (1975) and by Harvey et al (1975).

FLIGHT PROGRAM If work in ground-test facilities is to have relevance to the

prediction of transition in atmospheric flight, then some carefully planned flight
tests within the guidelines of the Transition Study Group are in order.
It is hoped generally that the transition-testing community will, in all its work,
adhere as much as possible to the aforementioned guidelines. For it is only in this way
that there is the possibility of generating reliable and interpretable transition data.

8 CONCLUDING REMARKS

The process of transition from laminar to turbulent flow remains as baffling as the
turbulence in the subsequent flow. However, significant inroads into the understand­
ing of transition are now possible because we are presently able to do sophisticated
theoretical and experimental studies of the stability of laminar boundary layers. Some
of the anomalies of the past have now been explained, and a greater sensitivity has
been developed to the details of the instability and growth that are at the foundation
of transition. It is hoped that this review will stimulate additional critical studies of
boundary-layer transition and its elements.

ACKNOWLEDGMENTS

This work was sponsored by the Air Force Office of Scientific Research under
Grant AF-AFOSR-74-2577A and by the Office of Naval Research under Grant
NOOO1 4-75C-0446.

Literature Cited

Barry, M. D. J., Ross, M. A. S. 1970. J. Fluid Stability of Parallel Flows. New York :
Mech. 43 : 813-18 Academic
Beckwith, !' E. 1975. AIAA J. 13 : 300-6 Cheng, S. 1. 1953. Aero. Eng. Lab. Rep. 2 1 1 .
Benek, J. A., High, M. D. 1973. Arnold Eng. Princeton Univ., Princeton, NJ ; Q. Appl.
Dev. Cent. AEDC- TR-73-158 Math. 1 1 : 34(rSO
Betchov, R., Criminale, W. O. Jr. 1967. Coles, D. E. 1 954. J. Aeronaut. Sci. 21 : 433-48
348 RESHOTKO

Deem, R. E., Murphy, 1. S. 1965. AIAA Mack, L. M. 1967. JPL Space Progr. Summ.
Preprint 65-128
Dunn, D. W. 1953. On the stability of the
37-48. III : 167--69
Mack, L. M. 1969 . JfL Preprint 900-277
laminar boundary layer in a compressible Mack, L. M. 1970. JPL Space Progr. Summ.
jluid. PhD thesis. Mass. Inst. Techno!., 37-66 III : 13-16
Cambridge, Mass. Mack, L. M. 1971. Proc. Boundary Layer
Dunn. D. W., Lin, C. C. 1955. J. Aeronaut. Transition Workshop. Aerosp. Rep. TOR
Sci. 22 ; 455-77 01 72 IV : 1-1-1-35
Gaster, M. 1962. J. Fluid Mech. .14 : 222-24 Mack, L. M. 1975a. AlA A J. 1 3 : 307-14
Gaster, M. 1974. J. Fluid Mech. 66 ; 465-80 Mack, L. M. 1975b. In AerodynamiC Analyses
Gortler, H. 1940a. Nachr. Akad. Wiss. Requiring Advanced Computers. NASA
Gottingen Math-Phys. K/' IIA. Math.­ Spec. Publ. SP-34 7
Phys.-Chem.-Abt. 2 ; 1-26. Trans!. June Moore, F. K. 1956. In Advances i n Applied
1954 in NACA Tech. Memo. 1375 Mechanics. ed. G. Kuerti, IV : 159-228
Gortler, H. 1940b. Z. Angew. Math. Mech. Morkovin, M. V. 1969. Air Force Flight Dyn.
20 ; 138-47 Lab. Rep. AFFDL-TR-68-149
Gregory, N., Stuart, J. T., Walker, W. S. 1955. Nayfeh, A. H., Saric, W. S., Mook, D. T. 1974.
Phi/os. Trans. R. Soc. London Ser. A 248 : Arch. Mech. 26 : 401--6
1 55-99 Owen, F. K., Horstman, C. C., Stainback,
Hama, F. R. 1964. JPL Space Progr. Swnm. P. c., Wagner, R. D. 1975. AIAA J. 1 3 :
37-29 IV : 163--68 266--69
Harvey, W. D., Stainback, P. c., Anders, Owen, P. R., Randall, D. G. 1953. R.
J. B., Carey, A. M. 1 975. AIAA J. 13 : 307- Aeronaut. Estab. Rep. Aero. 330
14 Pate, S. R. 1969. Arnold Eng. Dev. Cent.
AEDC- TR-69-1 72
Pate, S. R., Schueler, C. 1. 1969. AIAA J.
Jaffe, N. A., Okamura, T. T., Smith, A. M. O.
1970. AIAA J. 8 : 301-8
Kaplan, R. E. 1964. Mass. Inst. Techno/. 7 : 45(}-57
Aeroelastic Struct. Res. Lab. Rep. ASRL­ Potter, J. L. 1968. AIAA J. 6 : 1907-1 1
TR- 1 1 6- 1 Potter, J. L. 1975. AIAA J. 1 3 : 27(}-77
Kendall, J. M. 1967. Proc. Boundary Layer Pretsch, J. 1942. Ber. Aerodyn. Versuchsanst.
Transition Study Group Meet. Air Force Gottingen E. V. Insc. Forschungsjlugbecrieb
Rep. BSD- TR-67-213. Aerosp. Rep. TR Flugwesen Jahrb. Dtsch. Luftfahrtforsch.
0158 (S 381 6-63)-1 U : Sect. l0 Trans!. 1952 in NACA Tech. Memo. 1343
Kendall, 1. M. 1970. JPL Space Progr. Summ. Reshotko, E. 1962. NASA Tech. Note TN
37-62 III : 43-47 D-1220
Kendall, J. M. 1971. Proc. Boundary Layer Reshotko, E. 1963. Phys. Fluids 6 : 335-42
Transition Workshop Aerosp. Rep. TOR- Reshotko, E. 1969. Al AA J. 7 : 1086--9 1
0172 IV : 2-1-2- 16 Reshotko, E. 1974. ICAS Pap. 74-13. Congr.
Kendall, 1. M. 1975. AlA A J. 1 3 : 29{}-99 . Int. Council Aeronaut. Sci., 9th. Haifa.
Klebanoff, P. S., Tidstrom, K. D. 1972. Israel. Aug. 1 974
Phys. Fluids 1 5 : 1 173-88 Reshotko, E. 1975. AIAA J. 1 3 : 261--65
Klebanoff, P. S., Tidstrom, K. D., Sargent, RogIer, H. L., Reshotko, E. 1975. SIAM J.
L. M. 1962. J. Fluid Mech. 1 2 : 1-34 Appl. Math. 28 : 431--62. Presented at
Laufer; J. 1954. J. Aeronaut. Sci. 21 : 497-98 Symp. Modern Dev. Fluid Dyn., Haifa,
Laufer, J. 1964. Phys. Fluids 7 : 1 191-97 Israel, Dec. 1973
Laufer, J., Vrebalovich, T. 1960. J. Fluid Ross, J. A., Barnes, F. H., Bums, J. G., Ross,
Mecho 9 : 257-99 M. A. S. 1970. J. Fluid Mech. 43 : 819-32
Lees, L. 1947. NACA Tech. Rep. 876 Sanator, R. J., DeCarlo, J. P., Torillo, D. T.
Lees, L., Gold, H. 1966. In Fundamental 1965. AlA A J. 3 : 758-60
Phenomena in Hypersonic Flow. ed. J. G. Saric, W. S., Nayfeh, A. H. 1975. Phys.
Hall. Ithaca, NY : Cornell Univ. Press Fluids 1 8 : 945-50
Lees, L., Reshotko, E. 1962. J. Fluid Mech. Schubauer, G. B., Skramstad, H. K. 1948.
12 : 555-90 NACA Tech. Rep. 909. Originally issued
Liepmann, H. W. 1943. NACA Wartime 1943 as NACA ACR
Rep. W-107 Sheetz, N. W. 1965. AJAA Preprint 65-127
Liepmann, H. W. 1945. NACA ACR 4J28 Shen, S. F. 1961. J. Aeronaut. Sci. 28 : 397-
Ling, C. H., Reynolds, W. C. 1973. J. Fluid 404, 417
Mech. 59 : 571-91 Smith, A. M. O. 1955. Q. Appl. Math. 1 3 :
Lowell, R. L. Jr., Reshotko, E. 1974. Case 233-,62
Western Reserve Univ. FTAS/TR-73-93 Smith, A. M. 0., Gamberoni, N. 1956.
Mack, L. M. 1965. AGARDograph 97 Pt. I : Douglas Aircraft Co. Rep. ES26388. EI
329--62 Segundo, Calif.
 BOUNDARY-LAYER STABILITY AND TRANsmON 349

Squire, H. B. 1933. Proc. R. Soc. London Wagner, R. D., Maddalon, D. V., Weinstein,
Ser. A 142 : 621-28 L. M. 1970. AIAA J. 8 : 1664-70
Stetson, K. F., Rushton, O. H. 1967. AIAA J. Wazzan, A. R., Okamura, T. T., Smith,
5 : 899-906 A. M. O. 1968. J. Heat Transfer 90 : 109-14
Strazisar, A., Prahl, 1. M., Reshotko, E. 1975. Wazzan, A. R., Okamura, T. T., Smith,
Case Western Reserve Univ. Dep. Fluid A. M. O. 1970. Proc. Int. Heat Transfer
Thermal Aerosp. Sci. FTAS/TR-75-113 Conf.. 4th. 2 : FC1.4
Stuart, J. T. 1971. Ann. Rev. Fluid Mech. Wazzan, A. R., Taghavi, H., Keltner, O.
3 : 347-70 1974. Phys. Fluids 1 7 : 1655-60
Tani, I. 1969. Ann. Rev. Fluid Mech. 1 : 1 69- Wells, C. S. Jr. 1967. AI AA J. 5 : 172-74
96 Wilcox, D. C. 1975. AIAA J. 13 : 241-43
Taylor, O. I. 1923. Phi/os. Trans. R. Soc. Wygnanski, I. J., Champagne, F. H. 197J.
London Ser. A 233 : 289-343 J. Fluid Mech. 59 : 28 1-335
Van Driest, E. R., Blumer, C. B. 1963. AIAA Yang, K. T., Kelleher, M. D. 1964. Univ.
J. 1 : 1303-6 Notre Dame. Dep. Mech. Eng. Tech. Note
Van Driest, E. R., Blumer, C. B. 1968. AlA A 64-22
J. 6 : 603-7 Yates,J. E., Donaldson, C. D. 1972. Air Force
Van Ingen, J. L. 1 956. Dep. Aeronaut. Eng. Flight Dyn. Lab. AFFDL- TR-72-123.
Inst. Technol.. Delft. Holland. Rep. VTH Wright Patterson Air Force Base, Ohio
74