The Origin of Turbulent Spots

It has been suggested that a turbulent spot is formed when a transient separation occurs
M. W. Johnson
A. Dris
Department of Engineering,
The University of Liverpool,
Liverpool L69 3GH, United Kingdom
in the laminar boundary layer and this criterion has been successfully used by Johnson
and Ercan (1996, 1997) to predict bypass transition for boundary layers subjected to a
wide range of free-stream turbulence levels and streamwise pressure gradients. In the
current paper experimental results are presented that support the premise that the
formation of turbulent spots is associated with transient separation. Near-wall hot-wire
signals in laminar and transitional boundary layers are analyzed statistically to produce
probability distributions for signal level and trough frequency. In the laminar period the
signal level is normally distributed, but during the inter-turbulent periods in the transi-
tional boundary layer, the distribution is truncated at the lower end, i.e., the lowest
velocity periods in the signal disappear, suggesting that these are replaced during
transition by the turbulent periods. The number of these events (troughs) also correlates
with the number of turbulent spots during early transition. A linear perturbation theory is
also used in the paper to compute the streamlines through a turbulent spot and its
associated calmed region. The results indicate that a hairpin vortex dominates the flow
and entrains a low-momentum fluid stream from upstream with a high-momentum stream
from downstream and then ejects the combined stream into the turbulent spot. The hairpin
can only exist if a local separation occurs beneath its nose and the current results suggest
that this separation is induced when the instantaneous velocity in the near-wall signal
drops below 50 percent of the mean.

Introduction waves possess a shorter wavelength (typically 2 to 3 ␦) and also

The turbulent spot can be considered as the “building block” of
a turbulent flow. In the transition of a laminar boundary layer to a
turbulent one, the turbulent spot plays a pivotal role. The first
appearance of turbulent spots determines the start of transition
location and the subsequent growth of the spots dictates the length
of the transition region prior to a fully turbulent boundary layer
being achieved. Emmons (1951) first recognized the intermittent
nature of transitional flow and the role of the turbulent spot in the
transition process. Narasimha (1957) measured the variation in
intermittency through transition and showed that this was most
accurately represented by a “concentrated breakdown” model, i.e.,
a model in which all the turbulent spots are initiated at the same
streamwise location and the increase in intermittency through
transition results purely from the growth in size of these spots.
More recently, with the availability of more accurate measurement
techniques for intermittency, the validity of this model has been
challenged. Johnson and Fasihfar (1994) showed that a concen-
trated breakdown model did not give a good prediction of the
evolution of spot statistics through transition and that a “distrib-
uted breakdown” model, i.e., one where spots are initiated at
different streamwise locations resulted in improved prediction.
Johnson (1994) and Johnson and Ercan (1996, 1997) and more
recently Mayle and Schultz (1997) and Mayle et al. (1998) have
shown that the pretransitional boundary layer near wall velocity
fluctuations, induced primarily by pressure fluctuations associated
with the free-stream turbulence, grow more or less linearly in the
streamwise direction. These fluctuations are of long wavelength
(greater than 15 to 20 boundary layer thicknesses), but can be
many times greater in amplitude (typically 20 to 100) in the
near-wall region than in the free-stream. When these fluctuations
reach some critical level (approximately a local turbulence level of
23 percent according to Johnson (1994)) significant numbers of
turbulent spots are induced and transition commences. Natural
transition induced through Tollmien–Schlichting waves also com-
mences when the waves reach a similar critical amplitude. T–S
Contributed by the International Gas Turbine Institute and presented at the 44th
International Gas Turbine and Aeroengine Congress and Exhibition, Indianapolis,
Indiana, June 7–10, 1999. Manuscript received by the International Gas Turbine
Institute February 1999. Paper No. 99-GT-32. Review Chair: D. C. Wisler.
grow exponentially once the critical boundary layer Reynolds
number is exceeded. However, it would appear that the criterion
for the amplitude of fluctuation that induces a turbulent spot is
common to both modes of transition.
The structure of turbulent spots has been studied extensively
through both experiment (e.g., Cantwell et al. (1978); Seifert and
Wygnanski (1994); Gostelow et al. (1995) and numerically (Ber-
tolotti et al. (1992)); however, the mechanism leading to the
initiation of the spot and turbulence production is not clear. Var-
ious mechanisms for the generation of turbulence in a boundary
layer have been proposed. Sirovich and Karlsson (1997) believe
that turbulent bursts are generated by the interaction of spanwise
vorticity and oblique waves induced by the low frequency un-
steadiness. On the other hand, Smith et al. (1991) attribute the
turbulent bursts directly to the spanwise vorticity, produced
through the unsteadiness, which induces transient separation of the
boundary layer flow. This mechanism has also been proposed by
one of the current authors (Johnson, 1994).
The objective of the current paper is to investigate through both
experiment and theory the process of turbulent spot initiation.
Theoretical Model for Spot Initiation
Johnson (1994) considered the effect of the fluctuating pressure
field resulting from free-stream turbulence on the near-wall veloc-
ity profile within the laminar boundary layer. Experimental obser-
vations (Johnson and Ercan, 1996) of the near-wall velocity fluc-
tuations within laminar boundary layers show that the dominant
wavelengths are much greater than the boundary layer thickness.
Similar results have since been obtained theoretically by Mayle
and Schultz (1997). If this is the case, it is reasonable to assume
that the effect of the pressure field close to the wall is restricted to
two dimensions (i.e., the x–y plane) and that there will be negli-
gible phase shift over short distances from the wall. Experimental
observations also indicate that near the wall the turbulence level is
constant and hence as the mean velocity is proportional to y, it
follows that the instantaneous velocity must also be proportional to
y. With the assumptions so far made, a streamline in the near wall
region will be as shown in Fig. 1. In the unperturbed flow, the
streamline at a height y from the wall will carry fluid at a velocity
u. When the flow is perturbed by a local reduction in pressure,

88 / Vol. 122, JANUARY 2000 Copyright © 2000 by ASME Transactions of the ASME

Downloaded 10 Mar 2012 to 145.94.53.183. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm

Fig. 1 Near-wall streamlines perturbed through the pressure field in-
duced by the free-stream turbulence
This relation indicates that the minimum value of the measured
velocity u ⫹ u⬘ is 21 u when p( x, y) ⫽ 41 ␳ u 2 . If p( x, y) exceeds
this value, the streamline no longer has an equilibrium position and
an instability occurs. In fact, as the streamline is deflected outward,
the local pressure increases, decelerating the fluid on the stream-
line still further, resulting in further deflection of the streamline
from the wall, until ultimately the fluid stalls and flow separation
results. This simple model therefore predicts that when a pressure
perturbation induces the near-wall velocity to drop below 50
percent of the unperturbed local velocity, an instability arises,
which results in local separation of the flow. It is the normal
velocity v⬘ associated with this separation that is believed to
initiate the turbulent spot.

however, the fluid will accelerate and will also move closer to the Experimental Work
wall (to satisfy 2-d continuity). Conversely, when the pressure The experimental measurements were made in the boundary
increases, the streamline will move away from the surface. Along layer wind tunnel in the Department of Engineering at the Uni-
a streamline, assuming that the perturbation is time invariant and versity of Liverpool. A detailed description can be found in Fasih-
viscosity can be neglected, far (1992). The flat plate has a length and width of 1.2 m and
0.7 m, respectively and a flow velocity of 30 m/s, which was
p 0 共y兲 ⫽ p共x, y s 兲 ⫹ 12 ␳ u s2 , (1)
constant along the plate, was used for all the current experiments.
where the pressure field is assumed to be a function only of the A turbulence generating grid, placed 0.75 m upstream of the plate
free-stream perturbations and not the local conditions and u s is the leading edge, induced a free-stream turbulence level of 1 percent
velocity of fluid on the streamline at a height y s above the wall. at the plate leading edge. A Dantec 55C01 anemometer and 55P15
Continuity also requires that probe were used for the boundary layer measurements. The signal
was digitized over a sample length of 15 seconds using a sampling
uy ⫽ u s y s (2) frequency of 10 kHz. The signals were linearized using a digital
“look-up” calibration table. At each of 12 streamwise x locations,
Now with a general waveform w( x) for the perturbing pressure a boundary layer traverse consisting of 50 measurement points was
field made in order that the boundary layer integral parameters could be
1
␳
p共x, y s 兲 ⫽ ay s2 w共x兲 ⫽ a
uy
us 冉 冊 2
w共x兲 (3)
evaluated. One of the near-wall measurement points, which lay at
approximately y/ ␦ ⫽ 0.1, was selected for the signal analysis. The
turbulent and inter-turbulent periods and hence the intermittency
where a is a constant. Hence from Eq. (1), were determined using the method of Fasihfar and Johnson (1992),
whereby the signal was high pass filtered to remove the low-
u2
2
⫽a
uy
us 冉 冊 2
w共x兲 ⫹
u s2
2
(4)
frequency laminar fluctuations. The turbulent periods were then
defined as those periods when the signal amplitude exceeded a
specified threshold value. Full details, including the choice of the
and so, taking the root of this quadratic, which meets the require- filter setting and threshold are given in Fasihfar and Johnson
ment that u s ⫽ u when w( x) ⫽ 0, (1992). In the current work, the unfiltered signal was then analyzed

冉冊 冑
in the turbulent and inter-turbulent periods to evaluate probability
2
us 1 1 2ay 2 w共x兲 distributions for signal level and trough frequency.
⫽ ⫹ ⫺ (5)
u 2 4 u2
However, the velocity u ⫹ u⬘ measured at the fixed height y is Experimental Results
given by Boundary Layer Development. Figure 2 shows the boundary
yu s u s2 layer development along the plate. Laminar flow is maintained up
u ⫹ u⬘ ⫽ ⫽ (6) to Re x ⫽ 900,000, at which point the intermittency starts to
ys u increase, indicating that transition has commenced. In the current
and hence work transition is not fully completed at the last measurement
station. The near-wall local turbulence level increases approxi-
u⬘
u
1
⫽⫺ ⫹
2 冑 1 2p共x, y兲
4
⫺
␳u 2 冉冊 us
u
2
(7)
mately linearly through the laminar layer to a value close to 23
percent at start of transition, as discussed by Fasihfar and Johnson
(1992), and reaches a peak at 40 percent in the transition period

Nomenclature
a ⫽ constant in Eq. (3) T ⫽ Ut/ ␦ ⫽ dimensionless time v⬘ ⫽ normal fluctuating velocity
p ⫽ instantaneous static pressure u ⫽ time-averaged local velocity w( x) ⫽ waveform
p o ⫽ time invariant total pressure on a u៮ lam ⫽ time-averaged inter-turbulent pe- x, y, z ⫽ streamwise, normal, and span-
streamline riod velocity wise coordinates
Re ⫽ U ␦ / ␯ ⫽ boundary layer thickness u៮ tur ⫽ time-averaged turbulent period X, Y, Z ⫽ x/ ␦ , y/ ␦ , z/ ␦ ⫽ dimension-
Reynolds number velocity less coordinates
Re x ⫽ Ux/ ␯ ⫽ Reynolds number based u⬘ ⫽ fluctuating velocity y s ⫽ normal distance from wall to
on streamwise distance u i ⫽ u ⫹ u⬘ ⫽ instantaneous velocity streamline
Re ␪ ⫽ U ␪ / ␯ ⫽ boundary layer momen- u s ⫽ velocity on a streamline ␦ ⫽ boundary layer thickness
tum thickness Reynolds number U ⫽ free-stream velocity ␳ ⫽ fluid density
t ⫽ time

Journal of Turbomachinery JANUARY 2000, Vol. 122 / 89

Downloaded 10 Mar 2012 to 145.94.53.183. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 0.9 and hence the whole distribution curve is shifted to the left.
This asymmetry becomes greater as the transition proceeds with
very low probabilities below 50 percent of the local inter-turbulent
mean velocity at any streamwise location. This observation there-
fore lends strong experimental support to the suggestion by John-
son and Ercan (1996, 1997) that turbulent spots are induced when
the near-wall instantaneous velocity drops below 50 percent of the
mean, in that it is these parts of the signal which apparently vanish
from the inter-turbulent periods. Beyond Re x ⫽ 1,340,000, a
second peak resulting from the turbulent periods, which are asso-
ciated with a higher mean velocity, develops at a u i /u៮ lam ⬎ 1. This
peak becomes dominant as the end of transition is approached and
also adopts a normal distribution profile about a mean level of
u៮ tur/u៮ lam.

Trough Frequency. The trough frequency is defined as the

frequency at which minima occur in the signal. In the present
work, the distribution of the trough depths is computed within
the laminar or inter-turbulent period as the frequency with
which minima occur within each bin. This frequency is then
divided by the bin size to remove the effect of the number of
bins chosen. Frequency distributions for the troughs (Fig. 4)
again show that the instantaneous velocity rarely drops below
50 percent of the mean even though significant frequencies are
detected over 200 percent of the mean. The total trough rate in
the inter-turbulent period (Fig. 5) remains constant at about 800
Hz as the boundary layer develops.
If each time the instantaneous velocity drops below the 50
percent threshold, a turbulent spot is induced, the number of
observed threshold events should be equal to the number of spots
induced. Figure 5 shows that during the early part of transition, the
spot rate (⫽ number of turbulent periods/sample time) is indeed
approximately equal to the rate at which troughs disappear into the
Fig. 2 Boundary layer development

before decreasing to a level of about 25 percent as the end of

transition is approached. The reason for the peak at midtransition
is primarily because of the difference in the mean velocities in the
turbulent and inter-turbulent periods, and the local turbulence
levels within each period considered separately are considerably
lower, as discussed by Fasihfar and Johnson (1992).
Signal Level Probability. The significance of the pretransi-
tional development of the near-wall velocity fluctuations in the
triggering of turbulent spots has been recognized for a number of
years (Johnson (1994) and Mayle and Schultz (1997)). The mech-
anism through which the spots are induced has not clearly been
identified, however, and it was with this objective in mind, that the
near-wall hot-wire signals at y/ ␦ ⫽ 0.1 were analyzed to try to
identify those parts of the signal responsible for inducing spots.
The distribution of occurrence level for each sample (Fig. 3) was
determined by first nondimensionalizing the instantaneous veloc-
ities with the inter-turbulent mean velocity. The range of observed
instantaneous velocities between the minimum and maximum
value was then divided into approximately 25 equal subranges or
bins. The sample, which consisted of 150,000 instantaneous ve-
locity measurements, was then analyzed to determine the propor-
tion of the total sample time spent in each subrange. The results
were then divided by the bin size, such that the choice of the
number of bins did not affect the computed probability. The spread
of observed values increases necessarily as the rms fluctuation
level increases with distance downstream, as shown in Fig. 3. The
probability curve is almost symmetric in the laminar period up to
Re x ⫽ 900,000, but at the first measurement station within the
transition region, Re x ⫽ 1,030,000, the symmetry is lost. Al-
though there is a significant probability that the velocity will
exceed 150 percent of the local mean, the probability is almost
negligible that it will be below 50 percent. The loss of the low-
velocity events also increases the inter-turbulent mean u៮ lam, which
results in the most probable signal level u i /u៮ lam reducing from 1 to Fig. 3 Signal level probabilities

90 / Vol. 122, JANUARY 2000 Transactions of the ASME

Downloaded 10 Mar 2012 to 145.94.53.183. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm

Fig. 4 Trough frequency distributions
turbulent period (⫽ intermittency ⫻ trough rate). This is not the
case farther downstream, however, as the number of observed
spots will be less than the total number induced as the spots will be
less than the total number induced as the spots will begin to merge
with their neighbors as they grow in size. Nevertheless, the result
shown in the figure confirms that the induction of a spot is
associated with the disappearance of a trough in the near-wall
velocity signal.
Concentrated or Distributed Breakdown. Whatever the cri-
terion for the induction of turbulent spots, for concentrated break-
down to occur, the criterion can only be satisfied over a very short
streamwise distance (Narasimha, 1957). This would seem to ne-
Fig. 5 Minima and burst rates along the plate
Journal of Turbomachinery
Downloaded 10 Mar 2012 to 145.94.53.183. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
glect the statistical nature of the free-stream turbulence responsible
for inducing bypass transition. Consider, for example, a steady
flow over a flat plate, but suppose that the free-stream turbulence
level varies with time. Assuming the variation is made fairly
slowly, we would expect the transition location to move down-
stream as the turbulence level is decreased and to move upstream
as it is increased. This is not dissimilar from a wake-induced
transition experiment (e.g., Halstead et al., 1995) where periods of
high free-stream turbulence in the wakes result in large diversions
upstream in the transition location. This is an extreme example of
what happens when we have a nominally constant free-stream
turbulence level. If the turbulence level were measured over peri-
ods of only a few wavelengths, then its magnitude would be found
to vary considerably, because of its statistical nature. This varia-
tion would thus lead to a variation in the location at which
turbulent spots are generated and hence to a “distributed break-
down.” In the case of natural transition the amplitude of the
Tollmien–Schlichting waves also varies with time, but they grow
very rapidly in the streamwise direction and hence a large number
of spots will be induced over a relatively short streamwise distance
and thus can be modeled approximately as concentrated break-
down. However, for bypass transition the growth in amplitude is
approximately linear and hence the spot initiation sites will have a
significant streamwise distribution. This effect is apparent in most
measured intermittencies (e.g., Gostelow and Walker, 1991) where
the intermittency consistently exceeds the Narasimha concentrated
breakdown model value by between 2 and 5 percent in the early
transition period, but is accurately modeled by a distributed break-
down model (Johnson and Ercan, 1997). In the current experi-
ments occasional spots (at a frequencies less than 1 Hz and which
result in intermittencies less than 0.01 percent) were observed at
measurement stations in the laminar boundary layer Re x ⬍
900,000. These very early turbulent spots have a negligible effect
on the time mean boundary layer parameters and are generally
ignored by researchers, but their existence is evidence of occa-
sional extraordinary low troughs in the near-wall velocity signal,
which induce occasional turbulent spots in the laminar boundary
layer.
Theoretical Work
The theoretical results described in the current paper were
obtained using the method described by Johnson (1998a, 1999),
which is similar to the method of Li and Widnall (1989). A steady
nondeveloping boundary layer flow is perturbed by a fully three-
dimensional viscid linear perturbation. The method is therefore
only strictly accurate for small linear perturbations and turbulence,
which results from nonlinear perturbations, is not predicted. Nev-
ertheless, Johnson (1998b) showed that the geometric development
of the linear perturbation region was very similar to that of the
measured (and hence nonlinear) development of a turbulent spot
and its associated calmed region.
Figure 6 shows the streamlines through the perturbed region at
T ⫽ Ut/ ␦ ⫽ 60 for Re ⫽ 4000 (Re ␪ ⫽ 470) as seen by an observer
travelling at 50 percent of the free-stream velocity (approximately
the average velocity of the perturbation region). It should be noted
therefore that fluid with a velocity less than 0.5 U (i.e., Y ⬍ 0.25)
is seen by this moving observer to approach the perturbed region
from the front, whereas fluid with a greater velocity (Y ⬎ 0.25) is
seen to approach from the rear. The Y axis in the figure has been
scaled by a factor of 5 and streamlines originating from each Y
level are staggered by ⌬Y ⫽ 1 for clarity. In the figure all the
streamlines originating above Y ⫽ 0.1 travel around the core of a
hairpin vortex at X ⫽ 30. The streamlines beneath this level pass
beneath the nose of the hairpin vortex and are deflected along its
legs around the calmed region, which is bounded by the vortex.
These streamlines are in fact within the viscous sublayer (Y ⫽ 0.1
corresponds to y ⫹ ⫽ 9 at this Reynolds number). The streamlines
originating at Y ⫽ 0.2 are turned by the hairpin vortex through
180 deg to move away from the surface to mix with fluid carried
JANUARY 2000, Vol. 122 / 91

Fig. 6 Streamlines through the perturbed region
by streamlines originating behind the perturbed region at Y ⬎
0.25. The streamlines originating from Y ⫽ 0.4 and 0.5 dip
toward the wall through the calmed region to replace the fluid
ejected by the hairpin (Y ⫽ 0.2 streamlines). This results in an
increase in the boundary layer momentum in the calmed region
and hence a local thinning of the boundary layer. The streamlines
from Y ⫽ 0.4 and 0.5 pass through the calmed region to a point
below the nose of the hairpin vortex. At this point, the fluid slows
rapidly because of the increasing shear due to the wall below and
the hairpin vortex above. The fluid is then entrained by the hairpin
and is ejected forward along a similar path to the streamlines
originating from Y ⫽ 0.2. The fact that the fluid streams origi-
nating from upstream (Y ⫽ 0.2) and downstream (Y ⫽ 0.4 and
Y ⫽ 0.5) have significantly different energy levels will lead to
high local shear rates as the two streams mix out in the hairpin
vortex and forward of it following ejection and this leads to the
turbulence, which identifies this region as the turbulent spot.
In the current work, the region where the high-momentum
stream from downstream (streamlines from Y ⫽ 0.3, 0.4, and 0.5)
meets the low-momentum stream from upstream (streamlines from
Y ⫽ 0.1 and 0.2), is of particular interest. Where these two streams
meet, the high-momentum stream bifurcates with one part con-
tinuing to the spot and the remainder turning through 180 deg into
the calmed region. The bifurcation occurs along a line between the
calmed region and the hairpin vortex, which forms an open loop.
The front of the loop lies beneath the nose of the hairpin vortex and
travels with it (i.e., at approximately 50 percent free-stream ve-
locity); however, the legs of the loop trail behind the nose and end
on the wall at separation points. It is therefore a necessary condi-
tion, for the establishment of the hairpin and the turbulent spot and
its associated calmed region, that the boundary layer is perturbed
sufficiently to induce a local separation of the flow.
The hairpin vortex and bifurcation point have also been ob-
served in experiments (e.g., Cantwell et al., 1978).
Conclusions
1 A simple analytical model is presented, which suggests that
the pressure fluctuations induced by the free-stream turbulence
perturb the near-wall velocity in the laminar layer. If the instan-
taneous velocity is reduced to 50 percent of the unperturbed
velocity, transient separation occurs and a turbulent spot is initi-
ated.
92 / Vol. 122, JANUARY 2000
Downloaded 10 Mar 2012 to 145.94.53.183. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
2 Probabilities for the near-wall signal level in the laminar and
inter-turbulent periods show that instantaneous velocities below
approximately 50 percent of the mean are negligibly small even
though velocities in excess of 200 percent are present. This obser-
vation appears to confirm that occurrences of instantaneous veloc-
ities below this threshold are responsible for the initiation of
turbulent spots. This hypothesis is also supported by the fact that
the occurrence rate of such events (troughs) is approximately equal
to the burst rate within the early transition period. In mid and late
transition, merging of spots occurs and hence the burst rate be-
comes lower than the trough rate.
3 A linear perturbation model is used to predict the stream-
lines through a turbulent spot and its associated calmed region. The
results show that the hairpin vortex is responsible for sustaining
the turbulent spot and that the vortex depends on the existence of
a local separation of the flow. The separation is induced when the
local instantaneous velocity drops below 50 percent of the mean.
References
Bertolotti, F. P., Herbert, Th., and Spalart, P. R., 1992, “Linear and Non-linear
Stability of the Blasiuis Boundary Layer,” Jnl. of Fluid Mechanics, Vol. 242, pp.
441– 474.
Cantwell, B., Coles, D., and Dimotakis, P., 1978, “Structure and entrainment in the
plane of symmetry of a turbulent spot,” Jnl. of Fluid Mechanics, Vol. 87, pp.
641– 672.
Emmons, H. W., 1951, “The Laminar Turbulent Transition in a Boundary Layer—
Part 1,” Jnl. of Aero. Sciences, Vol. 18, pp. 490 – 498.
Fasihfar, A., 1992, “Mechanisms of boundary layer transition,” Ph.D. thesis,
University of Liverpool.
Fasihfar, A., and Johnson, M. W., 1992, “An improved boundary layer transition
correlation,” ASME Paper No. 92-GT-245.
Gostelow, J. P., Melwani, N., and Walker, G. J., 1995, “Effects of Streamwise
Pressure Gradient on Turbulent Spot Development,” ASME JOURNAL OF TURBOMA-
CHINERY, Vol. 118, pp. 737–743.
Gostelow, J. P., and Walker, G. J., 1991, “Similarity behavior in transitional
boundary layers over a range of adverse pressure gradients and turbulence levels,”
ASME JOURNAL OF TURBOMACHINERY, Vol. 113, pp. 617– 625.
Halstead, D. E., Wisler, D. C., Okiishi, T. H., Walker, G. J., Hodson, H. P., and
Shin, H.-W., 1997, “Boundary layer development in axial compressors and turbines.
Part 4. Computations and Analyses,” ASME JOURNAL OF TURBOMACHINERY, Vol. 119,
pp. 128 –139.
Johnson, M. W., 1994, “A bypass transition model for boundary layers,” ASME
JOURNAL OF TURBOMACHINERY, Vol. 116, pp. 759 –764.
Johnson, M. W., and Fasihfar, A., 1994, “Properties of turbulent bursts in transi-
tional boundary layers,” Int. Jnl. of Heat and Fluid Flow, Vol. 15, No. 4, pp. 283–290.
Johnson, M. W., and Ercan, A. H., 1996, “A boundary layer transition model,”
ASME Paper No. 96-GT-444.
Johnson, M. W., and Ercan, A. H., 1997, “Predicting bypass transition: A physical
model versus empirical correlations,” ASME Paper No. 97-GT-475.
Johnson, M. W., 1998a, “The structure of turbulent spots,” submitted for journal
publication.
Johnson, M. W., 1998b, “Turbulent spot characteristics in boundary layers sub-
jected to streamwise pressure gradient,” ASME Paper No. 98-GT-124.
Johnson, M. W., 1999, “Prediction of turbulent spot growth rates,” ASME Paper
No. 99-GT-31.
Li, F., and Widnall, S. E., 1989, “Wave patterns in plane Poiseuille flow created by
concentrated disturbances,” Jnl. of Fluid Mechanics, Vol. 208, pp. 639 – 656.
Mayle, R. E., and Schultz, A., 1997, “The path to predicting bypass transition,”
ASME JOURNAL OF TURBOMACHINERY, Vol. 119, pp. 405– 411.
Mayle, R. E., Dullenkopf, K., and Schultz, A., 1998, “The turbulence that matters,”
ASME JOURNAL OF TURBOMACHINERY, Vol. 120, pp. 402– 409.
Mayle, R. E., 1999, “A theory for predicting the turbulent spot production rate,”
ASME JOURNAL OF TURBOMACHINERY, Vol. 121, pp. 588 –593.
Narasimha, R., 1957, “On the Distribution of Intermittency in the Transition
Region of a Boundary Layer,” J. of Aero. Sciences, Vol. 24, pp. 711–712.
Seifert, A., and Wygnanski, I. J., 1994, “On Turbulent Spots in a Laminar
Boundary Layer subjected to Self-Similar Adverse Pressure Gradient,” Jnl. of Fluid
Mechanics, Vol. 296, pp. 185–209.
Sirovich, L., and Karlson, S., 1997, “Turbulent drag reduction by passive mecha-
nisms,” Nature, Vol. 388, pp. 753–755.
Smith, C. R., Walker, J. D. A., Haidari, A. H., and Sobrun, U., 1991, Phil. Trans.
R. Soc. Lond. A336, pp. 131–175.
Transactions of the ASME