You are on page 1of 10

PHYSICS OF FLUIDS VOLUME 13, NUMBER 3 MARCH 2001

A comparative study of near-wall turbulence in high and low Reynolds


number boundary layers
M. M. Metzgera) and J. C. Klewicki
Physical Fluid Dynamics Laboratory, Department of Mechanical Engineering, University of Utah,
Salt Lake City, Utah 84112
共Received 8 March 2000; accepted 8 December 2000兲
The present study explores the effects of Reynolds number, over three orders of magnitude, in the
viscous wall region of a turbulent boundary layer. Complementary experiments were conducted
both in the boundary layer wind tunnel at the University of Utah and in the atmospheric surface
layer which flows over the salt flats of the Great Salt Lake Desert in western Utah. The Reynolds
numbers, based on momentum deficit thickness, of the two flows were R␪ ⫽2⫻103 and R␪ ⬇5
⫻106 , respectively. High-resolution velocity measurements were obtained from a five-element
vertical rake of hot-wires spanning the buffer region. In both the low and high R␪ flows, the length
of the hot-wires measured less than 6 viscous units. To facilitate reliable comparisons, both the
laboratory and field experiments employed the same instrumentation and procedures. Data indicate
that, even in the immediate vicinity of the surface, strong influences from low-frequency motions at
high R␪ produce noticeable Reynolds number differences in the streamwise velocity and velocity
gradient statistics. In particular, the peak value in the root mean square streamwise velocity profile,
when normalized by viscous scales, was found to exhibit a logarithmic dependence on Reynolds
number. The mean streamwise velocity profile, on the other hand, appears to be essentially
independent of Reynolds number. Spectra and spatial correlation data suggest that low-frequency
motions at high Reynolds number engender intensified local convection velocities which affect the
structure of both the velocity and velocity gradient fields. Implications for turbulent production
mechanisms and coherent motions in the buffer layer are discussed. © 2001 American Institute of
Physics. 关DOI: 10.1063/1.1344894兴

I. INTRODUCTION ization refers to normalization by ␯ and the friction velocity


u ␶ (⬅ 冑␶ w / ␳ ) with ␳ and ␶ w symbolizing the mass density
As indicated in the recent review by Gad-el-Hak and and wall shear stress, respectively.
Bandyopadhyay,1 most of what is known concerning the Several interrelated factors exacerbate the task of quan-
characteristics of the viscous wall region of turbulent bound- tifying the structure of turbulent boundary layers at high R␪ .
ary layers has been derived from experiments and/or numeri- One factor stems from the significant demands placed on
cal simulations at low to moderate Reynolds number, R␪ experimental facilities designed for high Reynolds number
⬍5000 (R␪ ⬅U ⬁ ␪ / ␯ , where U ⬁ , ␪ , and ␯ denote the free boundary layer studies. Specifically, the production of a high
stream velocity, momentum deficit thickness, and kinematic Reynolds number boundary layer requires either a high free
viscosity, respectively兲. Numerous technologically important stream velocity, a thick boundary layer, a low kinematic vis-
cosity, or a suitable combination of the three. Because the
flows, most notably those within the atmosphere, are at con-
generation of a thick boundary layer demands a very long
siderably higher Reynolds numbers, R␪ ⬎100 000. Because
wind tunnel with a substantial development length, exacting
of this fact, there is significant motivation to obtain improved high construction and maintenance expenses, most high Rey-
instantaneous and statistical descriptions of the near-wall nolds number facilities compromise development length in
motions affecting the momentum and scalar transport pro- favor of higher flow speeds. Notable examples include the
cesses in high R␪ flows. These improved descriptions are National Transonic Facility at NASA Langley and the Na-
required to resolve scaling issues, advance physical models, tional Diagnostic Facility at the Illinois Institute of Technol-
and refine numerical models such as those associated with ogy. For a fixed development length 共i.e., fixed test section
large eddy simulations. The extrapolation of low R␪ results length兲, increasing the external flow speed generally results
to flows at high R␪ can be problematic. For example, contro- in a reduction of the boundary layer thickness ␦ . In the case
versy continues to exist regarding whether the peak value in of the Princeton University ‘‘Superpipe,’’ operating pres-
the inner normalized fluctuating streamwise velocity profile sures above atmospheric are utilized to decrease the kine-
is independent of Reynolds number,1–3 where inner normal- matic viscosity, effectively increasing R␪ . A decrease in ei-
ther ␦ or ␯ translates into a reduction in spatial resolution of
any fixed sized probe. Furthermore, as R␪ increases, the
a兲
Electronic mail: metzger@eng.utah.edu boundary layer thickness increases dramatically when mea-

1070-6631/2001/13(3)/692/10/$18.00 692 © 2001 American Institute of Physics


Phys. Fluids, Vol. 13, No. 3, March 2001 A comparative study of near-wall turbulence 693

sured in terms of the smallest scales of turbulent motion, TABLE I. Boundary layer integral parameters.
which are on the order of the viscous or inner unit of length ␪ 共m兲
R␪ U ⬁ 共m/s兲 u ␶ 共m/s兲 ␯ (m2 /s兲 ␦ 共m兲 H
defined as ␯ /u ␶ . For example, at R␪ ⫽5⫻103 , ␦ ⫹ ⬇2 000,
while at R␪ ⫽1⫻106 , ␦ ⫹ ⬎350 000, where the superscript 2⫻103 1.6 0.074 1.79⫻10⫺5 0.19 0.019 1.4

denotes normalization by inner units ␯ and u ␶ . Thus, even ⬃5⫻106 4.3 0.06–0.11 1.83⫻10⫺5 150 28 2.0
if ␦ remained constant over this R␪ range, the smallest scales
of turbulent motion still decrease by a factor greater than
175, posing severe challenges for sensor spatial resolution at
high Reynolds numbers. however, over the duration of any particular data acquisition
The present course of research attempts to overcome run, the mean wind direction remained essentially constant.
many of these difficulties by probing a high Reynolds num- Table I presents a summary of the boundary layer inte-
ber boundary layer (R␪ ⬇5⫻106 ) flow with extremely large gral parameters representative of the two experiments. In
length scales. A remote site on the salt flats of the Great Salt both cases, the friction velocity u ␶ was determined using the
Lake Desert was chosen because of the exceptionally smooth streamwise velocity gradient at the wall, as opposed to a
and barren terrain. An additional advantage of the site is the Clauser plot. The shape factor H (⬅ ␪ / ␦ * , with ␦ * denoting
highly predictable summer diurnal wind patterns, with ex- the displacement thickness兲 was determined by integrating
tended periods of constant wind speed and direction occur- the mean streamwise velocity profile. At the salt flats site,
ring in the late afternoon and continuing into the early mean horizontal wind speed profiles acquired from a tether-
evening. During these periods the atmospheric surface layer, sonde and miniSODAR were used to estimate ␦ , U ⬁ , ␪ , ␦ * ,
comprising roughly the lower tenth of the planetary bound- and R␪ .
ary layer, passes through neutral thermal stability, and, thus, Time-resolved streamwise velocity data were obtained
mimics a laboratory boundary layer. Comparisons are made using constant temperature hot-wire anemometry with the
between the flow over the salt flats and the low Reynolds five-wire rake depicted in Fig. 1. The rake remained at a
number (R␪ ⫽2⫻103 ) flow generated in a laboratory wind fixed location with the bottom of the probe resting on the
tunnel. The present results focus on the structure of the flow surface of the ground plane. In this position, the five hot-
in the viscous sublayer and buffer layer (2⬍y ⫹ ⬍30) where wires of the rake span a vertical distance of 0.5–5 mm above
turbulence production has been observed to reach a peak.4 the surface, and, thus, allow for detailed study of the viscous
The aim is to delineate Reynolds number effects on near- sublayer and buffer layer regions of the boundary layer. Ver-
wall turbulence statistics and turbulence production mecha- tical wire spacing was measured using a cathetometer
nisms. Single- and two-point streamwise velocity statistics, (⫾0.025 mm precision兲 with an overall estimated uncer-
single- and two-point ⳵ u/ ⳵ y statistics, spectra, spatiotempo- tainty of 4%. The hot-wires were comprised of 5 ␮ m diam-
ral correlations, and joint probability density functions were eter tungsten wire with a 1 mm active region. Regarding
analyzed. For purposes of comparison and consistency, the spatial resolution, it is important to note that for both the
same instrumentation and procedures were utilized in both field and laboratory measurements, l ⫹ ⬍6, where l ⫹
the R␪ ⬇5⫻106 and R␪ ⫽2⫻103 experiments; more impor- ⬅lu ␶ / ␯ with l denoting the sensor length. Data sets were
tantly, nearly identical inner normalized probe scales were sampled long enough temporally to capture over 10 000 in-
obtained in the two experiments which yielded high- tegral time scales in the wind tunnel and 1000–2000 integral
resolution streamwise velocity measurements at both R␪ . time scales, per run, for the field data 共ten independent runs
were acquired over a 3 h time frame兲. Note, in the former
II. EXPERIMENTAL METHODS
The low Reynolds number experiments were conducted
in the open circuit wind tunnel 共0.9⫻0.6⫻7.9 m3 ) located in
the Physical Fluid Dynamics 共PFD兲 laboratory at the Univer-
sity of Utah. Measurements were taken 5.3 m downstream
from the inlet. At this location, the axial free stream intensity
measured less than 0.3%. The high Reynolds number experi-
ments were conducted at a test site located on the southern
end of the salt flats of the Great Salt Lake Desert (113°26.5⬘
W, 40°8.1⬘ N兲. The terrain surrounding the test site is ex-
tremely smooth and flat with a roughness length estimated at
3⫻10⫺4 m. Surface elevation varies by less than 1 m over
the first 13 km north of the test site, and the upwind fetch
remains unobstructed for more than 100 km. The field ex-
periments were performed near sunset under conditions of
near neutral thermal stability, as verified from estimates of
the Monin–Obukov length. During the field trials, deviations
in the local mean wind direction, arising from low-frequency FIG. 1. Schematic of wall rake probe used in the present study. The lowest
motions of the order of 5–10 min, were not more than ⫾5°; wire lies 0.5 mm nominal above the surface. All dimensions are in mm.
694 Phys. Fluids, Vol. 13, No. 3, March 2001 M. M. Metzger and J. C. Klewicki

FIG. 2. Inner normalized plot of mean streamwise velocity vs position FIG. 3. Inner normalized plot of rms streamwise velocity vs position above
above the surface. The solid black line represents the log-law estimate of the surface.
Coles 共Ref. 7兲. The dotted lines mark the upper and lower bounds of the data
of Blackwelder and Haritonidis 共Ref. 8兲.

uncertainty in measuring ⌬U allowing for covariances of the


calibration fit parameters and the variance of the estimated
case, this is more than twice that needed to ensure conver-
mean value. The range of percent uncertainty in u ␶ for the
gence of the first four statistical moments of the streamwise
salt flats data reflects the decrease in u ␶ over the course of
velocity to within ⫾5% accuracy.3
the experiment, whereas the variance in u ␶ between indepen-
Calibrations were performed both prior to and following
dent runs remained fairly constant. Note, ten independent
each experiment. In the field trials, a 5.5 °C ambient tem-
runs were acquired at the salt flats site. The scatter between
perature drop occurred over the course of the experiments.
runs, obvious in the reported statistics, may be due to a num-
Since this constituted a non-negligible temperature drift, a
ber of factors including the inherent variability of the atmo-
temperature compensation technique, following Abdel-
sphere leading to unstationary statistics in the velocity when
Rahman et al.,5 was incorporated. Throughout the duration
averaged over any given time frame, slight misalignments of
of the field experiments, ambient temperature measurements
the probe with the mean wind direction, and a fairly rapid
were acquired concurrent with the hot-wire data. In the labo-
transition period between stably and unstably stratified con-
ratory, temperature compensation was not required and cali-
ditions at the test site.
bration data was fit to the model equation E 2 ⫽A⫹BV n . For
For comparison, the recent data of Folz,6 also acquired at
all experiments, the square wave frequency response of each
the salt flats site, are included in Fig. 2. Folz’s experiments
of the sensors on the probe was at least 15 kHz.
utilized an 8-wire probe, which was traversed vertically to
obtain profile data, and a 24-wire probe, which was held at a
III. RESULTS fixed vertical position. Only a single data set is reported for
A. Velocity and velocity gradient statistics each of Folz’s probes. Figure 2 indicates that mean data from
the field experiments tend to exceed Coles7 log-law line in
The mean, inner normalized, streamwise velocity profile the region y ⫹ ⬎20. As discussed by Blackwelder and
in the near-wall region is shown in Fig. 2. The notation Haritonidis,8 and manifest in their mean profile data, this
adopted throughout the remainder of the article is such that y observed trend exists when sublayer streamwise velocity
denotes the wall normal coordinate and ũ⫽U⫹u, where the data are used to determine u ␶ as opposed to a Clauser plot.
tilde denotes an instantaneous quantity, while the upper and The composite high and low Reynolds number data support
lower case letters symbolize the corresponding mean value inner scaling of the mean streamwise velocity profile.
and the zero-mean, fluctuating quantity, respectively. The su- In contrast, the inner normalized profile of the stream-
perscript ⫹ denotes normalization by the kinematic viscosity wise root mean square 共rms兲 velocity, presented in Fig. 3,
␯ and friction velocity u ␶ . For consistency, the value of u ␶ shows significant deviations between the low and high R␪
in both the wind tunnel and field studies was determined data, especially in the measured peak value near y ⫹ ⫽15. For
from the mean vertical velocity gradient at the wall, ⌬U/⌬y, the wind tunnel data, the uncertainty in u ⬘ ⫹ is estimated at
as opposed to a Clauser plot. Due to suspected heat conduc- less than 4.4% 共95% confidence interval兲 over the y ⫹ range
tion effects between the substrate and the wire closest to the measured. For the salt flats data, uncertainty estimates in u ⬘ ⫹
ground plane, u ␶ was determined using the second wire clos- range from 5.4% to 11% 共95% confidence interval兲 depend-
est to the ground 共see Fig. 1兲. Uncertainty in u ␶ , to within a ing on the particular run, with the highest uncertainties oc-
95% confidence interval, is estimated at 3.8% for the wind curring in the lowest two wires near the ground plane. To-
tunnel data and 4.0%–6.5% for the salt flats data. This ac- gether, the data in Fig. 3 depict a trend in the peak value of
counts for uncertainties in measuring ␯ and ⌬y as well as the u ⬘ ⫹ , increasing with increasing R␪ . Figure 4 displays the
Phys. Fluids, Vol. 13, No. 3, March 2001 A comparative study of near-wall turbulence 695

FIG. 4. Peak value of the inner normalized rms streamwise velocity vs FIG. 5. Skewness of the streamwise velocity vs inner normalized position
Reynolds number. The dotted line corresponds to the best fit line through the above the surface.
low R␪ data. The average value of the combined salt flats data is denoted
by 丢 .
2.73; and similarly,11 increasing l ⫹ from 14 to 32 produces a
peak u ⬘ ⫹ measurements from the present study as a function ⬘ ⫹ . The sensitivity of u ⬘ on probe
2.7 to 2.43 reduction in u max
of R␪ , along with data from Fernholz and Finley9 and sev- resolution could easily lead to erroneous conclusions regard-
eral other studies as referenced in Table II. The combined ing the R␪ dependence of u max ⬘ ⫹ , especially when based on
⬘ ⫹ value of 3.66. The hori-
salt flats data yield an average u max data that only cover a narrow R␪ range.1 This probably ac-
zontal error bars in Fig. 4 reflect the uncertainty in calculat- counts for a majority of previous work,2,9,15–17 contrary to
ing R␪ at the field site. As shown, the curve fit, derived only the present study, that supports inner scaling of u ⬘ in the
from the low R␪ data, extrapolates to well within the error near-wall region.
bounds and scatter of the field data. Overall, the data suggest Near-wall skewness and kurtosis profiles of the stream-
a logarithmic dependence of u max⬘ ⫹ on R␪ , indicating a failure wise velocity fluctuations are presented in Figs. 5 and 6. The
of inner scaling in the near-wall region. Interestingly, Maru- wind tunnel data from the present study generally show good
sic et al.10 have found that u ⬘ ⫹ measured at y ⫹ ⫽200 also agreement with the previous low R␪ data of Klewicki.18 The
exhibits a logarithmic dependence on R␪ with a slope of values of S(u) and K(u) near y ⫹ ⫽2 in the present wind
approximately 0.5. Their proposed similarity law for u ⬘ ⫹ , tunnel data are, however, slightly higher than the typical low
valid outside the buffer layer from y ⫹ ⭓100 to y/ ␦ ⭐1, pre- R␪ limiting values19,20 of 1 and 4, respectively. In contrast,
dicts an increase in u ⬘ ⫹ with increasing Reynolds number in both the skewness and kurtosis from the salt flats data
this region. achieve considerably higher values near the wall, reaching
Importantly, in all of the comparison cases listed in values of 1 and 4, respectively, at y ⫹ ⫽6 and continuing to
Table II as well as those of Fernholz and Finley,9 l ⫹ ⭐10. increase approaching the wall. Note, the low R␪ S(u) data
Previous studies8,11–14 have demonstrated the significant ef-
fect of reduced spatial resolution on the detection of small-
scale structures in the near-wall region and specifically on
the measurement of u ⬘ . For example,14 at a fixed R␪ increas-
ing the probe length 共or decreasing the spatial resolution兲
⬘ ⫹ from 2.81 to
from l ⫹ ⫽3.3 to 16 results in a decrease of u max

TABLE II. Low Reynolds number comparison data.

Paper Year R␪ ⬘⫹
u max

Ballint et al.24 1991 2685 2.68


Ching et al.37 1995 400, 1316 2.64, 2.71
Johansson et al.38 1987 4940 2.82
Johansson and Karlsson39 1988 2420 2.68
Klewicki and Falco3 1990 1010, 2870, 4850 2.65, 2.93, 3.05
Ligrani and Bradshaw12 1987 2620 2.81
Purtell et al.40 1981 465, 1840 2.41, 2.7
Spalart41 1988 300, 670, 1410 2.52, 2.64, 2.67
Ueda and Hinze42 1975 1244, 4248 2.73, 2.89
Wei and Willmarth13 1989 237, 1137, 3872 2.65, 2.68, 2.77 FIG. 6. Kurtosis of the streamwise velocity vs inner normalized position
above the surface.
696 Phys. Fluids, Vol. 13, No. 3, March 2001 M. M. Metzger and J. C. Klewicki

共95% confidence interval兲 over all runs. The dotted horizon-


tal lines represent the average u ⬘ ⫹ and S(u) values of the
unfiltered, wind tunnel data at y ⫹ ⬇20. The average, unfil-
tered values of the field data at this vertical location are
u ⬘ ⫹ ⫽3.6 and S(u)⫽0.39. As expected, increasing the cut-
off frequency results in a monotonic decrease in the filtered
u ⬘ ⫹ values from the collective field data. At a sufficiently
large cutoff frequency ( f ⫹ ⫹
c ⫽0.0018), the filtered u ⬘ values
attain the low R␪ value. A similar trend occurs in the S(u)
data, with the filtered salt flats data approaching that of the
low R␪ data near a cutoff frequency of f ⫹ c ⫽0.04. Above this
frequency, the filtered S(u) values steadily increase, appar-
ently asymptoting to zero within the scatter of the data. The
effect of the high-pass filter procedure on K(u) 共not shown兲
is negligible for f ⫹
c ⬍0.002, whereas, above this frequency,
the filtered K(u) values show dramatic amplification. Recall
FIG. 7. Effect of applying a high-pass filter, with cutoff frequency f c , to the from Fig. 6 that the field and wind tunnel K(u) values at
streamwise velocity statistics of the salt flats data at y ⫹ ⬇20. Dashed lines y ⫹ ⬇20 nearly overlie.
represent the unfiltered, wind tunnel results at the corresponding wall nor- The high-pass filtering procedure systematically elimi-
mal location. The average, unfiltered S(u) and u ⬘ ⫹ values of the cumulative nates any additive effect of large-scale, low-frequency mo-
field data are 0.39 and 0.34, respectively. Arrows denote the approximate
location at which the filtered salt flats data reach the respective low R␪ wind
tions on the measured statistics. Therefore, assuming that the
tunnel values. observed differences between the low and high R␪ flows re-
sult from the increased range of large scales inherent in the
high R␪ flow and assuming that the large and small scales of
exhibit a negative peak in the region 20⬍y ⫹ ⬍30, while the motion merely superpose, then one would expect the filtered
high R␪ data remain positive out to y ⫹ ⫽100. For y ⫹ ⬎20, u ⬘ ⫹ and S(u) values in Fig. 7 to approach their correspond-
the K(u) data from both the high and low R␪ flows appear to ing low R␪ values at identical cutoff frequencies. The fact
merge. that this does not exactly occur in the present data suggests
As discussed in Sec. I, a major distinction between high that there exists some direct coupling, albeit weak, between
and low Reynolds number flows lies in the range of intrinsic the large and small scales of motion. Note, the extent of the
turbulent length scales. Specific to the present study is the large-scale motions, as referred to in the present context, are
rather unusual condition that ␦ varies by a factor of about roughly taken as those motions with frequencies below ␻ ⫹
1000 between the high and low R␪ flows, whereas the sub- ⫽0.03, where ␻ ⬅2 ␲ f . Since each individual time series
layer thickness remains nearly identical for the two flows. from the salt flats exhibited the same trend, environmental
From this, one may hypothesize that the observed differ- factors, such as mean wind variations and/or surface layer
ences, as apparent from the statistics shown in Figs. 2–6, thermal stability conditions, are believed to have an inconse-
result primarily from the large scales inherent in the salt flats quential impact on the present observations.
boundary layer. Townsend21 originally posed the concept Owing to the relatively close spacing of the hot-wires on
that the motion at any point consists of two components the five element rake 共see Fig. 1兲, the simultaneously
共‘‘active’’ and ‘‘inactive’’ motions兲 which do not interact sampled data permitted measurement of ⳵ u/ ⳵ y time series.
directly. He further supposed22 that the large-scale, ‘‘inac- These were computed using a simple two-point finite differ-
tive’’ motions behave like a slowly varying mean flow to the ence, with the measurement point taken as halfway between
small-scale, stress-producing, ‘‘active’’ motions. Bradshaw23 adjacent wires. Figure 8 shows the present ( ⳵ u/ ⳵ y) ⬘ ⫹ data
showed, by considering short- and long-time averages using compared with the results from previous measurements3,20,24
a simple sinusoidal model of turbulence, that large-scale, of wall region spanwise vorticity ␻ z⬘ ⫹ . Note, the results of
‘‘inactive’’ motions can produce extensive streamwise veloc- Klewicki and Falco3 indicate that, independent of Reynolds
ity fluctuations in the inner layer without noticeable effect on number, ␻ z statistics are nearly identical to ⳵ u/ ⳵ y statistics
the mean flow. near the wall. The noticeable amplification in ( ⳵ u/ ⳵ y) ⬘ ⫹ at
This notion, that a range of large-scale, low-frequency high R␪ , in the region 2⬍y ⫹ ⬍10, emerges due to an in-
motions exist superposed on smaller-scale, high-frequency crease in spectral intensity from the entire range of scales of
motions, was explored in the context of the present data. A motion 共see Sec. III B兲. For y ⫹ ⬎10, no discernable Rey-
simple high-pass filtering procedure was applied to the field nolds number trend is apparent. Additionally, the present
data at y ⫹ ⬇20. This position corresponds closely to the data indicate, within tolerable scatter, that the wall value of
negative peak observed in S(u) at low R␪ and the peak u ⬘ ⫹ ␻ z⬘ ⫹ does not vary with R␪ . This result supports the recent
observed at high R␪ . Resultant u ⬘ ⫹ and S(u) values from high-resolution LDA pipe flow measurements by Durst et
the filtered salt flats data are plotted in Fig. 7 as a function of al.19 as well as the earlier wall shear stress measurements by
inner normalized cutoff frequency, f ⫹ c ⫽ f c ␯ /u ␶ . Note, the
2
Alfredsson et al.25 Both studies report a constant limiting

scatter in f c is due primarily to variations in u ␶ between value of ␻ z⬘ ⫹ ⬇0.4 at the wall.
runs, the uncertainty in f ⫹ c being estimated at less than 0.1% The sign of the near-wall instantaneous velocity gradient
Phys. Fluids, Vol. 13, No. 3, March 2001 A comparative study of near-wall turbulence 697

FIG. 8. Inner normalized plot of the rms of the vertical gradient of stream- FIG. 10. Power spectral functions of the streamwise velocity 共upper two
wise velocity vs position above the surface. plots兲 and streamwise velocity gradient 共lower two plots兲 near y ⫹ ⫽15 as a
function of inner normalized angular frequency. The exact wall normal po-
sitions corresponding to the above plots are as follows. For R␪ ⫽2⫻103 ,
⌿ u : y ⫹ ⫽20.6; ⌿ ⳵ u/ ⳵ y : y ⫹ ⫽16.7, and for R␪ ⬇5⫻106 , ⌿ u : y ⫹ ⫽16.9;
⌿ ⳵ u/ ⳵ y : y ⫹ ⫽14.1.
⳵ ũ/ ⳵ y 共or instantaneous spanwise vorticity ␻˜ z ) can provide
useful information regarding the nature of turbulent motions
near the wall. Figure 9 presents the probability of finding
positive ⳵ ũ/ ⳵ y as a function of inner normalized distance B. Spectra
above the wall. The low Reynolds number ␻ ˜ z results of Ra-
One-dimensional power spectra of the inner normalized
26
jagopalan and Antonia are also shown. The data presented streamwise velocity ⌽( ␻ ⫹ ,u ⫹ ) and velocity gradient
in Fig. 9 reveal Reynolds number similarity, at least within ⌽( ␻ ⫹ , ⳵ u ⫹ / ⳵ y ⫹ ) were calculated in order to investigate dif-
the region near the wall (y ⫹ ⬍25), in accord with the previ- ferences in the range of scales between the low and high R␪
ous ␻ ˜ z measurements of Klewicki et al.27 over a much nar- flows, here ␻ ⫹ (⬅ ␻␯ /u ␶ 2 ) denotes the inner normalized an-
rower Reynolds number range, 1010⭐R␪ ⭐4850. The com- gular frequency. Following the format of Perry and Abell,16
posite data indicate that, even out to R␪ ⬇5⫻106 , the data are presented in terms of spectral functions
instantaneous spanwise vorticity vector persistently remains
aligned with the direction of the mean shear very near the ⌿ u ⫽ ␻ ⫹ ⌽ 共 ␻ ⫹ ,u ⬘ ⫹ 兲
wall. This imposes constraints on the probable shape of the
and 共1兲
instantaneous streamwise velocity profile in the near-wall re-
gion, since ␻ ˜ z motions close to the wall correspond more to ⌿ ⳵ u/ ⳵ y ⫽ ␻ ⫹ ⌽ 共 ␻ ⫹ , ⳵ u ⫹ / ⳵ y ⫹ 兲 .
shear layers than solid body rotations.28
Power spectral functions acquired near y ⫹ ⫽15 are shown in
Fig. 10. The low Reynolds number wind tunnel data are
represented as black lines, while the high Reynolds number
salt flats data are drawn in light gray. The scales of motion
appear to be universal, in terms of inner normalization, only
in the low- to intermediate-frequency range, 0.03⭐ ␻ ⫹
⭐0.25. The present results taken together with the low R␪
channel data of Wei and Willmarth13 共their Fig. 21, also
acquired at y ⫹ ⫽15) indicate that this universal region ex-
tends further toward the high-frequency end of the spectrum
as the Reynolds number decreases. For example, their u
spectra, acquired over the Reynolds number range 200⬍R␪
⬍4000, begin to diverge at approximately ␻ ⫹ ⫽0.7, com-
pared to ␻ ⫹ ⫽0.25 in the present study. They did not observe
a noticeable deviation of the spectra at low frequencies; how-
ever, the lowest frequency reported in their paper was ␻ ⫹
⫽0.01. At this frequency, the present low and high R␪ val-
ues of ⌿ u differ by only ⬃0.07. This departure presumably
diminishes as the Reynolds number difference decreases, and
FIG. 9. Probability of encountering positive ⳵ ũ/ ⳵ y 共or negative, instanta- most likely, was not detectable over the Reynolds number
neous spanwise vorticity兲 in the near wall region. range investigated in the study of Wei and Willmarth.
698 Phys. Fluids, Vol. 13, No. 3, March 2001 M. M. Metzger and J. C. Klewicki

The present u spectra of Fig. 10 support the conclusion


that enhanced energy in the low-frequency motions is re-
sponsible for the observed Reynolds number discrepancy in
the u ⬘ ⫹ and S(u) profiles 共see Figs. 3 and 5兲. With high-pass
filtering, the low and high R␪ u ⬘ ⫹ and S(u) values at y ⫹
⫽15 coalesce at average cutoff frequencies of ␻ ⫹ c ⫽0.012
and ␻ ⫹ c ⫽0.11, respectively. From the spectra, it is apparent
that these cutoff frequencies lie well above the high-
frequency regime. On the other hand, the ⳵ u/ ⳵ y spectra re-
veal that enhanced contributions in the intermediate fre-
quency range 共roughly between ␻ ⫹ ⫽0.3 and 1.0兲 of the high
R␪ flow account for the slight Reynolds number dissimilarity
in the ( ⳵ u/ ⳵ y) ⬘ ⫹ profiles of Fig. 8 near y ⫹ ⫽15. For y ⫹
⬍10, the ( ⳵ u/ ⳵ y) ⬘ ⫹ profiles begin to diverge appreciably.
At y ⫹ ⬇10, the high R␪ spectrum of ⳵ u/ ⳵ y 共not shown兲 con-
tains more intensity, relative to the R␪ ⫽2⫻103 spectrum,
over the entire frequency range. Interestingly, in the R␪ ⬇5 FIG. 11. Spatial correlation of streamwise velocity vs inner normalized
⫻106 case, contributions associated with the lower frequen- distance from the reference position. The present study uses reference posi-
tions of y ⫹ ⫽1.6 (R␪ ⫽2⫻103 ) and 1.3⬍y ⫹ ⬍2.2 (R␪ ⬇5⫻106 ), while that
cies ( ␻ ⫹ ⬍0.03) have increased proportionately more than of Klewicki 共Ref. 18兲 uses a reference position of y ⫹ ⫽4.5.
that at all other frequencies in the spectrum.
Head and Bandyopadhyay29 observed in their flow visu-
alization studies that hairpin vortex-like motions in a turbu- present wind tunnel data coincide with the low R␪ results
lent boundary layer become more stretched at higher Rey- from the data base of Klewicki18 using a reference location
nolds numbers. The R␪ effects observed at high frequencies of y ⫹ ⫽4.5. The high and low R␪ results in Fig. 11 compare
in both the near-wall u and ⳵ u/ ⳵ y spectra are consistent with well for small ⌬y ⫹ , but begin to diverge significantly for
this notion. Owing to conservation of angular momentum, ⌬y ⫹ ⭓10. This behavior is consistent with the notion that a
the stretching of vortex loops into hairpin vortices necessar- considerable influence from large-scale motions at high R␪ ,
ily generates smaller-scale fluctuations with higher fre- which penetrate inside the near-wall region, causes both a
quency content, both of which should be manifest in the deviation from viscous scaling in the u statistics and a no-
velocity and vorticity spectra. Additionally, because of the ticeable increase in the correlated streamwise velocity fluc-
inherent random nature of turbulent boundary layer flow, tuations. Conversely, the R ⳵ u/ ⳵ y (⌬y ⫹ ) profiles of Fig. 12 dis-
stretching of one component of the vorticity vector will con- play relatively close agreement. The present results, using
ceivably lead to stretching of the other two components ow- fixed reference locations at y ⫹ ⫽2.8 (R␪ ⫽2⫻103 ) and 2.3
ing to the effect of the reorientation terms in the vorticity ⬍y ⫹ ⬍4 (R␪ ⬇5⫻106 ), are compared with the low R␪ ,
transport equations. The question remains whether the spec- R ⳵ u/ ⳵ y (⌬y ⫹ ) data of Klewicki18 as well as correlation coef-
tral amplification in the high-frequency ⳵ u/ ⳵ y motions at ficient profiles of spanwise vorticity from Klewicki and
high R␪ is a direct consequence of the stretching of spanwise
vorticity or, for example, streamwise vorticity 共characteriz-
ing hairpin vortices兲 that has become stretched and reori-
ented into the spanwise direction.

C. Correlations
Simultaneously sampled signals from the hot-wire rake
allowed the spatiotemporal structure of the near-wall flow to
be explored. In particular, the autocorrelation coefficient of
the streamwise velocity

具 u 共 y,t 兲 u 共 y⫹⌬y,t⫹ ␶ 兲 典
R u 共 y,⌬y, ␶ 兲 ⫽ 共2兲
u ⬘ 共 y 兲 u ⬘ 共 y⫹⌬y 兲
was determined for ⌬y spatial separation and ␶ time delay,
where 具 • 典 denotes a long-time average. The autocorrelation
coefficient for the streamwise velocity gradient (R ⳵ u/ ⳵ y ) was
also determined in a similar manner. In all calculations using
Eq. 共2兲, statistical stationarity was assumed so that t⫽0 and FIG. 12. Spatial correlation of the vertical gradient of streamwise velocity
time averages were performed over the entire record dura- vs inner normalized distance from the reference position. The present study
uses reference positions of y ⫹ ⫽2.8 (R␪ ⫽2⫻103 ) and 2.3⬍y ⫹ ⬍4.0 (R␪
tion. Figure 11 displays the present R u (⌬y ⫹ ) results using ⬇5⫻106 ). Comparison data utilize the following reference positions:
fixed reference locations of y ⫹ ⫽1.6 and 1.3⬍y ⫹ ⬍2.2 for Klewicki 共Ref. 18兲, y ⫹ ⫽7.5; Klewicki and Falco 共Ref. 3兲, y ⫹ ⫽7.5; Raja-
the R␪ ⫽2⫻103 and R␪ ⬇5⫻106 flows, respectively. The gopalan and Antonia 共Ref. 26兲, y ⫹ ⫽4.75; and Kim 共Ref. 26兲, y ⫹ ⫽4.75.
Phys. Fluids, Vol. 13, No. 3, March 2001 A comparative study of near-wall turbulence 699

FIG. 14. Probable instantaneous streamwise velocity profiles corresponding


to the quadrant contributions shown in Fig. 13. Tick marks on the vertical
axes are drawn in increments of y ⫹ ⫽20. Dotted lines represent the mean
boundary layer profile.
FIG. 13. Weighted joint probability density functions of inner normalized
⳵ u/ ⳵ y. Dotted lines represent the origin of ⳵ ũ/ ⳵ y. 共a兲 R␪ ⫽2⫻103 ,
y⫹ ⫹ 6 ⫹ ⫹
1 ⫽2.8, y 2 ⫽16.6. 共b兲 R␪ ⬇5⫻10 , y 1 ⫽4.0, y 2 ⫽23.5. in the viscous sublayer have ⫹⳵ ũ/ ⳵ y. In the buffer layer,
about one-third of Q4 contributions actually exhibit
⫺ ⳵ ũ/ ⳵ y, for the case of the low R␪ flow, while in the high
Falco,30 Rajagopalan and Antonia,26 and the DNS data of R␪ flow, about half of the Q4 contributions possess ⫺⳵ ũ/ ⳵ y.
Kim.26 In particular, all of the data indicate a negative peak This type of information provides a useful foundation to
between 12⬍⌬y ⫹ ⬍18. Although small, one apparent differ- construct likely streamwise velocity profiles that characterize
ence between the high and low R␪ data is a reduction in predominant near-wall turbulent events or coherent motions.
magnitude of the negative peak from ⫺0.42, at low R␪ , to Figure 14 presents a set of probable instantaneous boundary
approximately ⫺0.33, at high R␪ . Reasons for this disparity layer profiles consistent with the weighted joint pdf results of
become evident in contour plots of the weighted joint prob- Fig. 13. Mean profiles are denoted by a dotted line. Figures
ability density function 共pdf兲 of ⳵ u/ ⳵ y. 14共a兲, 14共b兲, and 14共d兲 largely resemble the conditionally
In order to examine the different quadrant contributions averaged velocity profiles of Blackwelder and Kaplan32
underlying the correlations in Fig. 12, weighted joint pdf’s of 共their Fig. 9兲, acquired at R␪ ⫽2550 and based on bursting
⳵ u/ ⳵ y were computed. For two random variables ␥ 1 and ␥ 2 , events detected at y ⫹ ⫽15. In particular, the profile in Fig.
the weighted joint pdf is expressed as ␥ 1 ␥ 2 p( ␥ 1 , ␥ 2 ). In the 14共a兲, associated with Q2 contributions, looks strikingly
present case, ␥ 1 ⫽ ⳵ u(y ⫹ ⫹
1 )/ ⳵ y ( ␯ /u ␶ ) and ␥ 2 ⫽ ⳵ u(y 2 )/
2
similar to their conditionally averaged profile at the time of
⫹ ⫹ ⫹
⳵ y ( ␯ /u ␶ ), where y 2 ⫽y 1 ⫹⌬y . The weighted joint pdf’s,
2
the detected burst event, i.e., at zero time lag ␶ ⫽0. The pro-
derived from the ⳵ u/ ⳵ y signals forming the basis for the files drawn in Figs. 14共b兲 and 14共d兲 correspond to their con-
observed peak negative correlation in R ⳵ u/ ⳵ y , are shown in ditionally averaged profiles occurring at ␶ ⬍0 and ␶ ⬎0, re-
Fig. 13. The corresponding reference positions are y ⫹ 1 spectively, characteristic of near-wall coherent structures.
⫽2.8, y ⫹ ⫹ ⫹
2 ⫽16.6 (R␪ ⫽2⫻10 ) and y 1 ⫽4.0, y 2 ⫽23.5 (R␪
3
Specifically, the decelerated flow of Q3 represents an ejec-
⬇5⫻106 ). The coordinate axes defining the origin of the tion, while the accelerated flow of Q4 signifies a sweep.33
instantaneous ⳵ ũ/ ⳵ y (⬅ ⳵ U/ ⳵ y⫹ ⳵ u/ ⳵ y), denoted by the The probability of ⫺ ⳵ ũ/ ⳵ y occurring in Q3 and Q4 implies
dotted lines in Fig. 13, were calculated using Van Driest’s31 that inflectional profiles, shown in Figs. 14共c兲 and 14共e兲, may
formulation. The present wind tunnel results agree remark- coexist in the buffer layer as well. Note, plausible shapes for
ably well with the near-wall, spanwise vorticity results of the inflectional profiles are portrayed in accordance with the
Rajagopalan and Antonia,26 R␪ ⫽1450 共their Fig. 10兲, and hydrogen bubble, side-view flow visualization results of Kim
those of Klewicki and Falco,30 R␪ ⫽1010 关their Fig. 17共b兲兴. et al.34 The fact that the probability of finding an instanta-
In particular, the negative peak correlation in R ⳵ u/ ⳵ y largely neous inflectional profile increases with R␪ may stem from
results from fourth quadrant 共Q4兲 interactions between an inherent amplification in the spanwise vorticity fluctua-
⫺⳵ u/ ⳵ y in the buffer region and ⫹⳵ u/ ⳵ y in the viscous tions or ( ⳵ u/ ⳵ y) ⬘ ⫹ at high R␪ in the region y ⫹ ⭐10 共see
sublayer. Substantial contributions also arise from second Fig. 8兲.
quadrant 共Q2兲 interactions between ⫹⳵ u/ ⳵ y in the buffer Space–time autocorrelations associated with
region and ⫺ ⳵ u/ ⳵ y in the viscous sublayer. In the high R␪ R u (⌬y ⫹ , ␶ ⫹ ) are shown in Fig. 15 with parts 共a兲 and 共b兲
flow, there is a larger positive contribution from Q3 which representing the low and high R␪ experimental results, re-
accounts for the fact that the negative peak correlation has a spectively. The y ⫹ reference locations are identical to the
slightly lower magnitude than that observed for the lower R␪ ones used to compute the corresponding autocorrelation pro-
flows. Relative to the instantaneous axes, all of the motions files in Fig. 11 for the case of ␶ ⫽0. Since upstream events
700 Phys. Fluids, Vol. 13, No. 3, March 2001 M. M. Metzger and J. C. Klewicki

a logarithmic increase in the peak value of u ⬘ ⫹ with increas-


ing R␪ . In the skewness profile, the high R␪ data remain
positive at least out to y ⫹ ⫽100; while at low R␪ , the skew-
ness becomes negative at y ⫹ ⬇10.
Certain aspects of the flow physics at high Reynolds
number can be inferred from these experimental observa-
tions. In particular, the attached eddy hypothesis of
Townsend,22 see also Marusic et al.,10 predicts that in a re-
gion outside of the buffer layer u ⬘ ⫹ depends logarithmically
on y/L o , where L o represents a characteristic length of the
largest attached eddy in the flow, the flow postulated as be-
ing comprised of a distribution of attached eddies of varying
scales. Since the range of eddy length scales increases with
FIG. 15. Space–time autocorrelations of u. Dotted lines denote the locus of
Reynolds number, so must u ⬘ ⫹ . If the peak value of u ⬘ ⫹
maximum correlation. Reference locations are identical to those used in Fig. inside the buffer layer does not grow accordingly with u ⬘ ⫹
11. 共a兲 R␪ ⫽2⫻103 . 共b兲 R␪ ⬇5⫻106 . outside of the buffer layer, then the mean gradient ⳵ 具 u 2 典 / ⳵ y
in this region could become zero or change sign from nega-
tive to positive. This would represent a dramatic shift in the
occur at a positive time delay, the flow in Fig. 15 can be location of the peak production of turbulent kinetic energy
thought of as moving from right to left. The angle of maxi- away from the wall and result in a gradient flux of turbulent
mum correlation, denoted by the dotted lines, in the region kinetic energy from the log layer toward the wall. The
0⭐y ⫹ ⭐10 is nearly identical in the two flows, measuring present data indicate that such a dramatic change in flow
60° (R␪ ⫽2⫻103 ) and 63° (R␪ ⬇5⫻106 ). For y ⫹ ⬎10, the physics does not occur with increasing R␪ . Furthermore, be-
angle of maximum correlation in the high R␪ case increases cause the skewness of u remains positive as the rms in-
to 76°. The tilt of the isocontours suggests that, on average, creases, the probability of local flow reversal, in the absolute
the predominant motion is characterized by the ejection of sense, does not necessarily grow. The present results regard-
low-momentum fluid away from the wall into the buffer
ing the unlikelihood of finding ⫺⳵ ũ/ ⳵ y near the wall also
layer. The weighted joint pdf’s between viscous sublayer and
supports this statement. Note, the possibility of local flow
buffer layer u motions 共not shown兲 illustrate that, indeed, the
reversal occurring in the sublayer could have appreciable im-
major contribution to R u arises from Q3, i.e., the most prob-
plications on turbulent production mechanisms at high R␪ .
able configuration stems from low-speed buffer layer fluid
Through high-pass filtering, the disparities between the
coupled with low-speed fluid in the viscous sublayer. This
low and high R␪ profiles of u ⬘ ⫹ and S(u) have been shown
observation agrees well with the dye injection flow visual-
to originate mostly from the large-scale, low-frequency mo-
ization results of Kline et al.4 obtained at R␪ ⫽1085, in
tions. The filtering procedure also demonstrated the potential
which the trajectory of low-speed streaks/‘‘eddies’’ away
from the wall was tracked using high-speed cinematography. that direct interactions between large and small scales of mo-
From their Fig. 17, and using inner normalization of space tion exist, which cannot be entirely removed by assuming a
and time scales, one finds that low-speed ‘‘eddies’’ are simple superposition of scales. Both the low- and high-
ejected from the wall at an angle of 58°, in remarkable ac- frequency regions associated with the spectra of u and ⳵ u/ ⳵ y
cord with the present results. Correlations associated with the displayed an increase in spectral intensity at high R␪ . A
high R␪ data generally exhibit larger magnitudes than the narrow region in the intermediate frequency range of the u
corresponding low R␪ values at any given (⌬y ⫹ , ␶ ⫹ ). This spectra scaled with inner variables. This universal region
deviation tends to increase with increasing spatial separation widens, extending further toward the high-frequency end of
⌬y ⫹ . the spectrum, as the Reynolds number decreases.13
Near-wall features evident in the space–time autocorre-
lations of u and, particularly, in the weighted joint pdf’s of
IV. SUMMARY AND CONCLUSIONS ⳵ u/ ⳵ y, were argued to be characteristic of turbulent bursts,
Streamwise velocity and velocity gradient data from sweeps, and ejections. More importantly, the present data
high-resolution hot-wire anemometry were used to explore revealed that, under inner normalization, many near-wall fea-
the influence of Reynolds number on viscous wall region tures are highly similar over a substantial Reynolds number
structure in a turbulent boundary layer. Low and high Rey- range, 2⫻103 ⭐R␪ ⭐5⫻106 . In addition, plan view flow
nolds number experiments were performed spanning an R␪ visualizations35,36 have demonstrated that inner normalized
range over three orders of magnitude. Importantly, in both spacing of low-speed streaks in the sublayer remains con-
cases, the same sensors and instrumentation were employed stant at ⬃100␯ /u ␶ , over a Reynolds number range similar to
and the hot-wire lengths measured less than 6 ␯ /u ␶ . Com- that of the present study. It is well known that the production
parison of the present low and high R␪ results reveal that, to of turbulent kinetic energy in a low Reynolds number bound-
within the scatter of the data, the mean streamwise velocity ary layer peaks in the buffer layer,4 and that, in the near-wall
profile, normalized by inner variables, exhibits Reynolds region, most of the net production occurs during turbulent
number similarity. On the other hand, the same data suggest burst events.4,34 Therefore, based on the cumulative data,
Phys. Fluids, Vol. 13, No. 3, March 2001 A comparative study of near-wall turbulence 701

16
many of the near-wall mechanisms involved in turbulent pro- A. Perry and C. Abell, ‘‘Scaling laws for pipe-flow turbulence,’’ J. Fluid
duction processes at low R␪ 共e.g., turbulent bursts, sweeps, Mech. 67, 257 共1975兲.
17
K. Harder and W. Tiederman, ‘‘Drag reduction and turbulent structure in
ejections, and low speed streaks兲 appear to remain largely
two-dimensional channel flows,’’ Philos. Trans. R. Soc. London, Ser. A
intact at R␪ ⬇5⫻106 . Time and length scales associated with 336, 19 共1991兲.
some of these coherent motions, however, are posited to vary 18
J. Klewicki, ‘‘On the interactions between the inner and outer region mo-
with Reynolds number based on an increased contribution tions in turbulent boundary layers,’’ Ph.D. thesis, Michigan State Univer-
sity, East Lansing, MI, 1989.
from ‘‘inactive’’ motions at high R␪ . The net effect of these 19
F. Durst, J. Jovanovic, and J. Sender, ‘‘LDA measurements in the near-
changes in scale on turbulent momentum transport across the wall region of a turbulent pipe flow,’’ J. Fluid Mech. 295, 305 共1995兲.
boundary layer remains to be determined. 20
J. Kim, P. Moin, and R. Moser, ‘‘Turbulence statistics in fully developed
channel flow at low Reynolds number,’’ J. Fluid Mech. 177, 133 共1987兲.
21
ACKNOWLEDGMENTS A. Townsend, ‘‘Equilibrium layers and wall turbulence,’’ J. Fluid Mech.
11, 97 共1961兲.
The authors would like to acknowledge the help of Jeff 22
A. Townsend, The Structure of Turbulent Shear Flow, 2nd ed. 共Cambridge
Adams in the construction of the five-element hot-wire rake. University Press, Cambridge, 1976兲.
23
P. Bradshaw, ‘‘‘Inactive’ motion and pressure fluctuations in turbulent
Also, the assistance provided by the personnel of the U.S. boundary layers,’’ J. Fluid Mech. 30, 241 共1967兲.
Army Dugway Proving Ground, West Desert Test Center, 24
J.-L. Balint, J. Wallace, and P. Vukoslavcevic, ‘‘The velocity and vortic-
Meteorology Division, and in particular the efforts of Chris- ity vector fields of a turbulent boundary layer. Part 2. Statistical proper-
topher Biltoft, is greatly appreciated. ties,’’ J. Fluid Mech. 228, 53 共1991兲.
25
P. Alfredsson, A. Johansson, J. Haritonidis, and H. Eckelmann, ‘‘The
This work was partially supported by the U.S. Army. fluctuating wall-shear stress and the velocity field in the viscous sub-
M.M.M. was supported by graduate fellowships from the layer,’’ Phys. Fluids 31, 1026 共1988兲.
26
University of Utah and the National Science Foundation dur- S. Rajagopalan and R. Antonia, ‘‘Structure of the velocity field associated
ing the course of this work. with the spanwise vorticity in the wall region of a turbulent boundary
layer,’’ Phys. Fluids A 5, 2502 共1993兲.
27
1
J. Klewicki, C. Gendrich, J. Foss, and R. Falco, ‘‘On the sign of the
M. G. el Hak and P. Bandyopadhyay, ‘‘Reynolds number effects in wall- instantaneous spanwise vorticity component in the near-wall region of
bounded turbulent flows,’’ Appl. Mech. Rev. 47, 307 共1994兲. turbulent boundary layers,’’ Phys. Fluids A 2, 1497 共1990兲.
2
S. Mochizuki and F. T. M. Nieuwstadt, ‘‘Reynolds-number-dependence of 28
J. Klewicki, J. Murray, and R. Falco, ‘‘Vortical motion contributions to
the maximum in the streamwise velocity fluctuations in wall turbulence,’’ stress transport in turbulent boundary layers,’’ Phys. Fluids 6, 277 共1994兲.
Exp. Fluids 21, 218 共1996兲. 29
M. Head and P. Bandyopadhyay, ‘‘New aspects of turbulent boundary-
3
J. C. Klewicki and R. E. Falco, ‘‘On accurately measuring statistics asso- layer structure,’’ J. Fluid Mech. 107, 297 共1981兲.
ciated with small-scales in turbulent boundary layers using hot-wire 30
J. Klewicki and R. Falco, ‘‘Spanwise vorticity structure in turbulent
probes,’’ J. Fluid Mech. 219, 119 共1990兲. boundary layers,’’ Int. J. Heat Fluid Flow 17, 363 共1996兲.
4
S. J. Kline, W. C. Reynolds, F. A. Schraub, and P. W. Runstadler, ‘‘The 31
E. Van Driest, ‘‘On turbulent flow near a wall,’’ J. Aeronaut. Sci. 23, 1007
structure of turbulent boundary layers,’’ J. Fluid Mech. 30, 741 共1967兲.
5 共1956兲.
A. Abdel-Rahman, C. Tropea, P. Slawson, and A. Strong, ‘‘On tempera- 32
R. Blackwelder and R. Kaplan, ‘‘On the wall structure of the turbulent
ture compensation in hot-wire anemometry,’’ J. Phys. E 20, 315 共1987兲.
6 boundary layer,’’ J. Fluid Mech. 76, 89 共1976兲.
A. B. Folz, ‘‘An experimental study of the near-surface turbulence in the 33
S. K. Robinson, ‘‘Coherent motions in the turbulent boundary layer,’’
atmospheric boundary layer,’’ Ph.D. thesis, University of Maryland, Col-
Annu. Rev. Fluid Mech. 23, 601 共1991兲.
lege Park, MD, 1997. 34
7 H. T. Kim, S. J. Kline, and W. C. Reynolds, ‘‘The production of turbu-
D. Coles, ‘‘Turbulent boundary layers in pressure gradients: A survey
lence near a smooth wall in a turbulent boundary layer,’’ J. Fluid Mech.
lecture,’’ in Proceedings of the 1968 AFOSR-IFP–Stanford Conference on
50, 133 共1971兲.
Computation of Turbulent Boundary Layers, edited by D. Coles and H. 35
J. C. Klewicki, M. M. Metzger, E. Kelner, and E. M. Thurlow, ‘‘Viscous
Hirst 共Stanford University, Stanford, CA, 1969兲.
8
R. Blackwelder and J. Haritonidis, ‘‘Scaling of the bursting frequency in sublayer flow visualizations at r ␪ ⬇1 500 000,’’ Phys. Fluids 7, 857
turbulent boundary layers,’’ J. Fluid Mech. 132, 87 共1983兲. 共1995兲.
36
9
H. Fernholz and P. Finley, ‘‘The incompressible zero-pressure-gradient C. R. Smith and S. P. Metzler, ‘‘The characteristics of low-speed streaks
turbulent boundary layer: An assessment of the data,’’ Prog. Aerosp. Sci. in the near-wall region of a turbulent boundary layer,’’ J. Fluid Mech. 129,
32, 245 共1996兲. 27 共1983兲.
37
10
I. Marusic, A. Uddin, and A. Perry, ‘‘Similarity law for the streamwise C. Ching, L. Djenidi, and R. Antonia, ‘‘Low-Reynolds-number effects in a
turbulence intensity in zero-pressure-gradient turbulent boundary layers,’’ turbulent boundary layer,’’ Exp. Fluids 19, 61 共1995兲.
38
Phys. Fluids 9, 3718 共1997兲. A. Johansson, J.-Y. Her, and J. Haritonidis, ‘‘On the generation of high-
11
A. V. Johansson and P. H. Alfredsson, ‘‘Effects of imperfect spatial reso- amplitude wall-pressure peaks in turbulent boundary layers and spots,’’ J.
lution on measurements of wall-bounded turbulent shear flows,’’ J. Fluid Fluid Mech. 175, 119 共1987兲.
39
Mech. 137, 409 共1983兲. T. Johansson and R. Karlsson, ‘‘The energy budget in the near-wall region
12
P. Ligrani and P. Bradshaw, ‘‘Spatial resolution and measurement of tur- of a turbulent boundary layer,’’ in Applications of Laser Anemometry to
bulence in the viscous sublayer using subminiature hot-wire probes,’’ Exp. Fluid Mechanics, edited by R. Adrian, T. Asanuma, D. Durao, F. Durst,
Fluids 5, 407 共1987兲. and J. Whitelaw 共Springer-Verlag, Berlin, 1989兲, pp. 3–22.
13 40
T. Wei and W. Willmarth, ‘‘Reynolds-number effects on the structure of a L. Purtell, P. Klebanoff, and F. Buckley, ‘‘Turbulent boundary layer at
turbulent channel flow,’’ J. Fluid Mech. 204, 57 共1989兲. low Reynolds number,’’ Phys. Fluids 24, 802 共1981兲.
14 41
P. Ligrani and P. Bradshaw, ‘‘Subminiature hot-wire sensors: Develop- P. Spalart, ‘‘Direct simulation of a turbulent boundary layer up to
ment and use,’’ J. Phys. E 20, 323 共1987兲. r ␪ ⫽1410,’’ J. Fluid Mech. 187, 61 共1988兲.
15 42
J. Laufer, ‘‘Investigation of turbulent flow in a two-dimensional channel,’’ H. Ueda and J. Hinze, ‘‘Fine-structure turbulence in the wall region of a
Tech Note 2123, NACA 共July 1950兲. turbulent boundary layer,’’ J. Fluid Mech. 67, 125 共1975兲.

You might also like