J. Fluid Mech. (2004), vol. 508, pp. 99–131.

c 2004 Cambridge University Press 99
DOI: 10.1017/S0022112004008985 Printed in the United Kingdom

Scaling of the streamwise velocity component

in turbulent pipe flow
By J. F. M O R R I S O N1 , B. J. M c K E O N2 , W. J I A N G3
A N D A. J. S M I T S2
1
Department of Aeronautics, Imperial College, London SW7 2AZ, UK
2
Department of Mechanical & Aerospace Engineering, Princeton University,
Princeton, NJ 08544-0710, USA
3
CARDC, PO Box 211 Mianyang, Sichuan 621 000, People’s Republic of China

(Received 7 March 2003 and in revised form 30 January 2004)

Statistics of the streamwise velocity component in fully developed pipe flow are
examined for Reynolds numbers in the range 5.5 × 104 6 ReD 6 5.7 × 106 . Probability
density functions and their moments (up to sixth order) are presented and their
scaling with Reynolds number is assessed. The second moment exhibits two maxima:
the one in the viscous sublayer is Reynolds-number dependent while the other, near
the lower edge of the log region, follows approximately the peak in Reynolds shear
stress. Its locus has an approximate (R + )0.5 dependence. This peak shows no sign of
‘saturation’, increasing indefinitely with Reynolds number. Scalings of the moments
with wall friction velocity and (Ucl − U ) are examined and the latter is shown to be
a better velocity scale for the outer region, y/R > 0.35, but in two distinct Reynolds-
number ranges, one when ReD < 6 × 104 , the other when ReD > 7 × 104 . Probability
density functions do not show any universal behaviour, their higher moments showing
small variations with distance from the wall outside the viscous sublayer. They are
most nearly Gaussian in the overlap region. Their departures from Gaussian are
assessed by examining the behaviour of the higher moments as functions of the lower
ones. Spectra and the second moment are compared with empirical and theoretical
scaling laws and some anomalies are apparent. In particular, even at the highest
Reynolds number, the spectrum does not show a self-similar range of wavenumbers
in which the spectral density is proportional to the inverse streamwise wavenumber.
Thus such a range does not attract any special significance and does not involve a
universal constant.

1. Introduction
It has long√been accepted that the motion in the viscous sublayer (y + = yuτ /ν < 30,
where uτ = τw /ρ, y is the wall-normal distance and τw is the wall shear stress)
is directly affected by viscosity, ν. Either a conventional overlap analysis that
uses asymptotic matching (Millikan 1938; Wosnik, Castillo & George 2000) or a
dimensional analysis that yields the logarithmic law for the mean velocity suggests
that the motion of the log region is independent of viscosity. However, such analyses
do not specify directly the range in y over which the log law applies, this being the
subject of much experimental investigation. Zagarola & Smits (1998) have shown
that the mean velocity exhibits a log-law dependence for 600 < y + < 0.07R + only
when R + > 9 × 103 (R + is the Kármán number based on pipe radius) and that
100 J. F. Morrison, B. J. McKeon, W. Jiang and A. J. Smits
for 60 < y + < 500, there is a power law when R + > 9000 which expresses the direct
influence of viscosity out to y + ≈ 0.15R + when R + < 9000 (see also Zagarola, Perry
& Smits 1997). Using an improved data analysis, McKeon et al. (2004) have recently
confirmed that complete similarity in the form of a log law (with slightly modified
constants) occurs for R + > 5 × 103 only, in the range 600 < y + < 0.12R + . These scalings
for the first moment therefore raise questions concerning the most appropriate choice
of velocity scale for the higher moments, particularly because the fluctuating motion
at a point comprises a range of scales, or equivalently, motion at a given wavenumber
receives contributions from the entire physical domain. Moreover, near walls much
of the turbulence information resides in the smaller scales which, in addition to
inhomogeneity and anisotropy, show significant departures from a Gaussian velocity
distribution. For example, close to the wall, the flatness of the wall-normal velocity
component becomes very large (Eggels et al. 1994) and can be attributed to the
spatially alternating behaviour of ejections and sweeps. Statistically, this suggests
inter-connection between many possible degrees of freedom, so contravening one of
the requirements of the central limit theorem. The behaviour of the higher moments
at very high Reynolds numbers is therefore very interesting.
In contrast to the mean velocity, it has been established for some time that the
Reynolds stresses near the wall scale neither with ‘wall’ (or ‘inner’) variables (ν/uτ , uτ ),
nor with ‘outer’ variables (R, uo , where uo is an outer velocity scale left undefined
at present). Of particular importance is the difference between the behaviour of
the wall-parallel (u, w) and wall-normal (v) components where the impermeability
constraint, which affects eddies out to a distance from the wall that is of the order
of the eddy size (‘blocking’ or ‘splatting’), is responsible for an increase in the wall-
parallel components at the expense of the wall-normal one. Thus the behaviour of the
statistics for the u- and w-components is different from those for the v-component,
and in the case of pipe flow, the streamwise and azimuthal components are subject to
specialized homogeneous boundary conditions. Failure of the u- and w-components
to scale on wall variables was explained by Townsend (1961, 1976) as the influence of
‘inactive’ motion (Bradshaw 1967; Morrison, Subramanian & Bradshaw 1992), that
of the large eddies inducing a ‘meandering or swirling’ on the near-wall motion. Being
largely confined to the (x, z)-plane and scaling on outer variables, the inactive motion
does not, to a first order, contribute to either the ρv 2 normal stress, or to the shear
stress, −ρuv, the ‘active’ component of near-wall motion. It is therefore supposed that
the two modes do not interact, the active one being modulated by an ‘irrotational
free stream’ (Bradshaw 1967), the result of both large-scale vorticity and irrotational
pressure fluctuations.
For ‘high’ Reynolds numbers, when y +  1 and y/R  1, an overlap region for
the first moment only of the streamwise component becomes apparent. Even then,
there are many measurements that show that the active component of the Reynolds
stresses does not scale on inner variables either. In particular, the ‘constant-stress’
region (−uv ≈ u2τ ) does not hold, except in the limit of very high Reynolds number.
But it is by no means clear just how high a Reynolds number is required: this
is particularly noteworthy since the constant-stress region can be deduced by the
same dimensional arguments that lead to the log law, but it has yet to be determined
whether or not the constant-stress region emerges at about the same Reynolds number
as that at which the log law does (Zagarola & Smits 1998; McKeon et al. 2004).
Morrison et al. (1992) show that the Reynolds-stress-bearing motion in a boundary
layer at Reθ ≈ 1.5 × 104 (equivalent to R + ≈ 5000) is associated with a Kolmogorov
scaling and this should be recognized as evidence of the failure of universal inner
 Scaling of the streamwise velocity component in turbulent pipe flow 101
scaling, that is, uτ = − (uv)1/2 with lengthscale y, y +  1. See also Antonia & Kim
(1994) and Wei & Willmarth (1989). These studies were performed at much lower
Reynolds numbers than those studied in the present investigation.
Nevertheless, in considering the Reynolds-number dependence of near-wall
turbulence, it is important to distinguish direct viscous influence (physically, the
eruptions of low-momentum fluid from the sublayer) from that of the outer-region
motion (large-scale inrushes that produce splats near the wall). The ratio of the two
relevant lengthscales is, of course, R + . Not only is the outer-scaling influence of
inactive motion more apparent at high Reynolds numbers, it is also more prevalent
in boundary layers in which the influence of inactive motion is larger than in internal
flows. This raises the question of the extent to which the active and inactive modes
interact, and indeed whether such a delineation is meaningful. In this context, it is
important to remember that Townsend’s original distinction was based on the basic,
but conceptual element of near-wall structure, the ‘attached wall eddy’. Recently, the
precise nature of so-called inactive motion in high-Reynolds-number boundary layers
has been examined by Hunt & Morrison (2000) who suggest that ‘top-down’, outer-
layer interactions are important to the dynamics in the near-wall region. Note that
any interaction between these two modes implies that, even at very high Reynolds
numbers when direct viscous effects are small, wall motion cannot be universal in
any meaningful sense. Outer-layer influences may also have ramifications for the
self-similarity of the mean velocity: both Bradshaw (1967) and Townsend (1976) use
a simple linear analysis to show that the effect of inactive motion is to make the von
Kármán constant, κ, Reynolds-number dependent. These considerations lead also to
the conclusion that simple arguments concerning the overlap of scales for the higher
moments in a particular region of wall turbulence are inappropriate. Crucial to the
understanding of these issues is the realization that the wall affects wall-normal and
wall-parallel components differently.
In this paper, we report hot-wire measurements of the streamwise velocity
component in the Reynolds-number range (based on pipe diameter, D and mean
velocity, U ) 5.5 × 104 6 ReD 6 5.7 × 106 . Statistics up to the sixth moment are
calculated as the moments of a probability density function (p.d.f.). Equivalent spectra
as a function of streamwise wavenumber, φ(k1 ), are also presented. The scaling of
both is investigated. It is becoming increasingly apparent that there are significant
differences between different flows of the same species: thus in the present context, we
distinguish between not only external and internal flows, but also between pipe and
channel flows (Nieuwstadt & Bradshaw 1997). Therefore reference to channel flows is
not made unless the results are specifically relevant to the present work. Early work
on pipe flow includes that of Laufer (1954) and Sandborn (1955) and more recently,
Durst, Jovanović & Sender (1995), Eggels et al. (1994) and Fontaine & Deutsch
(1995). However, all of these studies are limited in terms of the range of Reynolds
numbers over which data were obtained. The main influence, pervading much of the
work reported here, comes from Townsend’s seminal work concerning the self-similar
structure of attached wall eddies, which forms the basis of the supposed self-similarity
of the spectra for the surface-parallel velocities and the functional forms for the
normal stresses. This stimulated further work, principally that of Perry & Abell (1975,
1977) and Perry, Henbest & Chong (1986) (pipes) and Perry & Li (1990), Perry &
Marusic (1995), Marusic & Perry (1995), Marusic, Uddin & Perry (1997) and Jones,
Marusic & Perry (2001) (boundary layers). Marusic & Kunkel (2003) have recently
extended these ideas. In later sections, we interpret the data using the concept of
inactive motion. But, owing to the very significant potential benefits accruing from a
102 J. F. Morrison, B. J. McKeon, W. Jiang and A. J. Smits
self-similar description of wall turbulence, it seems prudent first to examine the exact
requirements for this to be so.

2. Similarity considerations
Zagarola & Smits (1998) show that asymptotic matching of the mean velocity
gradients in the overlap region leads to complete similarity (in the form of the log
law) for 600 < y + < 0.07R + when the Reynolds number is sufficiently high and when
the velocity scales for the inner and outer regions are the same, that is, given by
uτ . (See also Zagarola et al. 1997; McKeon, Li, Jiang, Morrison & Smits 2004) At
smaller y + , they argue that the ratio of inner to outer velocity scales, uτ /uo , is a
function of R + , and that simultaneous matching of both the velocities and velocity
gradients leads to a power law. The appearance of a power law may be regarded as
a form of incomplete similarity and is supported by the data of both Zagarola &
Smits (1998) for 60 < y + < 500 and McKeon et al. (2004) for 60 < y + < 300. The term
‘complete (or self-) similarity’ means that, first, the lengthscale used to normalize the
independent variable, y, in the log argument may be freely chosen, and second, the
von Kármán constant is universal. For this reason the log law is valid using inner
or outer scaling, or even using a rough-surface lengthscale. In fact, demonstration of
complete similarity requires simultaneous collapse using both inner and outer scaling.
Owing to the difficulties in scaling the Reynolds stresses near the wall, there have
been several attempts at finding a more suitable inner velocity scale. Prominent among
the alternatives to uτ is so-called ‘mixed’ scaling (where the velocity scale is (uτ Ucl )1/2 ,
see for example, DeGraaff & Eaton 2000) for the horizontal stresses. Zagarola &
Smits (1998) have suggested that a true outer velocity scale is Ucl − U , where Ucl
is the centreline velocity. For ReD > 2 × 105 , they show that (Ucl − U )/uτ −→ 4.34
although, with more data and a revised analysis, McKeon et al. (2004) suggest that
the constant is 4.28. The use of both mixed scaling and (Ucl − U ) as a velocity scale
is considered in § 5.
Since publication of Townsend’s seminal work, considerable attention has been
devoted to the deduction of spectral forms associated with the self-similar nature of
attached wall eddies. Such self-similarity manifests itself at ‘high’ Reynolds numbers
as a range of streamwise wavenumber, k1 , in which the spectrum φ11 ∝ u2τ k1−1 .
There are several derivations, the earliest provided by Tchen (1953), reappraised
by Hinze (1975), involving the balance between the spectral transfer of energy
by the mean shear and that by inertial interactions of the turbulence – a strong
interaction or ‘resonance’ condition. Tchen’s theory involves several assumptions that
are questionable in highly anisotropic wall turbulence, such as a constant strain rate
and a spherically symmetric eddy viscosity. As such, the individual components are
not distinguished. More pragmatically, it should be noted that a prescribed slope over
some region of wavenumber can usually be found in turbulence spectra on log–log
axes. The simpler theory for pipe flow was proposed by Perry & Abell (1977) and
Perry et al. (1986), but it is equally appropriate for boundary layers (see papers by
Perry and colleagues). The theory has been the subject of much attention, but it
appears that often the existence of self-similarity at practical Reynolds numbers is
taken for granted (Nikora 1999; Högström, Hunt & Smedman 2002), or that its proof
is the result of ab initio assumptions (Kader & Yaglom 1991). Given the prominence
of the theory and, in terms of the Reynolds number, the uniqueness of the present
data, a careful reappraisal is clearly needed (see also Morrison et al. 2002a, b). In this,
 Scaling of the streamwise velocity component in turbulent pipe flow 103
it would appear sensible to focus on that range of y in which the mean velocity is
known to exhibit self-similarity in the form of a log law.
‘Large’ scales (in which the direct effects of viscosity may be neglected) that
contribute to the streamwise velocity component may be scaled using either inner
or outer scales. Outer scaling suggests that y is not important and, taking uτ as the
appropriate velocity scale, dimensional analysis therefore yields
φ11 (k1 ) φ11 (k1 R)
= = g1 (k1 R), (2.1)
Ru2τ u2τ
while, alternatively, inner scaling suggests the exclusion of R as a relevant lengthscale
so that, at higher wavenumbers,
φ11 (k1 ) φ11 (k1 y)
= = g2 (k1 y). (2.2)
yu2τ u2τ
The veracity of these scalings is usually judged by the degree of collapse of the
spectra at wavenumbers lower than that at which spectral transfer (which at high
Reynolds numbers is given by the mean dissipation rate) becomes important. In the
range of wave numbers R −1 < k1 < y −1 over which both (2.1) and (2.2) are valid (that
is collapse is evident with both scalings, as required by asymptotic matching), it then
follows that
φ11 (k1 ) = Ru2τ g1 (k1 R) = yu2τ g2 (k1 y). (2.3)
Dimensional arguments and direct proportionality between g1 and g2 therefore imply
φ11 (k1 R) A1
2
= = g1 (k1 R), (2.4)
uτ k1 R
and
φ11 (k1 y) A1
2
= = g2 (k1 y), (2.5)
uτ k1 y
where A1 is a universal constant. Collapse with both length scales therefore suggests
a self-similar structure such that φ11 (k1 ) ∝ u2τ k1−1 . We will therefore call this situation
‘complete similarity’. In this situation, the only relevant lengthscale is k1−1 itself, and,
owing to the nature of the Fourier transform and because the foregoing analysis is
equally valid for the spanwise velocity component, a self-similar structure would have
to be space-filling in (x, z)-planes parallel to the surface. Now, it is possible that, for
example, while y and uτ might form a complete parameter set to define the motion
in the range of wavenumbers over which collapse is apparent with (2.2), these wave
numbers might, in fact, be too high for collapse to be possible using R and uτ as
in (2.1). Thus simultaneous collapse is not possible. We shall refer to this situation
as ‘incomplete similarity’, in which case the constant A1 in (2.4) and (2.5) cannot be
universal.
Note that this analysis is predicated on two principal assumptions. The first is that
the kinematic viscosity, ν, does not enter the problem. This requires that k1 ν/uτ  1.
In turn, this requires the Reynolds number to be sufficiently high, or equivalently that
y is sufficiently large, such the energy-containing scales are not affected directly by
viscosity. Taking the outer limit to the power-law region for the first moment to be
y + = 500, it would seem unlikely that higher moments would be free of direct viscous
effects below y + ≈ 1000, as shown by the conditional sampling results of Morrison
et al. (1992). The second assumption is that uτ is the correct velocity scale for both
the inner and outer regions. In particular, in conformity with Townsend’s theory,
104 J. F. Morrison, B. J. McKeon, W. Jiang and A. J. Smits
it supposes that inactive motion arises primarily through the influence of attached
eddies and that therefore uτ is the appropriate velocity scale. The analysis does not
specify uτ to be the velocity scale: rather, it specifies that the velocity scale should
be the same with both inner and outer scaling, without which complete similarity
would not be possible. Note also that this analysis does not apply to the wall-normal
velocity component which is blocked at wavenumbers, k1 ∼ y −1 .
In § 4, spectra are presented in premultiplied form on linear–log axes. A linear
ordinate enables a closer scrutiny of scalings than that afforded by a logarithmic one.
In addition, the use of non-dimensional axes ensures that not only the ordinate but
also the area under
+
the spectra is directly proportional to energy. Integration of the
spectra yields u2 = u2 /u2τ . Spectra are therefore in the form
k1 Rφ11 (k1 R)
= h1 (k1 R), (2.6)
u2τ
for outer scaling, and in the form
k1 yφ11 (k1 y)
= h2 (k1 y), (2.7)
u2τ
for inner scaling. In the context of assessing these scalings for data in the present
experiment, it is useful to clarify precisely what the foregoing analysis indicates.
Strictly, as long as ν/uτ  y  R (the Reynolds number is ‘high’), (2.4) and (2.5)
should both show a k1−1 range for R −1  k1  y −1 . However, in order to remove the
ambiguity concerning the relative values of y and R, one alternatively may fix y
in (2.4) and then R in (2.5). Equation (2.6) invites us to retain only R and uτ as
independent variables. Thus while y is fixed, uτ is varied by changing the pressure
drop along the pipe. In practice, this involves a change of Reynolds number (strictly
Kármán number) as changes of R are a little more problematical. This does not
pose a problem as long as the Reynolds number is sufficiently high such that the
wave number range of interest is not directly affected by viscosity. Alternatively, (2.7)
invites the use of y and uτ only as independent variables for any fixed R. In this
case, y can merely be varied (subject to ν/uτ  y  R) at a fixed Reynolds number,
although as long as ν can be neglected, a value of y at any Reynolds number might be
chosen. For brevity, we present spectra (obviously using both inner and outer scaling)
at different y/R for the lowest and highest Reynolds number. The self-similarity (or
not) of the k1−1 range is discussed in the light of the present results in § 5.
Accepting uτ as a velocity scale for both inner and outer regions makes possible
an overlap analysis with y (as well as k1 ) as the dependent variable. The use of uτ is
almost universal (e.g. see Perry & Abell 1977). Wosnik et al. (2000) suggest that this
has to be



R dP0 R dP
uτ = − = − ,
ρ dx ρ dx
as defined by the momentum equation in the outer layer. Note however that this
assumes that the viscous term is negligible and that streamwise homogeneity of v 2
leads to equality of the static-pressure and total-pressure gradients. Their analysis for
the overlap region leads to a logarithmic dependence for all the Reynolds stresses of
the form
+
u2 = A[R + ] ln[y + + a + ] + B[R + ], (2.8)
 Scaling of the streamwise velocity component in turbulent pipe flow 105
where A and B are functions of Reynolds number that are asymptotically constant.
The offset is given by 0 6 a + 6 − 16 and is required to account for the existence of a
‘mesolayer’ (Long & Chen 1981). However, the validity of equation (2.8) rests upon
the scaling of the Reynolds stresses being the same as that for the mean velocity
in the overlap region (hence the log dependence). But it is clear from the foregoing
arguments concerning active and inactive motion that first – as we shall show – there
is no simple scaling between the Reynolds stresses and the mean velocity and, second,
the behaviour of the individual components of the stress tensor depends crucially
on the velocity components involved. It is always useful to remember that the first
moment of a turbulence quantity says nothing about the higher moments: more
physically, this means simply that the frame-dependent mean velocity is never a scale
for the turbulence. These considerations are also relevant to the appropriateness of
mixed scaling.
Based on considerations of the self-similar structure of attached wall eddies,
Townsend (1976), Perry & Abell (1977) and Perry et al. (1986) have suggested other
logarithmic functional forms for the normal stresses of the surface-parallel velocities:
 
+ y
u = B1 − A1 ln
2 − C(y + )−0.5 . (2.9)
R
While (2.9) has a similar functional form to that of (2.8), the log term is obtained by
integrating (2.4) or (2.5) for R −1 < k1 < y −1 . For comparison with the present data, we
take the constants suggested by Perry, Henbest & Chong (1986): B1 = 2.67, A1 = 0.9
and C = 6.06. Previously, Perry & Abell (1977) used constants B1 = 3.53, A1 = 0.8 and
C = 9.54, but this makes no difference to our conclusions concerning the proposed
functional form. Marusic et al. (1997) extended (2.9) to include a wake deviation term,
Wg , appropriate for boundary layers:
   
u2 y y
2
= B1 − A1 ln − Vg [y ] − Wg
+
, (2.10)
uτ R R
which also has a log dependence. Vg [y + ] is a viscous deviation term operative at small
y + , while Wg is effective at large y/R. In § 4 we compare both (2.8) and (2.9) with the
present data.

3. Experimental techniques
Measurements are made in a closed-loop, compressed-air facility, in which an
extruded aluminium pipe with nominal diameter of 129 mm is mounted. The facility
is described in detail by Zagarola & Smits (1998), with particular attention paid to
assembly, alignment and surface finish of the test pipe. Air is driven by an impeller
mounted in a pumping section which is followed by a heat exchanger, the return leg,
a flow-conditioning section and a test section beginning at 160D downstream of the
contraction exit, with all measurements made at 164D. By comparison, the intensity
and spectral measurements of Perry & Abell (1975, 1977) at ReD = 3.0 × 105 , for
example, were performed at 71.9D and 86.2D downstream of the tripping device. Dean
& Bradshaw (1976) show that for channel flow, profiles of flatness of the streamwise
velocity fluctuation are independent of development length after 67 channel heights.
Zagarola & Smits (1998) estimate that the development length increases from 78D
to 131D for an increase of Reynolds number from ReD = 3.0 × 105 to 4.0 × 107 .
Therefore, in the case of the present measurements of the streamwise velocity
106 J. F. Morrison, B. J. McKeon, W. Jiang and A. J. Smits

Spatial resolution Averaging time

ReD R+
 Scaling of the streamwise velocity component in turbulent pipe flow 107
For comparison, Li et al. (2004) also use Freymuth’s solution for (3.1) (see also
Corrsin 1963) that ignores heat generation within and convection from the stubs.
This requires the assumption that the wire-end temperature is that of the wire–stub
junction, while the Freymuth ‘conduction-only’ model takes the wire-end temperature
to be the ambient temperature. The agreement is very good, except when l/d is so
small ( < 100) and K is large so that the model is very sensitive to the choice of wire-
end temperature. This agreement may be used as a justification for making estimates
of σ using the conduction-only model. This has also been shown to be accurate by
Morris & Foss (2003) who use a time-dependent thermal model of a hot wire with a
modelled ideal feedback amplifier to estimate σ .
Table 1 shows that, at the lowest Reynolds number, data are taken with a hot wire
for which l/d = 200. In this case, Rew ≈ 3 and therefore the criterion σ < 7% is met.
Table 1 further shows that end-conduction losses are most critical for the data set
ReD = 4.1 × 105 , the lowest Reynolds number at which wires with l/d = 100 are used.
For these data, 3.7 6 Rew 6 11.2. The calculations of Li et al. (2004) for a platinum
wire suggest that, in order to meet the criterion of σ < 7%, l/d > 145 at Rew = 3.7.
However, the conduction losses are increased by the use of, for this data set, tungsten
wire for which the thermal conductivity is about twice that of platinum. Yet, all the
hot wires used here had much longer stub lengths than those for which calculations
were performed. The calculations therefore suggest a worst-case estimate of σ ≈ 15%
for the data point closest to the wall.
Of more importance are the estimates of σ and σ for the data taken at the highest
Reynolds number, ReD = 5.7 × 106 , and for which Rew ≈ 250. Here, calculations
indicate that for platinum wire, l/d may be reduced to about 50, although precise
limits depend on the value of Rew . Use of the conduction-only model for a platinum
wire of length l/d = 100 operated with a = 0.82 and Rew = 250 gives σ = 2.3% and
σ = 0.84%. Therefore, end-conduction effects at the highest Reynolds number are
significantly less than those for wires with l/d = 200 at Rew ≈ 3. Li et al. (2004) provide
full details. Other less critical issues are related to the bridge frequency response. These
include the generation of heat waves along the wire and bridge stability at high output
voltages. These are also dealt with by Li et al. (2004).
All statistics are calculated as moments of probability density functions (p.d.f.).
Spectra are calculated using, typically, data records of 1800 s duration. Table 1 shows
the principal parameters governing the flow conditions as well as details concerning
the spatial and temporal resolution of the data. The spatial resolution is expressed
non-dimensionally as l + = luτ /ν as well as k1 η, where η is the Kolmogorov lengthscale
and k1 l = 2π, which is independent of the pipe flow conditions; η is deduced from the
local-equilibrium approximation of the dissipation rate,
, at mid log region and is
typically 10% larger than that deduced from the third-order structure function. The
sampling frequency was varied between 20 kHz and 100 kHz and was set so that the
Nyquist frequency expressed as a wavenumber, k1 = 2πf/U , exceeds that equivalent
to the limit of spatial resolution for the worst case situation of data at the pipe
centreline. Typically, the bridge frequency response was 60–75 kHz with a ≈ 0.8. The
signal was low-pass filtered at the Nyquist frequency and standard FFT algorithms
(Hanning window) are used to calculate the spectra as a function of k1 .
Convergence of p.d.f. moments is better than 1%, except in the case of the third
and fifth moments in the region 10 6 y + 6 30 where it is better than 10% only. This
increased error can plausibly be attributed to flow behaviour which induces large
changes of both magnitude and sign. The fact that the fourth and sixth moments
have smaller errors can be explained by the nonlinear behaviour of all anemometers.
108 J. F. Morrison, B. J. McKeon, W. Jiang and A. J. Smits
10 ReD
5.5 × 104
8 7.5 × 104
1.5 × 105
2.3 × 105
6 4.1 × 105
1.0 × 106
—2 + 3.1 × 106
u
4 5.7 × 106

0
100 101 102 103 104 105
y+

Figure 1. Second moment: wall scaling.

This should be regarded as a minor limitation of the thermal anemometry technique

only (klewicki & Falco 1990 find a similar result) rather than an indication of any
anomaly in the current data sets. Near the wall, convergence improves as the Reynolds
number increases, probably because of spatial filtering. However, outside of the range
10 6 y + 6 30, convergence does not improve with increasing Reynolds number and
this indicates that all the averaging times are sufficient (table 1). The reduced averaging
times at higher Reynolds number are simply the result of the faster sample rate and
the resulting larger data sets. Even these averaging times are significantly longer than
those suggested by Klewicki & Falco (1990) for measurements of up to fourth-order
moments in a boundary layer (T Ue /δ = 4000) and Blackwelder & Eckelmann (1979)
for fourth-order velocity statistics in pipe flow (T Ucl /R = 2870).

4. Results
4.1. Spectra and second moments
The second moment, normalized by u2τ and plotted against y + shows two
u2 ,
maxima (figure 1): the first, prevalent at low Reynolds numbers, is well-documented
(Laufer 1954; Sandborn 1955; Eggels et al. 1994; Durst et al. 1995). The data of
+
den Toonder & Nieuwstadt (1997) (obtained using LDA) suggest that u2 reaches a
maximum of about 7.3 that is constant with Reynolds number up to about 2.5 × 104
(see also Mochizuki & Nieuwstadt 1996; Durst et al. 1995). However, the present
data show that this maximum is, in fact, Reynolds-number dependent, reaching 8.6
at ReD = 7.5 × 104 . At higher Reynolds numbers (at ReD = 1.5 × 105 , l + ≈ 25), the
reduction in this peak with increasing Reynolds number is, of course, the result
of poorer spatial resolution. Ligrani & Bradshaw (1987) note that measurements
of u2 are significantly attenuated for l + > 20 − 25. The peaks at ReD = 5.5 × 104
and 7.5 × 104 have values of 6.6 and 8.5, respectively. For comparison, the near-
wall peak at ReD = 5.0 × 104 measured by Laufer (1954) has a value of 6.9, and
that measured by Perry & Abell (1975) at ReD = 7.8 × 104 has a value of 9.5. The
Reynolds-number dependence of the near-wall peak is, in some sense, consistent with
the concept of inactive motion, namely that it increases with Reynolds number –
our first principal conclusion. Mochizuki & Nieuwstadt (1996) suggest that the
position of this peak is also independent of Reynolds number at y + ≈ 15. The present
data do not contradict this, although owing to the effects of probe resolution, no
 Scaling of the streamwise velocity component in turbulent pipe flow 109
y/R y+
1.0 0.030 45
0.051 76
0.8 0.096 144
k1yφ11(k1y)/uτ2 0.203 303
0.279 417
0.6

0.4

0.2

0
10–3 10–2 10–1 100 101 102 103
k1y

Figure 2. Inner scaling, ReD = 5.50 × 104 , R + = 1.50 × 103 .

y/R y+
0.030 3.0 × 103
1.5
0.051 5.1 × 103
0.096 9.6 × 103
2.0 × 104
k1yφ11(k1y)/uτ2

0.202
1.0 0.279 2.8 × 104

0.5

0
10–3 10–2 10–1 100 101 102 103
k1 y

Figure 3. Inner scaling, ReD = 5.7 × 106 , R + = 105 .

firm conclusions may be drawn. In the context of pipe flow, the second maximum
at y + ≈ 500 and appearing only for ReD > 2 × 105 is altogether a new phenomenon
although its appearance in boundary layer data is well-documented: comparison
of boundary-layer laboratory data (Fernholz & Finley 1996) with those from the
atmospheric surface layer (Marusic & Kunkel 2003) suggest that the magnitude of
the second maximum increases indefinitely with Reynolds number. Its position, yp+ ,
also increases with Reynolds number. We examine the behaviour of both peaks in
more detail in § 5.
Using inner scaling, figure 2 shows φ11 (k1 y) in the form given by (2.7) for
ReD = 5.5 × 104 over the range in y for which collapse might be expected. Note
that this corresponds to a range of y + over which the first moment displays a power
law. Figure 3 shows equivalent data for ReD = 5.7 × 106 plotted in the same form.
In this case, the range of y + corresponds to that of the log law except for the two
points furthest from the wall. In figure 2, it is evident that the Reynolds number is
simply too low for collapse to be possible. Note that R + = 1500 only and that the
direct effects of viscosity permeate the whole layer. At the highest Reynolds number
(figure 3), there is some collapse in the region k1 y ≈ 1.0. However, the collapse is not
110 J. F. Morrison, B. J. McKeon, W. Jiang and A. J. Smits
y/R y+
0.030 3.0 × 103
1.5 0.051 5.1 × 103
0.096 9.6 × 103
2.0 × 104

k1Rφ11(k1R)/uτ2
0.202
1.0 0.279 2.8 × 104

0.5

0
10–1 100 101 102 103 104
k1R

Figure 4. Outer scaling, ReD = 5.7 × 106 , R + = 105 .

ReD y+
5.5 × 104144
1.5
3.1 × 106 5180
5.7 × 106 9620
k1Rφ11(k1R)/uτ2

1.0

0.5

0
10–1 100 101 102 103 104
k1 R

Figure 5. Outer scaling, y/R = 0.1.

along a horizontal line (corresponding to self-similar scaling), suggesting incomplete

similarity only.
Figure 4 shows the same data as in figure 3, but plotted using outer scaling. For
k1 R ∼ 1, there is no significant collapse, although the peaks for 0.033 6 y/R 6 0.107
lie closer together than the others. Inspection of figure 3 in the region of k1 y ∼ 0.1
shows that the same data (0.033 6 y/R 6 0.107) clearly do not collapse using
inner scaling. Instead, spectra for y/R = 0.033, 0.063 and 0.107 show discrete peaks,
appearing in a wavenumber sequence determined by y −1 , equivalent to the collapse
in figure 4 occurring at a point, k1 R ≈ 0.75. Since collapse occurs neither with outer
variables nor with inner variables, not even incomplete similarity is possible. At
best, incomplete similarity is apparent only for k1 R 6 1 using outer scaling, and for
k1 y > 0.06 using inner scaling.
As suggested in § 2, an alternative scrutiny of the data may be achieved by using
outer scaling at different Reynolds numbers while fixing y/R = 0.1. In figure 5, there is
some degree of collapse for k1 R ≈ 0.8 at the two higher Reynolds numbers. However,
the equivalent data scaled with inner variables (figure 3) do not collapse so that again,
at best, only incomplete similarity is possible. Note also that complete similarity would
 Scaling of the streamwise velocity component in turbulent pipe flow 111
10 ReD
5.5 × 104
7.5 × 104
8 1.5 × 105
2.3 × 105
4.1 × 105
6 1.0 × 106
3.1 × 106
—2 +
u 5.7 × 106
Eq.(2.9):5.5 × 106
4
Eq.(2.9):5.7 × 106

0 –3
10 10–2 10–1 100
y/R

Figure 6. Second moment: outer scaling.

imply that A1 takes a universal value, but it is clear from all the spectra that this is
not the case.
Interestingly, as has been shown by Kim & Adrian (1999) and Jiménez (1998), the
spectra show the presence of very long structures near the wall that give rise to a
bimodal shape at low Reynolds numbers, figure 2. Their wavelength increases as y
increases, reaching a peak of about 10R at y/R ≈ 0.1 before decreasing at larger y.
Figure 5 suggests that the length of these structures at a fixed y does not change with
Reynolds number (see Jiménez, Flores & Garcı́a-Villalba 2002).
Table 1 indicates that, at the highest Reynolds number, ReD = 5.7 × 106 , l + = 385. It
is therefore important to assess the effect of hot-wire spatial averaging on the spectra
of figures 3 and 4, for which the critical parameter is k1 l|max = 2π. For the spectra in
figure 3, this suggests full spectral resolution up to k1 y|max = 2πy/ l = 49, 83, 156, 330
and 454 for, respectively, positions y/R = 0.030 − 0.279. Our inference is therefore
that these spectral estimates (and, in particular, the deductions from them concerning
self-similarity) are free of the effects of limits to spatial resolution. The effect of spatial
+
resolution on u2 is dealt with in § 5.
Figure 6 shows the data of figure 1 re-plotted against y/R. There is a striking
collapse in the outer region, y/R > 0.4, except at the lowest Reynolds number. There
is no collapse in the overlap region, and, as figure 1 shows, no collapse anywhere
using wall variables, except presumably very close to the wall. Note that the near-wall
peak now depends on y/R: this provides a second principal conclusion, namely that
inactive motion depends on distance from the wall. This is a result of general validity
made trivial here because the near-wall peak in figure 1 occurs at approximately a
constant value of y + and R is a constant. Figure 6 also shows a comparison with
(2.9): at ReD = 5.7 × 106 when the viscous deviation term is small, the agreement of
the slope (determined by A1 ) is quite good, but the offset (determined by the additive
constant B1 ) is too small. The changes with Reynolds number even at y/R = 0.1
derive from the viscous deviation term, which qualitatively predicts the diminishing
influence of viscosity outside the sublayer as the Reynolds number increases. At
ReD = 5.5 × 104 however, the behaviour of (2.9) at small y/R is incorrect owing
to the large inactive contribution. This is one of the two reasons for the change
112 J. F. Morrison, B. J. McKeon, W. Jiang and A. J. Smits
in gradient of the data (and therefore A1 ) with Reynolds number. Moreover, for
0.02 < y/R < 0.1, this change in gradient is not monotonic, first decreasing before
increasing. This behaviour is therefore indicative of two effects: the increase in the
inactive contribution with increasing Reynolds number (which is partly obscured by
the poorer resolution as the Reynolds number increases), and the reduction in direct
viscous effects emanating from the sublayer as the Reynolds number increases. While
the latter is estimated quite well (but only for ReD > 106 ), no account of the former is
taken in (2.9), which appears therefore to require an additional term quantifying the
inactive contribution, the form of which is strongly dependent on the choice of outer
velocity scale.
If a spectral self-similar range exists, such that φ11 (k1 ) ∝ k1−1 (‘complete’ similarity),
the constant of proportionality, A1 , in (2.9) is universal. However, the evidence of
figure 6 is that the slope of the data in the vicinity of y/R ≈ 0.1 (where the viscous
deviation is negligible and where a k1−1 range is most likely) is still increasing at
the highest Reynolds number. One should also bear in mind that u2 , as the integral
of φ11 , is less sensitive to Reynolds-number scalings than the integrand itself. It is
possible that at even higher Reynolds numbers, the slope of u2 may asymptote to
a constant value indicative of complete similarity in φ11 . With the omission of the
viscous deviation term, (2.9) may be re-written as
+
u2 = B1 − A1 ln[y + ] + A1 ln[R + ]. (4.1)
This shows that the outer peak in figure 1 will increase indefinitely with Reynolds
number, regardless of considerations of the universality of A1 . One might suppose
that, at some stage, the outer peak might become larger than the inner peak in the
sublayer. However, this is unlikely and it is more likely that the inactive motion
near the wall will continue to increase with Reynolds number as long as its source
in the outer region does. Using data from the atmospheric surface layer for which
Reθ ≈ 5 × 106 , Metzger & Klewicki (2001) show that the near-wall peak does increase
without any apparent indication of an asymptote.
Similarly, (2.8) may be rewritten as
 
+ y
2
u = B + A ln + A ln[R + ], (4.2)
R
in which the offset, a + , being much smaller than y + is neglected. For clarity, the
Reynolds-number dependence of A and B is not made explicit and we take their
values to be approximately the same as those of Perry et al. (1986) since the functional
forms of (4.2) and (2.9) are very similar. Thus −A ≈ A1 = 0.9 to ensure that the slope
of
 Scaling of the streamwise velocity component in turbulent pipe flow 113

100 y+
60
255
598
10–1 2653
8559
χ

–2
P(α) 10

10–3

10–4
–4 –2 0 2 4
α

Figure 7. P.d.f.s of u for ReD = 4.1 × 105 . χ for data at the centreline.

be expected to be substantially free of resolution effects. The p.d.f. is defined by

Ni
P (ui ) = (4.3)
N u
where N is the total number of occurrences, Ni is the number of occurrences for
any velocity, ui , and u is the bin size. Here, normalization is performed with the
first moment subtracted so that α = u/σ (where σ is the standard deviation) and
therefore
 ∞
P (α) dα = 1. (4.4)
−∞

Also shown are the estimates of r.m.s. relative error, χ, at y + = R + = 8559 as defined
by Lumley (1970):

2T
χ= , (4.5)
Ni δt
where the integral timescale is estimated as T = R/uτ and δt is the sampling interval.
At any position, u is set by dividing the difference between the signal maximum
and minimum into one hundred equal intervals. Thus χ is bound to be large at the
extremes of the p.d.f., especially when the higher moments are numerically large, as
occurs both near the wall and the pipe centreline. Figure 8 shows p.d.f.s for the range
5.5 × 104 6 ReD 6 5.7 × 106 at y + ≈ 600, which is at the lower limit of the log-law
region except for those data at ReD < 4.1 × 105 .
Using DNS data equivalent to R + 6 395, Dinavahi, Breuer & Sirovich (1995) suggest
that p.d.f.s of velocity fluctuations in turbulent channel flow exhibit universality in
that, outside the viscous sublayer, they are independent of distance from the wall.
In contrast, figure 7 shows that these p.d.f.s do not collapse: for example, they
become progressively more negatively skewed with increasing distance from the wall.
Moreover, the maximum value of P changes from 0.38 at y + = 60 to 0.43 at the
centreline. Dinavahi et al. (1995) also suggest that p.d.f.s obtained by averaging
114 J. F. Morrison, B. J. McKeon, W. Jiang and A. J. Smits
100 ReD
5.5 × 104
7.5 × 104
1.5 × 105
10–1 2.3 × 105
4.1 × 105
1.0 × 106
3.1 × 106
5.7 × 106
P(α) 10–2

10–3

10–4
–4 –2 0 2 4
α

Figure 8. P.d.f.s of u at y + ≈ 600.

ReD
8 5.5 × 104
7.5 × 104
1.5 × 105
4 2.3 × 105
4.1 × 105
1.0 × 106
—3 + 0
u 3.1 × 106
5.7 × 106

–4

–8

100 101 102 103 104 105

y+

Figure 9. Third moment: wall scaling.

those at several distances from the wall in the range 63 6 y + 6 318, are ‘relatively
independent’ of Reynolds number. It is clear that the p.d.f.s are neither constant with
y + at a given Reynolds number (figure 7) nor constant at a given y + for a range of
Reynolds numbers (figure 8), although in the case of the latter, the collapse is better.
Generally, the differences become larger as the moment order increases.
Therefore a more precise assessment of the velocity field comes from examining the
moments of the p.d.f.s. Figures 9–12 show, in sequence, moments from the third to
the sixth scaled using wall variables. While the even moments are similar in shape to
each other (figures 1, 10 and 12), the odd moments are also similar to one another
+
(figures 9 and 11). Thus the behaviour of the peaks in u2 is very similar to that for
the fourth and sixth moments. There is no scaling with wall variables anywhere.
 Scaling of the streamwise velocity component in turbulent pipe flow 115
200 ReD
5.5 × 104
7.5 × 104
150 1.5 × 105
2.3 × 105
4.1 × 105
1.0 × 106
—4 + 100 3.1 × 106
u
5.7 × 106

50

0
100 101 102 103 104 105
y+

Figure 10. Fourth moment: wall scaling.

400 ReD
5.5 × 104
300 7.5 × 104
1.5 × 105
200 2.3 × 105
4.1 × 105
1.0 × 106
—5 + 100 3.1 × 106
u
5.7 × 106
0

–100

–200
100 101 102 103 104 105
y+

Figure 11. Fifth moment: wall scaling.

6000 ReD
5.5 × 104
5000 7.5 × 104
1.5 × 105
4000 2.3 × 105
4.1 × 105
—6 + 3000 1.0 × 106
u 3.1 × 106
5.7 × 106
2000

1000

0
100 101 102 103 104 105
y+

Figure 12. Sixth moment: wall scaling.

In order to compare statistics with those for a Gaussian p.d.f., the same moments
are scaled using the appropriate power of the second moment. Figures 13–16 show
the skewness, Su , flatness, Fu , superskewness, SSu and superflatness, SFu , respectively,
plotted against y + . As den Toonder & Nieuwstadt (1997) suggest, these statistics
116 J. F. Morrison, B. J. McKeon, W. Jiang and A. J. Smits
1.5 ReD
5.5 × 104
1.0 7.5 × 104
1.5 × 105
2.3 × 105
4.1 × 105
0.5 1.0 × 106
Su 3.1 × 106
5.7 × 106
0

–0.5

–1.0
100 101 102 103 104 105
y+

Figure 13. Skewness: wall scaling.

6 ReD
5.5 × 104
7.5 × 104
1.5 × 105
5
2.3 × 105
4.1 × 105
1.0 × 106
Fu 4 3.1 × 106
5.7 × 106

2
100 101 102 103 104 105
y+

Figure 14. Flatness: wall scaling.

20 ReD
5.5 × 104
15 7.5 × 104
1.5 × 105
10 2.3 × 105
4.1 × 105
1.0 × 106
SSu 5 3.1 × 106
5.7 × 106
0

–5

–10 0
10 101 102 103 104 105
y+

Figure 15. Superskewness: wall scaling.

are Reynolds-number dependent even though the changes are modest. However, Fu
and SFu show very good collapse for 30 6 y + 6 300 and, at the higher Reynolds
+
numbers, the position of the outer peak of u2 , yp+ , lies within the region of collapse.
 Scaling of the streamwise velocity component in turbulent pipe flow 117
80 ReD
5.5 × 104
7.5 × 104
60 1.5 × 105
2.3 × 105
4.1 × 105
1.0 × 106
SFu 40 3.1 × 106
5.7 × 106

20

0 0
10 101 102 103 104 105
y+

Figure 16. Superflatness: wall scaling.

1.5 ReD
5.5 × 104
1.0 7.5 × 104
1.5 × 105
2.3 × 105
0.5 4.1 × 105
1.0 × 106
Su 3.1 × 106
0 5.7 × 106

–0.5

–1.0
10–3 10–2 10–1 100
y/R

Figure 17. Skewness: outer scaling.

Note that, as elsewhere, in this region none of the statistics have Gaussian values
(Su = SSu = 0, Fu = 3 and SFu = 15). Near the wall, Su and Fu are both large and
positive, while near the centreline, Su is large and negative and Fu is again large and
positive. Physically, eruptions from the near-wall region arrive at the centreline with
low streamwise momentum, and conversely, wallward-moving eddies arrive at the
wall with high streamwise momentum. Any changes with Reynolds number of the
former effect cannot be deduced owing to the effects of probe resolution. However, at
the lowest Reynolds number, Su is still increasing at smaller values of y + than that at
which u2 reaches its sublayer maximum, at which point (y + ≈ 15), Su is close to zero
(±0.2) as is SSu (±1.0). Meanwhile, Fu ( ≈ 2.4) and SFu (≈ 9.0) both exhibit minima
here, as suggested by den Toonder & Nieuwstadt (1997) and Durst et al. (1995). Note
however, that these values are also non-Gaussian.
For brevity, we show only Su and Fu plotted against y/R in figures 17 and 18.
As with inner scaling, the degree of collapse is better for even moments than odd
ones, but generally, it improves with increasing moment order and increasing y/R.
At y/R ≈ 0.8, Su show minima and Fu show maxima. As will be seen later, this is
not an indication of asymmetry, merely that the point of maximum departure from
Gaussian occurs at 0.2R either side of the centreline. This is probably due to the
118 J. F. Morrison, B. J. McKeon, W. Jiang and A. J. Smits
6 ReD
5.5 × 104
7.5 × 104
5 1.5 × 105
2.3 × 105
4.1 × 105
1.0 × 106
Fu 4 3.1 × 106
5.7 × 106

2
10–3 10–2 10–1 100
y/R

Figure 18. Flatness: outer scaling.

influence of the ‘geometry effect’ (Wei & Willmarth 1989), or its equivalent in pipe
flow, in which eddies from all azimuthal locations converge on the centreline. The
changes in moment behaviour for 0.2R either side of the pipe axis are consistent with
the increasing influence of fast-moving eddies oscillating about the centreline with a
frequency of occurrence that is slightly less than that of near-wall eruptions reaching
the centreline. Interestingly, this effect appears not to diminish as the Reynolds
number increases, as evidenced by the persistence of outer scaling down to about
0.4R. Without reference to the wall-normal velocity component, it is not possible to
say over what range of radial positions the geometry effect is prevalent.
Tsuji & Nakamura (1999) have proposed the use of the Kullback Leibler Divergence
to measure the departure of p.d.f.s of the streamwise velocity from Gaussian in a
turbulent boundary layer. They suggest that the extent of the log law is defined
as that over which the p.d.f. is ‘similar’ when normalized by the second moment.
Given the difficulties mentioned in § 2 of using the first moment to make inferences
about the behaviour of the higher ones (or vice versa), here we test this assertion for the
present pipe flow in which the extent of the log-law region has been closely delineated
(Zagarola & Smits 1998; McKeon et al. 2004). In the range 600 6 y + 6 0.12R + , from
the results of figures 17 and 18, it is clear that, although Su and Fu are relatively
constant, their variation is such that this definition is unlikely to be as rigorous as
that provided by any form of overlap analysis (Zagarola & Smits 1998; Wosnik et al.
2000; McKeon et al. 2004). Indeed, all of the results here are consistent with the
notion that the turbulence, as exemplified by the second and higher moments of the
streamwise velocity, does not show any self-similarity. The departure of these statistics
from Gaussian is dealt with more fully in the next section.

5. Discussion
5.1. Alternatives to uτ as a velocity scale
Zagarola & Smits (1998) have used the data of den Toonder & Nieuwstadt (1997) to
show that Ucl − U collapses the second moment better than uτ for 0.2 < y/R < 1.0,
for the range 4.9 × 103 6 ReD 6 2.5 × 104 . Note that Ucl − U is frame-invariant. Here
we test this velocity scale using data over a much wider range of Reynolds numbers.
Given the lack of success in the use of uτ to collapse moments in the inner region,
mixed scaling (DeGraaff & Eaton 2000; Metzger et al. 2001) as an alternative is
 Scaling of the streamwise velocity component in turbulent pipe flow 119
0.4 ReD
5.5 × 104
7.5 × 104
0.3 1.5 × 105
2.3 × 105
4.1 × 105
—2 m 0.2 1.0 × 106
u 3.1 × 106
5.7 × 106

0.1

0 0
10 101 102 103 104 105
y+

Figure 19. Second moment: mixed scaling, y + abscissa.

also assessed. This is especially important because evidence of an inner velocity

scale that is different from the outer one immediately prevents self-similarity for the
velocity spectrum in terms of a k −1 region. It also raises questions concerning the
physical mechanisms that might give rise to velocity scales other than uτ : for instance,
Ucl −U as a velocity scale would suggest freely moving, convected large structures (see
Morrison et al. 1992). We do not consider further the analysis of Tsuji & Nakamura
+ +
(1999) who use an expression for u2 in the log-law region that only applies if u2
decreases monotonically.
5.1.1. Inner velocity scale
As a justification for mixed scaling, DeGraaff & Eaton (2000) suggest, following
Rotta (1962), that the total rate of energy dissipation in a boundary layer depends
on u2τ U . Rotta (1962) suggests that a good approximation to the total energy balance
close to the wall is given by
 2
2 dU dU dU
uτ = − uv +ν , (5.1)
dy dy dy
where the left-hand side may be taken to be the transport of energy from the outer
region to the wall region. However, the results of § 4 suggest that that the left-hand side
of (5.1) is more likely to scale as u3τ /y and not u2τ Ucl /y. Note that the local-equilibrium
hypothesis (a result of applying wall scaling to the turbulence kinetic energy equation)
suggests
∝ u3τ /y. Such a scaling is consistent with the notion that both active and
(to a first approximation) inactive components arise through the presence of attached
wall eddies. As we have seen in § 2, a particular problem with the choice of a velocity
scale involving any mean velocity is that it is frame-dependent. Moreover, there is no
physical reason why the geometric mean of two velocity scales might be appropriate.
Thus the use of (uτ Ucl )1/2 would appear to have little physical basis. Leaving aside the
m
issue of a corresponding mixed lengthscale, u2 (u2 normalized by uτ Ucl ) is plotted
against y + in figure 19 and against y/R in figure 20. In the case of the former, the
collapse is no better than that of wall scaling in figure 1, a result that is apparently
at odds with the conclusion of Metzger et al. (2001) using data from the atmospheric
surface layer. However allowing for the poorer resolution in the present data, the
changes in ordinate value at y + ≈ 15 (0.3−0.4) are similar in both cases. This suggests
that the choice of Ucl masks a more subtle effect. In the case of figure 20, the collapse
120 J. F. Morrison, B. J. McKeon, W. Jiang and A. J. Smits
0.4 ReD
5.5 × 104
7.5 × 104
0.3 1.5 × 105
2.3 × 105
4.1 × 105
—2 m 0.2 1.0 × 106
u 3.1 × 106
5.7 × 106

0.1

0
10–3 10–2 10–1 100
y/R

Figure 20. Second moment: mixed scaling, y/R abscissa.

10 ReD
4.9 × 103
8 1.0 × 104
1.8 × 104
2.5 × 104
5.5 × 104
6
7.5 × 104
—2 +
u 1.5 × 105
2.3 × 105
4

0
10–2 10–1 100
y/R

Figure 21. Second moment: outer scaling, u0 = uτ . Legend shows data in sequence of in-
creasing Reynolds number. Solid symbols indicate data from den Toonder & Nieuwstadt
(1997).

is demonstrably worse than the outer scaling of figure 6. Similarly, there is no collapse
when using the mean velocity as a velocity scale (not shown) except at the centreline.
5.1.2. The outer velocity scale
Since uτ and Ucl − U are related only by a constant of proportionality for
ReD > 2 × 105 (McKeon et al. 2004), the question of the better velocity scale relates
to those data for ReD < 2 × 105 only, that is, the data sets at the four lower Reynolds
numbers. In order to extend the comparison of velocity scales over a wider range
+ ∆
of Reynolds numbers, un and un = un /(Ucl − U )n are compared using the present
data for ReD 6 2.3 × 10 together with those of den Toonder & Nieuwstadt (1997),
5

for which 4.9 × 103 6 ReD 6 2.5 × 104 .

Figures 21–23 show, respectively, second-, third- and fourth-order moments, scaled
with outer scales taking uo = uτ . There is a striking collapse of all moments in the outer
region for y/R > 0.4. There is no collapse in the overlap region, as there is no collapse
using inner variables (figures 1 and 7–12) – the overlap region relates only to the
mean velocity and not to the turbulence statistics. Note that the inward progression
 Scaling of the streamwise velocity component in turbulent pipe flow 121
8 ReD
4.9 × 103
1.0 × 104
4 1.8 × 104
2.5 × 104
5.5 × 104
—3 + 0 7.5 × 104
u 1.5 × 105
2.3 × 105

–4

–8
10–2 10–1 100
y/R

Figure 22. Third moment: outer scaling, u0 = uτ .

ReD
160
4.9 × 103
1.0 × 104
1.8 × 104
120 2.5 × 104
5.5 × 104
—4 + 7.5 × 104
u 80 1.5 × 105
2.3 × 105

40

0
10–2 10–1 100
y/R

Figure 23. Fourth moment: outer scaling, u0 = uτ .

+ +
of the peak in both u2 and u4 is consistent with the Reynolds-number changes
over both data sets.
The same data are re-plotted with u0 = Ucl − U in figures 24–26 where estimates of
Ucl − U for the present data are estimated from a curve fit to the data of Zagarola &
Smits (1998). Use of the equivalent data from McKeon et al. (2004) gives a change in
uτ of about 0.3%, which is very close to the expected error in uτ (Zagarola & Smits
1998). Comparison of figures 21 and 24, of figures 22 and 25, and of figures 23 and
26 is especially revealing: it supports the assertion by Zagarola & Smits (1998) that
Ucl − U is a better outer velocity scale than uτ for the second moment for y/R > 0.35,
approximately. While the triple product appears to scale equally well with either
velocity scale, the fourth-order moment collapses better with Ucl − U . However, the
improved scaling only appears to affect the data for ReD 6 6 × 104 : that is, all of the
data sets from den Toonder & Nieuwstadt (1997) together with one of the present
data sets at the lowest Reynolds number. In figures 24 and 26, the behaviour almost
suggests two separate ranges of Reynolds number over which the data collapse for
y/R > 0.35: one for ReD < 6 × 104 , the other for ReD > 7 × 104 . Note that this result is
unlikely to affect the dimensional arguments for a k1−1 range in the spectra since the
122 J. F. Morrison, B. J. McKeon, W. Jiang and A. J. Smits
0.5 ReD
4.9 × 103
0.4 1.0 × 104
1.8 × 104
2.5 × 104
5.5 × 104
0.3 7.5 × 104
—2 ∆ 1.5 × 105
u 2.3 × 105
0.2

0.1

0
10–2 10–1 100
y/R

Figure 24. Second moment: outer scaling, u0 = Ucl − U . Solid symbols indicate data from
den Toonder & Nieuwstadt (1997).

0.10 ReD
4.9 × 103
1.0 × 104
0.05 1.8 × 104
2.5 × 104
5.5 × 104
7.5 × 104
—3 ∆ 0 1.5 × 105
u 2.3 × 105

–0.05

–0.10
10–2 10–1 100
y/R

Figure 25. Third moment: outer scaling, u0 = Ucl − U .

0.5 ReD
4.9 × 103
0.4 1.0 × 104
1.8 × 104
2.5 × 104
5.5 × 104
0.3 7.5 × 104
—4 ∆ 1.5 × 105
u
2.3 × 105
0.2

0.1

0
10–2 10–1 100
y/R

Figure 26. Fourth moment: outer scaling, u0 = Ucl − U .

 Scaling of the streamwise velocity component in turbulent pipe flow 123
ratio of Ucl − U to uτ reaches a constant for ReD > 2 × 105 and the choice of velocity
scale does not matter as long as it is invariant with Reynolds number. It would also
suggest that the outer region is more likely to be dominated by large, freely moving
detached structures, or at least those that do not conform to Townsend’s concept of
an attached wall eddy. This suggestion may well be realistic: Perry & Marusic (1995)
and Marusic & Perry (1995) show that prediction of the Reynolds stresses using
their attached wall eddy still requires an additional component more like a detached
structure.
5.2. Self-similarity
We have illustrated the importance of, first, the direct effects of viscosity outside
the viscous sublayer at low Reynolds numbers and, second, the influence of inactive
motion at high Reynolds numbers. Our most important result is perhaps that, while
the concept of inactive motion serves as a very useful qualitative description of the
low-wavenumber motion near the wall, it is somewhat of a misnomer. Westbury &
Morrison (1995) have previously noted that the distinction between the large-scale
inactive motion and the associated shear-stress-bearing sweeps is not necessarily
as clear cut as Townsend’s ideas might suggest. The evidence of figure 1 is that
+
inactive motion increases u2 just where the production of turbulence kinetic energy
reaches a maximum. Thus the outer eddies are, in fact, active in terms of producing
energy by the shear that they induce. This result underscores the importance of
the ‘top-down’ influence at high Reynolds numbers identified by Hunt & Morrison
(2000). While the alternative ‘bottom-up’ effect, in which it is supposed that near-wall
instabilities drive the momentum flux, may be an adequate description at very low
Reynolds numbers, the foregoing suggests that these two effects may be reconciled
by recognizing that the Reynolds number is a measure of the ratio of one to the
other. This interpretation suggests that the near-wall motion is not autonomous as
suggested by the studies of Jiménez & Pinelli (1999) and Jiménez et al. (2002) using
DNS data. More recently, Jiménez, del Alámo & Flores (2004) have suggested that
near-wall motion also consists of large structures that extend to the outer region.
However, these structures carry little Reynolds stress (they are ‘inactive’ as Townsend
suggested). This suggests that the technique used to mask the outer-region influence
does not mimic the Reynolds-number effects reported here.
It has been suggested to us by some reviewers that a self-similar k1−1 range is
more likely to exist closer to the wall than the range of y + represented by the data of
figure 3. This was examined in earlier data sets in Morrison et al. (2001) who show that
for y + < 1000, the direct effects of viscosity prevent any form of self-similarity. The
spectra of figures 3 and 4 therefore represent that range of y over which self-similarity
is most likely to be apparent.
It is interesting to note that, as suggested by figure 4, the lower limit to the region
in which complete similarity of the low-wavenumber motion is most likely to exist,
0.033 6 y/R 6 0.107, is equivalent to y + ≈ 5000 at ReD = 5.7 × 106 , and that collapse
of the spectra is significantly worse at lower Reynolds numbers. Thus it is very
unlikely that complete similarity in the form of the k1−1 range will be possible below
this Reynolds number. Note that an equivalent boundary-layer Reynolds number is
Reθ ≈ 300 000! Figure 6 shows that the direct effects of viscosity on u2 are apparent
in the outer region for ReD < 105 . On balance, it would appear that while collapse
of the velocity spectra may be possible with either inner or outer scaling (incomplete
similarity), it is unlikely that simultaneous collapse with both in the same wavenumber
range is possible (complete similarity), at least up to the maximum Reynolds number
124 J. F. Morrison, B. J. McKeon, W. Jiang and A. J. Smits
attained here. Thus spectra here do not exhibit a k1−1 range indicative of self-similar
structure, which should therefore be considered as a special case only. The behaviour
of u2 is consistent with the notions that (a), inactive motion increases with Reynolds
number and (b) inactive motion increases as y/R decreases (down to the sublayer).
On the basis of (a) and (b) alone, complete similarity as outlined above seems to
be unlikely because, while the active motion scales on y and uτ only (in the limit
of infinite Reynolds number) as Townsend proposed, the inactive component always
requires three scales, namely y, R, and a velocity scale, in compliance with (a) and
(b) above.
It is possible that a k1−1 range may appear at even higher Reynolds numbers typical
of the atmospheric surface layer, and a complete analysis of atmospheric data is
required (but see Marusic & Kunkel 2004). Its often-published appearance (see, for
instance, Kader & Yaglom 1991; Katul & Chiu 1998; Nikora 1999; Hunt & Carlotti
2001; Högström et al. 2002) leads to the obvious question of why this might be
so. Townsend (1956, p. 206) appears to have initially suggested the possibility of
a k1−1 range without distinguishing complete and incomplete similarity. In fact, it
would appear that Townsend (1976, p. 154) began to recognize that self-similarity
was unlikely: ‘It now appears that simple similarity of the motion is not possible
with attached eddies and, in particular, that the stress-intensity ratio depends to some
extent on position in the layer.’ Despite this later change of view, the acceptance
of a k1−1 range has continued with what now appears to be incomplete similarity
only. As we have suggested, complete similarity requires a single velocity scale. In
addition, it also implies that contributing structures are space-filling in x: given that
the theory of § 2 does not distinguish the streamwise and azimuthal directions, space-
filling structures are required in the latter direction also. Yet evidence to the contrary
abounds, namely that long structures only appear in the x-direction, induced by the
large strain rate. Even then, it is generally accepted that quasi-streamwise vortices are
confined to the near-wall region below that in which self-similarity might be expected.
A further requirement is that the self-similar motion arises as the result of a hierarchy
of non-interacting eddies so that the spectral transfer of energy (the dissipation rate
at high Reynolds numbers) is local, as required by the local-equilibrium hypothesis.
However, such a condition is inconsistent with the contribution to production by
large-scale, supposedly inactive, motion evident in figure 1 which makes the energy
processes non-local in both physical and wavenumber space. Bradshaw (1967) has
suggested that a near-wall energy balance for the inactive motion alone may be
expressed as
advection = diffusion + dissipation, (5.2)
in which, although there is direct viscous dissipation arising from the low-wavenumber
motion, there is no production. It now seems that (5.2) should be amended to include
a production term as well, so that the energy balance is the same as that for the active
turbulence. Bradshaw (1967) and Townsend (1976) have analysed the influence of the
large scales on the log law: they both suggest that the effective value of κ increases
with Reynolds number. The foregoing would suggest that, since the outer-region
influence increases indefinitely, κ will not asymptote to a constant value. However,
as the Reynolds number increases, the influence of the larger scales in the horizontal
velocity components becomes confined to smaller distances from wall. At R + ≈ 5000
(when the log law first appears), the effects of inactive motion, most evident at
y + ≈ 15, in fact occupy O(1%) of the pipe radius. Therefore any influence of inactive
motion on κ will be small.
 Scaling of the streamwise velocity component in turbulent pipe flow 125
3.0 ReD
2.3 × 105
4.1 × 105
2.9 1.0 × 106
3.1 × 106
5.7 × 106
2.8 Fu = 2.69 + 6.73 S2u
Fu
2.7

2.6

2.5
0 0.002 0.004 0.006
S2u

Figure 27. Fu as a function of Su2 .

5.3. Departures from Gaussian behaviour

The behaviour of the higher moments in channels has been investigated extensively
by Durst & Jovanović (1995) and Durst et al. (1995). In particular, Jovanović,
Durst & Johansson (1993) have shown that truncated Gram–Charlier series
expansions may be used to approximate p.d.f.s in regions in which the second
moment reaches a maximum. Representation of higher-order correlations in terms of
lower-order correlations is clearly desirable. In view of the fact that the present data
show that at high Reynolds numbers there are two maxima (figure 1), the behaviour
of the moments and, in particular, their departures from Gaussian over a large range
of Reynolds numbers, is especially revealing.
Jovanović, Durst & Johansson (1993) show that a Gaussian distribution may only
be obtained near such a maximum if
Sj2  ZZj(1) (Fj − 3), (5.3)
(10Sj − SSj )2  ZZj(2) (15Fj − 30 − SFj ), (5.4)
where ZZj(1) and ZZj(2) are functions representing the limiting case of a Gaussian
(normal) distribution:
Sj2
ZZj(1) = lim , (5.5)
Sj →0 Fj − 3
Fj →3

(10Sj − SSj )2
ZZj(2) = lim . (5.6)
Sj →0,SSj →0 15Fj − 30 − SFj
Fj →3,SFj →15

These expressions lead to simpler ones relating higher moments to lower ones. This
is potentially a very useful result if it might be shown that p.d.f.s are invariant with
both Reynolds number and position in the shear layer. In particular, they suggest a
linear relationship between odd moments, SSu and Su , and a quadratic relationship
between the even moments, SFu , Fu and Su . The sublayer peak in u2 is not amenable to
this detailed analysis: however, the peak at yp+ certainly is. Figures 27 and 29 show Fu
and SFu respectively, both plotted against Su2 while figure 28 shows SSu plotted against
Su : only the three data points both at and immediately adjacent to yp+ are used for the
126 J. F. Morrison, B. J. McKeon, W. Jiang and A. J. Smits

0.4

0.2
ReD
SSu 0 2.3 × 105
4.1 × 105
1.0 × 106
–0.2 3.1 × 106
5.7 × 106
SSu = 0.0954 + 7.22 Su
–0.4

–0.10 –0.05 0 0.05 0.10

Su

Figure 28. SSu as a function of Su .

12.0

11.5

ReD
SFu 11.0 2.3 × 105
4.1 × 105
1.0 × 106
3.1 × 106
10.5 5.7 × 106
SFu = 11.1 + 7.61 S2u

10
0 0.0025 0.0050 0.0075
S2u

Figure 29. SFu as a function of Su2 .

Reynolds numbers at which the peak is clearly evident, 2.3 × 105 6 ReD 6 5.7 × 106 .
Three points per Reynolds number are used to improve the statistical value of the
curve fit. The relationship between the odd moments appears to be close to linear
while that between the even moments and Su2 is not: figure 30 suggests that the
relationship between SFu and Fu is the most linear of all. In summary, it appears that
the relationships between the higher moments are too complicated to be represented
by these simple expressions. In addition, the constants are such that SFu , SSu , and Fu
have significantly non-Gaussian values as Su → 0. The p.d.f.s of figures 7 and 8 and
their moments also show that there are significant variations with both position and
Reynolds number.
+
5.4. Nature of the peak in u2 at yp+
The foregoing, and in particular the result of figure 1 which shows two maxima,
confirms the principal ideas of so-called inactive motion. The maximum at y + ≈ 15 is
Reynolds-number dependent while its position is only weakly so. Further discussion
of this near-wall peak is not possible owing to the poor resolution at higher Reynolds
numbers. This is not the case for the outer peak at yp+ . Using the momentum equation
 Scaling of the streamwise velocity component in turbulent pipe flow 127
12.0

11.5

ReD
SFu 11.0 2.3 × 105
4.1 × 105
1.0 × 106
3.1 × 106
10.5 5.7 × 106
SFu = –20.6 + 11.8 Fu

10
2.6 2.65 2.70 2.75 2.80
Fu

Figure 30. SFu as a function of Fu .

1000

800

600
y+p
400

200

0 5 10 (×104)
R+

Figure 31. Variation of yp+ with R + :– – – – –, 1.8(R + )0.52 .

and assuming the validity of the log law at quite small values of y + , the locus of
the peak shear stress is given by κ −0.5 (R + )0.5 . Using the values of κ given either by
Zagarola & Smits (1998) or by McKeon et al. (2004) gives 1.5(R + )0.5 while Sreenivasan
(1988), using data from a variety of sources, suggests that the locus is given by
2.0(R + )0.5 . Figure 31 shows the locus of the maximum in u2 , yp+ = 1.8(R + )0.52 .
Therefore, the mechanism for production of energy at y + ≈ 15 and of shear stress close
to yp+ are not the same and these differences may be traced back to the different ways
in which the wall affects the wall-parallel and wall-normal velocity components. While
the outer peak in u2 closely follows that of uv, a change in constant would suggest
that the uv-bearing scales are not necessarily the same as those bearing the turbulence
kinetic energy (dominated by u2 ). Such an interpretation is again consistent with the
distinction between active and inactive motion. At ReD = 5.7 × 106 , yp+ = 810 where
the wavenumber limit to the spatial resolution is given by 2πy/ l = 12. Reference
to figure 3 suggests that this estimate of yp+ is only slightly affected by the limit of
spatial resolution.
By analogy with a transitional critical layer, Sreenivasan (1988) has suggested that
a ‘critical layer’ in wall turbulent flow might be a useful model with a vortex sheet
128 J. F. Morrison, B. J. McKeon, W. Jiang and A. J. Smits
located at the maximum in Reynolds shear stress. The occurrence of a viscous critical
layer where the Reynolds stresses reach a maximum at singularities in the frictionless
stability equation leads to an obvious analogue of the peak in Reynolds shear stress
in wall turbulent flow at very high Reynolds number. This has implications both for
furthering our understanding of inner/outer interactions of wall turbulence as well
as for their control.
It is not until the highest Reynolds number that yp+ reaches the lower limit to the
range in which the mean velocity approaches complete similarity (Zagarola & Smits
1998; McKeon et al. 2004). Using the power law given by Zagarola & Smits (1998),
the mean velocity at yp+ is given by
Up+ = 9.43(R + )0.071 , (5.7)
and is more or less a constant fraction of Ucl+ for the range of data given in figure 31,
with 0.63 6 Up /Ucl 6 0.65 only (Sreenivasan 1988 finds 0.65). Such a result would
be expected for a transitional critical layer. Smith & Bodonyi (1982) have suggested
that a lengthscale or thickness of a nonlinear critical layer as a fraction of the pipe
diameter is proportional to (R + )−1/6 while the classical result (e.g. Stuart 1971) leads
to a thickness proportional to (R + )−1/3 . Unfortunately, while the width of the peak at
yp+ does appear to decrease as the Reynolds number increases, the data of figure 1
cannot be taken as conclusive in this regard. A characteristic velocity fluctuation for
the transitional critical layer is
U0 where U0 is a characteristic speed, which for our
analogy to the turbulent layer is uτ .
is the (small) amplitude of disturbance. While
the details depend on the nature of the analysis of the transitional critical layer, we
take
∝ (R + )−1/6 (Smith & Bodonyi 1982), in which case the turbulent ‘critical layer’
has a velocity scale proportional to uτ (R + )−1/6 . However, it is clear from figure 1 that
the peak at yp+ increases with Reynolds number. Further work is obviously required.

6. Conclusions
We have examined Reynolds-number effects on the second and higher moments
of the streamwise velocity component using Townsend’s distinction between active
and inactive motion. Such an analysis enables the direct effects of viscosity on the
near-wall motion to be distinguished from those of the outer-region large eddies.
+
Near-wall maxima in u2 are Reynolds-number dependent; u2 does not scale with
inner variables and this has to be interpreted as the influence of the outer-layer
eddies. The peak at y + ≈ 15 coincides with that of the production of turbulence
kinetic energy and therefore the description of this influence of outer-layer motion
as ‘inactive’ is inappropriate. In addition to so-called inactive motion increasing with
Reynolds number, we have shown that it also increases with decreasing wall distance,
down to the viscous sublayer, in accord with Townsend’s later observations. Statistics
show no simple scalings with either inner or outer variables in the same range of y:
thus there is no overlap region, a result that is consistent with the above interpretation.
Moreover, they also show significant departures from Gaussian over the whole layer,
these becoming large both near the wall and the centre line. A comparison of the
present outer-region statistics and those of den Toonder & Nieuwstadt (1997) scaled
with both uτ and Ucl − U shows that the latter is a better velocity scale, as long as
the ranges ReD < 6 × 104 and ReD > 7 × 104 are distinguished. Noting that, even just
for the present data, such a conclusion involves at least three sets of independent
measurements, it would appear to be a valid one not explainable by anomalies in
the data. The change in scaling behaviour appearing in, say, figure 24, occurs at
 Scaling of the streamwise velocity component in turbulent pipe flow 129
+
ReD ≈ 6 × 104 and may be traced to the change in Ucl+ − U (which occurs in the
range 6 × 104 < ReD < 2 × 105 , see McKeon et al. 2004) and the associated changes in
+
the scaling of the first moment. Note also that the second peak in u2 appears only
for ReD > 2 × 105 . Does this mean that there might be associated structural changes
of the form already referred to (perhaps related to hairpin vortex packets described,
for example, by Christensen & Adrian 2001) that might be attributable to a secondary
instability? Given the quite low Reynolds numbers involved, such a question might
soon be answerable by direct simulation as well as experiments.

The support of ONR under Grant Nos. N00014-98-1-0525 and N00014-99-1-0340 is

gratefully acknowledged. J. F. M. is indebted to the Engineering and Physical Sciences
Research Council (grants GR/M64536/01 and GR/R48193/01), the Royal Academy
of Engineering (England), and the Leverhulme Trust (grant F/07058/H) for financial
support.

REFERENCES
Antonia, R. A. & Kim, J. 1994 Low-Reynolds-number effects on near-wall turbulence. J. Fluid
Mech. 276, 61–80.
Blackwelder, R. F. & Eckelmann, H. 1979 The spanwise structure of the bursting phenomenon.
J. Fluid Mech. 94, 577–594.
Bradshaw, P. 1967 ‘Inactive’ motion and pressure fluctuations in turbulent boundary layers.
J. Fluid Mech. 30, 241–258.
Champagne, F. H., Sleicher, C. A. & Wehrmann, O. H. 1967 Turbulence measurements with
inclined hot wires part 1. heat transfer experiments with inclined hot wire. J. Fluid Mech. 28,
153–175.
Christensen, K. T. & Adrian, R. J. 2001 Statistical evidence of hairpin vortex packets in wall
turbulence. J. Fluid Mech. 431, 433–443.
Corrsin, S. 1963 Turbulence: Experimental methods. In Handbuch der Physik , vol. 8, pp. 524–590.
Springer.
Dean, R. B. & Bradshaw, P. 1976 Measurements of interacting turbulent shear layers in a duct.
J. Fluid Mech. 78, 641–676.
DeGraaff, D. B. & Eaton, J. K. 2000 Reynolds-number scaling of the flat-plate turbulent boundary
layer. J. Fluid Mech. 422, 319–346.
Dinavahi, S. P. G., Breuer, K. S. & Sirovich, L. 1995 Universality of probability density functions
in turbulent channel flow. Phys. Fluids 7, 1122–1129.
Durst, F. & Jovanović, J. 1995 Investigations of Reynolds-averaged turbulence quantities. Proc. R.
Soc. Lond. A 451, 105–120.
Durst, F., Jovanović, J. & Sender, J. 1995 LDA measurements in the near-wall region of a
turbulent pipe flow. J. Fluid Mech. 295, 305–335.
Eggels, J. G. M., Unger, F., Weiss, M. H., Westerweel, J., Adrian, R. J., Friedrich, R. &
Nieuwstadt, F. T. M. 1994 Fully developed turbulent pipe flow: a comparison between direct
numerical simulation and experiment. J. Fluid Mech. 268, 175–209.
Fernholz, H. H. & Finley, P. J. 1996 The incompressible zero-pressure-gradient turbulent boundary
layer: an assessment of the data. Prog. Aerospace Sci. 32, 245–311.
Fontaine, A. A. & Deutsch, S. 1995 Three component, time-resolved velocity statistics in the wall
region of a turbulent pipe flow. Exps. Fluids 18, 168–173.
Freymuth, P. 1979 Engineering estimation of heat conduction loss in constant temperature thermal
sensors. TSI Q. 5, 3–8.
Hinze, J. O. 1975 Turbulence, 2nd edn. McGraw-Hill.
Högström, U., Hunt, J. C. R. & Smedman, A.-S. 2002 Theory and measurements for turbulence
spectra and variances in the atmospheric neutral surface layer. Boundary-Layer Met. 103,
101–124.
130 J. F. Morrison, B. J. McKeon, W. Jiang and A. J. Smits
Hunt, J. C. R. & Carlotti, P. 2001 Statistical structure at the wall of the high Reynolds number
turbulent boundary layer. Flow, Turbulence Combust. 66, 453–475.
Hunt, J. C. R. & Morrison, J. F. 2000 Eddy structure in turbulent boundary layers. Eur. J. Mech.
B/Fluids 19, 673–694.
Jiménez, J. 1998 The largest scales of turbulent wall flows. In Annual Research Briefs, pp. 137–154.
Center for Turbulence Research.
Jiménez, J., del Alámo, J. C. & Flores, O. 2004 The large-scale dynamics of near-wall turbulence.
J. Fluid Mech. 505, 179–199.
Jiménez, J., Flores, O. & Garcı́a-Villalba, M. 2002 Organization of autonomous wall turbulence.
In Advances in Turbulence IX (ed. I. P. Castro, P. E. Hancock & T. G. Thomas), pp. 824–828.
Barcelona: CIMNE.
Jiménez, J. & Pinelli, A. 1999 The autonomous cycle of near-wall turbulence. J. Fluid Mech. 389,
335–359.
Jones, M. B., Marusic, I. & Perry, A. E. 2001 Evolution and structure of sink-flow turbulent
boundary layers. J. Fluid Mech. 428, 1–27.
Jovanović, J., Durst, F. & Johansson, T. G. 1993 Statistical analysis of the dynamic equations for
higher-order moments in turbulent wall-bounded flows. Phys. Fluids A 5, 2886–2900.
Kader, B. A. & Yaglom, A. M. 1991 Spectra and correlation functions of surface layer atmospheric
turbulence in unstable thermal stratification. In Turbulence and Coherent Structures (ed. O.
Métais & M. Lesieur). Kluwer.
Katul, G. & Chiu, C.-R. 1998 A theoretical and experimental investigation of energy-
containing scales in the dynamic sublayer of boundary-layer flows. Boundary-Layer Met. 86,
279–312.
Kim, K. C. & Adrian, R. J. 1999 Very large-scale motion in the outer layer. Phys Fluids. 11,
417–422.
Klewicki, J. C. & Falco, R. E. 1990 On accurately measuring statistics associated with small-scale
structure in turbulent boundary layers using hot-wire probes. J. Fluid Mech. 219, 119–142.
Laufer, J. 1954 The structure of turbulence in fully developed pipe flow. NACA Tech. Rep. 1174.
Li, J. D., McKeon, B. J., Jiang, W., Morrison, J. F. & Smits, A. J. 2004 The response of hot wires
in high Reynolds-number turbulent pipe flow. Meas. Sci. Technol. 15, 1–10.
Ligrani, P. M. & Bradshaw, P. 1987 Spatial resolution and measurement of turbulence in the
viscous sublayer using subminiature hot-wire probes. Exps. Fluids 5, 407–417.
Long, R. R. & Chen, T.-C. 1981 Experimental evidence for the existence of the ‘mesolayer’ in
turbulent systems. J. Fluid Mech. 105, 19–59.
Lumley, J. L. 1970 Stochastic Tools in Turbulence. Academic.
Marusic, I. & Kunkel, G. J. 2003 Streamwise turbulence intensity formulation for flat-plate
boundary layers. Phys Fluids. 15, 2461–2464.
Marusic, I. & Kunkel, G. J. 2004 Turbulence intensity similarity laws for turbulent boundary
layers. In IUTAM Symposium on Reynolds Number Scaling in Turbulent Flow (ed. A. J. Smits),
pp. 17–22. Kluwer.
Marusic, I. & Perry, A. E. 1995 A wall-wake model for the turbulence structure of boundary
layers. Part 2. Further experimental support. J. Fluid Mech. 298, 389–407.
Marusic, I., Uddin, A. K. M. & Perry, A. E. 1997 Similarity law for the streamwise turbulence
intensity in zero- pressure-gradient turbulent boundary layers. Phys. Fluids 9, 3718–3726.
McKeon, B. J., Li, J., Jiang, W., Morrison, J. F. & Smits, A. J. 2004 Further observations on the
mean velocity in fully-developed pipe flow. J. Fluid Mech. 501, 135–147.
Metzger, M. M. & Klewicki, J. C. 2001 A comparative study of near-wall turbulence in high and
low Reynolds number boundary layers. Phys. Fluids 13, 692–701.
Metzger, M. M., Klewicki, J. C., Bradshaw, K. L. & Sadr, R. 2001 Scaling the near-wall axial
turbulent stress in the zero pressure gradient boundary layer. Phys. Fluids 13, 1819–1821.
Millikan, C. M. 1938 A critical discussion of turbulent flows in channels and circular tube. Proc.
5th Intl Congr. Appl. Mech., pp. 386–392. Wiley.
Mochizuki, S. & Nieuwstadt, F. T. M. 1996 Reynolds-number-dependence of the maximum in
the streamwise velocity fluctuations in wall turbulence. Exps. Fluids 21, 218–226.
Morris, S. C. & Foss, J. F. 2003 Transient thermal response of a hot-wire anemometer. Meas. Sci.
Technol. 14, 251–259.
 Scaling of the streamwise velocity component in turbulent pipe flow 131
Morrison, J. F., Jiang, W., McKeon, B. J. & Smits, A. J. 2001 Reynolds-number dependence
of streamwise velocity fluctuations in turbulent pipe flow. In Turbulence and Shear Flow
Phenomena, TSFP-2 , pp. 43–48. Stockholm: Universitetsservice US AB.
Morrison, J. F., Jiang, W., McKeon, B. J. & Smits, A. J. 2002a Reynolds-number dependence of
streamwise velocity fluctuations in turbulent pipe flow. AIAA Paper 2002-0574.
Morrison, J. F., Jiang, W., McKeon, B. J. & Smits, A. J. 2002b Reynolds-number dependence of
streamwise velocity spectra in turbulent pipe flow. Phys. Rev. Lett. 88, 214501.
Morrison, J. F., Subramanian, C. S. & Bradshaw, P. 1992 Bursts and the law of the wall in
turbulent boundary layers. J. Fluid Mech. 241, 75–108.
Nieuwstadt, F. T. M. & Bradshaw, P. 1997 Similarities and differences of turbulent boundary-
layer, pipe and channel flow. In Boundary-Layer Separation in Aircraft Aerodynamics (ed.
R. A. W. M. Henkes & P. G. Bakker). Delft University Press.
Nikora, V. 1999 Origin of the ‘−1’ spectral law in wall-bounded turbulence. Phys. Rev. Lett. 83,
734–736.
Perry, A. E. & Abell, C. J. 1975 Scaling laws for pipe-flow turbulence. J. Fluid Mech. 67, 257–271.
Perry, A. E. & Abell, C. J. 1977 Asymptotic similarity of turbulence structures in smooth- and
rough-walled pipes. J. Fluid Mech. 79, 785–799.
Perry, A. E., Henbest, S. & Chong, M. S. 1986 A theoretical and experimental study of wall
turbulence. J. Fluid Mech. 165, 163–199.
Perry, A. E. & Li, J. D. 1990 Experimental support for the attached-eddy hypothesis in zero-
pressure-gradient turbulent boundary layers. J. Fluid Mech. 218, 405–438.
Perry, A. E. & Marusic, I. 1995 A wall-wake model for the turbulence structure of boundary
layers. Part 1. Extension of the attached eddy hypothesis. J. Fluid Mech. 298, 361–388.
Rotta, J. C. 1962 Turbulent boundary layers in incompressible flow. Prog. Aeronaut. Sci. 2, 1–220.
Sandborn, V. A. 1955 Experimental evaluation of momentum terms in turbulent pipe flow. NACA
Tech. Note 3266.
Smith, F. T. & Bodonyi, R. J. 1982 Amplitude-dependent neutral modes in the Hagen-Poiseuille
flow through a circular pipe. Proc. R. Soc. Lond. A 384, 463–489.
Sreenivasan, K. R. 1988 A unified view on the origin and morphology of the turbulent boundary
layer structure. In Turbulence Management and Relaminarization (ed. H. W. Liepmann &
R. Narasimha), pp. 37–62. Springer.
Stuart, J. T. 1971 Non-linear stability theory. Annu. Rev. Fluid Mech. 3, 347–370.
Tchen, C. M. 1953 On the spectrum of energy in turbulent shear flow. J. Res. Natl Bureau Stand.
50, 51–62.
den Toonder, J. M. J. & Nieuwstadt, F. T. M. 1997 Reynolds number effects in a turbulent pipe
flow for low to moderate Re. Phys. Fluids 9, 3398–3409.
Townsend, A. A. 1956 The Structure of Turbulent Shear Flow , 1st edn. Cambridge University Press.
Townsend, A. A. 1961 Equilibrium layers and wall turbulence. J. Fluid Mech. 11, 97–120.
Townsend, A. A. 1976 The Structure of Turbulent Shear Flow , 2nd edn. Cambridge University
Press.
Tsuji, Y. & Nakamura, I. 1999 Probability density function in the log-law region of low reynolds
number turbulent boundary layer. Phys. Fluids 11, 647–658.
Wei, T. & Willmarth, W. W. 1989 Reynolds number effects on the structure of a turbulent channel
flow. J. Fluid Mech. 204, 57–95.
Westbury, P. S. & Morrison, J. F. 1995 Bursts and low-Reynolds-number effects in turbulent
boundary layers. In Advances in Turbulence V (ed. R. Benzi), pp. 549–553. Kluwer.
Wosnik, M., Castillo, W. L. & George, K. 2000 A theory for turbulent pipe and channel flows.
J. Fluid Mech. 421, 115–145.
Zagarola, M. V., Perry, A. E. & Smits, A. J. 1997 Log laws or power laws: The scaling in the
overlap region. Phys. Fluids 9, 2094–2100.
Zagarola, M. V. & Smits, A. J. 1998 Mean-flow scaling of turbulent pipe flow. J. Fluid Mech. 373,
33–79.