Boundary Layer Vorticity and the Rise of

“Hairpins”

Peter S. Bernard

Abstract The downstream evolution of the vorticity field in the vicinity of hairpin-
shaped regions of rotational motion appearing in the transitioning boundary layer
is examined. It is shown that the dynamics of hairpins is inseparable from that of
the non-rotational vorticity out of which they develop in a self-reinforcing process
of ejection and reorientation. Widening the concept of structure to include the com-
plete localized vorticity that produces hairpins, allows for a more complete and self-
contained explanation of the boundary layer physics.

1 Introduction

By their nature, boundary layers in high speed flow contain prodigious amounts of
vorticity produced at the solid bounding surface by the action of viscosity. A vari-
ety of measurement techniques applied to either physical experiments or numerical
computations of boundary layers suggest the presence of “coherent” objects within
the vorticity field that make essential contributions to the dynamical behavior of the
flow. The possibility of arriving at a precise understanding of the boundary layer
structure depends on what is meant by “coherency,” a concept that is intrinsically
difficult to define. In recent times this has come to mean distinctive “regions of ro-
tational motion” and the coherent objects discovered by this criterion are generally
hairpin-shaped [1], meaning a flow volume with one or two streamwise oriented
“legs” attached to a spanwise “arch.”
Since much of the vorticity in the boundary layer does not lie in regions where
rotational flow is occurring, the assertion that rotational regions are structures begs
the question as to what role the remaining, non-rotational, vorticity has both in the
dynamics of the boundary layer in general, and the rotational regions in particu-

Peter S. Bernard
Department of Mechanical Engineering, University of Maryland, College Park, MD 20742, USA,
e-mail: bernard@[Link]

1
2 Peter S. Bernard

lar. The importance of this vorticity has been recognized previously, for example
in [5] where isosurfaces of spanwise vorticity reveal structural details that conform
to the rotational field in the form of hairpins. In addition, several recent studies
[2, 3, 4] using a vortex filament scheme visualized vortex structure in the bound-
ary layer without the a priori decision that the structures must occupy regions of
rotational motion. This showed that hairpin-like regions of rotation have a natural
association with uplifting furrow-like folds in the surface vorticity layer that develop
downstream into mushroom-like shapes prior to descending into chaotic forms. It is
evident from results such as these that there is much to be gained by examining the
vorticity environment surrounding individual hairpins with the goal of exposing the
role played by all of the vorticity that acts coherently to create the hairpin and that
is essential to the physics of the boundary layer. That is the focus of this study.

2 Vortex Filament Scheme

A vortex filament simulation of a spatially growing boundary provides the numerical

data for this study. Past work [2, 4] has described the numerical algorithm in detail.
It suffices for the present to mention that this is a hybrid approach in the sense that
vorticity determined from a finite volume solution to the viscous flow equations on
a thin wall mesh is converted to vortex filaments that represent the flow outside
the near-wall region. The mesh calculation is both more efficient and accurate at
resolving the largely 2D regions of intense vorticity diffusing out of the wall surface
than can be achieved using filaments. Within the filament field, hairpin removal
provides spatial and temporal intermittent dissipation at inertial range scales as well
as limiting growth in the number of vortices. The velocity is calculated using the
Biot-Savart law that takes into account the contributions from the vorticity in the
mesh and filaments. A velocity potential on the surface triangularization is used to
enforce the non-penetration boundary condition.
The simulation considered here occupies a larger spatial region and higher
Reynolds number than in previous work [4]. The boundary layer is computed on
the top and bottom surfaces of a flat plate with rounded edges of dimensionless
length 4 in the streamwise direction, 2.5 in the spanwise direction and 0.05 thick.
The region from the front of the plate until x = 1 is kept as an inviscid surface so
that the boundary layer starts at x = 1. Thin regions on the sides and rear of the plate
are also taken to be inviscid. The advantage of setting up the flow this way is that
it successfully stabilizes the boundary layer against large scale spanwise and wall-
normal flow asymmetries that may appear in the unfettered viscous motion past a
finite plate.
The Reynolds number at the end of the boundary layer it is Re = 225, 000. Just
after the start of the viscous computation at x = 1 the boundary layer is smooth
and is in excellent agreement with the Blasius boundary layer as shown in Fig. 1
comparing the computed velocity field with the Blasius result. Filaments produced
in this region are exactly aligned in the spanwise direction. Transition is induced
Boundary Layer Vorticity and the Rise of “Hairpins” 3

by the response of the many filaments and vortex elements in the mesh to slight
perturbations that are inherent in the discretization of the Biot-Savart law. Though
the perturbations are very small initially, and originate entirely within the near-wall
mesh, once provoked they grow quickly due to the mutual interactions between
vortices. The end result is transition to turbulence, though at Reynolds numbers
that are somewhat more typical of a heavily forced flow. For the present simulation
this means that the fully turbulent state is achieved at Re ≈ 100, 000. With finer
discretization this trend can be reversed, though at significant increase in the cost of
the simulation and without qualitative change in the observed vortex structure.
To obtain information about the vorticity field for this study, the velocity was
computed on a fine mesh covering the flow domain and then substituted into second-
order accurate finite-difference formulas to get the derivatives needed to compute
vorticity. The same data set was used to compute λ2 whose isosurfaces mark the
presence of rotational regions in the flow field including the hairpins.

3 Vorticity and Structure

The filament calculation produces a Klebanoff type transition bridging the gap be-
tween the Blasius boundary layer and a fully turbulent flow. This transition mode is
marked by the presence of low speed streaks and rotational structures as revealed
by λ2 in the form of hairpins. Figure 2 gives an overhead view of the computed
λ2 = −30 isosurfaces from x = 1.4 until the onset of the fully developed turbulent
field. Apart from some noise at the upstream end caused by the locally low ampli-
tude of λ2 as well as the coarse discretization used in its computation, the visual-
ization of λ2 reveals Λ -like vortices preceding the formation of rotational regions
in the shape of hairpins spaced approximately ∆ z+ = 350 that are not unlike those
seen in more traditional grid-based simulations.
To make sense of the vorticity field associated with the appearance of hairpins
it is useful to consider the streamwise behavior of the maximum amplitudes of the
vorticity components at fixed distances above the wall. Such data is shown in Fig. 3
at a point inside (y+ = 24.3) and a point outside (y+ = 51.3) the near-wall viscous
region. In these plots the maxima are restricted to the local spanwise region of a par-
ticular Klebanoff streak, in this case the one located between 0.15 ≤ z ≤ 0.25. Other

Fig. 1 Velocity in the Bla-

sius region at x = 1.3. –,
similarity solution; ◦, com-
puted.p The similarity variable
η = y Re /(x − 0.95) in-
cludes a slight adjustment to
the virtual origin.
4 Peter S. Bernard

streaks give qualitatively the same result with some small shifting in the streamwise
direction. For each figure, the appropriate Blasius spanwise vorticity ω3 is plotted as
a dashed line that shows how the boundary layer would behave in the same location
if it had remained laminar. In all cases the turbulent solution departs smoothly from
the Blasius values as the flow transitions.
Vertical lines in Fig. 3 indicate three relatively distinct zones in the evolution of
the vorticity as it impacts the hairpins. The first region, between x = 1.4 and 1.7,
is characterized by a steady increase in the streamwise (ω1 ) and wall-normal (ω2 )
vorticity components near the wall. This is significant if for no other reason than the
fact that such non-spanwise vorticity is absent in the Blaisius boundary layer. Down-
stream of the first zone there commences a number of significant changes to the vor-
ticity amplitudes including a dramatic drop in the magnitude of ω3 that effectively

Fig. 2 Overview of λ2 = −30 isosurfaces.

Fig. 3 Streamwise depen-

dence of the maximum vortic-
ity amplitude 0.15 ≤ z ≤ 0.25.
(a) at y+ = 24.3; (b),
y+ = 51.3. ω1 , (· · ·, blue);
ω2 , (− · −, cyan); ω3 , (—
, red). Dashed line is the
equivalent ω3 for the Blasius
boundary layer.
Boundary Layer Vorticity and the Rise of “Hairpins” 5

ends any remaining connection it has to the laminar form. The region 1.7 ≤ x ≤ 1.9
is singled out as being distinctive because of the structural transformation accom-
panying the vorticity behavior in Fig. 3. The end result of the developments in the
second zone is to enable the appearance of the fully formed hairpin-like rotational
regions located downstream of x > 1.9 that characterize transition until the break-
down to fully turbulent flow.
To give more context to the ensuing discussion, the trends in the vorticity magni-
tude both closer and further from the wall than is considered in Fig. 3 are displayed
in Fig. 4. Fig. 4(a) shows that after transition there is a very large increase in the
spanwise vorticity at y+ = 8.1 to values much higher than in the equivalent lami-
nar boundary layer. This state persists indefinitely downstream and underlies all the
activity that produces turbulent structure. In particular, this reservoir of very high
vorticity is undoubtedly the source of all “new” vorticity that enters into the cre-
ation of downstream structures after transition. The vorticity trends in Fig. 4(b) at
y+ = 89.1 − a region well beyond the direct reach of the viscous boundary layer −
shows the last stages in the creation of structure during transition.
Zone 1. The onset of the streamwise structure in Fig. 2 is in the same location
as the steady growth in streamwise vorticity near the wall in Fig. 3(a). Upstream
of x = 1.4 the magnitude of ω1 is a very slight fraction of the ambient spanwise
vorticity, but after x = 1.4 it grows significantly becoming more than one third of the
magnitude of ω3 by x = 1.7. There is also a somewhat smaller, but non-negligible,
rise in wall-normal vorticity accompanying that of ω1 . The location where these
vorticity components grow is very close to the boundary, entirely within the viscous
region of large spanwise vorticity out of which ω1 and ω2 develop by reorientation.
The physical boundary no doubt has an influence on suppressing ω2 in this region.
An idea of the structural form taken by ω1 as it appears in the flow is given
in Fig. 5(a) showing its isosurfaces at four locations in zone 1. The streamwise
vorticity has formed into oppositely signed concentrated pairs consistent with the
λ2 isosurfaces in Fig. 2, and is strengthening with downstream distance. Overlying
the ω1 isosurfaces, though not shown, is the smaller wall-normal vorticity.

Fig. 4 Streamwise depen-

dence of the maximum vortic-
ity amplitude 0.15 ≤ z ≤ 0.25.
(a) at y+ ≈ 8.1; (b), y+ = 89.1.
Definitions of curves are the
same as in Fig. 3.
6 Peter S. Bernard

It can be anticipated that there is some loss of ω3 as ω1 and ω2 develop from its
reorientation. On the other hand, the counter-rotating velocity created by ω1 causes
significant convection of ω3 away from the wall. This effect is apparent in Fig. 5(b)
giving a visualization of the ω3 contours at locations in zone 1. Spanwise vorticity
decreases near the wall accompanied by its simultaneous rise away from the wall.
As the spanwise vorticity is propelled upwards, its place is filled by low speed fluid
forming a streak.
The trends in zone 1 of Fig. 3 are consistent with the outward ejection process.
For example, in the region just beyond x = 1.55 the loss of spanwise vorticity shows
up at y+ = 24.3 while at y+ = 51.3 there is a gain. Note, as well, that there is little
non-spanwise vorticity at y+ = 51.3 suggesting that the reorientation process that
produces such vorticity has yet to rise up to this level above the wall.
Zone 2. Now consider the vorticity trends in Fig. 3 for 1.7 ≤ x ≤ 1.9 that is
taken to be the second development zone of the structures. A prominent feature
is a precipitous drop − to one fifth of its maximum value − of the spanwise vor-
ticity at y+ = 24.3. A significant decrease in ω3 also occurs at y+ = 51.3, though
delayed until x = 1.8. Near the wall the streamwise vorticity increases and levels
off, while there is a sudden and substantial growth in the wall normal vorticity at
y+ = 51.3 until it exceeds the local amplitude of ω3 . The picture that emerges from
these trends is that near the wall the reorientation process by which streamwise vor-
ticity develops from ω3 has run its course. At the same time, the prodigious amount
of ejected spanwise vorticity, now outside the near-wall domination of viscous dif-
fusion, rapidly reorients to establish a significant presence of wall-normal vorticity
that reaches a good distance from the wall. The latter process is also visible in Fig.
4(b) at points even further downstream where a sudden and significant growth in
spanwise vorticity is followed immediately by the production of wall-normal and
streamwise vorticity. With its rapid conversion to ω1 and ω2 the local dominance of
spanwise vorticity ends at these distances from the wall.

Fig. 5 Isosurfaces of vorticity at several locations in zone 1. (a), green, yellow, red correspond
to ω1 = −10, −20, −30, blue, cyan, magenta are ω1 = 10, 20, 30 (increasing inwards); (b), green,
blue, cyan, red correspond to ω3 = −25, −50, −75, −100 (increasing towards wall).
Boundary Layer Vorticity and the Rise of “Hairpins” 7

Some of the trends in zone 2 are illustrated in the isosurfaces of ω2 and ω3 shown
in Fig. 6. ω2 grows to prominence as an oppositely signed pair with a concentration
at its upper end coinciding with the λ2 isosurfaces at the same location. Though
not shown, there is a growing presence of ω1 in this region as well. Figure 6(b)
shows that the continued ejection of spanwise vorticity results in its collecting in
the region between the counter-rotating motion. Near the wall, ω3 continues to fall
in magnitude even though it strengthens in the fluid sublayer closest to the boundary
as seen in Fig. 4(a).
A summary of the state of affairs at the end of zone 2 is given in Fig. 7 that
shows, in one plot, isosurfaces of the three vorticity components. Prominent features
include the strong streamwise vorticity near the wall that forms a counter-rotating
pair, the wall-normal vorticity that has developed outside of the near-wall flow, and
finally, the significant spanwise vorticity that collects within the region between the
emerging “legs” of the hairpin.
Zone 3. Coincident with x ≈ 1.9 in Fig. 2 rotational regions with the character
of hairpin “legs” emerge out of the λ2 isosurfaces. Further downstream arch-like
structures can be found that cross between the streamwise structures. With increas-
ing x the isosurfaces reveal a breakdown of the rotational regions into a wide range
of more complicated rotational forms. A notable aspect of the vorticity maxima in
this region, seen in Figs. 3(b) and 4(b), is the development of relatively large sus-
tained peaks in the streamwise vorticity that now dominate the other components.
Evidently, this is the vorticity that is responsible for the presence of hairpin “legs,”
and it appears as the final phase of the conversion of the streamwise vorticity that
has been ejecting outwards from the wall.

Fig. 6 Isosurfaces of vorticity at several locations in zone 2. (a), green, yellow, red correspond to
ω2 = −10, −20, −30, blue, cyan, magenta are ω2 = 10, 20, 30 (toward the center); (b), green, blue,
cyan, red correspond to ω3 = −20, −45, −70, −95 (toward the wall).
8 Peter S. Bernard

Previous work that concentrated on examining the vortex filament field in tran-
sition [4] showed a direct connection between hairpin “legs” and the lobes of
mushroom-shaped structures in the vortex filaments that emerged out of the sur-
face vorticity layer. Such patterns are present among the developing hairpin regions
in the current simulation as well, and an example of this is shown in Fig. 8(b). Ac-
companying this image, in Fig. 8(a), is a view of the isosurfaces of the three vorticity
components at the same location. The streamwise vorticity occupies the “legs” that
are coincident with the lobes of the filament structure, while the wall-normal vor-
ticity extends upwards through the boundary layer encompassing the stem of the
mushroom. Finally, some streamwise vorticity persists at the top of the structure.
An extended streamwise view of the vorticity isosurfaces for this structure is
given in Fig. 9 that may be taken as a summary statement of what is entailed in a
complete view of the vorticity field associated with hairpins. Spanwise and wall-
normal vorticity accompany the prominent counter-rotating streamwise vorticity as
a legacy of the mechanisms by which the hairpin-like region developed. It should
be emphasized that because the mushroom-like form is unstable it often falls to one
side or the other producing single-legged hairpins [4]. For this common occurrence
elements of the vorticity arrangement in Fig. 9 are present even if they do not fully
resemble the somewhat idealized case depicted here.
Locations above the hairpin “legs” where the spanwise vorticity is concentrated
are where the λ2 signal is likely to indicate the presence of an arch vortex. This
agrees with the observation in previous work [3, 4] that the filament field forms
concentrated spanwise structure at the locations of arches.

Fig. 7 Isosurfaces of vorticity at the end of zone 2: ω1 = ±30, blue and green; ω2 = ±20, yellow
and cyan; ω3 = -40, red and magenta.
Boundary Layer Vorticity and the Rise of “Hairpins” 9

4 Summary and Conclusions

Some essential aspects of the development of the vorticity field in a transitional

boundary layer have been examined as it relates to the presence of hairpin-like struc-
tures. It is seen that by considering the complete local vorticity field surrounding
hairpins, and not focusing exclusively on the rotational motion by which they are
defined, the dynamics of the hairpins as well as that of the boundary layer itself can
be more clearly understood.
It is found that counter-rotating regions of streamwise vorticity develop in the
viscous sub-layer via a self-reinforcing process in which they are strengthened by
the spanwise vorticity that they cause to eject. Thus they fuel their own develop-
ment. As spanwise vorticity extends beyond the viscous sublayer it shears to create
wall-normal vorticity that concentrates away from the boundary. Out of this vortic-
ity, streamwise vorticity appears and accumulates to form the hairpins. The ejection
process near the wall proceeds until the available spanwise vorticity is depleted.
Remnants of the ejected spanwise vorticity, perhaps aided by a roll-up process, per-
sists in the outward flowing regions to produce isosurfaces of rotation that give the
impression of arch-type vortices connecting the “legs” of the structures.
A main conclusion is that while it is tempting to view the streamwise isosurfaces
of λ2 as forming hairpin “structure,” without also taking into account the active role
of the surrounding non-rotational vorticity, the analysis of hairpins misses essential
clues to the physics of the boundary layer. For example, the initial appearance of

Fig. 8 Comparison of vorticity isosurfaces in a hairpin with the local vortex filament field. (a),
isosurfaces: ω1 = ±30, blue and green; ω2 = ±15, yellow and cyan; ω3 = -15, red; (b), vortex
filaments intersecting 2.18 ≤ x ≤ 2.2, red and blue filaments are close to the streamwise direction.
10 Peter S. Bernard

streamwise rotation near the wall is only as a perturbation upon the dominant and
“invisible” spanwise vorticity field. Downstream, largely missed by λ2 isosurfaces,
is ejecting spanwise vorticity that provides the “fuel” out of which the hairpins ap-
pear. Shearing of non-rotational wall-normal and spanwise vorticity explains the
emergence of hairpin-shaped rotational regions that constitute just one aspect of a
much more intricate vortical structure than is visible as a ”hairpin.”
After the ejection process depletes spanwise vorticity, it regenerates from the
very high vorticity that persists at all times at the wall surface. It is likely that low
speed streaks in the fully turbulent region pinpoint locations where some variant
of the ejection mechanism described here acts to produce additional vortical struc-
tures that work their way to the outer flow. Vortical structures and their remnants
accumulate away from the wall and fill out the growing turbulent boundary layer.

Acknowledgements This research was supported in part by the National Science Foundation
through XSEDE resources provided by the Pittsburgh Supercomputing Center.

References

1. Adrian, R. J.: Hairpin vortex organization in wall turbulence. Phys. Fluids. 19 041301 (2007)
2. Bernard, P. S., Collins, P., Potts, M.: Vortex filament simulation of the turbulent boundary
layer. AIAA J. 48 1757-1771 (2010)
3. Bernard, P. S. The hairpin vortex illusion. J. Physics: Conf. Ser. 318 060024 (2011)
4. Bernard P. S.: Vortex dynamics in transitional and turbulent boundary layers. AIAA J 51
1828-1842 (2013)
5. Rist, U.: Visualization and tracking of vortices and shear layers in the late stages of boundary-
layer laminar-turbulent transition. AIAA Paper 2012-0084 (2012).

Fig. 9 Isosurfaces of vorticity associated with a hairpin as they develop in zone 3: isosurfaces have
the same interpretation as in Fig. 8.