Documenti di Didattica
Documenti di Professioni
Documenti di Cultura
2l
Ca
2
sin 1
where is the mean inclination angle, and is the wave-
length of the bottom contour. The capillary length is dened
as l
Ca
/(g), with the surface tension , the liquid den-
sity , and the acceleration of gravity g. We have seen that
the inclination angle changes the critical lm thickness. Thus
to check the inuence of surface tension on the vortices
alone for a given waviness without changing the liquid, we
x the inclination angle and vary the absolute scale of the
bottom undulation. To this end we have studied the case of
small and large Bond numbers at the waviness 0.2 for a
wavelength of 20 and 5 mm at two different inclination
angles. The results are shown in Fig. 6. Comparing the data
for a given inclination angle, Fig. 6a shows that the non-
dimensional critical lm thickness is larger for higher Bond
number. At an inclination angle of 45 the difference is rather
small; however, at an angle of 9.8 the gap has widened
considerably. On the other hand, comparing the data for dif-
ferent angles, we see that although the Bond number for the
20 mm waves at 9.8 is larger than that for 5 mm waves at
45, its nondimensional critical lm thickness is not. With
increasing lm thickness the difference in the nondimen-
sional vortex size diminishes so that for thick lms it be-
comes independent of the Bond number as well as from the
inclination angle. We further remark that the Reynolds num-
ber is largest for the case of high Bond number. For the
values shown in the graph, the Reynolds numbers range from
about 510
2
to about 0.2. The surface amplitude for the
different cases is shown in Fig. 6b. The surface amplitudes
apparently decrease exponentially for thick lms. For thin
lms, they cannot be described with an exponential law. Es-
pecially the amplitude for the lowest Bond-number case
studied seems to reach a plateau at a peakpeak amplitude of
about 90 m. Comparing the data for a given inclination
angle, the amplitude for high Bond numbers is higher than
that for the small Bond number for thin lms; however, this
is not the case when comparing different angles.
Up to now we have studied the vortices for a xed wavi-
ness by varying the lm thickness, the inclination angle, and
FIG. 7. Vortex with bent separatrix in the trough of a bottom wave as
visualized by adding up series of tracer images. The main ow direction is
from left to right. Bottom wavelength: 20 mm, amplitude: 9 mm, inclination
angle: 45. Film thickness: 5.3 mm.
FIG. 5. Critical lm thickness for the generation of a vortex as a function of
the inclination angle. The solid squares mark the minimum lm thickness
where the vortex has been observed and the open squares indicate the maxi-
mum lm thickness with no vortex observed. The curve is a t to the data of
the form: H
cr
H
0
(cot
0
cot ), with H
cr
, H
0
, and a
0
being the critical
lm thickness, and two t parameters. Bottom wavelength: 20 mm, ampli-
tude: 4 mm.
FIG. 6. Inuence of the Bond number on the vortex for given waviness. Distance of the separatrix to the center of the trough as a function of the lm thickness
a and peakpeak amplitude of the free surface b. The circles correspond to a wavelength of 5 mm, and the squares to a wavelength of 20 mm. The open
and solid symbols indicate the data for 9.8 and 45 inclination angle, respectively. Waviness: 0.2, Bond numbers: : 3.18; : 0.77; : 0.20; : 0.05.
430 Phys. Fluids, Vol. 15, No. 2, February 2003 Wierschem, Scholle, and Aksel
the absolute scale of the bottom wave to account for the
effect of surface tension. We now study the vortices for dif-
ferent waviness. First of all, we observe that for stronger
waviness the separatrix is not a straight line but is bent.
Figure 7 shows an example for the waviness 0.45. Second,
we detected the critical lm thickness for the vortex as a
function of the waviness. The results are given in Fig. 8
together with a logarithmic t to the data. At the inclination
angle of 45 considered here, the inuence of surface tension
is rather weak as shown in Fig. 6a. We see that the critical
lm thickness increases strongly with decreasing waviness.
For the higher wavy bottoms we could not nd a lm with-
out a vortex at the given inclination angle. Although we de-
creased the lm thickness down to less than 0.7 mm, corre-
sponding to about 1/30 of the wavelength.
In the cases of highly undulated bottoms, we found a
second vortex for sufciently thick lms. An example is
shown in Fig. 9. Like for the rst vortex, their critical lm
thickness decreases with increasing waviness. Figure 10
shows an example of the vortex size and the surface defor-
mation for very high waviness. Concerning the second vor-
tex, we see that it forms again beyond a critical lm thick-
ness and grows with the lm thickness. The velocities in the
second vortex are much smaller than that of the rst vortex.
For a lm thickness of about 0.47 times the wavelength, the
ratio of the velocities along the separatrices is about 3000:1
and it reduces to about 1250:1 for a lm thickness to wave-
length ratio of 0.62. In the region of thick lms, where the
second vortex is observed, the rst vortex shows the same
qualitative behavior as observed before for smaller waviness
in Fig. 3a, i.e., the growth of the vortex diminishes with the
lm thickness. However, for thinner lms the diminution of
the vortex lessens and the size seems to tend to a constant
value for zero lm thickness. For the thinnest lms the po-
sition of the vortex core moves slightly to the steep side of
the bottom contour. The surface amplitude of the thin lms
deviates from the exponential behavior and the surface shape
strongly differs from the sinusoidal form. The latter can be
seen in Fig. 11a. The lm is so thin that its free surface
invades the wavy region. At the falling edge of the bottom
the lm is bent backwards. Here the lm is extremely thin
and almost parallel to the bottom. The rising edge restricts
the level of the lm, since the lms free surface must be
monotonously falling. So in this region, the free surface de-
creases only slightly in ow direction. Thus, the rising edge
enforces a lower limit for the local lm thickness over the
trough. Only by increasing the inclination angle this limit
can be lowered. For the waviness of 0.45 studied here we
observed a ow without vortices at an inclination angle of
80 as shown in Fig. 11b.
We found that small vortices over highly undulated bot-
toms are sensitive to small perturbations. Perturbing the ow
by slight inclinations along the channel width or by a small
pulsation of the ow may deform the vortex into a spiral. An
example of a single vortex is given in Fig. 12a. Figures
12b and 12c show the second vortex. It can even be dis-
placed to the side as seen from Fig. 12c. We notice that in
this case the rst vortex is hardly affected by the perturba-
tion.
The ow separation due to the vortex modies the
boundary condition of the overlying ow downstream in the
sense that there is no no-slip condition for the downstream
ow along the separatrix. This may modify the overall trans-
port velocity of the ow. To clarify the inuence of the vor-
tices on the transport velocity, we measured the mean trans-
port velocity at the lms free surface over 1530 waves and
compared them to the theoretical lm surface velocity over a
corresponding at plate at same inclination angle, thus, for
the same volume ux. The result is shown in Fig. 13. We
remark that in this case there exists a vortex even for the
thinnest lms. For thin lms the transport velocity over the
wavy bottom is considerably smaller than that over a at
incline. With increasing lm thickness the difference tends to
zero. In a certain parameter range, we measured a ratio
slightly larger than one. However, the ratio one is still within
the experimental uncertainty of the data. Finally, for thick
lms, the ratio between the two velocities is about one.
FIG. 9. Film ow with two vortices in the trough of a
bottom wave as visualized by adding up series of tracer
images. The main ow direction is from the upper left
to the lower right. a shows the ow over one bottom
wave, and b is an amplication focusing on the sec-
ond vortex. Bottom wavelength: 20 mm, amplitude: 9
mm, inclination angle: 45, lm thickness: 11.6 mm.
FIG. 8. Critical lm thickness as a function of the waviness. The squares
indicate the critical thickness of the rst vortex and the circles that of the
second vortex. Solid and open symbols refer to the thickness where a vortex
is observed or not observed, respectively. The curves are logarithmic ts to
the data H
cr
/H
0
/B ln(A/A
0
/), with H
0
, B, and A
0
being t pa-
rameters. Inclination angle: 45.
431 Phys. Fluids, Vol. 15, No. 2, February 2003 Vortices in lm ow
IV. DISCUSSION
A. Discussion of the experimental observations
For strongly undulated bottoms, the vortices are gener-
ated beyond a critical lm thickness. We have observed these
vortices for small Reynolds numbers, however, covering sev-
eral orders of magnitude ranging from about 310
5
to
about 1.2. In this range, the Reynolds number apparently is
not responsible for the existence of the vortices. As shown in
Fig. 3a, the vortex takes the same size for the same lm
thickness irrespective of the Reynolds number, which varies
about one to two orders of magnitude. Furthermore, if the
Reynolds number were important for the generation of the
vortices, the critical lm thickness for the generation of a
vortex would be a function of the Reynolds number and
would diminish for increasing Reynolds numbers. However,
as shows Fig. 5, the critical lm thickness increases with the
inclination angle and so does the Reynolds number. Even
more striking is the fact that the critical lm thickness is
higher for larger absolute sizes, as shown in Fig. 6a, and
thus for higher Reynolds number. The Reynolds numbers for
20 mm bottom waves at the critical lm thickness is 5
10
3
and 4.510
2
for 9.8 and 45 inclination angle, re-
spectively. This is about two orders of magnitude higher than
the lowest Reynolds numbers at which we have observed
vortices over 5 mm bottom waves. Furthermore, inertia-
driven ow separation usually takes place at the lee side of a
ow. The vortices here, however, are generated in the very
trough of the bottom wave. We suppose that the reason for
the apparent independence of the vortices from the Reynolds
number lies in the fact that the velocities in the bottom
trough are even much smaller than at the free surface and
consequently as shown in Fig. 13 the mean free surface ve-
locity is about the same as that of a lm owing over a at
incline for the same volume ux. The maximum return ve-
FIG. 10. Vortex size as a function of the lm thickness a and peakpeak amplitude of the free surface b. In a, the distance of the vortex core and of the
separatrix to the center of the trough are shown as open and solid squares, respectively. Squares indicate the rst vortex and circles the second one. The thick
curve in b is an exponential t to the data for thick lms. Bottom wavelength: 20 mm, amplitude: 9 mm, inclination angle: 45.
FIG. 11. Film ow, as visualized by adding up series of tracer images, with a vortex in the trough of a bottom wave at 45 inclination angle a and without
a vortex at 80 inclination angle b. The main ow direction is along to mean inclination angle of the bottom contour. The lm is bent backwards at the
steeply falling edge. In a, the free surface is almost at at the rising edge. b shows the trough region only. The lm over the steeply falling edge is hardly
visualized since the free surface is almost parallel to the laser light. Bottom wavelength: 20 mm, amplitude: 9 mm, lm thickness: 1.1 mm.
432 Phys. Fluids, Vol. 15, No. 2, February 2003 Wierschem, Scholle, and Aksel
locity as depicted in Fig. 4 is about three to four orders
smaller than the mean surface velocity.
Instead of the Reynolds number, the critical lm thick-
ness depends on the inclination angle and on the surface
tension as shown in Figs. 5 and 6. The dependence on the
inclination angle in Fig. 5 can be tted properly with a co-
tangent function. This, on the other hand, corresponds to the
ratio of the force acting in the direction perpendicular to the
mean inclination to that acting in ow direction. It also cor-
responds to the ratio of the horizontal to the vertical dis-
tances between the crests of the bottom and thus restricts the
amplitude of the free surface. The atter the free surface is,
the easier the vortices are generated, i.e., at zero inclination
the free surface has to be at even for thin lms. Surface
tension increases this effect as we have seen in Fig. 6. Low-
ering the Bond number, i.e., increasing the surface tension,
lowers the critical lm thickness. Increasing the effect of
surface tension yields a higher local lm thickness over the
trough of the bottom wave but also a atter surface for the
respective critical lm thickness. Thus, the curvature of the
surface has an impact on the critical lm thickness of the
vortex. This is in accordance with the observation that the
critical lm thickness decreases with the bottom waviness, as
shown in Fig. 8.
We speculate that the strong increase in the critical lm
thickness for lowering the waviness may yield a minimum
waviness below which no vortices can be generated. Since
this would be the case for innite lm thickness, surface
tension is not supposed to play any role. One the other hand,
the region for which a critical waviness exists is restricted for
high waviness as shows particularly Fig. 10a. Thus, beyond
a certain waviness, there exists always a vortex in the trough
of the bottom wave, no matter how low the volume ux may
be. As depicts Fig. 11a this seems to be due to the fact that
the local lm thickness reaches a lower limit determined by
the difference in the vertical position of the rising edge and
the trough of the bottom undulation. Like the critical lm
thickness for the generation of a vortex, the maximum wavi-
ness for a nonzero critical lm thickness also depends on the
inclination angle as shows Fig. 11. The logarithmic t shown
in Fig. 8 was chosen assuming a minimum waviness for the
possible generation of a vortex and a maximum waviness for
the existence of a critical lm thickness. The fact that it ts
well to the data points underlines this conclusion.
Beyond the critical lm thickness, the size of the vortex
is very sensitive to the lm thickness as shown in Fig. 6a
and then converges apparently to an asymptotic maximum
size as depicted in Fig. 3a. The convergence takes place
when the surface is already almost at and thus the stream-
lines close to the bottom hardly change anymore with the
lm thickness. The asymptotic value is independent of the
Bond number as indicates Fig. 6a, which is clear since the
surface is at for thick lms. But it also seems to be inde-
pendent of the inclination angle. Comparing the results for
the waviness 0.2 in Fig. 6 to those for a waviness of 0.45 in
FIG. 12. Film ow, as visualized by adding up series of
tracer images, with a perturbed vortex in the trough of a
bottom wave. The main ow direction is along to mean
inclination angle of the bottom contour. A single vortex
at an inclination angle of 80 is shown in a; b and c
show a second vortex at 45 inclination angle during
different time intervals. Bottom wavelength: 20 mm,
amplitude: 9 mm, lm thickness: 3.67 mm a, and
11.35 mm in b and c.
FIG. 13. Mean transport velocity at the free surface of a lm over a wavy
bottom compared to that over a at incline. The curve is an exponential
decay t to the data. Bottom wavelength: 5 mm, amplitude: 1 mm, inclina-
tion angle: 9.9.
433 Phys. Fluids, Vol. 15, No. 2, February 2003 Vortices in lm ow
Fig. 10 shows that the vortex size seems to be a function of
the waviness.
Although the presence of a vortex changes the boundary
condition for the overlying lm ow downstream along the
separatrix from a no-slip condition to a slip condition, it does
not considerably increase the liquid transport as shown in
Fig. 13. For thin lms the transport velocity is signicantly
lower than that over a at plane, although there already ex-
ists a vortex in the trough. This is apparently due to an in-
crease of the mean lm thickness caused by the bottom un-
dulation as was shown in Ref. 5 for weakly wavy bottoms.
For thick lms on the other hand, the bottom undulation is
not supposed to play an important role anymore for the mean
surface velocity. Between these two extremes, we could not
nd clear-cut evidence for a possible enhancement within
experimental uncertainty.
In the bottom undulations of highest waviness we ob-
served a second vortex. It apparently shows the same quali-
tative behavior as the rst one. As shown in Fig. 8, there
apparently also exists a minimum waviness for its genera-
tion. The critical lm thickness also decreases with increas-
ing waviness. It also appears from Fig. 10a that its size
tends to an asymptotic value for thick lms.
B. Comparison to other studies of vortices at low
Reynolds numbers
We have seen that Reynolds number effects are not im-
portant in our study. Therefore, we conclude that the ob-
served vortices are not driven by inertia as those studied in
Refs. 68 and start comparing our ndings to Pozrikidis
numerical study of Stokes ow over sinusoidal inclined
planes.
9
He considered lm ow over bottom undulations
with a waviness of 0.01, 0.1, and 0.2 at inclination angles of
9 and 45. Furthermore, he studied the impact of the Bond
number. In our experiments, we covered the waviness rang-
ing from 0.13 to 0.5 at inclination angles ranging from 5 to
80. Both Pozrikidis numerical and our experimental study
coincide in the waviness 0.2 and the inclination angle of 45.
Although he mainly focuses on the free surface prole of the
lm while we center our attention on the vortices, we found
good agreement between the two studies whenever compa-
rable. As in our experiments, Pozrikidis observed from his
numerical calculations that the free surface is a nearly sym-
metric sinusoidal wave for thicker lms and it becomes al-
most horizontal over increasing slopes of the bottom for thin
lms. A comparison of Pozrikidis calculations for the free
surface amplitude at innite Bond number to our experimen-
tal data with Bo*4.5 shows good agreement and is repro-
duced in Fig. 14. Also the effect of the Bond number seems
to compare well qualitatively. Although the small Bond num-
bers are not the same in both studies, we observe as does
Pozrikidis that the surface amplitude does not tend to the
bottom amplitude for thin lms but to much smaller values.
The fact that the Pozrikidis calculations for a waviness
of 0.1 could not reveal any vortex coincides with our obser-
vation of a strong increase of the critical lm thickness by
lowering the waviness. For a waviness of 0.2, Pozrikidis did
not observe any vortex for thin lms but for thick lms.
Although he did not determine the critical lm thickness, at
least the points he studied t to ours, i.e., they show the same
qualitative behavior. Also the wall shear stress from which
he deduced the existence of a vortex is nearly symmetric
coinciding to our symmetric vortices for that parameters.
Pozrikidis stated that vortices occur beyond a critical wavi-
ness and that the critical lm thickness vanishes for very
large wave amplitudes, such as we have observed it in our
experiments. Unfortunately, however, he did not embark on
this subject and did not give evidence for it.
Like the systems studied experimentally by Taneda
10
we
found that the characteristic velocity of the second vortex is
several orders of magnitude smaller than that of the rst one.
In general, one may classify the different Stokes ows with
vortices studied by Taneda
10
into two groups: In one class,
vortices occur beyond a certain angle between the ow con-
ning boundaries such as Moffatts eddies.
13
There is no
length scale entering into this kind of problems and similar-
ity solutions are obtained. In the other group, the vortices are
generated below a certain length scale. Into this second class
fall the vortices between two spheres that are not in contact
of which the ow pattern has been calculated by Davis
et al.
11
The lm ow studied here belongs to this second
class with the waviness playing the role of a length scale
taking the lm thickness xed. Like, for instance, the case
described by Davis et al.,
11
increasing the waviness leads to
a generation of a further vortex. However, different from
those cases studied by Taneda
10
we have a free surface ow
and the lm thickness enters as a further parameter. Although
the experimentally accessible range for the lm thickness is
limited, it seems from our experiments that the maximum
number of vortices depends on the waviness and not on the
lm thickness.
In their experiments on free lm ow along vertical cor-
rugated surfaces, Zhao and Cerro
14
have observed vortices
between periodic convex half cycles independent of the lm
thickness. It seems to us that these are essentially those de-
scribed theoretically by Moffatt.
13
Besides these, they stated
that they found vortices only at the highest Reynolds num-
bers and Capillary numbers for walls shaped as triangles or
as concave half cycles but did not give evidence of these
FIG. 14. Comparison of the free surface amplitude with numerical calcula-
tions for Stokes ow by Pozrikidis. The experimental data are given as solid
squares; the numerical values, taken from Ref. 6, Fig. 5b, are indicated by
open squares. Parameters: waviness: 0.2 mm, inclination angle: 45, Bo*:
4.44 experimental and innity numerical.
434 Phys. Fluids, Vol. 15, No. 2, February 2003 Wierschem, Scholle, and Aksel
observations. However, Malamataris and Bontozoglou have
calculated numerically the ow along concave half cycles
and found recirculation zones behind the corner at the lee
side of the half cycles for high ow rate and surface
tension.
17
By varying the lm thickness, Reynolds number,
and the Capillary number, they showed that the eddy can
disappear. They found as a common feature of these calcu-
lations that in all three cases the local lm thickness dimin-
ished. Apart from this, however, we nd certain differences
to our study: The vortices that are generated in the concave
half cycles seem to be a ow separation due to a corner ow
while there is nothing like this in our experiments. The vor-
tices in our study are in general symmetric in the trough of
the bottom while those vortices are at the lee side. Finally,
we could show that in our case the Reynolds number is not
important.
Negny et al.
15
observed a bulge of the lm over the at-
test part of an undulated substrate and concluded in accor-
dance with numerical calculations
18
that the swelling is
caused by an underlying vortex. Different from our vortices,
these ones occur at rather high Reynolds numbers and Negny
et al. suggest that viscous friction prevents the ow from
overcoming the pressure gradient in these regions. From the
fact that these vortices occur in the attest region of the
corrugations it seems clear that these vortex are different
from ours.
V. CONCLUSIONS
We have presented an experimental study of vortices in
gravity driven lms owing down sinusoidal bottom proles
of rather high waviness. The vortices were visualized em-
ploying a particle image velocimeter with uorescent tracers.
They were observed in the troughs of the undulated bottom
prole at low Reynolds numbers down to the order of 10
5
.
We showed that the Reynolds number is not responsible for
their generation. From the experimental data, it seems that
there exists a minimum waviness below which these vortices
cannot be generated. Beyond this minimum waviness, the
vortices occur beyond a critical lm thickness. The minimum
lm thickness for their generation increases with the inclina-
tion angle and is lowered by surface tension. The critical lm
thickness diminishes with increasing waviness until it
reaches zero. This waviness depends on the inclination angle.
Beyond this waviness, there exists always a vortex irrespec-
tively of the lm thickness. Further increasing the waviness
results in the generation of a second vortex beyond a critical
lm thickness. This second vortex occurs again in the trough
of the undulation and seems to have the same features as the
rst one. The size of the vortices strongly increases beyond
the critical lm thickness and tends asymptotically to a nite
value for thick lms that is independent of the inclination
angle or surface tension. Finally, we could not observe a
signicant increase of the transport velocity due to the vor-
tices.
ACKNOWLEDGMENTS
The authors acknowledge the support of G. Jena, A.
Kammerer, A. Do
rnho