Sie sind auf Seite 1von 15

Under consideration for publication in J. Fluid Mech.

1
Quasi-steady corner ow
(Received 13 October 2009)
In the absence of signicant body forces the passive manipulation of uid interfacial
ows is naturally achieved by control of the specic geometry and wetting properties of
the system. Numerous microuidic systems on Earth and macrouidic systems aboard
spacecraft routinely exploit such methods and the term capillary uidics is used to de-
scribe both length scale limits. In this work a collection of analytic solutions are oered
for ows where a bulk capillary liquid is slowly drained by a faster capillary ow along at
least one interior edge of the container. The solutions are enabled by an assumed known
pressure (or known height) dynamical boundary condition. This boundary condition
can be in part determined a priori from the container dimensions and further quanti-
tative experimental evidence, but not proof, is provided in support of its expanded use
herein. In general, a small parameter arises in the scaling of the problems permitting a
decoupling of the local edge ow from the global bulk ow. The quasi-steady asymptotic
system of equations that result may then be solved in closed form for a useful variety of
geometries including uniform and tapered sections possessing at least one critically wet-
ted interior edge. Draining, lling, ullage migration and other imbibing ows are studied.
Preliminary experiments show favorable agreement with terrestrial and drop tower ex-
periments. The solutions are valued for the facility they provide in computing optimal
designs for select capillary uidics problems by way of passive transport rates. Because
geometric permutations of any given design are myriad, such analytic tools are capable
of eciently identifying and comparing critical design criteria (i.e. shape and size) and
the impact of various wetting conditions resulting from the uid properties and surface
conditions. The dimensional solutions are categorized along with their constraints and
sample optimizations are performed to demonstrate the utility of method.
1. Introduction and Scope
Microuidic processes on Earth such as ow circulation in capillary-driven heat pipes
(review art.) demonstrate the degree to which passive control may be maintained pro-
vided the characteristic system dimension is small, i.e. capillary length O(1mm). In the
microgravity environment aboard spacecraft, the reduced body force permits identical
control, but for systems with signicantly larger characteristic dimensions, i.e. capillary
length O(1m). The latter processes might be called macrouidic, and dramatically ex-
pand opportunities to exploit capillary forces to carry out tasks in unearthly ways, such
as to storage and manage of large quantities (tons) of liquid fuels or cryogens (chato),
or to circulate and separate large quantities of water in spacecraft life support systems
(Weislogel et al.). The term capillary uidics is adopted to describe both length scale
limits and is loosely dened as the management of uid interfacial systems by relying
strongly on the passive means aorded by system geometry and wetting conditions.
One geometric construct that has proven particularly useful in capillary uidic sys-
tems is the interior edge (or interior corner). Employed in, say, a polygonal container
or conduit, if the interior edge included angle and liquid-uid contact angle are small
2
enough, the edge is spontaneously lled by the wetting uid. This critical edge wetting
condition provides an attractive and eective means for truly passive uids transport
and control. For a variety of applications it can be used to drain large uid quantities
from one container to another, or to passively separate immiscible uid phases for further
processing. In such capillary uidic systems, the capillary edge ow governs the capillary
bulk liquid ows throughout the container or conduit.
In this paper, a brief description of capillary edge ow is provided along with a thread of
published works leading to the postulation of a dynamical boundary condition upon which
all of the subsequent analytical solutions depend. The boundary condition is restricted to
slender uniform or tapered sections with either sharp or rounded interior edges. Further
experimental evidence, but not proof, is provided in support of using the boundary
condition. Lists of local and global ow assumptions are then presented that reduce the
general system to one of j+3 equations, where j is the number of critically wetted interior
edges for the container: j local mass balance equations, 2 global mass balance equations,
and 1 global volume constraint. The equations are nondimensionalized by proven scales
giving rise to xx(?) geometric parameters, which are dramatically simplied with a low
saturation assumption applied to the local edge ow regions of the problems. Further
geometric restrictions are made such as n-regular-edge wetting containers of uniform or
tapering section (i.e. containers with n-regular polygonal sections), but only for clarity in
presentation. Sample solutions are then outlined to illustrate the general method. These
are followed by lists of results for several families of geometry types. Draining, lling,
ullage migration and other imbibing ows are studied. Dimensional forms and constraints
useful in design and analysis applications are provided along with cursory terrestrial
and drop tower experiments. The favorable agreement adds condence to the quick all-
analytic method for determining optimal designs for such capillary uidic systems prior
to the more time consuming task of full numerical computations. Directions for continued
research are highlighted in summary.
2. Review of Capillary-Driven Flows along Interior Edge
2.1. Edge Wetting Condition
A large literature exists for capillary driven ows in containers, along conduits, or across
surfaces that possess interior edges, channels, or grooves that are spontaneously wetted
by a uid. For sharp interior edges formed by planar walls, spontaneous ows occur when
the critical geometric wetting condition l/2 is satised, where is the equilibrium
contact angle of the wetting uid and is half the included angle of the planar walls. The
mathematical grounds on which this and other critical geometric wetting phenomena are
based are credited to mathematicians Concus and Finn (1969, 1974, 1990) who demon-
strated a discontinuous behavior of the uid surface as the critical value is crossed, and
established, through proofs of non-existence, that no other non-edge-lling surfaces are
possible. Because such critical wetting phenomena may also depend in a complex man-
ner on other geometric quantities such as overall container size, shape, ll fraction, and
wetting conditions, the discontinuous, or even nearly discontinuous, geometric wetting
condition for a particular system is referred to in general as the Concus-Finn condition
for the given geometry. In fact, under this denition, more than one Concus-Finn condi-
tion is possible for a single container and three are demonstrated by Tavan et al. (2010).
Earlier heuristic descriptions of critical wetting conditions have appeared in the literature
(e.g. cite and cite) with perhaps the most notable and uncirculated mathematical contribution
being that of Moisev (1968).
Quasi-steady corner ow 3
For partially wetting systems with nite contact angle hysteresis, at least (most) the
advancing (receding) contact angle should be used when evaluating Concus-Finn wetting
(dewetting) conditions in the short term. However, thermal and mechanical perturba-
tions over time appear to bring about the mathematically idealized conditions based on
the equilibrium contact angle alone; see Concus et al. (proboscis). Lastly, it is of interest
to note that no augmentation of the Concus-Finn condition has been reported for van
der Waals uids where the idealized three-phase contact line model breaks down (Wong
et al. 19xx,Wong et al. 20xx), or for micro/nano-scale systems where line tension and
contact line curvature signicantly alter local wetting conditions (e.g. Japanese, 20xx).
2.2. Spontaneous Edge Flows
For a record of model development for sharp and rounded interior edge ows leading to
the present work see Ayyaswamy et al. citeayya, Ransoho & Radke citeransohoradke,
Dong & Chatzis citedongchatzis, Kovscek & Radke citekovscekradke, Romero & Yost
citeromeroyost, Weislogel & Lichter citeweislichter, Langbein citelangbein, and Chen et
al. citechenetal. Extensive experimental verication is reported. An assortment of solu-
tions for open channels and conduits are collected by Weislogel citeweisadv, and gravity
draining problems are solved by Quere and xxxx citeQuere and Ram and Weislogel cit-
erameweis. Sample capillary uidics systems are demonstrated on Earth by Germans
(20xx) and xxxx (xxxx) and in low-g environments by Weislogel et al. citeweisthomas.
A typical spontaneous edge wetting ow is presented in Fig. 1 for a NASA 2.2s drop
tower test (NASA citenasadroptower of a perfectly wetting ( = 0

) liquid partially
lling a 30-60-90 right triangular cylinder (Weislogel citeweisirreg). Prior to the drop,
Fig. 1a, the interface is predominately at owing to the presence of gravity where the
Bond number Bo g
o
R
2
/ >> 1; is the density dierence across the interface, g
o
is
the acceleration of gravity (9.81m/s
2
), R is the characteristic dimension of the section
(e.g. radius of inscribed circle), and is the free surface tension. Following the release
of the experiment into free fall (the drop), Figs. 1b and 1c, net body forces are nearly
eliminated, and because the Concus-Finn condition < /2
j
is satised in each of
the j = 3 interior edges, spontaneous edge ows draw uid away from the bulk meniscus
by pumping it along the corners. Several important observations and their respective
model impacts are noted here (ref. Fig. 1):
The interior edge ows are driven by gradients in capillary pressure along the edge
and rapidly become slender, Fig. 1c-a situation that permits the increasingly appropriate
parallel ow and dominant cross-stream 1-dimensional curvature assumptions. The for-
mer assumption often negates inertia as a signicant player despite the high rates of ow
possible in certain large length scale capillary uidic systems. The parallel nature of the
edge ow also means that ows normal to the contact line are small compared to ows
parallel to it, and that the static contact angle provides a reasonable approximation for
the free surface boundary condition in planes perpendicular to the edge axes; see sketch
of section in Fig. 1c. Thus, only the z-component momentum (Poisson) equation must be
solved for the local ow rates

Q
j
(z, t). These results are in turn used in local volume bal-
ances to compute second order nonlinear PDEs for the evolution of the meniscus height
h
j
(z, t) for each of the edge ows. Solutions to these equations just require boundary
conditions.
The edge ows rapidly advance over pre-wetted portions of the edges and on to dry
regions where the height of the meniscus at the tip location is essentially zero, Fig. 1b.
Thus, h
j
(L
j
, t) = 0 serves as one boundary condition for the ow.
Each edge ow interface appears to pivot around a xed location near its source as
denoted by H
30
in Fig. 1, which display the 30

edge ow in near distortion-free prole


4
due to the geometry and orientation of the test cell (see top view section in Fig. 1a). This
condition is referred to as the constant pressure condition, but is equivalent to a constant
height boundary condition h
j
(0, 0) = H
j
. Its z-location provides a coordinate origin for
the ow, but is apparently only known by empirical means. With H
j
known a priori, a
second boundary condition is available to solve the governing system of equations. It is
noted that the constant height location exists outside the uid in the initial state, Fig.
1a.
For many related edge ows with eectively constant pressure boundary conditions,
the edge ows advance, and bulk menisci recede, with a long time t
1/2
behavior. A
formation time is required before the constant height condition may be accurately applied
(Weislogel and Lichter ? and Weislogel ?). Uncertainties of predictions employing the
constant height condition are frequently within 5% of experimental results. This does
not prove that the location and value of the constant height condition are facts, but it
certainly argues in support of their practical application.
Fortunately, the x-dimension location of the constant height boundary condition may
be determined explicitly for a variety of cylindrical container types using the method of
de Lazzer et al. ? with limitations noted by Finn and Neel ?. Unfortunately, no clear
analytic method for determining the z-dimension location is available. Because the z-
coordinate location for the constant height condition is central to the current analysis it
is discussed in greater detail here.
2.3. Constant Height Location: Discussion
For ease of discussion, the ow scenario of Fig. 1 is idealized in Fig. 2 for a right cylinder
with snow cone cross-section of area A, where critical edge wetting only occurs along the
single interior edge. The initial condition in Fig. 2a is for g yielding a perfectly at
surface. As observed in experiments, Fig. 2b displays overlays of surface proles at various
times following the cancellation of gravity. The constant height boundary condition is
quickly achieved at a distance H from the container edge and at z 0 where the interfaces
intersect, Fig. 2b and 2c. This location is the convenient stationary coordinate origin for
the ow. The mathematical assumptions of the problem are increasingly satised both
up- and downstream of this origin as time increases and the ow slows, becoming
increasing slender. The assumptions break down however in the vicinity of the bulk
meniscus because the curvature is two-dimensional there. A long time interface prole
at t4 is sketched in isolation in Fig. 2c. As velocities shrink t
1/2
, viscous normal
stresses vanish, and the bulk meniscus achieves a constant curvature ascribed to z z
b
,
and thus a constant pressure. The fact that the pressure is known anywhere in the ow
should serve as a means for applying at least one advantageous boundary condition
there. Either constant pressure conditions at the bulk meniscus or at the coordinate
origin would suce, and both yield t
1/2
predictions. But only the latter provides a
quantitative constant height boundary condition because it occurs in a region that is
well-characterized by one-dimensional curvature. (Note also that within experimental
uncertainty, the constant height location occurs at the same z-location for all j wetted
edge ows; ref. Fig. 1, or see Weislogel ?.)
Conserving volume, the amount of uid pumped past z = 0 is equal to that removed
from the bulk. For long times, the volume of the constant curvature bulk region ap-
proaches a xed value while everywhere else the ow volumes grow with time. This
implies that the similarly cross-hatched time-increasing uid volumes in Fig. 2c must
be equal, and that the similarly cross-hatched constant uid volumes in Figs. 2a and 2c
must be equal. If one could compute the volume of the long time bulk meniscus region
V
b
, one could easily compute the location of the z-coordinate origin, since V
b
= Az
o
for
Quasi-steady corner ow 5
this idealized ow. A useful method to estimate z
o
is provided in the Appendix, but a
rst principles analytic method remains to be seen.
Three nal observations are made here before experimentally demonstrating many of
the above arguments with a simple ow example. (1) For many slender containers types
the saturation levels in the edge ow regions of the container are small. Thus, the edge
ows advance far and fast while the bulk meniscus region recedes short and slow. This
suggests two very dierent time scales for ow. (2) Because the bulk interface moves only
slightly compared to edge ows, the constant height value at z = z
b
is only slightly larger
than at z = 0. In other words, the pressure at z
b
is only slightly higher than the pressure
at z = 0 and the majority of the pressure loss occurs across the edge ow between z = 0
and the advancing tip. (3) As time increases and the ow eld continues to grow at
t
1/2
, relatively speaking, the xed volume bulk curvature region shrinks to zero. Taken
together these observations suggest that one might further idealize such ows as sketched
in Fig. 2d, where the bulk meniscus is reduced to a line at z = z
b
(t), where H(z
b
) H.
Further support for such a model is provided in connection with a simple experimental
example.
2.4. Simple Example of a Capillary Edge Draining Flow
As Fig. 3 depicts, a 1mm thin-walled rounded corner square fused quartz tube is sealed
at it rightmost end and completely lled with a perfectly wetting low vapor pressure
silicone oil (uids properties measured at (porps). The tube is placed horizontally over
a backlight panel mounted to a level translation stage. A roll of absorbent tissue is
then brought into contact with the open leftmost end of the tube. The liquid near the
tube opening quickly wicks into the smaller pores of the tissue as the counter ow of
air is drawn through the tissue and into the tube. The bulk meniscus slowly recedes
as the interior edges pump liquid from the bulk into the tissue. The tissue roll may be
rotated to avoid saturation. The bulk meniscus region is tracked and recorded by video
microscope. For these long duration experiments the impacts of tube level and precision,
tissue resistance to air ingestion, saturation, and particulate contamination, and the
eects of temperature dependent uid properties including evaporation are all considered
and demonstrated to be within the constraints of the pending modeling assumptions. The
draining bulk meniscus location z
2
(?) is plotted as a function of dimensionless time (t

)
in Fig. 4 for a number of repeated experiments. The development of t

will be presented
in the analysis section. The long time behavior is expectedly t
1/2
and a theoretical
curve based on the proposed model assumptions is included on the gure and will be
discussed in detail shortly. To reduce optical distortions for comparisons of bulk interface
proles, the experiments are repeated using thick-walled rounded square tubes (material,
OD, square dim., roundedness, length). Select bulk interface proles from one such test
are overlaid in Fig. 5 such that bulk menscii centerlines are coincident. Clearly, the bulk
meniscus region rapidly establishes a constant curvature condition that maintains for
the duration of this 152.5hr experiment. (Note: a consistent asymmetry in the interfaces
is noticeable and caused by a correctible optical distortion amplied by a slight out-of-
vertical line of observation as depicted in the exaggerated sketch inset in Fig. 5.)
Because the tissue provides a large volume low capillary pressure sink, the boundary
condition at the open tube end is eectively a constant meniscus height condition for
all interior edges where h(0, t) 0. The question becomes one of what or where to
apply a second boundary condition? The bulk meniscus proles in Fig. 5 are scaled by
(z z
b
)/t
1/2
and re-plotted in Fig. 6. As time and distance increase, the bulk region is
observed to shrink to nearly a line as conjectured by Weislogel and Lichter (1998) and
as sketched for the idealized ow in Fig. 2d. Representative interface proles along the
6
interior edges are lightly sketched using dashed lines where they fall outside the eld
of view of the camera. For the majority of the interface, as t it seems logical
that the edge ow interfaces approach the constant height condition of de Lazzer et
al. The pivotal assumption of this research is that the height h(z, t) of each interior
edge ow corresponds to the constant height conditions of de Lazzer et al. at the bulk
meniscus which at long times is compressed to a line. In other words, h(z
b
(t)) = H. With
such a solution dependent (dynamical) boundary condition an important variety of bulk
capillary ows governed by interior edge ows may be approximated and solved.
Fig. 1. Drop tower test of xx cs Si oil ( = 0

) in a 30-60-90 right triangular cylinder:


a) g = g
o
at t = 0, b) g = 0, t = xxs, c) g = 0, t = xxs. Section, dimensions, and 30

edge uid sketched in respective top views. Test cell is constructed to minimize optical
distortions for ow along 30

edge. The constant height location H


30
is identied for the
30

edge ow. After a brief transient, the interface pivots around this point. L
30
and L
60
identify advancing menisci in the 30

and 60

corners, respectively.
Fig. 2. Idealized edge wetting ow: a) g , t = 0, b) g = 0 overlay of interfaces at
sequential times, c) longtime dynamic interface t
4
from b) indentifying constant curvature
bulk interface quantities and equal volumes of transient displaced volumes and constant
curvature volumes (a and c), and d) further idealized ow model where bulk meniscus is
shrunk to a line where H(z
b
) H.
Fig. 3. Schematic of rounded square tube draining experiment. Flow is initiated once
tissue roll is brought into contact with leftmost open end of tube. Images of the bulk
meniscus at select z-locations shown below. Tissue particles are observed to diuse a
short distance upstream (ref. z
b
= 4mm).
Fig. 4. Draining bulk meniscus location z
2
(t

) (xx) for repeated experiments of 5cs Si


oil in thin-walled rounded square tube (xx tube properties: OD, ID, Rc). Analytic curve
shown for comparisons.
Fig. 5. Overlay of bulk meniscus region interface proles at dierent z-locations during
5cs Si oil ow in thick-walled rounded corner square tubes (xx tube properties: OD, ID,
Rc). Note that (1) optical distortions are not corrected and that (2) asymmetry in the
proles is attributed to a slight (2

) tilt in the viewing axis which is exaggerated in


the schematic-both are identical for each image and the interfaces collapse well for all z
observed. Interfaces for z < 4mm are prevent by the imaging method employed.
Fig. 6. Scaled proles of Fig. 5. As time and meniscus location increase, the bulk
meniscus region becomes increasingly small, shrinking to zero in the limit. The value of
the constant height location H is sketched using a light dashed line. A sketch of the edge
of the container and the out-of-view interface are shown with light solid and dotted lines,
respectively. The primary assumption of this research is that h(z
b
(t)) = H.
3. Quasi-Steady Corner Flows: Solutions
3.1. List of Assumptions Employed
The assumptions employed to solve the local edge ows are listed below.
(a) The corner ow is suciently slender. This assumption is justied assuming
2
1
where = H
2
/z
2
.
(b) Fluid inertia in the corner ow is negligible, a condition which is usually satised
under assumption a.
(c) Curvature in the streamwise direction is negligible. This requires
2
f 1, where
f is a dimensionless geometric function, f = sin cos sin, again usually satised
under assumptiona .
Quasi-steady corner ow 7
(d) Gravity in all directions is ignored.
(e) The uid pressure or meniscus height at the top of the container, h(0, t) is known.
(f) At z
2
the percent of the total cross sectional area occupied by the uid in the corners
is small. This assumption seems necessary to yield closed form analytical solutions, and
will be discussed in more detail shortly.
The global ow modeling assumptions are:
(a) The method of de Lazzer et al. (??) is sucient to determine the 2-D radius of
curvature R at the location of the bulk meniscus. Knowing R the interface height h(z, t)
at z
2
may be computed, serving as a boundary condition for the local ow.
(b) Coupled with assumption a it is necessary that the volume of gas occupied in the
bulk meniscus be much smaller than the volume occupied elsewhere. To this end the bulk
meniscus is modeled as at as shown in gure ?? ??.
(c) For expanding, contracting or otherwise non-uniform geometries, the local length-
wise coordinate z collinear with the edge is modeled as being identical to the z

coordinate
dening the distance from the at bulk meniscus to the origin.
3.2. Corner Flow Formulation
Figure XX depicts a section of a single corner forming an included interior angle 2.
The contact angle is and the z -axis lies along the corner. The meniscus height h(t, z)
is measured from the corner to the free surface in the (z, y) plane. The meniscus height
is scaled by H and the length of the uid column along the z -axis is scaled by L. All
other scales are listed in table XX. The analysis assumes a slender (
2
1), low inertia
(R 1) ow with small curvature along the z axis (
2
f 1). Here is the slenderness
ratio
= H/L (3.1)
and
R
2
H
f
2
_
T
2
c
1 +T
2
c
_
2
(3.2)
f =
sin
cos sin
(3.3)
where R is a modied Reynolds number with characteristic velocity /. The surface
curvature is characterized by f, such that R = hf where R is the radius of curvature. T
c
is a ratio of the y and x length scales to be dened momentarily. Under these constraints,
it can be shown that the leading order Navier Stokes equations only consist of the z -
component momentum equation in the form
1

P
z
=

2
w
x
2
+

2
w
y
2
(3.4)
where w is the z -axis velocity, P is the pressure, and is the dynamic viscosity. Equation
3.4 is subject to no-slip along the walls, no shear stress on the free surface, and a known
contact angle along the contact line. Under the present assumptions pressure can be
determined from the normal stress boundary condition
P =

R
=
1
hf
. (3.5)
8
Following the Laplacian scaling method detailed by Weislogel et al. (2008), equation 3.4
is normalized using the spatially and temporally varying scales
x h(z, t) = x
s
(3.6)
y htan() = y
s
(3.7)
where the overbar is used to distinguish the time and space dependence of these scales.
Using 3.4 a scale for the average velocity is determined
w w
s
=
x
2
s

P
z
_
T
c
2
1 +T
c
2
_
(3.8)
which when implemented gives the dimensionless equation
1 =
_
T
c
2
1 +T
c
2
_

2
w

x
2
+
_
1 +T
c
2
T
c
2
_

2
w

y
2
. (3.9)
Equation 3.9 can be used to nd the average velocity
w

=
x
s
y
s
A
_ _
w

dx

F
i
(3.10)
where A is the dimensional ow area, given by
A = F
A
h
2
(3.11)
F
A
= f
2
_
cos sin
sin

_
. (3.12)
Solutions for F
i
are possible for all values of and . Most often a numerical procedure
is required, but asymptotic solutions are possible for a few limiting cases. Because of the
scaling, F
i
(, ) is tightly bounded 1/8 F
i
1/6 for all values of and and may
be treated as an O(1) constant with errors usually < 5%. The problem of nding the
average velocity is not complete, because the scales used are unknown and z -dependent.
Converting back to dimensional form gives
w = w

w
s
=

F
i
sin
2

f
h
z
(3.13)
which is substituted into continuity in the form
A
t
=
(wA)
z
(3.14)
to give the dimensional evolution equation
h
2
t
=

F
i
sin
2

z
_
h
2
h
z
_
. (3.15)
3.3. Draining Flows
Figure XX shows the schematic used to formulate the governing equations for constant
tip location ows. The subscripts 1 and 2 refer to quantities evaluated at the top of the
container (z -origin) and the location of the bulk meniscus, respectively. This analysis
assumes that the volume of gas occupied by the bulk meniscus is small compared to the
volume occupied elsewhere, and is therefore modeled as being at. Also, it is assumed that
the constant height location coincides with the at bulk meniscus location. Experiment
seems to support these claims, results of which will be published when available. Finally,
Quasi-steady corner ow 9
the perfect pumping condition at the origin is implemented by specifying a zero height
boundary condition there, essentially satised if (H
1
/H
2
)
3
1.
3.3.1. Formulation
The analysis is simplied when considering a container with a constant cross sectional
area because the boundary conditions for the governing evolution equation are constant,
namely 3.15
h
j
(0) = 0 h
j
(z
2
) = H
2j
. (3.16)
The subscript j refers to the jth of m interior corners in the container. The total ow
rate Q out of the container is written as the sum of all individual ow rates due the
capillary pumping at the top of the container. Employing 3.11 and 3.13
Q
1
=
m

j=1
A
j
w
j
=
m

j=1

F
aj
F
ij
sin
2

j
f
j
h
2
j
h
j
z

0
. (3.17)
A global volume balance is introduced assuming that the rate of liquid leaving the con-
tainer must equal the rate at which gas enters
Q
1
=
V
t
=

t
_
_
A
s
z
2

j=1
z
2
_
0
F
Aj
h
2
j
dz
_
_
(3.18)
where V = V (t) represents the volume of gas in the container. Equation 3.18 needs one
boundary condition. While it is possible to specify z
2
(t = 0) for academic purposes, the
more practical boundary condition is calculated assuming that the initial volume of gas
in the container, V
i
is known. Then z
2
(t = 0) can be determined by solving
V
i
V (0) = A
s
z
2
(0)
m

j=1
z
2
(0)
_
0
F
aj
h
2
j
dz. (3.19)
Equations 3.17 and 3.18 are combined to form the governing global ow equation
m

j=1

F
aj
F
ij
sin
2

j
f
j
h
j
2
h
j
z

0
=

t
_
_
A
s
z
2

j=1
z
2
_
0
F
aj
h
2
j
dz
_
_
(3.20)
which in conjunction with 3.15 and their associated boundary conditions form a system
of m+1 equations and unknowns. The system is to be solved for the meniscus location
z
2
(t) from which most useful aspects of the ow may determined.
3.3.2. Scaling
The two independent variables and one dependent variable of the problem are normal-
ized by the following scales:
z
V
i
A
s
t
V
i
Q
s
h
j
H
2j
(3.21)
Where Q
s
is a total ow rate scale and W
j
is the jth local average velocity scale
W
j
=

F
i
sin
2

j
f
j
H
2j
V
i
/A
s
Q
s
=
m

j=1
F
j
H
2
2j
W
j
. (3.22)
10
Employing these scales in the governing system of 3.15 and 3.20 yields the jth local
corner ow equation
_
_
m

j=1

j
_
_
h

j
2
t

=
j

_
h

j
2
h

j
z

_
(3.23)
subject to the boundary conditions
h

j
(0) = 0 h

j
(z

2
) = 1 (3.24)
where

j
=
A
j
A
s

j
=
W
j
W
m
(3.25)
and W
m
is the maximum of the m local average velocity scales. The quantity

m
j=1
(
j

j
)
appearing on the LHS of 3.23 is perhaps best thought of as representing the ratio of total
local ow to global ow time scales. The condition

m
j=1
(
j

j
) 1 therefore signals
the quasi-steady ow scenario of fast local corner ows subject to boundary conditions
from slowly developing global ows. The parameter
j
is a ratio of the jth velocity scale
to the maximum velocity scale, thus 0 <
j
< 1. When
j
1 the jth corner ow does
not contribute signicantly to the overall ow rate and may be ignored at O(1). The
dimensionless global equation 3.20 becomes
_
m

j=1

j
h

j
2
h

j
z

0
_
m

j=1

j
=
z

2
t

_
_
_
m

j=1

j
z

2
_
0
h

j
2
dz

_
_
_ (3.26)
subject to the initial condition
z

2
(0) = 1 +
m

j=1
z

2
_
0

j
h

j
2
dz

. (3.27)
In 3.27
j
appears alone, and although expressed as a ratio of the jth corner ow area
to total container cross section,
j
is actually the result of a ratio of local to global time
scales. The limiting case of
j
1 represents ows of low saturation and the liquid
occupying the corners may be therefore be ignored in the global ow equations at O(1).
The parameter

j

m
j=1

j

j
appears in each of the m terms on the LHS of the global
equation representing the ratio of ow rate contributed by the j
th
corner to that of the
total ow rate.
The equation forms 3.23-3.27 are simplied for the special case of identical corner ow
velocity scales. For such containers
j
= 1 and the system is reduced to having only one
parameter
j
. A further simplication results if the ow area is also the same for each
corner, in which
j
is the same for each corner. Thus, for constant contact angle and
equal interior corner angles throughout,the local and global governing equations reduce
to

h
2
t

=

z

_
h
2
h

_
h
2
h

0
=
z

2
t

2
_
0
h
2
dz

(3.28)
Quasi-steady corner ow 11
subject to their respective boundary conditions
h

(0) = 0 h

(z

2
) = 1 (3.29)
z

2
(0) = 1
z

2
(0)
_
0
h
2
dz

(3.30)
where =

m
j=1

j
. The above simplied system governs constant tip location ows
in which the velocity scales and local corner ow areas are identical in every corner. This
simplication is not necessary to yield solutions, but is utilized here so the following
examples are concise.
3.3.3. Solution
The quasi-steady assumption requires 1 whereupon the local and global equations
3.28 are decoupled and may be solved in series. Anticipating an asymptotic solution, h

and z

2
are expanded naively in the form
h

= h

0
+h

1
+O(
2
) (3.31)
z

2
= z

20
+z

21
+O(
2
) (3.32)
and substituted into the governing system 3.28 yielding the O(1) local equation

_
h

o
2
h

o
z

_
= 0 (3.33)
subject to the boundary conditions
h

o
(0) = 0 h

o
(z

2o
) = 1. (3.34)
The zeroth order global equation is
_
h

o
z

o
2
_

=0
=
z

2o
t

(3.35)
subject to
z

2o
(0) = 1. (3.36)
The O() local equation is
h

o
2
t

=

z

_
2h

1
h

o
h

o
z

+h

o
2
h

1
_
(3.37)
subject to
h

1
(0) = 0 h

1
(z

21
) = z

20
h

o
z

=z

2o
(3.38)
and the O() global equation is
_
2h

1
h

o
h

o
z

+h

o
2
h

1
z

=0
=

t

_
_
_z

21

2o
_
0
h

o
2
dz

_
_
_ (3.39)
12
subject to the initial condition
z

21
(0) =
z

2o
(0)
_
0
h

o
2
dz

. (3.40)
Solving the O(1) system and applying boundary conditions gives the surface prole, ow
rate, and meniscus location
h

=
_
z

2
_
1/3
Q

1
=
1
3z

2
z

20
=
1/2
(3.41)
where = 1 +
2
3
t

. With the O(1) solution known, the O() local ow equation gives
the higher order surface prole
h
1
=
1
20
_
z

20
2
+
_
z

20
_
1/3
_
z

20
3z

21

1
20
_
_
(3.42)
Because the surface prole has been modied slightly from the zeroth order solution and
initial volume has been specied, z
2
(0) must be non-zero at O() to conserve initial vol-
ume in the corners. This consequence is captured in the O() global ow initial condition
equation 3.40
z

21
(0) = 3/5. (3.43)
The O() global ow equation is then solved to nd
z

21
=
3(3 + 5)
40
(3.44)
Substituting 3.44 into the assumed solution form 3.32 yields the O() accurate solution
for z
2
z

2
=
1/2
+
3(3 + 5)
40
(3.45)
which is plotted in gure ?? for various values of . Here it is evident that the O()
solutions are nearly indistinguishable from the O(1) solution. From 3.44 it is clear that
3
8
< z

21
<
3
5
for all . Note that for long times t

1 and t

so z
2
t
1/2
and
Q t
1/2
.
3.3.4. Extension to a General Cross Section
A cross section which does not vary in area along the section length is special because
the driving force for the corner ow, H, remains constant. By varying cross sectional area
H is made to vary with z
2
, the functional form of which is given by XX. To demonstrate
some of the possible consequences, an O(1) solution is given, still subject to the n-regular
assumptions noted above. The scales are in table XX. A general O(1) expression may
be wrote dening the ow rate as a function of meniscus location as well as an integral
equation governing meniscus location with time
Q

=
A
3/2
s
3z

2
t

=
_
3z

2
A

s
(z

2
)
1/2
dz

2
. (3.46)
Notice that a constant ow rate Q 1/3 is possible at long times if A

s
= (1 +z)
2/3
.
Quasi-steady corner ow 13
3.4. Ullage Migration
A tapering, xed volume container is shown schematically in Figure XX. The subscripts
1 and 2 refer to quantities evaluated at the trailing and leading menisci, respectively.
Similar to the tapering containers discussed for constant tip location ows, the local
z

-coordinate running parallel to any interior corner is approximated by using the global
z-coordinate, z

= z+O. The menisci are modeled as at at the location of the constant


height boundary condition. In general, the at meniscus assumption is improved for small
taper angles, 1. The boundary conditions h
1
(t, z
1
) and h
2
(t, z
2
) are calculated using
the method of (XX), and recent experiments seem to validate this assumption. Only
containers in which
A
s
= F
s
z
2
(3.47)
are considered. F
s
is simply a dimensionless geometric function, Fs = F
s
(, m). Fi-
nally, the following analysis applies only to containers with an m-regular polygonal cross
section.
3.4.1. Formulation
First it is necessary to calculate the boundary conditions at menisci locations 1 and 2.
Using ?? the bulk radius of curvature reduces to the simple function R = F
r
z where F
r
is self dened from ??. Using ?? the boundary conditions become
H
1
=
F
r
f
z
1
H
2
=
F
r
f
z
2
(3.48)
In order to derive the global equation, a mass balance is wrote at the location of each
bulk meniscus
Q
1
= (A
s
A)
dz
dt

1
(3.49)
Q
2
= (A
s
A)
dz
dt

2
. (3.50)
Combining 3.11,3.47,3.48 and 3.49 gives the global equations
Q
1
= (F
s
mF
af
)z
2
1
dz
1
dt
Q
2
= (F
s
mF
af
)z
2
2
dz
2
dt
(3.51)
where F
af
= F
a
(F
r
/f)
2
and the boundary condition is again the initial volume of gas in
the container,
V (0) =
z
2
_
z
1
(A
s
A)dz =
z
2
_
z
1
F
as
z
2
dz mF
a
z
2
_
z
1
h
2
dz (3.52)
which can be simplied slightly by carrying out the integration on the rst term on the
right hand side to give
V =
F
as
3
(z
3
1
z
3
2
) mF
a
z
2
_
z
1
h
2
dz. (3.53)
3.4.2. Scaling
The governing local and global equations with their boundary conditions are normal-
ized by the following scales
z L h
F
r
f
L t
F
as
L
mWF
af
(3.54)
14
where
L =
_
3V
F
as
_
1/3
W =
F
r
f

mu
F
i
sin
2

f
. (3.55)
The dimensionless local equation is

h
2
t

=

z

_
h
2
h

_
(3.56)
subject to h

(z
1

) = z
1

and h

(z
2

) = z
2

. The global mass balance becomes


h
2
h

z
1
= (1 )z

1
2
dz

1
dt

h
2
h

z
2
= (1 )z

2
2
dz

2
dt

(3.57)
with the volume constraint
1 = z

1
3
z

2
3
3
z

2
_
z

1
h
2
dz

. (3.58)
3.4.3. Solution
Making the quasi-steady assumption 1 the transient term in the local equation is
neglected giving the O(1) surface prole
h

0
=
_
z

20
3
+
_
z

20
3
z

10
3
z

20
z

10
__1/3
. (3.59)
The volume constraint gives the fortunate result
z

1
3
z

2
3
= 1. (3.60)
because the uid volume in the corners is neglected. The ow rates at each location can
then be found to be redundant,
Q

0
= Q

1
= Q

2
=
1
3(z

2
z

1
)
(3.61)
and nally the global mass balance gives the implicit solution for meniscus location
t

= 3/4
_
(1 +z

10
3
)
4/3
z

10
4
1
_
. (3.62)
Quasi-steady corner ow 15
4. Eect of Geometry
4.1. Draining Flows
4.2. Ullage Migration
5. Comparison to Experimental Data
5.1. Draining Flows
5.2. Ullage Migration
6. List of Solutions and Constraints
7. Summary
REFERENCES
M.M. Weislogel, Y Chen, and D Bolledulla. A better nondimensionalization scheme for slender
laminar ows: The laplacian operator scaling method. Physics of Fluids, 20(2):163 170,
2008. ISSN 0273-1177.

Das könnte Ihnen auch gefallen