. .. .

CWf
L&lJll -80-75

WC.%J+ -c-l

TITLE:

A COMPUTATIONAL f4~THoD FORFREE SURFACEyDRoDyr~AMIcs H

AUTHOR(S):

C. W. Hirt and B. D. Nichols

SUBMllTED

TO:

ASME 1980 Pressure Vessels and Piping Conference August 12-15, 1980 St. Francis Hotel, San Francisco, CA

nlacl Nut n -

~-

.-......_ .,-.-....7 ....

., IIUII Itw

UY DI:L1IIIWICI1 (II ltll~ Wlldl,

Itll

IIIIIIh\lIIII IIIImIIIIfI%

U.S. (kwrmittwm IIvirm m IHIIIII+IIm rvllimhlw tttrll,Ill 1{1dllW

Iklw

q mrIIIIaI

Iuw.w, ,IWd Iy III*ISIIIWIW

llw lIIIIIII*hwl 1111111 1111+ III IXIIIIIIIIII

(illwt,lwlll,l!l 11111

11111,11%Ill) %11,I* I (J,!; Ill I

1 hI I Ui AliIIIIIh :hwIllIf II,

lldlm ldwlllty III!,

[.dM)lilllllv

111111111 Ihal 1+

lhv

Illlt, ml\

1111( fllrk n

m Wlllh Iwlhll

, VI

11111,1lw I

IIII!II$ of

(J :; lhIIIat IIIWIII I:f FIIIIIIIJ

LOS ALAMOS SCIENTIFIC LABORATORY

Post OfficeBox 1563 Los Akmos, New Mexico 87545 An Atfinnatlve Actkm/Equal Opportunity Employer

Form No, S30R3 SI.NO, 12/?s

h

?Ozfl

nlSIRfWllOH 00CUMEHT UHUMITlh OFTHIS It {

A COMPUTATIONAL METHGC FOR FREE SURFACE HYDRODYNAMICS C. U. Hlrt and B. D. Nichols Theoretical Division, Group T-3 University of California Los Alamos Scientific Laboratory Los Alamos, NM 87545 ABSTRACT There are numerous flw tms phenomena in pressure vessel and olping sys-

that involve the dynamics of free fluid surfaces. For example, fluid

interfaces must be considered during the draining or filling of tanks, in the fomlation and collapse of vapor bubbles, and in seismically shaken vessels that are partially filled. To aid in the analysis of these types of

flow phenomena, a new technique has been developed for the computation of complicated free-surface motions. This technique is based on the concept

of d local average volume of fluid (VOF) and is emijodiedin a computer program for two-dimensional, transient fluid flow called SOLA-VOF. The basic

approach used in the V(IFtechnique is briefly described, and compared to other free-surface methods. Specific capabilities of the SCLA-VOF program

are illustrated by generic examples of bubble growth and collapse, flows of immiscible fluid mixtures, and the confinement of spilled liquids.

-2-

1.

INTRODUCTION Numerical simulations of fluid flows undergoing large deformations are

most easily performed using Eulerian representations for the flow variables. That is, the flow is computed relative to a grid of small control However, where free surfaces or other

volumes that remain fixed in sPace.

free boundaries are present, special techniques must be devised to track these surfaces in the Eulerian grid. The need for a special treatment can be readily understo~d from the following argument. After one time step of calculation all fluid elements that find themselves in a given cell of the grid must be averaged together to define cell values needed for the next time step. This averaging procedure introduces a smootning of all variaIn particular, surfaces of dlscontinuit.ysuch

tions in thu flow variables.

as free surfaces can be smoothed to the point of being unrecognizable. To overcome the numerical smoothing of Interfaces a special interface tracking method is needed that satisfies three basic requirements. First, it must provide a numerical description of the location and shape of the boundary. Second, there must.be an algorithm for advancing the boundary Finally, a scheme must be provided for imposing the

description in time.

desired boundary conditions on fluid In the surrounding computational grid. A variety of interface tracking methods satisfying the above requirements are available, but most have limitations of one sort or another. example, the d~flnition of a surface by Its height above some reference level (y
u

For

h(x,t) In two dimensions) is a simple definition requiring a min-

imum of stcred information, hut It is limited to single-valued surfaces.

-3-

Lagrangian marker particles llnked together by straight line segments can be use to define (in two dimensions) arbitrary surfaces. tage of this nthod The disadvan-

is that the intersection of surfaces becomes a diffi-

cult computational problem and such intersections are usually allowed only with special, problem dependent, logic statements. Alternatively, marker particles can be used to mark all fluid-occupied regions rather than the boundaries of the regions. In this way the inters~ction of surfaces is no

problem. Unfortunately, this method needs considerably more storage and cunputational time than the other methods because it requires several marker particles in each cell occupied by fluid and each particle must be moved every time step. To determine where a boundary Is located also requires

keeping track of all particles in a cell so that an average surface location can be computed.

In this paper anotlermethnd is described that is simple and yet incorporates all the desirable features of che other methods. This new nwth-

od, referred to as the volume of fluid (VOFJ method, is based on a function whose value is unity at any point occupied by fluid and zero elsewhere. The average value of this function, F, +n u grid cell then represents the fractional volume of the cell occupfcd by fluid. Thus, a unit value of F

Indicates the cell Is full of fluid, while a zero value indicates an empty cell. ary. Cells with F values between zero and one must then contafn a boundComputer storage for the VOF method Is a minimum at one word per

cell, which Is equivalent to the storage requirement for all other flow variables. Because It follows flulclregions rather than surfaces It works for a~bitrarlly Interscctlllg surfaces.
diS()

In addition, the F dlstribu-

..

-4-

tion used in the VOF method has all the remaining properties desired of an interface tracking scheme. Surface locations, slopes, and curvatures are

easily computed fsr the setting of boundary conditions, and the F distribution can be advanced In time by advection through the Eulerian grid. However, to avoid the type of numerical smoothing noted earlier it is necessary to use special advection algorithms. In the SOLA-VOF program, de-

scribed in the next section, a type of donor-acceptor fluxing is used to compute the advection of F. for most applications. The VOF-based program, SOLA-VOF, that is described in Sec. 11 is a general purpose solution algorithm for a wide class of fluid dynamics problems. Originally the program was developed to solve time-dependent problems involving an incompressible Navier-Stokes fluid containing free surfaces. In its present form, however, SOLA-VOF is also applicable to probThis technique is simple and works quite well

lems involving two immiscible fluids. Additionally, It has an option for including surface tension with wall adhesion, an option for limited compressibility effects, and it has an internal obstacle capability. SOLA-VOF is an easy to use program because of rlumerous automatic features. For example, it has a flexible grid generator, built-in time step controls, and some self-testing features that automatically detect numerical stability problems and correct them. A variety of sample calculations illustrating the power and usefulness of the SOLA-VOF program are presented in Sec. 111. These examples cover a

wide range of fluid phenomena associated with pressure VCSCPIS and piping systems.

-5-

JI.

THE S!3LA-VOFPROGRAM The governing differential equations are the Navier-Stokes momentum

equations [1],

~+ at

$+

v ~=

ay

--Ue+gx+v

P ax

1 ati a%+&+~_.--L
[

2X2

ajf2

x ax

Xz )]

(1)

Fluid pressure is here denoted by p.

Velocity components (u,v) are in the

Cartesian coordinate directions (x,y) or axisymmetric coordinate directions (r,z). The choice of coordinate system is controlled by the value of L, where C = O corresponds to Cartesian and E = 1 to axisymmetric geometry. Body accelerations are denoted by (gx,gy), v is the coefficierlLof kinematic viscosity, and P is the fluid density. If the fluid is to have limited compressibility the appropriate mass continuity equation is [2]

(2)

where C Is the adiabatic speed of sound in the fluid. fluids l/C2 is set to zero.

For incompressible

In the limited compressibility model density

changes are assumed to be small (say less than 10%) and the p appearing in th~ Fressure gradier~t toms In Eq. (1) can be treated as constant. [When

two immlscib fluids are present this p is an appropriate lccal mixture of e the constant p values fcr each fluid,]

-6-

Equations (1} and (2) are discretized with respect to an Eulerian grid of fixed rectangular cells. Grid cells may have variable sizes, say 6x,. i

for the ith column and dyi for the jth row, as shown schematically in Fig. 1. Dependent variables are located at the staggered grid locations indi-

cated for a typical cell In Fig. 2. The basic procedure for advancing a solution through one increment in tilw, t, consists of three steps: (1) Explicit finite difference approximations of Eq. (1) are used to compute first guesses for the new time-level velocities. In this step the initial dependent variable values, cr the values from the prevfous time-level, are used to evaluate all advective, pressure, and viscous accelerations. (2) To satisfy the continuity equation, Eq. (2), pressures are iteratively adjusted in each cell. As each pressure value is changed the veThis pressure itera-

locities dependent on this pressure are also changed.

tion is continued until Eq. (2) is satisfied to a prespeclfied level of accuracy. (3) Finally, the F function defining fluid regions is updated to give the new fluid configurtitlon. After all necessary bookkeeping adjustments are completed, including data output, this three-step process can be restarted for the n~xt time-level ca!cu?ation. At each step, of course, suitable boundary conditions must be Imposed at all boundaries. The actual finite difference approximatlon~ used in SOLA-VOF for Eqs. (1) and (2) are not a crucial part of the algorithm. That is, various approximations could bc used without affecting the basic solution procedure.

-7-

The reader Is referred to Refs. [3,4] for the particular approximations use : in the present program. This flexibility does not apply, however, to the way in which the F distribution is advanced in time. Because F is a

scalar quantity fixed in the fluid Its evolution Is governed by pure advect~on,

(3)
where r = x when & = 1 and r = 1 when E!= O. This equation Is strictly

valid only for Incompressible flow, but Is also acceptable for the limited compressibility approximation. Nmerical approximations to Eq, (3) must bc constructed with special care to avoid numerical smoothing of the F distribution. There are several ways to do this. SOLA-VOF employs a type of do-

nor-acceptor fluxing using the fact that F values should be either one or zero. The basic idea can be grasped by considering the amount of F to be The total

fluxed across the right boundary of a cell during one time step. volume of both fluid @

void crossing the boundary, per unit cross-sec-

tional area, Is V = u6t, where u Is the normal velocity at the boundary. The sign of u determines which cell Is loslng F (the donor) and whlsh is gaining F (the acceptor). The amount of F crossing the boundary depends on how F Is distributed in the donor cell. Uhen the flux is primarily In the

direction normal to the F s~rface the fractional area of the flux boundary across which F Is flowlr,gis determined by the acceptor cell F value. the flux is prlmarlly tangent to the surface the donor cell F value Is used. In both cases the amount of F flcjxedIs computed as the product of Uhen

the cross-sectional area of the flux boundary times hF where,

.. ....

-8-

F=

MIN{FADIVI + CF , FD 6XD}

and where CF= MAX{ (l.O - FAO)IVI - i.O - FD) 3XD , 0.0} . (4)

Subscripts denote acceptor (A) and donor (D) cell values. The double subscript (AD) is equal to A when the flux is normal to the free boundary and equal to D otherwise. The MINfeature lnEq. fluxed than is available in the donor cell. (4) prevents more F being , The MAX feature accounts for

an additional flux of F if more than the amount of void volume available in the donor cell is f.uxed. typical cases. Figure 3 illustrates these features for several

The fluid is assumed distributed in the donor and acceptor

cells as shown depending on the orientation of the surface normal with respect to the flux direction.

In Fig. 3a the donor cell, acceptor tell, and

the flux volume are defined. Then Fig. 3b il!listrates situation in which a the donor cell value of F is used to define the fractional area of the flux boundary open for fluxing F. In case c of Fig. 3 the acceptor cell value In this case all the F

of F has been used to define the fractional area.

region in the donor cell is fluxed, but it is less than the total flux possible, which illustrates the use of theF!IN test in Eq. (4). Finally, in Fig. 3d, more F than the amount determined by the acceptor cell defined area must be fluxed. The extra flux contribution to defined inEq. (4). F is the quantity CF

-9-

The choice of the acceptor or donor cell F value to define a flux area, which depends on the orientation of the free boundary, is a feature not used in other schemes of this type. It is essential to do this, how-

ever, otherw~se boundaries advecting more or less parallel to themselves will develop step ~rregularities. Additional details of the SOL~,-VOFprogra~ ~elating to boundary conditions, numerical stability requirements, etc., can be found in Refs. [3,4]. The best way tn assess the strengths and weaknesses of the SOLA-VOF algorithm is to examine the calculations tt can perform. This is done in the next section, where several applications are used to illustrate its power for a wide variety of difficult problems.

III.

SAMPLE APPLICATICNS Pressure vessels and piping systems are subject to many kinds of com-

plex flw

phenomena. The following examples have been chosen to illustrate

how some of these phenunena can be addressed with the new SOLA-VOF program. These ex:;,,Plesover problems involving the growth and collapse of vapor c bubbles, problems associated with mixtures of intnisciblefluids. and problems involving extreme deformations of free surface dominated flows. J. Bubble Gynamics in pressurized systems for liquid transport it is sometimes possible for vapor bubbles to form. Under most circumstances bubbles are undesira-

ble as their growth or collapse can result in significant pressure fluctuations and local material damage. The theoretical prediction of bubble dy-

namics is complicated by the generally large free surface deformations lnvolved. his, then, is an excellent area where the capabilities of the

-1o-

SOLA-VOF program can be put to use.

For purposes of illustration we shall

consider what happens when steam is forced through a pipe submerged in a

poul of water. The pipe is located axisymmetrically in a cylindrical vessel approximately half filled with water (see Fig. 3). Experimental stud-

ies [5 ] indicate that when sufficient steam IS injected into the pipe bubbles may repeatedly form and collapse at the end of the pipe causing large pressurls transients to be generated in the water pool. Presumably the in-

crease in liquid surface area and stirring associated with the formation of a bubble increases the condensation of steam to the point where it can no longer support the bubble. Mhen this happens the bubble collapses and water rushes back into the pipe until sufficient steam pressure is again bui]t up to generate a new bubLle. To simplify the problem we shall not attempt to model all the processes associated with actual steam condensation, but shall use a simple prescribed pressure history for the steam. In particular, the steam pressure

is approximately linearly increased until a bubble has been generated then it is rapidly reduced to the saturation pr~ssure of the water in the pool. A one millisecond time interval was arbitrarily chosen for the depress~lrization time. The time at which the depressurization is started determines

the size of the bubble transient. This crude model approximates what would happen with a more detailed condensation model in which condensation proceeds more rapidly than the inertial response time for the bubble. 1P any

case, It Is sufficient to illustrate how the SOLA-VOF program can be used to study the complete history of bubble birth and death.

-11-

Starting from an ?nitial vapor pressure of 1.1 psia in the pipe (0.146 ft 1.D.) th~ pressure is increased to about 3.16 psia over a period of 190 ms then is ramped down in 1 ms to 0.38 psia. A sequence of computed The earliest

velocity plots and fluid configurations are shown in Fig. 4.

time shown corresponds to the time at which the vapor pressure is ramped down to the liquid saturation pressure. Because of Inertia in the liquid the bubble continues to grow (Fig. 4c). When the bubble begins to collapse

it does so asymmetrically, pulling liquid in from the top of the pool. This causes a detached bubble to form shortly after Fig. 4d, which then disappears some time between Figs. 4e and 4f. rapidly up the pipe in the final frame. Pressures computed at the center of the vessel floor are shown in Fig. 5. The initial rapid increase in floor pressure occurs shortly after the At 0.19 s the pressure drops because of A relatively violent pressure U~ter is seen to be moving

pipe has been Cledred of water.

the decrease in vapor pressure at that time.

transient develops when water reenters the pipe end and the detached bubble collapses. This transient, denoted by the dashed line in Fig. 5, is shown inan expanded scale in Fig. 6. To obtain this result the fluid must be

treated as a compressible medium because the pressure transients have characteristic times short compared to the time needed for acoustic waves to travel across the pool. Except for this short, violent transient the water can be treated as incompressible, but to correctly estimate the pressure pulse generated by the final bubble collapse requires compressibility. The limited compressibility model available in the SOLA-VOF program provides this capability.

-12-

B.

r~is~ible Fluids , mixture of oil and water provides an excellent example of a two-

fluid system often encountered in practical situations. Because of their slightly differing densities and the action of interracial surface tension forces the mixture behaves dynamically quite different than either fluid separately. Using the two fluid and surface tension options in the SOLA-VOF program a variety of interesting mixture problems can be investigated. To illustrate, Fig. 7 presents results from a calculation of the passege of a liquid drop through a constriction in a tube. The drop has a density equal

to 9/10 oi the density of the surrounding fluid. Surface tension at the interface between the two fluids is such that the Weber n~mber (pV2r/o) is equal to 0.192, based on the drop radius and average flow rate through the tube. This nwans that surface tensio~ forces are more significant than In this example viscous for~es are also relatively

those of inertia.

strong for the Reynolds number (Vr/v) was chosen to be 1.25, and it was assumed that both fluids have the same kinematic viscosity. Flow entering the flow channel is uniform, implying that the constriction is near the channel entrance where boundary layers have had little tine to develop. this were an oil-water mixture, it wuld (r= 3x lC-4 cm) forced rapidly (V
s

If

correspond to a small oil drop

46.1 cm/s) through a hole in 3 thin

plate. To force the drop through the constriction requirw extra wrk to de-

form the drop against its surface tension forces, Figs. 7a-7c. Much of this work, however, is recovered as the drop emerges from the constriction, Figs. 7d-7f.

-13-

C.

Larqe Distortion Dynamics The advantage of formulating SOLA-VOF in terms of an Euleriar repre-

sentation is its ability to treat flows undergoing exireme deformations. To illustrde this capability consider what happens when a tank containing fluid collapses. Suppase a dike is to te constructed around tbe tank to contatn the spilled fluid. The problem is how high to build the dike. A

specific example is illustrated in Fig. 8.

The fluid is initially a circu-

lar column having a heig$t equal to its diameter. The dike is a low axisymmetric wail whose radius has been arbitrarily chosen to be 24 column radii. As the column col?apses fluid rushes radially outward along the ground. Upon striking the dike the leading edge of the fluid is deflected upward, but if the dike isnt high enough it retains sufficient radial momentum to splash down outside the dike. This is seen t.obe the situation

in Fig. R where a signficant amount of fluid has been lost from the dike region. When the dike h~;qht is increased to approximat~ly 2 times the height shown in Fig. 8 additonal calculations indicate that virtually all

the fluid is contained within the dike. Although this is a conceptually simple problem, it is obviously one that involves highly complicated free surface dynamics. Nevertheless, the Other

SOLA-VOF program does a remarkable job in representing the flow.

variations involving different initial fluid and obstacle configurations are easily imagined. Of course, SOLA-VOF could also be used for similar problems involving two immiscible fluids with or without interracial surface tcn5ion. !t is this flexibility, in fact, that makes the SOLA-iOF With thoughtful use it provides a means of

program such a powerful tool.

investigating many previously intractable prublems associated with pressurized fluid systems.

-14-

ACKNOWLEIXMENTS We wish to thank R. S. Hotchkiss for his efforts in adding the surface tension capability to the SOLA-VOF program and for running the second example problem. This work was supported by the Electric Power Research Institute under contract RP-965-3.

-15-

REFERENCES 1, 2. L. D. Landau and E. M. Lifshitz, Fluid Mechanics, Pergamon Press, London, 1959. C. W. Hirt and B. D. Nichols, Adding Limited Compressibility to Incompressible Hydrocodes, J. Comp. Phys. 33, 1979.

3.
4.

C. U. Hirt and B. D. Nichols, Yolume of Fluid (VOF) Method for the

Dynamics of Free Boundaries, submitted to J. Comp. Phys., 1979. B. Il.Nichols, C. W. Hirt, and R. S. Hotchkiss, SOLA-VOF: A Solution Algorithm for Transient Fluid Flow with PlultipleFree Boundaries, I.os Alamos Scientific Laboratory report in preparation. S. B, Andeen and J. S. Marks, A~alysis atidTesting of Steam Chugging in Pressure Systems, Electric Power Research Institute report NP-908, 1978.

5.

Ui++,j

i, j-*

Fig. 1.

(A) Mesh setup showing variable cel Ca,lability. (II) Locat on of dependent variables In a typical mesh cell.

+Udl J : ~, ..-
-.

[2-

-----..-. .. ___ .. /,t, m

.~NoR ;

ACCFJTOR -. . .....

a ()

.. . -.. .4
m <~;[,;u WI
;;(&,;>

(h) AD--D

..

..- .. .._,

., (;,;;;.

11,,

.::(

j;l, /~jl;, ):;

$

..

!. $/1;//, /,,;./ /~,.; , ,,,!/,,, . /

.1AU4LL .-

(c!)

AD-A

Fig. 2.

Schematic used in calculating 1 advectlon. llonor-acc~ntor cnlls arc deflncflIn (a) where dashed line indicates left slhc of flux volumc. Cross-hatched regions In (b-d) are the amounts of 1ativecteci.

DIMENS1ONS In Feet

3.948 -0.146 0.167

ID
OD

0m833 1 1.729

.,. .. --.., ... ,,.,. ...

!,..... .J . .. ... ---.

. . . ./

.-

..- ..-

.-

A
Fig. 3.

Schem~tlc of stem Injection apparatus USCCIby S1!1lntcrnat, ional, Plus signs In (a) Indicate mesh CP1l centers that Initially contain fluid. Illmcnsionsof apparatus aro shown In (h).

T
,!

,,
I

i ii
I

i:

1:
1 I 1:
1, ~,

l!
/

I
I
I

!
I
1

I
1

I
I

,.

I I

I i

1:

I
;1 !;

Ii ,,

, 1

I d
...-

I
n-m= . ... ... ... . ..,.. . ..... ,.. ,.. . ....m .,. . ..,,, ,. .-. . ..,.. . ...,, .,. ..-

I I ,..,.,.. ,-,-,..,.,. .,.. ,..,., ,,, ,.. :I:A ,.. ,.. . -,.. -

k=
!
!

..-

.,...*..--

l , -,, , ,.. ,.. ,., :::= ,.. . ,-,., ,.. ,.. ....-

r: -. k-. -.

.. -. k
... . 4

,.. ,1
II --..

L..

t == } .

.
-----

...,,

.-

. ---

,--

, --

:=.=::: ., .. . ... .. .. . .. .. ,..

,.. . . .. . .. . ,.. . ... ,.. . ,.. .

,.. ..

4=. . .

,.. ,M . . ....-. -

,..

. .. . . .. . ,.

...-

... ,.. i

,.. -. ... , ,.. . ... ...

... . .,. ... . ... . ... ,

,.. -,. ... 1

., . .. .,.

... ., .: ... .. t-.,.

.-.. .. . I
. .. . .. ... . . . ., -.

,..

,.. . f

... ,.. .

... .

Fig, 4.

vectors and fluid conflqurations at selected times ~ftcr Initiation of pressllrlzation. Times In seconds are (a) 0,19, (b) 0.21, (c) 0.22, (d) 0.24, (c) 0.75, and (f) 0.30.

Veloclty

*PM

PRESSURE

JL
TE 1s 0.3 Tkm h) 0.4 0.5

TI

-o
Fig. 5.

0. I

0.2

Floor pressure history computed during bubble growth and collapse, Dashed line region shown in exp]nded time scale ~n Fig, 6.

120

FLOOR PRESSU?E FWING

EWBLE

COLLAPSE

10

. .~.

15

Q256

02S7
Thfu (d

.-&~
0.2s9 0.260

Fig. (i.

Computed floor pr~ssure on an expanded time scale follGwing huhble collapse.

,Fr,

11,
!,

,,, t, ,,.,. #, ..,, ..,, .,,, .,, ).,,]!( t,. ,., O )4 ,, ,.,,,, ,,.,, ,,,,,. .,.,,,, 0
,,, ,,, ,,, i,,
,,0, .,, ., ,,, ,,, ,,, ,,,

,, ,. ,, ,,
,, ,.

,! ,,,
,,

..,!,. ,,

!,,

.,, .,,,,,,,,, ,,, ,.

.!,,,

,,
,,,

.,.
,!? i!.
0,. .,,

,,,

,4,!,,,,,,

,It ,..

i..!.

,,, ,,,

,1,,,9 .,1., !,!,.

,41 ,,, ,,, ,,,

*I, ,!. ,,, .!, .,,

.I. !,,
,,. ,,

i,, ,.

i,., .,,,,
,,,

,Ir.

lt ,!

.,,

,,

,,, ,,, .,, .,, ,,,

,,

,,, ,,, ,, ,. ,,, ,,, ,, .,, ,,, .,

,1,... ,0,,,, ,,

,,,

,,,

,,,

,,0

.,.,,,,,,,

,.,

.,,,

,1.

,, ,, ,.

,,l.

,,,

,,

,o, ,,,
,! ,, ,,,

n,#,

,,, ,,, O ,,, m ,ict,

,,, ,,

cto ,,, ,,, ,.,

.,,

,!!!

1,,,,,,,,

,,,

...

.,, ,,,

,,, ,,, ,,, , 1#1,

,,, ,,,

,, ,1,,,,, .,,,,,,,,

1,, ,,,

,,!

,,,

C,

I,.

.,,

,,1

;,,,

,1 ,0 ,,

,1,,,. ,0,,,. .i

.0,,

...

,,0, ,,,

,,

,,!! ,,,1

,1

c,.

,,.

,:

:;;
,)

,,,

,,,
,,, ,,

s,,
,,,

,,,

,1 ,,

,,,

,,,

.,

.#m

.9,

,..

,1

,0,,,.

,,,

,1

,,,

,rtclt,

!,.

,,

l!r, ,#, ,., ,,, ,, ,1,,,. ,,, ,!, ,,, ,, ,OC, -! ,,,

c c ,,
I

rlll *,, !,

::\ 1
i

i~o,,., Itlrn Ill #.,.. 11 l,,..

If,

I #,c. .~ ,,, ,.. I

a.,
,,,

,,, ,,,

, .

,,,,,.,.,, . I,,,,,,, I ,,
,, .,,,,. I
,,
,, .,,,.,1 .,, ,0. i!
i..!

!!.

,,!,

!.

..0

.!

I
!

,!

...!!.!

,, ,,
,.,

,,,

,.

#.1.

..i.

.,,,,,,,
,,,

.,1,,,,,

,tm.

,,,

,,,

.,

r!.,

.,,.

l,o, ,,,

lm*,

.o
,,,

,,, ,,,
.,,,,

,,,
,,!

..1! ,11,

1.!..! ,,t
,,,

i,,
,,,

,:

,,,

#i,,,

I
,,, .,. .,

,1,

i t

,,
,,, ,,,

,..

!,,.

,,, ,,,

,,1111 ,.lllil

,,, ,,.

,,

,., ,, . ...* .,, .1

,1 :::

,,!1! . ...111
...

1,1,!, ,,, 1,

.,..

1,

..,.

11,

,..

... .. i,:

y:
I i

; :

1.1 ,.

,0,,,,

!,

,tioll ,,,

,,,

Bi ,, ,,

:;

,,,

l,,

0,, ,,, ,,,

.,,,, ,,, ,,,

,,, ,,

,,0,

,,,

,,

,,

...

,;! +41 .,111

,,, .,,
,.

: ,:~ \\: .%
1, .,.

[1
. ... n,
,..
,. ,*. . ,,,,. , .,, ,, ,,,

1,. f,. l,. t.

, .

1
I
I

i I
I

111
#nil ,4, ,1,

.r, ,i l#aall#l

ii

,,,
l,,

.,c
,,,

#o,

,,,

,.,1#1

,4,,,,,,, ,,,

,,, ,,

,,,

,,, ,0

,,,

,!,

,,,

,,

,,10

,,,

,,,

,,,

,0, ,,, ,,, ,

,,,

,,,

,,, ,,,

.,0,,

,,,

,,,

.,,

,,

,0!,

,,,

,,,

,,

,,, ,,, ,,

,,,

.,,.,, (,,, ,,, ,,,n ,,, , ,, , ;b#*, #,m, ,*mm#.

1111

,,, II

,,,

,,,

,,

,.,

,.,

.!,,,,,,,,

,,,

.,,

,#,

,, ,,,
,,,

,,,
,,

l,o, ,,,
.,1,

,,,

,#,

,,,

,,,

,,,

, , ,

!,,

,,

,1,,,,

,,!,

I!

*,,....

11.

!.

,1,0 ,!

.,,

,11,,,,01 ,,1,

,,,

.,,

,, ,llil ,*,

...

.,

I
.,!
,,! .,, !,,

!,

,.,,,,,,,, ,,,,,, . ,,,,,,, $u )

,1/ ##l Ill>. 1111

,1;

,,, ,,,

,,

,1,.,1!

,,1,

.,,,,,,,.

, .,,,,,,,,
! .,,,,,,,,, .,, ,.,
,,, ,,,

,., ,.i
!,
!.

,,,

,,,

,!,

,!
,,,

..!

., 4 ./1 J . ,/ ,., ..0/ .,, ,1 .,,1 ,,, ,, ,11 .,, ,!1 ,,, ,,, ,,, ,,. ,,, ,,, .,, ,,, .,, rl, l ,., !.. ,,, ,,, ,,, ,,, ( ... ,,, , ,,, ,., ,,

.;

{. ,,. \,.

0,.

,1,

.,.

111 ,,, ,,, ,,

89. ,1,,.1.. ,,, ,.. ,,,

1 ,1,,,.,.

,,,

.
,

!,,,.,,,

,,,

.,,,,!

1,1.

!ll,

!.

i,lici!o ,, !,,,,,
,,,

,,,

,,,

-.

.,,8

,,,

.!,

1!

,,,

,,,

,,,

.,.,!,,,

,.,

,1,,,!,,,,

,,,

..1?,.,,,, ,. !,,.,, ,,,

,,,

,,,

,,,

! ,1,

.,,,,,,,,., ,,t,

;,;,;,

,,,
,,, ,,, .,, !,, ,,, ,,,

,,,
,,, ,,, ,,, ,,, ,,, ,,, ,,1

tot ,,,
,,, ,1,,, ,, ,1, ,,, ,,, ,, .,,

,,,
!, ,,1, . ,,, ,,,

,,,
,,, ,., ,,,

,,,
,,, ,

r,,
,, ,, ,,, ,,, ,,,

,,,
,,, ,,, ,,,

,
, .

,,,
,,,1 ,,, ,0! ,., ,., ,., ,.* ,,,

::;,, ., [n .,,., Mi7k ,,, ,. i,, ,,

I:w,-,m,::m ,,, ,,, ,,, ,,, ,,,
Ito, l,lt, 11 ,111,18.,,,, ,.1, ,,, ,!, ,,, ,,, ,,, ,1!!,,

,,,

,,,

,,,

J
,,, ,,, ,1. ,,,
,,,1

,,,

,1

hy
8 ,,,
,,, ,,

!,!,!.

,,

,,,

,0,,,1,1,,,
,,,

ml
!,

,,,

,)

!,,

,),

.,,,!

,,,

,,,

,,

,.,

!,1,

,,,

,,,

,,,

,,

.,,,

,,,

,,,

.,,,,,

,)1!

1$1

,,,

,,,

,,,

,,,

,,, ,,

,,,

,,,

,,,

,,,

* ,

# # # #
1

,8

i: ,: / \: \i,, \8
tu, tt, ,,, h,, ,, ,. ,,
,.C ,,, ,,J 1,, ,,, 1,, Ile, , ,1 ,,,0 o,l, a,nlt, a,l, ,,9, ,,, ,,, D,l, *,,, ,,, ,,,

t,, ),
1 ,4,! ##mm ##,
0,,

.,!

,1,.,,,, ,,,

,,,

1,,

,,0 1,, ,,, ,,,

,1,,, ,, ,,

,,

,1018 ,,, ,1, ,,, ,#

,.

,
,

II, t,,

,.

J
l,, 1,, ,,

, ,

1,,

)*

;!,:; ,,,.
14. .

(
o,
,, ,,,

9 :
9

0 I.m.
181, 1,, -1,,10 991,4, II

,,,
,,, ,,,

,,,
,,, ,,,

a ea. m 180
,,, ,,, ,

,,, ,,,
,,,111,31

I*, ,11,,,, ,,. s,! ,,, ,,, 9#1 111 ,

II
!1 In ,

l#*h4, , ,0,,,,,,,,,0, 119,91aol

I#o, nb,

,*el*lllb 19am,,
@#,

901

II II

981000 *. .,,, lon oa,

11#1,1,
it, ,,, i,. i,i, ,.,

.#oa
.!l,
,1,0

9,11
.,!

s,o
i,

.,... 180 ,11*. ,,,

,1, ,11!6, I

,,, ,,, ,,, ,,, ,, ,,, ,

Il#ltl*lll#l

n,,

al, ,,18

010111,.1118011 DI, ,. ,,, , .,11,,1,011 ,,, ,,, ,,,

,,, ,,,

,..
,.,

,,,
,,1

##l ,,,
,00,,,0, ,,, ,,, ,,, ,,,

011
, ,,, *#b

,s,

,,,
,,, Iltilltblll

,,,

loll,

11,108,1+018,11 #l#llnl# ,., #,, 1#, 1,#, a,a, 8181 *I, II II

111
111, ba* Ian, II Ill

IB

Fig.

7.

Distortion of an Immiscible fluid drop passing thliougha cotlstrlctlonm

l.,

.,

2.

7
I

.,, .,

Fly. 8.

Collapse of a cylindrical column of fluid and its Dartfal confinement by a low dike. Cylindrical axis is at left edge of figures.