A Cascade of Structure in a Drop Falling from a Faucet

Author(s): X. D. Shi, Michael P. Brenner and Sidney R. Nagel

Source: Science, New Series, Vol. 265, No. 5169 (Jul. 8, 1994), pp. 219-222
Published by: American Association for the Advancement of Science
Stable URL: http://www.jstor.org/stable/2884174
Accessed: 19-09-2016 23:42 UTC

JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted
digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about
JSTOR, please contact [email protected].

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at
http://about.jstor.org/terms

American Association for the Advancement of Science is collaborating with JSTOR to digitize,
preserve and extend access to Science

This content downloaded from 129.59.95.115 on Mon, 19 Sep 2016 23:42:14 UTC
All use subject to http://about.jstor.org/terms
 REPORTS

low viscosities. On the other hand, Pumir,

A Cascade of Structure in a
Siggia, and co-workers (12) have analyzed
Drop Falling from a Faucet several mathematical models of singulari-
ties with a high Reynolds number that
showed nonsteady corrections to Eq. 1.
X. D. Shi, Michael P. Brenner, Sidney R. Nagel
However, to the best of our knowledge
A drop falling from a faucet is a common example of a mass fissioning into two or more those models did not have an experimen-
pieces. The shape of the liquid in this situation has been investigated by both experiment tal realization. For the case of the drip-
and computer simulation. As the viscosity of the liquid is varied, the shape of the drop ping faucet, we will show that both views
changes dramatically. Near the point of breakup, viscous drops develop long necks that have some validity: the similarity solution
then spawn a series of smaller necks with ever thinner diameters. Simulations indicate that provides a basis for the underlying struc-
this repeated formation of necks can proceed ad infinitum whenever a small but finite ture, but there are time-dependent correc-
amount of noise is present in the experiment. In this situation, the dynamical singularity tions to this solution that alter the shape
occurring when a drop fissions is characterized by a rough interface. dramatically.
There are three independent length
scales that characterize the hydrodynamics
of the dripping faucet (4): (i) the diameter
What happens when liquid drips from a where h describes the radius of the drop as of the nozzle D; (ii) the capillary length
faucet? As it falls, its topology changes from a function of the vertical coordinate z and
a single mass of fluid into two or more time t; f(t) is an arbitrary function of time; ly z1/2
drops. This common phenomenon is one of P is a constant; and z0 is the position
4 pg)
the simplest hydrodynamic examples of a where the droplet breaks. This scaling
singularity (1) in which physical quantities hypothesis has worked well in describing which gives the balance between surface
become discontinuous in a finite time. other types of singularities, ranging from tension, -y, and the gravitational force pg,
Here, we investigate the shape of this sin- critical phenomena (8) to droplet breakup where p is the fluid density and g is the
gularity for fluids of varying viscosity drip- in a two-dimensional Hele-Shaw cell (9, acceleration of gravity; and (iii) the viscous
ping through air from a cylindrical nozzle.10). Also, for our problem here Eggers length scale
Scientific studies of this system date back (11) to constructed a similarity solution that I2
Lord Rayleigh's stability analysis of a liquid showed good agreement with numerical
cylinder (2) and Plateau's analysis of a solutions to the Navier-Stokes equation at P'Y
hanging pendant droplet (3). More recent
work has begun to address the shape of the
interface near the singularity. Eggers and A B E Fig. 1. The shape close to the
_U _ _time of breakup
Dupont (4) simulated falling droplets by
y_ _ different viscosities. The liquids,
means of the Navier-Stokes equations and
with rq = 10-2 P (A), 1 0 - P (B), 1
showed that their solutions agreed with the P (C), 2 P (D), and 12 P (E), were
experimental shapes of water photographed allowed to drain slowly through a
by Peregrine et al. (5). glass tube with a nozzle diameter
Although our study focuses on a specific D of 1.5 mm. (A) and (E) show
experimental system, the dynamics near a pure water and pure glycerol, re-
singularity should be insensitive to changes spectively. We used an 80-mm
Hasselblad lens attached to a
in the initial conditions and external forcing
bellows and a still camera; the
within a wide range of parameters. Thus, the
drop was illuminated from behind
shape of the interface near the breakup point
by a fast (5 p.s) flash from a
should not depend on whether the drop is strobe (EG&G model MVS 2601,
falling in a gravitational field or being pulled Salem, Massachusetts) that was
apart by shear forces (6). This expectation triggered with a variable delay
arises from the realization that as the inter- from the time the drop intersected
face breaks, its thickness must eventually C D & X 1 | 0 a laser beam incident on a photo-
become much smaller than any other length y ' X diode.
in the problem (until microscopic atomic
scales are reached). In this regime, the only
length (7) that can affect the dynamics is the
thickness of the fluid itself, so that the
simplest assumption for the dynamics is self-
similarity-that is, the shape near the sin-
gularity changes in time only by a change in
scale. The mathematical definition of a sim-
ilarity solution is

h(z,t) = f(t)H[Zf(tO] ( )

James Franck Institute and Department of Physics,

University of Chicago, 5640 South Ellis Avenue, Chi-
cago, IL 60637, USA.

SCIENCE * VOL. 265 * 8 JULY 1994 219

This content downloaded from 129.59.95.115 on Mon, 19 Sep 2016 23:42:14 UTC
All use subject to http://about.jstor.org/terms
where 'r is the viscosity. For water (5), L, with different viscosities draining slowly
brass plate) to the main part of the drop. A
100 A. For that system, it is not pos- from a nozzle of diameter D = 1.5 mm varycloseup photograph (Fig. 2B) gives a clearer
sible to photograph the shape of the inter- view of the region just above the main drop
as the viscosity is varied (14) (Fig. 1). As
face where the thickness is asymptotically the value of Xq increases, the neck of the where the neck decreases in width, forming
smaller than L, because current optical drop elongates and forms structures not a secondary neck. At a slightly later time
techniques cannot resolve objects much observable in dripping water. Herein, we (Fig. 2C), this secondary neck suffers an
smaller than 1 ,um. concentrate our attention on the structure instability and forms a third neck. We
To find the asymptotic shape of the of a drop of liquid with intermediate viscos-estimate the thickness of the smallest neck
singularity, we studied droplet snap-off as ity
a and report a combination of photo- in Fig. 2C to be about 5 ,um. Through
function of viscosity. By mixing glycerol graphic studies of the singularity and nu- high-speed photography, we have verified
into water, we can increase the viscosity by merical simulations of the Navier-Stokes that both the second neck and the third
a factor of 103 so that L, can be increased equations. neck develop by a similar process: an initial
by 106 while the surface tension is not As a drop of an 85% glycerol and 15% thinning near the drop is followed by a
varied by more than 15% (13). We thor- water mixture [1 = 1 poise (P), where 1 P rapid extension of the neck upward away
oughly mixed the water and glycerol, taking = 1 dyne-s cm-2, and L = 0.13 mmi from the drop. This leads us to suspect that
care to eliminate air bubbles. The shapes breaks (Fig. 2), it forms a long thin neck the repeated
of formation of such necks could
close to the breakup point of five liquids fluid connecting the nozzle exit planehappen (a flatindefinitely. Our simulations de-
scribed below can produce such a phenom-
enon ad infinitum (with as many as seven
A B C successive necks at the resolution limit) as
long as a noise source is present. On the
other hand, in the absence of noise, our
simulations indicate that near the singular-
ity the interface is smooth. We argue that
noise, which is always present in the exper-
iment, invariably leads to a singularity with
complex structure.
Numerical simulations of the Navier-
Stokes equations have provided a well-

~~~~~~~~~~~ I_
developed tool for reproducing the large-
scale structures and velocities of fluid flows.
In addition, they can investigate structure
on finer scales than can be seen in the
figures here. Our simulations used the equa-
tions of Eggers and Dupont (4), which are
approximate versions of the Navier-Stokes
equations (for a Newtonian fluid) that ne-
glect non-axisymmetric effects (4, 15, 16):

(av av\ 3 da /av\ dp

P at az h2 aZ az/ dz
(2)
ah 2 a(h2v)
_ + - =0 (3)
at dz

Iah 2 -3/2 a2h

_ + kaz J z2 (4)
where h(z) is the radius of the fluid neck a
distance z from the nozzle, and v(z) is the
fluid velocity. Equation 2 expresses force
Fig. 2. A photograph of a drop of a glycerol in water mixture (85 weight %). (A) The initial stages of balance: the acceleration is due to viscous
the breakup as the droplet separates from the nozzle by a long neck. A secondary neck is visible stresses, a pressure gradient, and gravity;
just above the main drop. (B) A magnified view of the region near the breakup point (obtained from Eq. 3 expresses local mass conservation;
a different photograph). The thickest region in this picture corresponds to the long neck of (A). Onlyand Eq. 4 specifies that the pressure jump
one secondary neck is visible. (C) The same region as in (B) but at a slightly later time. The secondacross the interface is determined by its
neck and a well-developed third neck are clearly visible in this picture. Using the Eggers similarity
mean curvature. These equations were
solution, we estimate that (B) and (C) are 2 x 10O3 and 2 x 10Q4 s, respectively, before the time
solved with a modified version of a finite-
of breakup. The photographs were taken as in Fig. 1, except with a 35-mm lens. Because the
clearest demonstration of the structure of the singularity was obtained with the use of the largest difference scheme supplied by Eggers and
drop, we found it advantageous to replace the glass nozzle by a brass plate riddled with numerous Dupont (4). An adaptive grid, similar to
small holes on which the liquid would slowly collect before falling. The diameter of the drop at that
theof Bertozzi et al. (17), allowed high
plate was approximately 20 mm. resolution of the singularity. At prescribed

220 SCIENCE * VOL. 265 * 8 JULY 1994

This content downloaded from 129.59.95.115 on Mon, 19 Sep 2016 23:42:14 UTC
All use subject to http://about.jstor.org/terms
 m REPORTS

time intervals, the grid was adjusted. of necks that themselves form necks downshape of the interface near this singularity is
In Fig. 3, we present the results of to the smallest level detectable. therefore rough, with a time-dependent
numerical simulations for a liquid with a Noise fluctuations in the shape of the structure in contrast to the scaling hypoth-
viscosity of 1 P falling from a nozzle with a interface become more effective at creating esis of Eq. 1 (20). The description of this
diameter D = 4.5 mm. Figure 3A shows the additional structures as the thickness of the singularity is more complicated, with new
drop profile approximately 10-8 s before interface decreases (19). The amount of structures evolving progressively more fre-
rupture. The shape of the interface near the noise introduced in the simulations above quently in time (21). In a real fluid, such
drop is similar to that shown in Fig. 2A corresponds to angstrom-size fluctuations of repeated formation of necks can occur only
(18). In a magnification of the region of the the interface when the thickness of a neck a finite number of times. Once the neck
secondary neck (Fig. 3B), the interface is still on the order of 1 ,im. Thus, even becomes thin enough, a hydrodynamic de-
thins near the main drop. This thinning thermal fluctuations may be sufficient to scription breaks down as long-range van der
continues smoothly to arbitrarily small produce these effects. The only caveat is Waals forces and molecular effects become
scales. Near the singularity, the interfacial that because this is a strongly nonequilib- important. In this regime, simulators have
shape is well approximated by the similarity rium situation, we do not understand the used molecular dynamics techniques (22,
solution proposed by Eggers and-first seen by spectrum of the noise in the experiment. 23) to study such a breakup,
Eggers and Dupont (4). However, these However, there are many sources of noise, In our experiments, after the drop de-
simulations show no evidence of the repeat- ranging from surfactants on the interface, tached from the bottom of the neck a satel-
ed formation of necks, as shown in Fig. 2. to pressure fluctuations in the nozzle, to air lite droplet formed from the long neck itself
To induce the repeated formation of currents in the room, shear heating in the and separated from the fluid near the nozzle.
such necks in the simulations, it was -nec- liquid, and thermal fluctuations (for exam- The singularity in this case had a structure
essary to introduce a noise source explicitly. ple, excitation of capillary waves). In our similar to the one described above (where
We did this by placing fluctuations, on the experiment, we might also expect noise the drop separated from the bottom of the
order of 0-4 the minimum thickness of the generated from fluctuations in the concen-neck), and in both cases the thinnest necks
neck, on the interface during the simula- tration of the two liquids as well as from were attached to a large, almost spherical
tion. Both the temporal and spatial separa- evaporation from the surface. On repeating mass of fluid. However, for the upper singu-
tion between subsequent fluctuations were the experiment, we found that the details of larity, the diameters of the necks increased
small compared to the characteristic scales the interfacial shape can vary: the precise with distance from the nozzle (in the direc-
of the Eggers similarity solution. Our results position of the necks can change from drop tion of gravity), which is opposite to what
(10-8 s before break-off) show a magnifica- to drop and blobs can form on the interface happens in the situation at the bottom of the
tion of the secondary neck (Fig. 4A) and at irregular intervals. Nevertheless, the re- neck where the diameters decreased (against
the highest resolution attainable with our peated formation of necks appears to be the direction of gravity). The form of the
present code (Fig. 4B). The smallest neck robust. (Our simulations with a noise singularity thus appears to be inVariant to
shown in Fig. 4 is only 2 A wide. At each source also produce such variations when the direction of the external forcing.
level, before a new neck starts to form, the different realizations of the noise are used.) We have shown that the singularity
shape of the interface was approximated by Taken together, our results show that found in the fission of a viscous drop of
the similarity solution of Eggers. This result the singularity in the fission of a drop is liquid is different from what had been ex-
demonstrates that, at the point of breaking, characterized by the repeated formation of pected from earlier observations of less vis-
fission proceeds by means of the formation necks, which may proceed ad infinitum in cous fluids. Instead of a smooth interface,
the presence of a small noise source. The we found (to the limit of our resolution) a
rough interface with necks "growing" other
necks. The existence of this phenomenon
A hID B hID(x 10-2)
-1 0 1 -7.5 Q 5 A hID(x103) B hID(x104) appears to depend on the presence of
-4 0 4 -2.5 0.0 experimental noise. However, it should be
pointed out that because our simulations
2.2 used only an approximation of the Navier-
- ~~~~2.2-
2.46 4 Stokes equations, it is possible that such a
cascade of necks could form by a determin-
2.3-
istic mechanism in the full equations. Our
2.3-
results raise many questions: To what ex-
tent does the regular repetition of the
2.48- formation of these necks depend on the
'2.4-
2-4 characteristics of the noise source? Under
what circumstances can the interfacial
shapes develop even more structure? And
2.5 what is the spectrum of noise in a typical
2.50. experiment?
Fig. 3. Simulation of the profile for a drop of a
glycerol in water mixture (85 weight %; viscosityFig. 4. Simulation of the same drop as in Fig. 3
= 1 P) falling from a nozzle of diameter D = 4.5 in the presence of noise. A view of the entire
REFERENCES AND NOTES
mm. (A) A view of the entire drop, at approxi- drop is indistinguishable from that in Fig. 3A.
mately 10o-8 s before snap-off, shows that near (A) An enlargement of the first neck (as in Fig. 1 A. Pumir and E. D. Siggia, Phys. Rev. Lett. 68,
the bottom of the long neck there is a region 3B) shows a third, fourth, and (very thin) fifth 1511 (1992).
where the thickness decreases and forms a neck sprouting from the second neck. (B) An 2. W. S. Rayleigh, Proc. London Math. Soc. 4, 10
secondary neck. (B) An enlargement of the enlargement of the bracketed region of (A) (1878).
3. J. Plateau, Statique Exp6rimentale et Theoretique
region in (A) enclosed by brackets. Note that shows the fourth and fifth necks of the previous
des Liquides Soumis aux Seules Forces Mol6cu-
the scales are different along the two axes, so figure, as well as the sixth neck. In the inset, a
laires (Gauthier-Villars, Paris, 1873).
that the shape can be easily seen. The inset magnification of thve sixth neck shows a seventh
4. J. Eggers and T. F. Dupont, J. Fluid Mech. 262,
shows another, greater magnification. neck. 205 (1994).

SCIENCE * VOL. 265 * 8 JULY 1994 221

This content downloaded from 129.59.95.115 on Mon, 19 Sep 2016 23:42:14 UTC
All use subject to http://about.jstor.org/terms
 5. D. H. Peregrine, G. Shoker, A. Symon, ibid. 212, Applied Mathematics Series, New York, 1993). quickly to reach sexual maturity within an
25 (1990). 18. The boundary conditions used at the top of the
6. M. Tjahjadi, H. A. Stone, J. M. Ottino, ibid. 243, drop in the simulation are different from those that
ecologically feasible length of time. The
297 (1992). actually apply in the experiment. We believe that high metabolic rates of endotherms and the
7. We note that, nevertheless, the solution near the this accounts for the differences in the two figures high efficiency of energy production of ec-
singularity must still be measured in terms of a near the nozzle.
length scale relevant to the dynamics in the 19. M. P. Brenner, X. D. Shi, S. R. Nagel, unpublished
totherms are contrasting physiological strat-
problem. This scale is the viscous length defined results. egies for solving this problem (2). Argu-
below. 20. From the Eggers similarity solution, one can show ments for particular strategies of thermoreg-
8. C. Domb and M. S. Green, Eds., Phase Transi- that the Reynolds number is -1. This situation is
tions and Critical Phenomena (Academic Press,
ulation have been based on posture, bone
distinct from other models considered (1, 12).
London, 1972). 21. Tjahjadi et al. (6) observed a different hydrody- histology, predator-prey relations, feeding
9. P. Constantin etal., Phys. Rev. E47, 4169 (1993). namic phenomenon that also had a fractal spatial mechanics, brain size, postulated behav-
10. R. E. Goldstein, A. I. Pesci, M. J. Shelley, Phys. structure. By allowing a stretched drop to relax,
iors, and paleobiogeography. Though inno-
Rev. Lett. 70, 3043 (1993). they found a sequence of droplets with sizes that
11. J. Eggers, ibid. 71, 3458 (1993). decreased in a geometric progression. Our situ- vative, these lines of evidence have ulti-
12. A. Pumir, B. I. Shraiman, E. D. Siggia, Phys. Rev. ation differs from theirs in that the repeated for- mately proven to be inconclusive.
A 45, R5351 (1992); T. Dombre, A. Pumir, E. D. mation of necks is a property of the singularity In this study, we used the oxygen isoto-
Siggia, Physica D 57, 311 (1992). itself and not the result of multiple breakup events.
13. G. W. C. Kaye and T. H. Laby, Tables of Physical 22. J. Koplik and J. R. Banavar, Phys. Fluids A 5, 521 pic composition of bone phosphate (8p) to
and Chemical Constants (Longman, New York, (1993). calculate the body temperature variability
1986); J. B. Segur and H. E. Oberstar, Ind. Eng. 23. U. Landman, W. D. LuedtKe, N. A. Burnham, R. J. of Tyrannosaurus rex. The phosphate-water
Chem. 43, 2117 (1951). Colton, Science 248, 454 (1990).
14. E. A. Hauser, H. E. Edgerton, W. B. Tucker, J. 24. We thank J. Eggers and T. Dupont for many helpful
isotopic temperature scale was developed
Phys. Chem. 40, 973 (1936); H. E. Edgerton, E. A. discussions and for generously sharing their numer- for marine invertebrates (4) and later con-
Hauser, W. B. Tucker, ibid. 41, 1017 (1937). ical code and for showing us how to use it. We thank firmed to be accurate for fish (5) and mam-
15. Similar equations have been derived indepen- R. Leheny for initiating the experiments on the drop-
dently by S. E. Bechtel, J. Z. Cao, M. G. Forest [J. let breakup, R. Jeshion for help in the photographic mals (6). Vertebrate 6p is a function of
Non-Newtonian Fluid Mech. 41, 201 (1992)] in the development, 0. Kapp for help in obtaining a fast the body temperature at which bone forms
context of a general mathematical study of the camera,.and E. Siggia for useful comments. D. Grier and of the isotopic composition of body
dynamics of viscoelastic fluids. Their derivation and D. Mueth helped us analyze the photographs
water (7, 8). The isotopic composition of
includes Eqs. 2 and 3 as a special case, with a digitally, and N. Lawrence made measurements of
slightly different form for Eq. 4. drop mass. We also thank A. Bertozzi for advice and body water (8bw) depends on the 8180 of
16. A similar equation (without the correct coefficient L. Kadanoff for, among many other things, stimulat- water ingested during feeding and drinking
in front of the term with viscosity) was proposed ing our interest in this subject. Supported by NSF
as well as on the metabolic rate relative to
by R. W. Sellens [Atomization Sprays 2, 236 grants DMR-MRL 88-19860 and DMR 91-11733,
(1992)]. and Department of Energy grant DE-FG02- water turnover rates (8, 9). The body
17. A. L. Bertozi, M. P. Brenner, T. F. Dupont, L. P. 92ER251 19. water of air-breathing vertebrates is at a
Kadanoff, in Centennial Edition, Applied Mathe-
physiological steady state but is not in
matics Series, L. Sirovich, Ed. (Springer-Verlag 1 February 1994; accepted 4 April 1994
equilibrium with environmental wa'ter.
Several studies have demonstrated that 6p
is linearly offset from the 8180 values of
Thermophysiology of Tyrannosaurus rex: the local meteoric water, from the relative
humidity, or from the ambient water (in
Evidence from Oxygen Isotopes the case of marine mammals) (10). With-
out knowledge of the 8180 values (1 1) of
Reese E. Barrick* and William J. Showers body water for each fossil species, it is not
possible to calculate actual body tempera-
tures. However, all body water compart-
The oxygen isotopic composition of vertebrate bone phosphate (8p) is related to ingested
water and to the body temperature at which the bone forms. The ap is in equilibrium ments
with in individual mammals have essen-
tially identical 8180 values (12). The
the individual's body water, which is at a physiological steady state throughout the body.
Therefore, intrabone temperature variation and the mean interbone temperature differ- variations in 6p among skeletal elements
ences of well-preserved fossil vertebrates can be determined from the ap variation. in an individual should therefore reflect
Values
of ap from a well-preserved Tyrannosaurus rex suggest that this species maintained relative differences in body temperature.
homeothermy with less than 40C of variability in body temperature. Maintenance of ho-Temperature differences between the core
meothermy implies a relatively high metabolic rate that is similar to that of endotherms.
body (the trunk) and the extremities can
be calculated by use of the bone isotopic
differences (A8P) and the slope of the
phosphate-water paleotemperature equa-
Dinosaurs dominated the terrestrial land- some of which are not represented in tion (4, 5).
scape for 163 million years, from their extant terrestrial vertebrates. An assessment of the isotopic integrity of
origin in the mid-Triassic to their extinc- Endothermy is a pattern of thermoregu- the fossil bone material being examiried is
tion at the end of the Cretaceous (1). lation in which body temperature depends crucial for paleophysiological interpretations.
Whether they were warm-blooded (ta- on a high and controlled rate of metabolic We analyzed a T. rex specimen from the
chymetabolic endotherms), cold-blooded heat production (3). In ectothermy, body Maastrichtian Hell Creek Formation for 6p
(bradymetabolic ectotherms), or some- temperature is dependent on behaviorally and for the isotopic composition of both the
thing in between (2) has been uncertain. and autonomically regulated uptake of heat structural bone carbonate (8) and the coex-
A group as successful and diverse as the from the environment. These patterns of isting calcite cements (8c6) (13). This speci-
Dinosauria may have employed a wide thermoregulation are end members along a men was chosen because of its excellent pres-
range of thermal physiological strategies, continuum of physiological strategies. In ervation (14). Two lines of evidence demon-
homeothermy, cyclic body temperature strated that the bone phosphate analyzed was
Department of Marine, Earth, and Atmospheric Sci-
variation, either nyctohemerally or season- likely to reflect the original isotopic signal of
ences, North Carolina State University, Raleigh, NC ally, is maintained within ?2?C despite growth. One test compared 6p and the isoto-
27695-8208, USA. much larger variations in ambient temper- pic composition of structural carbonate (6c)
*To whom correspondence should be addressed. ature (3). Large dinosaurs must have grown with the isotopic composition of the diage-

222 SCIENCE * VOL. 265 * 8 JULY 1994

This content downloaded from 129.59.95.115 on Mon, 19 Sep 2016 23:42:14 UTC
All use subject to http://about.jstor.org/terms