Download as pdf or txt
Download as pdf or txt
You are on page 1of 14

European Journal of Physics

Related content
- Physics of liquid jets
Fluid flow up a spinning egg and the Coriolis force - Dynamics of vortices and interfaces in
superfluid 3He
To cite this article: J C Martinez et al 2006 Eur. J. Phys. 27 805
- Are inertial forces ever of significance in
cricket, golf and other sports?

View the article online for updates and enhancements. Recent citations
- Eggs and milk: Spinning spheres partially
immersed in a liquid bath
Kenneth R. Langley et al

This content was downloaded from IP address 203.110.242.24 on 14/10/2020 at 12:30


INSTITUTE OF PHYSICS PUBLISHING EUROPEAN JOURNAL OF PHYSICS
Eur. J. Phys. 27 (2006) 805–817 doi:10.1088/0143-0807/27/4/012

Fluid flow up a spinning egg and the


Coriolis force
J C Martinez1, E Polatdemir2, Ankita Bansal3,
Wang Yifeng4 and Wang Shengtao4
1 National Institute of Education, Nanyang Technological University, 1 Nanyang Walk,
Singapore 637616
2 Centre for Science and Mathematics, Republic Polytechnic, 1 Kay Siang Road,
Singapore 248922
3 Saint Andrew’s Junior College, 55 Potong Pasir Ave 1, Singapore 358389
4 Hwa Chong Junior College, 661 Bukit Timah Road, Singapore 269734

E-mail: jcmartin@nie.edu.sg

Received 20 February 2006


Published 8 May 2006
Online at stacks.iop.org/EJP/27/805

Abstract
We study the dynamics of a spinning sphere whose south pole is in touch with
the surface of a still body of liquid. When the sphere is turning fast enough,
the fluid rises up the sphere, reaches the equator and is flung out as a fountain
of droplets. Although the fountain forms for water a weakly viscous fluid,
and for propylene glycol a much more viscous fluid, the dynamical situation is
different for each. For flows at mid-latitudes on the sphere, we formulate the
dynamical equations for the two liquids in terms of Newton’s law in a rotating
frame, noting that the Coriolis force plays an essential role in both liquids, and
obtain qualitative agreement with observations. We also discuss the possible
roles played by other forces.
(Some figures in this article are in colour only in the electronic version)

1. Introduction

Without exception, illustrations of the Coriolis force focus on large-scale systems (hundreds of
kilometres in the case of the weather [1], or thousands of kilometres in the case of Jupiter’s red
spot [2]) or long-observation time dynamics (24 h in the case of the Foucault pendulum [3]).
No wonder the Coriolis force seems unimportant for the smaller length scale, or short-time
mechanics of everyday life. Yet through the media the viewing public is all too familiar with
the characteristic counterclockwise circulation of a hurricane in the Northern Hemisphere and
is well informed that this is due to the Coriolis force. Thus during the recent devastation
wrought by hurricane Katrina, NASA reported that ‘because air currents are influenced by
the Coriolis force (caused by the rotation of the Earth), Northern Hemisphere hurricanes are
0143-0807/06/040805+13$30.00 
c 2006 IOP Publishing Ltd Printed in the UK 805
806 J C Martinez et al

(a) (b)

Figure 1. (a) Fountain of (coloured) water droplets formed at the equator of a sphere rotating at
600 rpm. Water on a shallow tray rises in threads from the sphere’s south pole up the sphere. The
water never rises above the equator but is flung out from there. The photograph was taken at low
shutter speed. (b) High shutter speed close-up photograph for propylene glycol, a viscous liquid,
at 390 rpm. The flow lines can be clearly seen. The sphere’s angular velocity is straight down,
so the fluid lines appear to turn in the same direction. As the speed is increased more flow lines
appear. In the foreground we see the vortex formed below the sphere. There are also flow lines for
water as we will see in section 4.

characterized by an inward counterclockwise (cyclonic) rotation towards the centre [4]’. It is


easy to see why the average student’s familiarity with the Coriolis phenomenon can only be at
the theoretical level, limited to a few esoteric examples, and demonstrations of it necessarily
exploit its non-intuitive nature to provoke hilarious behaviour [5]. Here we discuss the
unexpected phenomenon of incompressible viscous fluid flow up the lower half of a spinning
egg which is a direct consequence of the Coriolis force. Indeed, when the bottom pole of
an upright rotating sphere (which for convenience replaces the egg) is placed in touch with
a quiescent body of viscous fluid or water, the fluid trails in sinuous rivulets up the sphere,
and collects at the equator from where it is flung off in a spectacular fountain of droplets
(figure 1). The fluid never goes beyond the equator no matter what the rotational speed of the
sphere. Also when the sphere is replaced by a V-shaped funnel, the fluid fountain re-appears;
but if the funnel is replaced by its inverted  image or by a cylinder, the effect is not seen
at all. In explaining this surprising phenomenon we provide the reader with a quantitative
understanding of the Coriolis effect that can be verified against direct observations at the
level of an elementary physics experiment (and at length scales of 10−2 m and time scales of
seconds). Therefore, we hope that the gap noted above in providing students with an intuitive
as well as an experiential grasp of the Coriolis effect can be filled [6].
When discussing the Coriolis effect, it is customary to go to a reference system that rotates
with the sphere. The photographs in figure 1 (and elsewhere in this paper) correspond to this
reference system. If the angular velocity of the sphere is Ω as seen in the laboratory, the
acceleration of a particle moving on the sphere due to the Coriolis force is ac = −2Ω × u,
where u is the velocity of the particle relative to the sphere [3, 6, 7]. In figure 2, we indicated
the directions of the Coriolis acceleration for a tiny fluid parcel on the surface of four common
geometric solids, all of them rotating counterclockwise when seen from above. For each solid,
the directions of ac are shown for two orthogonal velocity directions: (a) along a circumference
and (b) along an upward tangent in the same plane as Ω. On examining these figures, it is easy
to see why the fountain effect does not show up for the cylinder or the inverted  funnel: the
component of ac in the upward direction along the surface is absent. Similarly, one can show
why there is no fluid flow above the equator of the sphere, as noted in our experiments.
Fluid flow up a spinning egg and the Coriolis force 807


u
ac
u
ac u
ac = 0
u
ac

(a) (b)

Ω Ω

ac u ac u

u u
ac

ac
(c) (d)

Figure 2. Directions of the Coriolis acceleration ac for circumferential and vertical velocities
(dotted lines) for simple solids rotating about the vertical symmetry axis. The blob represents a
fluid parcel.

The phenomenon we have just sketched seems to have been first observed by Gutierrez,
Fehr, Calzadilla and Figueroa in 1998 [8]. Besides describing it and providing photographic
evidence of the spinning egg and the fountain, they also coined the epithet ‘fountain effect,’
and measured the speed of the water droplets spun off by a sphere, noting that the velocity
fitted well with the rotational speed of the sphere’s equator. But they did not pursue the subject
much further than a dynamical theory based on elementary particle mechanics; they attribute
the rise of the water up the sphere to an ‘adhesive’ force aided by the centrifugal force. We
are not aware of other contributions to the subject before or after this paper. As for the fluid
dynamics of a sphere rotating in a volume of otherwise still fluid, there is ample literature
available ([9, 10] and references therein). However this is not the focus of our report, which
is concerned with the fountain effect. In this report we will emphasize the role of the Coriolis
force and viscous friction which Gutierrez et al did not consider [8]. We will also argue that
the surface tension force plays a significant role only close to the liquid surface and near the
equator. The material we present is accessible to a beginning student in theoretical mechanics
who has some acquaintance with non-inertial forces, and would be a good introduction to the
fascinating subject of rotating fluids [10].
In section 2 we give a qualitative overview of the phenomenon we are studying, estimate
the magnitudes of the forces possibly involved and divide the hemisphere into three regions.
We will argue that the mid-latitude region between the south pole and the equator is the simplest
one where the importance of the Coriolis force is also clearest. We present a semi-qualitative
model of the fluid flow at mid-latitudes for a viscous liquid in section 3. The corresponding
model for water is discussed in section 4. Our conclusions are summed up in section 5. Some
ancillary details are briefly taken up in three appendices.
808 J C Martinez et al

2. Forces: qualitative discussion

It is convenient at this point to go over some of the gross details of the experiments we had
carried out (figure 1) and to give estimates of the magnitudes of the forces involved. We
used two liquids, a low viscosity liquid (water at room temperature) and a high viscosity
liquid (propylene glycol, C3 H8 O2 , also at room temperature). A wooden sphere of radius R =
4.8 cm was fitted with an axis which was mounted onto a mechanical stirrer with variable
rotation speeds. For both liquids a threshold rotational speed 0 had to be exceeded before
the fountain effect could be observed; for water this was about 500 rpm (≈50 rad s–1), while
for propylene glycol it was about 350 rpm (≈35 rad s–1). Pictures taken at high shutter speeds
showed that the thickness of the fluid trailing up the sphere was roughly 1 mm thick for water
and for propylene glycol. Although the sphere used was smooth, it did wobble slightly at high
rotation speeds: this was desirable because it would ensure that individual fluid flow lines could
be observed (we use the term flow line instead of streamline because the flow goes in ‘rivulets’
as in figures 1(b) and 7). Otherwise it would have been very difficult to distinguish flow
lines if the fluid shot up as a film uniformly spread out on the lower hemisphere. Finally, we
should mention the following important observation: for water, when the fountain effect was
taking place, the rivulets always travel opposite to the sphere’s sense of rotation (see figures 7
and 8); it is the reverse with propylene glycol so that the rivulets move along with the sphere
as it spins (see figure 1(b)).
In the following discussion it will be helpful to distinguish three regions on the sphere.
The region close to the bottom or south pole of the sphere is in contact with the pool of fluid;
the dynamics in this region involves the vortex caused by the rotating sphere and is fairly
complicated. Very close to the equator, one will note the formation of long threads as droplets
are formed and then flung out. The Coriolis force is either very small or vanishes in this region
and is outside our present interest (cf figure 2). It will be the region in between these two, or
the mid-latitude region, that will be the main focus of attention for us.
At the outset, five forces might possibly be relevant to the dynamics of the fountain
effect: gravity, surface tension, the Coriolis force, the centrifugal force and viscous drag. The
importance of the Coriolis force is clear since the rotational motion of the sphere can only
impart a circumferential motion to the water: a transverse force must be present to set the
water moving up the sphere. In this paragraph we estimate the approximate magnitudes of
the first four forces. Because the fluid trails up the sphere in thin rivulets, it is reasonable to
model this motion, not in terms of fluid flow or of a spreading thin film, but as a train of small
parcels wending their way up the surface of the sphere. For simplicity we then assume that
we have parcels of spherical shape with a typical diameter of d ≈ 1 mm for both liquids. (The
exact shape and size are not required.) Such a parcel would have a volume of v = 16 π d 3 and
a weight of Fg ≡ mg = ρvg, where ρ is the fluid density. The surface tension force is given
by the standard formula Fs = π dγ , where γ is the surface tension. (The relevant physical
constants are given in table 1.) This force acts on the parcel while it is close to the bottom of
the sphere in contact with the liquid surface; once the parcel has lost this contact, it is replaced
by a surface tension force with the sphere (R = radius of the sphere). The centrifugal force
Fr ≈ m2 R acts outwards in the radial direction. Together with formula F c = −2mΩ ×u for
the Coriolis force, we also remind ourselves that its direction is perpendicular to the velocity.
For estimating the Coriolis force we will use the approximate formula Fc ≈ −2m2 R instead.
Finally, the viscosity force can largely be ignored for water but it is important for propylene
glycol (see also the following paragraph). Estimates of the magnitudes of the first four forces
(based on the foregoing formulae and figures) are displayed in the first four rows of table 1.
It is clear that the weight may be safely ignored. The threshold angular speed mentioned in
Fluid flow up a spinning egg and the Coriolis force 809

U U

Figure 3. Growth of the boundary layer, indicated by the dashed curve, along a flat semi-infinite
plate located in an otherwise uniform flow with velocity U. The velocity field is zero on the surface
of the plate and grows gradually, resuming its uniform flow value beyond a fluid thickness of δ.

Table 1. Estimates of the magnitudes of the forces involved in the experiment under steady-state
conditions.
Water Propylene glycol
(ν ≈ 10−6 m2 s−1 , γ = 7 × 10−2 N m−1 , (ν ≈ 4 × 10−5 m2 s−1 , γ = 4 × 10−2 N m−1 ,
Force 0 ≈ 50 rad s−1 , ρ ≈ 103 kg m−3 ) 0 ≈ 35 rad s−1 , ρ ≈ 103 kg m−3 )

Fg (N) 0.5 × 10−5 0.5 × 10−5


Fs (N) 2.2 × 10−4 (close to the water surface) 1.3 × 10−4 (close to the liquid surface)
Fc (N) 0.5 × 10−5 0.6 × 10−4
Fr (N) 0.5 × 10−5 0.35 × 10−4
Fν Fν ≈ 0 (away from the water surface) Fν ≈ Fc (at mid-latitudes)

the previous paragraph has to do with the surface tension because the Coriolis and centrifugal
forces must overcome the attachment of the parcel to the fluid surface where the sphere touches
it. The magnitudes of Fs , Fr and Fc are about the same at the threshold angular velocity 0 .
As  is increased, the Coriolis force also increases whereas surface tension remains constant.
Once the fluid parcels have started to climb up the sphere the surface tension force with the
sphere balances Fr since they act along the same line. However when the parcels approach
the equator they slow down to zero relative velocity and the surface tension with the fluid
becomes important once again because the parcels collect there before being spun off. We
give photographic evidence of this later.
It remains to elaborate on the role played by the viscous drag for propylene glycol. Recall
that viscous flow refers to the growth of the velocity field from a zero value on the surface
of the boundary to the steady-state value characterizing the flow (see figure 3). This growth
takes place over a boundary layer that is in contact with the sphere. Because of the thinness
of the threads moving up the sphere it is important to compare their thickness with their
corresponding boundary layer. To calculate the thickness of the layer, we first compute the
Reynolds number Re from Re = RU/ν ≈ R 2 /ν, where ν is the kinematic viscosity [11].
The inverse of Re is also called the Ekman number (see p 7 of [10]). For the cases given
in table 1 we obtain Re ≈ 105 for water and Re ≈ 2 × 103 for propylene √ glycol. The
corresponding thickness δ of the boundary layer is given by δ = R/ Re, or 0.13 mm for
water and around 1 mm for propylene glycol. Because the typical thickness of the threads
is about 1 mm for both fluids, we see that viscosity is relatively unimportant for water but
should be very important for propylene glycol. Thus for the case of water, the flow above the
boundary layer is essentially viscosity free and we expect to describe the dynamics without
having to include viscosity, centrifugal force and surface tension (and gravity as noted in the
previous paragraph). Similarly for propylene glycol we exclude effects due to surface tension,
centrifugal force and gravity once the fluid parcels have started climbing the sphere (see last
row, table 1).
810 J C Martinez et al

R
z α
Fc . ρˆ cosα

ρ
2mΩρϕ& ρ
ˆ

Figure 4. For a spherical parcel of mass m moving up the sphere, the radial component of the
Coriolis force in cylindrical coordinates is given by Fc · ρ̂ = 2mρ ϕ̇. The radius is set to unity so
that the radial distance is x = sin α.

Two more clarifications have to be made to complete our qualitative picture. The first is
on the form of the viscous force. Indeed it is unlikely that the textbook definition of viscosity
in terms of stress and the velocity gradient is practical in our case. But although we are
unable to quantify the viscous force from experimental data, we can appeal to steady-state
conditions. When the sphere is rotating steadily and the fountain maintained, we can conclude
that steady-state conditions prevail. To be exact, the liquid threads wander about, appearing
and disappearing as the spinning sphere wobbles. However, because of mass conservation, the
rate of mass flow up the sphere must equal the rate of mass spun off the equator as droplets and
these should have ‘on average’ a steady value. In this sense, we can speak of the steadiness of
the velocity for threads at any given latitude. Under this condition we may also say that the
velocity at any latitude of the sphere is constant in time so that the Coriolis force must balance
all other forces. Since we are neglecting the surface tension, the centrifugal force and gravity,
we conclude that the steady-state condition implies the equalityFc ≈ Fν for propylene glycol
(see row 5 of table 1). In the following section we will see this equality utilized quantitatively.
The foregoing discussion shares something in common with the hypothesis of continuity that
is a cornerstone in fluid mechanics and we will leave the issue as it is here [12].
The second and final point is that, while we have been neglecting the viscous force for
water all along, it cannot be neglected any longer when the threads are close to the equator.
This is because water slows down as it approaches the equator (attaining zero relative velocity
at the equator so U → 0 there) and also collects there. The Reynolds number decreases
markedly while the boundary layer increases in thickness. Moreover the surface tension
operates here as the photographs show (see figures 6 and 8).

3. The case of propylene glycol for mid-latitudes

In this section we give a semi-quantitative treatment of the flow lines at mid-latitudes for the
case of the viscous liquid propylene glycol based on the discussion of section 2. Our task will
be to write and solve a simple dynamical equation for the fluid parcel moving up the sphere.
Following Erdos we find it more convenient to employ the cylindrical coordinates (ρ, ϕ, z)
[13]. In this system the acceleration is given by a = ρ̂(ρ̈ − ρ ϕ̇ 2 ) + ϕ̂(ρ ϕ̈ + 2ρ̇ ϕ̇) + ẑ z̈. The
Coriolis force can be cast in cylindrical coordinates as F c = 2(−ρ̇ ϕ̂ + ρ ϕ̇ ρ̂); there is no
z-component. Consider first the azimuthal motion. As can be seen from figure 4, the
component of the Coriolis force upward and tangent to the sphere is F c · ρ̂ cos α =
mR ϕ̇ sin 2α. (Note that this vanishes at the equator.) As discussed in section 2, we
Fluid flow up a spinning egg and the Coriolis force 811

-1
1
-0.5
0 0.5
0.5 0
1
0 -0.5
-1
-0.25

-0.5

-0.75

-1

Figure 5. The trajectory of the parcel given by equations (2) and (3); r = 4.4 and x0 = 0.23. The
photograph corresponds to propylene glycol and a sphere with a rotational velocity of 390 rpm.
The rotation direction is clockwise when seen from above.

1
0.5
-0.5 0
-1
0

-0.25

-0.5

-0.75
1
0.5
-1 0
-0.5

-1

Figure 6. Same as figure 5; r = 7.5 and x0 = 0.22. The photograph corresponds to a rotational
speed of 500 rpm. Note the long threads flung out from the equator. They are also visible close to
the equator (though less spectacular) in figure 5.

exclude gravity, surface tension and the centrifugal forces once the parcel has started to climb
the sphere. Assuming the simplest form for the component of the viscous force, Newton’s
second law in the azimuthal direction takes the form
mα̈ = mR ϕ̇ sin 2α − bR α̇. (1)
That is, bR α̇ is the (azimuthal) viscous drag force, linear in the velocity with constant
coefficient b (with units M T−1). (See also the first paragraph of section 4 for a comment on
equation (1).)
As suggested by an examination of the photographs accompanying figures 5 and 6, the
circumferential motion is much faster than the azimuthal motion associated with the angle α:
therefore we may set α̈ ≈ 0 in equation (1), which allows us to integrate it. Thus
   
b tan α tan(sin−1 x)
ϕ≈ ln = r ln , (2)
2m tan α0 tan(sin−1 x0 )
in which r = b/2m is a dimensionless constant and α0 (x0 ) is the initial value of the angle
α (radius x, in units of R). Hence r encapsulates the interplay between the viscous drag force
and the Coriolis force. The right-hand side of equation (1) shows clearly that the Coriolis
force is approximately balanced by the viscous force during much of the motion of the parcel
(see last row of table 1). That is, the Coriolis force is essential in making the fountain effect
812 J C Martinez et al

possible. Thus, together with the equation of the lower hemisphere,



z = − 1 − x2. (3)

Equation (2) provides us with the parametric equations of the trajectory of a parcel rising
up the sphere in terms of the radial distance x. If we are not interested in time itself these
suffice to solve our problem. Assumptions such as α̈ ≈ 0 often occur in multi-dimensional
problems as, for instance, in the Born–Oppenheimer approximation in atomic physics (where
it is also called the adiabatic approximation) [14].
We might attempt considering next the equation for circumferential motion.
Unfortunately, the motion in this direction must take into account the driving force, an element
that we do not have direct access to. That is, the continuous rotation of the sphere (spun by
an external agent) generates not only a vortex in the liquid pool beneath it (see foreground
of figure 1(b)) but also, through the spray of droplets flung from the equator, a confluence
of flows returning to the vortex which helps to drive the train of parcels circumferentially.
We note in passing that, because the driving force is in the circumferential direction, it is not
required in equation (1). We will have to be content with equations (2) and (3). In appendix A
we briefly discuss the effect of a periodic driving force with long period compared with −1 .
Plots of equations (2) and (3) are given in figures 5 and 6 for two values of  (to compare
with the photograph,  must have negative sign). A program in Mathematica for plotting these
figures is given in appendix B. Comparison with the photographs shows qualitative agreement.
The dimensionless constant r = b/2m must be a function of the only dimensionless constant
available, namely, the Reynolds number Re. Taking a quadratic dependence on Re, we find
that r is quadratic in , a relation roughly consistent with the two figures. This is further
corroborated when we observe that the first term on the right-hand side of equation (1) is
quadratic in  (since ϕ̇ is expected to be linear in the rotational velocity); the second term, by
contrast, is at most independent of (or possibly inversely proportional to) . Thus the viscous
force increases in step with . But as a parcel rises up the sphere, the Coriolis force decreases
due to the decrease in velocity. In figure 6 we also note the long threads spun off from the
equator; they resemble the long drops from a capillary tube which is attributed to the interplay
between gravity and surface tension [15]. In our case gravity is replaced by the centrifugal
force. We did not undertake a study of these long threads in our report.

4. The case of water for mid-latitudes

In this section we present a semi-quantitative model for the trajectory of water parcels at
mid-latitudes up the sphere which is qualitatively different from that for propylene glycol
because of the notable absence of viscous drag. (Indeed the rivulets of the two liquids move
in opposite directions, as the photographs suggest.) If we consider only the motion of a water
parcel out of the water surface and away from the equator, then the surface tension, centrifugal
force and viscosity may also be neglected. Thus the only force of importance is the Coriolis
force, as suggested in table 1. But the Coriolis force, like the magnetic force, does not change
the total energy: that is, v · F c = 0. Hence the role of the Coriolis force here is to constrain
the parcel to turn to its left. It follows that the kinetic energy (or equivalently, the speed of the
parcel) is a constant of the motion. (This property holds for the viscous case also but now the
presence of drag greatly diminishes the ability of the Coriolis force to ‘bend’ the trajectory of
a parcel upward. This is the content of equation (1).)
In their experiments with water, Gutierrez et al concluded that the water droplets flung
from the equator had speeds equal to the rotational velocity of the sphere at the equator,
Fluid flow up a spinning egg and the Coriolis force 813

i.e. v ∼= R [8]. To allow for greater generality, we will assume that the velocity ratio of the
(constant) speed of a water parcel v to R is equal to a constant e  1,
v
e= . (4)
R
This means that a parcel starts climbing the sphere with a velocity v = eR relative to the
sphere at that position. As photographs suggest, relative to the sphere, v is directed opposite
to the sphere’s tangential velocity. Following Erdos [13] we write the element of arc length
on a spherical surface in cylindrical coordinates as
R2
ds 2 = ρ 2 dϕ 2 + 2 dρ 2 . (5)
R − ρ2
Newton’s law, written for the circumferential motion, takes the form (see the paragraph before
equation (1))
ρ ϕ̈ + 2ρ̇ ϕ̇ + 2ρ̇ ≈ 0. (6)
Assuming that ϕ̇ ≈ 0 at the equator (as suggested by Gutierrez et al [8]), where ρ = R,
we may integrate equation (6) and obtain
 2 
R
ϕ̇ = − −1 . (7)
ρ2
The negative sign reflects the remark made about equation (4).
It is convenient to define the following dimensionless variables: S = s/R and x = ρ/R.
We can eliminate time through equation (7) and the relation dS = e dt. Hence combining
equations (4), (5) and (7) we find
 
e2 1
1 − 2 (1 − x ) dS 2 =
2 2
dx 2 . (8)
x 1 − x2
This can be further simplified to
1 x
dS =  dx, (9)
e (x 2 − a 2 )(x 2 − b2 )(x 2 − 1)
where the constants are  
1 1 1 1 1 1
a =1+ 2 +
2
+ 4 > 1, b =1+ 2 −
2
+ 4 < 1.
2e 2e e2 2e 2e e2
The integration of equation (9) is standard and can be carried out in terms of elliptic functions
provided that x  b; in other words bR is the smallest (or initial) value of radial distance of
the parcel from the rotation axis. (A summary of useful basic facts about elliptic functions is
provided in appendix C.) With the aid of the Mathematica software we readily obtain from
equation (9) the distance travelled on the sphere,
  

2 − 1  b2 − 1
1 1 x 
S= √ F sin−1 , (10)
e a2 − 1 b2 − 1  a 2 − 1
where F (φ |m ) is the elliptic integral of the first kind and m  1. Our convention is that
S = 0 at the equator. Inverting this for the radial distance x, we have
   2 
b − 1
x = 1 − (1 − b ) sin am e a − 1S  2
2 2 2  (11)
a −1
in which am(u|m) is the (Jacobi) amplitude function. Finally we can combine equations (7)
and (9) to derive
 
1 1 1 − x2
dϕ = −e dS = −  dx. (12)
x2 − 1 x (x 2 − a 2 )(x 2 − b2 )(x 2 − 1)
Integration, via the Mathematica software once again, yields the angular displacement about
814 J C Martinez et al

0
-1
-0.5 -0.25
0
0.5
1 -0.5

-0.75

-1
-1 -0.5 0 0.5 1

Figure 7. Comparison of the calculated trajectory (centre) for water (e = 0.9) with photographs
for angular velocities of  = 600 rpm (left) and 700 rpm (right).

the z-axis,
       
 

−1   −1  −1 +  −1 + b   −1 + b2
1 − b ; −1 + x
2
x2 2
  ,
ϕ=√ F sin  + 2
 −1 + a 2
a2 − 1 −1 + b2 −1 + a 2 −1 + b2
(13)
where we introduced the elliptic√integral of the third kind (n; φ|m). To complete our
description, we write z/R = − 1 − x 2 for the vertical height. One can also show that
F (0|m) = 0 and (n, 0|m) = 0. As before our convention is that at the equator ϕ = 0 and
negative at radial distance x.
A plot of the above results is shown in figure 7 and compared with high shutter-speed
photographs for two values of . Unlike the case of propylene glycol, the calculated trajectory
for water is independent of the angular velocity, provided that the threshold rotational velocity
has been exceeded. The only parameter available is the velocity ratio e which is assumed to
be close to unity. The similarity between the calculated trajectory and the actual one improves
with higher . This is easy to explain since the kinetic energy of a parcel at a higher angular
velocity is less affected by gravity than one of lower angular velocity. The Coriolis force is
strong at mid-latitudes above the south pole and below the equator as shown by the fairly
large angle between the actual path and the intersecting latitude. In contrast with the case of
propylene glycol, a half turn or less is sufficient for a water parcel to arrive at the equator.
This is clear evidence of the effect of viscous drag in the former. Since the parcel moves in
the opposite direction to the direction of the driving force, the latter need not be considered
further, unlike the case of propylene glycol.

5. Discussion and outlook

We have now seen how the Coriolis force can be important and significant in a small length-
and short time-scale system. It remains to put it in perspective. We had distinguished three
distinct regimes in the problem on hand: (a) very close to the south pole, (b) very close to
the equator and (c) the region between the south pole and the equator. We had confined our
analysis above exclusively to the last regime. In the first regime, four forces, namely, surface
tension, centrifugal force, Coriolis force and viscous drag are important. Gravity may still
be neglected since there is little depth involved. There is also a vortex here which acts as
an external driving agent. A similar vortex had been recently studied and found to contain
unexpected features [16]. This region clearly falls under fluid mechanics and is outside the
scope of our study.
Fluid flow up a spinning egg and the Coriolis force 815

Figure 8. Water threads being spun off the equator which is due to the interplay of surface tension
and centrifugal force.

The second regime corresponds to the approach of the parcel to the equator. In the case
of propylene glycol, x → 1− , and equation (2) takes the form ϕ = − 2r ln(1 − x) + O(1 − x) +
constant, so that a parcel circles the sphere many times as it gains only slightly in radial
displacement. Although this has intuitive appeal, we would have to regard it with suspicion
because we had neglected surface tension and equation (2) was based solely on a comparison
of the viscous drag with the Coriolis force. In fact surface tension does operate in this regime
as figure 6 (figure 8) clearly shows for propylene glycol (water). On the other hand, the
photographs and the graph in figure 7 suggest that a water parcel is also moving slowly at the
equator. In this case the velocity ratio e → 0 and both surface tension and the centrifugal
forces are at work in this instance. Neither the approach of section 3 nor of section 4 really
applies to this regime.
In conclusion, we have shown how simple dynamical analyses based on the interplay
between the Coriolis force and viscous drag in the case of propylene glycol and the constancy
of energy even in the presence of the Coriolis force for water yielded qualitative agreement
with observations. Our analyses took advantage of steady-state conditions and, unfortunately,
failed to include time; it would be interesting to obtain motion-picture observations to guide
one in understanding this complex phenomenon. We also observed that close to the equator
both liquids essentially are governed by the surface tension and the centrifugal force: the
Coriolis force no longer operates here since the fluid has attained almost zero relative velocity.
Thus a study of the second regime falls out of the scope of this report but should be an
interesting area to do further work in.

Acknowledgment

We are grateful to Lim-Ang Ah Buan and Lim Chee Sin for making available their laboratory
facilities to us.

Appendix A. Circumferential motion

We consider circumferential motion in the viscous case briefly. To simplify the equations we
will assume ρ̇  ρ, in keeping with the observation that the parcel makes many circumferential
turns before arriving at the equator, having started at the bottom. Then the acceleration in this
direction is ρ ϕ̈ + 2ρ̇ ϕ̇ ≈ ρ ϕ̈. Similarly, the sum of the Coriolis force and the viscous force in
816 J C Martinez et al

this same direction is −2mρ̇ − bρ ϕ̇ ≈ −bρ ϕ̇. (Physically the first term is the contribution
of the azimuthal velocity to the Coriolis force; because this azimuthal velocity was generated
by the Coriolis force acting on the circumferential motion, we expect its contribution in
the circumferential direction to be small.) Imitating equation (1), the approximate form of
Newton’s second law is

F0 eiωt − bρ ϕ̇ ≈ mρ̇ ϕ̈,

where the first term on the left is the driving force. The steady-state solution is
F0
ϕ= eiωt .
ωρ(−mω + ib)
We assume that the viscous force is strong so b  mω. Then for ω  −1 and short
intervals of time such that ωt ≈ 0 (whose parallel conditions corresponding to the photographs
of figures 5 and 6), we find a steady value for the angular velocity at a given radial distance ρ,
ϕ̇ ≈ iF0 /bρ.

Appendix B. Program in Mathematica

The program in Mathematica given below is based on equations (2) and (3). This was used
to generate the graphs in figures 5 and 6. Note that as the parcel trajectory approaches the
equator, smaller and smaller intervals are taken in order to plot the trajectory points more
accurately.

x0 = 0.23
g = 3.75
ϕ = g Log[Tan[ArcSin[x]/Tan[ArcSin[x0]]]

z = − 1 − x2
fig1 = ParametricPlot3D[{Cos[t]Cos[u], Sin[t]Cos[u], Sin[u], {t, 0, 2P i}, {u, −Pi/2, 0},
Shading- > False]
fig2 = ParametricPlot3D[{xCos[ϕ], x Sin[−ϕ], z}, {x, x0,.9}, ViewPoint- > {−0.5, 1.0, 0}]
fig3 = ParametricPlot3D[{xCos[ϕ], x Sin[−ϕ], z}, {x,.9,.95}, ViewPoint- > {−0.5, 1.0, 0}]
fig4 = ParametricPlot3D[{xCos[ϕ], x Sin[−ϕ], z}, {x,.95,.97}, ViewPoint- > {−0.5, 1.0, 0}]
fig5 = ParametricPlot3D[{xCos[ϕ], x Sin[−ϕ], z}, {x,.97,.98}, ViewPoint- > {−0.5, 1.0, 0}]
fig6 = ParametricPlot3D[{xCos[ϕ], x Sin[−ϕ], z}, {x,.98,.99}, ViewPoint- > {−0.5, 1.0, 0}]
Show[fig1, fig2, fig3, fig4, fig5, fig6, ViewPoint- > {−0.8, −1.5, −0.1}, Shading- > False]

Appendix C. Elliptic integrals



Integrals of the form Q(x, f ) dx, in which Q is a rational function, and f 2 is a cubic or
quartic polynomial in x, are known as elliptic integrals. Any elliptic integral can be cast in
terms of the three standard types of Legendre–Jacobi elliptic integrals [17]. Here we will be
needing only the first and third. We define the elliptic integral of the first kind for m  1 by
 φ   sin φ
F (φ|m) = dθ/ 1 − m sin2 θ = [(1 − t 2 )(1 − mt 2 )]−1/2 dt,
0 0
Fluid flow up a spinning egg and the Coriolis force 817

where − 12 π < φ < 12 π. The complete elliptic integral of the first kind is given by
 
K(m) = F 12 π |m . The elliptic integral of the third kind is defined for − 12 π < φ < 12 π by
 φ
(n, φ|m) = (1 − n sin2 θ )−1 (1 − m sin2 θ )−1/2 dθ .
0
The amplitude for Jacobi elliptic functions is the inverse of the elliptic integral of the first
kind. Hence if u = F (φ|m), then we define the inverse
φ = am(u|m).

References
[1] Andrews D G 2000 An Introduction to Atmospheric Physics (Cambridge: Cambridge University Press) chapters
4 and 5
[2] Encrenaz T and Bibring J-P 1990 The Solar System (Berlin: Springer) chapter 7
[3] Marion J and Thornton S 1995 Classical Dynamics of Particles and Systems (Fort Worth, PA: Harcourt Brace)
chapter 10
Kittel C, Knight W D, Rudermann M A, Helmhotz A C and Moyer B J 1973 Mechanics (Berkeley Physics
Course, vol 1) (New York: McGraw-Hill) pp 114–5
[4] http://www.nasa.gov/vision/earth/lookingatearth/h2005 katrina.html
[5] http://www.fas.harvard.edu/%7Escdiroff/lds/NewtonianMechanics/CoriolisForce/CoriolisForce.html
[6] The physics of the Coriolis force is well explained in Feynman R P, Leighton R B and Sands M 1963 The
Feynman Lectures on Physics vol 1 (Reading, MA: Addison-Wesley) pp 19–8 ff
Stommel H M and Moore D W 1989 An Introduction to the Coriolis Force (New York: Columbia University
Press)
See also Vandenbrouck F, Berthier L and Gheusi F 2000 Coriolis force in geophysics: an elementary introduction
and examples Eur. J. Phys. 21 359–66
McIntyre D H 2000 Using great circles to understand motion on a rotating sphere Am. J. Phys. 68 1097–105
Mohazzabi P 1999 Free fall and angular momentum Am. J. Phys. 67 1017–20
Lewowski T, Lewowska L and Mazur P 1999 Measurement of the effect of Coriolis and centrifugal forces on
the trajectory of a body in a rotating frame Eur. J. Phys. 20 109–16
[7] Vandegrift G 1995 On the derivation of Coriolis and other noninertial accelerations Am. J. Phys. 63 663
[8] Gutierrez G, Fehr C, Calzadilla A and Figueroa D 1998 Fluid flow up the wall of a spinning egg Am. J.
Phys. 66 442–5
[9] Garrett S J and Peake N 2002 The stability and transition of the boundary layer on a rotating sphere J. Fluid
Mech. 456 199–218
[10] Greenspan H P 1968 The Theory of Rotating Fluids (Cambridge: Cambridge University Press) section 3.6
[11] Guyon E, Hulin J-P, Petit L and Mitescu C 2001 Physical Hydrodynamics (Oxford: Oxford University Press)
pp 71ff
[12] Batchelor G K 1967 An Introduction to Fluid Mechanics (Cambridge: Cambridge University Press) p 4
See also Guyon E, Hulin J-P, Petit L and Mitescu C 2001 Physical Hydrodynamics (Oxford: Oxford University
Press) pp 89–90
[13] Erdos P 2000 Spiraling the Earth with C G J Jacobi Am. J. Phys. 68 888–95
[14] Messiah A 1970 Quantum Mechanics (Amsterdam: North-Holland) p 783
[15] McMahon T A and Bonner J T 1983 On Size and Life (New York: Scientific American Inc/Freeman) p 215
[16] Andersen A, Bohr T, Stenum B, Juul Rasmussen J and Lautrup J 2003 Anatomy of a bathtub vortex Phys. Rev.
Lett. 91 104502
[17] Milne-Thomson L M 1972 Jacobi elliptic functions and theta functions Handbook of Mathematical Functions
9th edn ed M Abramowitz and I A Stegun (New York: Dover) chapters 16 and 17

You might also like