What is the speed of a perturbed

vortex ring?

J. L. Helm

Report submitted for mas8091

MMath Project

Supervisors: Prof. C. F. Barenghi & Dr. A. J. Youd

School of Mathematics & Statistics

Newcastle University
Newcastle upon Tyne
United Kingdom

April 29, 2010

 1

Abstract
It has been known since the time of Lord Kelvin that a circular
vortex ring of radius R has a self induced velocity which is inversely
proportional to R. Real vortex rings are often perturbed by Kelvin
waves (see figure on title page [1]). In this report we ask if these per-
turbed rings go faster or slower than their unperturbed counterparts.
Contents

1 Introduction 3

2 Classical vortex rings 4

3 Vortex rings in BECs 8

3.1 Overview of superfluids and the NLSE . . . . . . . . . . . . . 8
3.1.1 The Madelung transformation . . . . . . . . . . . . . . 8
3.1.2 The healing length . . . . . . . . . . . . . . . . . . . . 9
3.2 Characteristics of vortices . . . . . . . . . . . . . . . . . . . . 10
3.2.1 The quantum of circulation . . . . . . . . . . . . . . . 10
3.2.2 Flowspeed around a vortex . . . . . . . . . . . . . . . 11
3.2.3 Vortex rings as a Hamiltonian system . . . . . . . . . 11

4 The program 13
4.1 Overview of the program . . . . . . . . . . . . . . . . . . . . . 13
4.1.1 General characteristics . . . . . . . . . . . . . . . . . . 13
4.1.2 The dimensionless NLSE . . . . . . . . . . . . . . . . 13
4.1.3 The dimensionless velocity . . . . . . . . . . . . . . . . 14
4.2 Testing the model . . . . . . . . . . . . . . . . . . . . . . . . 15
4.2.1 Vortex core density . . . . . . . . . . . . . . . . . . . . 15
4.2.2 Perturbed ring generation . . . . . . . . . . . . . . . . 16
4.2.3 Vortex ring speed as a function of radius . . . . . . . . 16

5 Perturbed rings 20
5.1 Helical perturbations . . . . . . . . . . . . . . . . . . . . . . . 21
5.1.1 Translational velocity . . . . . . . . . . . . . . . . . . 21
5.1.2 Motion and breathing . . . . . . . . . . . . . . . . . . 21
5.1.3 Period of oscillation and breathing . . . . . . . . . . . 22
5.1.4 Stability . . . . . . . . . . . . . . . . . . . . . . . . . . 28
5.2 Planar perturbations . . . . . . . . . . . . . . . . . . . . . . . 31

6 Conclusions and further work 34

2
Chapter 1

Introduction

Vortex rings have recently emerged as the fundamental element of turbu-

lence [2], but questions about their nature have captivated physicists and
mathematicians alike for many years. Lord Kelvin postulated the existence
of vortex atoms as fundamental particles. He believed that vortex rings
had fundamental modes of oscillation which could account for certain prop-
erties of matter, such as emission spectra [3]. This was all part of the vortex
theory of matter. With all that we now know about particle physics - from
Rutherford to Heisenberg - Kelvins theories may seem far fetched, but for
a long while many questions remained unanswered: How do vortex rings
move? How does disturbing a rings structure change this motion? Due to
the limitations of the mathematical sciences at the time, Kelvin was unable
to formulate equations to answer these questions and so did not have enough
information to determine the real effects of perturbations on the motion of
vortex rings.
In this report we will first discuss some of the well known properties of
classical vortex rings in viscous and inviscid fluids. This is followed by a brief
overview of the two key methods of analysing their motion (vortex filament
models vs. NLSE models) and the results of investigations into the motion of
perturbed rings using vortex filament models (Chapter 2). Then we will give
a more in-depth description of NLSE fluids, upon which the experimental
part of the report is based (Chapter 3). After this will follow a section
detailing the characteristics and testing of the computational model used
to obtain our results (Chapter 4). Finally there will be a section detailing
a description of the motion of perturbed vortex rings obtained from the
computational model (Chapter 5) and a section of conclusions (Chapter 6).

3
Chapter 2

Classical vortex rings

A much celebrated and widely used system of equations for motion V , energy
H and momentum p is used to describe the properties of vortex rings. This
system relates the rings self induced velocity to the circulation of the fluid
around the vortex core, the radius of the ring R, the density of the fluid
and the vortex core size a. For a classical vortex ring with a hollow core
with constant pressure we have [4]:

H = 21 2 R[ln(8R/a) 3/2] (2.1)

V = (/4R)[ln(8R/a) 1/2] (2.2)
p = R2 (2.3)
This is one of a number of models for a vortex ring, many of which (but not
all) behave as Hamiltonian systems [5].
An interesting quality of the motion of vortex rings arises when they
are compared to particles. As we know, the kinetic energy of a particle of
mass m with velocity v is given by E = mv 2 /2 and so the velocity increases
for larger energies. For vortex rings, the converse is true. By increasing
R, which increases the energy, we actually reduce the velocity of the ring.
The higher the energy of a vortex ring the more slowly it moves. This is
less counter-intuitive when we consider that larger rings can be thought to
contain more fluid, and so the energy is spread more thinly.
Equation (2.2) describes the motion of classical vortex rings in perfect
fluids (i.e. fluids with zero viscosity). This has been experimentally verified
in superfluid 4 He [6]. Viscous fluids like air and water are more complex
and present issues such as the difficulty in defining the core radius and the
fact that vorticity diffuses through the fluid. Such rings can be generated
using a piston arrangement as shown in figure 2.1. A solution for these
rings describes the vortex core parameter in equation (2.2) in terms of the
viscosity and the stroke time of the piston T such that a = 4T , and the
numerical constant of 1/2 becomes 0.558, namely the Euler constant [7].

4
CHAPTER 2. CLASSICAL VORTEX RINGS 5

Figure 2.1: Diagram of a device used to generate a vortex ring in viscous fluids.

There are two main methods of studying the motion of vortices. In this
report we focus on a method which describes superfluid flows in a fundamen-
tal sense. This method utilises the non-linear Schrodinger equation (NLSE)
to model superfluids as BoseEinstein condensates (or BECs). The spatially
continuous flow allows us to determine how vortices interact with one an-
other without losing information about how they interact with the fluid in
which they exist. This method currently only describes a superfluid in the
limit of zero temperature, although new models are being developed which
incorporate a thermal cloud of excitations, thus making better predictions
about fluids described by the two fluid model, like real superfluids.
The second method is the vortex filament method. Here, we simply
take the vortices to be interacting space curves of infinitesimal thickness
and evolve the interactions of these curves over time. This method loses
some information, such as kinetic energy dissipation through the so called
vortex sound. This effect is caused by the compressibility of the fluid, a
characteristic which cannot be reproduced through interacting space curves
alone. Vortex filament models are also unable to accurately reproduce vortex
reconnections, as reconnection events are an intrinsic characteristic of the
NLSE and also cannot be reproduced directly from simple interacting space
curves. In practice, ad-hoc reconnections must be made algorithmically
based on the proximity of two vortices. This is a simplification of the true
behaviour of superfluid vortices. Vortex filament models may have no way of
incorporating the compressible nature of fluids or vortex reconnections, but
they can take into account some thermal effects, approximating interactions
with a normal fluid. In this sense, the two methods are quite different and
suit different situations well [8].
The foundation of vortex filament models is the BiotSavart Law, which
determines the self induced velocity v SI at point x of a vortex filament by
integrating the filaments interaction with itself along its length:

(x x )
I
v SI (x) = dx
4 |x x |3
CHAPTER 2. CLASSICAL VORTEX RINGS 6

Care must be taken to exclude the divergence which occurs as x x .

This equation can be approximated at leading order using the Local
Induction Approximation (LIA)

v SI (x) ln(R/aef f )s s
4
where s is the curve representing the vortex filament, primes denoting deriva-
tives with respect to arc length (given the tangent and normal vectors of the
curve, the cross product of these quantities being the binormal vector).We
also have the effective core radius aef f and the local radius of curvature
R = 1/|s | [8].
The more versatile Schwarz equation [9] can be used to incorporate the
drag effects of thermal excitations, but this case is beyond the scope of this
report.
The LIA has been used both analytically and numerically to determine
the effects of perturbations (often called Kelvin waves) on a vortex ring
[10, 11]. These investigations provide some interesting results to be verified
in the context of this report, but all agree that Kelvin waves have the effect of
slowing down and even reversing the direction of the velocity of translation
of a vortex ring. For details of the results of the numerical simulations based
on the vortex filament model see figure 2.2.
CHAPTER 2. CLASSICAL VORTEX RINGS 7

Figure 2.2: Results obtained using the vortex filament model [11]. a) A circular
vortex ring of radius R = 0.1 cm (blue) and a vortex ring perturbed
by N = 10 Kelvin waves of amplitude A/R = 0.05 (red) which move
in the z-direction. Note that the perturbed ring moves slower. b)
Kelvin waves of large amplitude (A/R = 0.35) on a vortex ring. Note
that the perturbed vortex ring moves backwards in the negative z-
direction. c) Top panel: x y view of the large amplitude vortex at
t = 0 (blue) and t = 26 s (red, outermost). For comparison, a non-
disturbed vortex is shown with dashed line (green). d) Bottom panel:
the average location of the vortex ring as a function of time. From
top to bottom the curves correspond to A/R = 0.0, 0.05, 0.10, . . . ,
0.35.
Chapter 3

Vortex rings in
BoseEinstein condensates

3.1 Overview of superfluids and the NLSE

This report is primarily concerned with the behaviour of vortex rings in
superfluid condensates like liquid 4 He at zero temperature (i.e. zero degrees
Kelvin). Superfluid heliums large interatomic spacing and low density sug-
gest a description in terms of a weakly interacting Bose gas (a BEC) [12],
and so can be described by the GrossPitaevskii or nonlinear Schrodinger
equation (referred to as the GPE or NLSE):

~2 2
i~ = + g ||2 (3.1)
t 2m
Here, is a complex quantum particle wave function for Bosons of mass
m, is the chemical potential (or single particle energy) and g denotes the
strength of the particle-particle interaction. ~ is, as usual, Plancks constant

3.1.1 The Madelung transformation

Let us look at the results of taking a solution of the form = n ei . Here
(r, t) is the phase of the wave function and the amplitude is expressed in
terms of the number density of bosons in the condensate n(r, t). This can
be shown by normalising :
Z Z

dr = n dr = N
allspace allspace

where N is the total number of bosons in the condensate. Therefore the

density of the condensate at any point is

= mn (3.2)

8
CHAPTER 3. VORTEX RINGS IN BECS 9

If we take a mass current q = v we can show

i~
q= ( )
2
= ~n = v

i.e. the velocity of the flow is given by the gradient of the phase

~
v= (3.3)
m
By substituting this Ansatz and equations (3.2) and (3.3) into (3.1) we can
derive a set of equations which describe an inviscid fluid

+ (v) = 0 (3.4)
t
 
vj vj 1 P 1 jk
+ vk = + (3.5)
t xk xj xk
Here P is the pressure defined by the equation of state P = g2 /2m2 .
Equation (3.4) is an equation of conservation and equation (3.5) describes
the motion of an inviscid Euler fluid with an extra term containing the
quantum stress tensor jk . This term is vital over small scales and explains
many of the properties inherent in superfluids which are not present in
normal Euler flows. Perhaps most importantly the quantum stress tensor
allows vortex reconnections.

3.1.2 The healing length

We can define a typical length scale (such that 2 can be written as 2 )
for fluids governed by the NLSE and take this to be the scale over which
kinetic energy and particle interactions are of the same magnitude. This is
essentially when the second and third terms of equation (3.1) balance out

 2 2 
~2
  
~   2 
  
1=   g||  =  2
 
2m   2 mg||2 
 2 
2
 ~ 
=  
2mgn 
We can say that in general the particle interactions determine the average
energy per particle and so gn = . With all of these quantities being real
and positive, we obtain the healing length

~
= (3.6)
2m
This quantity can be seen in a number of places. For instance, the core
radius of a vortex is of the order of .
CHAPTER 3. VORTEX RINGS IN BECS 10

3.2 Characteristics of vortices

3.2.1 The quantum of circulation
The vorticity of a superfluid differs to that of normal fluids in a profound
way. The circulation is computed by integrating the flow speed v around a
closed path c. In normal fluids this circulation can take any real value:
I
v dl =
c
However through application of Greens theorem and equation (3.3) we can
see that under the Madelung transformation this integral must be zero:

I Z
v dl = ( v) dS
c s
Z
~
= ( ) dS
m s
=0

as the curl of the gradient of a scalar function is identically zero. This is the
same as saying that the fluid has zero vorticity: = v = 0
This only holds true when s (the surface bounded by the curve c) is
simply connected. If there is a topological pole within the contour, as there
is when the contour encloses a vortex, then the integral need not be zero
[13]:
I I
~ ~
v dl = dl =
c m c m
By using this with the fact that is a single valued function, i.e.
i i(+)
ne = ne
1 = cos() + i sin()

we can constrain the values of such that

I
~
v dl = 2n = n n = 0, 1, 2, . . . (3.7)
c m
This tells us that the circulation around a closed loop in a superfluid can
only take discrete values n where = h/m is defined to be the quantum of
circulation. Larger values of circulation are achieved by adding vortices and
so n is related to the number of vortices within the contour, not forgetting
to take their direction of rotation into account.
CHAPTER 3. VORTEX RINGS IN BECS 11

3.2.2 Flowspeed around a vortex

Working in cylindrical polar coordinates r = (r, , z) and assuming that
the flowspeed is a function of radius and has azimuthal symmetry, such that
we can determine an equation for the fluids flowspeed
v = (0, v(r), 0) = v ,
from equation (3.7) by taking the contour c to be a circle of radius r centred
on the vortex core:

I Z 2
v dl = r d v
c 0
= 2rv

and so the rotational velocity is

n
v(r) = (3.8)
2r
A natural question to ask is what happens to the flow speed as r tends to
zero. Equation (3.8) indicates a singularity in this limit, but this is physically

impossible. Happily, we can show that n r as r 0, meaning that the
density drops in the vortex core and so the core of a superfluid vortex is
hollow [15].

3.2.3 Vortex rings as a Hamiltonian system

As has been previously mentioned, vortices in superfluids can be considered
to have hollow cores, but the nature of this core is qualitatively different
from the model described in the introduction by equation (2.2). The den-
sity in the core of a superfluid vortex approaches zero smoothly while in
the previously described model the density drops instantly to zero. For this
reason the numerical parameters in the following equations are slightly dif-
ferent. Despite this, the core is still defined as hollow and so the following
system also forms a Hamiltonian system.
The velocity of a vortex ring in a superfluid can be determined from an
asymptotic analysis to be
   
8R
V = ln 0.615 (3.9)
4R a
Again, R and a are the rings radius and vortex core radius respectively.
The full derivation of this result is beyond the scope of this report [16].
Similarly we can determine the impulse of the ring to be

p = R2 (3.10)
where is the fluid density outside of the core.
CHAPTER 3. VORTEX RINGS IN BECS 12

The kinetic energy of a vortex ring in a BoseEinstein condensate can be

determined in a more straightforward manner by expanding on the notion
of the kinetic energy per unit length (E) of a classical rectilinear vortex line.
E can be expressed as the sum of internal (EI ) and external (EE ) energies.
Again, the derivation of the internal energy of a vortex line is beyond the
scope of this report, but the total external energy of a vortex of length h can
be determined quite easily by considering the volume integral of the local
energy. Working in cylindrical polars with a vortex line aligned along the z
axis and using equation (3.8):

v 2
Z Z Z
hEE = r dr d dz
2
Z 2
 2 h
Z C
1
Z
= dz d dr
2 2 0 0 a r
2
 
C
EE = ln
4 a

Here, the vortex is taken to be in a cylindrical container of height h and

radius C. The expression for the total energy per unit length of the vortex
line is

E = EE + EI
2
   
C
= ln + 0.385
4 a

The logarithmic term corresponds to the energy exterior to the vortex core
EE and the numerical constant corresponds to the energy within the core
EI . This constant was obtained through a numerical calculation [16].
By working with small values of a/R (in that the vortex core radius is
much smaller than the radius of the ring) we can simply define the energy
of a vortex ring as E = 2RE with C = 8R/ e2 and so we obtain
   
1 2 8R
E = 2 R ln 0.615 (3.11)
a
This result can be obtained more rigorously using a method similar to that
involved in determining V .
It is easily seen that equations (3.9) (3.11) obey the Hamiltonian rela-
tion V = E/p. It is also worth pointing out that the results for E and V
are to be regarded as providing the leading pairs of terms in the asymptotic
expansions of these quantities as a/R 0 and so are in principle exact.
Chapter 4

The program

4.1 Overview of the program

4.1.1 General characteristics
The results discussed in the next chapter of this report are obtained from a
program written in Fortran 90 to numerically solve the 3D NLSE [14]. The
code is quite versatile and can solve a number of forms of the equation using
various time stepping methods for various initial and boundary conditions.
The code is parallelised with the Message Passing Interface (MPI). This
allows the simulations to be run on a cluster, allowing a shorter run time.
Here, the program is set to use fourth order centred finite differences in
space, fourth order adaptive explicit RungeKuttaFehlberg time stepping
and reflective boundary conditions. The spatial grid is a 1283 mesh in a
cubic box which is also of size 1283 (i.e. 64 < x, y, z < +64).

4.1.2 The dimensionless NLSE

With many equation solving programs it is necessary to work with a di-
mensionless form of the equation you wish to solve. Here, we must make
the NLSE dimensionless using dimensionless variables r , t and scaled in
terms of , g, m and ~. The exact nature of this scaling must be determined
by substitution into (3.1), starting with unknown scaling parameters ,
and 0 such that

r = r t = t = 0
After substitution, cancellation and division by we obtain

~2
     2
~ g0
2i = 2
2
+ | |2
2 t 2m
By taking the terms in brackets to be dimensionless and equal to one we
obtain scales for the length, time and wave function,

13
CHAPTER 4. THE PROGRAM 14

r
~ ~
= = 0 =
2m 2 g
and a dimensionless equation

= 2 + 1 | |2


2i (4.1)
t

We can see that the length scale is equal to the healing length of the super-
fluid. The above equation is identical to that which is used in the Fortran
code.

4.1.3 The dimensionless velocity

In order to compare the behaviour of the program with known results, we
need to determine the dimensionless form of the equation for a rings self
induced velocity. By constructing equation (3.9) as a function of dimen-
sionless variables (and a parameter to describe the core size in terms of the
effective core radius aef f )

2
V = V R = R a = aef f =


we obtain the equation for the dimensionless velocity V :
 
8R
V = ln 0.62
4R aef f
We can see from this that the units in the equation are all self consis-
tent. As such, the quantities V and R are those which will be measured
directly from the model. The quantity must be defined in terms of the
dimensionless form of Plancks constant h :
h 1 2m
= = h
m m
2
= h

= h = 2
We can also argue that the effective core radius is roughly equal to 1.5 (see
section 4.2.1), and so aef f = 1.5, yielding
1

VN = ln R 1.059 (4.2)
2R

Similarly, by starting with equation (2.2) we can develop a dimensionless

equation for the velocity of a ring with a hollow core and constant volume.
1

VH = ln R 1.174 (4.3)
2R
CHAPTER 4. THE PROGRAM 15

4.2 Testing the model

Here we compare some of the known characteristics of superfluids to results
we obtain from the program. One of the main parameters to be tested is the
resolution, which we define as the the number of grid points j at which the
program evaluates the equation. The results obtained in the next chapter
are been obtained by running the program with a 1283 spatial mesh, i.e.
j = 128, in a 1283 box. Another useful quantity to consider is the distance
between neighbouring grid points x. Our simulations are run with x = 1.

4.2.1 Vortex core density

As stated above in section 3.2.2, the density of a superfluid tends to zero
in the core of a vortex. We verify this by producing a plot of the density
through the core of a straight line vortex. Figure 4.1 illustrates that the
model reproduces this drop in density.
It is important to determine whether or not the program is running at a
high enough resolution to accurately model the behaviour of the fluid in the
core. We must also show that changing the resolution does not drastically
alter the way that the program models the core. We ensure this by showing
that there are a large number of grid divisions in the vortex core, and that the
same core shape is generated by a number of different resolutions. Figure 4.1
shows that there are up to 14 grid points within the vortex core for j = 128,
which is a number high enough to ensure realistic behaviour. We also see
that the model behaves similarly for resolutions greater than j = 64, with
the points generated by j = 128 lying on same curve as those generated
by j = 256. This shows that j = 128 is sufficiently high to ensure a good
qualitative description of the vortex core.
Figure 4.1 also gives us an idea of what the vortex core radius is. Many
of the diagrams in this report show isosurfaces of the number density ||2 of
the flow, similar to part a) of figure 4.2. The isosurfaces for these diagrams
(and for some velocity calculations) were taken at ||2 = 0.4 and so the core
radius in these diagrams can be measured to be 1.3. For the majority of
the velocities calculated later similar isosurfaces, with ||2 = 0.3, are used
to determine the position and displacement of the ring, giving a radius of
1. These figures are useful for calculating errors. When ||2 = 0.5 the
flow has dropped to half of the bulk density and so the effective core radius
aef f can be taken to be 1.5.
CHAPTER 4. THE PROGRAM 16

1.2

1.0

0.8

0.6
||2

0.4

0.2

0.0
20 10 0 10 20
x

Figure 4.1: Graph showing the density ||2 of a slice of the fluid through a vortex
core at x = 0, generated by the program. The +, and symbols
correspond to j = 64, 128 and 256 respectively. We can see that the
points all map out the same curve within the vortex core, and so all
three resolutions model the vortex in the same way.

4.2.2 Perturbed ring generation

We must also show that our program is able to produce something which
qualitatively resembles a vortex ring. By plotting the density and phase
profiles of a flow containing a vortex ring and an isosurface plot of the flow,
we can see that the program reproduces known characteristics of a superfluid
vortex (see figure 4.2). Given that the model is able to create vortex lines
and vortex rings, we must also show that it can deform these structures into
helical and planar perturbed configurations. This is shown in figure 4.3.

4.2.3 Vortex ring speed as a function of radius

The majority of the work in the following chapter is concerned with anoma-
lous translational velocities of vortex rings, so we must first establish that
our program is able to produce results consistent with standard unperturbed
situations. We must test the programs ability to produce results which agree
with equations (3.9) and (2.2) using equations (4.2) and (4.3). Figure 4.4
shows that our model fits analytical predictions well.
CHAPTER 4. THE PROGRAM 17

Figure 4.2: a) The end on view of a 3D isosurface plot of the fluid density for
||2 = 0.4. The vortex ring is clearly visible. b) A contour plot of the
density of a slice down the middle of the ring. The two dark circles
clearly show where the ring passes through this slice, illustrating the
density drop in the vortex core. c) A contour plot of the phase through
the same slice as in b). The phase clearly varies from to around
the vortex cores, retrieving the result from section 3.2.1.
CHAPTER 4. THE PROGRAM 18

Figure 4.3: Numerically calculated perturbed vortex structures generated by the

program, shown from end, side and perspective views. The diagram
shows isosurfaces of the number density ||2 = 0.4 for flows con-
taining various perturbed structures. a) A vortex line with a planar
perturbation. b) A vortex line with a helical perturbation. c) A vor-
tex ring with a planar perturbation d) A vortex ring with a helical
perturbation.
CHAPTER 4. THE PROGRAM 19

0.25

0.20

0.15
V

0.10

0.05

0.00
5 10 15 20 25 30 35 40
R
Figure 4.4: Graph to compare the ring velocities obtained from the program com-
pared to velocities derived from (4.2) and (4.3), taking the effective
core radius to be aef f = 1.5. Solid line with + symbols: Numerically
generated data from program. Short dash line: Predictions based on
a dimensionless form of classical vortex ring motion with a hollow
core at constant pressure [4]. Long dash line: Predictions made by
P. H. Roberts based on a dimensionless form of vortex ring motion
derived from the NLSE [16].
Chapter 5

Perturbed rings

In this chapter we will study two types of perturbation: helical and planar.
These perturbations both have a regular periodic form, but one could be
considered to be the general form of a regular perturbation, while the other
is a particular case.
We can describe a ring of radius R, with a perturbation of amplitude A
and mode M , lying in the x-y plane as a parametric curve in rectangular
cartesian coordinates, given by (s) = (x(s), y(s), z(s)). The parameter s
runs from 02.
The helical perturbation (general form of a regularly perturbed ring)
takes a form described by

(R + A sin (M s)) cos (s)
H (s) = (R + A sin (M s)) sin (s) (5.1)
A cos (M s)
The planar ring has the same structure but with no z component:

(R + A sin (M s)) cos (s)
P (s) = (R + A sin (M s)) sin (s) (5.2)
0
The forms of these parametrisations can be seen in sections d) and c) re-
spectively of figure 4.3.
As our goal is to quantify the motion of these vortex rings for various
perturbations, we must first select a basic, unperturbed form. We will work
using rings with unperturbed radius R = 25. These rings move with self
induced translational velocity V0 = 0.0857. Later we will quantify a per-
turbations amplitude by scaling it against the vortex rings radius, so the
quantity A/R will often be stated as a key variable. We will also look at the
perturbed rings velocity V scaled by the unperturbed velocity, i.e. V /V0 .
Note we are now working exclusively in dimensionless quantities

20
CHAPTER 5. PERTURBED RINGS 21

These velocities are obtained from graphical representations of the nu-

merically generated data in the form of isosurface plots. By making isosur-
face plots of ||2 = 0.3 we can see the structure of the vortex core (c.f. figure
4.2 part a)). By observing the rings initial position (from a side on view)
and then the comparing this to its position at a suitably chosen later time,
we can determine the displacement and resultant average velocity. This dis-
placement is determined within an error of the order of the core radius. The
core radius for ||2 = 0.3 is 1.

5.1 Helical perturbations

5.1.1 Translational velocity
It can be shown quite easily that in a system governed by the NLSE, helically
perturbed vortex rings behave in a manner similar to the predictions made
by vortex filament models. Perturbations with higher modes and amplitudes
cause the ring to slow more and eventually cause the direction of motion to
reverse. The higher the mode and amplitude, the more pronounced this
effect becomes (see figure 5.1 and table 5.1). For each value of M studied
here the largest amplitude corresponds to the point at which the vortex ring
reconnects before undergoing a single breath (see section 5.1.2 for a descrip-
tion of the breathing phenomenon). For the results to be meaningful it is
necessary to take readings of the rings displacement at the point between
breaths corresponding to parts a) or e) in figure 5.2.

5.1.2 Motion and breathing

Large amplitude helical Kelvin waves have an interesting characteristic in
their motion. Just as the ring has self induced motion, the curves of the helix
induce yet more complicated motion. Instead of simply travelling around the
ring, the helical waves undergo a periodic distortion. This periodic distortion
will hereafter be referred to as a breath, as the ring has an appearance of
breathing as it moves.
As the ring breathes, the perturbations elongate outwards and forwards
in the direction of unperturbed translational velocity, forming pronounced
lobes. These lobes have a structure similar to vortex rings, so it is not sur-
prising that they have their own self induced motion. As this elongation
increases, the lobes twist such that the plane in which they lie is perpendic-
ular to the plane of the ring. When fully extended the lobes are distorted
radially outwards from the ring, no longer extended in front of the ring. The
lobes then begin to collapse backwards and the perturbation returns to its
original helical configuration. During their collapse, the lobes extend quite
considerably in the direction opposite to the velocity of the unperturbed
ring. It is during this phase of the breathing process that the rings back-
CHAPTER 5. PERTURBED RINGS 22

wards motion occurs. The lobes actually drag the ring backwards. This is
illustrated in figure 5.2.
This periodic change of motion is analogous to the waves seen on the
lines in part d) of figure 2.2. We also find that for smaller perturbations
this breathing is less pronounced, a quality also illustrated by figure 2.2.
Although rings with a smaller perturbation do not exhibit this breathing
so dramatically, the motion of the Kelvin waves is still not simply regularly
around the ring.

5.1.3 Period of oscillation and breathing

Helically perturbed vortex rings have two distinct types of oscillation.
The first is the standard oscillation expected of a Kelvin wave and corre-
sponds to the speed with which the wave moves around the circumference of
the ring. The frequency of these waves has been determined for straight line
vortices in the case of vortex lines. However, this result is only applicable
for suitably large wavelengths, and is not applicable here [8].
The second is the aforementioned breathing process. As the ring breathes
the elongation of the lobes causes the total length of the vortex line to
increase and decrease periodically. Figures 5.3 to 5.5 illustrate this.
The breathing period can be seen to be longer for larger perturbations.
This seems to be because larger perturbations have larger lobes, which can
be considered to be vortex rings in their own right. As these lobes are larger,
they move more slowly and so the rings breaths are longer. This can be
seen in table 5.1
CHAPTER 5. PERTURBED RINGS 23

1.2
1
0.8
0.6
V/V0

0.4
0.2
0
-0.2
-0.4
0 0.1 0.2 0.3 0.4 0.5 0.6
A/R

Figure 5.1: The graph shows how the average translational velocity of a vortex
ring changes when the amplitude A of the perturbation increases.
The quantities plotted are the rings average velocity scaled by the
unperturbed velocity V /V0 and the perturbation amplitude scaled
by the rings unperturbed radius A/R . From top to bottom the
lines show perturbations with modes M = 2(), 4(), 6(), 8(),
10(+). Only rings with higher modes of oscillation were able to move
backwards.
CHAPTER 5. PERTURBED RINGS 24

M A/R V /V0 PB P TD TR
4 0.120 0.920 1040 >750 >750
4 0.182 0.814 1060 >750 >750
4 0.260 0.623 1088 >750 >750
4 0.324 0.434 350 1248 >750 >750
4 0.396 0.369 445 1428 >750 >750
4 0.464 0.184 565 1720 >750 >750
6 0.044 0.979 780 >1000 >1000
6 0.101 0.895 780 >1000 >1000
6 0.149 0.735 800 >1000 >1000
6 0.210 0.467 860 >1000 >1000
6 0.276 0.274 265 960 >1000 >1000
6 0.348 0.072 340 1110 >1000 >1000
6 0.410 0.003 485 1397 >1000 >1000
8 0.110 0.733 640 770 >1000
8 0.139 0.619 667 480 >1000
8 0.168 0.489 667 440 >1000
8 0.198 0.316 680 480 650
8 0.231 0.183 190 746 370 680
8 0.242 0.115 210 782 440 730
8 0.263 0.035 230 817 460 620
8 0.273 -0.030 240 832 510 630
8 0.290 -0.062 265 896 540 680
8 0.302 -0.093 295 960 610 730
10 0.029 0.969 840 >1000 >1000
10 0.058 0.890 800 >1000 >1000
10 0.073 0.827 666 480 >1000
10 0.101 0.721 640 300 >1000
10 0.118 0.556 600 270 >1000
10 0.154 0.395 580 240 470
10 0.174 0.191 560 260 440
10 0.202 0.052 145 560 270 310
10 0.233 -0.149 190 540 300 450
10 0.258 -0.240 225 540 475 540
10 0.284 -0.231 280 540 330 430

Table 5.1: Table of results showing how the scaled velocity V /V0 of a helically per-
turbed vortex ring changes for perturbations of mode M and amplitude
to radius ratio A/R. Also listed is the the period of the perturbations
breath PB and the period of standard oscillation P . denotes per-
turbations which did not show discernable breathing. Also shown are
TD , the time after which the ring begins to distort, and TR , the time
at which the ring reconnects with itself. Where distortions or recon-
nections are not observed for the duration of the simulation TD and
TR are given to be >runtime of simulation.
CHAPTER 5. PERTURBED RINGS 25
CHAPTER 5. PERTURBED RINGS 26

Figure 5.2: (see previous page) Numerically calculated isosurfaces generated by

the program, illustrating the breathing motion of a large amplitude
helically perturbed vortex ring with M = 8. End, side and perspective
views shown. The diagram shows isosurfaces of the number density
||2 = 0.4. a) The initial configuration at t=0. The shape of the
Kelvin wave is regular and helical. b) The the ring at t=100. The
perturbations have begun to become distorted due to the non-linear
self induced motion. Lobes have begun to form and are extended
radially outwards and in the direction of unperturbed motion. c) The
ring at t=150. The perturbation lobes have reached their maximum
elongation, extended out in the plane of the ring. d) The ring at
t=200. The lobes are becoming less distorted and are now extended
radially outwards and in the opposite direction to the direction of
unperturbed motion. e) At time t=300 the ring has returned to its
original configuration.

600
550
500
450
400
L

350
300
250
200
150
0 100 200 300 400 500 600
t

Figure 5.3: Vortex line length L over time t for vortex rings of radius R = 25
perturbations of mode M = 10. From top to bottom the lines cor-
respond to perturbations of amplitude A/R = 0.284, 0.258, 0.233,
0.202, 0.174, 0.154, 0.118, 0.101, 0.073, 0.058, 0.029. The sudden
drops on the sixth and seventh lines from the bottom indicate where
the perturbation has become unstable and the structure of the helix
has collapsed and become irregular. The sharp cusp in the second
line at t 540 down corresponds to a reconnection event.
CHAPTER 5. PERTURBED RINGS 27

500

450

400

350
L

300

250

200
0 100 200 300 400 500 600
t

Figure 5.4: Vortex line length L over time t for vortex rings of radius R = 25
perturbations of mode M = 8. From top to bottom the lines corre-
spond to perturbations of amplitude A/R = 0.302, 0.290, 6.83, 0.273,
0.242, 0.231, 0.198, 0.168, 0.139, 0.110. Notice that the oscillations
decrease in both size and period as the amplitude of the Kelvin wave
decreases.

500

450

400

350
L

300

250

200

150
0 100 200 300 400 500 600
t

Figure 5.5: Vortex line length L over time t for vortex rings of radius R =
25 perturbations of mode M = 6. From top to bottom the lines
correspond to perturbations of amplitude A/R = 0.410, 0.348, 0.276,
0.210, 0.094, 0.0440. Notice that for this smaller value of M there is
no discernible breathing in most cases
CHAPTER 5. PERTURBED RINGS 28

5.1.4 Stability
As one would expect, the more drastically the ring is deformed the more un-
stable it becomes. As the size of the perturbation increases, the structure of
the helix becomes susceptible to distortions and eventually the Kelvin waves
decay into more random perturbations (see figure 5.6). When instabilities
on the ring cause it to become distorted in this way some of the Kelvin
waves grow in amplitude. These waves are likely to reconnect and separate
from the main ring.
For a helical perturbation the larger the amplitude the more the lobes
twist and so when the lobes are fully extended half way through each breath,
the ring is very susceptible to reconnections. This is why the rings are less
stable when the perturbation is great enough to cause backwards motion.
With higher modes the lobes are more densely packed, also making the rings
more likely to undergo reconnections (see figure 5.7).
It is interesting to note that, up to a point, rings with larger perturba-
tions can retain their structure for longer. This is because instabilities and
distortions tend to occur at particular points in the breathing process, so
the slower this breathing the longer the ring can retain its regular helical
perturbation. This can be seen in table 5.1.
CHAPTER 5. PERTURBED RINGS 29

Figure 5.6: Numerically calculated isosurfaces generated by the program, illus-

trating the distortion of a large amplitude helically perturbed vortex
ring. End, side and perspective views shown. The diagram shows
isosurfaces of the number density ||2 = 0.4 of a ring which originally
has perturbation mode M = 8 and amplitude A/R = 0.231. a) The
ring at t=420 after undergoing two breaths. Minor asymmetries have
begun to arise. b) The ring at t=590. The asymmetry has become
a large scale distortion c) The ring at t=650. The distortions have
caused a new complex structure to arise. There is still some symme-
try to the perturbation structure. d) The ring at t=680. Small loops
have formed. e) The ring at t=720. The loops have reconnected. The
original regular helical structure is, by this point, completely lost.
CHAPTER 5. PERTURBED RINGS 30

Figure 5.7: Numerically calculated isosurfaces generated by the program, illus-

trating the reconnection of a large amplitude helically perturbed vor-
tex ring. End, side and perspective views shown. The diagram shows
isosurfaces of the number density ||2 = 0.4 of a ring which originally
has perturbation mode M = 10 and amplitude A/R = 0.303. a) The
ring at t=0. The perturbation initially has a regular helical structure.
b) The ring at t=50. The Kelvin waves have begun to elongate in the
normal breathing motion. c) The ring at t=150. The lobes have ex-
tended and twisted to the point that reconnections are beginning to
occur. d) The ring at t=200. The lobes have fully separated from the
ring, leaving behind a ring with small regular helical perturbations.
CHAPTER 5. PERTURBED RINGS 31

5.2 Planar perturbations

Planar perturbations do not exhibit the drastic reduction in velocity that
their helical counterparts show. Figure 5.8 shows that any observed reduc-
tion in motion is often within the error of the observation and so cannot
be taken to be significant. Even if this were not the case and the errors
were much smaller, the reduction in velocity is considerably smaller than
that shown by rings with helical perturbations. Planar perturbed rings are
unlikely to reverse their direction as a suitably high amplitude would un-
doubtedly cause immediate reconnections (see table 5.2 for details).
Planar perturbations also lack the breathing motion exhibited by helical
perturbations. In fact, the waves do not travel around the circumference of
the ring at all. The amplitude of the perturbation does, however, move from
being positive or negative in the plane of the ring to being in the direction
of motion, and then continues to being negative or positive in the plane of
the ring. In other words, the perturbation oscillates around the vortex core,
but retains the same amplitude and position on the ring. This motion is
illustrated in figure 5.9, and the periods of the oscillation PO are given in
table 5.2.
This simpler motion may explain why there is little or no change in
translational velocity of the ring. Looking at the end on view of the ring
we can see that the parts of the perturbation which increase and decrease
the rings area change in size as the perturbation oscillates. These areas
cancel each other out and the perturbed ring has a roughly constant net
area comparable to an unperturbed ring of the same radius. As the rings
area and radius are linked, the area and translational velocity can be thought
to be similarly linked. A more in depth analysis of the rings local velocity
could confirm this, but is beyond the methods of this report.
Another result of the simpler motion resulting from the perturbations
structure is an increased stability. The helical perturbed rings complicated
oscillations made it susceptible to instabilities and resultant reconnections,
but planar perturbed rings have a far simpler mode of oscillation, and so
are far less susceptible to instabilities and reconnections. Indeed, whenever
the rings do not immediately reconnect upon generation, they retain their
structure until the end of the simulation.
CHAPTER 5. PERTURBED RINGS 32

1.2

1.1

0.9
V/V0

0.8

0.7

0.6

0.5
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5
A/R

Figure 5.8: The lines show how the average translational velocity of a vortex ring
changes when the amplitude of the planar perturbation increases. The
velocity and amplitude are scaled by the unperturbed velocity and
radius. From top to bottom the lines show perturbations with modes
M = 2(),4(), 6(), 8(+). Notice how none of the rings ever achieve
backwards motion, and how the variation is not significantly greater
than the errors involved.

M A/R V /V0 PO M A/R V /V0 PO

2 0.080 1.021 860 6 0.080 0.982 140
2 0.160 0.983 880 6 0.160 0.919 160
2 0.240 0.958 900 6 0.240 0.870 190
2 0.320 0.922 920 6 0.320 0.794 220
2 0.400 0.884 980 6 0.400 0.758 260
2 0.480 0.899 >1000 6 0.480 0.785 300
4 0.080 1.020 250 8 0.080 0.981 90
4 0.160 0.979 270 8 0.160 0.943 110
4 0.240 0.915 310 8 0.240 0.880 130
4 0.320 0.835 350 8 0.320 0.854 160
4 0.400 0.792 400
4 0.480 0.738 460

Table 5.2: Table of results showing how the scaled velocity V /V0 of a planarly per-
turbed vortex ring changes for perturbations of mode M and amplitude
to radius ratio A/R. Also listed is the the period of the perturbations
oscillation PO .
CHAPTER 5. PERTURBED RINGS 33

Figure 5.9: Numerically calculated isosurfaces generated by the program, illus-

trating the oscillatory motion of a planar perturbed vortex ring. End,
side and perspective views shown. The diagram shows isosurfaces of
the number density ||2 = 0.4. a) The starting configuration at
t=616. b) The ring at t=633. c) The ring at t=650. d) The ring at
t=665. e) The ring at time t=682.
Chapter 6

Conclusions and further

work

We have shown through numerical simulations of perturbed vortex rings as

solutions of the NLSE that Kelvin waves on vortex rings have the effect of
reducing the translational velocity of the vortex ring.
For helical perturbations, this reduction is dramatic to the point that the
direction of motion can be reversed, if only briefly before instabilities cause
the ring to break down and reconnect. In the case of planar perturbations
this reduction is small in magnitude, if it exists at all.
The method (the NLSE) has served to confirm previously observed re-
sults obtained from vortex filament models. We have also given more accu-
rate information in terms of stability and reconnections, as vortex filament
models cannot inherently reproduce these effects.
Further areas of work could include investigating the following:

The effects of much higher modes on larger rings, to see if reversed

motion can be achieved with more stable rings. This was not possible
for this report due to computational limitations.

The effects of noisy perturbations. The perturbations covered here

were simple sinusoidal waves. In turbulent flows, the Kelvin waves are
unlikely to be this regular.

The behaviour of perturbed vortex bundles. Large scale vortex rings

in BECs are essentially bundles of of quantised rings. It may be inter-
esting to consider the stability and motion of these large scale config-
urations.

34
Bibliography

[1] http://www.eng.warwick.ac.uk/es2089/turntabpics/builtank.html

[2] K. Shariff: Vortex rings, Annu. Rev. Fluid Mech. 24, 235 (1992)

[3] Proceedings of the Royal Society of Edinburgh, Vol. VI, 1867, pp. 94-
105.Reprinted in Phil. Mag. Vol. XXXIV, 1867, pp. 15-24.

[4] R. J. Donnelly: Quantized Vortices in Helium II, s1.6 (1991)

[5] R. J. Donnelly, I. S. Sullivan, J. J. Niemela, R. E. Hershberger,

D. Bolster: Dynamics of thin vortex rings, J. Fluid Mech. 609, 319
(2008)

[6] G. W. Rayfield, F. Reif : Quantized vortex rings in superfluid helium,

Phy. Rev. 136 : 5A, A1194 (1964)

[7] P. G. Saffman: The velocity of viscous vortex rings, Stud. Appl. Math.
49, 625 (1970)

[8] C. F. Barenghi, R. J. Donnelly: Vortex rings in classical and quantum

systems, Fluid Dyn. Res. 41, 051401 (2009)

[9] K. W. Schwarz: Three-dimensional vortex dynamics in superfluid 4 He:

Homgeneous superfluid turbulence, Phys. Rev B 38.4, 2398-2417 (1988)

[10] L. Kiknadze Y. Mamaladze: The waves on the vortex ring in HeII, J.

Low Temp. Phys. 126, 321 (2002)

[11] C. F. Barenghi, R. H anninen, M. Tsubota: Anomolous translational

velocity of vortex ring with finite-amplitude Kelvin waves,
Phys. Rev E74, 046303 (2006)

[12] P.H. Roberts, N. G Berloff: The nonlinear Scrodinger equation as a

model of superfluidity, Quantized vortex dynamics and superfluid tur-
bulence part V p235 (2001)

[13] R. J. Donnelly: Quantized Vortices in Helium II, section 2.8.3 (1991)

35
BIBLIOGRAPHY 36

[14] S. Z. Alamri, A. J. Youd, C. F. Barenghi: Reconnections of superfluid

vortex bundles, Phys. Rev. Lett. 101, 215302 (2008)

[15] E. P. Gross: Hydrodynamics of a superfluid condensate, J. Math. Phys

4, 195 (1963)

[16] P. H. Roberts: Motions in a Bose condensate, J. Phys. A: Gen. Phys.

4, 55-72 (1971)