UNIVERSITY OF CINCINNATI
______, 20 _____
I,______, hereby submit this as part of the requirements for the degree of:
______in: ______It is entitled: ______
Approved by: ______CURVATURE DEPENDENCE OF CLASSICAL SOLUTIONS EXTENDED TO HIGHER DIMENSIONS
A dissertation submitted to the
Division of Research and Advanced Studies of the University of Cincinnati
in partial fulfillment of the requirements for the degree of
DOCTORATE OF PHILOSOPHY (PH.D.)
in the Department of Physics of the College of Arts and Sciences
2003
by
Athula Herat
B.Sc., University of Colombo, Sri Lanka, 1994 M.S., University of Cincinnati, 1996
Committee Chair: Dr. Peter Surnayi Abstract
I study the curvature dependence of extended classical solutions of interacting field theories. The main motivation behind my study of classical solutions is confinement.
It has become apparent that classical objects (monopoles and vortices) are responsible for color confinement in QCD. Therefore, studying the properties of these objects is
extremely important. The main goal of this investigation is to accurately determine
the energy of extended classical solutions. In most applications it is assumed that the
energy depends simply on the length of the object, irrespective of the shape. I show
that the curvature effects are highly non-trivial.
I start with the kink solutions of the (2+1) dimensional linear sigma model. In
particular, I show that the curvature energy of a kink in two spatial dimensions, as
a prototype of extended classical solutions, is always negative. Assuming that the
deviations of the kink from the straight line are small, I derive a closed form for
the curvature energy. This investigation clearly demonstrates that the energy of the
(2+1) dimensional kink has a positive length term and a negative curvature term.
Next I look at extended vortex solutions. Vortices are (2+1) dimensional soliton
solutions of gauge theories coupled to complex scalar Higgs field. In the trivially
extended vortices the core of the vortex forms a straight line. Recent lattice studies
show that vortices appear not be straight. Therefore, it is important to study the
curvature effects of vortex solutions. However, unlike in the case of the kink, no
analytical vortex solutions exist, which makes the study of curved vortices an ex-
i tremely complex one. This problem can be made less daunting if circular vortices are considered. Therefore, I study the circular vortex of the U(1) gauge theory. I use numerical techniques to obtain the circular vortex configurations that minimize the Hamiltonian. I show that the curvature energy of the circular vortex is negative, which clearly demonstrates that the total energy of the (3+1) dimensional circular vortex is less than the length energy. I.e. the extended abelian vortex prefers the curved state to the straight one.
ii iii Acknowledgments
Throughout my graduate student life at the Department of Physics, University of
Cincinnati, I have have been extremely lucky to have had the guidance, support, and the friendship of many people. I would like to take this opportunity to extend my gratitude to all of them.
First and foremost, I would like to thank my advisor Dr. Peter Suranyi for giving me the opportunity to work with him. His underlying support through the years, academically, professionally, as well as personally, has made it possible for me to achieve this goal. It has been a pleasure to work under his guidance and I am indebted him for his patient help throughout the course of my research. I am extremely grateful to Dr. Rohana Wijewardhana for being both a mentor and a friend. His guidance and assistance from the very first day I set foot in Cincinnati, especially during times when the going was tough, has been invaluable to me. I would like to extend a special thanks to Dr. Frank Pinski for his invaluable advice and assistance with the numerical aspects of the research. I thank Dr. Brian Meadows and Dr. Mark Jarrell as well as all of the above for serving in my dissertation committee.
It has been a great privilege to work and interact with my talented and gifted colleagues and friends. I thank them all for their cooperation and their entertaining company. I am grateful to David Will for helping me solve every single computer issue that came up. I would also like to thank Dr. Mike Sokoloff for his help in various aspects of my stay at the Department of Physics.
iv Without the dedication and the support of my family, I would not have had the courage to undertake this endeavor. I owe a special debt of gratitude to my wonderful
wife, Manori, for her loving support and for the countless sacrifices she has made over
the years so that I could achieve my dream. To my beautiful daughter Hiruni, I
am so glad that you came into my life during this entire process. I could not have
completed this effort without the both of you. I am also grateful to my sister and
brother-in-law (Tamara and Sampath Jayawardane) and my parents-in-law (Walter
and Indrani Singhabahu) for their love and support.
I would like to dedicate this work to my parents (Samson and Ramani Herat) who
have been a constant source of encouragement and inspiration to me throughout my
life. I am grateful to them for their warmth, support and guidance.
v Contents
1 Introduction 5
1.1 Confinement ...... 6
1.2 Models of Confinement ...... 7
1.2.1 Dual superconductor model of confinement ...... 8
1.2.2 Abelian Projection ...... 11
1.2.3 Abelian Projection on a Lattice ...... 14
1.2.4 Center vortex model of confinement ...... 18
1.2.5 Monopoles vs. Vortices ...... 23
2 The Undulating Kink 25
2.1 Introduction ...... 25
2.2 Kink solution ...... 27
2.3 Extension to 2+1 dimensions ...... 29
2.4 Exact solutions for small deviations from the straight line ...... 32
2.5 Non-infinitesimal undulation ...... 36
1 3 The Circular Vortex 44
3.1 Introduction ...... 44
3.2 The Straight Vortex ...... 46
3.2.1 Solutions of the straight Vortex ...... 52
3.2.2 Results ...... 55
3.3 The Circular Vortex ...... 57
3.4 Numerical Treatment ...... 62
3.4.1 Variational Method ...... 63
3.4.2 Jacobian Iteration Method ...... 64
3.5 Results ...... 65
4 Conclusions 68
4.1 Classical Objects and Confinement ...... 69
4.1.1 ’t Hooft’s dual superconductivity model ...... 69
4.1.2 Abelian Projection ...... 70
4.1.3 Center Vortex Model of Confinement ...... 72
4.2 The Undulating Kink ...... 75
4.3 The Circular Vortex ...... 77
2 List of Figures
1.1 (a) The Abrikosov string between the monopoles in the superconductor and
(b) an analogue of the Abrikosov string between the chromo-electrically
charged particles in the dual superconductor...... 9
1.2 Abelian and nonabelian potentials (with self energy V0 subtracted), Ref. [17]. 16
1.3 The abelian potential (diamonds) in comparison with the photon contribu-
tion (squares), the monopole contribution (crosses) and the sum of these
two parts (triangles), Ref. [17]...... 17
1.4 (a) The ”Spaghetti vaccum” of condensed vortices (b) The ”monopole vacumm”
appears as a result of abelian projection [25] ...... 24
2.1 (a) A plot of the static kink(anti-kink) solution (2.4) (b) The energy density
of the kink (2.5). It is localized...... 29
3.1 Plot of the straight vortex field configurations for different values of the
coupling constant. (a) β = βc = 1/4, critical coupling. (b) β = 3βc. (c)
β = 2βc. (d) β = βc/2. (e) β = βc/3...... 56
3 3.2 Plot of the variation method approximate circular vortex field configurations
for critical coupling, β = βc = 1/4 ...... 64
3.3 Plot of the circular vortex field configurations for different values of the
coupling constant. (a) β = βc = 1/4, critical coupling. (b) β = 3βc. (c)
β = 2βc. (d) β = βc/2. (e) β = βc/3...... 66
4 Chapter 1
Introduction
One of the major developments of twentieth century physics has been recognition that all the known interactions can be described in terms of gauge theories [1]. Gauge the- ories in general exhibit a rich spectrum of finite energy classical solutions. Of these vortices, monopoles, and instantons are the best known topological solutions in 2, 3, and 4 dimensions. These classical objects play a major role in describing a wide vari- ety of phenomena such as magnetic field lines in type II superconductors [3], cosmic strings in grand unified theories [4], and objects responsible for confinement in dy- namical models of nonabelian gauge theories [5][6], to name a few. Our investigations of classical solutions extended to higher dimensions have been manily motivated by the recent success of these dynamical models of confinement.
5 1.1 Confinement
Ample experimental evidence exists to support the fact that the strongly interacting particles, hadrons (i.e. protons, neutrons,etc.), are made of more elementary con- stituents called quarks. There is further evidence that, in addition to their electric charge, quarks carry an additional quantum number called color. In fact, quarks come in three colors. A non-abelian gauge theory, Quantum ChromoDynamics(QCD) ex- plains the interactions between quarks as due to the exchange of a set of particles called the gluons. This theory is similar to Quantum Electrodynamics (QED) that explains the Electromagnetic interactions between electrically charged particles as due to the exchange of a particle, the photon. The main difference is that gluons interact among themselves while the photons do not. The self interactions between gluons play an important role in determining the nature of the color force.
The strength of the strong force between quarks is given by the gauge coupling strength. Perturbation theory, which is the mathematical technique used in solving quantum theories, can only be applied in the regime where the coupling strength is relatively week. Perturbative calculations in Quantum Chromodynamics indicates that the gauge coupling becomes weak at short distance scales. At zero separation, quarks act like a set of free particles. This behavior is termed asymptotic freedom.
QCD is well understood in this perturbative regime and agrees with all known ex- periments. On the other hand, at long distances the coupling becomes large and
6 perturbation theory fails. Understanding the long distance behavior of QCD is ex- tremely important if we want to consider it as the theory of strong interactions.
Experimental evidence indicates that any viable theory of strong interactions should also contain one important feature, confinement. Free quarks or any other ob- jects carrying the quantum number color have never been observed outside hadrons.
It seems that quarks are confined inside hadrons at a distance scale of about one
15 Fermi, i.e, 10− meters. This feature is referred to as confinement. There is still no analytical proof of confinement in QCD. Confinement is a long distance phenomena
and perturbation theory cannot be used to explore it as the gauge coupling becomes
strong in this regime. Therefore, it is extremely important to understand how QCD
could be analyzed in its non perturbative regime.
Although an analytical proof is absent it is evident from results of numerical
simulations that confinement is indeed present in QCD. Wilson [2] in his elegent
formulation of QCD on a space-time lattice illustrated the existence of confinement
in the strong coupling limit.
1.2 Models of Confinement
As mentioned previously, confinement is one of the most important, yet unsolved,
aspects of non-abelian gauge theories. Since no analytical proof of confinement ex-
ist, it becomes essential to look at models that capture the dynamical degrees of
freedom that are responsible for confinement in order to understand the physical un-
7 derpinnings. Perhaps the most popular of such theories of quark confinement is the dual-superconductor picture of ’t Hooft.
1.2.1 Dual superconductor model of confinement
In mid 1970’s, ’t Hooft, Nambu and Mandelstam proposed an intriguing idea that confinement in non-abelian gauge theories could be due to the dual version of super- conductivity [7, 8, 6]. The BCS theory of superconductivity requires the existence of phonon mediated electron-electron pairs, namely the Cooper pairs. These spin zero bosons undergo Bose-Einstein condensation at low temperatures, and the condensate exhibits infinite conductivity. In the superconducting phase photons acquire a mass, thus, external magnetic field lines are either completely expelled from the supercon- ductor (type I) or squeezed into thin flux tubes (type II). This phenomenon is known as the Meissner effect. The magnetic flux between a pair of imaginary monopole- antimonopole pair inside the superconductor, will then be squeezed into a flux tube
(Abrikosov string) [3] and will give rise to a linear attractive potential between them
(see Figure 1.1a). Therefore, we can say that the monople-antimonopole pair is con-
fined.
This is very similar to what we desire in QCD, i.e the existence of a linear, attractive potential between a pair of chromo-electric charged objects. However, the role of the chromo-electric and chromo-magnetic fields are reversed compared to the supercon- ductor, and hence referred to as a dual-superconductor model. Consider the QCD
8 (a) (b)
Figure 1.1: (a) The Abrikosov string between the monopoles in the superconductor and
(b) an analogue of the Abrikosov string between the chromo-electrically charged particles in the dual superconductor. vacuum to be a condensate of monopoles. Then, in this state, gluons would become
massive, giving rise to a dual version of the Meissner effect. Suppose a pair of objects
with opposite chromo-electric charge (quarks or gluons) is placed in this medium. The
chromo-electric field between the pair would get squeezed into flux tubes (see Figure
1.1 b) and they would experience a linear, attractive potential. This is confinement.
However, this picture has two problems with it:
1) The dual-superconductor model is based on a theory, in which electro-magnetic
duality is manifest, while QCD is a non-abelian theroy.
2) The existence of monopoles is essential in this scenario, where as in non-abelian
gauge theories (without Higgs bosons), there are no monopole solutions.
9 As a solution to these two problems, ’t Hooft proposed the concept of abelian gauge
fixing [9]. By definition, abelian gauge fixing reduces QCD to an abelian gauge theroy, where one fixes the non-Abelian part of the gauge, such that the Cartan subgroup remains an exact symmetry. In the case of SU(2), U(1) remains exact, and this left-
over symmetry corrresponds to magnetic fields. As a remarkable fact, in this abelian
gauge, magnetic monopoles appear as topological objects corresponding to the non-
N 1 N 1 trivial homotopy group Π2(SU(N)/U(1) − ) = Z − . By abelian gauge fixing, QCD
is reduced to an abelian gauge theory with monopoles, and thus provide a theoretical
basis for the dual superconductor model of confinement.
Different abelian projections of non-abelian theories lead to different abelian theo-
ries, and the most popular of these is the maximal abelian projection (MA) [10, 11].
The MA gauge is the abelian gauge in which the diagonal components of the gauge
fields are maximized by gauge transformations. In this gauge, the physical informa-
tion of the gauge configuration is concentrated in the diagonal components. Lattice
studies of theories in the MA gauge give a strong indication of abelian dominance,
which means that the string tension and the chiral condensate are mostly given by
the abelian variables [12, 13]. There is also lattice evidence that in the MA gauge
only the monopole part of the abelian variables contribute to non-perturbative QCD,
and this feature is refered to as monopole dominance [17, 16]. The presence of abelian
10 and monopole dominance is considered to be strong evidense in favour of the dual-
superconductor model of confinement [14, 15].
1.2.2 Abelian Projection
The abelian monopoles arise from non-abelian gauge fields as a result of the abelian
projection suggested by ’t Hooft [9]. The abelian projection is a partial gauge fixing under which the abelian degrees of freedom remain unfixed. Since the original SU(N)
gauge symmetry group is compact, the remaining abelian gauge group is also com- pact. However it is known that abelian gauge theories with compact gauge symmetry group contain abelian monopoles. Therefore SU(N) gauge theory in the abelian gauge has monopoles.
As we have mentioned before, there are infinite number of abelian projections. How- ever, to demonstrate the method of abelian projection let us first consider the simplest example, i.e. the case of F12 abelian projection for SU(2) gauge theory.
Under the gauge transformation the component of the field strength tensor, Fˆ12,
transform as
Fˆ (x) Fˆ0 (x) = Ω†Fˆ (x)Ω. (1.1) 12 → 12 12
The F12 gauge is defined as:
Fˆ (x) Diagonal Matrix. (1.2) 12 ≡
11 Then in this gauge, the field strength tensor is invarient under U(1) gauge transfor- mations. i.e.,
ˆ ˆ F12(x) = ΩU† (1)F12(x)ΩU(1). (1.3) where,
eiα(x) 0 ΩU(1)(x) = (1.4) iα(x) 0 e− Therefore, the gauge condition (1.2) fixes the SU(2) group up to the diagonal U(1) subgroup.
The SU(2) gauge field transforms under the gauge transformation as:
i Aˆ Ω†Aˆ Ω Ω†∂ Ω. (1.5) µ → µ − g µ
where the SU(2) gauge field is given by,
A3 A+ ˆ µ µ 1 2 Aµ = , where Aµ± = Aµ Aµ. (1.6) 3 ± Aµ− Aµ − If we use the F12 gauge condition, then under the residual U(1) gauge transformation
the components of the nonabelian gauge field Aˆµ will transform as:
1 A3 A3 ∂ α, and A+ e2iαA+. (1.7) µ → µ − g µ µ → µ
3 + Here, we can see that Aµ transforms like an abelian gauge field, where as the field Aµ
plays the role of a charge 2 abelian vector matter field.
12 Thus, we obtain abelian fields from a non-abelian theory. Now, let us look at the
appearence of monopoles. For simplicity let us use a = A3 . Since, H~ = ~ ~a, if µ µ ∇ × a is regular, then ~ H~ = 0. i.e. there are no monopoles. However, Ω, the gauge ∇ · transformation matrix may contain singularities. In that case the abelian field a
may be non-regular. The non-abelian field strength tensor Fˆµν is not invariant under
singular gauge transformations. i.e.:
Fˆ Fˆ0 = Ω†Fˆ Ω + singular part, (1.8) µν → µν µν where,
i singular part = Ω†(x)[∂ ∂ ∂ ∂ ]Ω(x). (1.9) −g µ ν − ν µ
If, we define the abelian field strength tensor as that we obtain from singular gauge
transformation (1.5), i.e. f = ∂ a ∂ a , then f may contain singularities (Dirac µν µ ν − ν ν µν strings). i.e.
f = f r + f s , where ε ∂ f r = 0 and ε ∂ f s = 0. (1.10) µν µν µν µναβ ν αβ µναβ ν αβ 6
= ~ H~ = 0. Therefore, we can conclude that the abelian projected theory can ⇒ ∇ · 6 contain monopoles. To find the charge of the monopole, consider a monopole and
surround it by an infinitesimally small sphere S. Then the charge is given by the
integral,
1 m = H~ d~σ (1.11) 4π I
13 It has been shown that monopole charge can take values, m = 0, 1 , 1 ,...... [10, 11]. ± 2g ± g The quantization of the monopole charge is due to topological reasons. i.e. the sur- face integral shown in equation (11) is equal to the topological winding number of
SU(2) over the sphere S surrounding the monopole. And, the quantization comes
about due to the electric charge being fixed by equation (1.5), and consequently the
magnetic charge having to obey the Dirac quantization condition.
According to the dual superconductor model, the abelian gauge freedom remain-
ing after the gauge transformation, together with the monopoles, are responsible for
confinement. Therefore, once the gauge fixing is done all the off-diagonal elements of
the gauge field are discarded. This process is referred to as Abelian projection.
To solve the problem of confinement it is insufficient to prove the existance of monopoles,
but one needs to further establish that the monopoles condense. However, the only
known method of investigating this feature is through lattice calculations.
1.2.3 Abelian Projection on a Lattice
In the previous section the Fˆ12 gauge was used to illustrate the process of abelian
gauge fixing. However, there are an infinite number of abelian projections. Instead
of diagonalizing the component Fˆ12, one can choose to diagonalize any hermitian
operator G, that transforms as the adjoint representation of the gauge group, i.e. G
Ω†GΩ. Each operator G defines an abelian projection. →
14 Recent lattice studies have shown that infrared phenomena such as confinement and chiral symmetry breaking are to a very good extent reproduced in the Maxmial
Abelian(MA) gauge [18, 15]. This method was first devised by Kronfeld, Laursen,
Schierholtz, and Wiese [10]. The SU(2) gauge fields, Ul, on the lattice, are defined
by SU(2) matrices attached to the liks l. Under gauge transformation, the field, Ul,
transforms as Uµ0 (x) = Ω†(x)Uµ(x)Ω(x + aµˆ). In the SU(2) lattice formalism, the MA gauge is defined so as to maximize the functional,
1 F [U] = T r U (x)σ U †(x)σ (1.12) 2 µ 3 µ 3 x,µ X
Using the standard parametrization for the link matrix Ul,
iθl iχl cosϕle sinϕle Ul = (1.13) iχl iθl sinϕle− cosϕle− − where θ, χ [ π, +π] and ϕ [0, π] ∈ − ∈
In this parametrization,
R [Ul] = cos2ϕl. (1.14) Xl Here it can be observed that the maximization of R corresponds to the maximiza- tion of the diagonal elements of the link matrix. Once the gauge fixing is done one is free to perform the abelian projection. This is done be replacing the full gauge variables U = a I + i~a ~σ that are defined on the links in the lattice by the abelian 0 ·
15 links A given by;
a I + ia σ A = 0 3 3 (1.15) 2 2 a0 + a3 p
In order to test this theory one has to check whether the abelian projected theory
0.6 V-V0 0.5 ab ab V -V0 0.4
0.3
0.2 V(R) 0.1
0
-0.1
-0.2 2 4 6 8 10 12 14 16 R
Figure 1.2: Abelian and nonabelian potentials (with self energy V0 subtracted), Ref. [17]. is indeed equivalent to the original non-abelian theory. This is acheived by looking for abelian dominance in the projected theory. The notion of the abelian dominance means that the expectation value of the physical physical quantity < χ > in the non-abelian theory coincides with the corresponding expectation value in the abelian theory obtained by the abelian projection. Among the well-studied problems is the
abelian dominence for string tension σ [12, 13, 17]. In this case, χ(Uˆnonabelian) = σSU(2)
and χ(Uˆabelian) = σU(1). Both string tensions σSU(2) and σU(1) are calculated by means
16 0.8 Vab 0.7 Vph Vmon,fs 0.6 Vmon,fs+Vph
0.5
0.4 V(R) 0.3
0.2
0.1
0 2 4 6 8 10 12 14 16 R
Figure 1.3: The abelian potential (diamonds) in comparison with the photon contribution
(squares), the monopole contribution (crosses) and the sum of these two parts (triangles),
Ref. [17]. of Wilson loops. In ref. [17] the authors have carried out an accurate study of SU(2) gluodynamics on a 344 lattice. The abelian and non abelian potentials are shown in
Figure (1.2), while the photon and monopole contributions are shown in Figure (1.3)
[17]. From the slopes of Figures (1.2) and (1.3) one can derive the following results:
σU(1) = 92%σSU(2), σj = 95%σU(1), where σj refers to the monopole contribution to the string tension. From the first result one sees clear evidence of abelian dominence while the later indicates monopole dominance, which means that the monopoles are responsible for most of the SU(2) string tension.
Other calculations performed have shown condensation of monopoles in the confining
17 phase. At the same time mesearments have also revealed that chiral symmetry is also
broken int the abelian projected theory and the size of the chiral condensate takes
the desired values. The main conclusion is that the QCD vacuum behave like a dual superconductor, where the monopole currents are condensed and are responsible for confinement.
1.2.4 Center vortex model of confinement
In the preceding section I looked at the dual superconductor model of confinement, where the QCD vacuum is considered to be a monopole condensate, and as a con- sequence, the abelian charges confined. Evidence of abelian dominance in lattice calculations is indicative of the successfulness of this theory.
The idea that extended vortex configurations are responsible for confinement in SU(N) gauge theories was recognized by several people, including Nielsen and Olesen [20], ’t
Hooft [21], Cornwall [22], and Mack [23]. This picture is referred to as the spaghetti vacuum. Here, the QCD vacuum is understood to be a condensate of vortices of some
finite thickness, carrying magnetic flux in the center of the gauge group, hence refered to as center vortices. Recently, results of numerous lattice studies have indicated that the center vortex picture to be a strong candidate for describing confinement.
In center vortex model of SU(N) gauge theories, the gauge fixing conditions is choosen to reduce the full SU(N) gauge symmetry to the discrete center group, ZN . This is
18 reffered to as center gauge fixing. The most popular center gauges include, two ver- sions of the Maximial Center (MC) gauge [24, 25, 26], and the Laplacian Center(LC) gauge [27, 28].
I will first describe the ”indirect” maximal center gauge. One starts by fixing to the maximal abelian gauge [10]. As shown in the privious section (equation 1.12), for
SU(2) this amounts to maximizing the quantity,
T r Uµ(x)σ3Uµ†(x)σ3 (1.16) x,µ X Then the abelian projection is carried out where the full link variables are replaced by
the abelian link variables according to equation (1.15). These link variables transform
under the residual symmetry as U(1) gauge fields. The indirect maximal center gauge
uses this residual symmetry to bring the abelian link variables as close as possible to
SU(2) center elements I. This is acheived by writting ±
eiθ A = (1.17) iθ e− and using the residual U(1) symmetry to maximize
2 cos (θµ(x)) (1.18) x,µ X
This leaves a remnant Z2 symmetry. This is the (indirect) maximal center gauge.
Then Center Projection is carried out by replacing the full link variable U by the
19 center element ZI, where
Z sign(cosθ) = 1 (1.19) ≡ ±
This is refered to as the ’indirect’ maximal center gauge because the center is maxi- mized in the abelian link variable, rather than directly in the full link variable. Lattice
results show evidence of center dominance in this gauge. However, the agreement of the string tension extracted from this theory compare with the original string tension
before gauge fixing is not as good as one would desire. Better results have been re-
ported in the ”direct” center gauge.
The direct maximal center gauge fixing is accomplished by maximizing the functional
F [U] over the gauge transformation V(x), where
3 1 V 2 1 V V F [U] = T rU (x) = T r σ U (x)σ U † . (1.20) 4 µ 6 i µ i µ x,µ x,µ " i # X X X 1 3 = T r σ V (x)U (x)V †(x + aµˆ)σ V (x + aµˆ)U †(x)V †(x) . 6 i µ i µ x,µ " i # X X re-arranging this,
1 3 F [U] = T r V †(x)σ V (x)U (x)V †(x + aµˆ)σ V (x + aµˆ)U †(x) 6 i µ i µ x,µ " i # X X this can be written in the form
1 3 F [U] = g (x)g (x + aµˆ) T r σ U (x)σ U †(x) 6 i i i µ i µ x,µ i X X Write F [U] in terms of the adjoint representation quantities,
1 3 F [U] = g (x)U ab(x)g (x + aµˆ) (1.21) 3 ia µ ia x,µ iab X X 20 Consider the standard form of the gauge field,
U (x) = a + i~a ~σ, where a2 + ~a2 = 1 µ 0 · 0
2 then F [U] = a0. Therefore by maximizing a0, ~a is minimized, and results in the gauge field becomingP as close as possible to the elements of the center, U 1. µ ∼ ± This is the direct Maximal Center (MC) gauge. Once the gauge is fixed, the center projection is carried out by substituting, a + i~a ~σ sign a . 0 · → 0
Recently, lattice studies of MC theories have investigated whether the ”projection vortices” that appear in direct MC theories due to the excitations of Z2 gauge fields have anything to do with the existence of center vortices in the full, unprojected lat- tice configurations. L. Del Debbio et al.[26] have shown that indeed theses projection vortices locate center vortices in the full theory. In their calculations the string tension of the vortex contribution to the Wilson loops is found to match, quite accurately, the string tension extracted from the full Wilson loops. This is center dominence. These results seem to support the view that center vortices are responsible for confinement.
However, recently there has been criticism of the methods used to fix the gauge on the lattice in MC projection models [28]. The relationship between the ZN excita- tions which are located on a plaquette and the center vortices of the original SU(N) theory proceeds through the gauge fixing scheme. It is the gauge fixing technique used and not the maximal center projection that has raised doubts about the valid-
21 ity of this model. Gauge fixing is accomplished by iterative local maximization of
a gauge functional; and this process stops once a local maximum is reached. There
is no guatrantee that a gobal maximum has been reached, and as a result the local
maximum achieved depends on the starting gauge configuration. This situation has
been shown to have harmful effects in the MC gauge [30].
As a solution to this problem, Alexandrou, d’Eliya, and de Forcrand introduced the
Laplacian Center (LC) gauge [31, 27]. Consider the adjoint form of the functional
F [U]. 1 3 F [U] = g (x)U ab(x)g (x + aµˆ) (1.22) 3 ia µ ia x,µ iab X X One only needs two orthogonal vectors (say g1a and g2a) in order to uniquely define the adjoint representation gauge transformation. Therefore, it is suffiecient to keep only two terms i = 2 and i = 2 in the sum over i. g1a and g2a are normalized and are orthogonal to each other. In the LC gauge the normalization and the orthogonality
conditions on the vectors gia are relaxed. At certain points ’x’, g1a is parallel to g3a,
and these points form codimension 2 set. Then at such points the gauge freedom
is locally enlarged from Z2 to U(1), and the gauge trnsformation become singular.
These define the location of vortices. At points where the gauge freedome is further increaded to SU(2) for codimension 3 sets that can be identified with monopoles.
Therefore, in the LC gauge, center vortices and monopoles appear together as local gauge defects. The LC guage fixing is then followed by center projection where the
22 links are replaced by center element. Recent lattice simulations [28] have shown center dominance where the string tensions agree to a better precision than in the monopole picture with that calculated from the full theory.
1.2.5 Monopoles vs. Vortices
The strong evidence in support of a vortex condensation theory of confinement pre- sented in the previous section, is reminiscent of the data that support the monopole condensation theory. However, these seem to be mutually exclusive propositions, for one thing, monopoles are of codimension 3, while vortices are codimension 2 objects.
So, this raises the natural question: Which one of these models are correct? i.e. if the QCD vacuum is dominated by vortec configurations (”spaghetti vacuum”), then how do we explain the numerical success of the abelian projection in the maximal abelian gauge?
Recently, there has been several exhaustive studies carried out on this problem
[25]. The current opinion is that a center vortex configuration, transformed to max- imal abelian gauge and then abelian-projected, will appear as a chain of monopoles alternating with antimonopoles. Therefore, these monopoles are essentially an arti- fact of the projection and they are condensed because the long vortices from which they emerge are condense. If this picture is accurate, then the ”spaghetti vaccum”
(of condensed vortices) appears, under abelian projection, as a ”monopole vacuum”
(see figure 1.4).
23 +
+ - + - - + + + -
-
+
- -
+
(a) (b)
Figure 1.4: (a) The ”Spaghetti vaccum” of condensed vortices (b) The ”monopole vacumm” appears as a result of abelian projection [25]
From the preceding discussions, clearly there is convincing numerical evidence that
monopoles and mostly vortex condensation picture of confinement is valid. although
lattice simulation evidence is still very important, it is also important to use all means
possible to investigate the properties of classical objects responsible for confinement.
Few important aspects of these classical solutions that can possibly be investigated
in Higgs theories include, the relationship between monopole and vortex induced confinement, interaction of vortices, and the very ambitious goal of studying the condensation of vortices. However, in order to carry out such investigations one needs to know how to correctly define the energy of a single object. i.e. we need to be able to determine the energy dependence of extended classical solutions on the shape as well as the length of the object.
24 Chapter 2
The Undulating Kink
2.1 Introduction
Classical solutions of field theories play an important role in a host of applications.
In particular vortices, abelian and nonabelian, acquired central importance, among
others, in superconductors [3], cosmic strings [4], and as objects responsible for con-
finement [5] [6]. Classical solutions of recent interest are p-branes, appearing in string
theories [32].
Vortices and some other solutions of field equations distinguish themselves by the
fact that they are extensions of finite energy classical solutions (solitons) into one
or more additional spatial dimensions. As such, they have infinite energy, but finite
energy per unit length if the additional space dimensions are infinite. The energy of
these solutions becomes finite if the additional spatial dimensions are finite. In what
follows I will only consider finite additional dimensions that make analytical studies
of these objects possible.
25 To the best of my knowledge, analytical studies of these objects have not been ex-
tended to classical solutions. One important aspect of classical, finite energy solution
lifted into higher dimensions, is that their energy is dependent on the shape and length of these objects. In most applications it has been assumed [33], [34] that the energy of a 3 dimensional vortex is simply its length multiplied by the energy of unit length (which is equal to the energy of the two dimensional soliton), irrespective of its shape.
Before proceeding any further I need to clarify what I mean under the dependence of the energy on shape. Topologically stable classical solutions are characterized by the
vanishing of an order parameter (usually a scalar field) at a finite coordinate, which
is, by definition, the location of the center or core of the object. In trivially extended
classical solutions the zeros form a straight line. These extended objects are solutions
of higher dimensional field equations. Suppose now I impose a constraint that the
locus of zeros lies on a predetermined curve and minimize the Hamiltonian with this
constraint. Then, in general, the value of the Hamiltonian will depend on the shape
and length of the locus of zeros. This dependence is important, because before one
can attempt an analytic or numerical study of the interaction and condensation of
vortices, or of other classical extended objects, one needs to clarify the energy of a
single object. As I show in this chapter, the dependence of the energy on the length
and curvature of an undulating object is highly nontrivial.
26 One might think that the deviation of an extended object from the minimum energy straight line is a purely quantum phenomenon and should be treated in per- turbation theory. Such a treatment would not, however, adequately describe large scale features of extended objects. Such, macroscopic features of vortices are clearly seen in superconductors or simulations of cosmic strings. [34] Therefore it should be possible to calculate the properties, at least in some approximation, studying the clas- sical equations of motion. Naturally, if the energy of curved vortices can be calculated their contribution to the functional integral, i.e. their full quantum contribution, can possibly be evaluated.
One of the ultimate targets of this investigations is vortex solutions in gauge theories. As no analytic vortex solutions exist, the problem of curved vortices is fairly difficult. Though my aim is to understand issues pertaining to the curvature dependence of classical solutions in more complicated quantum theories, I will first discuss the curvature energy for a simpler problem, the kink solution of the 2+1 dimensional linear sigma model. I need to gain an understanding of this problem so
I can be apply such information to solve more difficult problem of vortices and other extended classical objects as described in chapter 3.
2.2 Kink solution
One of the simplest examples of classical solutions of interacting field theories the one dimensional kink, which is a soliton solution of the two dimensional linear sigma
27 model. Kinks are toplogically stable finite energy solutions that can be extended to 2 dimensions in a straightforward manner. I start with a brief description of these, well
known solutions, [35] with the intention of establishing notations. As the solution can
be chosen to be static, the Lagrangian coincides with the negative of the Hamiltonian.
The Hamiltonian for a time independent solution is solution is
1 ∂Φ 2 λ m2 2 H = dξ + Φ2 . (2.1) 2 ∂ξ 2 − λ Z " # The vacuum expectation value can be scaled out if I use the transformation
m Φ(ξ ξ0) = Ψ(x x0), (2.2) − √λ −
where m(ξ ξ )/√2 = x x . The rescaled Hamiltonian takes the form − 0 − 0 m3 ∂Ψ 2 H = dx + Ψ2 1 2 = dxh(x). (2.3) 2√2λ ∂x − Z " # Z with a minimizing kink (anti-kink) solution
Ψ (x x ) = tanh(x x ). (2.4) c − 0 ± − 0
The energy density of the kink solutions is
m3 1 h(x) = (2.5) √2λ cosh4(x x ) − 0 The value of the Hamiltonian at this minimizing solution is
m3 4 H = H0 = . (2.6) √2λ 3
In what follows I will set m2 = λ = 2. The correct units can be easily restored.
28 (a) (b)
Figure 2.1: (a) A plot of the static kink(anti-kink) solution (2.4) (b) The energy density
of the kink (2.5). It is localized.
2.3 Extension to 2+1 dimensions
In the three dimensional linear sigma model there is an extra term in the Hamiltonian
density, (∂Φ/∂y)2. Solution (2.4) also satisfies equations in 2 dimensions, but it would
have infinite energy if the space is infinite in the y direction. As indicated earlier, I choose a finite y dimension, with periodic boundary condition. In other words, I seek
a kink solution on an infinite strip of width L.
The minimum energy kink on the strip is a straight line at a fixed value of x. It
is interesting to study, however, the dependence of the energy on the shape of the
kink. For this purpose I distort the shape of the kink by assuming that the zero of
the solution is not at x0 = 0 (this value is chosen arbitrarily) but rather at x = (y).
The periodicity condition, with the translation symmetry along the x axis allows me
to choose (0) = (L) = 0.
29 The problem I pose is the following. Suppose a close curve, defined by x = (y), winds through the strip. This curve is the locus of the zeros of Φ(x, y). In other words, I impose the constraint on the field, Φ((y), y) = 0. We wish to find the field configuration, Φ(x, y), satisfying this constraint and minimizing the Hamiltonian. I am mostly interested in the dependence of the minimum of the Hamiltonian on the shape of the curve.
The two-dimensional Hamiltonian has the form
1 dΦ 2 dΦ 2 H = dx dy + + (Φ2 1)2 . (2.7) 2 dx dy − Z " #
It is assumed here that Φ is a configuration minimizing H for a given shape of the core.
Define the ‘length-energy,’ Hl, of a two a dimensional curved kink solution as
Hl = lH0, where l is the total length of the kink,
2 l = dy 1 + 0 . Z p and H0 is the energy of the one dimensional kink (energy of unit length of straight two dimensional kink). Then the curvature energy of a two dimensional kink solution is defined by subtracting the ‘length-energy,’ from the total energy, H = H H . I c − l can then prove the following theorem:
The curvature energy of the two dimensional kink is non-positive.
My detailed calculations will show that the curvature energy is negative definite. In
30 other words, discounting the length energy, the kink prefers the curled state to the
straight one.
Proof: Consider a class of solutions that depend on the coordinate x and y only C through the combination x (y). The Hamiltonian minimized on functions of this − class will have a minimum that is not smaller than the unrestricted minimum. To
prove the theorem I will show now that the Hamiltonian minimized on is exactly C the length energy.
Assume that Φ depends on x and y through x (y). Then upon substitution − into (2.7) and shift of coordinate x by (y) I obtain
2 1 2 dΦ 2 2 H = dx dy (1 + 0 ) + (Φ 1) . (2.8) 2 dx − Z " #
The substitution x x√1 + 2 leads to → 0
2 1 2 dΦ 2 2 H = dx dy 1 + 0 + (Φ 1) = lH0. (2.9) 2 " dx − # Z p where,
2 l = dy 1 + 0 . Z p Φ depends only on x. As the one dimensional Hamiltonian is minimized by the original kink solution, (2.4), we obtain H = 4l/3, as asserted above. This completes the proof of the theorem.
31 2.4 Exact solutions for small deviations from the
straight line
In the previous section illustrated that the curvature energy of an arbitrary curved kink is negative. I am able to calculate the curvature energy analytically only if it is assumed that the deviation of the kink from the straight line, (y), is small.
The general form of the two dimensional kink solution is written in the form
Φ(x, y) = Φ (x )+χ(x , y), where Φ = tanh(x). As the constraint is Φ((y), y) = c − − c
0, and having Φc(0) = 0 one gets
χ(0, y) = 0. (2.10)
Since at = 0 the solution Φc is exact, χ = O(), as well.
Now I can vary the Hamiltonian with respect to χ with the subsidiary condition
(2.10). Keeping terms of O() I obtain the following equation of motion for χ:
2 χ + χ 00(y)Φ0 2(3Φ 1)χ = δ(x)λ(y), (2.11) xx yy − c − c − where the Lagrange multiplier λ(y) = χ ( + 0, y) χ ( 0, y) and where the x − x − subscripts x and y indicate partial derivatives. The appearance of the delta function in (2.11) has no other significance in our subsequent calculations other than removing a condition on the continuity of the derivative of χ at x = 0.
32 Notice that the inhomogeneous driving term 00(y)Φc0 in (2.11) is an even function of x. Then due to the x x symmetry of (2.11) we also have χ(x, y) = χ( x, y). It → − − is sufficient to solve (2.11) for x > 0, with boundary conditions χ(0, y) = χ( , y) = 0. ∞ Note that though χ(x, y) is even χ (0, y) = 0 because it is not continuous. x 6 To find the appropriate solution one must find the solutions of the homogeneous and inhomogeneous equations that vanish at x and find a combination of these → ∞ two solutions that also vanishes at the origin.
Let me expand χ(x, y) in a Fourier series of y as shown below.
1 χ(x, y) = e2πıny/LF (x), (2.12) √ n L n X ? where F n = Fn . Using Φc = tanh x the modes satisfy the one-dimensional inhomo- − geneous Schr¨odinger equation
2 2 2 2 6 π n 4π n n F 00 + Fn 4 1 + Fn = , (2.13) n cosh2 x − L2 − L2 cosh2 x where n is the Fourier expansion coefficient of (y).
The solution of the inhomogeneous Schr¨odinger equation satisfying boundary con- ditions F (0) = 0 and F (x) 0 at x is n n → → ∞
1 αx 3α 3 2 Fn(x) = n e− 1 + tanh x + tanh x , (2.14) cosh2 x − α2 1 α2 1 − − where
π2n2 α = 2 1 + 2 . (2.15) r L 33 The O(2) deviation of the Hamiltonian to that of a straight kink becomes
1 2 2 2 2 ∆H = dxdy (χ ) + (χ 0Φ0 ) + 2((3Φ 1)χ . (2.16) 2 x y − c c − Z After integrating by parts and utilizing (2.11) the Hamiltonian can be written as,
1 ∆H = dxdy0Φ0 (0Φ0 χ ) (2.17) 2 c c − y Z Substituting solution (2.14) into (2.17) the integral can be easily evaluated to give
4n2π2 2 ∆H = 2 (2.18) n L2 α(α2 1) X − using (2.15) the final result can be derived,
2 2 2 2 π n 8n π 1 + L2 ∆H = 2 (2.19) n L2 q4π2n2 L2 + 3 X Compare (2.19) with (2.9). Given that, ∆l = l L is the increase of the length − of the kink due to undulation, the increase of length energy, as obtained from the
expansion of (2.9) is
2 2 H0 2 2 8π n ∆H = H ∆l dy[0(y)] = . (2.20) 0 0 ' 2 n 3L2 n Z X
Then comparing (2.20) with (2.19) shows that ∆H0 > ∆H, or in other words, the
curvature energy is always negative. This statement is analogous to the general statement that the second order perturbation to the ground state is always negative.
Assuming that long wavelength modes dominate I can expand the expression of the excess energy in π2n2/L2. Then using
1 + 2 1 R = 0 , ' | 00| | 00| 34 where, R is the curvature radius we get the leading correction term in long wavelength
modes
8π2n2 2π4n4 1 ds ∆H 2 + . . . = H ∆l + . . . . ' n 3L2 − 3L4 0 − 32 R2 n X Z Omitted higher order terms will contain derivatives of the curvature radius. The
minimum of the total Hamiltonian is
1 2 H = H ds ds(x00) + . . . , 0 − 32 Z Z where x is the 2-dimensional coordinate vector of the undulating kink core in the path-length gauge. Clearly, the curvature term is negative, as required by theorem proven in the previous section.
To reconcile this result with another to be obtained in the next section, consider the low frequency approximation from the beginning. I will take the infrared limit of the small approximation and then take the small limit of the infrared approxima- tion. Note that for this comparision it is better to use the view that the unperturbed solution is Φ (x ) rather than Φ (x) = tanh x. Then the method is modified and one c − c needs to solve an inhomogeneous Schrodinger equation, rather than a homogeneous one. The differential equation becomes
2 2 2 2 2 4π n 4π n F 00 2(3Φ 1)F F = Φ0 (2.21) n − c − n − L2 n − L2 n c
After integration by parts, the Hamiltonian transforms to
1 ∆H = dx dy 0Φ0 (0Φ0 χ ). (2.22) 2 c c − y Z 35 Using the approximation that at most O(δ4) terms are kept, where δ = nπ/L,
2 only O(δ ) terms in χ are needed. But then, following from (2.21), the function Fn
is proportional to δ2 anyway so correction terms of O(δ2) are not needed. Therefore
to order δ4 in the Hamiltonian one can use a simplified equation
2 2 2 2 2 4π n 4π n n F 00 2(3Φ 1)Fn = nΦ0 = . (2.23) n − c − − L2 c − L2 cosh2 x
It is feasible to check how this works. One solution of the homogeneous equation is
obviously 1 F 0 = . n cosh2 x
Then the second solution can be determined as stated in (2.37). The solution is
given by (2.39). Now, a linear combination that decreases at infinity and vanishes at
the origin is
2 2 2x 4x 4π n 8e− + e− 12x 9 Fn = n − − (2.24) − L2 48 cosh2 x
This solution substituted into (2.22) results in the following contribution
2 5 ∆H = 00 . (2.25) − 36
2.5 Non-infinitesimal undulation
Now consider results in the interesting domain of finite, O(1), 0(y), restricted our-
2 selves to the low frequency domain, when 00 << (0) /L. This implies that derivatives
of the field with respect to y are small and decrease with the number of derivatives.
36 In this domain the low frequency region is the most interesting, because this leads to the most probable macroscopic deformations.
The trick for the analytic solution of this approximation is that the solution will
be dominated by the trivially transformed solution of the straight vortex,
x (y) Ψ0(x, y) = tanh − tanh u, √ 2 ≡ 1 + 0
where the variable u is defined as u = (x )/√1 + 2. This function alone would − 0
give an energy of E = H0l, where l is the length of the kink, up to a correction term
2 of O(00 ). I will assume a small deviation from this solution, which, admittedly, gives
only a small deviation in the energy, as well. One would guess that increasing 00
would lead to larger deviations. The importance of this result is that it is completely
beyond the capability of calculating quantum corrections to the straight string, as the starting configuration deviates already from the straight string by a finite amount.
First calculate the energy following from substituting Φ0 into the Hamiltonian.
2 1 0 u 00 2 2 2 H = dx dy + + Φ0 (u) + (Φ 1) 2 2 2 3/2 0 0 1 + 0 √1 + (1 + 0 ) − Z (" 0 # ) 2 2 2 u 00 2 2 2 = dx du √1 + 1 + Φ0 (u) + (Φ 1) . 0 (1 + 2)3 0 0 − Z 0
2 Neglecting the subleading term, of O(00 ), the Hamiltonian is clearly optimized by the function tanh(u). Then, using this form the following expression for the leading
order approximation to the Hamiltonian can be obtained
π2 6 ds H = H ds + − , (2.26) 0 9 R2 Z Z 37 2 2 where s is the arclength and also ds = dy√1 + 0 , and R = (1 + 0 )/00. Notice that
(2.26) also gives an exact upper bound to the energy of the kink, as no approximation was made, other than the assumption of the form tanh(u).
Next write the equation of motion for (2.7), assuming that Φ depends on x and y through the variables z = x (y), and y. Then, as before Φ(z, y) vanishes at z = 0 − and z . Furthermore, Φ is a periodic function of y, with period L. This is so → ±∞ because (y) is assumed be a periodic function, as well. Then the y-component of the kinetic energy term takes the form
2 (Φ 0Φ ) , y − z where as usual lower subscripts signify partial derivatives. The equation of motion becomes
2 2 Φ + Φ + 0 Φ 20Φ 00Φ 2Φ(Φ 1) = 0. (2.27) zz yy zz − yz − z − −
If 00 << 1/L then differentiations with respect to y are small. In leading order of the frequency derivatives of Φ with respect to y can be neglected. Then in the leading order (2.27) reads as
2 2 (1 + 0 )Φ 2Φ(Φ 1) = 0. (2.28) zz − −
The solution of (2.45) is exactly the form that was used previously to derive the energy (2.26).
z Φ(0) = tanh u tanh , (2.29) ≡ √ 2 1 + 0 38 where the variable u can be defined as
z x u = = − . (2.30) 2 2 √1 + 0 √1 + 0
(0) Φ when substituted into the Hamiltonian gives H = H0l.
Now perform a perturbative expansion around Φ(0). Writing Φ = Φ(0) + χ. The
leading driving terms are clearly
(0) (0) 20Φ 00Φ . − zy − z
Then keeping first order terms in χ only and omitting derivatives of χ with respect
to y one can obtain the following equation for χ:
2 (0)2 (0) (0) (1 + 0 )χ χ2(3Φ 1) = 20Φ + 00Φ . (2.31) zz − − zy z
(0) Note that the term Φyy or derivatives of χ with respect to y would contain at least
2 (00) or 000, negligible compared to 00.
The solution of (2.31) requires to find two solutions, one of the homogeneous
equation, one of the inhomogeneous equation, both vanishing at infinity and then
taking a linear combination that vanishes at z = 0, as well. First rewrite (2.31) using
variables u and y.
2 2 6 00 20 40 χuu 4 χ = 1 + u tanh u . (2.32) 2 2 2 2 2 − − cosh u u√ − 1 + 0 1 + 0 cosh 1 + 0
The solution of (2.32) satisfying the boundary conditions has the form
2 2 1 00 000 00 2000 χ = g1(0) + g2(0) + g1(u) + g2(u(2.33)), 2 2 2 3/2 2 2 3/2 −cosh u √ (1 + 0 ) √ (1 + 0 ) 1 + 0 1 + 0 39 where the first term on the right hand side is a solution of the homogeneus equation,
while g1(u) and g2(u) satisfy the equations
6 1 g00 4 g1 = (2.34) 1 − − cosh2 u cosh2 u
and
6 2u tanh u 1 g00 4 g2 = − , (2.35) 2 − − cosh2 u cosh2 u with g (u), g (u) 0 when u . 1 2 → → ∞ One solution of the homogenous equation is
1 g0 = (2.36) cosh2 u
Then it is easy to generate another solution of the homogeneous equation as
du0 1 g˜0(u) = g0(u) = [12u + 8 sinh(2u) + sinh(4u)] . (2.37) g2(u ) 32 cosh2 u Z 0 0
Having known a solution of the homogeneous equation like
g00 + h(u)g(u) = 0
it is easy to write a solution of the inhomogeneous equation,
g00 + h(u)g(u) = f(u).
It is given by
u u0 du0 g(u) = g (u) du00g (u00)f(u00), (2.38) 0 g2(u ) 0 Z 0 0 Z 40 where f(u) is the inhomogeneity. Using the above definition of g0 inhomogeneity can be written as 1 f(u) = cosh2 u
8 cosh(2u) + cosh(4u) g¯1(u) = , (2.39) 48 cosh2 u
where a multiple of g0 has been added.
Note that the second solution of the homogeneous equation is
8 sinh(2u) + sinh(4u) + 12 u f 2(u) = (2.40) cosh2 u
Thus the solution of the inomogeneous equation with the right asymptotic behavior is
2u 4u 8 e− + e− 12 u g1(u) = − , (2.41) 48 cosh2 u
For an inhomogneity 2u tanh(u) 1 f(u) = − cosh2 u
I obtain 24u2 + 8 cosh(2u) + cosh(4u) g¯2(u) = . − 96 cosh2 u
Now the solution with the correct asymptotic behavior is
2 2u 4u 24u + 8e− + e− 12u g2(u) = − . (2.42) − 96 cosh2 u
41 Then, in view of (2.43) one can write
1 3 3 2 χ = 00 00 0 2 2 2 3/2 −cosh u 16√1 + 0 − 32(1 + 0 ) 2u 4u 2 2 2u 4u 00 8 e− + e− 12 u 2000 24u + 8e− + e− 12u + − − 2 2 2 3/2 2 √1 + 0 48 cosh u − (1 + 0 ) 96 cosh u 2 00 0 2 2u 4u = 9 24u 8e− e− + 12u 2 2 2 √ u 1 + 0 − − − 96 1 + 0 cosh 2u 4u 18 + 16e− + 2e− 24u . (2.43) − − }
Now the relevant Hamiltonian density for the field χ, which is second order in χ
or derivatives with respect to y is
1 2 2 2 (0)2 (0) (0) (0)2 h = (1 + 0 )χ + 2χ (3Φ 1) 20(Φ χ + Φ χ ) + Φ . (2.44) χ 2 x − − y x x y y Note that it is necessary to add the contribution of the pure length term to the
Hamiltonian density. The last term of (2.44) corresponds exactly to the last term of
(2.26).
Now integrating by parts in (2.44) and using the field equation for χ, (2.31),
1 (0) (0) (0)2 1 (0) (0) (0)2 h = χ(20Φ + 00Φ ) + Φ = 0(Φ χ Φ χ ) + Φ . (2.45) χ 2 yz z y 2 yz − z y y When one integrates this expression, one obtains a correction term of the form
1 ds . R2 Z This must agree with the expression that was derived in the previous sectionr, as the
0 limit of this must be the same. After differentiation →
(0) (0) 00 1 2 2 20Φyz + 00Φz = 2 3/2 2 1 0 + 40 u tanh u . (1 + 0 ) cosh u − 42 Furthermore, the term
2 2 2 (0)2 00 0 u Φy = 2 2 4 . (1 + 0 ) cosh u
So finally, the hamiltonian density takes the form
2 00 1 2 2 1 2 2 hχ = u 0 + 1 0 + 40 u tanh u 2(1 + 2)2 cosh4 u 96 − 0 2 0 2 2u 4u 2u 4u 9 24u 8e− e− + 12u 18 + 16e− + 2e− 24u(2.46). × 1 + 2 − − − − − 0 Substituting this expression into the expression of the integrated Hamiltonian it can
be shown that
2 2 00 5 1 2 2 5 0 ∆H = + (π 6)0 + . (2.47) (1 + 2)3/2 −36 72 − 72 1 + 2 0 0
2 where I have included the multiplier √1 + 0 coming from the change of variable
from x to u. In the limit of 0 0 this agrees with the infrared limit of the small → solution, (2.25). The interesting fact about the above expression for the energy is that
the curvature contribution is negative only if 0 < 1.32674. In effect, the functional | |
integration then cuts off at 0 = 1.32674. ±
43 Chapter 3
The Circular Vortex
3.1 Introduction
In the preceding chapter, we investigated the properties of classical solutions of inter-
acting scalar field theories in (2+1) dimensions. We now turn to more complicated systems in higher dimensions. i.e we will investigate classical solutions of gauge field
theories in (3+1) dimensions.
Such interacting field theories give rise to non-linear equations of motion and such systems of equations often contain classical solutions [35]. The solutions of the field equations of these models can be classified by the number of dimensions of the object.
One type of classical solutions, referred to as solitons are particle like objects that
carry finite energy. The kink of the preceding chapter is a (1+1) dimensional soliton.
In this chapter I will look at another type of classical solution - the Vortex. Vortex
solutions exist in field theories where gauge fields are coupled to a complex scalar
field with a topologically nontrivial degenerate vacuum. The vortex solutions are
44 particle-like finite energy (2+1) dimensional solitons. If we extend the vortex into
(3+1) dimensions it becomes a two dimensional object like a string, and in this case the energy will in general be infinite, given that the extension of space is infinite. In other words, vortices carry finite energy per unit length.
One of the first relativistic vortex models was developed by Nielsen and Olsen [38].
They investigated an example of a field theory with a vortex solution by considering an
Abelian gauge field coupled to a charged scalar. Nielsen and Olsen were motivated to study this system because it stimulates the dual string picture of elementary particles
[8, 39, 40]. Their system is also a relativistic analogue of what happens in a type-
II superconductor, i.e. the Landau-Ginzburg model[3]. Nielsen and Olsen found approximate analytical solutions of their abelian model. [38].
As I mentioned in the previous chapter, the energy of vortex solutions of (3+1) dimensional gauge theories is always determined by simply multiplying its length by the energy of the two dimensional vortex[33], [34]. In most of the literature on the subject of classical solutions lifted into higher dimensions, the energy dependance on the shape of the solitons has been neglected. However, as observed previously in the case of the kink, the curvature of the soliton affects the energy in a highly nontrivial fashion. One ultimate goal of this investigations is to determine the curvature de- pendance of vortex solutions. However, unlike in the case of the kink, no analytical vortex solutions exist, making the study of curved vortices a very difficult task. The effects of curvature can be studied easiest on circular vortices. In this chapter I will
45 investigate circular vortex solutions of large radius of the U(1) gauge theory. How- ever, before solving the equations of motion of the circular vortex, first it is necessary to study and obtain a complete straight vortex solution. This is necessary since the straight vortex solution appears in the field equations of the circular vortex. In the next section I shall describe the development of the straight Abelian vortex solution, and subsequently, develop the circular-vortex. Circular vortices are not stable if they are allowed to shrink. Therefore, I will investigate vortices the core of which is pinned down along a circle.
It may seem logical to consider the fluctuations of the extended vortex solutions from its straight line minimum is purely quantum mechanical. Unstable configura- tions are important in a quantum field theory as they contribute to partition functions when one integrates over all configurations. However, integrations over small fluctua- tions around classical solutions, such as the straight vortex would not include curved vortices. i.e. in order to capture the macroscopic features of extended classical solu- tions one needs to study the classical equations of motion.
3.2 The Straight Vortex
Consider a (2+1) dimensional U(1) gauge theory with a complex scalar Higgs (Φ).
For a static solution the Lagrangian corresponds to the negative of the Hamiltonian,
46 given by,
1 1 H = d2x (∂ A ∂ A )2 + (∂ + ıeA ) Φ 2 + λ Φ 2 η2 2 , (3.1) 2 2 ν µ − µ ν | µ µ | | | − Z where λ is the scalar coupling constant and η is the Higgs vacuum expectation value.
Aµ is the U(1) gauge field.
µ The corresponding equations of motion in the ∂ Aµ = 0 gauge take the form,
1 2A = ıe (Φ?∂ Φ Φ∂ Φ?) e2ΦA , (3.2) µ 2 µ − µ − µ 2Φ + 2ıeA ∂µΦ e2A AµΦ = λ Φ2 η2 . (3.3) µ − µ − Finding finite energy solutions of gauge theories is a daunting task, specially if stability of the solutions are to be investigated. However, this can be facilitated by employing symmetry considerations and topological conservations laws.
It can be seen from equation (3.1) that zero energy configurations must have
Φ(x, t) = η and hence in order to ensure finite energy solutions the fields should | | statisfy,
Φ η as x , and | | → → ∞ (∂ + ıeA ) Φ 0 as x . µ µ → → ∞
The phase of the complex scalar field Φ Φ eıα is not fixed by these boundary ≡ | | conditions at infinity. Since the phase angle α takes values on a circle, S1, the bound- ary conditions map the surface at spatial infinity onto this circle. Now, consider the
47 following results on homotopy groups.
πn(Sn) = Z (3.4)
πn(Sm) = 0 for n < m, (3.5)
and πn(S1) = 0 for n > 1, (3.6)
where πn(Sm) refers to the homotopy group for the mapping of Sn into Sm, and Z refers to the group of integers. i.e., according to (3.4) the mapping of S S come n → n in a discrete infinity of homotopy classes, each characterized by an integer. The zero on the right hand side of (3.5) and (3.6) implies that the group is trivial, i.e. in those cases all mappings can be deformed into one another. These topological relationships are important when investigating classical solutions [41], because theories that contain nontrivial homotopy classes of mappings are guaranteed to have at least one stable
finite energy solution in each of the homotopy classes.
In the Abelian model the boundary conditions at requires the symmetry group ∞ of the target space to be S1. Consider the spatial dimension to be D = 3, then the surface at infinity is a sphere S2. However, (3.6) tells us that π2(S1) = 0, and therefore there are no non-trivial homotopy sectors of (3+1) dimensional solutions for this model. For the abelian model this is true for all D 3 since πD 1(S1) = 0 ≥ − ∀ D 3. However, when D = 2 the surface at infinity becomes a circle, giving rise to ≥ mappings of circles into circles which leads to the non-trivial homotopy group π1(S1).
This is the group of winding numbers. Therefore, in this theory stable finite energy
48 solitons will be found only in D = 2. These solutions can be characterized by the winding number which is a conserved quantity. However, consider stable solutions in (3+1) dimensions where the solution is independent of the third direction. In this case, unless the third dimension is compact, the energy of the classical solution becomes infinite, only the energy per unit length remains finite.
Given that the solutions are two dimensional, the z component of the vector potential can be chosen to be zero. Since only time independent solutions are of interest, I can choose the A0 = 0 gauge. Also note that the winding number describes the number of times the phase α of the Higgs is wound around the circle at infinity,
α(2π) α(0) − n = 2π . Therefore, we may seek nontrivial classical solutions of the form,
a(ρ) A (ρ, ψ) = a(ρ)∂ ψ = ψˆ, (3.7) µ µ ρ Φ(ρ, ψ) = ηeınψχ(ρ), (3.8)
The solution only depends on the polar coordinates (ρ, ψ) and is independent of z. Now, integrating over ψ the Hamiltonian (3.1) can easily be rewritten in terms of the two fields a(ρ) and f(ρ) as,
1 da 2 η2 dχ 2 H = π dρρ + (ea + n)2 χ2 + η2 + λη4 χ2 1 2 (3.9) ρ2 dρ ρ2 dρ − Z " # The appropriate boundary conditions for finite-energy vortex solutions are
χ(0) = a(0) = 0,
lim χ(ρ) = 1, and lim a(ρ) = n/e. (3.10) ρ ρ →∞ →∞ −
49 These conditions ensure that each one of the terms of the integrand of (3.9) is finite.
For these cylindrically symmetric solutions, equations of motion (3.2, 3.3) take the
form,
d2a 1 da eη2 (ea + n) χ2 = 0 and dρ2 − ρ dρ − (3.11) d2χ 1 dχ χ + (ea + n)2 2λη4χ χ2 1 = 0. dρ2 ρ dρ − ρ2 − − Given the boundary conditions (3.10), the flux Ψ of the vortex line can be derived,
a(ρ) 2πn Ψ = A (x)dxµ = dψρ = . µ ρ e I I
2π Thus, the flux of the vortex line is quantized, e being the quantum.
The field equations (3.11) can only be solved numerically, and these calculations provide monotonic solutions for both a(ρ) and χ(ρ), as shown in figure(3.1). However, a lower bound for the energy can be found by adopting the Bogomol’nyi linearization technique [43]. Assuming that n > 0, the Hamiltonian can be written as,
1 da 2 dχ (ea + n) χ 2 H = π dρρ + γeη2 χ2 1 + η2 + 2γ (3.12) ρ dρ − dρ ρ Z "
2 2 2 2 2 2 4 2 2 (ea + n) η χ (1 4γ ) 1 2 d 2 + λ γ e η χ 1 + 2 − 2γη (ea + n) χ 1 − − ρ − ρ dρ − # √λ where the parameter γ > 0. It can be observed that by choosing γ < 1/2 and γ < e
fourth and third terms in (3.12) become positive. Also, the last term can be integrated
50 and using the boundary conditions one obtain,
H = 2πγη2n + non negative terms
This shows that the energy of the soliton is greater than 2πγη2n. Using the limits imposed on γ, the lower bound of the hamiltonian will be the smaller number of either
H πη2n or ≥ H 2π√λη2n/e. (3.13) ≥
The limiting case is where the Higgs mass mH = 2η√2λ becomes identical to the
√ λ 1 gauge mass, mA = 2eη. i.e. when e2 = 4 . At this critical coupling limit the vortex- vortex forces vanish. Below this point the vortices attract each other and above it they repel each other. The Hamiltonian in the critical coupling case reduces to
1 da 1 2 dχ (ea + n) χ 2 H = πη2n + π dρρ + eη2 χ2 1 + η2 + ρ dρ 2 − dρ ρ Z " # which yields an exact value for the energy of the vortex,
H = πη2 n (3.14) | | and the fields can be shown to satisfy linearized Bogoml’ny type equations,
dχ χ = (n + ea) , and dρ − ρ da η2eρ = 1 χ2 dρ 2 −
51 3.2.1 Solutions of the straight Vortex
I am interested in determining the solutions of the coupled system of equations (3.11).
However, as mentioned previously, solving these nonlinear coupled second order dif-
ferential equations necessitate the use of numerical techniques. Begin by going over
to dimensionless variables
r ρ = , and eη (n + ea) = α. (3.15)
λ Now, by defining β = e2 the hamiltonian (3.9) can be rewritten using the new vari-
ables,
2 2 1 dα dχ 1 2 H = πη2 drr + + α2χ2 + β χ2 1 . (3.16) r2 dr dr r2 − Z " # The equations of motion (3.11) of the straight vortex become,
d2α 1 dα αχ2 = 0 dr2 − r dr − d2χ 1 dχ 1 + α2χ 2βχ χ2 1 = 0, (3.17) dr2 r dr − r2 − − with the boundary conditions (3.10) transforming to,
χ(0) = 0 and α(0) = n,
lim χ(r) = 1, and lim α(r) = 0 (3.18) r r →∞ →∞
The numerical technique used to solve the field equations (3.17) is the fourth order
Runge-Kutta method with an adaptive step size [42]. First, it is necessary to reduce
52 the second-order differential equation (3.17) to a set of first-order ODEs,
dα = v(x), dr dv v(r) = + e2α(r)χ(r), dr r dχ = w(r), and dr dw w(x) α2(r)χ(x) = 2βχ(r) χ2(r) 1 dr − r − r2 − − where two new variables v(x) and w(x) have been introduced. The problem of solving
the field equations is thus reduced to studying a set of 4 first order ordinary differential
equations for the function yi, i = 1, ..., 4, having the general form
dy (r) i = z (r, y , y , y , y ), i=1,2,3,4 dr i 1 2 3 4 where the functions zi are known. The Runge-Kutta method uses information from previous steps as well as the values of the functions zi to propagate the solutions (yi)
over an interval from r r , where the step size is h = r r . Each stage in n → n+1 n+1 − n the fourth order Runge-Kutta method requires the following calculations,
k k k k y = y + 1 + 2 + 3 + 4 + O(h5), n+1 n 6 3 3 6 where,
k1 = hz(ρn, yn) h k k = hz ρ + , y + 1 2 n 2 n 2 h k k = hz ρ + , y + 2 3 n 2 n 2
k1 = hz(ρn + h, yn + h)
53 Now, the solutions (yi) can be propagated starting at r0, i.e. at r = 0 where the
boundary conditions gives the values of fields at the starting point. However, given
the nature of the differential equations and the boundary conditions this process has
two complications;
1. Boundary conditions (at r = 0) are only known for two of the fields, α(r) and
dα dχ χ(r). i.e. dr r=0 = v(0) and dr r=0 = w(0) are not known.
2. Two of the functions zi are singular at ρ = 0, and therefore the simulation will
diverge near this point.
The second issue can be addressed by displacing the starting point of the Runge-Kutta method by a small amount to r = , so that the singular nature of the equations at
r = 0 does not affect the simulation. However, this trick will only work if the values
of the fields at the new starting point can be determined, i.e. I need to resort to a
different technique to propagate the solution from r = 0 . This is done using a → power series method. Taylor series expansion of the fields together with the boundary
conditions at r = 0 can be used to analytically solve the field equations (3.17). This
technique generates expressions for the fields α(r) and χ(r) in powers of r.
However, note that these solutions are only valid for very small values of r, and
therefore, has to be kept very close to zero. Also, the derivatives of the two fields at
r = 0, α0(0) and χ0(0), appear in the solutions as two unknown parameters, a point
54 previously mentioned as one of the complication that arise when using the Runge-
Kutta method. This issue can be overcome by making use of the the fact that the form of the fields are known at large r, i.e. use the boundary conditions at r , → ∞ (3.18). I use a parameter fitting algorithm to pin down the two unknown boundary values at r = 0 that would produce the desired large r behavior.
3.2.2 Results
I solved the field equations numerically using the technique mentioned previously for the case where the winding number n = 1. The resulting field configurations of the
λ straight vortex for different values of the coupling constant β = e2 are shown in figure
(3.1).
Once the fields configurations are known I can calculate the energy of the straight vortices by evaluating the hamiltonian (3.16). The numerical integration is done using the adaptive Gaussian quadrature method. The values of the Hamiltonian calculated for each case is given in the graphs. Figure (3.1a) shows the field configurations at critical coupling β = 1/4, and the Hamiltonian yields
H = πη2 (1.0000027) ×
As stated previously in (3.14) the Hamiltonian for the critical coupling can be deter- mined analytically to be exactly H = πη2 n . Given that n = 1, the analytical value | | of the hamiltonian is H = πη2(1), and the numerical evaluation is in agreement to
6 decimal places. It can also be observed that the energy calculation of the remain-
55 HaL b=bc=1ê4
1 cHrL 0.8
0.6 H=ph2 äH1.0000027L 0.4
0.2 aHrL
r 2 4 6 8 10 12 14
HcL b=2bc HbL b=3bc
1 1 cHrL cHrL 0.8 0.8
0.6 0.6 2 H=ph2 äH1.260910L H=ph äH1.156767L 0.4 0.4 aHrL 0.2 aHrL 0.2
r r 2 4 6 8 10 12 14 2 4 6 8 10 12 14
b bc HdL b= ÅÅÅÅcÅÅÅ HeL b= ÅÅÅÅÅÅÅ 2 3
1 1 cHrL 0.8 0.8 cHrL
0.6 0.6 H=ph2 äH0.867911L H=ph2 äH0.800865L 0.4 0.4
0.2 aHrL 0.2 aHrL
r r 2 4 6 8 10 12 14 2 4 6 8 10 12 14
Figure 3.1: Plot of the straight vortex field configurations for different values of the coupling constant. (a) β = βc = 1/4, critical coupling. (b) β = 3βc. (c) β = 2βc. (d) β = βc/2. (e)
β = βc/3.
56 ing cases agrees with the lower bounds stated in (3.13). These agreements give us confidence in the numerical method we employ for the solutions.
Next I turn to the circular vortex. The field configurations of the straight vortex
developed previously will be used in the numerical treatment of the circular vortex
solutions.
3.3 The Circular Vortex
I start with the Hamiltonian of the U(1) model in 3 dimensions.
1 1 H = d3x (∂ A ∂ A )2 + (∂ + ieA ) Φ 2 + λ Φ 2 η2 2 (3.19) 2 2 ν µ − µ ν | µ µ | | | − Z Going over to dimensionless variables
x Φ ηΦ, A ηA, and x → → → eη
Since I am only interested in time independent solutions, the Hamiltonian takes the
form,
η2 (∂ A ∂ A )2 ∂Φ 2 H = d3x i j − j i + + iA Φ + β( Φ 2 1)2 , (3.20) 2 2 ∂x i | | − Z " i #
where I have introduced β = λ/e2. A toroidal coordinate system can be used to studt
this system. In other words, measure the distance ρ from a circle of radius R in the xy plane, the angle φ is the inclination out of the xy plane measured from the circle,
and the angle ψ is the angle of the projection in the xy plane on to the x axis. Then
57 the coordinate transformation from the Cartesian one is
x = cos ψ(R + ρ cos φ),
y = sin ψ(R + ρ cos φ),
z = ρ sin φ. (3.21)
The metric tensor takes the form,
cos ψ cos φ cos ψρ sin φ sin ψ(R + ρ cos φ) − − g = . (3.22) sin ψ cos φ sin ψρ sin φ cos ψ(R + ρ cos φ) − sin φ ρ cos φ 0 This implies that the Jacobian is
det g = ρ(R + ρ cos φ). (3.23) −
Given the form of the determinant, the measure is
2π 2π ∞ d3x dψ dφ dρ ρ(R + ρ cos φ)Θ(R + ρ cos φ), (3.24) → Z Z0 Z0 Z0
Since the action is independent of ψ, the integration can be performed. The leading term of the integral will be of the form
2πR ρ dρdφ... Z which is just the two dimensional action multipled by the length of the circular vortex, 2πR. The purpose of the subsequent calculations is the extraction of the leading correction terms of total energy dependent on the curvature radius, R.
58 The Θ-function under the integral is irrelevant to any finite power correction in
R to the straight vortex contribution, as the the integrand (powers of the vortex field
and its derivatives) decreases exponentially away from the circle where the vortex is pinned down. I use the following ansatz for the gauge field and the higgs field. The ansatz for the gauge field is
1 A (x) = φˆ a(ρ, φ), (3.25) µ µ ρ
ˆ where φµ is the unit vector in the azimuthal direction,
ˆ φµ = ρ∂µφ.
The ansatz for the Higgs field is
Φ(x) = ηeinφf(ρ, φ), (3.26)
where the function f is real. The gauge was fixed such that the phase of the higgs
field is exactly nφ.
Now I calculate the action using our ansatz. First of all, I need to calculate ∂ ∂ µ · µ in toroidal coordinates. This becomes equivalent to performing the calculation in two
dimensional polar coordinates as there is no dependence on angle ψ.
(∂ A ∂ A )2 (∂ φ∂ ρ ∂ φ∂ ρ)2 ∂a 2 1 ∂a 2 µ ν − ν µ = µ ν − ν µ = . (3.27) 2 2 ∂ρ ρ2 ∂ρ
The kinetic term for the higgs field is the following
∂Φ 2 1 ∂f 2 ∂f 2 + iA Φ = (a + n)2f 2 + + . (3.28) ∂x i ρ2 ∂ρ ∂φ i
59 Thus, the complete action, using this ansatz, takes the form
2π 2 2 2 ∞ ρ 1 ∂a ∂f 2 2 H = πη R dφ dρ ρ 1 + cos φ 2 + + (a n) f 0 0 R (ρ " ∂ρ ∂φ − # Z Z ∂f 2 + + β(f 2 1)2 (3.29) ∂ρ − )
This action differs from the one for the straight vortex in the driving correction
term in the measure and in the term containing the derivative of the Higgs field with
respect to the asimuthal angle. Again, it is easy to see, just like for the d-dimensional
1 1 kink, that to second order in R− only the leading correction, of O(R− ), to the Higgs
and gauge fields contributes. Thus, these fields can be written as
1 a = a0(ρ) + a1(ρ, φ), R 1 f = f 0(ρ) + f 1(ρ, φ). (3.30) R
The correction term contributes in the form
π H = H + H = H + (A + A ), (3.31) 0 1 0 R 1 2 where
2π 1 2 1 2 ∞ 1 ∂a ∂f A = dφ dρ ρ + + (a0 + n)2(f 1)2 + (a1)2(f 0)2 1 ρ2 ∂ρ ∂φ Z0 Z0 ( " ∂f 1 2 + 4(a0 + n)a1f 0f 1 + + β(f 1)2[3(f 0)2 1] , ∂ρ − ) 2π 1 0 ∞ ∂a ∂a A = 2 dφ dρ cos φ + (a0 + n)a1(f 0)2 + (a0 + n)2f 0f 1 2 ∂ρ ∂ρ Z0 Z0 ∂f 1 ∂f 0 + ρ2 + ρ2βf 1f 0[(f 0)2 1] (3.32) ∂ρ ∂ρ −
60 Clearly the variation gives inhomogeneous equations for the two unknown functions, a1 and f 1, where the inhomogeneity is proportional to cos φ. Then obviously, the
solution is such that both of these functions depend on φ such that
a1 = cos φ b(ρ),
f 1 = cos φ g(ρ). (3.33)
In terms of b and g the correction term to the hamiltonian is (using the notation
a0 a, f 0 χ to denote the straight vortex solutions) → →
2 π ∞ 1 2 2 2 2 2 2 H = ... + dρ [b0 + g + (a + n) g + b χ + 4(a + n)bχg] R ρ Z0 2 2 2 2 + ρg0 + 2βρg (3χ 1) + 2a0b0 + 2(a + n)bχ −
2 2 2 2 + 2(a + n) χg + 2ρ χ0g0 + 4ρ βχg(χ 1) . (3.34) −
The boundary conditions that ensure finite-energy solutions are
b(0) = g(0) = χ(0) = a(0) = 0,
lim b(ρ) = 0, lim g(ρ) = 0, lim χ(ρ) = 1, and lim a(ρ) = n (3.35) ρ ρ ρ ρ →∞ →∞ →∞ →∞ −
Now the equations of motion for b and g can be determined.
1 2 2 b00 b0 bχ 2(a + n)χg = ρa00 + ρ(a + n)χ , − ρ − − −
1 1 2 2 g00 + g0 [g + (a + n) g + 2(a + n)bχ] 2βg(3χ 1) ρ − ρ2 − −
1 2 1 d 2 2 2 = (a + n) χ (ρ χ0) + 2ρ βχ(χ 1). (3.36) ρ − ρ dρ −
61 Using the straight vortex field equations (3.11) for a and χ the right hand side of
(3.36) can be simplified to get
1 2 b00 b0 bχ 2(a + n)χg = a0, − ρ − − −
1 1 2 2 g00 + g0 [g + (a + n) g + 2(a + n)bχ] 2βg(3χ 1) = χ0. (3.37) ρ − ρ2 − − −
3.4 Numerical Treatment
I begin as I did in section(3.2.1) by redefining the straight vortex field
(n + a) = α.
The correction term H1 to the hamiltonian take the form,
2 π ∞ 1 2 2 2 2 2 2 H = dρ [b0 + g + α g + b χ + 4αbχg] 1 R ρ Z0 2 2 2 2 + ρg0 + 2λρg (3χ 1) + 2a0b0 + 2αbχ −
2 2 2 2 + 2α χg + 2ρ χ0g0 + 4ρ λχg(χ 1) , (3.38) − and the field equations change to
1 2 b00 b0 bχ 2αχg = α0, − ρ − − −
1 1 2 2 g00 + g0 [g + α g + 2αbχ] 2βg(3χ 1) = χ0, (3.39) ρ − ρ2 − − −
The boundary conditions for α can be stated as
α(0) = n and lim α(ρ) = 0. ρ →∞
62 Compared to the equations of motion of the straight vortex (3.17), the above system of equations for b and g are more complex. Accordingly, the numerical techniques that is used to solve (3.39) differ from that employed previously. Next, I will describe the two techniques used to solve the field equations of the circular vortex.
3.4.1 Variational Method
I start by using a variational method to obtain approximate solutions for the fields b and g that minimize the hamiltonian (3.38). Subsequently, I will use these approxi- mate forms of the fields as the initial ”guess” in an iterative technique to solve the
field equations (3.39). As indicated by the boundary conditions (3.35) both b an g tend to zero as ρ 0 and when ρ . The two trial functions given below has all → → ∞ these features.
2 γρ δρ b(ρ) = ρ b1e− + b2e−