You are on page 1of 29

Confinement

Contents
1 Introduction 1

2 Prerequisites 2
2.1 Yang-Mills Theories . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
2.2 Area Law for Wilson Loops . . . . . . . . . . . . . . . . . . . . . . . . . . . 5

3 Asymptotic Freedom 7
3.1 Derivation of QCD β-function . . . . . . . . . . . . . . . . . . . . . . . . . . 7
3.2 Implications of QCD β-function . . . . . . . . . . . . . . . . . . . . . . . . . 9

4 Toy Models in Low Dimensions 10


4.1 Maxwell Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
4.1.1 Pure Behaviour . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
4.1.2 Theta Term . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
4.2 The Abelian Higgs Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
4.2.1 d=1+1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
4.2.2 d=2+1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18

5 t’Hooft’s Dual Superconductor Model 21

6 Conclusion 26
1 Introduction
The grand success of perturbation theory is pervasive in modern quantum physics. Where
exact solutions to our theories are difficult or impossible to obtain, the successive
approximations provided by perturbative analysis allow us to connect these theories to
experimental data and so gain meaningful knowledge. We need not worry about exact
analytical solutions because our approximations are ‘close enough’.

Like Taylor series, perturbative expansions can only converge for sufficiently small
expansion parameters. In quantum field theories (QFTs) where the perturbative
expansions are typically drawn as Feynman diagrams, these expansion parameters have
physical meaning; they are the coupling constants which describe the interaction strength
between particle fields. Renormalization techniques used to account for field self-energies in
QFT lead to an understanding of how these coupling constants depend on energy scale. In
quantum electrodynamics (QED), for example, the coupling constant between photons and
electrons is known to increase as energy increases; therefore QED accurately models the
low-energy systems of nature, such as atoms and scattering electrons.

However the coupling constants of quantum chromodynamics (QCD), the theory of the
strong nuclear interaction, can be shown to diverge at low energies. This remarkable
property is known as asymptotic freedom, and while it has the benefit of providing a
convenient free-theory description of high-energy quarks, it requires that non-perturbative
techniques be applied to understand QCD at the energy levels of familiar bound-state
hadrons and mesons.

A full analytic understanding of the non-perturbative regime of QCD still escapes


reckoning. However we are confident from many experiments and from lattice simulations
of QCD that certain qualitative properties hold. One of these properties is that quarks and
gluons are confined : they are never observed in singlet states, but rather in colour-neutral
multiplet states such as mesons and hadrons. Furthermore, the interquark potential is
known to be linear with distance [1]: V (r) = σr.

In this essay I will first give an overview of Yang-Mill theories and the Wilson loop, the
necessary machinery for much of the following analysis (Section 2). I will then discuss
asymptotic freedom and outline a derivation of the beta function for QCD (Section 3).

1
Having motivated the concept of confinement, I will discuss a compact gauge theory known
as the Abelian-Higgs model, which exhibits confinement in both 1+1 dimensions and 2+1
dimensions (Section 4). Finally, I will explore a model due to t’Hooft wherein the QCD
vacuum is described as the electromagnetic dual to a Landau-Ginzburg superconductor
(Section 5).

2 Prerequisites
Gauge theories constructed from non-Abelian gauge groups enjoy a great diversity of physical
applications; including both the QCD and electroweak sectors of the Standard Model. Why
are they so effective at describing aspects of reality which are qualitatively very different?
The answer is in the name: gauge symmetries, which are not symmetries at all, give infinitely
redundant descriptions of each physical state. Given freedom to choose whichever description
we like for each state, we can therefore construct theories which ‘make manifest ... Lorentz
invariance and locality and, in the quantum theory, unitarity’ [2] but which make more
specific predictions upon identification of the states with our chosen gauge group.

2.1 Yang-Mills Theories


This procedure is historically a generalization of Maxwell theory, where the gauge field Aµ
which incorporates all electromagnetic fields is redundant under any gauge transformation
Aµ → Aµ + ∂µ α, with α a smooth scalar field. Any interacting theory of electromagnetism
with some other fields must retain this invariance or lead to contradictory descriptions of
the same interacting state. Though not the historical route, a geometric path of reasoning
following [3] gives great insight into the classical structure of the gauge theory and is
presented here.

We work with an SU (n) gauge symmetry group G as an example, although a gauge theory
can be constructed using any compact, simple Lie group. The group can then be
represented faithfully by matrices V acting on an n-plet field ψ. Typical local Lagrangian
terms such as ψ̄ψ and ∂µ ψ̄∂ µ ψ are naturally invariant under any such symmetry, along
with higher-order interaction terms. We would like to further require that the symmetry
holds locally:
VG : ψ a (x) → ψ 0b (x) = V ab (x)ψ a (x) (1)

2
Allowing the group action to be independent at each spacetime point distorts the terms
which involve derivatives. Explicitly, a directional derivative of the field along η µ ,
1
η µ ∂µ ψ ≡ lim [ψ(x + η) − ψ(x)], (2)
→0 

attempts to compare two field values which transform in different embedded groups G,
thought of as forming a bundle over spacetime. However, we can define a nonlocal
‘comparator’ operator U which relates two distant group transformations, and transforms
under G as:
VG : U (y, x) → V (y)U (y, x)V −1 (x) (3)
Geometrically, any such comparator allows for consistent comparison of fields living at x
and y even as they transform under the gauge group. The resultant covariant derivative
lim→0 1 [ψ(x + η) − U (x + η, x)ψ(x)] ≡ η µ Dµ ψ is seen to transform in a simple way:

VG : η µ Dµ ψ → V (x)η µ Dµ ψ, Dµ ψ̄Dµ ψ → Dµ ψ̄Dµ ψ (4)

where the latter expression shows just one of many gauge-invariant derivative terms which
may now be included in a Lagrangian. In order to write down such expressions in terms
of the generators of G, we restrict U to act unitarily on group representations V, and set
U (y, y) = 1 in order to Taylor expand:

U (x + η, x) = 1 +  · iη µ Aaµ ta + higher order terms (5)

Here ta are the infinitesimal generators of G which form a basis for Lie algebra L[G]. All
we have really said is that infinitesimally, the comparator U is some linear combination of
L[G] basis elements with one-form-valued coefficients Aa . These Aa contain all information
necessary to describe our new locally gauge-invariant terms, and are accordingly called
gaugef ields. Manipulations with the above Taylor expansion [3] show that A transforms
as in Maxwell theory for gauge group U (1).

A natural next step is to consider whether invariant terms can be constructed which have
no ψ-dependence; again this will have a geometric interpretation. From the comparator
transformation law (3), a product of comparators which form a closed path at some point z
will transform as:

VG : U (z, x1 )U (x1 , x2 )...U (xn , z) ≡ U → V (z)UV (z)−1 (6)

Though this quantity is not gauge-invariant, it may be made so by taking a trace over the
closed path and applying the cyclic identity on traces. Comparators in this context are also

3
All the ingredients necessary to define a Yang-Mills theory have now been developed. A
Lagrangian theory which is locally invariant under gauge group G has been constructed by
introducing gauge field(s) Aa and modifying derivative terms so that they are covariant
with G. All that remains is to define the self-interaction of the gauge field(s), by
supplementing the Lagrangian with the invariant term we have derived:
1 a
LY M = T r(Fµν F aµν ) (8)
2
where the constant coefficient comes from normalizing the trace of the Lie algebra generators
as T r(ta tb ) = 12 δ ab [1].

2.2 Area Law for Wilson Loops


By introducing Wilson loops, we arrived at a quantity (the field strength tensor) which is
the key observable in example Maxwell theory; it remains equally important in more
involved gauge theories. Quantitatively describing confinement will require examining how
field strength depends on the separation between source particles: the Wilson loop will
thereby provide a simple test for confinement known as the area law.

Analagous to the Ricci curvature in general relativity, the field strength geometrically
measured an ‘angular deficit in internal field space’ [1] for particles. We can similarly
interpret the Wilson loop as a parallel transport operator, following a trajectory in internal
field space which satisfies Dµ (U ) = 0. (5) already gives an infinitesimal solution to this
differential equation; we obtain a finite solution by treating the Taylor series in its
exponential form and integrating over a path from x to y:
Y Z
µ
UP (y, x) = exp[−ηi (iAµ (x + ηi ))] = exp[−i dxµ Aµ (x)] (9)
i

To be truly consistent for non-Abelian gauge groups, the integral should be constructed so
that the gauge field terms are multipled in order along the path, but this notation is
suppressed here for readability.

So far we have not strayed from the classical realm, but conveniently, (9) has another
interpretation: the partition function obtained by considering the Hilbert space for
complex-vector quark states [3] and integrating out the internal variables specific to quarks.

5
Substituting the known linear interquark potential E(r) = σr, the resultant exponentiated
factor is simply the area A = r · T of the Wilson loop, up to a constant factor. The final
proportionality relation is known as the area law, and will be taken in this essay as the
defining characteristic for a confined theory. Conversely, non-confined theories or phases of
theories, such as the Higgs phase to be discussed in Section 4, have a ”perimeter law” where
the Wilson loop expectation value goes exponentially with the length of the closed path. [2]

3 Asymptotic Freedom
Having thoroughly defined Yang-Mills theories, we can further define QCD as the Yang-Mills
theory with gauge group SU(3), where the fundamental quark fields are fermionic. Peskin
and Schroeder [3], as well as Tong [2], derive the beta function for QCD via the background
field method. The significance of this beta function is that it is negative for all energy
scales g, implying that coupling constants of QCD become stronger at low energies. This
phenomenon is known as Asymptotic Freedom, and it will motivate our discussion of quark
properties at the energy scales where they are confined in familiar cold matter. I will follow
[2] in studying a ”pure” Yang-Mills theory without matter terms, though the method and
result generalize well with the inclusion of fermionic fields [3].

3.1 Derivation of QCD β-function


The background field method proceeds by separating the gauge field Aµ into components of
frequency below (Āµ ) and above (δAµ ) some cutoff k, then integrating over the
higher-energy modes to generate an effective field theory for the lower-energy modes (the
background field). Renormalization counterterms in the one-loop correction are calculated,
and the resultant gauge coupling constant of the effective theory is taken to be the
appropriate running of this coupling with scale. The derivative of the running function
with respect to the log of the scale returns the beta function of the theory.

First, the Yang-Mills Lagrangian (8) is expanded in terms of Āµ and δAµ . The algebra is
tedious but benefits from the antisymmetry properties of the adjoint action of the gauge

7
field, and the result (with gauge coupling restored and gauge indices suppressed) is [2]:
1 1
LY M = [T r( F̄µν F̄ µν + 2F̄ µν Dµ δAν
g2 2
+ Dµ δAν Dµ Aν − Dµ δAν Dν δAµ − iF̄ µν [δAµ , δAν ] +O(δA3 )] (14)
| {z }
quadratic terms

where all covariant derivatives are taken along the background gauge field Āµ , and F̄
represents the field tensor with only background gauge field contributions. Formal path
integration with a Yang-Mills action requires that the gauge be fixed, and here the
Faddeev-Popov method comes into play [3]. A zero-condition chooses a slice through the
gauge space:
G(A) = Dµ δAµ − ω = 0 (16)

where ω is an arbitrary scalar field, which is removed by a convenient Gaussian integration


within the partition function to produce a separable gauge-fixing term Sgf :
Z  Z  
iSY M (A) δG(A) iSY M (A) 1 µ 2
DA e = det DαDA e · exp 2 T r(D δAµ ) (17)
δα g
| {z }
Sgf

where α is the parameter of the gauge transformation Aµ → Aµ + g1 ∂µ α, along which the


condition G[A] is varied in the Faddeev-Popov method. After expressing the functional
determinant as a path integral over Grassmann-valued (complex) field c, the complete gauge-
fixed action is a sum of three terms: S = SY M + Sgf + Sghost . Now, in order to produce an
effective action for the background field alone, both δAµ and c must be integrated out. At
the one-loop level, only terms quadratic or lesser in these fields need be considered; the linear
term in (15) vanishes because this term is equivalent to the classical equation of motion for
F̄ after integration by parts. Fortunately, there are only three quadratic terms in the full
action S:
Z
1
Squad = 2 d4 x T r[Dµ δAν Dµ δAν − 2iF̄ µν [δAµ , δAν ] + Dµ c† Dµ c] (18)
g

Integrating out fields now is purely an exercise on Gaussian functional integration [3] over
bosonic gauge operator −D2 δ µν + 2i[F̄ µν , ·] and fermionic ghost operator −D2 . Each
contributes a constant coefficient to the background-field Yang-Mills action, as a result of a
sum of the contributions from Aµ and c as one-loop diagrams. These functional
determinants are multiplicative in the exponential of our path-integral partition function

8
(17), and so represent additive contributions to the gauge coupling [2]:
 
   2
1 1 Cadj  1 1 4 Λ
= 2+ − + − 8  log (19)

2
g (k) g 2
(4π) |{z}3 2 3  k2
| {z }
ghost gauge

where Cadj is a constant of the gauge Lie group which normalizes its generators in the adjoint
representation, in which the gauge fields reside, and Λ is the renormalization scale necessary
to cancel logarithmic divergences from the one-loop computation [3]. Taking the case of
QCD, where Cadj = 3, the beta function for g is then:

11g 3
β(g) = − (20)
(4π)2

Introducing n fermions into the theory, the constant 11 is countered by a − 23 n factor [3].
But unless n is quite large, we have proven that the beta function is negative not only for
QCD, but for all non-Abelian gauge theories. Therefore the gauge coupling constant runs
downward at high frequencies, and these theories are asymptotically free.

3.2 Implications of QCD β-function


We should venture back one step and examine the contributions from the ghost and gauge
determinants in (19). The first two terms are small compared to the third, which is largely
responsible for the overall negative sign of the beta function, and therefore for asymptotic
freedom. In physical terms, why do different types of fields want to pull the coupling in
opposing directions? What’s going on here?

In electromagnetic materials, we recall perhaps the most familiar example of a field theory
whose effective strength is modulated by other fields (matter fields) which it traverses.
Electric fields experience screening by both free and bound charges in a material; magnetic
fields can experience both screening and strengthening depending on the magnetisation of
a material. In QED, where the beta function is positive due to vacuum polarization [2],
screening is a very intuitive explanation for the decrease in electron charge / coupling
strength at large distances.

It is not unreasonable to interpret asymptotic freedom as an ’anti-screening’ of quark charge


analagous to paramagnetism. While a small diamagnetic contribution arises in part of the

9
gauge determinant contribution – perhaps some screening effect comparable to QED vacuum
polarisation – it is clearly overpowered by a paramagnetic contribution which arises [2] from
the ’spin 1 nature of the gauge field’. As we continue onward to examine specific mechanisms
for confinement in a variety of (not totally proven) models, I find it personally motivating
to know that a firm explanation is known for the strong-coupling phenomenon which gives
rise to confinement.

4 Toy Models in Low Dimensions


Confinement is most easily manifested as a feature of simple physical systems in 2 and 3
spacetime dimensions. The regimes we will explore bear few fundamental similarities to
QCD (especially as they are Abelian), but the incentive for exploring them is that the
manner in which confinement appears might prove similar to the manner in which it
appears in physical QCD.

Dimensional analysis of the sourced Maxwell action in d dimensions [2] will further
motivate our choice of low-dimensional systems:
Z
1
S = dd x − 2 F ab Fab + Aµ jµ (21)
2e
where A is the gauge field over d indices and F is the field strength tensor
Fµν = ∂µ Aν − ∂ν Aµ . In natural units, [S] = 0, and the definition of the covariant derivative
gives [A] = 1. Therefore [F ] = [A] + 1 = 2 and [e2 ] = 4 − d.

In our usual 3+1 electromagnetism, the gauge coupling e2 is therefore marginally relevant;
such couplings can give rise to interesting behaviour at either end of the energy spectrum,
but we have been aware since the era of silk and glass rods that electric charge is not
confined. Instead we will study 1+1 and 2+1 Maxwell theory, where the gauge couplings
are definitely relevant and therefore ensure strong-coupling behaviour in the IR region. We
will then further introduce a complex scalar ”Higgs” field couplied to the gauge field,
exploring the appearance of a confined phase in what is known as the Abelian Higgs model.
Significantly, this confined phase will occur in disagreement with perturbation theory
intution regarding the behavior of such a theory.

10
4.1 Maxwell Theory
What follows is not your grandmother’s Maxwell theory. Simple evaluations in 1+1 and 2+1
pure theory will confirm confinement of electric charge. However, topological features unique
to each dimensionality will provide a richer description of confinement properties, and will
play an important role in the generalization to the Abelian Higgs model.

4.1.1 Pure Behaviour

Applying the Euler-Lagrange equation to action (21), the classical motion of the gauge field
is derived:
∂L ∂L 1
∂µ = =⇒ ∂µ F µν = j v (22)
∂(∂µ Aν ) ∂Aν e2
But in 1+1 dimensions, the antisymmetry condition on F µν leaves it with only one degree
of freedom, the electric field E ≡ F 01 , such that (setting e = 1)

∂1 E = j 0 and − ∂0 E = j 1 =⇒ ∂µ j µ = 0 (23)

where the right-hand equation gives conservation of source charge. From these we can
intuitively understand that a point charge maintains a discontinuity of E (in x1 ), while an
instantaneous line current source produces an instantaneous jump in E (in x0 ). These
correspond to the boundary conditions observed in 3+1 electrodynamics, e.g. infinite
charged planes which partition the space. There are no wavelike solutions which can persist
without a sustained source.

Electric fields store energy, and a single point charge or line current configuration in 1+1
therefore stores infinite energy; this is not a stable configuration. Viewed from a sufficient
distance, the field due to any charge configuration is approximated well by a point source
with the net charge of the configuration – therefore only neutral configurations will have an
electric field which vanishes at infinity. A prototypical configuration is a pair of opposing
charges with separation L. By (23), point charges j0 (x) = ±qδ(x) generate electric field
discontinuities which cancel, leaving zero field at infinity in both directions. The electric
field has magnitude q in the intercharge region, and evaluating the Hamiltonian from (21)
gives [2]:
1 2 q2
Z
H = dx E = L (24)
2 2
This is the simplest theory of confined charges which reflects the observed linear interquark
potential. It emerges by requiring that the electric field is conserved on a path between the

11
4.1.2 Theta Term

Although Maxwell theory describes classical electrodynamics completely, it does not do so


uniquely. ‘Topological’ terms may be added to the Lagrangian (21) which are total
derivatives; therefore they do not impact the classical theory except at the (infinite)
boundary. However, they will have a major impact in theories on closed manifolds, and
even more drastically so in the quantum theory. Most relevantly, they will complicate and
enrich our description of confinement.

The Hodge star is a bijective operation on differential forms which expressses the duality
between p-forms and (n − p)-forms [4]. The relevance to Maxwell theory is clearest in 3+1
dimensions, where the two-form field tensor F µν has a dual field tensor ?F µν = 12 µνρσ Fρσ
which physically switches the electric and magnetic fields in the theory. The resultant
representation of Maxwell’s equations is manifestly dual. Because ?F µν is necessarily gauge-
invariant, the natural combination ?F µν Fµν is both gauge- and Lorentz-invariant [2], and is
called the theta term for 3+1 Maxwell theory:
Z
Sθ = constants · θ d4 x (?F µν )Fµν (26)

Here θ is introduced as a dimensionless paramater which tunes the strength of the term.
Evaluating this expression explicitly in terms of Aµ , and integrating by parts:
Z
Sθ ∝ d4 x µνρσ ∂ρ Aσ ∂µ Aν
Z
= d4 xµνρσ [∂ρ (Aσ ∂µ Aν ) − Aσ ∂ρ ∂µ Aν ]
Z
= d4 xµνρσ ∂ρ (Aσ ∂µ Aν ) (27)

where the only non-boundary term vanishes due to symmetry. Thus the theta term is
indeed a total derivative.

In 1+1 dimensions, the situation is even simpler because the dual field tensor is in fact a
Lorentz scalar ?F = 12 µν Fµν = E, and by default contributes a boundary term to the action
[2]: Z  
2 1 2 θ
Sextended = dx 2
E + E (28)
2e 2π
As an aside, this construction is not possible in 2+1 dimensions because the dual field
tensor is a Lorentz vector Gρ = ρµν Fµν containing the two electric field and one magnetic

13
field variables of the theory. As discussed in Section 2.1, there are no other gauge-invariant
Lorentz vectors in the theory, so the only candidate theta term would be inner product
Gρ Gρ , which clearly is not a trivial physical quantity.

Returning to (28), our Maxwell theory has been shown to have no propagating solutions;
quantization will proceed more meaningfully by requiring the theory to live on a closed
manifold. The natural choice is to identify the spatial dimension as a circle S 1 with radius
R, following [2]; freedom from boundary conditions applied at spatial infinity, this setup
allows the theta term to affect the quantum spectrum. We work in A0 = 0 gauge, so that
the only physical degree of freedom is the total A1 configuration:
Z 2πR
φ(t) = dxA1 (x, t) (29)
0

which is called a zero mode by virtue of its zero-approaching frequency. In this gauge,
E = ∂0 A1 , so extended action (28) can be expressed in terms of φ [2]:
Z
1 θ
Sextended = d2 x 2
(∂0 φ)2 + ∂0 φ (30)
R×S 1 4πe R 2π
Basic quantum mechanical manipulations of the φ-dynamics lead to an energy spectrum:
 2
θ
Ĥ |ψl i ∝ R l − |ψl i , |ψl i = eilθ |0i (31)

where the energy eigenstates have l ∈ Z. The only complication from a standard
potential-well problem is that rather than boundary conditions, it is the enforced gauge
invariance of the quantum theory which leads to plane-wave solutions with quantized
momentum l. Gauge invariance of φ(t) must hold even under large gauge transformations
A1 → A1 + ∂1 ω where the eigenstate transforms in a well-defined manner (ψ → eiω ψ), but
ω is not necessarily periodic. The result is that periodic values φ ∼ φ + n2π are identified,
and the eigenstates given in (31) are the only plane waves which respect this identification.

The action of the electric field operator on energy eigenstates can be extracted by comparing
(28) and (31), and reveals the relationship between the theta parameter and a ‘background
electric field’ [2]:  
θ
Ê |ψl i ∝ l − |ψl i (32)

Because this result is not dependent on the compactification radius R, it can be interpreted
to remain valid for a non-compact spatial dimension.

14
It follows that the imaginary part of the instanton action contribution, which must match
inθ
up with the theta term in (34) to generate this background flux, is 2π
. Again we see that
the presence of the theta term diversifies the Hilbert space of the system under study. As
we are unable to shed light on the interaction of this gauge field with charges by classical
reasoning, the Wilson loop is employed similarly as in Section 2.2 to insert probe particles
of charge q:
 Z  Z
µ µ
T r(UP ) = exp −iq dx A (x) =⇒ hT r(UP )i = DADφ exp[−SE [A, φ]] (35)

The approximations made in taking this path integral are that the instantons, while
numerous, are well-separated ant non-interacting. Furthermore, we assume no instantons
intersect path C, so that the above expression can be expressed as a product of two
spacetime partition functions normalized to the general partition funciton:

hT r(UP )i = (Zinside [θ] · Zoutside ) [θ]/Z[θ] (36)

Temporarily considering as before a compactified spacetime with finite area A, the partition
function over all single-instanton configurations neglecting the Wilson loop takes the form
inθ
Zsingle [θ] = A · K · e 2π (37)

where the real part of the action exponential has been absorbed along with quantum
corrections to the instanton into a constant K [2], and instantons are taken to be
homogeneous throughout spacetime. To account for the full combinatorial range of
multi-instanton configurations, [2] assumes only n = ±1 contributions to the partition
function, and finds that
Z[θ] = exp[2A · Kcosθ] (38)
This may be substituted into (36) by noticing that the outside partition function is not
influenced by the Wilson loop; it is the general partition function (38) but with subtended
area less the Wilson loop area AC . Inside the loop, however, the Wilson loop integrates over
the asymptotic winding number of contained instantons, contributing an imaginary action
term iq per instanton; interpreting this change as a theta term, it is as though parameter
θ has been increased by 2πq, which may be substituted in (38) over loop area C. Now
evaluating (36) gives:

hT r(UP )i = exp[2AC · K(cos(θ + 2πq) − cos(θ))] (39)

Comparing with (13), immediately we can conclude that charges are confined by a linear
potential (in fact even if θ = 0). The caveat is that test charge q must not be an integer

17
monopole operators which are compatible with the theory. In pure Maxwell theory, there
exists a conserved quantity:

µ 1 µνρ µ
Jtop =  Fνρ , ∂µ Jtop =0 (40)

However, we may study a larger but still consistent theory space, in which sources for the
current J µ may be inserted as monopole operators M(x) transforming under a
‘topological’ U(1) symmetry group [2]. In the Coulomb phase of the Abelian-Higgs model,
where m2 >> e4 in analogy to the d = 1 + 1 case, charged scalar particles are too massive
to interfere with the U(1) symmetry of Maxwell theory. These excitations clearly are not
magnetic monopoles, and therefore the monopole U(1) symmetry is broken to its default
no-monopole state.

This monopole can be approximated as a pure phase source M(x) ∼ eiσ(x) of the topological
current, in order to evaluate the partition function for the no-monopole theory. Integrating
out Maxwell fields F to obtain an effective partition function for the source σ(x), [2] finds:

e2
 Z 
3 µ
Zeffective [σ] = exp − d x 2 ∂µ σ∂ σ (41)

The result is an action formulation for an effective ‘dual’ photon which propagates in
spacetime and enforces the no-monopole configuration. Importantly, it is not distinct from
the single-polarization photon which necessarily exists in d = 2 + 1 Maxwell theory; it is
the same phenomenon, but expressed in a rich way which will generalize to the Higgs phase
of the theory. With no monopoles to impact large-scale charge interactions, this phase of
the theory remains log-confined as previously found for pure Maxwell theory.

In the Higgs phase (m2 << −e4 ), again neutral Higgs particles appear due to a breaking of
the Maxwell gauge symmetry; and again, the vector boson which appears to be identified
with the Maxwell photon has a finite mass [2]. This implies, with more explicit evidence to
follow, that magnetic monopole states are no longer forbidden, for (41) clearly no longer
holds true. Then the topological U(1) symmetry which these monopoles inhabit is
unbroken, and the monopoles are charged. We can expect that they will have a dramatic
impact on the confinement of charges in this phase of the theory.

Rather than defining suitable monopoles from thin air, we can arrive at them naturally by
considering a non-Abelian SU(2) variety of the Higgs model. This leap is justified because

19
SU(2) is the smallest gauge group for which instanton configurations on a 2-sphere at
spacetime infinity occur with an integer winding number ν, in the same sense as Figure 6
on a 1-sphere. In close parallel, a gauge field vortex is required to compensate for divergent
effects of this monopole upon the action; the resultant ‘t’Hooft-Polyakov’ monopole has
fixed magnetic flux 4πν [2]. A similar computation of the path integral over configurations
of such monopoles is not uniquely enlightening, and is complicated by the need to include
1
r2
interactions between charged monopoles. However, the resultant expression can be
rearranged as a path integral over a scalar field σ in order to compare to the massless dual
photon degree of freedom in the Coulomb phase [2]:
Z  Z 
3 2
Z = Dσ exp − d x Cσ (∂µ σ) − Kcos(2σ) (42)

where Cσ is a constant of the monopole Coulomb interaction, and K is again the exponetiated
real portion of the single-monopole action, with quantum corrections. We no longer have a
massless dual photon, but rather a massive one which lives in a periodic ‘winding’ potential.
The equation of motion for this effective action is:

Cσ ∂ 2 σ = −Ksin(2σ) (43)

which is called the Sine-Gordon model. This dynamical system is famed in the study of
solitons, stable field configurations which behave as pseudoparticles. Finite energy solutions
which connect the chain of vacua in the theory are found by Bogomolnyi’s method [7];
they are domain walls, planes in three dimensions which divide spacetime into two vacuum
regions. For our particular example, they have a constant energy density:
p
E∝ Cσ · K. (44)

For the massless dual photon (41), the conserved Noether current under topological U(1)
µ e2
symmetry is Jtop = (2π)2
∂ µσ [2]. Comparing with the original physical definition of the
topological current (40) reveals the field observable associated with spatial variance in σ
(choice of the ∂2 direction is arbitrary):

e2
E= ∂2 σ (45)

The domain walls described then contribute zero field outside of their active region, but
constant energy E implies a constant electric field in the plane of the kink. Returning to a
Minkowskian spacetime picture, monopoles inserted by σ are connected by an electric field
string, of constant flux, which propagates in time to form a local Sine-Gordon domain wall.

20
This results in a pair potential between charged monopoles (the only charged particles in
the phase) which is linear with separation, and we find that the Higgs phase of the
Abelian-Higgs theory is linearly confined.

It is noteworthy that while we left the Abelian-Higgs model behind somewhat in order to
introduce t’Hooft-Polyakov monopoles into the theory, this SU(2) gauge theory reduces to
an Abelian-Higgs theory at low energies, and can be interpreted as ”U(1) gauge theory, with
a UV cut-off” [2].

5 t’Hooft’s Dual Superconductor Model


We have now examined how confinement properties can emerge in two- and
three-dimensional theories with specific topological features. However our (macroscopic)
universe has four spacetime dimensions, and it is well-established that electromagnetism
and its variants are not confined theories. t-Hooft’s ”Dual Superconductor” model sets out
to explain, by analogy with the Landau-Ginzburg model of superconductivity, why
chromoelectric charges in Yang-Mills theories have a linear pair potential [1]. In fleshing
out the details of this description, tools which were useful in the above toy models will
again play a key role: the Wilson loop area-law corresponds to a linear potential, and again
topological vortices provide the physical mechanism for separated charges to form a bond.

Although this essay will not elaborate at length on the shortcomings of t’Hooft’s model,
shortcomings are not lacking in abundance. While charge-singlet states (quarks) are confined,
the model does not provide a mechanism for the confinement of colour charge in gluons; this is
seen by an extension of the argument which will be given in the adjoint representation rather
than the fundamental representation of the gauge group [6]. The theory does agree well with
some properties observed in lattice simulations of QCD, for example the appearance of flux
tubes, and the finding that monopole condensation and confinement have the same critical
temperature and therefore are fundamentally related. However other key lattice features,
in particular the scaling of the confining potential with the Casimir operator of its particle
representation, are not accounted for. T’Hooft himself acknowledged [8] that a ‘precise
calculational procedure [is a] more distant aim’, and that the value of the model rather
lies in the enforcement of electromagnetic duality in the particle spectrum while avoiding
unphysical ‘phantom solitons’.

21
Following [1] we turn again to the Abelian-Higgs model, due to the vortex solutions which
it permits, as a candidate for a confined theory. Unlike the low-dimensional examples, this
theory is an established working model of an oft-studied physical phenomenon:
superconductivity, in particular modelled as a condensate of electrons. Called the
Landau-Ginzburg Model in this nonrelativistic context, the theory’s vortices will
differentiate between Type I and Type II superconductors. More relevant for our purposes,
they will manifest the confinement of magnetic charge within the bulk of the
superconductor.

Superconductivity is thermodynamically described by a second-order phase transition at its


boundary from the ‘standard’ conducting phase. In such a phase transition between
significantly less- and more-ordered states, an order parameter which numerically probes
the transition is well-defined. For Landau-Ginzburg theory this parameter is a complex
field φ, where physically |φ|2 ‘indicates the fraction of electrons that has condensed into a
superfluid’ [9]. Wilson loops will also take on an interpretation as an order-parameter for
this theory.

Following [1], the Abelian-Higgs action in d = 3 + 1 is:


2
v2
Z   
4 1 µν ∗ µ ∗ ∗
S = d x − Fµν F + (Dµ φ) (D φ) − V (φ φ) where V =λ φ φ− (46)
4 2
V is the familiar Higgs potential, with minima at |φ| = √v forming the classical vacuum.
2
Solutions must have zero energy at spatial infinity or cause the total energy of the system
to diverge. It is instructive to examine the topologically-distinct classes of such solutions.

Given some vacuum state φ0 = 0 at spatial infinity, other stable vacuum states φvac cannot
have a continuous deformation φ0 ' φvac , or one would minimize to the other during
evolution of the system; that is to say they must be homotopy inequivalent. The φ minima
under condition |φ| = √v are topologically a circle S 1 . In three spatial dimensions, the
2
topological structures which are non-trivial at infinity may be classed as point, line, and
plane defects. For plane defects, the bulk is partitioned; as the manifold of vacuum
solutions is simply connected, both sides of the partition are simultaneously minimized to
φ0 . For point and line defects, spatial infinity is topologically S 2 or S 1 respectively. The
homotopy groups of spheres in low dimensions are well known [10]:

π2 (S 1 ) = 0 π1 (S 1 ) = Z (47)

22
vortices.

The similarity in appearance betweeen the t’Hooft operator (48) and the Wilson loop (9) is
clear: in fact the two are related by electromagnetic duality, which interchanges electric
and magnetic fields. Just as a linear potential on magnetic charges has emerged in the
above argument via magnetic flux vortices, we can define a dual-superconductor theory
where the Wilson loop creates electric flux vortices, confining electric charges. All that is
required is that these electric charges live in a vacuum which is defined by the
electromagnetic dual to Landau-Ginzburg theory.

In order to thoroughly express the dual-superconductor formulation for a Yang-Mills theory


such as QCD, rather than an Abelian U(1) theory, we would have to define equivalent t’Hooft
operators for every U(1) subgroup of the gauge group G = SU (n) [1]. For fields which
transform in the adjoint representation of G, such as the gauge fields, this construction
proceeds smoothly and leads to a set of n vortices. However the computation is dependent
upon fields being Zn − invariant, implying that the gauge group for these fields is really
SU (n)/Zn , where Zn is the cyclic group with n elements. Unfortunately, quark fields in QCD
are not invariant in this way, as they live in the fundamental representation of the gauge group
[1]. Stable vortices do not emerge from the same creation operators. One interpretation of
this phenomenon is that quark-antiquark pair production does not allow long vortices to
persist; instead they will break into locally-confined mesons. Regardless, while a dual-
superconductor model for the QCD vacuum does incorporate confinement properties in a
natural way, it is not possible to exactly apply the Wilson area law to quarks.

25
6 Conclusion
We have completed a tour of confinement in gauge theories. Motivated by the principle of
asymptotic freedom in non-Abelian Yang-Mills theories, we have explored the causes and
characteristics of confinement through models in two, three, and four spacetime
dimensions. This tour has been by no means exhaustive; in [2] and [1] alone there are
several more complete models with distinct features which exhibit confinement, and the
literature on this topic is incredible extensive.

All of models detailed here have relied upon topological properties to produce interesting
non-perturbative effects. These properties are not defined at any single spacetime point of
the theory, but depend on large-scale or even global configurations of dynamical fields; thus
it is not surprising that perturbation theory cannot capture them. The charged instanton
and vortex solutions we observed in these theories served to anti-screen charged source
particles, an effect opposite to familiar vacuum polarisation in QED. For the d = 1 + 1
Abelian-Higgs model in particular, we were able to study how two distinct topological
features, the theta term as well as instanton solutions, produce competing varieties of
confinement in a two-dimensional theory space.

There are also, of course, many more nuances to the three models discussed in this essay
than could be discussed in these pages. Confinement is a continually evolving area of study,
closely tied as it is to the frontiers of both experimental hadronic physics and theoretical
understanding of non-perturbative gauge theories. When an analytic proof for confinement
is found at last, I suspect on intuition that it will include many of the qualitative elements
overviewed here.

26
References
[1] Nair, V.P. Quantum Field Theory: a Modern Perspective. Springer, 2011.
Chapters 10, 19.

[2] Tong, David. Gauge Theory. 2018.


http://www.damtp.cam.ac.uk/user/tong/gaugetheory/gt.pdf Sections 1, 2, 7, 8.

[3] Peskin, Michael Edward, and Daniel V. Schroeder. An Introduction to Quantum Field
Theory. CRC Press, 2018.
Chapters 9, 12, 15, 16.

[4] Lahiri, Amitabha . Lecture Notes on Differential Geometry for Physicists. 2011.
http://bose.res.in/∼amitabha/lecnotes.html
Chapter 17.

[5] Coleman, Sidney. Aspects of Symmetry: Selected Erice Lectures of Sidney Coleman.
Cambridge University Press, 1999.
Chapter 7, Section 4.

[6] Ripka, Georges. Dual Superconductor Models of Color Confinement. Lecture Notes in
Physics, 2004, doi:10.1007/b94800.

[7] Chua, Dexter. Part III - Classical and Quantum Solitons, based on lectures by Manton,
N.S. and Stuart, D. 2017.
https://dec41.user.srcf.net/notes/III E/classical and quantum solitons.pdf

[8] T’Hooft, G. Topology of the Gauge Condition and New Confinement Phases in Non-
Abelian Gauge Theories. Nuclear Physics B, vol. 190, no. 3, 1981, pp. 455478.,
doi:10.1016/0550-3213(81)90442-9.

[9] Ginzburg, Vitaly L. On Superconductivity and Superfluidity (What I Have and Have
Not Managed to Do), As Well As on the Physical Minimum at the Beginning of the 21st
Century. Nobel Lecture, 16 July 2004, doi:10.1007/978-3-540-68008-6 1.
https://onlinelibrary.wiley.com/doi/full/10.1002/cphc.200400182

[10] Hatcher, Allen. Algebraic Topology. Cambridge University Press, 2018.

27

You might also like