Symmetry and Confinement

B. Caudy & JG, arXiv:0712.0999

B. Lucini & JG, in progress

Jeff Greensite
Quarks and Hadrons in Strong QCD
St. Goar, Germany
 What is Confinement?

Juliet:

"What's in a name? That

which we call a rose
By any other name would
smell as sweet."

Romeo and Juliet (II, ii, 1-2)

What are people trying to prove, in order to “prove” confinement?
And what do they mean by that word?

1 linear static quark potential, rising to infinity most order parameters

2. colorless asymptotic particle states common terminology

These are not quite the same thing, which raises some
semantic issues:
What are people trying to prove, in order to “prove” confinement?
And what do they mean by that word?

1 linear static quark potential, rising to infinity most order parameters

2. colorless asymptotic particle states common terminology

These are not quite the same thing, which raises some
semantic issues:

against #1 - in real QCD, with quarks, the static potential rises

and then levels off, due to string breaking.

so is real QCD not confining?

What are people trying to prove, in order to “prove” confinement?
And what do they mean by that word?

1 linear static quark potential, rising to infinity most order parameters

2. colorless asymptotic particle states common terminology

These are not quite the same thing, which raises some
semantic issues:

against #1 - in real QCD, with quarks, the static potential rises

and then levels off, due to string breaking.

so is real QCD not confining?

against #2 - asymptotic particle states are also colorless in a

Higgs theory, where there is no linear potential at all.

so are Higgs theories confining?

The Fradkin-Shenker-Osterwalder-Seiler Theorem

Consider an SU(2) gauge-Higgs theory with lattice action

!1 !1
S=β Tr[U U U U ] + γ
† †
Tr[φ† (x)Uµ (x)φ(x + µ
")]
2 x,µ
2
plaq

It has a phase diagram something like this: (Campos, 1997)

The theorem says that

there is no complete
separation between the
Higgs-like and the
confinement-like regions.
 More precisely: between a point

”a” deep in the confinement-like regime ( β, γ ! 1 ) , and a point

“b” deep in the Higgs regime ( β, γ ! 1) ,
there is a path from a to b such that all Green’s functions of all local,
gauge-invariant operators

!A(x1 )B(x2 )C(x3 )..."

vary analytically along the path.
b

This rules out an abrupt

transition from a colorless to
a color-charged spectrum.
a
 Creation operators for three colorless vector mesons:

! " ! "
!† ! + µ) ,
φ (x)Uµ (x)φ(x !†
Re φ (x)Uµ (x)!
ϕ(x + µ) , !†
Im φ (x)Uµ (x)!
ϕ(x + µ)

where

!
! (x) = σ2 φ(x)
ϕ

Higgs-like region: “W-bosons”

Confinement-like region: Mesons (with scalar constituents)

 So “color confinement” in the asymptotic spectrum doesn’t
distinguish between Higgs and “real QCD”-like dynamics.

Question

Can the confinement phase of a gauge theory be regarded

as the symmetric (or broken) realization of a gauge symmetry?

This idea appears in a number of popular approaches, in particular:

i) Dual superconductivity

ii) the Kugo-Ojima criterion

iii) Coulomb-gauge confinement

Naively, symmetry breaking of gauge invariance violates

Elitzur’s Theorem:
Local gauge symmetries do not break spontaneously. In the absence
of gauge fixing, !ϕ" = 0 regardless of the shape of the Higgs potential.

However, although local symmetries can’t break spontaneously, it is still

possible to break a global subgroup of the local symmetry.

Subgroups of this kind are typically what remains of the local symmetry
after a gauge choice.
Landau gauge
Remnant symmetries are homogenous gauge transformations

g(x) = g
There are also inhomogenous transformations

1
g(x) = exp[iΛa (!; x) σa ] (Hata, 1983)
2
1
Λa (!; x) = !aµ xµ − g 2 (Aµ × !µ )a + O(g2 )
∂

analogous to abelian transformations Aµ → Aµ + ∂µ φ

with
φ(x) = c + "µ xµ

〈φ〉≠0 ߭ remnant gauge symmetry breaking.

Coulomb gauge
In addition to g(x) = g there is a much larger, time-dependent
remnant symmetry

g(x, t) = g(t)
Define ! " #
T
L(x, T ) = P exp i dtA0 (x, t)
0

Then Tr[L(x, T )] = Tr[gL(x, T )g † ] for constant transformations, but

Tr[L(x, T )] != Tr[g(0)L(x, T )g † (T )]

〈Tr[L]〉 in Coulomb gauge probes the breaking of a remnant gauge

symmetry which is different from that probed by 〈φ〉 in Landau gauge.
It is insensitive to the breaking of the homogenous g(x)=g symmetry.
In principle, 〈Tr[L]〉 and 〈φ〉 could show transitions (0 → non-zero) at
different places in the space of couplings.

Do different global subgroups of the gauge group break in different

places in the phase diagram?

If they do, then there is an ambiguity is the phrase “spontaneously

broken gauge symmetry”. Precision requires specifying which
symmetry is actually broken.

With this in mind, we revisit three confinement criteria which are

based on the symmetric or broken realization of a gauge
symmetry.
I. The Kugo-Ojima Criterion

Kugo and Ojima introduce a function uab (p2 ) defined by

! "
pµ pν
u (p ) gµν
ab 2
− 2 =
p
#
d4 x eip(x−y) "0|T [Dµ ca (x)g(Aν × c)b (y)|0$

where ca(x) is the ghost field in a covariant gauge. They then show that
the expectation value of charge vanishes in any physical state

!phys |Qa |phys" = 0 (confinement?)

providing the following conditions are satisfied:

1. Remnant symmetry under g(x) = g is unbroken

2. The criterion uab (0) = −δ ab is satisfied.

It turns out that (2) implies that the spatially inhomogenous

remnant symmetry in Landau gauge is also unbroken (Hata, Kugo).

Therefore, the Kugo-Ojima scenario requires that the

entire remnant gauge symmetry in Landau gauge is
unbroken, i.e. 〈φ〉 = 0 .
II. The Coulomb Criterion Marinari, Parisi, Paciello, Taglienti (1993)
Olejnik, Zwanziger, JG (2004)

The idea is to show that

޼ The Coulomb energy of an isolated color charge is infinite;

޽ The color Coulomb potential is confining.

It turns out that both of these are implied by unbroken remnant gauge
symmetry

g(x, t) = g(t)
which means that
! " #$
Tr L(x, T ) = 0
Isolated Charge Ψaq = q a (x)Ψ0
propagation in time

G(T ) = !Ψaq |e−(H−E0 )T |Ψaq "

! " #$
∝ Tr L(x, T )

infinite energy if G(T)=0 , which implies〈Tr[L]〉 = 0

Color-Coulomb Potential

d ! "
Vcoul (R) = − lim log Tr[L(x, T )L† (y, T )]
T →0 dT

Vcoul(R) goes flat at G(R) → ∞ (no confinement) if 〈Tr[L]〉 ≠ 0



So both conditions require unbroken remnant gauge symmetry.

 Remnant Symmetry Breaking in the Gauge-Higgs Model

Landau Gauge Coulomb Gauge

order parameters

! 1 " ! 1 "
φ= φ(x) U (t) = U0 (x, t)
V x V3 x

! # Lt " $
1 1 ! 1
QL = Tr[φ"φ"† ] QC = # (t)U
Tr[U # † (t)]
2 Lt t=1 2

symmetric phase
1 1
QL ∝ QC ∝
V V3
broken phase

lim QL > 0 lim QC > 0

V →∞
V →∞
 We see a thermodynamic transition (or sharp crossover) for β > 2.0

"

" !
↑ ↑ /(001!,(-)

$%&'(&)*)&+!,(-)

.
. " #
1.2 2.2
!

Plaquette Energy, "=2.2, 164 Lattice

Plaquette Energy, "=1.2, 164 Lattice
0.64
0.55
0.63
0.5
0.62
0.45 0.61

0.6
P

0.4
P

0.59
0.35
0.58
0.3 0.57

0.25 0.56
0.5 1 1.5 2 2.5 3 0.4 0.6 0.8 1 1.2 1.4
! !
For β > 2.0 , the Landau and Coulomb gauge transitions happen at about
the same γ . For β < 2.0 , these transitions happen at different γ .

"=1.2, 144 lattice

0.6
QL
QC
0.5

0.4
Q

0.3

0.2

0.1

0
0.8 1 1.2 1.4 1.6 1.8 2 2.2 2.4
γ!

We need to pinpoint the location of the transitions...

 Plotting Q vs lattice extension L shows that indeed
1 1
QL ∝ and QC ∝
V V3
below the transition, and Q ≈ constant above....

Landau gauge, !=1.2

10

0.1
QL

0.01
"=1.0
"=1.1
"=1.2
0.001 "=1.3
"=1.35
"=1.4
"=1.45
0.0001
10
L

but a better method is to look for peaks in the susceptibilities.

 Let ! and define susceptibilities
Q = !Q"

! #
χL " 2L " − Q2L
= V 2 !Q
! #
χC " 2C " − Q2C
= V32 !Q

Landau gauge, #=1.2 Coulomb gauge, #=1.2

6
1.0*10 7.0*103
L= 6 L= 6
9.0*105 L= 8 L= 8
L=10 3
6.0*10 L=10
8.0*105 L=12 L=12
L=14 L=14
7.0*105 5.0*103 L=16
6.0*105 3
4.0*10
5.0*105
!

!
4.0*105 3.0*103
3.0*105
2.0*103
5
2.0*10
3
1.0*105 1.0*10

0.0*100 0
1.1 1.2 1.3 1.4 1.5 1.6 1.7 1.8 0.0*10
1.2 1.4 1.6 1.8 2 2.2 2.4 2.6 2.8 3
γ" γ"
The final result:

Gauge Symmetry-Breaking Transition Lines

3
Coulomb transition
Landau transition
2.5

2
higgs-like region
1.5
!

0.5 confinement-like region

0
0 0.5 1 1.5 2
"β

Coulomb and Landau remnant gauge symmetries break in

different places, and they break in the absence of any
discontinuity in the spectrum, or in the Green’s functions.

Order parameters for confinement?? Unlikely!

III. Dual Superconductivity Mandelstam and ‘t Hooft, mid-1970’s

In compact U(1) gauge theories there is a conserved magnetic current

jµM = ∂ ν F!µν
associated with a dual U(1) gauge symmetry.

Spontaneous breaking of the dual gauge symmetry leads to confinement

via a dual Meissner effect.

How to detect spontaneous breaking of a dual (global) gauge symmetry?

Pisa Proposal Di Giacomo, Paffuti, D’Elia, Lucini, del Debbio...

The order parameter for dual symmetry breaking is a monopole

creation operator, denoted μ, which doesn’t commute with magnetic
charge.
The monopole operator inserts a monopole field centered at a given point x

µ(x)|Ai ! = |Ai + AM
i !

accomplished by
! " #
µ(x) = exp i d3 y AM
i (y)Ei (y)

(In a non-abelian theory, an abelian subgroup is picked out by abelian

projection.)

In practice one computes

∂
ρ= log!µ" = !S"S − !SM "SM
∂β

A large negative peak in ρ at some β=βc , growing with lattice volume, is the
sign that !µ" = 0 , and dual superconductivity disappears, for β>βc .
 In case after case, a symmetry restoration transition
ρ → −∞ , 〈µ〉→ 0
occurs at the deconfinement temperature.

Pure SU(2), NT=4

Di Giacomo et al. (1999)

But what about the behavior of ρ near other types of transitions;

e.g. in the gauge-Higgs model, at zero temperature?
 Dual Gauge Symmetry Transitions
(Lucini & JG)

There is strong evidence of µ→0 (dual symmetry restoration)

transitions in the absence of any transition from a confining to a non-
confining phase, and even in the absence of any change of phase
whatever.

We find such µ→0 transitions, at zero temperature, in

1. SU(5) gauge theory

2. mixed fundamental-adjoint SU(2) gauge theory

3. pure SU(2) (Wilson action)

4. gauge-Higgs theory

5. G(2) gauge theory (Cossu et al.)

Example 1 - SU(5)

Pure SU(5) gauge theory is known to have a first order transition

around β = 16.3. And there, as it turns out, µ→0 . But the transition is
not deconfining.

Pure SU(5)
20
0
-20
-40
-60
-80
!

-100
-120
-140
-160
L=8
-180 L=10
L=12
-200
12 13 14 15 16 17 18 19 20
"
Example 2 - SU(2) mixed action

The mixed fundamental adjoint action is

!1 !1
S=β Tr[U (P )] + βA TrA [U (P )]
2 2

and many years ago Creutz and Bhanot found the phase structure

βA=1.5
There is a µ→0 transition along the (non-deconfining) Bhanot-Creutz
transition line

SU(2) Mixed Action with fixed "A=1.5

200

100

-100

-200
!

-300

-400
L=8
-500 L=10
L=12
L=16
-600
0.4 0.6 0.8 1 1.2 1.4 1.6 1.8
"
Example 3 - pure SU(2) Wilson action

There even appears to be a µ → 0 transition in pure (Wilson action)

SU(2) LGT, although it requires rather large volumes to see the (much
broader) peak at β=2.3.

Pure SU(2)
0

-50

-100

-150

-200
!

-250
L=8
-300 L=10
L=12
-350 L=16
L=20
L=24
-400
1.2 1.4 1.6 1.8 2 2.2 2.4 2.6 2.8
"
 Example 4 - SU(2) gauge-Higgs action

We also find µ → 0 transitions in the gauge-Higgs model, where the

Fradkin-Shenker theorem tells us that the phase diagram is connected.

gauge-Higgs model, #=1.6 gauge-Higgs model, #=2.2

0 200

0
-50
-200
-100
-400
-150 -600
!

!
-200 -800

-1000
-250
L=6 -1200
-300 L=8 L=6
L=12 -1400 L=8
L=16 L=12
-350 -1600
0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2 2.2 0.4 0.5 0.6 0.7 0.8 0.9 1 1.1 1.2
γ Ќ " γ "
Ќ
β=1.6, γ=1.3 β=2.2, γ=0.84
off the thermodynamic on the thermodynamic
transition line transition line
Example 5 - G(2) gauge theory

Apart from us: the Pisa group themselves have found a µ → 0

transition, at zero temperature, in G(2) lattice gauge theory.

Cossu, D’Elia, Di Giacomo,

Lucini, Pica (2006)
Is it possible to find some other order parameter µ’ such that

i) µ’(β) = µ(β) at large β ≥ βlarge

?
ii) µ’(β) has no transition

Answer: No! Given (i), a transition µ→0 guarantees a

transition in µ’. Here’s why:

! βlarge
log[µ(βlarge )] = lim dβ ρ(β) + log[µ(0)]
V →∞ 0

If µ(βlarge)=0 then, by (i), also µ’(βlarge)=0 .

All that can happen is that, since ρ’(β) ≠ ρ(β) , the transition may
happen at a different β < βlarge , the peak may be broader, and the
transition harder to see at small (or even fairly large) volumes.
to summarize
Global gauge symmetries associated with the

a) Kugo-Ojima
b) Coulomb confinement
c) Dual superconductor

scenarios are found to have transitions where there is no

transition to/from a confinement phase, and even where
there is no change of phase whatever.

in consequence
Global gauge symmetries do not seem to provide us with
good order parameters for confinement.
So, what’s in a name?
If “confinement” means:

color-singlet spectrum

then there is probably no meaningful

distinction between the confined and
Higgs phases, at least in terms of
symmetries

But there is a difference in physics! Flux tube formation, linear potential,

Regge trajectories....as opposed to a Yukawa potential.

If we focus on these, rather than on color neutrality, then it is better to say

that confinement is the phase of

magnetic disorder

Then there is a relevant associated symmetry.

Magnetic Disorder

means: the existence of vacuum fluctuations strong enough to induce

an area law falloff in Wilson loops at arbitrarily large scales.

The vacuum of the gauge-Higgs theory satisfies this condition only as

γ→0, and there is a symmetry which distinguishes γ=0 from γ>0 :
Magnetic Disorder

means: the existence of vacuum fluctuations strong enough to induce

an area law falloff in Wilson loops at arbitrarily large scales.

The vacuum of the gauge-Higgs theory satisfies this condition only as

γ→0, and there is a symmetry which distinguishes γ=0 from γ>0 :

Center Symmetry When center symmetry is broken, either

spontaneously
deconfinement
adjoint rep matter fields

or explicitly,
fundamental rep matter fields

or doesn’t exist
G(2) gauge group
magnetic disorder is lost.
The traditional order parameters for confinement test center symmetry:

A. finite asymptotic string tension σ > 0 (implies linear potential)

! " #
W (C) = P exp[i dxµ Aµ ] ∼ exp[−σArea(C)]
C

B. vanishing Polyakov lines (isolated charge has infinite energy)

! " #
T
P (!x) = P exp[i dt A0 (!x, t)] =0
0

C. ‘t Hooft loop (center vortex creation operator)

B(C) ∼ exp[−µPerimeter(C)]
D. center vortex free energy:

if Fv = Lz Lt exp[−σ ! Lx Ly ] then σ ≥ σ!

None of these conditions are satisfied if global center symmetry is

broken, either spontaneously or explicitly.
Question:

“If the center is so important, then what confines gluons?”

↑
spectrum sense!

Answer:

The same thing that “confines” large electric charge in QED.

In QED it is impossible to have an object of nuclear size having an
electric charge much greater than |Qc| ≈ 170.

QED vacuum
In QED it is impossible to have an object of nuclear size having an
electric charge much greater than |Qc| ≈ 170.

QED vacuum

!"!
"!
In QED it is impossible to have an object of nuclear size having an
electric charge much greater than |Qc| ≈ 170.

QED vacuum

!!"%"#
$
!""!#
$
The same process goes on for adjoint charges in non-abelian theories,
given sufficient charge separation

Yang-Mills vacuum
The same process goes on for adjoint charges in non-abelian theories,
given sufficient charge separation

Yang-Mills vacuum
The same process goes on for adjoint charges in non-abelian theories,
given sufficient charge separation

Yang-Mills vacuum
 The same process goes on for adjoint charges, with sufficient
separation

QCD vacuum

I prefer to call this “color screening”, rather than color “confinement”.

 Group Disorder and Center Disorder
K. Langfeld, S. Olejnik, H. Reinhardt, T. Tok, & J.G. (2006)

Representation dependence of the string tension:

Casimir scaling at intermediate distances;

N-ality dependence asymptotically.

The latter is due to string-breaking by gluons and/or matter fields.

No big mystery.

But this is a “particle” explanation...

What is the “field” explanation, for both Casimir and N-ality

behavior, in terms of vacuum fluctuations which dominate the
relevant functional integral?
Basic idea: In a surface slice, the vacuum is dominated by
overlapping center domains on some scale R. Fluctuations within
each domain (beyond some confinement scale rc ) are subject
only to the weak constraint that the total magnetic flux adds up to
a center element of the gauge group.

Fluctuations within a domain Group disorder, Casimir scaling

Existence of domains Center disorder, N-ality
A simple model: center domain in the plane of a Wilson loop
contributes a factor

zn ∈ ZN z0 = 1

z = Ḡr [αn ]
1 ! n #
"
= χr exp[i#
α · H]
dr

where χr is the group character, H ! the generators of the

Cartan subalgebra, and the α ! n depend on the overlap of the
domain with the interior of the loop.
Ḡr [α] is proportional to the quadratic Casimir for small α , and
goes to a center element (which may be z0 = 1 ) for enclosed
domains.

! represents the average magnetic flux in the overlap

! ·H
α
region of loop and domain.

We suppose that fluctuations in different regions of each domain

are correlated only by the constraint that they add up to center
element. If AD is the area of the domain, and A is the
area contained in the loop, then for SU(2) we get

! "2 # 2
$ % &2
A A A
α (x)
1
= const. − 2 + 2π
AD AD AD
! "2 # 2
$
A A
α0 (x) = const. − 2
AD AD
 Difference between G(2) and SU(2): G(2) has only one type of
center domain, only α0 contributes, string tension is
asymptotically zero.

For SU(2), the domain model gives results for the static potential
like these:

f=.01,g=.03,p=2 f=.01,g=.03,p=2
3.5 7
j=1/2
3 j=1/2 6 j= 1
j= 1 j=3/2
2.5 j=3/2 5

2 4

V(R)
V(R)

1.5 3

1 2

0.5 1

0 0
0 10 20 30 40 50 0 50 100 150 200
R R

Casimir scaling Color screening

(short distance) (asymptotic)
 Conclusions
If “confinement” means:

color-singlet spectrum

then there is probably no meaningful

distinction between the confined and
Higgs phases, at least in terms of global
gauge symmetries

These symmetries show transitions in the wrong places!

But, if confinement means

magnetic disorder

Then the relevant symmetry is center symmetry.

A Question
to people who compute ghost/gluon propagators on the
lattice, and who see, e.g., an infrared enhanced ghost
propagator:

what happens to κ in the gauge-Higgs coupling plane?

is there a transition at the remnant-symmetry transition?