SPIN-CHARGE SEPARATION, CONFORMAL COVARIANCE

AND THE SU(2) YANG-MILLS THEORY

Ludvig Faddeev∗
S. Petersburg Branch of Steklov Mathematical Institute,
Russian Academy of Sciences, Fontanka 27, St. Petersburg, Russia
arXiv:hep-th/0608111v2 17 Aug 2006

Antti J. Niemi†
Department of Theoretical Physics, Uppsala University,
P.O. Box 803, S-75108, Uppsala, Sweden
Laboratoire de Mathematiques et Physique Theorique CNRS UMR 6083,
Université de Tours, Parc de Grandmont, F37200, Tours, France and
Chern Institute of Mathematics, Tianjin 300071, P.R. China
(Dated: February 2, 2008)

0
 Abstract
In the low energy domain of four-dimensional SU(2) Yang-Mills theory the spin and the
charge of the gauge field can become separated from each other. The ensuing field variables
describe the interacting dynamics between a version of the O(3) nonlinear σ-model and a
nonlinear Grassmannian σ-model, both of which may support closed knotted strings as stable
solitons. Lorentz transformations act projectively in the O(3) model which breaks global in-
ternal rotation symmetry and removes massless Goldstone bosons from the particle spectrum.
The entire Yang-Mills Lagrangian can be recast into a generally covariant form with a confor-
mally flat metric tensor. The result contains the Einstein-Hilbert Lagrangian together with
a nonvanishing cosmological constant, and insinuates the presence of a novel dimensionfull
parameter in the Yang-Mills theory.

∗
Electronic address: [email protected]
†
Electronic address: [email protected]; URL: http://www.teorfys.uu.se/people/antti

1
I. INTRODUCTION

Apparently the necessity of a mass gap in a pure Yang-Mills theory and the nature
of its particle spectrum were originally posed as problems by Wolfgang Pauli during a
1954 Princeton seminar by C.N. Yang [1]. However, despite over 50 years of effords
the physical particle content of a pure, interacting four dimensional Yang-Mills theory
remains a mystery. In particular, the explanation of color confinement is a theoretical
challenge [2]. Dimensional transmutation with its intimate relationship to the high
energy asymptotic freedom does open a door for a dimensionfull parameter to enter.
But since the high energy limit of the Yang-Mills theory describes asymptotically free,
massless gauge particles the precise relationship between this dimensionfull parameter
and the mass gap which secures confinement remains unclear.
During the last ten years [3]-[5] we have investigated the possibility that the low en-
ergy spectrum of pure Yang-Mills theory could comprise of closed and knotted strings
as stable solitons. This proposal is very natural from the point of view of QCD phe-
nomenology. If quarks are indeed confined by stringlike collective excitations of the gauge
field, in the absence of quarks these strings should close on themselves into stable and
in general knotted and linked configurations. In [5] we proposed how the properties of
such closed stringlike solitons could be related to the spectrum of the Yang-Mills theory.
In particular, we suggested that in the case of a SU(2) gauge theory the effective low
energy Lagrangian should relate to the following version of the O(3) nonlinear σ-model,
originally proposed by one of us [6]

m2 1
Lef f = (∂a s)2 + (s · ∂a s × ∂b s)2 + V (s) (1)
2 4
Here s is a three-component unit vector, and m is a parameter with the dimensions of
mass. The last term is a potential term. It breaks the global O(3) symmetry which is
present in the first two terms, removing the two massless Goldstone bosons from the
spectrum. Both numerical simulations [7] and formal mathematical arguments [8] have
confirmed our proposal [9] that (1) does indeed support closed knotted strings as stable

2
solitons.
Several approaches have been suggested, how to derive (1) directly from the Yang-
Mills theory. These derivations are commonly based on the following, incomplete Lie-
algebra expansion of the SU(2) gauge field [3]

Aa = Ca s + ∂a s × s + ρ∂a s + σ∂a s × s (2)

where ρ and σ are real scalars. The first two terms in (2) were originally introduced by
Duan and Ge [10], and subsequently by Cho [11], to describe the properties of Wu-Yang
monopoles. Indeed, this is a very natural decomposition of the gauge field in terms of
the SU(2) Lie algebra. When (2) is substituted to the Yang-Mills Lagrangian, one finds
that the structure (1) emerges after one-loop radiative corrections are taken into account
[12].
The apparent lack of off-shell completeness in (2) prompted us in [4] to propose
an alternative decomposition of the gauge field. This new decomposition is off-shell
complete, and (2) assumes a role in ensuring its manifest gauge covariance. Our new
decomposition leads to many unexpected consequences, some of which we shall reveal
in the present article. In particular, it admits a very intriguing physical interpretation:
The decomposition [4] is intimately related to the slave-boson decomposition [13] that
has been introduced independently in the context of strongly correlated electron systems
as an alternative to Cooper pairing and Higgs effect [14]. As a consequence our new
decomposition suggests that the separation between the spin and the charge could be
a general phenomenon that can be exhibited by a large variety of quantum fields [15],
[16].
A common feature of these spin-charge decompositions is that they all seem to in-
volve a real-valued scalar field. A nonvanishing ground state expectation value for this
scalar field describes a condensate. This corresponds to the density of the material en-
vironment, and the presence of a nontrivial condensate is a necessary condition for the
spin-charge decomposition to occur. The condensate can also yield an alternative to the
conventional Higgs effect [14].

3
 Here we find that in the case of SU(2) Yang-Mills theory the real-valued scalar field
admits another interpretation. It is the conformal scale of a metric tensor that describes
a conformally flat space-time. In terms of the spin-charge separated variables, the Yang-
Mills Lagrangian then contains both the Einstein-Hilbert Lagrangian and a cosmological
constant. Since conformal flatness is equivalent to the vanishing of the traceless Weyl
(conformal) tensor Wµνρσ , the gravitational contribution to the Yang-Mills Lagrangian
can be interpreted as the γ → ∞ limit of the higher derivative gravitational Lagrangian

1√ √ 2
LEH = g R − Λ · g + γ · Wµνρσ (3)
κ
It has been shown [17] that the Lagrangian (3) is part of a renormalizable higher deriva-
tive quantum theory of gravity. In particular, the one-loop β-function for γ does indeed
send this coupling to infinity in the short distance limit. This enforces asymptotically
the condition
Wµνρσ ∼ 0 (4)

Hence at short distances the space-time becomes asymptotically (locally) conformally

flat. We also note that the presence of the higher derivative Weyl tensor contribution
in (3) gives rise to a linearly increasing component in the large distance interactions.
Furthermore, our reformulation of the Yang-Mills theory in a manner that contains
the Einstein-Hilbert and cosmological constant terms has the additional peculiar conse-
quence that a novel dimensionfull parameter enters the Yang-Mills Lagrangian.

We start the present article by defining our notations. We then proceed to describe
the new decomposition [4] of the four dimensional SU(2) gauge field. This decomposition
singles out the Cartan direction of the SU(2) Lie algebra. This leaves us with a geometric
structure that involves the Grassmannian manifold G(4, 2) of two dimensional planes
that are embedded in a four dimensional space. The Grassmannian geometry guides our
subsequent decomposition of the gauge field into its spin and charge constituents. In
particular, it leads us to a O(3) symmetric order parameter, a three component internal
space unit vector field n akin the vector field s in (1).

4
 The Grassmannian structure also introduces an internal, compact U(1) gauge sym-
metry. In the context of lattice gauge theories [18] a compact U(1) gauge theory is
known to display a first-order phase transition between a strong coupling phase and a
weak coupling phase. The strong coupling phase exhibits confinement, which is absent
in the weak coupling phase.
It is quite natural to expect that in an abelian theory the gauge coupling decreases
when the distance increases. Thus the presence of the internal compact U(1) gauge
structure may explain why in short distance Yang-Mills theory the spin and the charge
can become strongly confined into asymptotically free and pointlike gauge particles, and
why the decomposition into independent spin and charge carriers can only occur in the
weakly coupled long-distance domain of the compact U(1) theory.
The internal U(1) gauge structure leads to a projective realization of the Lorentz
transformations in the O(3) unit vector n. The ensuing one-cocycle breaks the global
O(3) symmetry which is displayed by the first two terms in the corresponding Lagrangian
(1), even in the absence of an explicit symmetry breaking term such as the third term
in (1): Since the ground state of the theory can not violate Lorentz invariance, the
ground state direction of the vector n in the internal space becomes uniquely fixed. As
a consequence the requirement that the ground state is Lorentz invariant, removes the
two massless Goldstone bosons that are otherwise associated with the breaking of the
global O(3) symmetry in the dynamics of n.
We then proceed to inspect the detailed structure of the Yang-Mills Lagrangian in
terms of the separate spin and charge variables. In particular, we explain how the
Lagrangian (1) for the internal vector n is embedded in the tree-level Yang-Mills La-
grangian, even before any radiative corrections are taken into account. Since (1) supports
closed knotted strings as stable solitons [9], [7], [8], this endorses our proposal that knot-
ted and closed stringlike solitons are indeed the natural candidates for describing the
interacting spectrum of the pure SU(2) Yang-Mills theory [5].
It turns out that the functional form (1) relates both to the spin and the charge degree

5
of freedom. This suggests the presence of some kind of dual structure between the spin
and charge variables in the Yang-Mills theory. Moreover, since the spin conventionally
relates to magnetism while the charge relates to electricity, this could be the sought-after
electric-magnetic duality of the Yang-Mills theory.
We then argue that our decomposition is independent of the way how we choose
the Cartan direction in the SU(2) Lie algebra. The gauge covariance becomes manifest
when we introduce the structure of (2) in our decomposed gauge field.
We show that in terms of the spin and the charge variables the entire Yang-Mills
Lagrangian admits a manifestly generally covariant form with a conformally flat metric
tensor. Thus the SU(2) Yang-Mills theory describes the interactions between (1) and
the G(4, 2) Grassmannian nonlinear σ-model in a conformally flat space-time. Both the
Einstein-Hilbert Lagrangian and a cosmological constant term are present. This also
introduces a dimensionfull parameter in the Yang-Mills theory, that becomes visible
only when it is realized in terms of the spin-charge separated variables.
Our result suggests the tantalizing possibility that long distance Einstein gravity
metamorphoses into a renormalizable Yang-Mills theory at short distances.
Finally, we analyze the finite energy content of the spin-charge separated, static
Yang-Mills theory using a Hamiltonian formulation. We find that closed knotted strings
can also be supported by the G(4, 2) nonlinear σ-model, in a manner that involves
the structure of (1). Furthermore, we propose that for finite energy configurations the
space-time R4 becomes compactified into S3 × R1 .
We conclude by presenting some physical interpretations of our results, together with
suggestions on the directions for future research.

6
II. SOME NOTATION

We consider a SU(2) Yang-Mills gauge field A = Aia σ i dxa in R4 (a, b, ... = 1, 2, 3, 4)

with Euclidean signature. The Lie algebra generators coincide with the standard Pauli
matrices σ i (i, j, ... = 1, 2, 3) and we employ the complex combinations
1
σ ± = (σ 1 ± iσ 2 ) (5)
2
For the gauge field this gives

A = Aia σ i dxa = Aa σ 3 dxa + Xa+ σ − dxa + Xa− σ + dxa (6)

where
Xa± = A1a ± iA2a

The finite gauge transformation is

A → gAg −1 + 2igdg −1 (7)

Note that in our notation the Yang-Mills coupling constant appears as a factor in front
of the Lagrangian.
For an infinitesimal group element
i i
g = exp{ ǫ · σ} ≈ 1 + ǫ̂ + O(ǫ2 )
2 2
the gauge transformation takes the form

δǫ̂ Aia = δ ij ∂a ǫj + εijk Aka ǫk

When the gauge transformation is in the direction of the Cartan subgroup UC (1) ∈
SU(2)
i 3
g ∼ h = e 2 ωσ ∈ UC (1) (8)

the component A3a ∼ Aa transforms as a UC (1) gauge field

δh Aa = ∂a ω (9)

7
For the off-diagonal Xa± we get

δh Xa± = e∓iω Xa± (10)

Consequently, when we only consider gauge transformations in the direction of the Car-
tan subgroup, we can interpret the full SU(2) gauge field as a charged UC (1) vector
multiplet
Aia → (Aa , Xa± )

Eventually we shall argue that even though we here introduce a particular (global)
identification of the Cartan UC (1) in terms of the Pauli matrices, our results are in-
dependent of this particular choice. However, for the clarity of presentation we shall
momentarily proceed with this choice of the Cartan direction in the SU(2) Lie-algebra.
Finally, the Yang-Mills field strength tensor is

i
Fab = ∂a Aib − ∂b Aia + ǫijk Aja Akb

In terms of the charged UC (1) vector multiplet it decomposes according to

3 i
Fab = ∂a Ab − ∂b Aa + (Xa+ Xb− − Xb+ Xa− ) = Fab + Pab (11)
2

± 1 2
Fab = Fab ± iFab = (∂a ± iAa )Xb± − (∂b ± iAb )Xa± = D± ± ± ±
A a Xb − D A b Xa (12)

Here the first term Fab in (11) is the UC (1) (Cartan) field strength tensor, and (12)
involves the UC (1) covariant derivatives D± ±
A b of the charged vector fields Xa . These

terms are clearly consistent with the UC (1) multiplet structure. The sole term that lacks
an obvious physical interpretation in terms of the UC (1) multiplet, is the antisymmetric
tensor Pab in (11). We now proceed to interpret it geometrically.

8
III. GRASSMANNIANS AND SPIN-CHARGE SEPARATION

We shall now interpret the antisymmetric tensor Pab in (11). Explicitely we have

i
Pab = (Xa+ Xb− − Xb+ Xa− ) = A1a A2b − A1b A2a (13)
2

This is a real antisymmetric 4 × 4 matrix that obeys the quadratic relation

P12 P34 − P13 P24 + P23 P14 = 0 (14)

In fact, using simple linear algebra one can show that any real 4 × 4 antisymmetric ma-
trix Pab that is subject to the condition (14) can always be represented in the functional
form (13) in terms of some two vectors A1a and A2b . In projective geometry the relation
(14) is known as the Klein quadric. It describes the embedding of the real Grassman-
nian G(4, 2) in the five dimensional projective space RP5 as a degree four hypersurface.
This Grassmannian is the four dimensional manifold of two dimensional planes that are
embedded in R4 , and it can be identified with the homogeneous space

SO(4)
G(4, 2) ≃
SO(2) × SO(2)

It is convenient to describe the two dimensional planes in R4 in terms of a zweibein.

For this we introduce an orthonormal doublet eαa (α = 1, 2)

eαa eβa = δ αβ

that spans a generic two dimensional plane in R4 . We can then represent the off-diagonal
components of the gauge field as

Aαa = M α β eβa

where M α β is a 2 × 2 matrix. In terms of the complex combination

1
ea = √ (e1a + ie2a )
2

9
we have
ea ea = 0
(15)
ea ēa = 1
and we can write
Xa+ = A1a ± iA2a = ψ1 ea + ψ2 ēa (16)

Here the ψα are two arbitrary complex functions, they are linear combinations of the
matrix elements of M α β .
When we substitute (16) into (13) we get

i i 2
Pab = (|ψ1 |2 − |ψ2 |2 ) · (ea ēb − eb ēa ) = · ρ · t3 · (ea ēb − eb ēa ) (17)
2 2

We have here introduced the three component unit vector

 
  cos φ · sin θ
1  ∗ ∗  ψ1   
t = 2 ψ1 ψ2 σ =  sin φ · sin θ  (18)
 
ρ ψ2  
cos θ

where we employ the following angular parametrization

ψ1 = ρ eiξ cos θ2 e−iφ/2

(19)
ψ2 = ρ eiξ sin θ2 eiφ/2

The representation (17) has a motivation in terms of the properties of Pab . For this
we consider the action of the general linear group GL(2, R) on the coordinates A1a and
A2a that describe our generic two-plane
    
1
A G α β A1
 a  −→    a
A2a γ δ A2a

Here  
α β
G =  
γ δ

10
is the matrix realization of the G ∈ GL(2, R) on the two-dimensional plane in R4 . This
gives
G
Pab −→ (αδ − βγ)Pab = det G · Pab

Thus Pab supports a one-dimensional representation of GL(2, R), where the orbit is
parametrized by the prefactor in (17)

|ψ1 |2 − |ψ2 |2 = ρ2 · t3

and the G-action corresponds to a scaling of the density ρ according to

G
√
ρ −→ det G · ρ (20)

Clearly, the volume and orientation preserving subgroup SL(2, R) ∈ GL(2, R) is an

invariance group of Pab . The maximal compact subgroup UI (1) ∈ SL(2, R) is the internal
invariance group of the decomposition (16). It acts on the fields according to

ea → e−iλ ea
ψ1 → eiλ ψ1 (21)
ψ2 → e−iλ ψ2

The vector t in (18) is invariant under the external UC (1) gauge transformation. The
component t3 is in addition invariant under the internal UI (1) gauge transformation. But
the remaining two components transform nontrivially under the internal UI (1). With

t± = t1 ± it2

the UI (1) transformation sends

1
t± = (t1 ± it2 ) → e∓2iλ t± (22)
2

where the factor of 2 reflects the fact that t is a bilinear in ψα .

A priori, the r.h.s. of the decomposition (16) involves a total of nine independent
field degrees of freedom. These are the four real components of the functions ψα and

11
the five real components of the (complex) normalized vector ea . But due to the internal
UI (1) symmetry both sides of the decomposition (16) describe an equal number of eight
independent field degrees of freedom. This coincides with the number of independent
components in the off-diagonal gauge field and confirms that (16) yields a full, complete
field decomposition of the off-diagonal components Xµa of the gauge field.
According to (21) the complex scalar fields ψ1 and ψ2 are oppositely charged with
respect to the internal UI (1) gauge group. The vector field ea is also charged w.r.t. the
internal UI (1) group while Aa remains obviously intact under a UI (1) transformation.
The obvious choice of a UI (1) connection is the composite vector field

Ca = i ēb ∂a eb (23)

as it transforms according to
Ca → Ca + ∂a λ

under the internal UI (1) transformation (21).

We note that in terms of the explicit parametrization (19), the UI (1) transformation
(21) sends
φ → φ − 2λ

We also note that the connection (23) admits a geometric interpretation as a spin
connection that parallel transports the zweibein (e1a , e2a ).
According to (8), the Cartan subgroup UC (1) of the SU(2) gauge group acts on the
complex coefficients as follows,
h
ψ1,2 −→ e−iω ψ1,2

or in terms of the parametrization (19)

ξ →ξ−ω

while ea remains intact. Consequently we can interpret the SU(2) gauge field as the
UC (1) multiplet
Aia ∼ (Aa , ψ1 , ψ2 , ea )

12
where Aa is the UC (1) gauge field, the complex scalar fields ψα are equally charged w.r.t.
the UC (1), and the complex vector field ea is UC (1) neutral.
We can also interpret the gauge field as the UI (1) multiplet

Aia ∼ (Ca , ψ1 , ψ2 , Aa )

where Ca is the UI (1) gauge field, the complex scalar fields are oppositely charged w.r.t.
the UI (1) and the vector field Aa is UI (1) neutral.
Notice that the ψα are scalars and ea is a vector under SO(4) rotations (a.k.a. Lorentz
transformations) in R4 . Since ea is neutral under the UC (1) ∈ SU(2) gauge group
while the ψα transform nontrivially, we conclude that the decomposition (16) entails a
separation between the spin and the charge in the off-diagonal components Xa± of the
SU(2) gauge field. The spinless scalar fields ψα describe the UC (1) charge degrees of
freedom of the Xa± , and the UC (1) neutral vector field ea describes their spin degree of
freedom.
The present separation between the spin and the charge degrees of freedom in Xa± is
quite analogous to the slave-boson decomposition of an (nonrelativistic) electron, widely
employed in attempts to explain high-temperature superconductivity [14]. There, the
spin-charge separation is introduced as an alternative to the BCS superconductivity.
Instead of the Higgs effect in terms of the Cooper pairs, superconductivity emerges
when the analog of our variable ρ forms a condensate,

< vac| ρ |vac >= ∆ 6= 0 (24)

13
IV. ELECTRIC AND MAGNETIC COMPONENTS

Consider the tensor part of Pab in (17),

i
Hab = (ea ēb − eb ēa ) (25)
2
We define its “electric” (pi ) and “magnetic” (qi ) components in the usual manner, by
setting
pi = H0i = 2i (e0 e∗i − ei e∗0 )
(26)
qi = 21 ǫijk Hjk = 2i ǫijk ej e∗k
These two vectors are subject to the orthogonality relations

p·q =0
(27)
1
p·p+q·q = 4

Together with the Poynting vector

s = p×q

we then have an orthogonal triplet in R3 .

The relation (26) can also be inverted, the result is
 
e0
  √ 
e 
 1 2|p|
e =   = eiη   ≡ eiη ê (28)
 e2  2s+ip
√
  2|p|

e3

Here η is an overall phase of e. This phase is invisible to p and q since it does not
contribute to the bilinear combination (25). But this phase is subject to the internal
UI (1) gauge transformation (21) which sends

η →η−λ (29)

Thus the phase transformation (29) determines a rotation between the real and imagi-
nary components e1a and e2a of the complex vector ea in R4 .

14
 Now consider the action of SO(4) rotations (a.k.a. Lorentz transformations) on ea .
We are particularly interested in the effect of an infinitesimal SO(4) (Lorentz) boost in
a generic spatial direction εi. As a component of a four-vector, the ea should transform
in the following manner
Λε e0 = −εi ei
(30)
Λε ei = −εi e0
This is clearly a SO(4) rotation, it preserves the orthonormality relations (15).
On the other hand, we expect that when we realize this boost transformation on the
electric and magnetic components (26) we should get the familiar results

δε p = q × ε
(31)
δε q = p × ε

Curiously, we find that when we substitute this in the explicit realization (28) there is a
difference between (30) and (31): If we compare the action of Λε in (30) with the action
of δε on ea which is defined using (31) and (28), the two descriptions of the boost differ
from each other by the (infinitesimal) phase

(Λε − δε ) ea = −iΘ(p, q; ε) · ea

As a consequence the difference between the two Lorentz boosts is a (infinitesimal) shift
in the angle η in (28) according to

η → η − Θ(p, q; ε)

This is an infinitesimal internal UI (1) gauge rotation (29).

Explicitely, we have

p·ε ∂
Θ(p, q; ε) = 2
= εi ln |p| (32)
|p| ∂pi

One can verify that this quantity obeys the one-cocycle condition

δε1 Θ(p, q; ε2 ) − δε2 Θ(p, q; ε1 ) = 0

15
and as a consequence (32) is a one-cocycle.
The result means that the action of the boost δε on the vector field ea determines
a projective representation of the (Euclidean) Lorentz group. In particular, the vector
field êa in (28)
ê = e−iη e

transforms under the projective representation according to

δε ê = Λε ê + iΘ ê

We conclude that the phase η in (28) is a non-trivial field degree of freedom. If we set
η = 0 in (28), a spatial boost will generate a nontrivial η determined by the one-cocycle
(32). Moreover, since the cocycle depends only on the electric component of (25) we
propose that η can be viewed as a phase (angular) variable for electric circulation.
Consider the internal UI (1) connection (23). It admits the following explicit realiza-
tion

ˆ · ∂a ê − ∂a η = 2|q| (k × l · ∂a k) − ∂a η = Ĉ a − ∂a η
Ca = i ē · ∂a e = i ē (33)

Here k and l are two mutually orthogonal unit vectors in the electric and magnetic
directions respectively,
1
p = |p|k = √
2 2
cos ϑ · k
(34)
1
q = |q|l = √
2 2
sin ϑ · l
Since both k and l remain intact under both the external UC (1) and the internal UI (1)
gauge transformations we conclude that the vector field

2 p · ∂a s
ˆ · ∂a ê = 2|q| (k × l · ∂a k) =
Ĉ a = Ca + ∂a η = i ē (35)
p2

is gauge invariant both under the external UC (1) and under the internal UI (1) gauge
transformations; The internal UI (1) acts only on the phase variable η in (33) according
to (29).

16
 If we introduce the normalization
1
ωa = − Ĉa = −2k × l · ∂a k (36)
2|q|
we arrive at
∂a ωb − ∂b ωa = l · ∂a l × ∂b l (37)

This is the pull-back of the volume two-form on S2 that also appears in the second
term of (1). As a consequence (36) admits the geometric interpretation as the Kirillov
one-form for the co-adjoint orbit S2 = SU(2)/U(1), in the magnetic direction of l.
The result (37) follows directly from the formal properties of the unit vectors k and
l. Alternatively, it can be verified by using an explicit angular representation of these
vectors. If we denote by m the unit vector in the direction of the Poynting vector,

m = −k × l

we can represent the orthonormal triplet (k, l, m) as follows,

k = cos γ · ux + sin γ · uy
l = uz (38)
m = − sin γ · ux + cos γ · uy

where
     
cos α cos β − sin α cos α sin β
     
ux =  sin α cos β  & uy =  cos α  & uz =  sin α sin β  (39)
     
     
− sin β 0 cos β

We then have
1 1
AM = − ω = − (cos β · dα + dγ) (40)
4 2
and
1 1
FM = − dω = − sin β · dα ∧ dβ (41)
4 2
In (40) we recognize the connection one-form AM and in (41) the curvature two-form
FM (magnetic field) of the Dirac magnetic monopole.

17
 For (35) we get from (34)
√
Ĉ = − 2 sin ϑ · [cos β · dα + dγ] (42)

This vanishes when we go to the purely electric limit ϑ = 0, and coincides with (twice)
the Dirac monopole connection (40) when ϑ = π/4 and the strength of the electric and
magnetic fields are equal.

18
V. A HIGGS EFFECT

The three components of the unit vector t that we have introduced in (18) are bilinear
in the complex functions ψ1 and ψ2 . Since these functions are SO(4) a.k.a. Lorentz
scalars, the unit vector t is also a Lorentz scalar. But its components t± are not invariant
under the internal UI (1) gauge transformations.
If instead we introduce a new three component unit vector n such that

n± = e2iη t±
n3 = t3

then this new unit vector is invariant under both the external UC (1) and the internal
UI (1) gauge transformations. Explicitely, in terms of the angular variables in (18) we
have  
cos(φ + 2η) · sin θ
 
n =  sin(φ + 2η) · sin θ  (43)
 
 
cos θ
Here
φ + 2η

is a UC (1) × UI (1) invariant combination of the UI (1) dependent variables φ and η.

In the sequel we shall propose that êa in (28), Ĉa in (35) and n are the obvious
UC (1) × UI (1) gauge invariant variables for describing the Yang-Mills theory. For this
we note that the relation (35) between Ca and Ĉa is a version of the familiar (linear)
Higgs relation between the U(1) gauge vector field and the gauge invariant (massive)
vector field in the standard abelian Higgs model:
In the case of conventional Higgs mechanism, a U(1) gauge field combines with the
gradient of the phase of a complex scalar field into a gauge invariant vector field. The
modulus of the complex scalar field remains as an additional, independent gauge invari-
ant field variable. When this modulus develops a nonvanishing ground state expectation
value the gauge invariant vector field becomes massive.

19
 Here, the UI (1) gauge field Ca combines similarly with the phase η of the complex
vector field ea into the gauge invariant vector field Ĉa . This leaves the vector field êa in
(28) as an additional independent and UI (1) invariant field variable. Furthermore, the
transition from t to the UI (1) invariant n can be viewed as a nonlinear version of the
Higgs mechanism. Note that all of these three UI (1) invariant field variables are also
invariant under the UC (1) gauge transformations.
We remind that due to the presence of the one-cocycle (32) the vector fields Ĉa and
êa are not SO(4) vectors. Nor are the ±-components of n scalars under SO(4). Instead,
all of these UI (1) gauge invariant quantities transform under a projective representation
of the spatial SO(4) (a.k.a. Lorentz) group.
The breaking of the Lorentz invariance by the one-cocycle has important physical
consequences. For this we note that the two components A1a and A2a of the SU(2) gauge
field appear symmetrically in the Yang-Mills Lagrangian, they can be exchanged by a
global gauge transformation. Consequently we can expect that in terms of the spin-
charge separated variables the Lagrangian should display a similar global symmetry
between the two complex scalar fields ψ1 and ψ2 . This symmetry should translate into
a global O(3) rotation invariance when represented in terms of the unit vector n in
(43); See the first two terms in (1). But the presence of such a global O(3) symmetry
poses a problem. When we select a ground state direction of n we break the global
O(3) invariance explicitly, in a manner which in general leads to two a priori massless
Goldstone bosons.
The violation of Lorentz invariance by the one-cocycle appears to remove the Gold-
stone bosons: Since we have no reason to expect that the ground state of the theory
violates Lorentz invariance and since n3 is the sole Lorentz invariant component of n,
the only possible Lorentz invariant ground state direction for n is
 
0
 
n → ± 0 ≡ ẑ
 
 
1

20
Any nonvanishing ground state value for the components n± breaks Lorentz invariance
due to the presence of the one-cocycle.
In particular, we conclude that at large spatial distances the unit vector n should
become asymptotically parallel with the z-axis,

|x|→∞
n −→ ±ẑ (44)

Alternatively, in terms of the vector t the only UI (1) gauge invariant asymptotic ground
state direction is
t → ±ẑ

since any other asymptotic direction violates the internal UI (1) gauge invariance.

21
VI. YANG-MILLS IN GAUGE INVARIANT SPIN-CHARGE VARIABLES

We now proceed to inspect how the separation between the spin and the charge
manifests itself in the Yang-Mills Lagrangian. We start by recalling the tree-level gauge
fixed Euclidean space Yang-Mills Lagrangian

1 i 2 ξ + +2
LY M = (F ) + |DA a Xa | + Lghost (45)
4 ab 2
Note that we have here introduced a gauge fixing term only for the off-diagonal com-
ponents Xa± of the gauge field. For reasons that will eventually become transparent,
we do not introduce any gauge fixing term in the direction of the abelian subgroup
UC (1) ∈ SU(2). The last term Lghost denotes the ghost contribution. In the sequel its
explicit form will not be of importance to us. We only need to observe that it is entirely
independent of the gauge fixing parameter ξ [19], [20].
In our approach we do not introduce decomposed variables in the path integral. That
would only lead to unnecessary complications. Instead we propose that the appropriate
stage to implement the spin-charge separation is at the level of the effective Yang-Mills
action which has been computed in the covariant background field formalism. This
effective action accounts for all quantum fluctuations in the gauge field. But since its
explicit form is not available beyond a few leading terms in a loop expansion, we need
to resort to an indirect analysis.
By general arguments of gauge invariance we can expect that the full effective action
i
is a functional of the background field strength tensor Fab and its background covariant
derivatives. In the low momentum infrared limit we can ignore the derivative contribu-
tions, hence in this limit the effective action involves only the field strength tensor. Since
the full result is unknown to us, for simplicity we proceed by considering the infrared
limit only in its lowest order. This limit coincides with the classical Lagrangian (45), but
excluding the ghost contribution. Consequently our starting point will be the classical
Yang-Mills Lagrangian (45). Indeed, the classical Lagrangian should be an important
ingredient of the full quantum action. We now proceed to subject it to the separation

22
between the spin and the charge.
When we introduce (11) and (12) we find for the classical Yang-Mills Lagrangian

1 i 2 ξ
LY M = (Fab ) + |D+ + 2
A a Xa |
4 2

1 1 3 3 4 ξ−1 + + 2
= (Fab + 2ρ2 n3 Hab )2 + |D+ + 2 2 4
A a Xb | + (1 − n3 )ρ − ρ + |DA a Xa | (46)
4 2 8 8 2

The reason why we present the third and fourth terms in (46) in this particular manner
becomes evident as we proceed.
Note that we have here overlooked a surface contribution that originates from the
difference
D+ + − − + + − −
A a Xb D A b X a − D A a X a D A b Xb

Explicitely, the surface contribution is

 
1 + − − − + +
 1 + −
∂a X D X + Xa DA b Xb − ∂b (Xa Xb ) (47)
2 a Ab b 2

We first observe that in (46) there are two particularly interesting values for the gauge
fixing parameter ξ. These are the value ξ = 1, and the limit ξ → ∞.
If we select ξ = 1 the last term in (46) becomes absent. In particular, when ξ = 1
there are no terms present in the Lagrangian barring the surface term, where the Lorentz
index in the off-diagonal components Xa± becomes contracted with the Lorentz indices
that are carried by the other quantities, such as the derivative operator ∂a . For ξ = 1
the Lorentz indices in Xa± are only contracted internally between different contributions
of the Xa± .
Since the ghost Lagrangian is independent of ξ, by arguments of gauge invariance we
expect that this property persists to all orders of perturbation theory. In particular we
expect that in the full ξ = 1 quantum effective action the Lorentz indices carried by the
(background) fields Xa± are only contracted internally, between different contributions
of Xa± . This indicates that we can (crudely) analyze the feasibility of the spin-charge

23
separation by considering the Lagrangian (46), with ξ = 1 and ignoring the ghost con-
tributions. As a consequence we limit our interest to only the following four terms,

1 1 3 3 4
LY M = (Fab + 2ρ2 n3 Hab )2 + |D+ + 2 2 4
A a Xb | + (1 − n3 )ρ − ρ (48)
4 2 8 8
Similar conclusions can be drawn in the gauge that emerges when we send ξ → ∞.
In this limit we obtain in addition the maximal abelian gauge (MAG) condition

(∂a ± iAa )Xa± = D± ±

A a Xa = 0 (49)

It is known [19], [21] that this gauge condition is also the (Euler-Lagrange) variational
equation that describes the gauge orbit extrema of the following quantity,
Z
R = d4 x Xa+ Xa− (50)

In particular,
ρ2 = |ψ1 |2 + |ψ2 |2 = Xa+ Xa−

where ρ is the density that we have introduced in (19). Notice that with (49) the first
two terms in the surface contribution (47) vanish.
The variable ρ has an interpretation as a condensate; see (24). In the context of the
maximal abelian gauge this interpretation of ρ has been discussed widely in the literature
[19], [21]: The extrema values of ρ on the gauge orbit are obviously gauge invariant and
according to (50) correspond to gauge field configurations that are subject to the MAG
gauge condition (49) and we refer to [19], [21] for further discussion.
From (20) we also conclude that selecting the extrema value of ρ breaks the Grass-
mannian GL(2, R) into SL(2, R).
We now proceed to analyze the Lagrangian (48). In the present section our goal will
be to represent it in terms of the UC (1) × UI (1) gauge invariant variables. We shall
find that this can be achieved by a change of variables, with no need to any additional
explicit gauge fixing. This will justify a posteriori why in (45) we have introduced the
gauge fixing term only for the off-diagonal Xa± .

24
 We first consider the second term in (48). Using (18) we can write this term as
1 2 1 2
|D+ + 2 C 2 C 2 2 C 2 C 2 C 2
A a Xb | = |DA a ψ1 | + |DA a ψ2 | + ρ |DA a eb | + ρ t+ (D̄A a ēb ) + ρ t− (DA a eb ) (51)
2 2
Here DC
A a is the following UC (1) × UI (1) covariant derivative,

DC
A a ψ1 = (∂a + iAa − iCa )ψ1
DC
A a ψ2 = (∂a + iAa + iCa )ψ2
DC
A a eb = (∂a + iCa )eb

Note that even though the t± are not invariant under the internal UI (1) gauge transfor-
mations (22), since DC
A a eb transforms according to

DC
A a eb → e
−iλ C
DA a eb

the Lagrangian (51) is gauge invariant under both UI (1) and UC (1) gauge transforma-
tions.
We introduce the UC (1) × UI (1) invariant supercurrent [4], [22]
i
Ja = {ψ ∗ DC ψ1 − ψ1 D̄C ∗ ∗ C C ∗
A a ψ1 + ψ2 DA a ψ2 − ψ2 D̄A a ψ2 }
2ρ2 1 A a
From this we can solve for Aa in favor of Ja . The result is
i ←→ ←→
Aa = −Ja + 2
{ψ1∗ ∂a ψ1 + ψ2∗ ∂a ψ2 } + n3 · Ca
2ρ
When we substitute this in (51) we get for the first two terms
1 2 Ĉ 2
|DC 2 C 2 2 2 2
A a ψ1 | + |DA a ψ2 | = (∂µ ρ) + ρ (D a n) + ρ Ja (52)
4
Here we have defined the covariant derivative operator

(DĈa )ij = δ ij ∂a + 2ǫij3 Ĉa (i, j = 1, 2, 3) (53)

Note in particular that the middle term in the r.h.s. of (52) is Lorentz invariant even
though both the components n± and the connection Ĉa transform according to a pro-
jective representation of SO(4). The covariant derivative (53) compensates for the lack
of SO(4) invariance (a.k.a. Lorentz invariance) in the i = 1, 2 components of n.

25
 With (52) we have achieved our goal, in the sense that the r.h.s. of (52) involves only
quantities which are UC (1) × UI (1) invariant.
We now proceed to the third term in (51). For this we get

ρ2
ρ2 |(∂a + iCa )e|2 = ρ2 |(∂a + iĈa )ê|2 = {(∂a p)2 + (∂a q)2 }
2

ρ2  2
= cos ϑ · (∂a k)2 + sin2 ϑ · (∂a l)2 + (∂a ϑ)2 (54)
16
Clearly, this involves only manifestly UC (1) × UI (1) invariant quantities.
We observe that there is the following apparent structural similarity between a con-
tribution to the second and third terms in the r.h.s. of (52), and the r.h.s. of (54),

(∂a n)2 ↔ cos2 ϑ · (∂a k)2 + sin2 ϑ · (∂a l)2

Ja2 ↔ (∂a ϑ)2

In [4] it has been suggested that this structural similarity can be interpreted in terms
of an electric-magnetic duality. Here we propose that it suggests a duality between the
spin and the charge.
The last two terms in (51) can also be represented in terms of UC (1) × UI (1) gauge
invariant variables as follows,

1 2 1 2
ρ t+ (DC 2 ˆ 2
A a ēb ) = ρ n+ (∂a ēb )
2 2

1 ρ2 n+ √ √
= 2
< ∂a (p + q), p − q − 4 2 is > · < ∂a (p − q), p + q − 4 2 is > (55)
128 |s|
1 2 1 2
ρ t− (DC 2 2
A a eb ) = ρ n− (∂a êb ) =
2 2

1 ρ2 n− √ √
= 2
< ∂a (p + q), p − q + 4 2 is > · < ∂a (p − q), p + q + 4 2 is > (56)
128 |s|

26
Shortly we shall argue that these two terms admit a geometrical interpretation in the
Grassmannian framework. However, prior to this we consider the remaining contribu-
tions to the Yang-Mills Lagrangian.
We proceed with the first term in (51). When we eliminate Aa in favor of the super-
current Ja we get for this term

1 1
(Fab + 2ρ2 t3 Hab )2 = (Lab + Mab − n3 Kab − 2ρ2 n3 Hab )2 (57)
4 4

Here
Lab = ∂a Jb − ∂b Ja
1
Mab = 2
n · DĈa n × DĈb n (58)
Kab = ∂a Ĉb − ∂b Ĉa
In particular, (57) and (58) involve only quantities which are explicitely UC (1) and UI (1)
invariant. The covariant derivative (53) ensures that all quantities are also independently
SO(4) (Lorentz) invariant.
We note that we can write the second and third terms in the r.h.s. of (57) as follows,

1 1
n3 (∂a Ĉb − ∂b Ĉa ) − n · DĈa n × DĈb n = ∂a [n3 Ĉb ] − ∂b [n3 Ĉa ] − n · ∂a n × ∂b n (59)
2 2

The structure in (59) is reminiscent of the ’t Hooft tensor [23]. The last term is the
pull-back of the volume two-form on S2 , and if we introduce the corresponding Kirillov
one-form (36)
1
− n · ∂a n × ∂b n = ∂a Qb − ∂b Qa
2
we can combine the first three terms in the r.h.s. of (57) into

Lab + Mab − n3 Kab = ∂a (Jb − n3 Ĉb − Qb ) − ∂b (Ja − n3 Ĉa − Qa )

In summary, when we combine our results we find that in terms of the spin-charge
separated variables the Yang-Mills Lagrangian has the following UC (1) × UI (1) invariant

27
form

1 2 1 1 1 ρ2 
LY M = Fab + (∂a ρ)2 + ρ2 Ja2 + ρ2 (DĈa n)2 + (∂a p)2 + (∂a q)2
4 2 2 8 4

1  3 3
+ ρ2 n+ (∂a ēˆb )2 + n− (∂a êb )2 + (1 − n23 )ρ4 − ρ4 (60)
4 8 8
where

1
Fab = ∂a Jb − ∂b Ja + n · ∂a n × ∂b n − {∂a (n3 Ĉb ) − ∂b (n3 Ĉa )} − 2ρ2 n3 Hab
2

We find it noteworthy that the final Lagrangian (60) contains only UC (1) and UI (1)
invariant quantities, despite the fact that in (48) we have only introduced gauge fixing
for the off-diagonal components Xa± . In particular, the UC (1) ∈ SU(2) gauge invariance
has been eliminated explicitely by the introduction of gauge invariant variables. This
elimination of the UC (1) ×UI (1) gauge invariance has been at the expense of introducing
variables Ĉa , ê and n± which transform according to a projective representation of the
SO(4) (Lorentz) group. However, in (60) these variables appear only in SO(4) invariant
combinations.
The final Lagrangian (60) has a very interesting structure. It describes the interacting
dynamics between a version of the O(3) nonlinear σ-model that one of us introduced in
[6] and the G(4, 2) Grassmannian nonlinear σ-model.
Clearly, the natural interpretation of the real scalar field ρ is in terms of a condensate.
Since ρ is a positive definite quantity we can expect that it develops the non-vanishing
ground state expectation value (24) that characterizes a material background in (60);
see [19], [21].
Due to the presence of the third term in (60), a nonvanishing ∆ in (24) leads to an
effective mass to the vector field Ja . As a consequence this vector field is subject to the
Meissner effect. If we assume that at large distances we can ignore the contribution from
Ja , the remaining Lagrangian involves only variables that describe a the present version
of the O(3) σ-model and the G(4, 2) Grassmannian non-linear σ-model.

28
 In the London limit where we replace ρ by its ground state expectation value (24), the
version of the O(3) nonlinear σ-model that has been embedded in (60) has the following
Lagrangian,
∆2 Ĉ 2 1 n o2 3
(D a n) + n · ∂a n × ∂b n − 2 · { ∂a (n3 Ĉb ) − ∂b (n3 Ĉa ) } + ∆2 (1 − n23 ) (61)
8 16 8
This is in close resemblance with the effective Lagrangian (1), which we have proposed
previously could be an effective model for SU(2) Yang-Mills theory [3]-[5]. The difference
stems from the fact that here the ± components of the order parameter n lack Lorentz
invariance due to the one-cocycle (32). The Lorentz invariance of the Lagrangian is
restored by the presence of the similarly Lorentz invariance violating vector field Ĉa .
We note that the last term in (61) is an additional O(3) symmetry breaking potential
term. It is Lorentz invariant since n3 is the sole component of n that is a scalar under
Lorentz transformations.
The reason for the particular combination of the potential terms that we have in-
troduced in (46) (the third and fourth terms) becomes now obvious: This combination
ensures that the angular variable θ of n in the parametrization (43) acquires a positive
mass term. This choice still leaves the potential term involving only ρ with a negative
sign. Eventually, this sign will also find an explanation.
The original model (1) supports knotted closed strings as stable solitons. The version
(61) involves also the dynamical gauge field Ĉa that restores Lorentz invariance in the
present case. It would be very interesting to understand how the addition of this field
affects the soliton structure of the theory.
We now proceed to identify the G(4, 2) nonlinear σ-model that has been embedded
in (60). This embedding is determined by the kinetic term (54) that can be written as
ρ2 
|(∂a + iCa )e|2 = (∂a p)2 + (∂a q)2
2
This reveals the topological
SO(4)
G(4, 2) ∼ ∼ S2 × S2
SO(2) × SO(2)

29
structure of the Grassmannian. We conclude that when we subject p and q to the two
conditions (27), these two three-component vector fields describe the four dimensional
Grassmannian manifold G(4, 2). Indeed, a priori the two vector fields p and q have six
independent components. But due to the two conditions (27) only four of the components
are independent, and correspond to coordinates on the four dimensional Grassmannian
manifold G(4, 2).
Now, we return to the two terms (55) and (56) which together with the last term in
(57) describe the coupling between the Grassmannian model and the O(3) model. We
argue that the Grassmannian contribution in the interaction terms (55) and (56) can be
identified as the (anti)holomorphic one-form on the complex manifold G(4, 2) ∼ S2 × S2 .
For this we introduce the explicit parametrization (38), (39). When we specify to the
magnetic limit where ϑ → π/2 in (34) we find for the Grassmannian contribution in (55)

1 √ ϑ→0
· < ∂a (p+q), p−q −4 2 is > −→ −2eiψ (l+im)·∂a k = −2eiψ (dβ+i sin βdα) (62)
|s|

This is the (unique) holomorphic one-form on the magnetic two-sphere described by k.

Similarly we find in the electric limit ϑ → 0 that the Grassmannian contribution to (55),
(56) yields the (anti)holomorphic one-form on the electric two-sphere in G(4, 2) ∼ S2 ×S2
which is described by the unit vector l. These observations endorse our proposal that
the Grassmannian contributions in (55), (56) engage the holomorphic and the anti-
holomorphic one-forms on the complex manifold G(4, 2) ∼ S2 × S2 .
Notice that in the interaction terms (55), (56) the phase ψ in (62) can be combined
with the phase of n± into
φ + 2η → φ + 2η + ψ

Finally, for the last term in (57) we have in the Lorentz invariant ground state where
n3 = ±1
1 2 1 1
(2ρ n3 Hab )2 = n23 ρ4 ≈ ρ4
4 2 2

30
When we compare this with the last term in (60) we conclude that despite the negative
sign of this term we have an overall stability of the theory.

31
VII. GAUGE COVARIANCE

Our description of the spin-charge separation employs the Pauli frame (5), (6) that
identifies the diagonal matrix σ 3 with the direction of the UC (1) subalgebra in the
SU(2) Lie algebra. We now proceed to show that the spin-charge separation is frame
independent, instead of σ 3 we can select the direction of the Cartan subalgebra UC (1)
in SU(2) in an arbitrary and space-time dependent manner. For this we introduce an a
priori arbitrary g(x) ∈ SU(2) and perform the conjugation
g def
σ 3 −→ g σ 3 g −1 = mi σ i = m̂
g def def 1 1
σ ± −→ g σ ± g −1 = e±
i σ
i
= (e ± ie2i )σ i = ê±
2 i
Clearly, these matrices also satisfy the same algebra as (5), with m̂ the Cartan generator

[m̂, ê± ] = ±2ê±

[ê+ , ê− ] = m̂

The gauge transformation (7) by the matrix g maps (6) onto

1
Aa σ 3 + Xa+ σ + + Xa− σ − → Aa m̂ + Xa+ ê+ + Xa− ê− + aa m̂ + [∂a m̂, m̂]
2i
where
aa = −itr[σ 3 g −1 ∂a g]

We interpret this connection in the following manner: We define

1
Aa = Aia σ i = (Aa + aa )m̂ + [∂a m̂, m̂] = gAa σ 3 g −1 + 2ig∂a g −1 (63)
2i
1
= Ca m̂ + [∂a m̂, m̂]
2i
and
Xa = Xai σ i = Xa+ ê+ + Xa− ê− = g(Xa+ σ + + Xa− σ − )g −1 (64)

Here Aa is the connection originally introduced by Duan and Ge [10], and subsequently
by Cho [11]; see also [3].

32
 We introduce a generic h(x) ∈ SU(2) and redefine

g → gh

This determines a transformation under which A transforms as a connection

A → hAh −1 + 2ihdh −1

while X transforms as a tensor,

X → hX h −1

In (63), (64) we have superficially fourteen field degrees of freedom. These are the four
components of Ca , the eight components of Xa± and the two independent components
of m̂. However, if we impose the h-covariant condition [24]

D[A]ij j
a Xa = 0 (65)

this condition eliminates two of the field variables and we are left with only the twelve
independent components of a four dimensional SU(2) gauge field.
The condition (65) is a gauge covariant version of the maximal abelian gauge condition
(49). Explicitely, when we substitute (63) and (64) in (65) and use the identity

mi (δ ij ∂a + ǫikj Aka )Xaj = 0

we conclude that (65) is equivalent to the condition (49) for the original components
(Aµ , Xµ± ). In particular, when we choose m̂ ≡ σ 3 we find that (65) reduces to (49)
and we retain all our previous results. This confirms that our separation between the
spin and the charge in the Yang-Mills Lagrangian is gauge covariant, independent of the
direction of UC (1) in the SU(2) gauge group. Furthermore, the connection by Duan and
Ge and by Cho acquires a role in the gauge covariantization of our formalism.

33
VIII. CONFORMAL GEOMETRY

The Yang-Mills Lagrangian has a number of attractive features that become trans-
parent when we present it in terms of the independent spin and charge variables. For
example, the Lagrangian can be related to a two-gap superconductor model [20], and ρ
admits also an independent interpretation as a gauge invariant condensate [19], [21].
Here we shall propose an alternative interpretation of the Yang-Mills Lagrangian. We
shall propose that ρ can be viewed as the conformal scale of a conformally flat metric
tensor, and (60) describes the coupling between matter fields and the Einstein-Hilbert
gravity in the presence of a nontrivial cosmological constant.
The version of the spin-charge separated Yang-Mills Lagrangian that we shall employ
is the following,
(1) (2) (3) (4)
LY M = LY M + LY M + LY M + LY M

where
 2
(1) 1 1 Ĉ Ĉ 2
LY M = ∂a Jb − ∂b Ja + n · D a n × D b n − n3 (∂a Ĉb − ∂b Ĉa ) − 2ρ n3 Hab (66)
4 2

(2) 1 2 2 1 2 Ĉ 2
LY M = ρ Ja + ρ (D a n) (67)
2 8

(3) 1 2 1 2
LY M = ρ2 |DC 2 C 2 C
A a eb | + ρ t+ (D̄A a ēb ) + ρ t− (DA a eb )
2
(68)
2 2

(4) 1 3 3
LY M = (∂a ρ)2 + (1 − n23 )ρ4 − ρ4 (69)
2 8 8
Our goal is to write these terms in a manifestly covariant manner, with the conformally
flat metric tensor
 ρ 2
gµν = δµν (70)
∆
Here ∆ is a constant with dimensions of mass: Since ρ has dimensions of mass we
need to introduce ∆ so that the components of the metric tensor acquire their correct

34
dimensionality. The obvious choice is to identify ∆ with the vacuum expectation value
of the condensate ρ according to (24).
We introduce the vierbein
gµν = δab E a µ E b ν (71)

where
δ ab = g µν E a µ E b ν

and the vierbein E a µ is given explicitely by

ρ a
Eaµ = δ µ
∆

with
E a µ Eb µ = δ a b

The Christoffel symbol of the metric (70) is

1 1 √
Γµνσ = g µη (∂ν gησ + ∂σ g ην − ∂η gνσ ) = { δσµ δντ + δνµ δστ − δ µτ δνσ }∂τ ln g (72)
2 4

where
√  ρ 4
g=
∆
The spin connection is defined by demanding covariant constancy of the vierbein,

∂µ Ea ν + Γνµλ Ea λ − ωµb a Eb ν = 0

This gives
ωµab = E a ν ∇µ Eb ν = E a ν (∂µ Eb ν + Γνµλ Eb λ )
(73)
= −Eb ν ∇µ E a ν = −Eb ν (∂µ E a ν − Γλµν E a λ )
In these relations we also indicate how the covariant derivative ∇µ acts on the vector
and co-vector fields.
Explicitely we get from the metric tensor (70), (71) for the spin connection

1 a σ √
ωµab = { δ µ δb − δbd δ d µ δ ac δc σ }∂σ ln g
4

35
 We employ the vierbein E a µ and the complex Grassmannian zweibein (15) to intro-
duce the following complex zweibein

eµ = E a µ ea
ēµ = E a µ e∗a

This zweibein is then normalized w.r.t. the metric gµν according to

g µν eµ ē∗ν = 1
g µν eµ eν = g µν ē∗µ ē∗ν = 0

When we push forward the spin connection into

ωµab → ωµλν = Ea λ ωµab E b ν

we can introduce the generally covariant version of the connection (23),

Cµ = iēσ (∂µ eσ − Γλµσ eλ + ωµλσ eλ ) = iēσ ∇µ eσ + iēλ ωµσλ eσ (74)

and when we twist the covariant derivative operator with (74),

∇Cµ = ∇µ + iCµ (75)

we have an operator that parallel transports the zweibein.

We now proceed to employ this formalism to rewrite the Yang-Mills Lagrangian (66)-
(69) in a generally covariant manner. Our computations simplify considerably when we
observe that for the metric tensor (70)
eν
∇µ eν + ωµλν eλ = ∂µ eν − Γλµν eλ + ωµλν eλ = ρ · ∂µ ( )
ρ
and
i √
Cµ = iēν ∂µ eν − ∂µ ln g
4
We start with the second and third term in (66), which we write in a generally
covariant form as follows,
1 1
n · DĈa n × DĈb n − n3 (∂a Ĉb − ∂b Ĉa ) → n · ∇Cµ n × ∇Cν n − n3 (∂µ Cν − ∂ν Cµ )
2 2

36
We also write the last contribution in (66) as

−2ρ2 n3 Hab = −iρ2 n3 (ea e∗b − eb e∗a ) → −i∆ · n3 (eµ e∗ν − eν e∗µ ) = −2∆ · n3 Hµν

When we define

1
Fµν = ∂µ Jν − ∂ν Jµ + n · {∇Cµ n × ∇Cν n − 2ẑ[(∂µ Cν − ∂ν Cµ ) + 2∆ · Hµν ]}
2

where ẑ is a unit vector in the z-direction of the internal space, we conclude that we
can write the entire (66) in the following generally covariant form

(1) (1) 1 √ µν ρσ
LY M → LY M = gg g Fµρ Fνσ (76)
4

Similarly we can write (67) in the following generally covariant form,

(2) (2) √
LY M → LY M = ∆2 · g g µν (Jµ Jν + ∇Cµ n · ∇Cν n) (77)

We now proceed to (68). For this we send our flat space UI (1) covariant derivative of
the Euclidean metric Grassmannian zweibein to a generally covariant form as follows,

(∂a + iCa )eb → Ea λ {δ νλ ∇Cµ + ωµν λ }eν = Ea λ Dµν λ eν

We have here introduced the following twisted covariant derivative

Dµν λ = δ νλ ∇Cµ + ωµν λ

It extends the action of the twisted covariant derivative (75) to the vector fields eν .
With this, we can present the entire (68) in the following covariant form,

(3) (3)
LY M → LY M

 
√ 1 1
= ∆ ·2 µν λη
g·g g (D̄µσλ ēσ )(Dνκη eκ ) + t+ (D̄µ λ ēσ )(D̄ν η ēκ ) + t− (Dµ λ eσ )(Dν η eκ )
σ κ σ κ
2 2
(78)

37
 Finally, we proceed to (69). We introduce the Riemann tensor in terms of the
Christofffel symbol (72)

Rµνσ λ = ∂ν Γλµσ − ∂µ Γλνσ + Γηµσ Γλην − Γηνσ Γληµ

and the Ricci tensor

Rµν = Rνµ = Rµλν λ

and the Ricci scalar

R = Rµµ

which transforms according to (in D dimensional space)

R → R̃ = φ−2 {R − 2(D − 1)g µν ∇µ ∇ν ln φ − (D − 2)(D − 1)g µν (∇µ ln φ)(∇ν ln φ)} (79)

under the conformal scaling

gµν → g̃µν = φ2 gµν

of the metric tensor. For the metric tensor (70) this leads to the identification

1 1 √
(∂µ ρ)2 → gR
2 ∆2

This is the covariant interpretation of the first term in (69). For the remaining terms in
(69) (except for the surface term) we get from (70)

3 3 1 3 2 √ 1 3√
(1 − n23 )ρ4 − ρ4 → (1 − n3 ) g − g
8 8 ∆2 8 ∆2 8

and we conclude that the entire (69) can be presented in the following generally covariant
manner,

(4) (4) 1 √ 3 1 √ 3 1 √
LY M → LY M = 2
gR − 2
g + 2
(1 − n23 ) g (80)
∆ 8∆ 8∆

Here the first contribution is the standard Einstein-Hilbert Lagrangian, the second is
the standard (negative) cosmological constant term, and the third gives a (in general)
space-time dependent correction to the cosmological constant when n3 6= ±1.

38
 Note that the sign of the Ricci scalar is consistent with the sign proposed in [25],
ensuring that the Euclidean Einstein-Hilbert Lagrangian is bounded from below.
We conclude by summarizing, that when we combine (76), (77), (78) and (80) we find
that in terms of the spin-charge separated variables the Yang-Mills Lagrangian can be
written in the following generally covariant form

(1) (2) (3) (4)

LY M = LY M + LY M + LY M + LY M

Here
(1) 1 √ µν ρσ
LY M = gg g Fµρ Fνσ
4
which has the standard form of the generally covariant Maxwell Lagrangian,

(2) √
LY M = ∆2 · g g µν (Jµ Jν + ∇Cµ n · ∇Cν n)

is a generally covariant current-current interaction term,

 
(3) 2 √ 1 1
LY M = ∆ · g·g g µν λη
(D̄µ λ ēσ )(Dν η eκ ) + t+ (D̄µ λ ēσ )(D̄ν η ēκ ) + t− (Dµ λ eσ )(Dν η eκ )
σ κ σ κ σ κ
2 2

gives the kinetic term for the Grassmannian eµ together with two terms describing its
interaction with n where we recall that these interaction terms can be related to the
(anti)holomorphic one-forms on the Grassmannian. Finally,

(4) 1 √ 3 1 √ 3 1 √
LY M = 2
gR + − 2
g + 2
(1 − n23 ) g
∆ 8∆ 8∆

is the Einstein-Hilbert Lagrangian together with a cosmological constant with a space-

time dependent correction that vanishes in the ground state where n3 = ±1.
The final Lagrangian has a manifestly generally covariant form, and it coincides with
the spin-charge separated Yang-Mills Lagrangian (66)-(69) when we evaluate it using
the conformally flat metric tensor (70). Notice that from the present point of view the
nontriviality of the ground state expectation value in (24) becomes quite natural. When
ρ vanishes our conformal space-time just shrinks away.

39
IX. STATIC LIMIT

It is often instructive to inspect the static limit of the Lagrangian, it gives an in-
dication on the ground state properties of the theory. We reach the static limit when
we only retain the spatial derivatives and set the time component of the vector field e
to zero. This sends ϑ → π/2 in (34), and as a consequence the electric vector field p
vanishes and the only non-vanishing contribution to the tensor field Hab is

1
Hij = √ ǫijk lk
2 2
where l is the unit vector in the magnetic direction. Furthermore, from (36), (37), (42)
we get
1
∂iĈ j − ∂jĈ i = − √ l · ∂i l × ∂j l
2 2
and
1
(∂i ej )2 = · {( l + im) · ∂i k}2
16
With these, we get from (60) for the energy density in the static limit

1 1 1 1
Hstatic = (∂i ρ)2 + ρ2 Ji2 + ρ2 (DĈi n)2 + ρ2 (∂i l)2 (81)
2 2 8 32
1 2
+ ρ n+ e−2iψ ( [ l + im] · ∂i k)2 + n− e2iψ ( [ l − im] · ∂i k)2 (82)
64
1 3 3
+ Fij2 + (1 − n23 )ρ4 − ρ4 (83)
4 8 8

Here

1 1
Fij = ∂i Jj − ∂j Ji + n · DĈi n × DĈj n + n3 √ { l · ∂i l × ∂j l − 2ρ2 ǫijk lk } (84)
2 2 2
From this we can draw the following conclusions:
There is an apparent duality between the internal vector field n and the space-valued
vector field l. In particular, both are embedded in (81)-(83) in a manner that employs
the version (1) of the O(3) nonlinear σ-model. As a consequence both n and l have
the potential of supporting closed knotted strings as stable solitons. It is suggestive to

40
interpret l as a “magnetic” order parameter, and n as an “electric” order parameter in
the static limit [4].
The vector field n takes values in the internal space. Due to the one-cocycle (32)
in the Lorentz transformations, it has a unique rotation invariant ground state value at
large distances which is given by (44). But l is a space valued vector field, it transforms
as a vector under spatial SO(3) rotations. Consequently its only conceivable asymptotic
ground state value at large distances is the spherically symmetric

r→∞ x
l −→
r

From this we get the asymptotic behaviour

r→∞ xk 1
l · ∂i l × ∂j l −→ ǫijk 3
∼ 2 ǫijk lk
r r

for the third contribution to (84). Note that this is reminiscent of a magnetic monopole.
We combine the last two terms in (84) asymptotically into

r→∞ 1 xk
l · ∂i l × ∂j l − 2ρ2 ǫijk lk −→ ( − 2ρ2
)ǫijk (85)
r2 r

We now make the following proposals: For a finite energy, we can expect that each
of the positive definite terms in the static Hamiltonian are integrable. This means that
asymptotically at large distances, in an analytic power expansion in r we can expect

1
ρ(r) . O( )
r

In terms of the four dimensional metric tensor (70) this suggests that for finite energy
the space should be compact. But from the present static point of view we can also argue
as follows. We consider the space-time to have the topology of M × R1 where M is the
three dimensional space manifold and R1 is the time. If we define the three-dimensional,
spatial metric tensor by setting
(3) ρ4
gij = δij (86)
∆4

41
we find that some of the terms in the energy density admit an independent, three di-
mensional geometric interpretation: From (79), the first term in the r.h.s. of (81) can
be written in terms of the three dimensional Ricci scalar as

1 1 1 p (3) (3)
(∂i ρ)2 = g R
2 80 ∆2

The second term can be written as

1 2 2 1 p (3) (3) ik
ρ Ji = g g Ji Jk
2 2

Similarly we conclude that each of the terms in (81) and (82) admit a generally covariant
interpretation in terms of the present three dimensional conformal geometry.
For the terms in (83) the present three dimensional geometric interpretation appears
to fail. But when we demand that the quadratic terms are independently integrable,
since the asymptotic behaviour of n is dictated by (44) we can argue that it becomes
very natural to expect that asymptotically the two terms in (85) cancel each other. This
suggests that at large distances we have

1 1
ρ2 ∼
2 r + λ2
2

where λ is some parameter. When we substitute this in (86) we find that the spatial
part of our space-time becomes asymptotically compactified into the sphere S3 . It would
be very interesting if this proposal could be made more rigorous.

42
X. SUMMARY

We conclude our article with a summary of our results and a number of remarks on
their possible physical consquences.
We have introduced a novel, complete field decomposition in the Yang-Mills La-
grangian. The decomposition implements a separation between the spin and the charge
in the gauge field. The decomposition also introduces an internal, compact U(1) in-
teraction. A compact U(1) gauge theory is known to exhibit confinement in a strong
coupling domain which is separated from a weakly coupled and deconfined domain by
a first order phase transition. Since the coupling in an abelian theory should increase
when the distance scale decreases, the spin-charge separation is not in an apparent con-
flict with the high energy limit of the Yang-Mills theory, represented by asymptotically
free and massless gauge bosons.
The spin-charge separated Yang-Mills Lagrangian describes the interacting dynam-
ics between a version [6] of the O(3) nonlinear σ-model and a G(4, 2) Grassmannian
nonlinear σ-model, in a conformally flat spacetime and in the presence of both the
Einstein-Hilbert Lagrangian and a negative cosmological constant term.
The conformal scale of the metric coincides with the gauge invariant condensate that
has been studied previously in [19], [21]. Numerical lattice studies indicate that the
ground state value (24) of the condensate is nonvanishing. From our geometrical point
of view this is an expected result: If the conformal scale vanishes there is no space-time.
The metric properties of the classical Yang-Mills theory are consistent with the short
distance limit of a renormalizable higher derivative gravitation theory with a Lagrangian
of the form (3). It would be truly exciting if at short distance Einstein gravity metamor-
phoses into a Yang-Mills theory, as a single renormalizable quantum theory of material
interactions.
The presence of the higher derivative term in (3) gives rise to a linearly increasing
component in the large distance gravitational interaction. From the point of view of
the Yang-Mills theory this may have some obvious advantages. However, at distance

43
scales which are well beyond those that should be described by the Yang-Mills theory
as such, it may become desirable for the β-function of the coupling γ in (3) to force this
coupling to flow towards γ = 0. Such a large distance behaviour in the quantum theory
would then leave the conventional Einstein gravity as the sole surviving very long range
component of the spin-charge separated Yang-Mills theory.
Note that since γ flows towards γ → ∞ at short distances, the asymptotic condition
(4) dissolves the massless modes of the Einstein gravity from the (not too) short distance
spectrum. This leaves us with a gapped and spin-charge separated Yang-Mills theory
that describes asymptotically free gauge vectors as distance scale goes to zero.
Our results suggest that due to fluctuations in the gauge invariant condensate ρ
and the vector field n, at short distances both Newton’s constant and the cosmological
constant become variable.
The version of the O(3) nonlinear σ model that embeds the vector field n in the Yang-
Mills Lagrangian, essentially coincides with (1). This Lagrangian is known to describe
knotted strings as stable solitons. Our results then support the proposal [3]-[5] that such
strings are present in the spectrum of the SU(2) Yang-Mills theory.
In the absence of the potential term in (1), the spectrum of the O(3) model contains
two massless Goldstone bosons. These bosons originate from the asymptotic breaking of
the global O(3) symmetry, when we select the large distance direction for the unit vector
s. In the case of the Yang-Mills theory, these massless Goldstone bosons are removed
by the one-cocycle that breaks the Lorentz invariance of the pertinent order parameter
n. The requirement that the large distance ground state is Lorentz invariant uniquely
fixes the asymptotic direction of the order parameter.
The spin-charge separated Yang-Mills theory also describes the G(4, 2) Grassmannian
nonlinear σ-model. The S2 × S2 structure of the G(4, 2) manifold has an interpretation
in terms of electric and magnetic variables, the two spheres S2 are related to each other
by an electric-magnetic duality. In terms of the corresponding unit vector fields, the
Grassmannian contribution to the Yang-Mills Lagrangian admits a very transparent

44
realization. In particular, in the static (magnetic) limit we are left with only one three-
component unit vector field l, pointing in the magnetic direction. This vector field
is embedded in the Yang-Mills Lagrangian by a version of the Lagrangian (1). Conse-
quently we have an additional duality between two different embeddings of (1), described
by n and l respectively. In particular, this means that the vector field l has also the
potential of supporting closed knotted strings as stable solitons.
The interaction between the O(3) σ-model and the Grassmannian σ model involves
the (anti)holomorphic one-forms on the Grassmannian manifold. These appear in a com-
bination with the n± components of the vector field n. The presence of the one-cocycle
in the Lorentz transformation of n implies that at large distances these interaction terms
are absent.
Finally, we have argued that demanding finiteness of energy in the Yang-Mills theory
enforces a compactification of the space. We have proposed that it is natural for the
asymptotic topology of the space-time to coincide with that of the manifold S3 × R1 .

45
XI. ACKNOWLEDGEMENTS

The work by L.D.F. has been supported by RFBR grant 05-01-00922, CRDF grant
RUM-1-2622-ST-04, and the program ”Problems of nonlinear dynamics” of Presidium
of Russian Academy of Sciences. The work by A.J.N. has been supported by a Grant
from ANR, by a VR Grant, by a STINT Institutional Grant and by a STINT Thunberg
Stipend. L.D.F. thanks L. Lipatov and A. Slavnov for discussions. A.J.N. thanks M.
Chernodub, U. Danielsson, U. Lindström, M. Volkov and K. Zarembo for discussions,
M. Niedermaier, M. Paranjape and A. Tseytlin for communications, and S. Slizovskiy
for comments. A.J.N. also thanks the Yukawa Institute at the Kyoto University, the
Department of Physics at Tokyo University and the Asian Pasific Center for Theoretical
Physics for hospitality during the early part of this work.

46
 [1] A private communication by C.N. Yang.
[2] see the website http://www.claymath.org/millennium/ for a general discussion.
[3] L.D. Faddeev and A.J. Niemi, Phys. Rev. Lett. 82 (1999) 1624; L.D. Faddeev and A.J.
Niemi, Phys. Lett. B449 (1999);
[4] L.D. Faddeev and A.J. Niemi, Phys. Lett. B525 (2002) 195;
[5] L.D. Faddeev, A.J. Niemi and U. Wiedner, Phys. Rev. D70 (2004) 114033
[6] L. Faddeev, Quantisation of Solitons, preprint IAS Print-75-QS70, 1975; and in Einstein
and Several Contemporary Tendencies in the Field Theory of Elementary Particles in Rel-
ativity, Quanta and Cosmology vol. 1, M. Pantaleo, F. De Finis (eds.), Johnson Reprint,
1979
[7] R. Battye and P. Sutcliffe, Proc. Roy. Soc. London A455 (1999) 4305; Phys. Rev. Lett.
81 (1998) 4798; J. Hietarinta and P. Salo, Phys. Lett. B451 (1999) 60
[8] F. Lin and Y. Yang, Comm. Math. Phys. 249 (2004) 273
[9] L.D. Faddeev and A.J. Niemi, Nature 387 (1997) 58
[10] Y.S. Duan and M.L. Ge, Sinica Sci. 11 (1979) 1072
[11] Y.M. Cho, Phys. Rev. D21 (1980) 1080; Y.M. Cho, Phys. Rev. Lett. 46 (1981) 302
[12] E. Langmann and A.J. Niemi, Phys. Lett. B463 (1999) 252; Y.M. Cho, H.W. Lee and
D.G. Pak, Phys. Lett. B525 (2002) 347
[13] S.E. Barnes, J. Phys. F6 (1976) 1375; L.D. Faddeev and L.A. Takhtajan, Phys. Lett. A85
(1981) 375; P. Coleman, Phys. Rev. B29 (1984) 3035
[14] G. Baskaran and P.W. Andersson, Phys. Rev. B37 (1988) 580; P.W. Andersson, Science
235 (1987) 1196; for a review see P.A. Lee, N. Nagaosa and X.-G. Wen, cond-mat/0410445
[15] A.J. Niemi and N. Walet, Phys. Rev. D72 (2005) 054007
[16] M.N. Chernodub and A.J. Niemi, quant-ph/0604162
[17] K.S.S. Stelle, Phys. Rev. D16 (1977) 953; E.S. Fradkin and A.A. Tseytlin, Phys. Lett.

47
 104B (1981) 377; E.S. Fradkin and A.A. Tseytlin, Phys. Rept. 119 (1985) 233
[18] M.N. Chernodub and M.I. Polikarpov, in Confinement, duality, and nonperturbative as-
pects of QCD, P. van Baal, Ed. (Plenum Press, New York 1998) (hep-th/9710205); T.
Suzuki, Prog. Theor. Phys. Suppl. 131, 633 (1998); R.W. Haymaker, Phys. Rept. 315,
153 (1999).
[19] K.-I. Kondo, Phys. Lett. B514 (2001) 335
[20] A.J. Niemi, JHEP 0408 (2004) 035
[21] F.V. Gubarev, L. Stodolsky and V.I. Zakharov, Phys. Rev. Lett. 86, 2220 (2001); L
Stodolsky, P. van Baal and V.I. Zakharov, Phys. Lett. B552, 214 (2003).
[22] E. Babaev, L.D. Faddeev and A.J. Niemi, Phys. Rev. B65, 100512(R) (2002)
[23] G. ’t Hooft, Nucl. Phys. B 79, 276 (1974)
[24] S.V. Shabanov, Phys. Lett. B463 (1999) 263
[25] G.W. Gibbons, S.W. Hawking and M.J. Perry, Nucl. Phys. B138 (1978) 141

48