Geometry and Dynamics of a Coupled 4D-2D Quantum Field Theory

Geometric and dynamical aspects of a coupled 4D-2D interacting quantum field theory - the gauged nonAbelian vortex - are investigated. The fluctuations of the internal 2D nonAbelian vortex zeromodes excite the massless 4D Yang-Mills modes and in general give rise to divergent energies. This means that the well-known 2D CP(N-1) zeromodes associated with a nonAbelian vortex become nonnormalizable. Moreover, all sorts of global, topological 4D effects such as the nonAbelian Aharonov-Bohm effect come into play. These topological global features and the dynamical properties associated with the fluctuation of the 2D vortex moduli modes are intimately correlated, as shown concretely here in a U(1) x SU(N) x SU(N) model with scalar fields in a bifundamental representation of the two SU(N) factor gauge groups.


Introduction
NonAbelian vortices -vortex solutions carrying nonAbelian continuous orientational zeromodes {B i } -have been extensively investigated in the last decade, revealing many interesting features arising from the soliton and gauge dynamics, topology, and global symmetries [1]- [5]. Typically they occur in a system in the color-flavor locked phase, i.e., systems in which the gauge symmetry is broken by a set of scalar condensates that, however, leave a color-flavor diagonal symmetry intact. Color-flavor locked systems appear to be quite ubiquitous in Nature. Standard QCD at zero temperature exhibits some characteristic features of this sort, as can be seen in a hidden-symmetry perspective [6]. They occur in the infrared effective theories of many N = 2 supersymmetric theories softly broken to N = 1, and may carry important hints about the mechanism responsible for quark confinement [7]. In particular they could shed light on the mysteries of nonAbelian monopoles. They are realized in high-density QCD in the color superconductor phase [8], which may well be realized in the interiors of neutron stars.
Other fascinating aspects associated with these vortex solutions appear when further gauge fields are introduced, coupled to part or all of the color-flavor diagonal global symmetry. We shall refer to these systems as "gauged nonAbelian vortices" in this paper. All sorts of global effects, such as nonAbelian Aharonov-Bohm phases and scattering, an obstruction to part of the "unbroken" gauge symmetry, nonAbelian statistics under the exchange of parallel vortices, Cheshire charges, etc. make their appearance, depending on the vortex orientations, {B i }. These phenomena have been investigated in various general contexts [9]- [13] with gauge symmetry breaking, G → H, with π 1 (G/H) = 1, and more recently, in the context of concrete model, e.g., in a U 0 (1) × SU l (N) × SU r (N) gauge theory, with scalar fields in the bifundamental representation of the two SU(N) gauge groups [14]- [16] .
A sort of paradox or dilemma seems to arise, however. One of the most important, characteristic features of nonAbelian vortices is their collective dynamics. Quantum fluctuations of the vortex internal orientational modes are described by various 2D (vortex worldsheet) sigma models, such as È N −1 1 . The È N −1 interactions are asymptotically free: the vortex orientation {B i } fluctuates strongly at large distances. As a result, all of the global topological effects mentioned above would be washed away.
Let us remind ourselves that a characteristic feature of a color-flavor locked vacuum is the fact that all massless Nambu-Goldstone particles are eaten by the broken gauge fields, all of which become massive, maintaining mass degeneracy among them. No massless scalars or gauge bosons survive in the bulk. In the vortex sector, the only massless modes are those of the vortex orientational zeromodes, which are a kind of Nambu-Goldstone excitation, confined within the vortex worldsheet. In the case of gauged nonAbelian vortices, instead, some combinations of the gauge bosons remain massless in the 4D bulk, and their coupling to the fluctuations of the 2D zeromodes {B i } are expected to affect significantly the dynamics of the latter. Some preliminary studies of these issues have been done [14,15].
It is the purpose of the present article to examine more thoroughly the effects of the unbroken 4D gauge interactions on the gauged nonAbelian vortex collective dynamics in the 2D vortex worldsheet. A nontrivial task is that of disentangling the effects of the extra gauge fields on the static vortex configuration itself from those of residual dynamical effects of the 4D massless gauge modes and their couplings to the massless 2D orientational modes. These problems will be worked out systematically below. This paper is organized as follows. In Section 2 the model and the main properties of the vortex solution are presented together with a brief review of the associated global effects. The closely related question of topological and geometric obstructions is reviewed, generally and in our concrete model, in Section 3. Vortex fluctuations and the induced excitations of the 4D Yang-Mills modes are worked out in Section 4 and Section 5. Section 6 is dedicated to clarifying the connection between the global, topological aspects of the gauge-vortex system and the dynamics of the zeromode excitations of the vortex. Finally, in Section 7 we examine the apparent discontinuity in physics in the limit in which one of the nonAbelian gauge factors, SU r (N) (or SU l (N)), is decoupled. It is argued that it is essentially due to the noncommutativity of the two limits, g r → 0 and R → ∞, where R is an infrared cutoff introduced to regularize the energy divergences caused by the vortex fluctuations. Appendix A deals with the peculiarity of the solution of Gauss's equations in the particular g r = 0 case. Appendix B proves the uniqueness of the Ansatz Eq. (4.4) used to solve Gauss's equations in Section 4 and Section 5.

The model, vortex solutions and AB effect
Even though our study can be generalized to the case of an arbitrary gauge group of the type we shall choose, for concreteness, to work with G L = G R = SU(N). The matter sector consists of a complex scalar field Q in the bifundamental representation of the two SU(N) factors, with unit charge with respect to U 0 (1).

Vortex solutions
We shall work with a BPS-saturated action 2 where v 2 0 = Nξ and the covariant derivative is The scalar-field condensate in the vacuum takes the form leaving a left-right diagonal SU(N) gauge group unbroken. The fields remain massless in the bulk, whereas the orthogonal combination and the U(1) field a µ , become massive.
The nontrivial first homotopy group means that the system admits stable vortices, which are our main interest below. The vortex solutions can be found by the BPS completion of the expression for the tension (for configurations depending only on the transverse coordinates x and y): The BPS equations are accordingly: For a minimal vortex with a fixed orientation in color-flavor, for example (1, 1 N −1 ), one can take the scalar field Ansatz to be 3 whereas the nonAbelian and Abelian gauge fields can be written in the diagonal form and The boundary conditions are The BPS equations (2.8)-(2.9) then show that the profile functions satisfy which can be solved by numerical methods. These equations are identical to those found earlier for the global nonAbelian vortex, i.e., for g r = 0 or g l = 0, except for the fact that the gauge fields compensating the scalar winding energy ∂Q/∂θ are now shared between the left and right SU(N) fields. In other words, the static vortex profile remains basically unmodified as compared to the standard nonAbelian vortex (and for that matter, the ANO vortex), with a well-defined width, ∼ 1 √ ξ . The vortex tension is given by (see Eq. (2.7)) (2.20)

The Aharonov-Bohm effect
Whenever the (untraced) Wilson loop in some representation of a gauge field around the vortex is not equal to unity, particles belonging to that representation are transformed when encircling the vortex. The transformation is given by the (untraced) Wilson loop. This is called the Aharonov-Bohm (AB) effect.
These Wilson loops are easily calculated from Eqs. (2.11-2.13) To obtain the AB phase given a representation of G is now quite straightforward. For example, consider the scalar field Q, which is of charge 1 under U(1) 0 , transforms in the fundamental of SU(N) l and the antifundamental of SU(N) R . The corresponding AB phase is simply the product of (2.21) and (2.22) divided by (2.23), which is the identity, so Q does not transform. This is an important consistency check because the Q field condenses and any condensate must be single valued.
The AB phase can be calculated using the gauge fields in the ultraviolet A (l) and A (r) basis or in the mass eigenstate A and B basis. Of course the two answers must agree. To use the second basis, one needs the corresponding Wilson loops, which are the exponentials of The fact that A vanishes implies that the AB phase is independent of the A charge, it only depends on the charge under the massive gauge field B and U(1) 0 .
Rewriting (2.2) at spatial infinity as 26) one sees that in the mass basis Q has charge 1 under U(1) 0 and also under B. Therefore upon circumnavigating a vortex at large radius, Q is rotated by the exponential of This exponential is unity and so again we recover the fact that Q is invariant. The invariance of Q, which is necessary for the condensate to be well-defined, imposes that the matrix in Eq. (2.27) is integral. This in fact is the quantization condition on the nonAbelian vortex charge. In the low energy gauge theory it is the topological condition that the flux of a Higgsed gauge field is quantized. In particular, it is preserved by any continuous deformation of the vortex.

Vortex zeromodes (degeneracy)
The solution (2.10) further breaks the unbroken color-flavor diagonal H = SU l+r (N) group as implying the existence of degenerate solutions, corresponding to the coset, (2.29) The general solutions are related to (2.10) by global SU l+r (N) transformations, i.e., where the rotation matrix That the vortex (internal) moduli space is exactly a È N −1 in our model has been verified recently in [16], [17] by studying these vortex solutions in the large-winding limit, where vortex configurations can be analytically determined, and accordingly all zero modes can be determined by following the analysisà la Nielsen-Olesen-Ambjorn [19]- [21].
In contrast with the standard nonAbelian vortices (which appear in theories with g r = 0 or g l = 0), here the gauged vortex solutions with distinct moduli {B i } are related by global part of the SU(N) gauge transformations. The moduli can be promoted to collective coordinates if they are allowed to depend on z and t. The orthogonal parts of the collective coordinates can combine with each other, or with the (z, t)-independent parts of the fields, in gauge-invariant combinations. As a result the corresponding oscillations do correspond to physical degrees of freedom, which are eaten in a kind of mini-Higgs mechanism.
Even in the case of vortices with constant {B i }, the relative orientations of multiple vortex systems do have observable effects and hence are physical [12], [13], [16], [17]. Consider two or more parallel vortices with different orientations, {B i }. Generalizing Eqs. (2.21-2.23), particles carrying (for instance) fundamental charges with respect to G = U 0 (1) × SU L (N) × SU R (N), will experience various nonAbelian Aharonov-Bohm (AB) effects ("gauge transformations") when encircling one of them, ψ → Γ ψ, (2.34) If a particle is in a generic representation of G, it will experience an AB effect similar to the above but with appropriate charge and generators.
If there are multiple vortices with orientations {B 1 , B 2 , . . .}, various closed paths encircling these vortices give rise to nonAbelian AB effects: the gauge transformations experienced by a particle depend on the order in which various vortices are circumnavigated. These and many other beautiful features related to such systems have been discussed in [12,13,16,17].

Topological and geometric obstructions
In the vacuum our scalar condensate breaks the gauge symmetry G down to a smaller gauge group H. In the presence of a vortex the symmetry is broken yet further. However, far from the vortex, the symmetry group H is restored. There are natural gauges, such as the regular gauge used above, in which the scalar condensate is position-dependent and therefore so is the embedding of H in G. The AB effect potentially makes the embedding of H yet more complicated, as elements of H do not generally commute with the AB phase.

Topological obstructions in general
In various contexts it is useful to define a global symmetry corresponding to the gauge symmetry H. The fact that the embedding of H in G is, in many gauges, position dependent means that such a global, continuous definition of the generators of H may not exist for all values of the azimuthal angle θ.
Indeed there are many well-known cases where such an obstruction is known to exist. For example, consider nonAbelian monopoles. These are 't Hooft-Polyakov monopoles which preserve a nonAbelian symmetry group H which is a subgroup of the ultraviolet gauge group G. Consider a sphere S 2 which links a monopole. It is known [22] that it is not possible to continuously define a set of generators of H on this S 2 . The obstruction arises as follows. The connection defines a trivial G gauge bundle on S 2 , as the gauge field is continuous and globally defined inside of the sphere. The gauge symmetry G is broken to H, and so one obtains an H subbundle of the G bundle. However the H bundle is nontrivial. Indeed, it is the nontriviality of the H bundle that gives 't Hooft-Polyakov monopoles their topological charge.
This nontrivial bundle poses no problem for the existence of the monopole. H can be defined on northern and southern hemispheres of the S 2 and these hemispheres are related by a gauge transformation which corresponds to the generator of π 1 (H). However when H is nonAbelian, this gauge transformation acts nontrivially on any set of generators of H and so implies that no set of generators of H can be extended over the S 2 . This means that no global H symmetry exists. Colored dyons would be charged under such a global H symmetry, and so the topological obstruction to a global definition of H has the very physical consequence that no colored dyons exist.
A similar obstruction can occur in the case of vortices, as was discovered in Ref. [23] which introduced the Alice string. Again a symmetry group G is broken to H. In this case H = U(1). The vortex is linked by a circle S 1 and so the high energy theory is described by a trivial G bundle on S 1 and the low energy theory by an H bundle on S 1 . Again the H bundle is nontrivial. As the base space of the bundle is just a circle, the bundle can be trivialized on θ ∈ (0, 2π) and so it is entirely characterized by the transition function when passing θ = 2π.
The Alice string is particularly exotic because a particle encircling the Alice string negates its electric charge. Whenever particles encircling a vortex change the representation of H under which they transform, the H bundle is not principle. Nonprinciple U(1) bundles have transition functions in Iso(U(1)), the group of isometries of U(1). So each U(1) bundle corresponds to an element of Iso(U(1)). This is essentially an infinite dihedral group, it consists of multiplication by elements in U(1) and the inversion of an element in U(1). Group multiplication by a fixed element of U(1) yields the AB effect. The total space of the U(1) bundle is a torus and the multiplication by a fixed element simply means that the modulus of this torus is not purely imaginary, nonetheless the bundle is trivial. On the other hand, the Alice string corresponds to the bundle in which the U(1) fiber is inverted, corresponding to the negation of electric charge, when circumnavigating the vortex. The total space of the bundle is the Klein bottle, it is not homeomorphic to the torus and the bundle is not trivial. In particular, H cannot be globally defined, since the generator of the Lie algebra of H changes sign when circumnavigating the circle.

How can we classify such topological obstructions to the global definition of H?
The obstruction was a consequence of the nontrivial topology of the H bundle over S 1 and indeed an obstruction implies a nontrivial H bundle although the converse is not necessarily true. These bundles are classified geometrically by an element of Iso(H), the choice of monodromy when winding around the vortex. In the case of gauged nonAbelian vortices, no particles change representations when encircling the vortex so the only elements of Iso(H) which appear as monodromies are multiplication by elements of H. Therefore our bundles over S 1 are in one to one correspondence with elements of H. Similarly in the case of monopoles, nontrivial H bundles over S 2 are classified by π 1 (H), although, as the original 't Hooft-Polyakov monopole case illustrates, a nontrivial bundle is not sufficient for a topological obstruction to the existence of a dyon, it is also necessary that the transition functions act nontrivially on the generators of the Lie algebra of H.

Topological obstruction in our case?
To determine whether or not there is a topological obstruction in our case, we will first need a more careful global definition of G and H, paying particular attention to torsion elements, which are often responsible for topological obstructions. There are three kinds of gauge fields, the photon carrying U(1) 0 and the gluons carrying the left and right SU(N). Thus the gauge group could in principle be as large as U(1) × SU(N) l × SU(N) r . However the gauge symmetries which leave all of the fields invariant have no physical effect and so, to avoid confusion, should be quotiented out of the total gauge group. There are four kinds of fields. The three gauge fields and the scalar field Q. All three gauge fields leave U(1) 0 invariant. The left gauge fields SU(N) l are invariant under U(1) 0 × SU(N) r and also under their own center Z l N ⊂ SU(N) l . Similarly the right gauge fields are invariant under U(1) 0 × SU(N) l × Z r N . If these were the only fields, the total gauge symmetry would be SU(N) r /Z r N × SU(N) l /Z l N . However the scalar field Q is charge one under U(1) 0 , transforms in the fundamental of SU(N) l and the antifundamental of SU(N) r . This means that SU(N) l × SU(N) r acts freely on Q except for the central Z N 's. If one acts on Q with e ij/N ∈ SU(N) l and e ik/N ∈ SU(N) r then the total effect will be to multiply Q by e i(j−k)/N . Therefore one must quotient the gauge symmetry by the Z N diagonal symmetry of Z l N × Z r N , in other words by the elements for which j = k. Including U(1) 0 makes the story slightly more complicated. Again U(1) 0 generally acts freely on Q except for the Nth roots of unity. If one acts on Q by e im/N ∈ U(1) 0 as well as the central elements of SU(N) l × SU(N) r as above, then Therefore for any pair (j, m) ∈ Z 2 N there exists a k, given by (3.2), such that the corresponding gauge action leaves Q invariant. In conclusion, the total gauge group G needs to be quotiented by this unphysical Z 2 N , which leaves all of the fields in the theory invariant. We then conclude If in addition one includes matter transforming in the fundamental of SU(N) l or SU(N) r then only a single Z N should be quotiented, while with the inclusion of both there would be no quotient at all.
What about H? This is the symmetry group left invariant by a vacuum value of Q. This is not invariant under any rotation, so the U(1) 0 is eliminated. Furthermore, as Q is proportional to the identity, the SU(N) l and SU(N) r elements must be equal, leaving a single SU(N) l+r . The central Z N ⊂ SU(N) l+r consists of transformations such that j = k and m = 0 so they lie in the Z 2 N denominator in G and so need to be quotiented out, leaving the projective unitary group Note that if matter transforming under only SU(N) l and/or SU(N) r is included then this Z N is no longer quotiented out of G and so H = SU(N) l+r . In either case, so long as all fields have integral U(1) 0 charge, H is path connected and so π 0 (H) = 0. Therefore no H bundle over the circle can be nontrivial, so there is no topological obstruction to globally defining H.
In fact, it is not difficult to explicitly construct elements of H at arbitrary θ. H is the symmetry group at large r, where Eq. (2.10) reduces to This transformation preserves Q because Q times the first factor in the right hand side of (3.6) is proportional to the identity.
Summarizing, we have observed that π 0 (H) = 0 implies that all H gauge bundles over a circle are trivial. Therefore ours must be trivial and so no topological obstruction can exist to a global definition of H. However if we allow matter with nonintegral U(1) 0 charges then G and H will change, becoming covers of the definitions above. In such a case π 0 (H) is no longer necessarily trivial. Nonetheless as U(1) 0 is abelian we do not expect this to lead to a topological obstruction in the definition of H.
What does this all mean physically? In the monopole case, the fact that the 't Hooft-Polyakov monopole has no obstruction to a global definition of H but its nonAbelian generalization does implies that the former can be modified to carry the electric charge but the latter cannot. In the present case, the lack of a topological obstruction to the definition of H means that the color electric charge of a perturbed vortex solution is well defined. This color electric charge can be calculated by simply integrating the current multiplied by the H generators described in (3.6) and traced.

Geometric obstruction
We have seen that there is no topological obstruction to the global definition of H at all azimuthal angles θ. However the nonAbelian Aharonov-Bohm effect yields a closely related phenomenon: a geometric obstruction. It is not possible to globally construct generators of H which are covariantly constant with respect to the gauge field.
Recall that the scalar fields in the regular gauge have a nontrivial winding at spatial infinity where the rotated SU(N) generator is given by T (U ) A covariantly constant embedding of the unbroken symmetry group H -the little group of Q (θ) -inside the original symmetry group G becomes θ dependent, and as a result some of the generators are not globally defined. 4 We also recall that ζ l ≡ Let us rewrite (3.7) as where u(θ) can be read off from (3.7) for the various simple factors in G: This rewriting, equation (3.8), is more adequate for a discussion valid in general gauge In such systems, in order for the energy |D i Q| 2 to be finite, the gauge field must approach where the integral is computed along a circle at radial infinity.
Let T a be a basis of generators for Lie(G). Define the Lie algebra automorphism A covariantly constant embedding of the unbroken symmetry group H inside G varies with θ: In general, after a full circuit where O must be a real orthogonal matrix to preserve Hermiticity and normalization of the generators with respect to the invariant inner product (T a , T b ) := Tr(T a T b ). By a basis change in Lie(H) O can always be diagonalized (every orthogonal matrix is unitarily diagonalizable over C). The diagonal basis will involve in general complex linear combinations of the original generators. However, since O is real its eigenvalues come in conjugate pairs λ and λ * and corresponding eigenvectors v and v † , i.e., v (3.14) The covariantly constant generators for which ξ a = 0 are globally well-defined and generate the unbroken symmetry group H ⊂ H. The generators for which ξ a = 0 are not.

Zeromode excitations
Let us now allow the vortex orientation zeromodes {B i } to fluctuate. If {B i } are allowed to depend only on {z, t} the field equations (2.8)-(2.9) containing the x and y derivatives and the corresponding gauge fields components are unmodified. The other field equations however are modified, and in order to describe the zero mode excitations one must take into account the correct response of the gauge fields to the {z, t} modulation of the vortex orientation. The new gauge components, induced by the the nonAbelian Gauss and Biot-Savart effects, satisfy the following equations of motion: and It turns out that the form of the new gauge field components consistent with these equations is given by the following Ansatz where The profile functions ρ(r, θ) and η(r, θ), as we are going to see next, satisfy the equations that follow from the insertion of the Ansatz into equations (4.1) and (4.2).
For the standard, ungauged, nonAbelian vortices (with g r = 0 or g l = 0), it was convenient to develop the effective action using the vortex solution in the singular gauge, the one in which the scalar fields do not wind [1]. For the gauge nonAbelian vortex instead the use of the singular gauge is somewhat subtle, as the gauge fields in such a gauge develops Dirac sheet singularities [16]. Below we shall work in the regular gauge, Eq. (2.10) -Eq. (2.13). We shall also see that result obtained for the ungauged vortex in the singular gauge is correctly reproduced.
The price that we pay for using the regular gauge is that the equations of motion take a more complicated form as the background and quantum fields depend upon the azimuthal angle. The color structure of the gauge fields (4.4) is also richer than in the case of the standard nonAbelian vortices, where the only color structure (in the singular gauge) was where is the Delduc-Valent projection [18] on the Nambu-Goldstone direction in the tangent space. The form (4.6) means that the massless excitations correspond to the Nambu-Goldstone modes, whose direction must be kept orthogonal to the "rotating background", UQU † [5]. The more general form of the gauge field Ansatz (4.4) contains an additional term that arises from the gauge transformation to the regular gauge (see Eq. (4.18) below). The strongest justification for the Ansatz, however, comes from the fact it allows one to resolve the color structure of the equations in a closed form (4.1), (4.2).
To lowest order the excitations above the static, constant tension vortex are described by the effective action We neglect here the terms coming from F (l/r)2 αβ which would generate higher derivative terms in the effective action; we shall come back later to discuss the validity of this approximation. By inserting the Ansatz, a straightforward calculation [24] yields where (σ l ≡ 1 + 2g l ρ l , σ r ≡ 1 + 2g r ρ r ) and dzdt is the standard È N −1 sigma model action. The equations for the profile functions ρ and η can be determined by minimizing I. Alternatively they can be derived directly from the equations of motion (4.1) and (4.2).
Note that, in spite of the fact that terms with two different color structures W α = i∂ α T (U ) T (U ) and V α = ∂ α T (U ) appear in various contributions, the total action is simply proportional to Tr(∂ α T (U ) ) 2 , due to the following identities:

Equations for ρ and η
By minimizing I with respect to ρ r , η r , ρ l and η l , given the other functions Q 1,2 , f 0 , f N A fixed, one finds the following four coupled equations (4.14) It can be verified that the same equations follow directly from the Gauss equations, after factoring out the common color structures W α and V α .

Global (g r = 0) nonAbelian vortex
A nontrivial check of the equations found above is provided by the consideration of the ungauged vortex case. After setting the right gauge coupling to zero g r = 0 and η r = ρ r = 0, equations (4.13) reduce to (g l ≡ g, η l ≡ η and ρ l ≡ ρ) (4.15) In order to compare them to the equations studied earlier [1]- [5], it is necessary to gauge transform form the singular to the regular gauge. In the singular gauge the gauge field A (s) α is given by The gauge transformation from the singular to regular gauge is achieved by That is: where σ (s) ≡ 1 + 2gρ (s) . Apart from an irrelevant U 0 (1) transformation e iθ/N , (4.17) is just a gauge transformation which winds the scalar fields once. A useful relation is (4.20) The profile functions are accordingly transformed as (sin θ σ (s) ) . At this point it is a simple matter to verify that our equations in the regular gauge give the same profile function (for the g r = 0 theory) known from the earlier studies. We write the equations (4.15) in terms of ψ as a single complex equation: By substituting (4.24) into this equation one gets, after some simple algebra, This is precisely the equation for the profile function ψ (s) = σ (s) = 1 + 2gρ (s) found earlier [1]- [5] in the singular gauge, whose solution is given by as can be shown by using the BPS equations for Q 1 , Q 2 , f and f N A . The result for ψ in the regular gauge is then (4.28)

Solution of Gauss's Equations and Vortex Excitation Energy
To study the solutions of equations (4.13) and (4.14) and to obtain the associated excitation energy, it is convenient to use the complex combination of the profile functions (ρ, η) already introduced in Section 4.2, this time both for the left and right fields: Then equations (4.13) and (4.14) reduce to two complex equations

Partial-waves and asymptote
These equations can be solved via the partial wave decomposition: Taking the difference of (5.5) and (5.6), one can prove that φ m l → φ m r exponentially fast at spatial infinity. The asymptotic form of equation ( Similarly, the asymptotic form of equation (5.6) for φ m r is which is the same as equation (5.8) since ζ l + ζ r = 1.
Near the vortex core (r = 0), where Q 1 → 0 and A → 0, Eq. (5.5) behaves as This equation has two independent solutions: Similarly, from (5.6): For the m = 0 wave, the solution behaving as log r at the origin is excluded by the regularity requirement, and so are the negative-power solutions for m = 0.

Exact solution
Equations (5.2) and (5.3) can be solved exactly once the vortex profile functions are known. Define the function ϕ = ϕ(r) as follows: ϕ is regular at r = 0 and behaves as ϕ ≃ log r at large r. With f, f N A , Q 1 , Q 2 defined in Eq.(2.17)-(2.19), ϕ(r) satisfies the differential equations The properties of the function ϕ have been studied in detail in [25].
By using the first equation of Eq.(5.18), Eq.(5.2) and Eq.(5.3) can be rewritten as Note the following Z 2 symmetry (ψ l , ζ l ) ←→ (ψ * r , ζ r ) . (5.20) In order to simplify the equations we set ψ l = e −ζ l ϕ ψ l , ψ r = e ζrϕ ψ r , (5.21) and we use a complex formulation. Eqs. (5.19) then become It can now be easily seen that the holomorphic functions ψ l , ψ r that satisfy ∂z ψ l = 0 , ∂z ψ r = 0 , and ψ l = z ψ r (5.23) solve (5.22) (∆ ∝ ∂ z ∂z). Using the Z 2 symmetry, we find another set of solutions. The general solution to Eq.(5.2) and Eq.(5.3) is therefore given by the linear combinations ψ l = e −ζ l ϕ z χ r (z) + e ζ l ϕχ l (z) , ψ r = e ζrϕ χ r (z) + e −ζrϕzχ l (z) , (5.24) in terms of two arbitrary holomorphic functions χ r (z) and χ l (z). At large r, they behave as The solution of the minimum excitation energy (m = 1 wave for g r < g l , see the next subsection) corresponds to the particular choice above, χ r (z) = const. ;χ l (z) = 0 . The holomorphic and antiholomorphic terms correspond to the positive and negative angular momenta respectively in the partial wave decomposition, (5.4).

Divergences of the energy
To study the excitation energy we consider the integral I in (4.10) which appears in front of the È N −1 action in (4.9). In terms of ψ l and ψ r it simplifies considerably: First of all, we note that this expression is positive semidefinite: (5.32) (using |z| ≥ Re(z), Im(z)). As the integrand of I is homogeneous in ψ l and ψ r , the absolute minimum of the integral is given by I = 0, at ψ l = ψ r = 0. Going back to our Ansatz (4.4), we see that such minimum corresponds to constant profile functions ρ l = −1/2g l , ρ r = −1/2g r and η l = η r = 0, and thus, from Eqs. (4.4)-(4.5), one sees that is a pure SU(N) l+r gauge form, In addition, as it should. Our interest is in excitations of the above static vortex configuration, with T = 2πξ. Therefore we shall assume ψ l = 0, ψ r = 0 in what follows.
Inserting the partial wave expansion (5.4), I becomes the sum of different angular momentum excitations: For large r this becomes The contribution to the integral from a single mode where R is an infrared cutoff. As the contributions to the energy of the various modes are decoupled, the excitation of minimum energy has m corresponding to the minimum value of s.
For g r < g l we have and thus the mode with least energy is the one with m = 1 for which the divergence is The solution is given by ψ l = φ l e iθ , ψ r = φ r with φ l and φ r (dropping the index m = 1) satisfying the system In concluding that the energy diverges as (5.39), we have assumed implicitly (5.38), that is that the φ m l,r has the growing power law behavior. This does not necessarily follow from the general result (5.13)- (5.15). This can however be shown as follows. Multiply equation (5.5) by φ m l and (5.6) by φ m r , divide the first equation by g 2 l and the second by g 2 r and sum the two equations. Integrating this over the xy plane, one gets (5.42) Namely, we have reproduced the well known result that any action quadratic in the fields with the standard kinetic term, computed at its minimum, is a total derivative and thus is determined by the boundary values of the fields only. Since the contributions from the lower limit vanish on the left hand side (φ m l , φ m r must be regular) and as the right hand side is positive definite, φ m l ∼ φ m r cannot vanish at r = ∞.
We take the parameters normalized to ξ = g ′ 1. The minimum excitation energy (m = 1 wave for g r < g l ) corresponds to the particular choice χ r (z) = const. ,χ l (z) = 0 ,  in Eq. (5.24). The exact solution of (5.41) is then where γ is a constant. Note that in the limit g r → 0 (ζ r → 0, ζ l → 1) this minimumenergy solution smoothly approaches the known profile function for the standard, ungauged nonAbelian vortex, brought to the regular gauge form, Eq. (4.28).
We present below some numerical solutions for the vortex profile functions and the corresponding zero modes. We consider for simplicity the case (5.27) for which the profile functions simplify to (5.28). For this set of solutions, the vortex profile functions are given in Figure 1. Also, φ l and φ r given by (5.44) are plotted for several different values of g l and g r for constant g ′ , in Figure 2. In Figure 3 the planar density in I is plotted. The integral I can be expressed by using (5.42), up to a certain infrared cutoff R, as This is plotted in Figure 4 as a function of R for the particular solution (5.44). For R ≫ 1/g ′ √ ξ it can be approximated, up to exponential vanishing terms, by For g r > g l instead, ζ l < 1/2, so the excitation energy is minimum for the m = 0 mode, for which φ l ∼ const. and φ r ∼ r near r → 0, and φ l ∼ φ r ∼ r ζ l at r → ∞. One has also in this case I ∼ R 2ζr , (5.47) which is the same as in (5.40). The solution is ψ l = φ l and ψ r = φ r e −iθ with φ l and φ r satisfying

Infrared cutoff
We have introduced above an infrared cutoff R in the (x, y) plane, to regularize the integral I in (5.45). The divergences arise due to the presence of the massless SU(N) l+r gauge fields in the bulk. There are certain limitations to the parameter space where the effective action in (4.8)-(4.10) can be applied. First of all we neglected in (4.8) the terms coming from F l/r αβ and we have to check when this approximation is reliable. The extra term that we are neglecting in the action is TrF l/r αβ This is a higher derivative correction to the effective action, it is thus proportional to 1/λ 4 where λ is the typical wavelength of the fluctuation of the 2D sigma model we are considering. In order for this terms to be negligible with respect to the two-derivative term the effective action (4.9) we need (neglecting some irrelevant coefficients) This is thus a low-derivative condition, but related to the choice of the cutoff R: Another restriction comes from the fact that the SU(N) unbroken gauge theory is asymptotically free (unless, e.g., one introduces many fermions) and becomes strongly coupled at large distances. Our lowest-order calculations are thus valid only for wavelengths much less than the confinement length, 1/Λ 4D . At the same time, by definition of the vortex effective action, one is calculating the fluctuations at length scales much larger than the vortex size, 1/g ′ √ ξ. To summarize, one must assume for the validity of our analysis.

Origin of the non-integer power divergences
In this section it will be shown that the non-integer power divergences just found are intimately related to the geometric obstruction discussed in Section 3. Our theory has vortex solutions as a consequence of the bulk symmetry breaking G → H, π 1 (G/H) = 1. When H is nonAbelian, a given vortex configuration further breaks H into H ⊂ H: the action of the unbroken generators in H/ H generates the internal zeromodes.
We recall also that the presence of the vortex makes the embedding H ⊂ G positiondependent in the regular gauge. As a consequence, in a covariantly constant basis some of the generators become multivalued. Multivalued symmetries give rise to Aharonov-Bohm scattering of gauge bosons at large distances from the vortex. The corresponding zero modes lead to nontrivial effects, like cosmic string color superconductivity. They also explain the precise nature of the non-integr, powerlike divergences encountered.
Let us summarize these features in a model independent fashion following the reasoning of Alford et al. [9], [10]. Consider a generic Lagrangian which describes a gauge theory with gauge group G coupled to a scalar field Q, with D µ = ∂ µ − igA µ and A µ = A a µ T a , where T a are the generators of G in the Q representation, normalized as Tr(T a T b ) = δ ab /2. The scalar potential is chosen so that Q condenses, breaking the full gauge group G to a subgroup H. For a nontrivial π 1 (G/H) vortex configurations exist. The elementary vortex is oriented in some fixed direction in the Lie algebra that we denote as S. One may take the following Ansatz where α = z, t. Q 0 approaches a uniform vacuum configuration at spatial infinity and u(θ) describes the asymptotic winding of the scalar condensate, The Ansatz for the gauge fields is determined by the requirement of finiteness of the energy. In particular, a necessary condition is that |D i Q| → 0 as r → ∞, which implies A r → 0 and A θ → (1/g) S. Therefore, the boundary conditions on the profile function a θ are a θ → 1/g as r → ∞ and a θ → 0 as r → 0. Since the vacuum of the theory leaves an unbroken symmetry group, which however can act nontrivially on our Ansatz, the vortex carries orientational moduli, describing its embedding in the Lie algebra. Through a global gauge transformation U of H, one obtains a vortex with a generic orientation, where S (U ) = USU † . In the spirit of the moduli space approximation, U is taken to depend on the string worldsheet coordinates t and z. Then, As soon as U is taken to fluctuate along the string, nontrivial A α fields are induced, whose precise form is dictated by the equations of motion. Their behavior at spatial infinity, however, is fixed by the requirement |D α Q| → 0, which implies that A α approaches (the U transform of) the gauge fields belonging to H ⊂ G left unbroken by the vacuum configuration u(θ)Q 0 (∞).
We may now gauge transform by U † (t, z), going to what we shall call the static gauge. This gauge choice makes computation somewhat simpler. The new Ansatz is where A i = a i (r)S. Passing to the static gauge allowed us to eliminate the ∂ α A i piece. The equations of motion following from the Lagrangian density (6.1) are We focus now on the second of equations (6.8), which is the nonAbelian Gauss law. Substituting it into our Ansatz, the left hand side becomes This rewriting also follows from the fact that in static gauge F iα = D i A α and where the covariant derivatives in r reduce to the ordinary derivatives since A r = 0. At large r the right hand side of Gauss's equation vanishes as the covariant derivative of the scalar field approaches zero. One obtains thus a Laplace-type equation One may expand A α in eigenstates of the operator ∆ θ = ∂ θ − i[S, ·], which at spatial infinity commutes with ∆ r , ∆ θ ψ s α = isψ s α , (6.12) with s being a real contant. The eigenfunction in (6.12) can be written as using the fact that (∂ θ − i[S, ·])e iθS Me −iθS = 0 , (6.14) for any θ independent matrix M. 6 If a parallel transport by u(θ) describes the embedding of H inside G, we see that if A α at θ = 0 is oriented along some generator S a , at θ = 0 it is oriented along the rotated generator S a (θ).
One can now see the consequences of the requirement of single valuedness of ψ s α . If S a is covariantly constant and belongs to the globally defined subgroup H ⊂ H, then u(2π)S a u(2π) † = S a , (6.15) and the single-valuedness condition for ψ s α implies that s ∈ Z. If, on the contrary, S a is not globally defined, u(2π)S a u(2π) † = e 2πiξ a S a , (6.16) with some non integer constant ξ a . The single valuedness of ψ s α then implies that Using (6.13) in (6.11), one finds the asymptotic behavior, The energy grows like R 2|s| . Allowing for a nonvanishing angular momenta m ∈ Z in (6.9), (6.11), one gets in general Such a general argument, however, carries us only this far. In particular, the determination of the generators along which the fields A α are excited can only be done by studying the nonAbelian Gauss equation for all r and requiring that the color matrix structures of the left and right hand sides match precisely, as has been done in Sections 4 and 5. Related to this, the non-integr power s cannot be determined by a general argument based on the asymptotic behaviors of the fields alone. Both of these features carry information about the vortex configuration at finite r.
In our specific U 0 (1) × SU l (N) × SU r (N) theory, it was found that the gauge fields excited by the vortex modulation lie in the directions Knowing this, and knowing how these gauge fields are parallel-transported around the vortex (see (3.9)), (similarly for A (r) α ), one finds that upon encircling the vortex W α V α → cos(2πζ l ) sin(2πζ l ) − sin(2πζ l ) cos(2πζ l ) where the commutation relations have been used. The eigenvalues of the monodromy matrix (6.23) are thus e −2πiζ l and e 2πiζ l , with respective eigenvectors v = ∂ α T (U ) T (U ) + ∂ α T (U ) and its Hermitian conjugate v † . Thus in our model ξ a in equation (3.14) or (6.16) is given concretely by Our results in Section 5 indeed agree with the general expectations (6.19) with this specific index.
These discussions clearly illustrate the connection between the global features of the system discussed in Section 3 and the dynamical properties of the vortex excitation explored in Section 5. In particular, let us note that for the standard nonAbelian vortex (g r = 0 or g l = 0) ξ a reduces to an integer. Thus no geometric obstruction, no nonAbelian AB effects, no non-integr-power growth of vortex excitation energy, etc., occur. The 2D È N −1 dynamics becomes fully operative, and at the same time the massless, free SU r (N) (or SU l (N)) Yang-Mills system in 4D simply decouples from the vortex system.

Discussion
In this paper we discussed in some detail the topological and dynamical features of an interacting 4D-2D coupled quantum field theory, in the context of the underlying U 0 (1) × SU(N) × SU(N) gauge theory in an SU(N) l+r symmetric ("color-flavor" locked) vacuum. Of course, such a system can be considered ab initio, just by starting with a given 2D system possessing some global symmetry, and making it local, by introducing 4D gauge fields coupled to it [27,28], i.e., embedding the 2D system in 4D spacetime. Another venue, which we have opted to follow here, is to study such a 4D-2D coupled quantum field theory which emerges as a low-energy effective description of the vortex sector of the underlying 4D system. Here the 2D variables are the fluctuations of the vortex collective coordinates, whereas the 4D degrees of freedom are part of the original Yang-Mills fields, the coupling among them being fixed by the local gauge symmetries of the underlying theory. This is a new type of interacting quantum field theory, in which field variables living in different dimensions interact nontrivially. Our way of approaching the problem gives an existence proof that such a theory can be consistently defined, as the underlying 4D system is a consistent local gauge theory.
In a standard nonAbelian vortex, the modulation of the vortex internal orientation is described by a 2D sigma model ( È N −1 in the case of U 0 (1) × SU(N) theory) on the vortex worldsheet. The È N −1 model in two dimensions is asymptotically free and becomes strongly coupled at low energies Λ ≪ √ ξ. The renormalization group flow of two dimensional model "carries on" the evolution of the 4D gauge coupling which was frozen at the vortex mass scale √ ξ, into the infrared, showing a remarkable realization of 2D-4D duality [2,4,26].
When the orientational mode dynamics becomes strongly coupled, the orientation {B i } of the vortex in the Lie algebra fluctuates wildly, and the AB effects are washed away. This is consistent with the fact that there is no spontaneous symmetry breaking and no Goldstone bosons in 2D: the spectrum has a mass gap.
As soon as g r is turned on, however, the physics changes immediately, and drastically. The presence of massless fields in the bulk makes the coefficient in front of the effective È N −1 action divergent and the vortex fluctuations become nonnormalizable.
They become infinitely costly to excite. At the same time, the AB effects and all sorts of topological and nonlocal effects emerge.
One wonders however how such a discontinuous change of physics is possible at g r → 0. As in the phenomena of phase transitions, physics cannot really exhibit such a qualitative change at g r → 0, if the spacetime is finite 7 . The expression (5.40) indeed shows that the limits g r → 0 and R → ∞ do not commute, pointing clearly to the origin of the discontinuity. Stated differently, it is because one is accustomed to thinking about infinitely extended spacetime in relativistic-field-theory applications that physics looks discontinuous at g r = 0.
Let us look into the nature of this discontinuity more carefully. At large r the divergent effects such as Eq. (5.38)-Eq. (5.40) come from the dominant, massless SU l+r (N) gauge field excitations (the massive fields in the bulk Higgs mechanism give only exponentially suppressed contributions). They are nonetheless the result of the interactions with the 2D vortex collective coordinates, the fluctuations of {B i }, and in fact, the non-integr power and the precise direction in color space in which the massless gauge fields are excited, both reflect the details of the vortex configurations near the vortex core.
What happens in the g r → 0 limit is that most of these divergent effects smoothly transit into the continuum spectrum of the free SU r (N) Yang-Mills system, now decoupled from the vortex. The fact that they appear in the coefficient in front of the vortex effective action = I dz dt Tr(∂ α T (U ) ) 2 (7.1) does not necessarily mean that they are related to the vortex physics.
Summarizing, if one first takes the IR cutoff R → ∞, the screening of the massive gauge fields implies that the massless gauge fields dominate, leading to the power law scalings. On the other hand, if one takes the limit g r → 0 first, for the vortex effective action of minimum energy cost, i.e., the contribution of the partial m = 1 wave, one finds from (5.46) that lim gr→0 I(R) = π 2g 2 l + O e −g l √ ξR .

(7.2)
Taking then the R → ∞ limit we recover the physics of the standard nonAbelian vortices with a finite È N −1 coupling Taking the limits in this order, the only massless degrees of freedom that remains are the È N −1 orientational modes in the vortex worldsheet.
The renormalization of the È N −1 model, for the ungauged nonAbelian vortex, is given by 8πI(µ) = log µ Λ 2D (7.4) namely I is decreasing logarithmically as the RG scale µ decreases. For the gauged nonAbelian vortex, we find another effect: I is increasing as the cutoff scale 1/R gets smaller, this time as a power law (5.46), due to the interaction with the massless unbroken SU(N). We may thus conjecture the È N −1 model coupled to the 4D massless gauge fields will remain frozen in the Higgs phase, at least as long as the massless SU(N) field in 4D remains in the Coulomb phase. Note that Coleman's theorem on the absence of the Nambu-Goldstone modes in two-dimensional spacetime would not apply here due to the interaction of the 4D massless gauge fields. This would mean that the global and topological effects related to the vortex orientation {B i } would not be washed away by the 2D quantum effects.
To compute the effective action in this paper, we have integrated out all gauge fields, including the massless gauge fields in the bulk. This is formally not quite consistent from the renormalization group point of view. One should integrate away the massive gauge fields first, and obtain the effective action defined at the vortex mass scale √ ξ, which should play the role of an UV cutoff for the study of further quantum effects. The effective action at Λ U V = √ ξ would consist of two pieces: the full 4D action for the massless gauge fields living in the bulk and the 2D action describing the fluctuations of the internal orientation, {B i }, which is a sort of Nambu-Goldstone bosons living in the vortex worldsheet. Their coupling is dictated by the full H ⊂ G gauge invariance. By integrating out the coupled massless 4D gauge fields and the 2D fields {B i } simultaneously down to some infrared cutoff scale µ, and varying it, one would find an appropriate RG flow towards the infrared.
Although such a renormalization group program is still to be properly set up and be worked out, we believe that the number of results established and the subtle issues of decoupling / transitions to the g r → 0 limit clarified here, should provide us with a useful starting point for such a task.
The crucial difference is the last term in (A.12), (A.13). At r → 0, (A.13) gives ∆φ ∼ 1 r 2 φ , (A.14) whose regular solution behaves as φ ∼ r. At r → ∞, instead, ∆φ = µ 2 (φ − 1) , (A.15) which implies φ ∼ 1 + c e ±µr √ r . From the exact solution of (A.13) we know that actually φ behaves as The procedure which has led to (A.11) before yields this time Even though the LHS vanishes in this case also, the integrand of the right hand side is no longer positive definite, thus no contradiction arises.
We conclude that in the g r = 0 theory, the only physical mode in Eq. (A.6) is the m = 1 wave. This is to be contrasted with the powerlike divergent modes for the general gauged vortex, Eqs. (5.40), Eq. (5.47). Even though these latter solutions are nonnormalizable also, they represent the continuous spectrum of the theory and as such are to be regarded as physical excitation modes.

B Gauge fixing for ψ l and ψ r
We have seen in the main text that the Ansatzes (4.4)-(4.5), solve the Gauss equation, given the vortex configuration