Professional Documents
Culture Documents
R i c h a r d A. B R A N D T a n d Filippo N E R I
Department o f Physics, New York University, New York, New York 10003
The classical electrodynamics of electrically charged point particles has been generalized
to include non-Abelian gauge groups and to include magnetically charged point particles.
In this paper these two distinct generalizations are unified into a non-Abelian gauge theory
of electric and magnetic charge. Just as the electrically charged particles constitute the gen-
eralized source of the gauge fields, the magnetically charged particles constitute the gener-
alized source of the dual fields. The resultant equations of motion are invariant to the ori-
ginal "electric" non-Abelian gauge group, but, because of the absence of a corresponding
"magnetic" gauge group, there is no "duality" symmetry between electric and magnetic
quantities. However, for a class of solutions to these equations, which includes all known
point electric and magnetic monopole constructions, there is shown to exist an equivalent
description based on a magnetic, rather than electric, gauge group. The gauge potentials in
general are singular on strings extending from the particle position to infinity, but it is
shown that the observables are without string singularities, and that the theory is Lorentz
invariant, provided a charge quantization condition is satisfied. This condition, deduced
from a stability analysis, is necessary for the consistency of the classical non-Abelian theory,
in contrast to the Abelian case, where such a condition is necessary only for the consis-
tency of the quantum theory. It is also shown that in the classical theory the strings can-
not be removed by gauge transformations, as they st etirnes can be in the quantum theory.
An action principle is presented which leads to the particle and field equations of motion
except for extra contributions arising from the possibility that the strings carry electric
charge. Only solutions with electrically neutral strings are acceptable as monopoles. The
action includes unphysical electrically charged fields defined on particle trajectories and
on strings. The Lagrangian is gauge invariant and leads to electric current conservation via
Noether's theorem.
1. Introduction
* Resea==h supported in part by the National Science Foundation under Grant No. PHY74-
22218A01.
221
222 R.A. Brandt, F. Neri / Classical Lagrangian formalism
unify these two generalizations into a non-Abelian gauge theory of electric and mag-
netic charge. There have indeed been a number of interesting attempts to accomplish
this [4,5]. Most notable are the explicit constructions of classical non-Abelian gauge
fields which describe point magnetic monopoles - a by-product of the more popular
constructions of non-singular monopoles by 't Hooft [6] and Polyakov [7] and fol-
lowers [8]. Wu and Yang [9] have developed a very convenient non-singular topologi-
cal formalism for describing these singular monopoles in which space-time outside of
monopole world lines is divided into overlapping regions each carrying its own gauge
field. However, there has yet to be presented a complete and satisfactory Lagrangian
formalism for non-Abelian point monopoles or even a complete set of (particle and
field) equations of motion (valid in all space-time) to which the above mentioned
explicit constructions are solutions. Our purpose in this paper is to supply such a for-
malism and investigate some of its implications.
In the context of our formalism, we can ask some interesting questions. (i) Are
the explicit constructions in fact solutions of the field equations everywhere? The
answer is yes, but these solutions are all essentially Abelia_n (in the sense that there
exists a gauge in which the gauge covariant cross product A X A of the gauge fields
vanishes). (ii) Is the theory Lorentz invariant? The answer is yes, provided the charge
quantization condition eg = 27rn, where e is the "electric" gauge charge, g is a gener-
alized magnetic charge, and n is an integer. In the non-quantized Abelian theory, no
such strict condition is necessary for Lorentz invariance because of the absence of
the vector potential in the field equations. (iii) Are the explicit solutions stable against
small perturbations? In the case of the 0(3) Lie algebra, the answer is yes for n = 1
and no for n > 1. (iv) If the monopole position is ignored, the string singularity pre-
sent in the 0(3) gauge field can be removed by a gauge transformation for even n.
Does this remain true when the monopole position is taken into account? The answer
is no. (v) Do the equations of motion have other, perhaps not essentially Abelian,
solutions? We suggest that such solutions do exist, but cannot rigorously prove this.
(vi) In the Wu-Yang formalism, the field behavior is considered only outside of the
monopole positions. (This is sufficient for quantum mechanics, but not for classical
mechanics or quantum field theory). Can their analysis be extended to all of space-
time? We will give only a partial answer to this important question.
The "magnetic" field equation which we shall adopt is simply
.
R.A. Brandt, F. Neri / Classical Lagrangian formalism 223
limit K c -~ 0 of zero magnetic current, and becomes the usual Abelian magnetic
field equation in the limit D ~ 1 of the Abelian theory. But by adopting (1.1), the
(dual) symmetry between electric and magnetic quantities present in the Abelian
theory is lost. All the fields in the proposed theory transform as representations of
the original "electric" gauge group G, which remains as the total gauge symmetry
group, and there is no magnetic group. In contrast, in the Abelian theory charged
fields ~i transform (~ki -* ei(et~+gi(3) ~i) as representations of the electric plus magne-
tic gauge group U(1) × U(1), and there is complete dual symmetry. However, in
defense of (1.1), we will show that: (i) All of the known explicit constructions are
in fact solutions of (1.1) for suitable sources. (ii) A magnetic gauge group G' does in
fact exist. Eq. (1.1) is just the electric form of the field equation. There is also an
equivalent magnetic form invariant to G'. (iii) There is present at least a partial dual
symmetry and that is enough to deduce the generalized conservation of K c. Further-
more, a complete dual symmetry is not expected [10] and probably not even consis-
tent after quantization if G is compact *.
A main result in this paper is the deduction of (1.1) plus the other field and par-
ticle equations from an action principle. Our action contains unphysical "string"
singularities familiar from the Abelian theory (only the particle trajectories z(s), the
termination points of the strings, are measurable), unphysical charged fields ¢(s)
defined on the trajectories and required even in the absence of monopoles (only the
particle charges ¢ t ~T¢ are measurable), and further unphysical fields defined on the
strings and trajectories. It should be stressed that our action in general describes sys-
tems with electrically charged, and therefore observable, strings. The true monopole
systems must satisfy a constraint equation, as will be discussed in detail.
The main purpose of the action is to enable one to proceed from the classical
field theory to the quantum field theory. Once the theory is quantized, the field
strength F, even in the Abelian case, underdescribes the physical situation so that
the potential A itself plays an important dynamical role [9] **. Since in the presence
of magnetic charge A contains unphysical string singularities, the requirement that
the strings are unobservable (so that the theory is Lorentz invariant) leads to the
Dirac quantization condition [1 ] ***
where (ei, gi) are the (electric, magnetic) charges carried by the quantum particles in
* In a non-Abelian compact theory of monopoles, we expect that there will be two quantization
conditions, one arising from the compactness and one arising from the requirement of rotatio-
nal invariance. If in addition the theory is dual invariant, these quantization conditions should
imply that the magnetic charge can be rotated away. Note that the dual invariant theory of
ref. [5] is based on a non-compact group.
** More precisely, it is the line integrals ~A • dx which are measurable.
*** For a recent discussion, see ref. [ 11 ]. If the theory is dual invariant, the condition reads
(eigi - ejgi)Pl = 27rn (see ref. [12]).
224 R.A. Brandt, F. Neri / Classical Lagrangian formalism
the Abelian case. We find that the non-Abelian generalization of this condition is
QaMa
J = 2zrn, (1.3)
where K - (21rL) - l mini, jlQaMTI for some integer L. Given (1.4) or (1.5), the string
singularities are absent from physical quantities, but these classical theories remain
unsatisfactory because of the remaining unphysical point singularities at the positions
of the charged particles. In this respect the non-singular monopoles [ 6 - 8 ] are prefer-
able since they are physical solutions of the theories with finite energy. Once the
theory is quantized, this preference is lost since renormalization renders also the point
singular monopoles to have finite energy *. In fact, since it seems impossibly difficult
to quantize the non-singular monopoles, whereas the singular monopoles are easily
quantized by standard Feynman diagrammatic methods [14,15] **, it seems that the
singular monopoles are more tractable at the present time. This has been a major moti-
vation in our work.
It is worth remarking here that we are able to deduce, beyond the expected con-
dition (1.5), a quantum-mechanical type of quantization condition already in our
classical non-Abelian monopole theory. Demanding rotational invariance of the
equations which describe the scattering of classical gauge fields off of a fixed mono-
pole of magnetic charge M a is seen to require that
egr = 27mr, (1.6)
where e is the (dimensionless) gauge field coupling constant and igr (r = 1,2, ...) are
the eigenvalues o f M a. (In the Abelian case there are no non-vanishing eigenvalues
but in the non-Abelian case there is at least one.) Although this result leads to the
quantization of magnetic charge, the parameter e is not the "electric" charge of a par-
ticle. Only after first quantization are there particles, of electric charge eo = x/~e, etc.,
associated with the gauge fields. Then (1.6)becomes a special case of (1.3).
In the Abelian case, monopoles with different n [eq. (1.2)] are always physically
distinct, but in the non-Abelian cases, monopoles with different n [eq. (1.3)] can
often be transformed into one another by a gauge transformation defined everywhere
but at the monopole positions [6,9,17,18]. We find (subsect. 5.5) that such con-
clusions are not valid classically when the monopole positions are not excluded. If
the above gauge transformations are extended to the monopole positions so as to
remain symmetry transformations, they cannot charge n, as is expected since eq. (1.3)
is gauge invariant.
In sect. 2 we will attempt to motivate the equations which describe our theory by
noting that they are natural generalizations of the non-magnetic and non-Abelian
equations. In sect. 3 we review the Abelian theories of electric and magnetic charge.
We exhibit new actions for these theories from which the constant charge conditions
(2.2e,f) arise as equations of motion and the current conservation conditions are con-
sequences of Noether's theorem. In contrast, the usual actions involve only constant
charges from the beginning and possess no non-trivial global gauge groups and there-
fore do not give rise to Noetherian currents. We proceed to formulate an action prin-
ciple for the non-Abelian gauge theory of electric charge in sect. 4. We here find it
necessary to introduce new fields q~i which are present in the equations of motion
only in the combination Qa = ~ TaXi which defines the generalized electric charge.
Sect. 5 contains the development of the non-Abelian gauge theory of electric and
magnetic charge. We begin in part 5.1 by exhibiting the equations of motion and the
implied integrability constraint condition. The action formalism is presented in part
5.2. We find that only when a second constraint condition is satisfied will the equa-
tions of motion which result from the action be the desired ones. The symmetries of
the theory are considered in part 5.3. In particular, a magnetic form of the field equa-
tions is exhibited which is complementary to the original electric form. In part 5.4
we first consider the familiar essentially Abelian static solutions to the field equations.
By using the Hamiltonian method of quantization, we deduce the quantum mechni-
cal charge quantization condition (1.3), and by considering the scattering of non-
Abelian waves off the static sources, we deduce the classical charge quantization con-
dition (1.6). We conclude in part 5.5 by noting that the singular transformation
which removes the Dirac string is not a gauge transformation of the complete classi-
cal theory.
2. Survey
In order to explain our procedure, we will summarize here our assumptions and
results. The classical Abelian (Lorentz-invariant) electric Maxwell equations are *
a .F=J, (2.1a)
0 • F a = 0, (2.1b)
• Our notation suppresses all Lorentz indices. Thus eq. (2.1a) means a/aF/av= .Iv.
226 R.A. Brandt, F. Neri / Classical Lagrangian formalism
where F u r is the electromagnetic field, Fay = 1 e u m o F ~¢ is its dual, and Ju is the elec-
tric current. To describe magnetic monopoles, these equations are generalized to
~" F = J, (2.2a)
~ " F a = K, (2.2b)
where K u is the magnetic current. These equations are symmetric to the "duality"
transformation
(F, F a) -+ ( F cos 0 + F a sin 0, F sin 0 - F a cos 0 ) , (2.3a)
(~ + e A X ) F = J (2.4a)
(0 + e A X ) F a = 0 , (2.4b)
where the gauge potential A u satisfies ?
and J is the generalized electric current multiplet t t . Here the additional symmetry
is invariance to the non-Abelian gauge transformations [3]:
-+ F - eA X F , (2.6a)
field equations (2.1) do not have to exhibit the Abelian vector potential A u, defined
by
F = ~ ^ A, (2.7)
whose existence is guaranteed by (2.1b) and which is introduced at this level only
for convenience. Also, the fundamental non-Abelian fields F are gauge covariant [eq.
(2.6a)] and therefore not observable, whereas the Abelian F is invariant under the
Abelian gauge transformation:
At * -+ Au + ~uA . (2.8a)
To combine these generalization of (2.1), it is natural to adopt the field equations
(~ + e A X ) F = J , (2.9a)
(a + eA X)-F d = K, (2.9b)
where now Ku is the generalized "magnetic" current multiplet. It might be thought
that this simplest generalization is defective on several counts. All quantities ap-
pearing in (2.9b) carry non-Abelian electric charge just as in (2.9a), according to
(2.6a-c) and the necessary (for gauge invariance) transformation law of K :
~ K - eA X K . (2.6d)
There is no "magnetic group" as there would be if charged fields were present in the
Abelian case. Correspondingly, the Abelian dual invariance under (2.3) does not
generalize to an invariance of (2.9) because of the explicit electric vector potential.
We will argue that nevertheless (2.9) define a suitable theory of non-Abelian mono-
poles and probably the only sensible such theory. It is in particular immediate that
(2.9) are Lorentz and gauge invariant and reduce to (2.4) in the no-monopole limit
and to (2.2) in the Abelian limit. We will show furthermore that all known explicit
point monopoles are solutions of (2.9) for suitable currents, that a magnetic group
and magnetic objects do exist, and that the theory is at least "partially" dual inva-
riant.
To complete the definition of the generalized theory, we must specify the vari-
ables associated with the charged particles. The particle trajectories r i are given by
functions ziu(s), i = 1 ..... N (for N charged particles), of the proper time s. In the
Abelian case, the conserved currents are
N
J(x) = ~ eif dz84(x - z) = Q ( x ) L ( x ) , (2.2c)
i=1
ri
N
K(x) = ~ gi f dz 64 (x - z) = M(x) L (x), (2.2 d)
Pi
where dz u = zu ds, k = dz/ds. The second forms involve the conserved particle num-
228 R.A. Brandt, F. Neri / Classical Lagrangian formalism
ber current
N
t(x) = f - z), (2.1o)
i=I
Fi
with (Q, M) defined only on the particle trajectories and constant on each trajectory:
~" 0Q = 0, (2.2e)
~- 0M = 0. (2.2f)
The constant values are of course
(Q(zi), M(zi)) = (el, gi) . (2.11)
where m i is the mass of the ith particle. The classical Maxwell-Lorentz force law is
specified by
fir(z) = Q(z) F ( z ) , (2.1c)
and if monopoles are also present, this becomes [2]
fir = Q F + M F a , (2.2g)
whereas the non-Abelian generalization of (2.1c) is
fir = 6 " F . (2.4c)
The combined generalization is
9: = Q - F + ~1-F a . (2.9g)
* The electric equations seem to have been given first by Wong [19].
R.A. Brandt, F. Neri / Classical Lagrangianformalism 229
Eqs. (2.9a-g) will be valid in our non-Abelian classical gauge theory of electric and
magnetic charge.
The derivation of the Abelian electric Maxwell-Lorentz equations (2.1) from an
action principle is a classical result. The action explicitly involves the unphysical vec-
tor potential A. The derivation of the Abelian electric-magnetic equations (2.2) from
an action principle was given by Dirac [2], apart from the Dirac "veto" problem
which has been recently resolved [ 13]. Here the action involves, in addition to A, un-
physical Dirac string variables associated with each monopole. The derivation of the
non-Abelian electric equations (2.4) from an action principle is given in this paper
(sub__ssect__;3.2). The action here involves a further unphysical "potential" field tp(x)
for Q : Q = ie~tT¢. The derivation of the non-Abelian electric-magnetic equations
(2.9) from an action principle is also given in this paper (subsect. 4.2) *. The action
will be seen to involve both the unphysical potential field ¢ and unphysical string
variables, and even some further unphysical objects.
where 2i(0), the trajectory of the ith particle, is parametrized by a scalar parameter o
[In sect. 2 we took o = s = proper time, in which case (3.1c) simplifies as in (2.12).
Now we leave o unspecified and write ~(o) - a/boz(o).] and the electric current,
given by (2.2c), is explicitly conserved: a- J = 0. These equations follow simply from
minimizing the well-known action
* A Lagrfingian theory of non-Abelian classical Dirae monopoles has been previously discussed
by Ezawa and Tze [ 18 ]. However, in this work, the electric charge is associated with fields
rather than with point particles and so the theory is not in general Lorentz invariant, even with
a charge quantization condition.
230 R.A. Brandt, F. Neri / ClassicalLagrangian formalism
where a new field ¢(zi) defined on the trajectories has been introduced *. This action
is invariant under the local gauge transformations
Q = eCt~ ; (3.5c)
and as usual (from 6F)
F= 3 ^A, (3.5d)
which gives (3.1b); and from 8q~, 6q~t)
* More precisely, a new field q~i(o)should be defined on each trajectory. We have defined ¢(zi(o))
q~i(a).
R.A. Brandt, F. Neri / Classical Lagrangian formalism 231
~" F = J , (3.6a)
a. F a= K , (3.6b)
~ri~.i)= [eiF(2i) + giFd(gi)] " ~ . (3.6c)
A simple action for these equations, apart from the Dirac veto problem, is *
where
F = a ^ A - Ga , (3.8a)
where
* We are using here the Schwinger-Yan [20] form of the action. It is shown in ref. [14] that this
action is equivalent to that of refs. [ 15,2 ].
232 R.A. Brandt, F. Neri / ClassicalLagrangianformalism
the action (3.7) can be altered as in ref. [13] to give (3.6c) instead of (3.9) which is
only correct if trajectories z i don't intersect strings w i (the Dirac veto). This leads to
the weak classical charge quantization conditions (1.4).
We again wish to consider a more complicated action. We define
where Ei is a semi-infinite surface bounded by the particle trajectory I"i and gener-
ated by the trajectory and strings. The position yiu(o, r) on ~,i is parametrized by
_oo < o < '~ and 0 < r < oo and we use r/to denote o or r. The field *
F a r ( y ) --yUy'UFauv (3.13b)
is the restriction o f F a to ~, and the fields X~ and Mi are only defined on ~i. [We
could define X n as the restriction ** Yv, n X v of a four-vector field Xv(y), and then
the action would look more symmetric:
The form (3.13a) is simpler to deal with and so we will not explicitly use (3.14)].
Variation of A, ~b, and ~b~" gives the previous equations (3.5) for these quantities,
and variation o f F again gives (3.8a) where now
where
Variation of XP gives
anMi(o, r) = O, or Y,n " aM(y) = O, (3.16)
so that we need only vary zi and not Yi. This variation gives
Since
O" S = L, (3.21)
and so
a " G =ML = K, (3.22)
we again obtain (3.6b), where K has the correct form (2.2d) if we solve (3.16) by
Ml( o, r) = M(zi) = gi . (3.23)
Substitution of (3.23) in 12 gives back/'l [eq. (3.7a)] and substitution into G and E
reduces them to their previous forms (3.8b) and (3,12). The force law (3.19) is
and is incorrect as usual, but again changing I as in ref. [13] gives the correct result
(3.6c). [Note that the second term in (3.24) is ~q(ei& - giei)Si, which vanishes when
ei = cgi, as it must since then the magnetic charges can be rotated away].
We thus see that I2 is a suitable action for eqs. (3.6). This remains true if a general
surface not necessarily satisfying (3.18) is used. Our purpose in introducing 12 is that
it will be seen to have a natural generalization t o t h e non-Abelian case, whereas the
simpler action/'2 has no such generalization.
234 R~I. Brandt, F. Neri / Classical Lagrangian formalism
3.3. S y m m e t r i e s
To conclude our analysis of the Abelian case, we will discuss the symmetries of the
various theories, both of the equations of motion and of the actions. We begin with
the purely electric equations of motion (3.1). These equations are of course Poincar~
invariant, but they possess no non-trivial gauge invariance since they involve only the
gauge group scalars F and z i. However, a symmetry-related aspect of the equations is
the fact that the electric current is conserved, as follows explicitly [eq. (2.2c)] or
from the anti-symmetry o f F [eq. (3.1a)]. The action/'1 [eq. (3.2)] is also Poincar~
invariant, but is in addition invariant to the non-trivial local gauge group (2.8a)
because J is explicitly conserved. The Poincar~ invariance leads to the usual associated
conserved Noetherian currents, but since the global gauge group (A = const.) is trivial,
the electric current J is not a Noetherian current in this formalism. Our second action
11 [eq. (3.3)] is more interesting in this regard. Here the local gauge group, eq. (3.4),
has a non-trivial global subgroup A = const., and the associated conserved current
(3.5b) is Noetherian. Its conservation is not explicit but depends on the equations of
motion [(3.5a) or (3.5g)]. This formalism thus more closely resembles the situation
in quantum field theory.
The symmetry analysis is more complicated in the presence of magnetic charge
because the Poincar6 invariance is less obvious and the additional symmetry of duality
invariance is present. We begin again with the equations of motion (3.6), where every-
thing remains simple. These equations are Poincar~ invariant and also invariant to the
duality transformation (2.3). There is no non-trivial gauge group, but electric and
magnetic current conservation are explicit and also follow from the antisymmetry of
F and F a.
Because of the explicit current conservation, the action I~ [eq. (3.7a)] is invariant
to the enlarged group of non-singular local gauge transformations
A ~ A + ~X, B ~ B + a~ . (3.25)
[To be precise, we must imagine here that the B-field is given by
B = B w + ab, (3.26)
of the independent variables.] For the same reason as before, the currents J and K
are not Noetherian. Poincar6 invariance is more subtle in that I~, as opposed to the
equations of motion, is not invariant to the usual Poincar6 transformations [because
~2-1Bw(E~x) = Bs2w ( x ) ~ B ( x ) for a Lorentz transformation I2]. If the quantization
condition (1.4) is satisfied, the Lorentz invariance of the theory follows from the
R,4. Brandt, F. Neri / Classical Lagrangian formalism 235
invariance of/'2 to a string change together with a singular gauge transformation [14]:
3' ~ 3", A --* A + ~X, F ~ F, Fi -* Pi , (3.28a)
(3.29)
and now the electric current is the Noetherian current corresponding to A = const.
Its conservation requires the equations of motion, (3.6a) or (2.2e). On the other
hand, the magnetic current K is not Noetherian and/2 is not dual invariant. The con-
servation of K however does follow in a general way from the global gauge invariance
of the action and the dual invariance of the equations of motion [21].
* Furthermore, since (2.3) involves parameter' rotations as well as field rotations, Noether's theory
would in any case not be relevant. An exception is the free field case where all the charge param-
eters vanish. Then the action o f ref. [ 15 ] is dual invariant and there exists a conserved current
whose charge generates (2.3a).
236 R.A. Brandt, F. Neri / ClassicalLagrangianformalism
times written a s Qi(zi)] which couple gauge-invariantly to the gauge fields. These
equations of motion, together with equations for some unphysical particle variables,
will then be deduced from an action principle.
The first equation of motion simply relates tile gauge field to the gauge potential:
= 0 ^ .~ + e~, X ,~. (4.1 a)
The second equation states that the particle currents are the generalized sources of
the gauge field:
(a + e A X ) F = ]*. (4.1b)
Here the particle current is by definition [19]
(0 + e.~X) ~a = 0 . (4.2)
This is the Bianchi identity of the theory and states that the particles do not carry
"magnetic" charge. Note that (4.1a) cannot be replaced by (4.2) in the set of equa-
tions of motion since (4.2) does not imply (4.1a). It likewise follows simply from
the structure of (4.1b)that
(0 + e,~ X) J* = O. (4.3)
This generalized "conservation" law also follows explicitly from (4.1 c) and (4.1 d).
The corresponding charge
on Fi, which imply that (4.7)satisfies(4.1 d). Finally, variation of I'i gives the force
law
a 2
m, to A + cA" Ti)g)iluv, (4.11)
238 R.A. Brandt, F. Neri / Classical Lagrangian formalism
which reduces to (4.1e) if (4.10) and (4.7) are used. Ia thus leads to all the eqs. (4.1)
and is therefore a suitable action. The scalar fields are unphysical in that only the
combination (4.7) is coupled to an applied field, and their equations of motion (4.10)
are superfluous. I 3 is a direct generalization of the action 12 used in the Abelian case
but, unlike in the Abelian case where the simple action I~ can be used, the presence
of the unphysical fields is here necessary. These fields are required for an action
formalism and for a quantum mechanical description, but give an overdescription of
the physics.
The simplest solutions of the equations of motion (4.1) are the static Coulomb-
type configurations. Provided
6i X Q/= 0 , (4.12)
the potential
N
For N = 1, Mandula [22] has shown that this solution is stable against small classical
fluctuations in the fields if e is sufficiently small.
We assume now that our classical point particles can carry both generalized electric
and magn__eeticcharges. The electric charge of the ithparticle is again characterized by
the field Qi(o), which will also be written Qi(gi) =- Q(zi), def'med on the trajectory Fi.
The physical magnetic charge of the ith particle will be similarly characterized by a
field Mi(o) or Mi(zi) =- M(zi). As in theAbelian case, it will be necessary to consider
Mi(o) as the boundary value of a field Mi(o, r) defined on the surface Zi = Fi × 7i,
where 7i is the string attached to the ith particle, and we will sometimes write this
field as ~¢li(Yi) = M(Yi).
The first field equation relates the field strengths F to the potentials A and charged
surfaces ]~i:
with
where S is defined in eq. (3.15). [cf. eqs. (3.8) and (4.1a)]. Thesecond field equation
states that the electric charges are the (generalized) sources of F:
-+
[cf. eq. (4.1 d)]. The final equations describe how the fields determine the motion of
the charged particles:
= eFo-r X M. (5.4)
No such constraint arises from (5.1 d) since that equation only holds on (one-dimen-
sional) trajectories. All of the field equations and the above relations are invariant to
A "~ U - x A U + I u - I a U _ ; (5.10a)
e
O-~U-IOU, O=F,Q,M; (5.lOb)
where _Uare the matrices in the adjoint representation of the gauge group.
5.2. Action formalism
In order to deduce (5.1e) from an a~ion principle, we are compelled to introduce
yet another unphysical field multiplet × ia(o, z) on each surface ~i as Lagrange multi-
pliers. Our action is thus
14 = l a + 1
This expression is locally gauge invariant and reduces to Ia in the limit (M = 0) of zero
magnetic charge and reduces to I2 in the Abelian limit. It is, however, immediate that
14 describes physical systems which are more general than those desired. In the final
term in (5.11), which can be written
can correspond to true systems of point electric and magnetic charges. But we do not
at present want to invoke (5.14) at the Lagrangian level since that would lead to the
equations 3n~l = 0, instead of (5.1e). [In fact, this equation is satisfied for all known
solutions of (5.1), but we will be more general in this subsection].
We proceed to exhibit the equations of motion which result from setting to zero
the coefficients of the independent field and trajectory variations. The 8F equation
is (5.1a,b), which implies (5.2). The 8A equation is
(3 +e,~X) F = 7 +P,, (5.15)
as anticipated above. The 8¢, 5~ t equations are_~4.10), which imply (5.1d) with i~
given by (4.7). The 6X equation is (5.1e). The 8M equations are
or
(3,/ + eA~
-" X) -* -+d ,
X~ = 2For (5.17)
and
i"G(o, o) = o . (5.18)
where
[cf. eqs. (3.15a,b,c), (3.20), and (3.24).] is the expected term to be disposed of by
imposing a generalized Dirac veto, and
describes the unphysical influence of the charged string on the charged particle at its
endpoint. The field equations (5.1e) have been used to simplify (5.22).
The action 14 is thus seen to describe in general a theory of non-Abelian charged
particles and strings. The field R u, which has no Abelian analogue, is the electric cur-
rent of the strings and renders the strings observable. Likewise, c~ u in (5.19) repre-
sents the force of the charged strin__~son the charged particles. We therefore must
only consider solutions for which R u - 0 and c~u = 0. Consequently, we must sup-
plement our action (5.11) with the constraint equation (5.14). It then follows from
(5.1 e) and (5.17) that the dynamical fields involved in (5.1) are constrained by
Fo5 x = 0. (5.23)
It is in fact sufficient to take
~n(o, r) = an(o, r) ~'1(o, r) . (5.24)
Eqs. (5.1e) and (5.17) then give
which fixes anan. Choosing an appropriate 0% gives a Xn from (5.24) and leaves only
the constraint
1~ cx F--+da t , (5.26)
which is a special case of (5.23). Our original equations of motion (5.1), supplemented
by the new constraint (5.23) or (5.26), are thus deduced from an action principle.
Conversely, any solution of (5.1) which satisfies (5.26) will satisfy the extra Lagrangian
equation (5.17) and constraint (5.14) simply by defining Xn by (5.24) with any o~
satisfying (5.25).
It will turn out that all known exact solutions of (5.1) satisfy (5.26) anyway (see
subsect. 5.4) and so we have been unnecessarily general here. This excessive generality
has been motivated by our hope for the eventual discovery of other solutions.
In summary, our field equations (5.1) have led to two constraint equations - the
integrability constraint (5.4) and the Lagrangian constraint (5.23):
= -'-~d
ForXM FozXfi =0. (5.27)
R.4. Brandt, F. Neri / Classical Lagrangian formalism 243
These can be thought of as boundary conditions necessary for (5.1) to describe true
monopoles. Although it is conceivable that there might exist an action different from
ours for which it is not necessary to impose the second constraint, the similarity of
the two constraints leads us to believe that the second constraint, as the first, is a
property of the physical system and not of our particular action which formulates
the theory.
Given (5.14), the equations deduced from 14 agree with (5.1) apart from the con-
tribution of (5.21) to the force law (5.19) *. To obtain the desired force law (5.10,
14 must be modified as in the Abelian case treated in sect. 3. Fi~tly, to be a continu-
ous function of the trajectories Pi, 14 must be defined modulo Q(x). ~l(x) for
X E {Pi) A (Y,]}. [This can beseen most simply ~ the second order formalism, where
the field Lagrangian is ¼F 2 + J • A + K " B with F given explicitly by (5.9b). If I"i
sweeps through a string ~,], the action changes by
fdz,6.X"=ffta (5.28)
D
This requires that the classical quantization condition (1.5) be satisfied. Secondly, 14
must be modified by specifying a contour prescription as in ref. [13] if any trajec-
tory Fi hits a string "0" The so-modified action, constrained by (5.14), will then imply
the correct equations of motion (5.1), constrained by (5.23).
5.3. Symmetries
The equations of motion (5.1) are invariant to the non-Abelian gauge transforma-
tions (4.5) plus (5.5) for non-singular functions •. Unlike the Abelian equations (3.6),
eqs. (5.1) are not invariant to the usual Poincard transformations because of the string-
dependence of (5.1b). [Eqs. (5.1 c - f ) a n d (5.2) do not determine the theory. Eqs.
( 5 . 1 a - f ) do determine the theory and imply (5.2).] But eqs. (5.1) are invariant to
string rotations together with singular gauge transformations. In fact, if a surface Zi
changes even locally by 6 Zi, the simultaneous infinitesimal change (4.5a) in A induces
the covariant change (4.5b) in F if A is chosen to satisfy
a ^ aX(x) -
f
dodr(.j, A y ')aM(y)~4(x _ y ) . (5.29a)
Eqs. (5.1) are thus invariant to Zi ~ Zi + 6~i plus (4.5) plus (5.5) given (5.29a). It
follows that (5.1) are invariant to the usual infinitesimal Lorentz transformations
x ~ ~2x (with the usual associated field transformations) together with the string
rotation n -~ ~ n in eachyi(o, r) = zi(o ) + n'r. This conclusion is, however, completely
* Note that (5.21) vanishes if ~i = cMi for some constant c. This is perfectly analogous to t h e
situation in the Abelian case given in eq. (3.24). This is in spite of the fact that even if Qi = cMi,
the absence of dual symmetry in the non-Abelian case does not allow one to conclude that
the magnetic charge can be rotated away.
244 R.4. Brandt, F. Neri / Classical Lagrangianformalism
formal because, although (5.29a) implies that .~(x) iseffectively non-vanishing only
on the infinitesimal surface 8 Zi, it also implies that A(x) has a large discontinuity
through ~ I~i. To be precise, we must deal with the finite gauge transformations (5.10)
which are the integrated forms of the infinitesimal transformations. Then the condi-
tion that the change G G m G [eq. (5.1b)] induced by a surface change is compen-
sated by a gauge transformation is
1
-0 ^ _U-1)U = G 'a - G a . (5.29b)
e
Using the parametrization
x
and that there exist new fields ~ , and c ~ in the adjoint representation ** of some
* A dual invariant monopole theory based on a non-compact group is considered in ref. [5].
** We use the same vector notation for elements of the adjoint representations of both G and G'.
R.A. Brandt, F. Neri / Classical Lagrangian formalism 245
(o + g ~ x ) ~ ~ = 6 L , (5.31c)
(a+g~X)Q =0. (5.31d)
To continue, we assume that
M, (5.38)
which is also the case for all known solutions. Then
62 = 62 (5.39)
and
(5.40)
so that (5.35) becomes
Now eqs. (5.31) are precise magnetic analogues of eqs. (5.1) and the relationship is
reciprocal in that the same procedure applied to (5.31) leads back to (5.1). The point
that F a need not be a gauge field may be resolved in this formalism by the fact that
the "generalized" dual tensor (5.33) is a gauge field.
Much work remains to be done to show that the above construction is physically
meaningful. In particular, the role of the "magnetic" group G' and its relation to
magnetically charged fields should be explored in quantum field theory. Also, it
would be interesting to investigate the connection of our magnetic~roup and the one
defined by Goddard et al. [23]. The importance of the vectors M/M 2 above actually
suggests the existence of such a connection.
We turn next to the symmetries of the action (5.11). The gauge group transforma-
tion laws are (4.5a,b), (4.8), (5.5), and
Cv) -->X (y) + e~ (y) X A(y). (5.41 )
The conserved Noetherian current corresponding to the global (space-time indepen-
dent) gauge group is
Using the field equation (5.1c), this can be put in the manifestly conserved form
~u ~ F~v. (5.42b)
or, by (5.2)
k u = v~v~'a
-uv • (5.43b)
This magnetic current is, however, not a Noetherian one since there is no magnetic
symmetry group. In the Abelian case treated in sect. 3, we deduced the conservation
of magnetic current from the electric gauge invafiance of the action plus the dual
invariance of the equations of motion. In the present case, even the equations of mo-
tion are not dual invariant, but we can proceed nevertheless because of the fact that
the equations of motion are at least partially dual invariant. Eqs. (5.1 c - f ) and (5.2)
are invariant to dual rotations in which (F, F a) and (Qi, Mi) rotate as two-vectors
and e and A are invariants. The conservation of the Noetherian electric current (5.42a)
follows just from the dual covariant field equation (5.1c), and correspondingly the
conservation of the magnetic current (5.43a) which arises from (5.42a) by a dual
rotation, follows from the dual covariant field equation (5.2). The field equation (5.1)
which breaks the dual invariance is irrelevant in this analysis [21 ].
Other conserved currents in our theory are the electric and magnetic source cur-
R.A. Brandt, F. Neri / ClassicalLagrangianformalism 247
rents
5.4. Solutions
Simple static solutions to the field equations (5.1) can be found. These solutions
are essentially Abelian in that there exists a gauge for which
7'. X 7'u = 0. (5.46)
-+
Ju = 0 , -+= 0 ,
K K,o (r) = M~3(r) , (5.47d)
corresponding to a static trajectory and surface:
z~(o) = 8~oO, yu(o, r) = zu(o ) + n u t . (5.47e)
F~o = 0. (5.47f)
This solution obviously satisfies the integrability constraint (5.4) and the Lagrangian
constraint (5.23) (For = 0, ~ a = bin. r/4~3).
All of the known singular non-Abelian monopoles have essentially the above form,
but we suspect there exist other, essentially non-Abelian, solutions to (5.1). This will
be explained below.
The simple form of the above known solutions makes possible their derivation
from the simpler action
f, = - f dodrfi, . (5.48)
where the Mi are given constant vectors. Variation of the independent fields in (5.48)
leads to the equations of motion (5.1a-d,f), and eq. (5.1e) becomes the constraint
equation
ACvi) X bi = 0. (5.49)
eeM'--T= 1 (5.50)
is satisfied. We can construct a self-contained derivation of this result by performing
a first-quantization of the non-relativistic limit of our theory. The non-relativistic
limit of the Lagrangian density in the second piece of the expression (4.9) for I3 is
(taking one charged particle for simplicity)
which imply
eifi'~ = 1, (5.57)
or eq. (1.3).
It is remarkable that a strict charge quantization condition can be deduced even
for the classical non-Abelian theory. Consider the matrix form [of. eqs. (5.7)]
A =MD, Du(r;h ) = eouik~k/4n(r - r.h~;
on our solutions. For further simplicity, we will actually consider only solutions for
which a vanishes on the fixed string.
To simplify our analysis, we will use the background gauge condition
ava v + e[Av,_a v] = 0, (5.61)
and choose a new basis t__r of our Lie algebra such that
Such a basis always exists, with at least one non-vanishing eigenvalue gr, in any com-
pact semi-simple Lie algebra. We can thus write
a u = ~arut r . (5.63)
¥
It then follows from the field equation (5.7) that, to first order in_a,
(~ + iegrD)2aru + 2iegr(a A O)Uua~ = O. (5.64)
250 R.A. Brandt, F. Neri / ClassicalLagrangian formalism
This is the equation that describes the scattering of vector mesons of type r in the
field of a monopole. But, because of the n dependence of (5.58), the solutions of
(5.64) will not, in general, be rotationally invariant. Only if the string direction n
can be arbitrarily rotated by a gauge transformation will (5.64) describe rotationally
invariant physics. As usual (see, e.g., ref. [11]. A more detailed analysis of (5.64)
will be given in ref. [25]), the condition for string independence is a quantization
condition. The condition in this case reads [25]
The reason we are able to deduce a quantization condition even in the classical case
is that the vector potential is present in the basic field equation (5.7b). In contrast,
in the Abelian case or for the essentially Abelian solutions (5.58) to the non-Abelian
equations, there is no vector potential in the field equations.
The solutions to eq. (5.64) give rise to approximate solutions of the non-Abelian
field equations (5.7). The existence of these non-essentially Abelian approximate
solutions suggests that exact non-essentially Abelian solutions also exist. The solutions
to (5.64) also determine the stability of the essentially Abelian solutions (5.47). We
find that the nr = 1 monopole is stable but that the nr > 1 monopoles are not stable.
The derivation of these and related results will be given in another paper [25].
Consider thesolution (5.58) of the previous subsection in the case of SU(2), with
the generators T = ½T m terms of the Pauli matrices L, and with
/~ = ~ = (0, O, g ) . (5.68)
* The existence of the Al~elian classical action requires only that the weak quantization condi-
tion (1.5) is satisfied. Use of the action (3.13a) actually leads to a strong Abelian quantization
condition, but this is clearly only an artifact of the unnecessary complexity of that action.
R.A. Brandt, F. Neri / Classical Lagrangian formalis m 251
The matrix
I COS 10e i~ sin 10 ]
_V(f) = (5.69)
- s i n 30 cos ~ O e - i ~ J '
where ~ = (sin 0 cos ~, sin 0 sin ~b, cos 0), has the property that
_ V ( f ) g V - l ( ~ ) = g~" ~ . (5.70)
If we consider _Vas a transformation matrix, and choose e = 41r/g, the transformed
field satisfies [6,17]
F[7 g rark
= eiJk 47r r 4 ' (5.73a)
'a_ g rb
A i - eia b 4rrr 2 . (5.73b)
We now pose the question: Is _Va gauge transformation? I.e., does it leave the
equations of motion invariant, or, in other words, does it take solutions into solutions?
Applying _Vto the original field equations
(a + e,~ ×) ~ = 0 , (5.74a)
(a +e~,X)~d=g.o = "-"
ga 3(r), (5.74b)
we find
(a + e,~'X) F ' = 0 , (5.75a)
Vis too singular for K' = V K V - l to exist, G' = V G V -1 = ti • ~ S (see eq. (5.1b))
d-oes exist and (a + e.~' X)-~'= 0.] The field (5.73) is, in fact, one of the Wu-Yang
[26] solutions which, as they recently stressed [9], is n o t a gauge field at the origin.
[For example it fails to satisfy the Bianchi identity at the origin.]
We can attempt to generalize the class of allowed solutions to our field equations
beyond temperared distributions. One possibility is to define K' = 0 in our (assumed
to exist) extended class of generalized functions. I f this is possible, then (5.73) can
be legitimate gauge fields with zero generalized magnetic source. These fields are then
physically distinct from the original ones (5.47), and _Vis not a gauge transformation•
For example, the magnetic charge
3
~K - f d rK0(r) (5.76)
does not define an invariant ~ 2 since ~:2 =g2 for (5.47) and ~K,2 = 0 for (5.73). It
• • "~t . • -
is, however, also conceivable that K could be interpreted to not vanish and satisfy
cK 2 = cK 2 = g2. Then (5.73) could be equivalent to (5.47) and so _Vwould be a gauge
transformation. However, (5.73) would then provide an extremely singular (although
string free) description of the physics. For example, it fails to satisfy the Bianchi iden-
tity even through it is string-free.
It should be noted that the problems with _Valluded to above are n o t present in the
first-quantized theory of non-Abelian monopoles. This is because the given external
fields always appear multiplied by Schr6dinger wave functions and these functions
vanish (sufficiently fast) at the origin. The fields (5.47) and (5.73) are thus gauge
equivalent in non-relativistic quantum mechanics.
Because of the irrelevance of the origin in quantum mechanics, a Wu-Yang [9]
formulation of the theory is possible in which the space outside of the monopole is
divided into overlapping regions each carrying its own vector potential. This is a string-
free description of the monopole even in the "Abelian" gauge (5.58), as long as
eg/4n = ~n for integer n. Ifn is even, then the transformation (and its generalizations
for n > 2) which removes the string can be used to fuse together the Wu-Yang regions
into a single region in which the vector potential is given by (5.73) everywhere outside
of the origin [9,18,23,24]. It is, however, presently unknown if such a description is
possible in classical mechanics or in quantum field theory, where it may not be pos-
sible to ignore the monopole position.
We note finally that (5.47) and (5.73) are equivalent starting points for the con-
struction of non-singular monopoles. This is because the non-singular solutions (of
the source-free equations including I-Iiggsscalar fields) are obtained by multiplying
(5.47) or (5.73) by functions which again vanish sufficiently fast at the origin [6-8].
References
[3] C.N. Yang and R.L. Mills, Phys. Rev. 96 (1954) 191.
[4] A. Joseph, Phys. Rev. D5 (1972) 313;
N. Murai, Prog. Theor. Phys. 47 (1972) 678;
L.R. Thebaud, Phys. Rev. D14 (1976) 1673.
[5] N. Christ, Phys. Rev. Lett. 34 (1975) 355.
[6] G. 't Hooft, Nucl. Phys. B79 (1974) 276.
[7] A. Polyakov, ZhETF Pisma 20 (1974) 430 (JETP Lett. 20 (1974) 194).
[8] S. Coleman, in Proc. 1975 Int. School of subnuclear physics "Ettore Majorana" (Academic
Press).
[9] T.T. Wu and C.N. Yang, Phys. Rev. D12 (1975) 3845.
[10] Gu Chao-hao and C.N. Yang, Sci. Sin. 18 (1975) 483.
[11] R.A. Brandt and J. Primack, Phys. Rev. D15 (1977) 1175.
[12] D. Zwanziger, Phys. Rev. 176 (1968) 1489.
[13] R.A. Brandt and J. Primack, Phys. Rev. D15 (1977) 1798.
[14] R.A. Brandt and F. Neri, Phys. Rev., in press.
[15] D. Zwnaziger, Phys. Rev. D3 (1971) 880, and references therein.
[16] R.A. Brandt, F. Neri and D. Zwanziger, Phys. Rev. Lett. 40 (1977) 147, and to be published.
[17] J. Arafune, P.G.O. Freund and C.J. Goebel, J. Math. Phys. 16 (1975) 433.
[18] Z.F. Ezawa and H.C. Tze, Phys. Rev. D15 (1977) 1647.
[19] S.K. Wong, Nuovo Cim. 65A (1970) 689.
[20] J. Schwinger, Phys. Rev. 114 (1966) 1087;
T.M. Yan, Phys. Rev. 150 (1966) 1349.
[21] R.A. Brandt and K. Young, to be published.
[22] J. Mandula, Phys. Rev. D14 (1976) 3497.
[23] P. Goddard, J. Nuyts and D. Olive, Nucl. Phys. B125 (1977) 1.
[24] F. Englert and P. Windey, Phys. Rev. D14 (1976) 2728;
E. Corrigan and D. Olive, Nucl. Phys. B l l 0 (1976) 237.
[25] R.A. Brandt and F. Neri, to be published.
[26] T.T. Wu and C.N. Yang, in Properties of matter under unusual conditions, ed. H. Mark and
S. Fernbach (Wiley, New York, 1969) p. 349.