Classical Four-Vector Fields in The Relativistic Longitudinal Gauge

JOURNAL OF MATHEMATICAL PHYSICS VOLUME 41, NUMBER 7, (pp.
4622−4653)# JULY 2000
Classical four-vector fields in the relativistic longitudinal gauge†
Dale A. Woodsidea)
Department of Physics, Macquarie University-Sydney, New South Wales 2109, Australia
(Received 21 September 1999; accepted for publication 24 March 2000)
Minkowski four-space uniqueness theorems are used to develop further the author’s “relativistic
longitudinal gauge” [J. Math. Phys. 40, 4911 (1999)] for four-irrotational classical four-vector
fields. A theorem is developed which distinguishes between two and only two “physical” classes of
classical four-vector fields. One must satisfy the “relativistic transverse gauge”, i.e., the Lorentz
condition, while the other must satisfy this new relativistic longitudinal gauge where its four-curl,
the Maxwell field tensor itself, is set to zero. The Lagrangian density of the new four-irrotational
four-vector field is distinguished from the usual Lorentz constrained Lagrangian density by the
incorporation of an additional overall minus sign. Application of the relativistic longitudinal gauge,
in the four-irrotational four-vector field case, eliminates the badly behaved terms associated with the
spatial degrees of freedom from a most general, fully quadratic, Lagrangian density. The resulting
constrained Lagrangian density is bounded from below and therefore a relativistic longitudinal
classical four-vector field has the possibility of a physical interpretation.
c 2000 American Institute
of Physics. [S0022-2488(00)04907-0]
†
This version of the article was typeset by the author using LATEX 2e and REVTEX v3.1.
#
Note, page numbering differs slightly from the original journal article. [See heading at top of page.]
I. INTRODUCTION
In a previous article by the author,1 a Minkowski four-space “Helmholtz” identity was proved using a Minkowski
space retarded Green’s function. This identity was then used to prove uniqueness theorems for sufficiently smooth
four-vector fields in the Minkowski space R3+1 . It was found that the specification of the four-curl and four-divergence
of the four-vector field throughout the four-volume V4 , as well as the four-tangential and four-normal projections of
the four-vector field everywhere on the bounding three-surface Σ, are sufficient to obtain a unique four-vector field. A
further result was that a four-vector field is uniquely specified by the sum of a four-irrotational and a four-solenoidal
part. This latter theorem corresponds to a four-space generalization of Helmholtz’s uniqueness theorem.
These results are now applied to two general classes of four-vector fields. These two classes of four-vector fields
are distinguished by their different gauge related properties. It is found that in addition to the usual (pure real)
four-vector field in a “relativistic transverse gauge,” i.e., an electromagnetic-type field in the Lorentz gauge, there is
also the possibility of a physical interpretation for a new four-vector field whose Lagrangian differs from the former
by an overall minus sign. In order for this new four-vector field to be classically bounded from below, it turns out
that the field must be constrained by a new relativistically covariant gauge that the author has previously referred to
as the “relativistic longitudinal gauge,”1 where the Maxwell field tensor is set to zero while the four-divergence is in
general non-zero. This results in a four-vector field which is “four-irrotational.”
New uniqueness statements for these two classes of four-vector fields are then developed in Secs. III C and III D for
sufficiently smooth massless fields in unbounded Minkowski space using the author’s Minkowski four-space Helmholtz
identity.1 In the standard (pure real) four-vector field case, the Lorentz condition is applied to the identity and leads to
what is called a “restricted relativistic transverse gauge.” In the new four-vector field case, the relativistic longitudinal
gauge condition is applied to the identity and leads to what is called a “restricted relativistic longitudinal gauge.”
These restricted relativistic gauges have the notable property that the usual gauge transformation of the second kind
reduces to an identity transformation.
A covariant canonical formulation of the classical four-vector field is presented in Sec. IV. The approach is sufficiently
general so as to include both the usual (Lorentz constrained) four-solenoidal fields and the new four-irrotational fields
in one comprehensive formalism. Both four-vector and pseudo-four-vector current coupling is explored. Theorem
III of Sec. IV A proposes that there are two and only two classes of classical four-vector fields, associated with a
most general, fully quadratic, Lagrangian density, which are bounded from below under covariant constraint and are
therefore potentially “physical”. One class is comprised of four-vector fields in the Lorentz gauge and the other is
comprised of these new four-vector fields in the author’s relativistic longitudinal gauge.
4622
J. Math. Phys., Vol. 41, No. 7, July 2000 Dale A. Woodside : Classical four- vector fields in the ... 4623
II. MATHEMATICAL BACKGROUND
Before proceeding, a few preliminary definitions are made. The nonzero components of the flat space Min-
kowski metric tensor ηµν are taken as −η00 = η11 = η22 = η33 = −1. So, the ordinary four-vector derivatives are
taken as ∂µ = ((1/c)∂/∂t, ∇) and ∂ µ = ((1/c)∂/∂t, −∇). Similarly, the position four-vector xν = (ct, x, y, z), and so
xν = (ct, −x, −y, −z) in this (+ − −−) signature metric.
Next, it is assumed that the most general form of Lagrangian density for a four-vector field, which is no more than
quadratic in its variables and their derivatives, is given by the so-called Stueckelburg Lagrangian density2,3 (in SI
units where 0 is the free space permittivity and c is the speed of light),
o c2 λo c2 2 o c2 µ2
L=− Fµν F µν − (∂µ Aµ ) + (Aµ Aµ ) − jµ Aµ , (1)
4 2 2
where j ν = (ρc, j) is the usual four-vector current, where the positive real constant λ is a Lagrange multiplier for the
Lorentz constraint term, and where µ = 2π/λC = 2πmc/h is the Compton wave number for photons of mass m. A
choice of λ = 0 and µ = 0 yields what many physicists believe to be the electromagnetic theory, with its massless
photons, i.e., when an appropriate constraint is externally imposed. However, the choice of λ = 0 has the distinct
disadvantage of implying a vanishing momentum canonically conjugate to the zeroth component of the four-vector
potential Aν = (φ/c, A). The incorporation of the Lorentz constraint term, with its ∂φ/∂t functionality, eliminates
this deficiency, and yields an added bonus in terms of the ease of renormalization of the theory. A particularly simple
choice of λ = 1 (and µ = 0), then yields a Lagrangian density which is equivalent (i.e., differs by no more than a four-
divergence), to the so-called Fermi Lagrangian density.4 The Fermi Lagrangian density is the most straightforward
take off point for field quantization in terms of harmonic oscillators which correspond to massless photons (cf. Ref.
4). The Stueckelburg Lagrangian density (1) also has the advantage of explicitly including the four-divergence and
four-curl of Aν , which are in turn sufficient for the unique specification of a four-vector field as is reviewed in Sec.
III C. The inclusion of the four-divergence in particular is what makes (1) the only suitable choice for the analysis of
four-vector fields in the author’s new relativistic longitudinal gauge.
III. APPLICATION OF THE MINKOWSKI SPACE UNIQUENESS THEOREMS OF REF. 1 TO TWO

CLASSES OF FOUR-VECTOR FIELDS
A. Introduction of a new class of four-vector fields
In this chapter the Minkowski space uniqueness Theorem V of Ref. 1 will be applied to two general classes of classical
four-vector fields. One class is comprised of fields of the same general form as a (massive) electromagnetic field. In
the classical domain this class is characterized by (ostensibly pure real valued) fields constrained by a relativistic
transverse gauge, i.e. by the Lorentz condition. The second class is comprised of fields whose Lagrangian density
differs by an overall minus sign from the former. It turns out that the Minkowski space uniqueness Theorem V of Ref.
1 sheds new light on the distinction between these two classes of four-vector fields. The theorem leads to a unique
selection of covariant constraints which not only render the electromagnetic field classically bounded from below in
the usual way, but in addition render this new class of classical four-vector fields bounded from below as well.
Consider (1), the classical free space Lagrangian density of the electromagnetic field (with a small mass), now
written as (2) in terms of the electric field E and the magnetic field B, and with its Lorentz constraint term now with
λ = 1, as follows:
2
o c2 E2 o c2 1 ∂φ o c2 µ2 φ2

2 2
Lem = − B − + ∇ · A + − A − ρφ + j · A. (2)
2 c2 2 c2 ∂t 2 c2
Although the photon Compton wave number µ is certainly very tiny, if not actually zero, for the sake of generality,
let us suppose that it is nonzero.
Now, introducing the four-vector potential Aµ = (φ/c, A) in the usual way
∂A
E = −∇φ − , B = ∇×A, (3)
∂t
allows one to rewrite (2) entirely in terms of the potentials as
" 2 # 2
o c2 o c2 1 ∂φ

1 ∂A 2
Lem = −∇φ − − (∇×A) − + ∇·A
2 c2 ∂t 2 c2 ∂t
o c2 µ2 φ2

2
+ − A − ρφ + j · A. (4)
2 c2
It is interesting to note that due to the inclusion of the Lorentz constraint term, every possible derivative in four-space
of φ and A appear in (4) as meaningful squared quantities, i.e., to quadratic order.
It is important to note that the time derivative of A appears in the Lagrangian density (4) with a plus sign, since it
is in a squared term. This is required for the classical action (i.e., the four-space integral of the Lagrangian density),
to be bounded from below (i.e., have a finite minimum). What is not usually remarked about is the initially surprising
fact that the time derivative term involving the scalar potential φ does not enter the Lagrangian density with a plus
sign. Therefore, for a sufficiently rapid change in time, over a given time interval, one can make the action negative,
thus preventing it from having a finite minimum under variation of its parameters, and thus violating the principle of
least action. The Lorentz constraint addition to the Lagrangian density is therefore not bounded from below! Another
consequence of this violation is that negative energy terms appear in the reciprocal space Hamiltonian, (cf. Ref. 5, p.
378, Eq. (A64)).
In the quantized version of (4), this leads to a commutation relation between creation and annihilation operators
of the scalar photons associated with the scalar potential φ as follows:
0 0
as (k), a+

s (k ) = −δ(k − k ) (5)
(cf. Ref. 5, p. 381, Eq. (B4b)). The minus sign on the right-hand side (rhs) of (5) leads to an indefinite metric
Fock space. [It is easily shown that an anti-Hermitian massive scalar field (i.e. adding an overall minus sign to the
usual massive scalar field Lagrangian), leads to the same result, Eq. (5). Therefore, one might interpret a massive
electromagnetic-type four-vector field as containing within it an anti-Hermitian scalar field as its zeroth component.]
However, (for µ = 0) it is well known2,6–8 that through the introduction of a subsidiary condition, the positive
energy terms of the third component of the vector potential (i.e., the longitudinal photons), compensate for these
negative energy scalar photons in the Hamiltonian. This allows a positive square norm Fock space to be developed
after all. Indeed, this process is valuable in that it preserves the Lorentz covariance of the theory.
It is appropriate at this point to mention some historical aspects of the indefinite metric quantization procedure
to drive home the point. A key issue for this procedure is the space-time coordinate parameterization itself. Two
essentially different parameterizations of space-time coordinates exist which therefore lead to two different ways of
approaching this procedure.
The current, fairly universal definition is to parameterize space-time with a contravariant four-vector xµ in a
Riemannian space, here assumed to be without gravitational fields (i.e., a flat space),
xµ = x0 , x1 , x2 , x3 = (ct, x, y, z).

(6a)
The covariant four-vector xµ is then obtained by contraction with the flat space Minkowski metric tensor ηµν as
xµ = ηµν xν = (x0 , x1 , x2 , x3 ) = (ct, −x, −y, −z). (6b)
(Here ηµν is the diagonal form, as defined in Sec. 1, of the more general metric tensor gµν for the case of vanishing
gravitational fields.)
Minkowski on the other hand, in 1908, had already introduced the concept of an imaginary valued fourth coordinate
for time.9 Thus, an intriguing complex space-time parameterization has frequently been used instead of Eqs. (6a) and
(6b) as follows:
xµ = (x1 , x2 , x3 , x4 ) = (x, y, z, ict), (7)
with a pure imaginary valued fourth coordinate x4 = ict, and where only lowered indices are used in tensor equations.
The basic difference between the two parameterizations is that no allowance for a possible geometric description of
gravitation was included in the space-time parameterization of (7). Therefore, due to the success of Einstein’s General
Theory of Relativity, the use of this space-time definition has for the most part been abandoned.
In the middle of the century, however, authors still frequently used or at least referred to the complex space-time
definition. Indeed, it would appear that Gupta7 and Bleuler,6 the developers of the indefinite metric quantization
procedure for the electromagnetic field, were greatly influenced by it. Gupta used Aµ = (A0 , A1 , A2 , A3 ) with A4 = iA0
for his four-vector potential, while Bleuler used Aµ = (A, iV ) = (A1 , A2 , A3 , A4 ) for his four-vector potential, and
they both used all lowered indices indicative of a flat space-time. So the vector potential components A1 , A2 , and
A3 , as well as the scalar potential A0 = V are Hermitian operators for Gupta and Bleuler, but the key point is that
A4 = iA0 = iV is thereby an anti-Hermitian operator! This fact is obvious with the space-time parameterization of
(7), but is unfortunately submerged in the signature of (6). Thus, for Gupta and Bleuler the problems associated
with the fourth component of the four-vector potential of electromagnetism stemmed from considering Aµ to be a
Hermitian operator, despite the fact that the fourth component was giving a negative energy. Their solution was
essentially to do a unitary transformation using an indefinite Dirac metric10 in order to force the anti-Hermitian A4 to
give pure imaginary expectation values. Insightful discussions of the method using the complex space-time definition
(7) are given in Mandl8 and Källén.11
Returning to the current investigation, it becomes interesting to turn the whole problem around and introduce an
overall minus sign into a most general, fully quadratic Lagrangian of the form (2). Underscored variables are all taken
as pure real valued. [In a purely formal way one could however incorporate the overall minus sign in the Lagrangian
into the quadratic fields by defining a pure imaginary valued four-vector potential K µ = iK µ = i(κ/c, K), where
K µ is a pure real valued four-vector potential, since i2 = −1. Similarly, for the source terms one could define pure
imaginary valued sources, i.e., jgµ = ij µg = i(cρg , jg ), where j µg is a real four-current. Although purely formal, this does
not in principle present a problem for a classical four-vector field where the force, Hamiltonian, etc., always contain
products of pairs of these variables and are thereby real number fields.] With the different constants, o → go and
µ → µg , one has the hypothetical free space Lagrangian density,
2
go c2 G2 go c2 1 ∂κ

2
Lgi = − −I + +∇•K
2 c2 2 c2 ∂t
go c2 µ2g κ2

2
− − K + ρg κ − jg • K. (8)
2 c2
In analogy with the subscripts “em” in (2) which refer to the electromagnetic field, the subscripts “gi” in (8), and
throughout this article, label the theory of the fields G and I (or the pure imaginary valued G and I).
The real four-vector potential K µ of (8) is related to the real vector fields G and I as
∂K
G = −∇κ − , I = ∇×K, (9)
∂t
so that upon substitution one obtains
" 2 # 2
go c2 1 go c2 1 ∂κ

∂K 2
Lgi = − −∇κ − − (∇×K) + + ∇ • K
2 c2 ∂t 2 c2 ∂t
go c2 µ2g κ2

2
− − K + ρg κ − jg • K. (10)
2 c2
One can then take K = 0 and ρg = 0 to yield
2
go c2 1 go c2 1 ∂κ go c2 µ2g κ2

2
Lgi (K = 0, ρg = 0) = − (−∇κ) + −
2 c2 2 c2 ∂t 2 c2
" 2 #
go 2 1 ∂κ
=− (∇κ) − + µ2g κ2 . (11)
2 c ∂t
That is, when K = 0 and ρg = 0, the theory reduces to the equivalent of a pure real (i.e., Hermitian), scalar field
theory! Note that the time derivative terms in the Lagrangian density (11) are positive and so this reduced Lagrangian
is classically bounded from below. Of course, the vector potential components K (i.e., the effectively anti-Hermitian
K), still have negative time derivative terms in the Lagrangian (10) and are therefore not bounded from below.
However, the bottom line is that if one can eliminate the vector potential K in a satisfactory and nontrivial way,
while preserving the effectively pure real degree of freedom (i.e., the scalar potential κ), then one will have available
a new class of classical field theories.
B. Gauge considerations in classical field theory
A first method for eliminating the effectively anti-Hermitian degrees of freedom from (10) follows from an exami-
nation of the parallels between the electromagnetic theory and this new theory. In the case of electromagnetism, a
noncovariant method for eliminating what is effectively an anti-Hermitian scalar potential iφ exists in the form of
the Coulomb or transverse radiation gauge. A covariant method entails instead the use of a Lorentz gauge or the
inclusion of Lorentz constraint terms in the Lagrangian density as in (2).
The transverse and longitudinal aspects of vector fields in general and the electromagnetic field in particular yield
important insights for the present investigation, so a short review of the topic is warranted, (cf. Refs. 5, 12, and 13).
It is possible to decompose a vector field over all of a Euclidean three-space into the components
A(r) ≡ AL (r) + AT (r), (12)
which are longitudinal, i.e., irrotational,
∇×AL (r) = 0, (13a)
or transverse, i.e., solenoidal,
∇ · AT (r) = 0, (13b)
for all r. The decomposition into (13a) and (13b) is unique over all R3 by Helmholtz’s theorem. A discussion of the
use of Helmholtz’s theorem with fields which fall off only as fast as 1/r (e.g., potentials), is given in connection with
the modified three-space Helmholtz identity (11b) in Ref. 1. The Fourier spatial transform pair A(r) ↔ A(k) is now
defined as follows:
Z
1
A(k, t) = A(r, t)e−ik · r d3 r, (14a)
(2π)3/2
Z
1
A(r, t) = A(k, t)eik · r d3 k. (14b)
(2π)3/2
The time dependence is shown here for reference, but will be subsequently suppressed. Additionally, one has the
Fourier spatial transform pair ∇ ↔ ik which thereby yields for the longitudinal component,
ik×AL (k) = 0, (15a)
and for the transverse component
ik · AT (k) = 0, (15b)
for all k. Thus, the longitudinal vector field, AL (k), is parallel to the wave number k, while the transverse vector
field, AT (k), is perpendicular to k. This decomposition of a vector field is however not generally Lorentz covariant
(since, e.g., a vector field which is transverse in one inertial reference frame is not necessarily transverse in another).
Next consider the gauge transformation of the second kind as defined by
A(r, t) → A0 (r, t) = A(r, t) + ∇F (r, t), (16a)
∂F (r, t)
φ(r, t) → φ0 (r, t) = φ(r, t) − , (16b)
∂t
or in covariant notation as
A0µ (xν ) = Aµ (xν ) − ∂ µ F (xν ) . (17)
The gauge transformation (16) leaves E(r, t) and B(r, t) unchanged. In defining A via the Maxwell’s equation
∇ · B = ∇ · ∇×A ≡ 0, a certain arbitrariness is left in A due to the vector identity ∇×∇F ≡ 0, where F = F (r, t)
is the arbitrary sufficiently smooth scalar function appearing in (16a). So, in order to leave the electric field E(r, t)
as defined by (3) unchanged, the transformation equation (16b) must be introduced as well. Note however that by a
suitable extension of Theorem U of Ref. 1 to vector potential fields A which typically fall off only as fast as 1/r (see
Ref. 1), one can uniquely specify the vector potential field A over all of a Euclidean three-space by specifying both
its divergence and its curl. Consequentially, once the ∇×A is specified (in R3 ), one must also specify the ∇ · A in
order to specify A uniquely.
One way to specify the ∇ · A is with the Lorentz condition defined by
1 ∂φ(r, t)
∇ · A(r, t) + = 0. (18)
c2 ∂t
Equation (18) has the advantage that it is a relativistically invariant relation as can be seen by writing it in covariant
notation as
∂µ Aµ (xν ) = 0. (19)
Indeed, (19) is a four-divergence which coincidentally specifies the scalar potential φ as well. On the other hand, the
Coloumb gauge defined as
∇ · A(r, t) = 0 (20)
has the disadvantage of being a noncovariant condition and consequently is not suitable for covariant uniqueness
statements in Minkowski space. It is nevertheless interesting in that it is always possible to choose a gauge such that
(20) is true in some particular inertial frame of reference (cf. Ref. 14). Indeed ∇ · A(r, t) = 0 and ∇ · A0 (r, t) = 0
when A0 = A + ∇F implies that F is harmonic. And if the harmonic function F vanishes at infinity one therefore
has F vanishing for all xµ as well. The condition (20) has therefore traditionally been considered to determine the
gauge transformation uniquely.2
Now, using the gauge transformation comprised of (16a) and (16b), potentials A0 and φ0 can always be found to
satisfy the Lorentz condition (18) with
1 ∂φ0 (r, t) 1 ∂φ(r, t) 1 ∂ 2 F (r, t)

0 = ∇ · A0 (r, t) + = ∇ · A(r, t) + + ∇ 2
F (r, t) − (21)
c2 ∂t c2 ∂t c2 ∂t2
leading to the associated requirement that
1 ∂φ(r, t)
∇ · A(r, t) + = 2F (r, t), (22)
c2 ∂t
assuming of course that such a gauge function F (r, t) can be found. If one takes the further restriction
2F (r, t) = 0, (23)
then the Lorentz condition would be invariant (provided ∂µ Aµ (xν ) = 0 initially) under a restricted gauge transfor-
mation defined by Eqs. (16) and (23) and typically referred to as the Lorentz gauge.13
The gauge transformation of the second kind (16) is transformed to reciprocal space as
A(k) → A0 (k) = A0L (k) + A0T (k) = AL (k) + AT (k) + ikF(k) (24a)
⇒ A0L (k) = AL (k) + ikF(k), A0T (k) = AT (k),
∂F(k)
ϕ(k) → ϕ0 (k) = ϕ(k) − , (24b)
∂t
where the Fourier spatial transform pair φ(r) ↔ ϕ(k) is defined analogously to (14), and where the transform pair
F (r, t) ↔ F(k, t) represents the arbitrary scalar gauge function. It is important to note the last result of (24a) which
shows that the transverse vector potential AT ↔ AT is gauge invariant, i.e., A0T = AT or A0T = AT . Only the
longitudinal and scalar potentials are therefore changed by a gauge transformation of the second kind. Indeed, the
Lorentz condition (18) similarly only relates the longitudinal and scalar potentials since with (15b), the reciprocal
space version reduces to
1 ∂ϕ(k)
ik · AL (k) + = 0. (25)
c2 ∂t
For the Coulomb gauge, its reciprocal space version is
ik · AL (k) = 0 ⇒ AL (k) = 0 ∀k; k 6= 0. (26)
Here, the longitudinal vector potential is actually discarded, being set to zero. The vector potential in the Coulomb
gauge is therefore entirely transverse.
Consider now the requirements for the present theoretical investigation of the Lagrangian density (10). In this case,
one wants to keep the degree of freedom associated with the scalar potential κ of (9) since it is the only component of
the four-vector field which is well behaved in the classical domain, and also acts like a Hermitian field in the quantum
domain. Therefore, one certainly can not use the Coulomb or transverse gauge as this would eliminate the desired
scalar potential κ. The Coulomb gauge is associated with a solenoidal or transverse vector potential as implied by
(13b). However, by a suitable extension of Helmholtz’s theorem (i.e., Theorem H1 of Ref. 1), to vector potential fields
A which typically fall off only as fast as 1/r, (see Ref. 1), one is free to specify the longitudinal vector potential as
nonzero instead via a relation like (13a). Therefore, taking the vector potential K as irrotational, i.e.,
∇×K(r, t) = 0, (27)
for all r, yields what will be referred to as the “longitudinal gauge”. [P. W. Anderson already in 1958 referred to the
∇×A = 0 as the defining relation in his longitudinal gauge in an investigation of the Meissner effect.15 ] In reciprocal
space, with the spatial Fourier transform pair of the real field K(r, t) as K(r, t) ↔ K(k, t), one has
0 = ik×K(k) = ik×KL (k) + ik×KT (k) = ik×KL (k) ⇒ KT (k) = 0 ∀k; k 6= 0, (28)
which following an inverse Fourier spatial transform implies that the vector potential is purely longitudinal, that is
KT = 0 implies K = KL , hence the name longitudinal gauge. In passing note that under a gauge transformation of
the second kind that the transverse vector potential KT is gauge invariant (by a relation analogous to (24a)). So, for
the longitudinal gauge, the implication that the transverse vector potential equals zero is therefore a gauge invariant
statement, i.e., K0T = KT = 0. It is therefore possible to completely remove the transverse vector potential from the
Lagrangian, Hamiltonian, and equations of motion in a gauge invariant fashion as would be required for consistency
in this new longitudinal gauge. Naturally, as in the Coulomb gauge, one still has the noncovariant limitation of the
approach so that one must choose a particular inertial reference frame for the sake of calculations. Nevertheless, the
longitudinal gauge appears to be a logical first step because it removes the two badly behaved transverse degrees of
freedom from K µ . Now, only the badly behaved longitudinal degree of freedom remains to be removed. This will be
accomplished with a relativistically invariant constraint in Sec. III D which is a natural extension of the longitudinal
gauge (27) to Minkowski space.
C. Uniqueness of the massless classical four-vector field in the relativistic transverse gauge
The Minkowski space generalization of the Helmholtz identity, Theorem II of Ref. 1 (but in the (+ − −−) signature
metric of this paper), as well as the Minkowski space uniqueness Theorem V of Ref. 1, will now be applied to the
case of a massless classical four-vector field in the relativistic transverse gauge (i.e., the electromagnetic field in the
Lorentz gauge). This is the first of the two main Lagrangians as described at the beginning of Sec. III A, but now
specialized to the massless case. The second case, the classical four-vector field of Lagrangian (10) in the relativistic
longitudinal gauge (with µg = 0), will be addressed in section III D. Examination of the electromagnetic case will
provide a basic framework from which this new classical four-vector field will be subsequently approached.
The present analysis will be limited to the case of an unbounded region in Minkowski space. A theorem will
therefore be stated as follows:
Theorem I: Given a suitable covariant scalar two-point retarded Green’s function G (x, x0 ), the following identity
holds for sufficiently smooth (massless) four-vector fields Aµ (xσ ) which vanish sufficiently rapidly at infinity in the
unbounded Minkowski space R3+1 :
Z Z
Aµ (x) = ∂ µ (∂ν0 Aν (x0 )) G (x, x0 ) d4 x0 + ∂α (∂ 0α Aµ (x0 ) − ∂ 0µ Aα (x0 )) G (x, x0 ) d4 x0 . (29)
V40 V40
Proof: Theorem I follows from Theorem II of Ref. 1 for four-vector fields which vanish sufficiently rapidly at infinity
in unbounded Minkowski space (i.e., the three-surface integral terms of the previous identity vanish in the limit).
[Note, the unprimed derivatives have been factored out of the integrands of (29) for convenience and there is a sign
change due to the metric signature used in the present paper.] ♣
Theorem I is an unbounded Minkowski space generalization of the Helmholtz identity.
One can now proceed in a manner that parallels the use of the Helmholtz identity in R3 by stating a Helmholtz
uniqueness theorem for four-vector fields in unbounded Minkowski space as follows:
Theorem II: A sufficiently smooth four-vector field Aµ (xσ ) which vanishes sufficiently rapidly at infinity in the
unbounded Minkowski space R3+1 and which satisfies identity (29) (i.e., Theorem I) is uniquely specified by giving its
four divergence and its four-curl. That is one must specify the following:
∂ν Aν (xσ ) ≡ s, (30a)
i.e., a rank zero source density and
∂ α Aµ (xσ ) − ∂ µ Aα (xσ ) ≡ cαµ , (30b)
i.e., a rank two circulation density, for all xσ .

Proof: Theorem II is based on Theorem I and follows in a similar fashion as did Theorem V of Ref. 1. ♣
Condition (30b) coincidentally also implies that cαµ = −cµα , i.e., that cαµ is anti-symmetric. Substitution of (30a)
and (30b) into (29) now yields
Z Z
Aµ (x) = ∂ µ s (x0 ) G (x, x0 ) d4 x0 + ∂α cαµ (x0 ) G (x, x0 ) d4 x0 , (31)
V40 V40
for all xσ . So far the statements made are applicable to either of the two cases of four-vector fields previously defined.
If one now defines the four-vector field Aµ specifically to be the electromagnetic four-vector potential, it is at once
obvious that the rank two circulation density cαµ defined in (30b) is just the electromagnetic field tensor F αµ . Similarly
the condition (30a) bears a striking resemblance to the Lorentz condition, but with a nonzero rhs. Now, drawing from
the comments made in the previous two sections, if one is to eliminate the effectively anti-Hermitian scalar potential
φ from the Lagrangian density (4) in a relativistically invariant way and thus obtain a result that is classically
bounded from below, one is lead to the specification of the relativistically invariant Lorentz condition (19). The
Lorentz condition (19) and the Maxwell field tensor F αµ are both invariant under the restricted gauge transformation
defined by Eqs. (19) and (23) and which is usually referred to as the Lorentz gauge. Therefore, substituting the
electromagnetic field tensor with cαµ = F αµ , while choosing the Lorentz condition (19) with s = ∂ν Aν = 0, reduces
(31) to the following:
Z
Aµ (x) = ∂α F αµ (x0 ) G (x, x0 ) d4 x0 3 ∂ν Aν (xσ ) = 0 ∀xσ ∈ R3+1 . (32)
V40
The uniqueness statement (32) satisfies the requirement that the Lagrangian density is bounded from below as desired,
through the Lorentz condition, and in addition places an additional restriction on allowable gauge transformations as
is shown below.
Performing a gauge transformation of the second kind as defined in (17) on the integral in (32) then gives
Z
0µ µ µ
A (x) = A (x) − ∂ F (x) = ∂α F 0 αµ (x00 ) G (x, x00 ) d4 x00 , (33)
V40
where, to avoid confusion, single primed variables indicate the gauge transformation and double primed variables are
used to denote source point space-time variables. It is easy to demonstrate that the electromagnetic field tensor is
gauge invariant under the gauge transformation of the second kind (17) as follows:
F 0 αµ = ∂ α A0µ − ∂ µ A0α = ∂ α Aµ + ∂ α ∂ µ F − ∂ µ Aα − ∂ µ ∂ α F = F αµ , (34)
and so (33) with (34) substituted reduces to

Z
A0µ (x) = Aµ (x) − ∂ µ F (x) = ∂α F αµ (x00 ) G (x, x00 ) d4 x00 . (35)
V40
Comparison of the integral in the uniqueness statement (32) and the gauge transformed statement (35) then leads
one to the surprising conclusion that
∂ µ F (xσ ) = 0 ∀xσ ∈ R3+1 . (36)
In other words, instead of the arbitrary scalar gauge function F (xσ ) being interpreted as a free field which satisfies the
wave equation (23), and thereby forms part of the Lorentz gauge, one has instead by (36) that the gauge function F (xσ )
is constant or zero throughout Minkowski space. Therefore, the gauge transformation itself, under the restriction (36),
is reduced to
A0µ (xσ ) = Aµ (xσ ) , (37)

which is no more than an identity transformation. This result is related to the fact that the transverse vector
potential is always gauge invariant as shown in (24a), while the longitudinal vector potential and the scalar potential
are changed by a gauge transformation of the second kind and are therefore non-unique. Although these later two
degrees of freedom are effectively canceled by the Lorentz condition, they are not uniquely specified until an additional
gauge restriction like (36) is applied to restrict the arbitrary gauge function F (xσ ).
Definition: This combination of (32), (17), and (36) will be referred to as a restricted relativistic transverse gauge.
One concludes therefore that specifying Aµ via the four-curl of Aµ (i.e., the Maxwell field tensor F αµ ) in the
uniqueness statement (32), while simultaneously specifying the four-divergence of Aµ via the Lorentz condition (18),
indeed leads to a unique four-vector potential Aµ in that it reduces the gauge transformation of the second kind
(17) to the identity transformation (37) via the gauge restriction (36). One is tempted to say that the uniqueness
statement (32) is an over-specification of the vector potential Aµ since it eliminates the gauge freedom evident in
the gauge transformation (17). Of course, the above uniqueness
statement (32) is predicated on four-vector fields
Aµ which are sufficiently smooth, i.e., which are C 2 V 4 . However, these twice continuously differentiable functions
would appear to be the appropriate ones for Lagrange equations of motion which in the present context are second
order differential equations, e.g., the four-vector potential wave equation.
In the quantum field theory context, the gauge transformation (17) when combined with a local phase transformation
ψ(x) → ψ 0 (x) = exp(−iqF (x))ψ(x), (38)
where q is the charge, allows one to minimally couple the electromagnetic field to the electron’s Dirac field ψ(x),
(cf. Ref. 16). However, the integral equation in the uniqueness statement (32) is in fact a vector identity when the
Lorentz condition is assumed, as is the case in (32). The implication is that as soon as one imposes the Lorentz
condition, the classical electromagnetic vector potential is unique via (32)! Before the reader becomes too alarmed,
one should note that one is not actually allowed to impose the Lorentz condition in a strong fashion as an operator
identity in the quantum domain since it would conflict with a certain canonical commutation relation, (cf. Ref. 16).
The most one can do is to impose the Lorentz condition in a weaker fashion as a restriction on the allowable states
of the electromagnetic field as for example with
∂ν Aν(+) (x) |0i = 0, (39)
which restricts the positive frequency parts of ∂ν Aν , (cf. Ref. 16). Therefore, the uniqueness statement (32) simply
provides more evidence in support of the conclusion that the Lorentz condition is too strong a constraint in some
situations. However, as remarked before, if one does not impose the Lorentz condition in the classical domain for
Lagrangian densities like (4), which contain a Lorentz constraint term, the classical electromagnetic field is not
bounded from below.
D. Uniqueness of the massless classical four-vector field in the relativistic longitudinal gauge
In Sec. III A, a new class of classical four-vector fields was introduced which was characterized by the Lagrangian
(10) which differed from the usual electromagnetic-type Lagrangian by the introduction of an overall minus sign. In
this section, Theorems I and II will now be applied to this new class of classical four-vector fields in the relativistic
longitudinal gauge (again limited to the massless case).
The analysis parallels the electromagnetic case in Sec. III C. It will be convenient for comparing subsequent results
to their electromagnetic counterparts to substitute (in a purely formal way) the pure imaginary valued version of the
four-vector field, namely K µ = iK µ as previously defined in Sec. III A in connection with the Lagrangian density
(8). This notation will suppress the overall minus sign in the results. Therefore, one obtains for the four-divergence
condition, instead of (30a), the following:
∂ν K ν (xσ ) ≡ sgi , (40a)
i.e., a rank zero source density, and for the four-curl condition, instead of (30b), the following:
∂ α K µ (xσ ) − ∂ µ K α (xσ ) ≡ cαµ

gi , (40b)
i.e., a rank two circulation density, for all xσ . Similarly, instead of (31), one has the following:
Z Z
K µ (x) = ∂ µ sgi (x0 ) G (x, x0 ) d4 x0 + ∂α cαµ 0 0 4 0
gi (x ) G (x, x ) d x , (41)
V40 V40
for all xσ .
Now, (40b) is analogous to the electromagnetic field tensor and so one can define the new field tensor
Gαµ ≡ ∂ α K µ (xσ ) − ∂ µ K α (xσ ) . (42)
For consistency with the definition (9), the components of Gαµ in SI units can be defined as
0 −Gx /c −Gy /c −Gz /c

 
 G /c 0 −Iz Iy 
Gαµ = x . (43)
Gy /c Iz 0 −Ix 
Gz /c −Iy Ix 0
Next, one can define a Lagrangian density (as in (8) but retaining the pure imaginary variables) for the K µ field
analogous to the relativistically invariant Stueckelburg Lagrangian density (1), but with the Lagrange multiplier
λ = 1 and with the mass parameter µg = 0, as follows:
go c2 go c2 2
Lgi = − Gµν Gµν − (∂µ K µ ) − jgµ K µ , (44)
4 2
where jgµ is the lowered index (pure imaginary valued) four-current density. As in the previous section for the
electromagnetic field, one now seeks to eliminate any effectively anti-Hermitian degrees of freedom from (44), so as to
obtain a Lagrangian density that is classically bounded from below. However, one can not specify a Lorentz condition
in this case by specifying sgi = 0 in (40a) since this would eliminate the only effectively Hermitian degree of freedom,
i.e., associated with the positive sign in the κ time derivative term in (8). In addition, specification of the longitudinal
gauge (27), while eliminating the two transverse degrees of freedom KT in a gauge invariant fashion, still leaves the
longitudinal degree of freedom KL intact. Therefore, one of the three effectively anti-Hermitian degrees of freedom,
i.e., associated with the negative sign in the K time derivative term in (8), would still remain. In addition, the
longitudinal gauge (27) is not a relativistically covariant constraint.
The longitudinal gauge (i.e., ∇×K = 0) will therefore be generalized to a relativistically invariant condition. A
condition between KL and κ, analogous but quite different than the Lorentz condition, will be shown to follow from
this relativistic generalization of the longitudinal gauge. This condition will enable one to eliminate the longitudinal
degree of freedom in a relativistically invariant way. In addition, this condition will still eliminate the transverse
degrees of freedom just as in the longitudinal gauge. This generalization will also shed new light on the theory of the
relativistic transverse four-vector field.
Consider the relativistic generalization of the Coulomb or transverse gauge of (20) to the Lorentz condition or
“relativistic transverse gauge” of (18), as follows:
1 ∂φ
∇ · A = 0 −→ ∇ · A + = 0, (45a)
c2 ∂t
or in tensor notation as
∂k Ak = 0 −→ ∂µ Aµ = 0. (45b)
The important thing to note about Eqs. (45a) and (45b) is that the three-divergence of A is generalized into a
relativistic four-divergence of Aµ . Thus, a solenoidal vector field A (i.e., the transverse vector field of (13b)), is
generalized into a four-solenoidal four-vector field Aµ .
In analogy with (45), one therefore surmises that the longitudinal gauge of (27), which includes (13a), has the
relativistic generalization
∇×K = 0, i.e., ∂ i K j − ∂ j K i = Gij = 0 −→ ∂ µ K ν − ∂ ν K µ = Gµν = 0, (46)
where the association of I = ∇×K via (9) with the space part of the field tensor Gµν of (43) has been used.
Definition: The name “relativistic longitudinal gauge”1 will therefore be chosen for gauges which use this new
covariant gauge constraint
Gµν = ∂ µ K ν − ∂ ν K µ = 0. (47)
[Kalb and Ramond already in 1974 used a condition like (47) in their often quoted investigation of string
interactions.17 ]
The important thing to note about (46) is that the three-curl of K is generalized into a relativistic four-curl of K µ .
Thus, an irrotational vector field K (i.e., the longitudinal vector field of (13a)), is generalized into a four-irrotational
four-vector field K µ . Now, it already follows from I = ∇×K = 0 via (27) and (28) that KT = 0. Setting the
remaining components of Gµν to zero yields the additional condition
∂K
G = −∇κ − = 0, (48)
∂t
with its reciprocal space version,
∂KL (k, t) ∂KT (k, t)
G(k, t) = −ikκ(k, t) − − = 0, (49)
∂t ∂t
where κ(r, t) ↔ κ(k, t). Since −ikκ is a longitudinal component, (49) implies that the longitudinal and transverse
components of G(k, t) are each separately equal to zero, giving as a result
∂KL (k, t)
G L (k, t) = −ikκ(k, t) − = 0, (50a)
∂t
∂KT (k, t)
G T (k, t) = − = 0. (50b)
∂t
Now, taking the inverse Fourier spatial transforms of Eqs. (50a) and (50b) yields
∂KL
−∇κ = , (51a)
∂t
∂KT
= 0. (51b)
∂t
Equation (51a) is the sought after condition relating the longitudinal vector potential and the scalar potential analo-
gous to the Lorentz condition (18) but with the space and time derivatives switched with respect to KL and κ ! Also,
equation (51b) is consistent with the previous requirement that KT = 0 in the nonrelativistic longitudinal gauge.
Now, all three of the effectively anti-Hermitian degrees of freedom, i.e., associated with the negative sign in the K time
derivative term in (8), are successfully eliminated from the Lagrangian density (44) by the relativistic longitudinal
gauge (47). Substituting (47) using (42) and (40b), while also substituting (40a), into (41) then yields
Z
K µ (x) = ∂ µ (∂ν0 K ν (x0 )) G (x, x0 ) d4 x0 3 Gµν = ∂ µ K ν − ∂ ν K µ = 0 ∀x ∈ R3+1 . (52)
V40
The result (52) is a preliminary uniqueness statement for the four-vector field K µ . Although satisfying the requirement
that the Lagrangian density (44) is bounded from below through the relativistic longitudinal gauge, an additional
restriction on allowable gauge transformations is required in order to yield a unique four-vector field as is shown below.
Performing a gauge transformation of the second kind, defined in analogy to the electromagnetic field case (17), as
K 0µ (x) = K µ (x) − ∂ µ Fg (x), (53)
on the integral in (52) then gives

Z
K 0µ (x) = K µ (x) − ∂ µ Fg (x) = ∂ µ (∂ν00 (K ν (x00 ) + ∂ 00ν Fg (x00 ))) G (x, x00 ) d4 x00 , (54)
V40
where, to avoid confusion, single primed variables indicate the gauge transformation and double primed variables are
used to denote source point space-time variables. In order for (54) to be identical in form to (52), and thus gauge
invariant, it is sufficient to take as a gauge restriction the following:
∂ν ∂ ν Fg (xσ ) = 0. (55)
Interestingly, (55) yields a gauge function Fg (xσ ) which satisfies the same requirement, i.e., that it be a free field,
as for the case of the Lorentz gauge in electromagnetism (cf. (23)). In the electromagnetic case the restriction on
the gauge function F (xσ ) in (23) followed from a requirement that the Lorentz condition (18) be gauge invariant in
some particular gauge (since it is not gauge invariant in all gauges). In the present case, however, the rank zero scalar
∂ν K ν (xσ ) 6= 0 initially, or in general. Therefore, the gauge restriction (55) only implies that the four-divergence of
K µ is gauge invariant in this special gauge as the following demonstrates:
∂µ K 0µ (x) = ∂µ K µ (x) − ∂µ ∂ µ Fg (x) = ∂µ K µ (x). (56)
The trivial gauge restriction
∂ µ Fg (xσ ) = 0, (57)
which makes the gauge transformation (53) an identity transformation, also leads to the result (56). Recall that the
transverse vector potential is always gauge invariant and that in the present case the constraint (47) requires that is
is always zero. On the other hand, the longitudinal vector potential and the scalar potential are changed by a gauge
transformation of the second kind when the arbitrary gauge function Fg (xσ ) remains unrestricted or even partially
restricted as in (55). These later two degrees of freedom are therefore non-unique until an additional gauge restriction
like (57) is applied. Therefore, a complete uniqueness statement for the four-vector field K µ is comprised of the
preliminary uniqueness statement (52), the gauge transformation (53), and the trivial gauge restriction (57).
Definition: This combination of (52), (53), and (57) will be referred to as a restricted relativistic longitudinal gauge.
It should also be noted that the relativistic longitudinal gauge condition (47) is of course gauge invariant in any
gauge defined by (53) as a calculation analogous to the electromagnetic one (34) would clearly show. The relativistic
longitudinal gauge condition (47) is therefore both relativistically invariant and gauge invariant as desired. However,
judging from the quantum electrodynamic case as discussed in relation to (38) and (39), it is likely that the gauge
condition (47) is too strong a constraint to impose quantum mechanically. One would then impose the relativistic
longitudinal gauge constraint (47) in a weaker fashion as a restriction on the allowable states of the (formally anti-
Hermitian) field K µ as for example with
(∂ µ K ν(+) (x) − ∂ ν K µ(+) (x)) |0i = 0, (58)
which restricts the positive frequency parts of Gµν . The restriction (58) amounts to several constraint equations.
These constraint equations are needed to eliminate the three effectively anti-Hermitian degrees of freedom associated
with the three-vector potential K, thereby leaving only the effectively Hermitian degree of freedom associated with
the scalar potential κ, as desired.
IV. CANONICAL FORMULATION OF THE RELATIVISTIC LONGITUDINAL CLASSICAL

FOUR-VECTOR FIELD
A. The case of four-vector current coupling
Based on the gauge considerations in Sec. III, a classical free space Lagrangian density for a massive four-vector
field in the relativistic longitudinal gauge would most likely be of the type (8), having a built in Lorentz constraint
term (now multiplied by the arbitrary parameter λg ) and a small g-photon mass term with a Compton wave number
µg . Also, the constant go is the “permittivity” of free space in SI units for this four-vector field. Therefore, one has a
so-called Stueckelberg Lagrangian2,3
2
go c2 G2 go c2 1 ∂κ go c2 µ2g κ2

2 2
Lgi = − I − λg + ∇·K + − K − ρg κ + jg · K, (59)
2 c2 2 c2 ∂t 2 c2
which is no longer multiplied by an overall minus sign as in (8), since all the fields and sources are now taken for
notational convenience as pure imaginary valued. Consequentially, relativistic longitudinal and transverse fields are
covered here in one comprehensive formalism with only a change of variables required to switch between the cases.
The case of pseudo-four-vector current coupling is discussed in Sec. IV C.
Now, the four-vector potential K µ is related to the vector fields G and I in the same way as in (9), i.e.,
∂K
G = −∇κ − , I = ∇×K. (60)
∂t
Also, the field tensor Gµν is defined in (42) and (43), and so the Lagrangian density (59) in covariant notation is
go c2 go c2 2 go c2 µ2g
Lgi = − Gµν Gµν − λg (∂µ K µ ) + (Kµ K µ ) − jgµ K µ . (61)
4 2 2
The Euler-Lagrange equations of motion as defined by

∂Lgi ∂Lgi
∂µ − = 0, (62)
∂ (∂µ Kν ) ∂Kν
follow from the Lagrangian density (61) as
∂α ∂ α + µ2g K µ − (1 − λg ) ∂ µ (∂σ K σ ) = jgµ /go c2 .

(63)
Taking the four-divergence of (63) yields
∂α ∂ α + µ2g ∂µ K µ − (1 − λg ) ∂µ ∂ µ (∂σ K σ ) = ∂µ jgµ /go c2 .

(64)
Simplifying (64) and assuming a conserved (four-vector) current ∂µ jgµ = 0 yields
λg ∂α ∂ α + µ2g ∂µ K µ = ∂µ jgµ /go c2 = 0.

(65)
Thus, for λg 6= 0 and a conserved current the four-divergence ∂µ K µ forms a free scalar field which satisfies the
Klein-Gordon equation with a “mass” (i.e., wave number),
p
ms = µg / λg , λg > 0, (66)
where λg > 0 has been assumed to ensure that m2s > 0.

One can now define, by analogy with what has been referred to for a massive electromagnetic-type theory as a
“transverse” field,2 the following:
λg
KµT = Kµ + ∂µ (∂σ K σ ) . (67)
µ2g
Taking the four-divergence of (67) gives
λg µ
∂ µ KµT = ∂ µ Kµ + ∂ ∂µ (∂σ K σ ) = 0, (68)
µ2g
which is zero by (65). Therefore, one finds that the field KµT is as a result divergenceless. Rearranging (67) slightly
into
λg
Kµ = KµT − ∂µ (∂σ K σ ) , (69)
µ2g
implies that K µ can be split into so-called “transverse” and “scalar” parts. After canonical quantization, the usual
analysis of the massive electromagnetic-type theory leads to an indefinite metric with different masses for the vector
and scalar components of its creation and annihilation operators.2 In the present case, however, one should expect,
in light of this paper, that the three spatial vector components will be the ones leading to an indefinite metric, as
opposed to the zeroth or scalar component in the usual analysis.
The classical canonical formalism now proceeds with a calculation of the covariant canonical momentum, defined
in terms of the Lagrangian density (61), as follows:

∂Lgi K µ , ∂ β K µ , t
Πνgi (xµ ) ≡ , (70)
∂ (∂0 Kν )
which yields [cf. (73) for index µ = 0, or (3-100) in Ref. 2],
Πνgi (xµ ) = −go c2 ∂ 0 K ν − ∂ ν K 0 − λg go c2 g 0ν ∂σ K σ .

(71)
Clearly, the covariant canonical momentum Πνgi is a function of both ∂σ K σ and the time components of Gµν . This
can be brought into sharper focus by defining a second rank canonical momentum tensor density as follows:18

σ 2
2 ∂ (∂σ K )
α β α

∂L K , ∂ K , t 2 ∂ G Gλρ
gi go c λρ go c
Πµν α
gi (x ) ≡ =− − λg . (72)
∂ (∂µ Kν ) 4 ∂ (∂µ Kν ) 2 ∂ (∂µ Kν )
Only terms with λ = µ and ρ = ν or with λ = ν and ρ = µ survive the contraction on λ and ρ, which since
Gµν = −Gνµ yields
Πµν α 2 µν
gi (x ) = −go c G − λg go c2 (∂σ K σ ) g µν (73)
as the second rank canonical momentum tensor density. From this point of view, the relativistic transverse gauge (i.e.,
the Lorentz condition) causes the symmetric part of Πµν gi , here the diagonal elements, to vanish. On the other hand,
the relativistic longitudinal gauge (47) (i.e., Gµν = 0) causes the anti-symmetric part of Πµν gi to vanish. In passing, it
is interesting to note that (72) and (73) can be used to express the Lagrange equation of motion (63) via (62) in the
more compressed notation:

µν ∂Lgi ∂Lgi
∂µ Πgi = ∂µ = = −jgν + go c2 µ2g K ν . (74)
∂ (∂µ Kν ) ∂Kν
Thus, (73) can be thought of as a generalization of the Maxwell field tensor Gµν to include a symmetric part, i.e.,
including diagonal elements associated with the momentum canonically conjugate to the scalar component K 0 = κ/c.
A further insight into the distinction between the “gauges” can be obtained from a calculation of the stress energy-
momentum tensor of the four-vector field K µ . One first calculates the canonical stress tensor
∂Lgi
e µν =
Θ ∂ ν Kλ − g µν Lgi = Πµλ ν µν
gi ∂ Kλ − g Lgi . (75)
∂ (∂µ Kλ )
With the result (73) and the Lagrangian density (61) of the field coupled to an external current, one obtains
2
e µν = − go c2 Gµλ (∂ ν Kλ ) − λg go c2 (∂σ K σ ) g µλ (∂ ν Kλ ) + g µν go c Gλρ Gλρ
Θ
4
2 2 2
go c µν 2 go c µg µν
+ g µν (jgσ K σ ) + λg g (∂σ K σ ) − g (Kσ K σ ) . (76)
2 2
The first term of (76) is nonsymmetrical and nongauge invariant and so in the usual analysis of the electromagnetic
field, one carries out a symmetrization process where, due to the nonuniqueness of the Lagrangian, a four-divergence
is added which is equal to a vanishing surface term by the four-dimensional form of Gauss’ divergence theorem. In
the present case, the second term also has the appearance of a nonsymmetrical term. But, in light of the relativistic
longitudinal gauge (47), this term in the stress tensor is nonzero and is actually symmetrical, i.e., ∂ ν K µ = ∂ µ K ν ,
for the relativistic longitudinal field case. In the relativistic transverse field case, one of course imposes the Lorentz
condition and this term vanishes. Nevertheless, in order to show the consistency of the entire extended Lagrangian
formulation (61) under either relativistically transverse or longitudinal gauge constraints, the usual symmetrization
procedure will be carried out by adding a four divergence to (76) as follows:
Θµν = Θ
e µν + go c2 ∂ρ (Gµρ K ν ) = Θ
e µν + go c2 Gµρ (∂ρ K ν ) + go c2 (∂ρ Gµρ ) K ν . (77)
Substitution of the canonical stress tensor (76) into (77) then yields
go c2
Θµν = − go c2 Gµλ (∂ ν Kλ − ∂λ K ν ) + g µν (jgσ K σ ) − λg go c2 (∂σ K σ ) (∂ ν K µ ) + g µν Gλρ Gλρ
4
go c2 µν 2 go c2 µ2g µν
+ λg g (∂σ K σ ) − g (Kσ K σ ) + go c2 (∂ρ Gµρ ) K ν . (78)
2 2
Substitution of (42) into (78) with − Gµλ Gνλ = − Gµλ Gνλ = Gµλ Gλν = Gµλ Gλν then yields
g µν

Θµν = go c2 Gµλ Gλν + Gλρ Gλρ − λg go c2 (∂σ K σ ) (∂ ν K µ ) + g µν (jgσ K σ )
4
2
go c µν 2 go c2 µ2g µν
+λg g (∂σ K σ ) − g (Kσ K σ ) + go c2 (∂ρ Gµρ ) K ν (79)
2 2
as the new “symmetrized” energy-momentum tensor density. One now breaks up Θµν using the usual definition of
the pure electromagnetic-type energy-momentum tensor Θµν T , which could be regarded as representing a “massless”
transverse part of the field, along with the definition of a new Θµν L , which could be regarded as representing a “massive”
longitudinal part of the field. External current terms in general and certain terms associated with the transverse field’s
mass in particular are left separate, along with one other term yielding
jgµ

Θµν = Θµν T + g µν
(jgσ K σ
) + Θµν
L + go c2
λ g ∂ µ
(∂σ K σ
) + µ2 µ
g K − Kν , (80)
go c2
where the equation of motion (63), rewritten using (42) as
jgµ
∂ρ Gµρ = λg ∂ µ (∂σ K σ ) + µ2g K µ − , (81)
go c2
has been substituted in the last term of (79), and where the following definitions are used:
g µν

µν 2 µλ ν λρ
ΘT ≡ go c G Gλ + Gλρ G , (82)
4
go c2 µν 2 go c2 µ2g µν
Θµν 2 σ ν µ
L ≡ −λg go c (∂σ K ) (∂ K ) + λg g (∂σ K σ ) − g (Kσ K σ ) . (83)
2 2
In the relativistic transverse gauge, the Lorentz condition reduces (80), yielding the total transverse energy-
momentum tensor density
go c2 µ2g µν jgµ

µν µν µν σ σ 2 2 µ
ΘT,total ≡ ΘT + g (jgσ K ) − g (Kσ K ) + go c µg K − Kν , (84)
2 go c2
where the third term of (84) is an additional mass term that comes from Θµν L . For µg = 0, (84) is the usual
electromagnetic-type energy-momentum tensor in the presence of an external current, (cf. pp. 22-25 of Ref. 2).
Taking a four divergence of the pure electromagnetic-type “massless” transverse energy-momentum tensor (82) yields
the force density fTν as
g µν

ν µν 2 µλ ν λρ
−fT ≡ ∂µ ΘT = go c ∂µ G Gλ + Gλρ G
4
go c2 λρ ν
= go c2 ∂µ Gµλ Gλν + go c2 Gµλ (∂µ Gλν ) + 2

G (∂ Gλρ ) , (85)
4
where the minus sign on the left-hand side (lhs) of (85) is chosen in the (+ − −−) signature so as to produce the right
sign in the Lorentz-type force law for the transverse field. Using the equation of motion (81) and the anti-symmetry
of Gµλ , Eq. (85) becomes
!
ν 2
jgλ λ σ 2 λ go c2 λρ ν
−fT = go c 2
− λg ∂ (∂σ K ) − µg K Gλν + go c2 Gµλ (∂µ Gλν ) + G (∂ Gλρ ) . (86)
go c 2
But, the last two terms of (86) can be rewritten as
go c2 λρ ν
go c2 Gµλ (∂µ Gλν ) + G (∂ Gλρ )
2
go c2
Gµλ (∂µ Gλν ) + Gµλ (∂µ Gλν ) + Gµλ (∂ ν Gµλ )

=
2
go c2 go c2
Gµλ ∂µ Gλν − ∂λ Gνµ = Gµλ ∂µ Gλν + ∂λ Gµν = 0,

= (87)
2 2
where the last equality is zero since it is a contraction of an anti-symmetric and a symmetric factor in µ and λ, and
where the second line of (87) has been transformed into the third line using the source free Maxwell-type equations
written in covariant form and contracted by gµρ and gλσ as follows:
0 = gµρ gλσ (∂ ν Gρσ + ∂ ρ Gσν + ∂ σ Gνρ ) = ∂ ν Gµλ + ∂µ Gλν + ∂λ Gνµ . (88)
Therefore, with the result (87), Eq. (86) reduces to
fTν = −jgλ Gλν + go c2 λg ∂ λ (∂σ K σ ) + µ2g K λ Gλν .

(89)
ν
The first term of (89) is a Lorentz-type force density fLorentz = −jgλ Gλν . Naturally, when in the Lorentz gauge
and when µg = 0, this is the only surviving term. A slightly more involved four-divergence of the total transverse
energy-momentum stress tensor (84), that is including the other terms associated with the external currents and field
mass, yields
ν
≡ ∂µ Θµν λν
+ ∂ ν jgλ K λ − jgλ ∂ λ K ν = (∂ ν jgµ ) K µ ,

−fT,total T,total = jgλ G (90)
which is of the same form as the result for the electromagnetic case, (cf. Eq. (1-118) on p. 25 in Ref. 2), except for
an implicit factor of −1 due to the quadratic appearance of the pure imaginary variables (in this notation).
In the relativistic longitudinal gauge, where Gµν = 0, (80) simplifies to
Θµν = Θµν
L +g
µν
(jgσ K σ ) , (91)
where the last term of (80) vanishes due to the vanishing of the lhs of the equation of motion (81) in the relativistic
longitudinal gauge. [It should be noted in passing that (91) can be obtained directly without using the symmetrization
procedure (77) since all the terms containing Gµν vanish in the relativistic longitudinal gauge.] Taking a four-
divergence of the pure field part of the energy-momentum tensor (91), i.e., of (83), gives
−fLν ≡ ∂µ Θµν 2 σ ν µ 2 σ ν µ
L = − λg go c (∂µ ∂σ K ) (∂ K ) − λg go c (∂σ K ) (∂µ ∂ K )
go c2 go c2 µ2g
+ λg 2 (∂σ K σ ) g µν ∂µ (∂σ K σ ) − 2Kσ g µν ∂µ K σ . (92)
2 2
Upon commuting some of the four-vector derivatives, the second and third terms of (92) cancel (in flat space-time)
yielding
−fLν ≡ ∂µ Θµν 2 σ ν µ 2 2 ν µ
L = −λg go c (∂σ ∂µ K ) (∂ K ) − go c µg Kµ ∂ K
= −go c2 λg ∂σ ∂ σ + µ2g Kµ (∂ ν K µ ) ,

(93)
where gµν Gνσ = 0 has been used on the first term. The result can be simplified more if one rewrites (81) as
jgµ jgµ
0 = ∂ν Gµν = λg ∂ µ (∂σ K σ ) + µ2g K µ − = λ ∂
g σ ∂ µ σ
K + µ2 µ
g K − , (94)
go c2 go c2
where commuted derivatives and (47) are used, so that with the application of the relativistic longitudinal gauge in
the form ∂ µ K σ = ∂ σ K µ one has
jgµ
λg ∂σ ∂ σ + µ2g K µ =

(95)
go c2
as the new field equation of motion in the relativistic longitudinal gauge. Therefore, using (95), Eq. (93) reduces
finally to
−fLν ≡ ∂µ Θµν ν µ µ ν
L = −jgµ (∂ K ) = −jgµ (∂ K ) , (96)
as the force density of the pure field energy-momentum tensor (83) in the relativistic longitudinal gauge, where either
of the expressions in (96) are equivalent since Gµν = 0. Taking the four-divergence of (91) instead, which includes
the effect of the external current while using the result (96) for the longitudinal pure field part gives
−f ν ≡ ∂µ Θµν = ∂µ Θµν µν σ ν µ ν µ ν µ
L + g ∂µ (jgσ K ) = (∂ jgµ ) K + jgµ (∂ K ) − jgµ (∂ K ) , (97)
which yields the same result as for the transverse field (90), namely,
−f ν ≡ ∂µ Θµν = (∂ ν jgµ ) K µ . (98)

It is clear, however, that the relativistic longitudinal gauge has radically changed the nature of the force law associated
with the pure field case to (96). Compare with the relativistic transverse pure field case (89) with its first term as the
familiar Lorentz-type force law.
It is instructive to write out the components of the second version of (96) as
1 ∂K ν
fLν = jgµ (∂ µ K ν ) = ρc − jg · (−∇K ν )
c ∂t
dK ν

∂
=ρ + u · ∇ Kν = ρ , (99)
∂t dt
where jgµ = (ρc, jg ) = (ρc, ρu), with u as the ordinary velocity of the charge density ρ, and where use has been made
of the convective derivative,
d ∂
= + u · ∇. (100)
dt ∂t
Although one can express the four-current as jgµ = ρ(c, u) = ρv µ using a non-invariant four component velocity
v µ ≡ (c, u), .
it is more convenient to express the four-current in terms of the invariant four-velocity uµ ≡ (γc, γu),
p
where γ ≡ 1 1 − |u|2 /c2 , as
jgµ = ρo (γc, γu) = ρo uµ . (101)

Use has been made of the formula ρ = γρo in (101), which reflects a Lorentz contraction in the direction of motion of
the volume which contains the invariant proper rest charge density ρo of the g-charges. The form (99), with ρ = γρo
substituted for the charge density, is then especially convenient for expressing the power density or zeroth component
of the four-force law as
dK 0

0 γρo dκ d E γ dE
fL = γρo = = = , (102)
dt c dt dτ c c dt
where τ is the proper time and where d/dτ = γ d/dt. The energy density then follows easily from (102) as E =
ρo κ + constant. Next consider the first of the two equivalent versions of (96) written out in components as
fLν = jgµ (∂ ν K µ ) = ρc ∂ ν K 0 + jgj ∂ ν K j = ρ ∂ ν κ + jgj ∂ ν K j . (103)
One can then express the four-force f ν in terms of the proper rest mass density ζo as
duν duν
f ν = ζo = ζo γ . (104)
dτ dt
Equation (104) follows from the canonical formalism if the term − 21 ζo uµ uµ is added to the Lagrangian density.
Combining (104) and (103), while using the formula ρ = γρo and Eq. (101), then yields
duν
ζo γ = fLν = γρo ∂ ν κ + ρo uj ∂ ν K j . (105)
dt
This form of the pure field four-force in the relativistic longitudinal gauge is especially convenient for expressing the
three-force density or space components of fLν as follows:
duk
ζo γ = fLk = γρo ∂ k κ + ρo uj ∂ k K j . (106)
dt
[Although (106) looks different than the space components of (99), i.e., the second version of (96), they are really the
same in the relativistic longitudinal gauge where the second terms are equal by the three-vector version of (46), and
the first terms are equal by (51a).] Using the three-vector version of (46) on the second term of (106) and writing the
result in three-vector notation yields
d
ζo (γu) = −ρo ∇κ + ρo (−u · (−∇)) K = −ρo ∇κ + ρo (u · ∇)K, (107)
dt
where a factor of γ has been canceled in writing (107). Multiplying (107) by the rest volume Vo which contains all of
the g-charges and masses, while simultaneously assuming only one particle is present in Vo , allows one to obtain the
single particle force equation using the formulas g = ρo Vo and mo = ζo Vo as follows:
d
mo (γu) = −g∇κ + g (u · ∇) K. (108)
dt
It is desirable, however, to eliminate the effectively anti-Hermitian vector potential K from (108) so as to express the
force law entirely in terms of the effectively Hermitian component κ = cK 0 . In this regard, it is convenient to set the
Lorentz constraint parameter in the Lagrangian density (61) to λg = 1, so that the Lagrange equation of motion (63)
becomes
∂α ∂ α + µ2g K µ = jgµ /go c2 ,

(109)
where it is entirely analogous to the case of a massive electromagnetic-type field in the Lorentz gauge. If one takes
the case where the mass parameter µg = 0 in the wave equation (109), one can obtain the usual retarded potential
solutions. In the case of a point charge, it is well known that the retarded potential solutions reduce to the Liénard-
Wiechert potentials, which in the g-charge case are written as follows:
g u
K(r, t) = 2
, (110a)
4πgo c r − (u · r)/c
g 1
κ(r, t) = , (110b)
4πgo r − (u · r)/c
where u is the ordinary velocity of the charge g at the source point. [Note however that the Liénard-Wiechert potentials
themselves are known to satisfy the Lorentz condition, i.e., the relativistic transverse gauge. Additionally the Liénard-
Wiechert potentials lead in general to a nonzero field tensor Gµν . Therefore the Liénard-Wiechert potentials do not
appear to be appropriate for the case of a relativistic longitudinal field where Gµν = 0 and one wants ∂σ K σ 6= 0.
Nevertheless, it is interesting to explore their use in the present situation to see if it leads to some kind of useful
ansatz. In Sec. IV B a four-vector potential example which does satisfy the relativistic longitudinal gauge will be
explored.] Rewriting (110a) as
g u uκ(r, t)
K(r, t) = = (111)
4πgo c2 r − (u · r)/c c2
demonstrates that, at least for the case of a point charge and a massless field, the vector potential K can be eliminated
in favor of the scalar potential κ in a straightforward manner. However, it was shown in (27), (28), and (46) that
in the relativistic longitudinal gauge the transverse vector potential KT = 0, and so K is purely longitudinal, i.e.,
K = KL and must satisfy (27). Equation (111) therefore reduces to
κ
b = uL κ ,

K = KL = k b·K k b= k b·u k (112)
c2 c2

where the longitudinal velocity uL ≡ k b·u k b is the projection of the ordinary velocity u in the direction of the wave
number unit vector k. b Substitution of (112) into (108) yields

b2 · u2 k
k b 2 κ2
d
mo1 (γ1 u1 ) = −g1 ∇κ2 + g1 u1 · ∇ (113)
dt c2
as the force equation as following from the ansatz (111) for a point particle with rest mass mo = mo1 and charge
g = g1 in the presence of a massless four-vector field (and presumably still in the relativistic longitudinal gauge), but
written entirely in terms of (110b), the retarded scalar field κ2 of a second point charge g = g2 . Note that since each
g-charge is pure imaginary valued in this notation, the force law (113) includes a factor of i2 = −1, indicating that
the force is predominantly attractive between like g-charges.
It is interesting at this point to compare the preceding results with a force law derived directly from an interaction
Lagrangian formulation. Consider the single particle Lagrangian
q
Lsp = −mo1 1 − u21 /c2 + LI . (114)
The interaction Lagrangian density of a four-vector field (κ2 /c, K2 ) of a second particle interacting with a particle of
charge g1 and rest mass mo1 is presumably of the form
LI = −g1 κ2 + g1 u1 · K2 (115)
(i.e., LI = −jgµ K µ ). The single particle Lagrangian (114) with (115) substituted yields as its Euler-Lagrange equations
of motion a Lorentz-type force law

d(γ1 mo1 u1 ) ∂K2
= g1 −∇κ2 − + g1 u1 ×(∇×K2 ). (116)
dt ∂t
In the relativistic longitudinal gauge this Lorentz-type force vanishes since the two bracketed terms on the rhs of (116)
ν
vanish via (46) and (51), (or by (47) substituted into a Lorentz-type force density fLorentz = −jgλ Gλν ). Therefore the
force law (116) is not consistent with the force densities (108) or (113) which followed from the manifestly covariant
stress tensor formulation in the special case of a massless field. One could regard this inconsistency as ruling out an
interaction of a relativistic longitudinal field with a conserved g-charge four-current jgµ , arguing that the canonical
momentum density associated with the K 0 = κ/c component of the field Lagrangian density (59), namely
Pgi0 ∂Lgi
= = −λg go ∂µ K µ , (117)
c ∂(∂κ/∂t)
is a free field due to (65). This is in contradistinction to a relativistic transverse field where the canonical momentum
density
∂Lgi
Pgik = = −go Gk , (118)
∂ (∂K k /∂t)
associated with the K k components is not a free field since
∂α ∂ α + µ2g Gµν = ∂α ∂ α + µ2g (∂ µ K ν − ∂ ν K µ ) = ∂ µ jgν − ∂ ν jgµ /go c2 ,

(119)
which follows from (63) in a Lorentz gauge. The G0i components (defined in (43)) are proportional to the momenta
canonically conjugate to the K i components and so by (119) they are not associated with a free field. Also, the
Stueckelburg Lagrangian density (59), despite being the most general quadratic Lagrangian for a four-vector field, is
not gauge invariant due to the field mass term and the Lorentz constraint term. As a result the force density (96)
in the relativistic longitudinal gauge is not gauge invariant either, and is dependent on the gauge function Fg of the
gauge transformation (53), even in a relativistic longitudinal gauge defined by (52), (53), and (55), as the following
demonstrates:
fL0ν = jgµ (∂ ν K 0µ ) = jgµ (∂ ν K µ + ∂ ν ∂ µ Fg ) , (120a)
fL0ν = jgµ (∂ µ K 0ν ) = jgµ (∂ µ K ν + ∂ µ ∂ ν Fg ) , (120b)
where both of the equivalent forms of (96) have been gauge transformed in (120). This dependence on the gauge
function Fg (when Fg is nonconstant), can only be removed by removing the interaction term −jgµ K µ from the
Lagrangian density (61), thereby precluding an interaction with a conserved four-vector current jgµ . The inconsistency
of the force laws is then also removed as well.
It appears then that an interaction with a conserved four-vector current of g-charges jgµ would be ruled out and
consequently a relativistic longitudinal four-vector field would appear to be best described as a free field. A study of
a theory including a fourth-order self-interaction term
(4) go c2 (4) 2
LI =+ λ (Kµ K µ ) , (121)
4! I
(4)
analogous to a φ4 scalar field theory, is therefore warranted (but is not done here). [The sign of the LI term is
chosen so that the zeroth or scalar components agree with the φ4 scalar field theory.]
Now, consider again Θµν , the new “symmetrized” energy-momentum tensor density (80) composed of (82), the pure
electromagnetic-type energy-momentum tensor Θµν T representing a massless transverse part of the field, along with
(83), the tensor Θµν
L representing the longitudinal part of the field. [Note, however, that the relativistic longitudinal
stress tensor Θµν
L defined in (83) also suffers from the problem outlined in (120), i.e., that it is dependent on the gauge
function Fg of the gauge transformation (53), even in a relativistic longitudinal gauge defined by (52), (53), and (55).
Despite this problem, a few calculations will be made for the sake of completeness.] When terms associated with the
external current are set to zero, the momentum density (in energy units) of (80) in the relativistic transverse gauge is
PTj = Θ0j = go c2 G0λ Gλj + go c2 µ2g K 0 K j

= go c2 ∂ 0 K i − ∂ i K 0 ∂i K j − ∂j K i + go c2 µ2g K 0 K j ,

(122)
since G00 = 0. The “transverse” field momentum (in momentum units) reduces to
Z
PT = d3 x go (GT ×I) + µ2g κK ,

(123)
which is the usual result plus an added mass term. On the other hand, the tensor Θµν of (80) reduces in the relativistic
longitudinal gauge to (91) due to the equation of motion (81). Then when terms associated with the external current
are set to zero, the momentum density (in energy units) follows from (83) as
PLj = Θ0j 2 σ j 0

L = −λg go c (∂σ K ) ∂ K . (124)
The space integral of (124) times 1/c is the longitudinal field momentum
λg go c2
Z Z
3 σ 0 3 1 ∂κ
PL = d x (∂σ K ) ∇K = d x (λg go ) 2 + ∇ · KL ∇κ, (125)
c c ∂t
since ∂ j → −∇. Interestingly, in the longitudinal case since the mass term in (83) is diagonal, there is no mass contri-
bution to the field momentum density (124), while in the previous case of the transverse field there is a nondiagonal
mass term which therefore shows up in (122). A similar calculation for a massive pure real scalar field would yield
only the partial time derivative term in (125), as well as an overall minus sign (cf. p. 45 of Ref. 4). The analogous
minus sign is however implied in (125) since κ is pure imaginary valued and so the first term in (125) is consistent
with a scalar field with coupling constant λg g0 . The second term of (125) is then basically a relativistic extension of
a massive scalar field.
Next, one can calculate the angular momentum of the K µ field in the relativistic longitudinal gauge. In an isolated
region, free of currents (assuming that they could be consistently coupled to the field in the first place), the longitudinal
energy-momentum stress tensor Θµν L is conserved, i.e.,
−fLν ≡ ∂µ Θµν ν µ µ ν
L = −jgµ (∂ K ) = −jgµ (∂ K ) = 0, (126)
by (96) when jgµ = 0. It is also symmetric by its definition (83), i.e., Θνµ µν
L = ΘL since G
µν
= 0 implies ∂ µ K ν = ∂ ν K µ .
These two properties allow one to define a covariant third-rank tensor (anti-symmetric in µ and ν)
ν µλ
M λµν ≡ xµ Θνλ
L − x ΘL , (127)
which is conserved (as ∂λ acting on (127) while using the symmetry of Θµν L clearly shows). The conservation of the
symmetric longitudinal energy-momentum stress tensor (83) and of the anti-symmetric third rank tensor (127) (which
together comprise the ten generators of the Lorentz group) implies that the classical theory of the four-vector field
K µ in the relativistic longitudinal gauge is invariant under the inhomogeneous Lorentz group. Next, the total angular
momentum JLµν is then the volume integral of the λ = 0 component of (127), that is,
∂JLµν
Z
JLµν = d3 x M 0µν = −JLνµ , = 0. (128)
∂t
The i, j components of this anti-symmetric tensor JLµν , with (83) substituted, gives
Z Z
ij 3 i j0 j i0 2
d3 x (∂µ K µ ) xi ∂ j − xj ∂ i K 0 .

JL = d x x ΘL − x ΘL = −λg go c (129)
In the center of mass frame of the particle this is also the intrinsic spin angular momentum, i.e.,18
SL1 = JL23 , SL2 = JL31 , SL3 = JL12 , SL0 = 0, (130)
so that in three-vector notation one has finally

Z
1 ∂κ
SL = λg go c d3 x + ∇ · K L (x×∇)κ. (131)
c2 ∂t
Note, it is possible to eliminate KL from (131) in favor of κ using (112) as an ansatz. However, it has not yet been
proved whether or not (112), which was derived only for the special case of a point particle interacting with a massless
field using Liénard-Wiechert potentials, holds in general. As in the earlier calculation of the field momentum (125),
a similar calculation for a massive pure real scalar field would yield only the partial time derivative term in (131), as
well as an overall minus sign. The analogous minus sign is implied in (131) again since κ is formally pure imaginary
valued and so the first term in (131) is consistent with a scalar field with coupling constant λg go . The second term
of (131) can then be interpreted as a relativistic extension of a massive scalar field.
Next, the Hamiltonian density of the four-vector field K µ is furnished by a calculation of the zero-zero component
of the energy-momentum tensor Θµν in the relativistic longitudinal gauge starting with (91) as follows:
Hgi ≡ Θ00 = g 00 (jgσ K σ ) + Θ00

L
λg go c2 00 2 go c2 µ2g
= −λg go c2 (∂µ K µ ) ∂ 0 K 0 + g (∂µ K µ ) − g 00 (Kµ K µ ) + jgµ K µ , (132)
2 2
or in component form with K = KL in the relativistic longitudinal gauge,
2
1 ∂K 0 1 ∂K 0 go c2 µ2g

Hgi = −λg go c2 − λg go c2 (∇ · KL ) (K 0 )2 − KL · KL

−
c ∂t c ∂t 2
2 0 2 0
λg go c2

λg go c 1 ∂K 1 ∂K
+ + λg go c2 (∇ · KL ) + (∇ · KL )2 + ρg cK 0 − jg · KL . (133)
2 c ∂t c ∂t 2
Canceling and combining terms yields the energy density,
2
λg go ∂K 0 go c2 µ2g λg go c2

Hgi = − − (K 0 )2 − KL · KL + (∇ · KL )2 + ρg cK 0 − jg · KL , (134)
2 ∂t 2 2
which is bounded from below since an implicit factor of minus one due to the pure imaginary valued field (i.e.,
(K 0 )2 = i2 (K 0 )2 = −(K 0 )2 , where K 0 is pure real) cancels the minus sign on the time partial derivative term. A
time derivative term involving KL , the badly behaved part of the four-vector field in this case, is not present in the
relativistic longitudinal gauge! In passing note that if one applies the ansatz (112) to Eq. (134), while substituting
jg1 = ρg1 u1 and κ2 = cK20 to distinguish the field from the current density as in (113), one would then obtain
2
go µ2g u22L

λg go ∂κ2 2 λg go u2L κ2 2 u1 · u2L
Hgi = − 2 − (κ2 ) 1 − 2 + ∇· + ρ g1 κ 2 1 − , (135)
2c ∂t 2 c 2 c2 c2
which is expressed entirely in terms of the well behaved degree of freedom κ.
In an entirely similar fashion one could apply the relativistic longitudinal gauge (47) to the Lagrangian density (61)
thus obtaining
go c2 2 go c2 µ2g
Lgi = −λg (∂µ K µ ) + (Kµ K µ ) − jgµ K µ . (136)
2 2
Since the individual terms in (136) are of the same functional form as in (132), then by the same arguments that
lead to (134) (or by setting the first bracketed term in (59) to zero) it is clear that the relativistic longitudinal
gauge Lagrangian density (136) is also bounded from below, and so is its associated action integral. And since the
above Hamiltonian and Lagrangian densities in the relativistic longitudinal gauge are bounded from below, the (pure
imaginary valued) classical four-vector field K µ satisfies the principle requirement for a physical field in the classical
domain.
A theorem can therefore be stated as follows:
Theorem III (Physical classes of classical four-vector fields): There are two and only two nontrivial classes
of sufficiently smooth classical four-vector fields in unbounded Minkowski space which under covariant constraint are
potentially physical, i.e., with a Lagrangian density that is bounded from below. The Lagrangian density itself is
assumed to be sufficiently general in the sense that it is fully quadratic in its variables and its derivatives as for
example with the Stueckelberg Lagrangians (1) and (61). These two classes are characterized as follows:
1. Four-Vector Fields Aµ : An ostensibly pure real valued four-vector field Aµ (xν ) which is defined by the Lagrangian
density (1), must be constrained by the relativistic transverse gauge ∂σ Aσ (xν ) = 0, i.e., by the Lorentz condition, in
order for it to be potentially physical. In other words a pure real valued four-vector field must be four-solenoidal in
order for it to be classically bounded from below.
2. Four-Vector Fields K µ : A formally pure imaginary valued four-vector field K µ (xν ) which is defined by the
Lagrangian density (61), must be constrained by the relativistic longitudinal gauge Gµν (xσ ) = ∂ µ K ν (xσ )−∂ ν K µ (xσ ) =
0 in order for it to be potentially physical. In other words, the four-vector field K µ (xν ) must be four-irrotational in
order for it to be classically bounded from below.
Proof: Consider that the Stueckelberg Lagrangians (1) and (61) are quadratic in quantities which in turn are linear
in the partial derivative ∂µ , namely F µν or Gµν and ∂µ Aµ or ∂µ K µ , respectively. Also, the specification of these two
types of linear combinations of ∂µ and the four-vector field, i.e., the four-curl and the four-divergence, over the entire
(unbounded) four-space volume of R3+1 uniquely specifies a sufficiently smooth four-vector field by Theorem II of
Sec III C. [And there is a corollary requiring only the additional specification of the (real) nonzero mass parameters
µ and µg which follows from Theorem XII of Ref. 1.] Specifically, there are two and only two linearly independent
constraints which are linear in the four-curl or the four-divergence and which remove time derivative terms from a
Stueckelberg Lagrangian. These two are the relativistic longitudinal gauge where one sets the four-curl to zero, and
the relativistic transverse gauge, i.e., the Lorentz condition, where one sets the four-divergence to zero. [The cases
where the four-curl or the four-divergence are set to a constant will be considered trivial (or linearly dependent).]
Now, only the following cases of the application of these two constraints can occur.
The four-vector field K µ in the relativistic longitudinal gauge has already been proved to be classically bounded
from below, (see the discussions accompanying the Hamiltonian density (134) and the Lagrangian density (136)).
Also, the field momentum (125) demonstrates that this case can be looked at as a relativistic extension of a scalar
field and so this is indeed a new four-vector field case (and not just the usual theory of a scalar field). For the
ostensibly pure real valued four-vector field Aµ in the relativistic transverse gauge, the Lorentz condition ∂µ Aµ = 0
reduces the Lagrangian density (1) to
o c2 o c2 µ2
L=− Fµν F µν + (Aµ Aµ ) − jµ Aµ . (137)
4 2
All the badly behaved time derivative terms, i.e., those involving the scalar potential A0 = φ/c with their leading
minus signs, are now eliminated from the Lagrangian density as can be seen by applying the Lorentz condition to
(4). The pure real four-vector field in the relativistic transverse gauge is therefore classically bounded from below,
as is well known. Applying both the relativistic longitudinal gauge and the Lorentz gauge simultaneously eliminates
all four of the degrees of freedom from a Stueckelberg Lagrangian and therefore leads to a trivial case. Applying the
relativistic longitudinal gauge to a pure real valued four-vector field, or applying the Lorentz gauge to a formally pure
imaginary valued four-vector field, leads to Lagrangian densities that are not bounded from below because the badly
behaved degrees of freedom from a Stueckelberg Lagrangian would be retained. This covers all the possible cases of
the application of the two covariant constraints. Therefore, only the covariantly constrained four-vector fields in the
two and only two classes specified in Theorem III are bounded from below, thus proving Theorem III. ♣
A few additional comments are in order. It should be apparent that a simultaneous application of the (nonrela-
tivistic) longitudinal gauge (with I = ∇×K = 0 eliminating all but the mass term in (123)), and the Lorentz gauge
(with ∂σ K σ = 0 causing the longitudinal field momentum (125) to vanish), leads to a trivial case since for a (massless
field) there would be no field momentum transport at all. Similarly, a simultaneous application of the (nonrelativistic)
longitudinal gauge (which implies KT = 0), and the Coulomb gauge (with ∇ · K = 0 implying KL = 0 and φ = 0),
leads to a trivial case. On the other hand, a massless real valued four-vector field in the Coulomb gauge leads to a two
degree of freedom theory which is classically bounded from below. But the Coulomb gauge is nonmanifestly covariant
and leads to essentially the same result in the massless field case as the covariant Lorentz gauge and this type of
four-vector field (e.g., electromagnetism) is already covered (covariantly) in Theorem III. It is reasonable therefore to
conclude that Theorem III covers all of the potentially physically interesting cases of classical four-vector fields.
B. Pure gauge and ghost fields
A four-vector field which does not need to involve the Liénard-Wiechert potentials will be briefly considered. In
fact this four-vector field actually satisfies a relativistic longitudinal gauge condition F µν = 0. Consider the “pure
gauge” field [i.e., a field which can be removed by the gauge transformation (17)]
Aµ ≡ ∂ µ Λ (138)
as discussed in connection with the Meissner effect deep in a superconductor, where the magnetic field is required to
vanish (cf. Ref. 19). This can be seen in the present context as following from the vanishing of the field tensor,
F µν = ∂ µ Aν − ∂ ν Aµ = ∂ µ ∂ ν Λ − ∂ ν ∂ µ Λ = 0, (139)
so that both the magnetic and electric fields vanish. Substituting the pure gauge field K µ ≡ ∂ µ Λ into a massless wave
equation gives
∂α ∂ α K µ = (∂α ∂ α ) ∂ µ Λ = jgµ /go c2 . (140)
An inhomogeneous solution of (140) can be obtained from a Green’s function technique as follows:
Z
0 0 1 ν 0ν
α µ (4)
(∂α ∂ ) ∂ Gµ (x, x ) = δ (x − x ) = d4 k e−ikν (x −x ) . (141)
(2π)4
The four-vector Green’s function Gµ then follows from the calculation

Z Z
−1 1 −ikν (xν −x0ν ) 1 kµ −ikν (xν −x0ν )
Gµ (x, x0 ) = ((∂α ∂ α ) ∂ µ ) 4
d 4
k e = 4
d4 k e , (142)
(2π) (2π) −i3 k 4
yielding the integral form of the Green’s function as

−ikµ −ikν (xν −x0ν )
Z
1
Gµ (x, x0 ) = d4 k e . (143)
(2π)4 k4
This integral can be performed as an integration in the complex plane in a manner similar to that used to obtain the
electromagnetic retarded potentials. The result is not required here. It is enough to say that the resulting Green’s
function yields the inhomogeneous solution for Λ via the integral
jgµ (x0 ) 4 0
Z
Λ(x) = Gµ (x, x0 ) d x. (144)
V40 go c2
However, as is well known, the pure gauge field does not couple to a conserved current. At any rate, the Green’s
function approach has yielded the momentum space Green’s function following from (143) as
e µ (k) = −ikµ ,
G (145)
k4
which is purely longitudinal in the direction of k µ .
“Ghost” fields in nonabelian gauge theory also appear to be of this “pure gauge” form (cf. Sec. 15.6 in Ref.
19). Ghost fields are useful in studying the renormalization properties of nonabelian gauge theories. Currently in
QCD these pure gauge ghost fields are enigmatically labeled as fermions due to the sign of the contributions to their
Lagrangian. This amounts to a violation of the spin-statistics theorem which is however ignored since the ghost fields
are not considered to be physical degrees of freedom. A reclassification of these states as relativistic longitudinal
gauge (formally anti-Hermitian) four-vector fields in SU(3) might restore the boson nature of pure gauge fields to
these particles, eliminating a current paradox. In passing note that the relativistic longitudinal gauge condition in
SU(n) would be defined as follows:
0 = Fµνa (x)ta = Fµν (x) ≡ ∂µ Aν − ∂ν Aµ − [Aµ , Aν ] , (146)
where the ta are elements of the Lie algebra. The condition (146) is routinely taken as implying a pure gauge field.2
The use of pure gauge fields in nonabelian gauge theory is discussed in Sec. 12-1 in Ref. 2.
C. The case of pseudo-four-vector current coupling
The results of Sec. IV A suggest that a coupling of the four-vector field K µ to a conserved four-vector current
appears to be problematic. Indeed, K 0 , the only effectively Hermitian component of the potential K µ , acts like
a free field. However, there are examples of four-currents which are not conserved. The four-current coupling of
a nucleon iso-spinor to a pseudo-scalar π meson isotopic (three) vector exhibits what is referred to as “differential
current conservation”, (cf. Ref. 20). In this case the proton and π + meson currents are not separately conserved, but
their combination is conserved. It is interesting, therefore, to consider the case of pseudo-four-vector current coupling
to the K µ field in its own right.
While one could just take K µ and jgµ in the Lagrangian (61) as pseudo-four-vectors, the essential features of the
coupling to a fermion current would not be illustrated. Therefore, although the analysis of the present article is
ostensibly limited to classical field theory, it is necessary to present a few quantum mechanical calculations so as to
obtain a preliminary force law for this case. The pseudo-four-vector current will be taken as
jgµ ≡ g ψ † γ 0 γ µ γ5 ψ = g ψγ µ γ5 ψ, (147)
where γ 0 , γ µ , and γ5 are the Dirac matrices, where ψ is a Dirac spinor, and where the adjoint spinor is defined
as ψ ≡ ψ † γ 0 . Note that a factor of i is implicit in (147) via the formally pure imaginary valued classical charge
g = ig, where g is a real number. As a result (147) is anti-Hermitian. Also, the pseudo-four-vector nature of (147)
follows from the inclusion of the factor of γ5 . A suitable Lagrangian density (of the Stueckelberg type) will be chosen
(dropping go in favor of rationalized Heaviside units and with ~ = c = 1) using (147) as follows:
1 † µν λg µ2g
∂µ K µ† (∂µ K µ ) − Kµ† K µ + ψ (iγ µ ∂µ − m) ψ − g ψ6K γ5 ψ,

Lgi = Gµν G + (148)
4 2 2
where the Feynman slash notation K 6 ≡ γ µ Kµ has been used. Note that an explicit minus sign has been inserted
in the first three terms of (148) since the field is being taken as anti-Hermitian for convenience with the relation
K µ† (x) = −K µ (x), and so, e.g., Kµ K µ = −Kµ† K µ , [compare with (61)]. The last term of (148) is the interaction
Lagrangian density
LI = −g ψ6K γ5 ψ = −g ψγ µ γ5 ψKµ = −jgµ Kµ , (149)
where Kµ commutes with ψ and γ5 . [Equation (149) could also be written as the symmetrized sum
jgµ† Kµ + jgµ Kµ†

LI = , (150)
2
in which case the explicit minus sign for an anti-Hermitian field is factored into these terms, too.] The Euler-Lagrange
equation of motion for K µ which follows from the Lagrangian density (148) is
∂α ∂ α + µ2g K µ − (1 − λg ) ∂ µ (∂σ K σ ) = g ψγ µ γ5 ψ = jgµ ,

(151)
where the minus sign in the first three terms of (148) [and in the symmetrized interaction Lagrangian density (150)]
gets reabsorbed in obtaining (151), which in turn is similar in form to (63). The Euler-Lagrange equation of motion
for the spinor ψ is
(iγ µ ∂µ − m) ψ = g6K γ5 ψ. (152)
Dividing (152) by i while setting K µ = 0 gives
0 = (γ µ ∂µ + im) ψ = γ 0 ∂0 ψ + γ j ∂j ψ + imψ. (153)
The Hermitian conjugate of (153) is

†
0 = (γ µ ∂µ ψ + imψ) = (∂0 ψ † )γ 0† + (∂j ψ † )γ j† − imψ † = (∂0 ψ † )γ 0 − (∂j ψ † )γ j − imψ † , (154)
since γ 0† = γ 0 and γ j† = −γ j . Now, taking the four-divergence of (147) (times 1/g) using (153) and (154) gives
(1/g)∂µ jgµ = ∂µ (ψ † γ 0 γ µ γ5 ψ)
= (∂0 ψ † )γ 0 γ 0 γ5 ψ + (∂j ψ † )γ 0 γ j γ5 ψ + ψ † γ 0 γ 0 γ5 (∂0 ψ) + ψ † γ 0 γ j γ5 (∂j ψ)
= (∂0 ψ † )γ 0 γ 0 γ5 ψ − (∂j ψ † )γ j γ 0 γ5 ψ − ψ † γ 0 γ5 γ 0 (∂0 ψ) − ψ † γ 0 γ5 γ j (∂j ψ)
= (imψ † )γ 0 γ5 ψ − ψ † γ 0 γ5 (−imψ) = 2imψ † γ 0 γ5 ψ = 2imψγ5 ψ, (155)
since {γ µ , γ5 } ≡ γ µ γ5 + γ5 γ µ = 0 and {γ 0 , γ j } ≡ γ 0 γ j + γ j γ 0 = 0, and so the pseudo-four-vector current (147) is not

conserved as expected. This is significant in the present case since the four-divergence of the wave equation (151),
after cancellation of terms, is
λg ∂α ∂ α + µ2g ∂µ K µ = ∂µ jgµ = 2mg(ψiγ5 ψ),

(156)
and so the four-divergence ∂µ K µ is no longer a free scalar field. [Contrast with (65) where it is easy to show in a
similar fashion using (153) and (154) that the four-divergence of a four-vector current jgµ = gψγ µ ψ is zero.]
The covariant canonical momentum (71) follows as before (in the units of this section) as
Πνgi (xµ ) = − ∂ 0 K ν − ∂ ν K 0 − λg g 0ν ∂σ K σ .

(157)
Now, in order to simplify calculations it is desirable to use the relativistic longitudinal gauge constraint (47), Gµν = 0.
However, in Sec. III D it was suggested that this constraint would most likely be imposed as a restriction on the
positive frequency parts of Gµν as per (58). But, the restriction (58) still implies that
hψ|Gµν |ψi = 0, (158)
and since the expectation value of an operator is the relevant quantity in the classical mechanical limit, the author
will cavalierly set Gµν = 0 in the calculations which follow. Therefore, the only surviving component of (157) in the
relativistic longitudinal gauge is the momentum canonically conjugate to the scalar field component K 0 , namely,
Π0gi = −λg g 00 ∂σ K σ = −λg ∂σ K σ , (159)
and so by (156), the (potentially observable) canonical momentum Π0gi has a pseudo-scalar wave equation,
!
α
µ2g
∂α ∂ + Π0gi = −2mg(ψiγ5 ψ), (160)
λg
p
with a “mass” as defined previously in (66), namely ms = µg / λg , where λg > 0. Equation (160) is of a similar
functional form as the wave equation for a π 0 meson (pure real pseudo-scalar field) coupled to a proton via an attractive
Yukawa potential (cf. Eq. (10.10) in Ref. 20). However, in (160) g is taken as a pure imaginary valued coupling
constant and Π0gi is an anti-Hermitian operator (i.e., (160) is a wave equation for an anti-Hermitian pseudo-scalar field
which can be rewritten to look like it is for a Hermitian pseudo-scalar field). Also, there is that curious appearance
of the fermion mass m (as well as a factor of 2) on the rhs of (160).
The energy-momentum tensor formulation follows as before from (75). However, in the present case when one
calculates the four-force densities one must use the pseudo-four-vector current (147). The four-force density in the
relativistic longitudinal gauge (96) becomes
fLν ≡ −∂µ Θµν ν µ ν µ µ ν µ ν

L = jgµ (∂ K ) = g ψγµ γ5 ψ (∂ K ) = jgµ (∂ K ) = g ψγµ γ5 ψ (∂ K ) , (161)
where either of the two general forms in (161) are equivalent if one assumes Gµν = 0 in anticipation of the calculation
of an expectation value for the force law. There will be higher order interaction force terms of quantum mechanical
origin as well. Also, it should be emphasized that the Lorentz-type force law (116), which vanishes in the relativistic
longitudinal gauge, does not follow from a single particle Lagrangian approach for this pseudo-four-vector current case
due to the factor of γ5 in the interaction Lagrangian (149). Therefore, the inconsistency present in the four-vector
current case of Sec. IV A with respect to obtaining a force law does not arise here.
V. CONCLUSION:
The author is currently studying the interactions of relativistic longitudinal gauge fields with electromagnetic-type
fields. For example, consider the enigmatic 1/2 quantum infinite zero-point vacuum contribution to the spin one
electromagnetic field, which is currently ignored using normal ordering of quantum operators. Upon first inspection it
appears likely that this vacuum contribution can be canceled via a coupling with an appropriate relativistic longitudinal
gauge four-vector field, which has an analogous zero-point contribution but of opposite sign. This type of cancellation
was discussed by Pauli in 1943 for a system of two harmonic oscillators (one Hermitian with positive energy and
one anti-Hermitian with negative energy) [see Eq. (34) of Sec. 4 in Ref. 21]. He found that the ground state of the
system was not uniquely determined when an indefinite metric Hilbert space was used (which is not surprising since
his system was not bounded from below). On the other hand, the constrained system of one Hermitian four-vector
field in the Lorentz gauge and one formally anti-Hermitian four-vector field with one effectively Hermitian component
surviving in the relativistic longitudinal gauge, would be bounded from below and so this would be the first plausible
use of the technique.
ACKNOWLEDGMENTS:
The style and presentation of this article has benefited from detailed discussions with the author’s Ph.D. thesis
adviser J. V. Corbett of Macquarie University-Sydney, Australia. This work is both an extension and a partial report-
ing of the results of the author’s Ph.D. thesis (Ref. 18). Tuition exemption from the Higher Education Contribution
Scheme (HECS) was obtained via the Ph.D. by Research program. The author would like to thank R. E. Collin of
Case Western Reserve University (CWRU), R. Coquereaux of the Centre de Physique Théorique, Centre National de
la Recherche Scientifique, Campus de Luminy, France, and C. A. Hurst of Adelaide University, Australia, for their
insightful comments on the author’s Ph.D. thesis. The author would also like to thank L. L. Foldy, R. E. Collin, and
Y. H. Pao of CWRU for encouraging a speculation that the author made some twenty four years ago regarding pure
imaginary valued four-vector fields which led to a Master’s project (Ref. 22) (i.e., thesis) at CWRU, and in the course
of time to this article.
a)
Electronic mail: [email protected]
1
D. A. Woodside, J. Math. Phys. 40, 4911 (1999).
2
C. Itzykson and J. Zuber, Quantum Field Theory (McGraw-Hill, New York, 1980), pp. 10-12, 136, 576, 562-572.
3
E. C. G. Stueckelberg, Helv. Phys. Acta 11, 225 (1938); 11, 299 (1938).
4
F. Mandl and G. Shaw, Quantum Field Theory, revised ed. (Wiley, New York, 1994).
5
C. Cohen-Tannoudji, J. Dupont-Roc, and G. Grynberg, Photons & Atoms, Introduction to Quantum Electrodynamics (Wiley,
New York, 1989), pp. 378-399, 13-22.
6
K. Bleuler, Helv. Phys. Acta 23, 567 (1950).
7
S. N. Gupta, Proc. Phys. Soc., London, Sect. A 63, 681 (1950); 64, 850 (1951).
8
F. Mandl, Introduction to Quantum Field Theory (Wiley, New York, 1966).
9
H. A. Lorentz, A. Einstein, H. Minkowski, and H. Weyl, The Principle of Relativity (Dover, New York, 1952), pp. 73-91.
10
P. A. M. Dirac, Proc. Roy. Soc., London, Ser. A 180, 1 (1942); Comm. Dublin Inst. for Adv. Studies A 1, 1 (1943).
11
G. Källén, Quantum Electrodynamics (Springer, New York, 1972).
12
R. Plonsey and R. E. Collin, Principles and Applications of Electromagnetic Fields (McGraw-Hill, New York, 1961), pp.
29-36.
13
J. D. Jackson, Classical Electrodynamics, 2nd ed. (Wiley, New York, 1975), pp. 608-612, 219-223.
14
B. R. Holstein, Topics in Advanced Quantum Mechanics (Addison Wesley, New York, 1992), pp. 117-119.
15
P. W. Anderson, Phys. Rev. 110, 827 (1958).
16
J. R. Aitchison, An Informal Introduction to Gauge Field Theories (Cambridge University Press, Cambridge, 1982).
17
M. Kalb and P. Ramond, Phys. Rev. D 9, 2273 (1974).
18
D. A. Woodside, Ph.D. thesis, Macquarie University-Sydney, 1998.
19
S. Weinberg, The Quantum Theory of Fields (Cambridge University Press, Cambridge, 1996), Vol. II, pp. 333-334, 24-27.
20
J. D. Bjorken and S. D. Drell, Relativistic Quantum Mechanics (McGraw-Hill, New York, 1964).
21
W. Pauli, Rev. Mod. Phys. 15, 175 (1943).
22
D. A. Woodside, Master’s project thesis, Case Western Reserve University, 1977.

Classical Four-Vector Fields in The Relativistic Longitudinal Gauge

Uploaded by

Copyright:

Available Formats

Classical Four-Vector Fields in The Relativistic Longitudinal Gauge

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Classical Four-Vector Fields in The Relativistic Longitudinal Gauge

Uploaded by

Copyright:

Available Formats

JOURNAL OF MATHEMATICAL PHYSICS VOLUME 41, NUMBER 7, (pp.

4622−4653)# JULY 2000

Classical four-vector fields in the relativistic longitudinal gauge†

II. MATHEMATICAL BACKGROUND

III. APPLICATION OF THE MINKOWSKI SPACE UNIQUENESS THEOREMS OF REF. 1 TO TWO

A. Introduction of a new class of four-vector fields

xµ = ηµν xν = (x0 , x1 , x2 , x3 ) = (ct, −x, −y, −z). (6b)

xµ = (x1 , x2 , x3 , x4 ) = (x, y, z, ict), (7)

B. Gauge considerations in classical field theory

A(r) ≡ AL (r) + AT (r), (12)

which are longitudinal, i.e., irrotational,

∇×AL (r) = 0, (13a)

or transverse, i.e., solenoidal,

ik×AL (k) = 0, (15a)

and for the transverse component

A(r, t) → A0 (r, t) = A(r, t) + ∇F (r, t), (16a)

A0µ (xν ) = Aµ (xν ) − ∂ µ F (xν ) . (17)

1 ∂φ0 (r, t) 1 ∂φ(r, t) 1 ∂ 2 F (r, t)

ik · AL (k) = 0 ⇒ AL (k) = 0 ∀k; k 6= 0. (26)

i.e., a rank zero source density and

∂ α Aµ (xσ ) − ∂ µ Aα (xσ ) ≡ cαµ , (30b)

i.e., a rank two circulation density, for all xσ .

F 0 αµ = ∂ α A0µ − ∂ µ A0α = ∂ α Aµ + ∂ α ∂ µ F − ∂ µ Aα − ∂ µ ∂ α F = F αµ , (34)

and so (33) with (34) substituted reduces to

∂ µ F (xσ ) = 0 ∀xσ ∈ R3+1 . (36)

A0µ (xσ ) = Aµ (xσ ) , (37)

ψ(x) → ψ 0 (x) = exp(−iqF (x))ψ(x), (38)

∂ν Aν(+) (x) |0i = 0, (39)

∂ν K ν (xσ ) ≡ sgi , (40a)

∂ α K µ (xσ ) − ∂ µ K α (xσ ) ≡ cαµ

Gαµ ≡ ∂ α K µ (xσ ) − ∂ µ K α (xσ ) . (42)

0 −Gx /c −Gy /c −Gz /c

∇×K = 0, i.e., ∂ i K j − ∂ j K i = Gij = 0 −→ ∂ µ K ν − ∂ ν K µ = Gµν = 0, (46)

K 0µ (x) = K µ (x) − ∂ µ Fg (x), (53)

on the integral in (52) then gives

∂µ K 0µ (x) = ∂µ K µ (x) − ∂µ ∂ µ Fg (x) = ∂µ K µ (x). (56)

The trivial gauge restriction

(∂ µ K ν(+) (x) − ∂ ν K µ(+) (x)) |0i = 0, (58)

IV. CANONICAL FORMULATION OF THE RELATIVISTIC LONGITUDINAL CLASSICAL

A. The case of four-vector current coupling

follow from the Lagrangian density (61) as

∂α ∂ α + µ2g K µ − (1 − λg ) ∂ µ (∂σ K σ ) = jgµ /go c2 .

Taking the four-divergence of (63) yields

∂α ∂ α + µ2g ∂µ K µ − (1 − λg ) ∂µ ∂ µ (∂σ K σ ) = ∂µ jgµ /go c2 .

Simplifying (64) and assuming a conserved (four-vector) current ∂µ jgµ = 0 yields

λg ∂α ∂ α + µ2g ∂µ K µ = ∂µ jgµ /go c2 = 0.

where λg > 0 has been assumed to ensure that m2s > 0.

Taking the four-divergence of (67) gives

which yields [cf. (73) for index µ = 0, or (3-100) in Ref. 2],

Πνgi (xµ ) = −go c2 ∂ 0 K ν − ∂ ν K 0 − λg go c2 g 0ν ∂σ K σ .

where the equation of motion (63), rewritten using (42) as

But, the last two terms of (86) can be rewritten as

0 = gµρ gλσ (∂ ν Gρσ + ∂ ρ Gσν + ∂ σ Gνρ ) = ∂ ν Gµλ + ∂µ Gλν + ∂λ Gνµ . (88)

Therefore, with the result (87), Eq. (86) reduces to

fTν = −jgλ Gλν + go c2 λg ∂ λ (∂σ K σ ) + µ2g K λ Gλν .

−f ν ≡ ∂µ Θµν = (∂ ν jgµ ) K µ . (98)

jgµ = ρo (γc, γu) = ρo uµ . (101)

∂α ∂ α + µ2g K µ = jgµ /go c2 ,

associated with the K k components is not a free field since