INSTITUTE OF PHYSICS PUBLISHING EUROPEAN JOURNAL OF PHYSICS

Eur. J. Phys. 24 (2003) 397–401 PII: S0143-0807(03)54074-0

Imaginary in all directions: an elegant

formulation of special relativity and
classical electrodynamics
Martin Greiter and Dirk Schuricht
Institut für Theorie der Kondensierten Materie, Universität Karlsruhe, Postfach 6980,
D-76128 Karlsruhe, Germany

Received 30 September 2002, in final form 15 November 2002

Published 23 June 2003
Online at [Link]/EJP/24/397

Abstract
A suitable parametrization of space–time in terms of one complex and three
quaternionic imaginary units allows Lorentz transformations to be implemented
as multiplication by complex quaternionic numbers rather than matrices.
Maxwell’s equations reduce to a single equation.

1. Introduction

One of the most established symmetries of our universe is its invariance under Lorentz
transformations, or the principle of special relativity [1]. Lorentz originally discovered that
Maxwell’s theory of electromagnetism was invariant under these tranformations. Einstein
subsequently interpreted them as a principle of relativity, which is by no means specific
to electromagnetism but a fundamental and far-reaching invariance of our universe. The
transformations were formulated in terms of 4 ×4 matrices (
νµ ) that act on contravariant four-
vectors denoting space–time (q µ ) ≡ (t, x, y, z), energy–momentum ( pµ ) ≡ (E, px , p y , pz ),
or the like. The introduction of the Minkowski metric (gµν ) and another set of covariant
vectors such that (qµ ) ≡ (t, −x, −y, −z) provided a highly convenient notation for relativistic
theories, which has since been taught and referred to as relativistic notation. Einstein once
even joked that his most important contribution to physics was to introduce this notation.
Apart from the success of the notation, it is taught to students of physics at such an
early stage that they hardly question it. They learn in other classes, for example in alternating
current circuit theory, the enormous technical advantage that can result if one employs complex
numbers to describe the phase shifts in the time dependencies of real but alternating voltages
and currents, but nothing suggests that any of this formalism, or a generalization of it, can
simplify Lorentz transformations in a similar way. The only obvious application of imaginary
units appears to be a rotation from Minkowski to Euclidean space by replacing the time t by
an imaginary time τ ≡ −it, with the effect that the Minkowski metric (gµν ) is replaced by a
simple Euclidean metric (δµν ).
In this paper, we will show that a suitable imaginary parametrization of space–time or
other Lorentz contra- or covariant quantities, using both one complex and three quaternionic
0143-0807/03/040397+05$30.00 © 2003 IOP Publishing Ltd Printed in the UK 397
398 M Greiter and D Schuricht

imaginary units [2, 3], yields a tremendous simplification of relativistic invariance [4–6]:
Lorentz transformations will no longer require multiplication by matrices, just multiplication
by complex quaternionic (CQ) numbers. Classical electrodynamics will then provide us with
the first orchard in which to harvest the gain in formal elegance [7, 8]. To be specific, Maxwell’s
four equations will reduce to a single identity between two CQ numbers.

2. Special relativity

Let us now set up the formalism. We introduce a complex algebra with generator 1, @ ∈ C,
such that
@2 = −1 (1)
as well as an quaternionic algebra with generators 1, i, j, k ∈ H, such that
i 2 = j 2 = k 2 = −1,
(2)
i j = − ji = k, j k = −k j = i, ki = −i k = j,
which mutually commute:
[i, @] = [ j, @] = [k, @] = 0. (3)
∗
We further introduce a complex conjugate operation , which takes the form
@ → @∗ = −@
but leaves i , j and k unchanged, as well as a quaternionic conjugate operation − , which leaves
@ unchanged but takes the form
i → ī = −i, j → j̄ = − j, and k → k̄ = −k.
Note that, if o1 , o2 ∈ C⊗H are two CQ numbers, then the order of the product o1 o2 is reversed
under quaternionic conjugation only:
(o1 o2 )∗ = o1∗ o2∗ but o1 o2 = ō2 ō1 . (4)
No simple product rule would exist if we had two or more mutually commuting quaternionic
algebras.
We label space–time (and other Lorentz contravariant quantities usually denoted by four
vectors) by a purely imaginary CQ number,
q ≡ @t + i x + j y + kz (5)
where t, x, y, z ∈ R. We identify this subspace of C ⊗ H using Minkowski space and denote it
by M. Complex conjugation ∗ and quaternionic conjugation − correspond within this space to
time reversal (T) and parity (P) transformations, respectively. Note that q ∗ = −q̄ for q ∈ M.
The corresponding covariant quantity is given by its quaternionic conjugate or parity
reversed CQ number,
q̄ = @t − i x − j y − kz, (6)
yielding the proper time interval
−q̄q = −q q̄ = t 2 − x 2 − y 2 − z 2 . (7)
Let n = i n x + j n y + kn z with n n̄ = n 2x = 1 and n x , n y , n z ∈ R be a quaternionic
+ n 2y + n 2z
imaginary unit vector (n x , n y , n z ). Then a Lorentz transformation is given simply by
q → q  = ωq ω̄∗ , (8)
with either
1 θ θ
ω = e 2 nθ = cos + n sin (9)
2 2
Imaginary in all directions: an elegant formulation of special relativity and classical electrodynamics 399

for a rotation through an angle θ around n, or

1


ω = e 2 n@
= cosh + @n sinh (10)
2 2
for a boost by a Lorentz angle
in direction n. Equations (8)–(10) are readily verified using
n 2 = −1 and (n@)2 = 1.
Clearly the covariant CQ number q̄ transforms as
q̄ → q̄  = ω∗ q̄ ω̄. (11)
With ωω̄ = ω̄ω = 1, the Lorentz invariance of the proper time −q̄q is evident.
At this point we assume that the reader is aware of how awkward it is to write down a
rotation around (or a boost along) an arbitrary axis using 4 × 4 matrices, so that there is no
need to dwell on the elegance of the formulation proposed here.
It is convenient at this point to use the norm given by the proper time interval to define a
scalar product between two imaginary or Minkowski CQ numbers M × M → R, according to
 p, q ≡ 14 (( p − q)( p − q) − ( p + q)( p + q))
= − 12 ( p̄q + q̄ p)
= − 12 ( p q̄ + q p̄). (12)
The proper time interval (7) is given by q, q = −q̄q. With p = @E + i px + j p y + kpz and
q = @t + i x + j y + kz, we obtain
 p, q = Et − px x − p y y − pz z. (13)
Furthermore, it is convenient to define the contravariant differentiation operator
∂ ∂ ∂ ∂
D ≡ @ −i − j −k = @∂t − i ∂x − j ∂ y − k∂z , (14)
∂t ∂x ∂y ∂z
which likewise transforms according to
D → D  = ωD ω̄∗ . (15)
Note that −D q̄ = − D̄q = 4, as expected. The covariant differentiation operator is, of course,
given by
D̄ = @∂t + i ∂x + j ∂ y + k∂z , (16)
and transforms
D̄ → D̄  = ω∗ D̄ ω̄. (17)

3. Classical electrodynamics

We proceed by applying this formalism to classical electrodynamics. To begin with, we

introduce a contravariant vector (or imaginary CQ field)
A ≡ @φ + i A x + j A y + k A z , (18)
and require the theory to be invariant under electromagnetic gauge transformations
A → A + Dλ(q), (19)
where λ(q) is an arbitrary real-valued scalar function of space–time q. We proceed by defining
the electromagnetic field strength
F ≡ 12 ( D̄ A − D̄ A). (20)
Clearly F is invariant under (19). Note that we cannot replace D̄ A by ĀD in the second term
in (20), since the derivative operator has to act on A. Under a Lorentz transformation, F = − F̄
transforms as
F → F  = ω∗ F ω̄∗ . (21)
400 M Greiter and D Schuricht

At this point it is propitious to introduce

e1 ≡ i, e2 ≡ j, e3 ≡ k (22)
such that
ei e j = −δi j + i j k ek with i, j, k ∈ {1, 2, 3}, (23)
where the indices i, j, k in (23) are not to be confused with the quaternionic generators i, j, k
in (22). With summation over repeated indices but no hidden minus signs implied, we write
A = @φ + ei Ai D = @∂t − ei ∂i etc. (24)
Writing out the field strength (20) in components, we obtain
F = ei i jk∂ j Ak − @ei (−∂t Ai − ∂i φ). (25)
Note that
D F = (−@∂i + ei ∂t )(−∂t Ai − ∂i φ) − ei i j k ∂ j klm ∂l A m (26)
is a purely imaginary CQ number (D F ∈ M). This implies that
∗
D F + D F = 0, (27)
which will prove useful below.
We proceed by defining magnetic and electric field strengths according to
Bi ≡ i j k ∂ j Ak and E i ≡ −∂t Ai − ∂i φ, (28)
and use them to rewrite (25) as
F = ei (Bi − @E i ). (29)
We may hence write the Lagrangian density for the electromagnetic field coupled to an external
current (ρ, Ji ) as
L ≡ 12 (E i2 − Bi2 ) + ρφ − Ji Ai
= 1 (F 2 + (F ∗ )2 ) + 1 ( J¯ A + Ā J ),
4 2 (30)
where we have defined
J ≡ @ρ + ei Ji . (31)
Clearly, L and F =2
E i2
+ 2@E i Bi + Bi2
are Lorentz invariant.
Variation of L with respect to δ Ā yields, after integration by parts,
∗
4 (D F − D F ) + 12 J = 0.
1
(32)
Together with (27), we obtain
D F + J = 0. (33)
This is Maxwell’s equation. (In the standard formulation, we need to speak of equations
(plural), but we shall see now that this single equation replaces all four of them.) Note
that (33) transforms contravariantly under Lorentz transformations. Evaluation of D F, starting
from (29), yields
D F = ∂i Bi − @∂i E i + ei (∂t E i − i j k ∂ j Bk ) + @ei (∂t Bi + i j k ∂ j E k ). (34)
Since (33) implies that the coefficients of D F + J in all eight orthogonal directions
{1, @, ei , @ei } of C ⊗ H vanish, substitution of (34) and (31) into (33) immediately yields
∂i Bi = 0
−∂i E i + ρ = 0
(35)
∂t E i − i j k ∂ j Bk + Ji = 0
∂t Bi + i j k ∂ j E k = 0.
Imaginary in all directions: an elegant formulation of special relativity and classical electrodynamics 401

We assume that the reader is familiar with these equations.

In the Lorentz gauge,
∂t φ − ∂i Ai = − 12 ( D̄ A + D̄ A) = 0. (36)
Maxwell’s equation (33) then reduces to
D D̄ A + J = 0, (37)
which is readily recognized as a wave equation.

4. Conclusion

In conclusion, we have shown that a suitable CQ parametrization of Lorentz contra- and

covariant quantities can greatly enhance the elegance and simplicity of relativistic theories.
We believe that this language is likely to yield new perspectives on quantum field theories [9–
12], which we are currently investigating.

Acknowledgment

One of us (MG) is deeply indebted to S C Zhang [13] for countless discussions on his four-
dimensional generalization of fractionally quantized Hall states that have inspired the present
work.

References
[1] See, for example, Weinberg S 1972 Gravitation and Cosmology: Principles and Applications of the General
Theory of Relativity (New York: Wiley) ch 2
[2] Hamilton W R 1969 Elements of Quaternions (New York: Chelsea)
[3] Dixon G M 1994 Division Algebras: Octonions, Quaternions, Complex Numbers and the Algebraic Design of
Physics (Dordrecht: Kluwer)
[4] Synge J L 1972 Quaternions, Lorentz Transformations, and the Conway–Dirac–Eddington Matrices (Dublin:
Institute for Advanced Studies)
[5] De Leo S 1996 J. Math. Phys. 37 2955
[6] Edmonds J D (ed) 1997 Relativistic Reality: A Modern View (Singapore: World Scientific)
[7] Imaeda K 1995 Clifford Algebras and Spinor Structures ed R Ablamowicz and P Lounesto (Dordrecht: Kluwer)
[8] Edmonds J D 1972 Int. J. Theor. Phys. 6 205
Edmonds J D 1973 Found Phys. 3 313
Edmonds J D 1978 Am. J. Phys. 46 430
[9] Morita K 1981 Prog. Theor. Phys. 67 1860
Morita K 1982 Prog. Theor. Phys. 68 2159
Morita K 1983 Prog. Theor. Phys. 70 1648
[10] Sharma C S and Almeida D F 1990 J. Math. Phys. 31 1035
[11] Adler S 1995 Quaternionic Quantum Mechanics and Quantum Fields (Oxford: Oxford University Press)
[12] De Leo S 1996 J. Phys. G: Nucl. Part. Phys. 22 1137
De Leo S 1998 Int. J. Theor. Phys. 37 1511
[13] Zhang S C and Hu J 2001 Science 294 823