math-ph/0307038

Physical Space as a Quaternion Structure, I:

Maxwell Equations. A Brief Note.
Peter Michael Jack∗
Hypercomplex Systems
Toronto, Canada †
(Dated: July 18, 2003)
This paper shows how to write Maxwell’s Equations in Hamilton’s Quaternions. The fact that the
quaternion product is non-commuting leads to distinct left and right derivatives which must both be
included in the theory. A new field component is then revealed, which reduces part of the degree of
freedom found in the gauge, but which can then be used to explain thermoelectricity, suggesting that
the theory of heat has just as fundamental a connection to electromagnetism as the magnetic field
has to the electric field, for the new theory now links thermal, electric, and magnetic phenomena
altogether in one set of elementary equations. This result is based on an initial hypothesis, named
arXiv:math-ph/0307038v1 18 Jul 2003

“The Quaternion Axiom,” that postulates physical space is a quaternion structure.

PACS numbers: 03.50.De, 72.15.Jf, 72.20.Pa, 73.50.Lw, 74.25.Fy

I. THE QUATERNION AXIOM. and ‘operate to the left’, respectively. Then, for example,
if the ‘a’ term is the operator, and the ‘b’ term is the
The quaternions variable acted upon by the operator, we have;

a = a0 + a1.i + a2.j + a3.k (1) a→b = a0.b0 − a1.b1 − a2.b2 − a3.b3

where i, j, k are the anti-commuting hypercomplex + a0.(b1.i + b2.j + b3.k)
roots of -1, and the a0, a1, a2, a3 are elements of the + (a1.i + a2.j + a3.k).b0 (3)
real number set, can be used to write down Maxwell’s + (a2.b3 − a3.b2).i
Equations. We postulate that physical space is a quater-
+ (a3.b1 − a1.b3).j
nion structure, so that the units { i, j, k} represent space
dimensions, while the scalar { 1} represents time, and + (a1.b2 − a2.b1).k
the space units obey the product rules given by W. R.
Hamilton in 1843[1] ; b←a = a0.b0 − a1.b1 − a2.b2 − a3.b3
+ a0.(b1.i + b2.j + b3.k)
+ (a1.i + a2.j + a3.k).b0 (4)
i2 = j 2 = k 2 = −1
− (a2.b3 − a3.b2).i
(2)
− (a3.b1 − a1.b3).j
i = jk = −kj, j = ki = −ik, k = ij = −ji
− (a1.b2 − a2.b1).k
We shall refer to this postulate as The Quaternion Ax- Note, we only need the symbols, → and ← , between
iom. A position vector in this quaternion space will take the quaternions themselves. Once we have resolved the
the form, r = ct + ix + jy + kz, with c the characteris- algebra to the level of the components, we can revert to
tic speed that links clock ticks to metre rules, equivalent the usual convention of putting the operator on the left
here to the speed of light, so that all measures within this and the variable acted upon on the right.
quaternion 4-vector are ultimately in the same units-of- If there is no physical reason to select one product over
length. the other, both products must be accounted for equally
in the expressions used to model phenomena in any given
theory, otherwise such expressions will have an inherent
A. Ambiguous Product. left-hand or right-hand bias.
When dealing with operator products therefore, we de-
Now, because quaternions don’t commute, we need to fine the right product by, a → b, and the left product by,
recognise the distinction between left and right actions. b ← a. Then the two combinations of these, the symmet-
Consider two quaternion variables a & b. Let us define ric product, {a,b}, and the antisymmetric product, [a,b],
the symbols ‘→’ and ‘←’ to mean ‘operate to the right’ are defined correspondingly by;
{a, b} = (1/2)(a → b + b ← a) (5)

∗ Alumnus of the Physics Department of Columbia University, NY.

† Electronic address: math@hypercomplex.com [a, b] = (1/2)(a → b − b ← a) (6)
 2

Which product we employ in our theory, and where, Now, although we could invent constructions using the
is dictated by the symmetries inherent in the physical quaternion conjugate to eliminate the time component
problem. from all our calculations and so obtain the exact results
of the 3-vector calculus, this would make the calculations
more contrived and less natural than the simple structure
II. MAXWELL EQUATIONS. presented here. We therefore let the natural structure of
the algebra lead us instead, to the results that follow
Now, let the Electromagnetic Potential be from accepting this new field component, rather than try
to eliminate it from our framework. Then, by inspection,
A = U + A1.i + A2.j + A3.k (7) the reformulated Maxwell Field Equations are,

and the differential operator (d/dr) be defined by,

[d/dr, B] = +{d/dr, E} (13)
d 1 ∂ ∂ ∂ ∂
= + i+ j+ k (8)
dr c ∂t ∂x ∂y ∂z
[d/dr, E] = −{d/dr, B} (14)
then by inspection, the Electric and Magnetic fields as
quaternions are, The antisymmetric derivative of the magnetic field is
the positive of the symmetric derivative of the electric
E = −{d/dr, A} = −(1/2)(d/dr → A + A ← d/dr) (9) field. And, the antisymmetric derivative of the electric
field is the negative of the symmetric derivative of the
magnetic field. The first represents a real physical law,
B = +[d/dr, A] = +(1/2)(d/dr → A − A ← d/dr) (10) while the second is easily proven to be an algebraic
identity when given the definitions of the electric and
That is, the electric field is the negative symmetric magnetic fields above.
derivative of the potential, and the magnetic field is the
positive antisymmetric derivative of the potential. The When written in the vector notation of Heaviside-
space components of these quaternion fields correspond Gibbs, these two 4-vector (quaternion) equations become
exactly to the electric and magnetic fields in the usual the usual four 3-vector equations,
3-vector calculus. However, the electric quaternion field
now has a time component, which we label, T, so that, curl(B) = +1/c.∂E/∂t + grad(T ) (15)
E = T + E, while the magnetic quaternion field has no
time component, so that, B = 0 + B. And if we allow our curl(E) = −1/c.∂B/∂t (16)
notation to alternate between Heaviside-Gibbs 3-vector div(E) = +1/c.∂T /∂t (17)
and that of Hamilton’s Quaternion 3-vector, taking care div(B) = 0 (18)
to match up only the components of the appropriate ex-
pressions, we can write the quaternion derivative in terms
of the more familiar vector notation, T = −1/c.∂U/∂t + div(A) (19)
E = −grad(U ) − 1/c.∂A/∂t (20)
d 1 ∂U 1 ∂A B = curl(A) (21)
→A= − div(A) + + grad(U ) + curl(A)
dr c ∂t c ∂t
(11)
provided we now identify the electric charge density, ρ,
and electric current density, J, with the terms involving
d 1 ∂U 1 ∂A T. Thus, 4πρ = +1/c.∂T /∂t, and 4πJ/c = grad(T ).
A← = − div(A) + + grad(U ) − curl(A) The scalar quantity, T, we shall call the “Temporal
dr c ∂t c ∂t
(12) Field.” This scalar field has the same units-of-dimension
as the electric and magnetic vector fields in our Gaussian
where we observe that the quaternion derivative has system of units. And thus for a given charge, q, the
“five” distinct parts, much like the five fingers on the quantity, qT, has units of force, similar to the electric
human hand. The electric field is then obtained from force, qE, and the magnetic force, qv/c × B. However,
clapping the hands together, so that the distinction this scalar force has no space direction. Instead, it
between the two vanish (and consequently giving rise, as acts along the timeline, since that is the scalar axis
we shall see, to “heat” as well as electricity). While, the under The Quaternion Axiom used as the basis of this
magnetic field is obtained from the opposite action of derivation.
pulling the hands apart, so that the distinction between
the two become more pronounced. What effect is produced by a scalar force acting along
the timeline?
 3

III. THERMOELECTRICITY. When we introduce the electric charge density and elec-
tric current source terms, as additional inhomogeneous
The electric force moving a charge, q, through a dis- parameters to the temporal terms, instead of identify-
placement, dx, does work, dW = qE·dx. And work is a ing the electric source directly with the temporal terms
form of energy. This energy is delivered to the external themselves, the second equation is unchanged, and the
environment, and is available, for example, as mechan- first equation says,
ical work, capable of moving the parts of a mechanical
system against a frictional resistance. d/dr → (d/dr → A) + (A ← d/dr) ← d/dr = 8πJ
(24)

A. Heat. where, J = (ρ, J/c). In the Heaviside-Gibbs 3-vector for-

mat, the new inhomogeneous electromagnetic equations
Similarly, the temporal force acting on a charge, q, therefore become,
for a time interval, cdt, produces an energy-like term,
dW = −qT cdt (the minus sign reflecting the opposite curl(B) = +1/c.∂E/∂t + grad(T ) + 4πJ/c (25)
signs in the squares of the unit time, 12 = +1, and curl(E) = −1/c.∂B/∂t (26)
unit space, i2 = j 2 = k 2 = −1, measures, in the div(E) = +1/c.∂T /∂t + 4πρ (27)
quaternion spacetime, needed here because E and dx are div(B) = 0 (28)
Heaviside-Gibbs vectors; thus a positive charge, q > 0,
under the influence of a positive value temporal field,
T > 0, produces the equivalent of negative work, i.e. T = −1/c.∂U/∂t + div(A) (29)
the charge-field interacting system will absorb energy
from its surroundings, positive charges thus effectively E = −grad(U ) − 1/c.∂A/∂t (30)
appearing “cold,” while negative charges effectively B = curl(A) (31)
appearing “hot”).
These equations then make a clearer distinction
In this case, over the given time interval, energy is between the “thermal” and the “electric” source contri-
absorbed or evolved from the charge-field interacting butions to the electromagnetic fields.
system accordingly as the signs of the charge and the
temporal field are the same or opposite. Since this scalar P. W. Bridgman[2] observed, in 1961, that thermoelec-
energy does not require the charge to move in space, tric phenomena require the phenomenological description
in order to materialize as some observable physical of e.m.f to allow for two different kinds of electromo-
phenomena the energy that is absorbed and/or evolved tive force, one that provides what he calls the “working”
must manifest as a form of heat. Moreover, this heat is e.m.f, and the other that provides the “driving” e.m.f,
proportional to the first power of the charge, and thus for the thermoelectric system. The “working” e.m.f is
reverses sign with the change in sign of the charge, or responsible for the production of the total energy that
correspondingly a change in sign of the electric current, emerges from the system, while the “driving” e.m.f is re-
making this a reversible heat, corresponding to the sponsible for moving the charges in the system, giving
experimental observations already known as Peltier and rise to the electric current.
Thomson Heats in thermoelectricity. These two e.m.fs, traditionally considered the same
normally in electricity, are not the same when including
Thus, the temporal field, T = −1/c.∂U/∂t + div(A), thermoelectric effects.
represents the total heat energy per unit charge evolving Bridgman invents a thermodynamic construction to
per unit time (i.e. time measured in units of length) define these two phenomenologically required e.m.fs,
from the charge-field interacting system due to both the but he emphasises that since these are constructions
loss in electrostatic potential energy at the location and they are not directly observable. Here we find an
the flow of electrodynamic momentum out of the same alternative explanation of Bridgman’s idea of the two
location, much like the diffusion heat flow being due e.m.fs, on grounds much more fundamentally linked to
to the sum of two different processes, conduction and the electromagnetic equations and not requiring his ad
convection, in nonequilibrium thermodynamics. hoc thermodynamic arguments.

Now, instead of using the symmetric and antisymmet- If we take our new Coulomb-Gauss law, eqn (27), and
ric derivatives, the two quaternion electromagnetic equa- separate the sources so that we isolate each effect in or-
tions can also be written using the left and right deriva- der to consider the impact of one independently of the
tives of the potential. other, we find the corresponding result to Bridgman’s
two e.m.fs.
d/dr → (d/dr → A) + (A ← d/dr) ← d/dr = 0 (22) We set the free electric charge density to zero and label
d/dr → (A ← d/dr) − (d/dr → A) ← d/dr = 0 (23) the electric field, whose source is given by the temporal
 4

term, +1/c.∂T /∂t, alone, with a subscript to indicate dT /dK = −1/(qc).d(−qT c)/dK, and -qTc is the heat
this part arises from the temporal source; ET . energy absorbed per unit time by the charge, q, we see
that dT/dK is effectively the measure of a type of “heat
capacity” of the unit charge within the particular mate-
div(ET ) = +1/c.∂T /∂t (32) rial.
Then, by subtracting this electric field from the to- Consequently, from the new Coulomb-Gauss law, we
tal field we obtain the Coulomb-Gauss law we are more infer that an electric field, ET , is induced by a time rate
familiar with, consisting of just the free electric charge of change in temperature, the magnitude of the field be-
density as source. ing determined by this special “heat capacity” of a unit
charge within the material and the rate of change of tem-
div(E − ET ) = 4πρ (33) perature with time. The direction of the field is radially
away from the point at which the temperature change
Here, we can define E, as the “driving” field, since this occurs.
is clearly responsible for moving the charges through the
region of space. Then, E−ET , is the “working” field, that In an homogeneous isotropic medium, the neighbour-
determines the total energy (i.e. net energy) delivered ing points also produce similar fields, but being oppo-
by the field in the act of moving those charges—part of sitely directed, they cancel each other, leaving no net
the energy that would normally be observed due to a field within the medium. However, where there is an
charge moving under the influence of that total electric anisotropy introduced into the medium, such as a tem-
field being now reabsorbed instead by the mechanics of perature gradient within the material, a net field will
the reversible thermoelectric heat in the internal energy emerge along the direction of that anisotropy. This is
conversion process. consistent with Bridgman’s thermoelectric e.m.f induced
It is not clear to me, however, whether these two by changing temperature with time, which requires the
e.m.fs should simply replace Bridgman’s constructions, presence of an isothermal electric current, thus providing
or whether Bridgman’s e.m.fs should be considered to the requisite anisotropy, and also consistent with the py-
exist independently, thus masking the presence of the roelectric effect where the mechanical stress and strain
quaternion results. provide the anisotropic conditions. Thus, here again, we
find the effects of the temporal field being masked by
Bridgman also notes that symmetric thermoelectric ar- other theoretically anticipated and experimentally known
guments require that a time varying temperature will in- phenomena.
duce a related thermoelectric electromotive force in the
material undergoing the variation of temperature, an ef-
fect that has not yet been experimentally observed.
Now, we would expect that the temporal field, T, in an
C. Seebeck Effect
homogeneous material, would itself be uniform through-
out the material medium, and show no direct indepen-
dent variation in space or time, except that, given the Now, consider an electric charge moving from one ma-
close relation to heat energy, the temporal field would terial medium into another, say two different metals.
vary directly with temperature, K, and would conse- From the rest frame of the moving charge, we find that
quently show an indirect space variation and time varia- there is a jump in the temporal field, T, at the time the
tion to the extent that the temperature itself varies with charge is seen to cross the boundary from one metal and
space and time.[3] enter the other metal in the lab frame. This time rate of
change of temporal field is seen as an electric field, ET ,
by the charge, according to our new Coulomb-Gauss law,
B. dT/dK and this produces an electric force on the charge that ac-
celerates it from rest in its instantaneous rest frame, or
This relationship between the temporal field and tem- just accelerates the charge in the lab frame, and this is
perature can be more clearly expressed functionally, T = the source of the Seebeck thermoelectric e.m.f.
T (K). And with this in mind, we conclude that the time Note, no static surface charges are required at the
rate of change of the temporal field is proportional to the boundary between the two metals to establish an electro-
time rate of change of the temperature, static field that will then accelerate the charges to sustain
1 ∂T 1 dT ∂K the current. The jump in the temporal field between the
= (34) two metals, which is essentially related to the jump in the
c ∂t c dK ∂t
specific heat capacities per unit charge and the jump in
In fact, for any given homogeneous material medium, electrical conductivity, is the sole source of the manifest
the measure of the temporal field is simply a measure e.m.f. Randomly moving charges that cross the bound-
of the temperature, because the parameter dT/dK is a ary will be accelerated by this e.m.f, and a closed circuit
characteristic of the medium. Given that we can write, will thus sustain a current.
 5

D. Thomson Effect from the “irreversible heat” due to Joule heating by

measuring the total heat produced by the current
Consider the new form of Ampere’s law. If we multi- flowing in one direction, then reversing the direction of
ply both sides of the equation (25) by the ratio of cur- the current and measuring the total heat again. The
rent to electrical conductivity, J/σ, using the vector “dot difference between the two total heats is then twice the
product,” then write the gradient of the temporal field Thomson Heat.
in terms of the temperature gradient, and re-arrange the
terms, we obtain four quantities that sum to zero, We are now in a position to interpret the special “heat
capacity” parameter, dT/dK, introduced above. It is es-
J2 c dT sentially equivalent to the product of the Thomson Spe-
+ J · grad(K) (35) cific Heat and the Electrical Conductivity of the medium,
σ 4πσ dK
c 1 hT · σ, and thus it is indeed a characteristic of the ma-
− J · curl(B) + J · ∂E/∂t = 0. terial.
4πσ 4πσ

The first term we recognise as the Joule Heat produced IV. POLARIZATION AND MAGNETIZATION.
by an electrical current. It is proportional to the square of
the electrical current density, J2 , and is thus independent We can identify the net electric field with the electric
of the direction of the current or the sign of the charge displacement, D = E − ET . Then, if we define the
carriers. magnetic field, BT , to be that which solves the equation,
The second term we recognise has the form of the curl(BT ) = +1/c.∂ET /∂t + grad(T ), we can write,
Thomson Heat produced by an electrical current flowing H = B − BT , and obtain the more familiar macroscopic
up a temperature gradient. It is linear in the electrical equations.
current density, J, and thus reverses sign when the cur-
rent reverses direction or the signs of the charge carriers We can then argue that in a macroscopic medium, the
are reversed. effects due to the presence of these new terms, grad(T )
These two terms must just balance the last two terms and +1/c.∂T /∂t, are normally hidden in the complexities
in this equation, for a thermally isolated electromagnetic of the Polarization and Magnetization parameters that
system. But if we design an experiment to reduce the lat- describe the material, and thus, in many instances, the
ter two terms to zero, say, for example, we arrange things effects of the temporal field are not easily separated to
so that the electric field is constant in time, ∂E/∂t = 0, be identified and measured as independent phenomena.
and the circulation of the magnetic field is perpendicu-
lar to the current flow, J · curl(B) = 0, then the system
can no longer be thermally isolated, in general, and we V. CONCLUSION.
must either pump heat energy into or extract heat out
to maintain our particular requirements.
When placing this electromagnetic system in contact When James Clerk Maxwell[4] wrote the second
with heat reservoirs, therefore, and allowing heat to be edition of his Treatise on Electricity and Magnetism
exchanged between the system and the reservoirs, in such he included a quaternion representation of his elec-
a way that the last two terms vanish, we obtain the con- tromagnetic equations, but he did not include both
ditions that characterize the Thomson Heat experiment. left-hand and right-hand derivatives, and the differential
The rate of net heat energy exchanged with the reserv- operator nabla was restricted to the 3-dimensional space
iors, dQ/dt, is now equal to the sum of the first two heats, form lacking a time component, and so his work is
and the equation becomes the usual Thomson Heat equa- fundamentally different from that presented here.
tion of thermoelectricity, which says the rate of produc-
tion of heat by the system is the sum of the irreversible Indeed, in the calculus of quaternions the differential
Joule Heat and the reversible Thomson Heat, operator almost always appears on the left acting
towards the variable on the right, ignoring the other al-
dQ ternative. And even though, Charles Jasper Joly[5] notes
= J2 /σ − hT J · grad(K) (36) the distinction in his book A Manual of Quaternions,
dt
c dT the importance of the idea goes unnoticed, unexplored,
hT = − (37) and unused. As a consequence of this, an important
4πσ dK
field component went missing in Maxwell’s Equations,
Where now, hT , is the Thomson Specific Heat of and all of modern physics has developed from there
the material, which is defined to be the quantity of perpetuating one of the consequences of this oversight,
“reversible heat” absorbed per unit time by an electrical namely, that the electromagnetic field possesses six
current of unit strength flowing up a temperature components, whereas, as we have shown, there should
gradient of one degree per unit length in a wire of unit be seven.
area cross section. The “reversible heat” is separated
 6

Our macroscopic experience tells us that heat is pro- particle is complete—we infer such a particle must have
duced by two opposing agents acting, the one against the an extended structure with a variable intrinsic pulse in
other, rubbing as it were, as in the familiar case of me- addition to its quantum mechanically determined fixed
chanical friction, to produce the gross fire that manifests intrinsic spin.
as heat when there is contact with matter. So, when
we find the electric field is also the sum of two oppos- This brief note introduces the essential ideas that will
ing principles, the left-handed tension acting against the be explored more fully in a future paper.
right-handed tension, we should not be surprised to find
a heat component, a more subtle fire, hiding within the
field, and emerging as heat when the field comes into
contact with charged matter.
Indeed, the action of a material body can be generally
Acknowledgments
described in terms of three transformations: translations,
rotations, and pulsations. The electric field, E, tends to
induce translations in a test charge, while the magnetic I wish to acknowledge the encouragement from the
field, B, tends to induce rotations when the charge is usenet community of sci.physics, sci.physics.particle, and
moving, we are then left to conclude that the temporal sci.math, where the lively discussion my ideas began in
field, T, tends to induce the pulsations. 1995, and to acknowledge Prof. Pertti Lounesto in partic-
Thus, with the inclusion of the missing temporal ular, who kept prompting me to get these ideas published
field the description of the action of a charged material in a more formal medium.

[1] W. R. Hamilton, 1844, On a new species of Imagi- [3] We use the letter, K, for temperature, instead of the more
nary Quantities connected with the Theory of Quaternions usual letter, T, since the latter is being used here for the
[communicated November 13, 1843], Ir. Acad. Proc., II, temporal field, and we can remember this as K for Kelvin
424-434. instead of T for Thomson, given that William Thomson
[2] P. W. Bridgman, 1961, The Thermodynamics of Elec- became Lord Kelvin, hence the promotion of the letter.
trical Phenomena in Metals and a Condensed Collec- [4] J. C. Maxwell, 1954, Treatise on Electricity and Mag-
tion of Thermodynamic Formulas, Dover Publications. – netism, 3rd ed., 2 vols, Dover, New York.
for definitions of “working” and “driving” electromotive [5] C. J. Joly, 1905, A Manual of Quaternions, London
forces, see comments in preface page v, and text pages Macmillan. In Art 57, Joly recognises the two different left
19,61,63,69,129ff. And for comments relating to the ex- and right differentiations; pp.74-77, and Exercises Ex.5,
pected, but as yet unseen, “E.M.F. produced by temper- Ex.11, on pg.76.
ature varying with time,” see preface page vi, and text
pages 144-145.