# Doing Physics with

Quaternions
Douglas B. Sweetser
©2005 doug <sweetser@alum.mit.edu> All righs reserved.
INDEX
Introduction
2 What are Quaternions?
3 Unifying Two Views of Events
4 A Brief History of Quaternions
Mathematics
6 Multiplying Quaternions the Easy Way
7 Scalars, Vectors, Tensors and All That
11 Inner and Outer Products of Quaternions
13 Quaternion Analysis
23 Topological Properties of Quaternions
28 Quaternion Algebra Tool Set
Classical Mechanics
32 Newton’s Second Law
35 Oscillators and Waves
37 Four Tests of for a Conservative Force
Special Relativity
40 Rotations and Dilations Create the Lorentz Group
43 An Alternative Algebra for Lorentz Boosts
Electromagnetism
48 Classical Electrodynamics
51 Electromagnetic Field Gauges
53 The Maxwell Equations in the Light Gauge: QED?
56 The Lorentz Force
58 The Stress Tensor of the Electromagnetic Field
Quantum Mechanics
62 A Complete Inner Product Space with Dirac’s Bracket Notation
67 Multiplying quaternions in Polar Coordinate Form
69 Commutators and the Uncertainty Principle
74 Unifying the Representations of Integral and Half−Integral Spin
79 Deriving A Quaternion Analog to the Schrödinger Equation
83 Introduction to Relativistic Quantum Mechanics
86 Time Reversal Transformations for Intervals
Gravity
89 Unified Field Theory by Analogy
101 Einstein’s vision I: Classical unified field equations for gravity and electromagnetism
using Riemannian quaternions
115 Einstein’s vision II: A unified force equation with constant velocity profile solutions
123 Strings and Quantum Gravity
127 Answering Prima Facie Questions in Quantum Gravity Using Quaternions
134 Length in Curved Spacetime
136 A New Idea for Metrics
138 The Gravitational Redshift

140 A Brief Summary of Important Laws in Physics Written as Quaternions

155 Conclusions
1

What Are Quaternions?
Quaternions are numbers like the real numbers: they can be added, subtracted, multiplied, and
divided. There is something odd about them, hinted at by the Latin. Quaternions are composed of
four numbers that work together as one. They were discovered by several people back in the
eighteen hundreds. Some enthusiasts thought quaternions would be able to express everything that
could happen in our three dimensions of space and one for time because quaternions naturally had
that form too. Math accidents do not happen − they revel deep things about how Nature works.
Unfortunately, the big fans of quaternion mathematics claimed far more than they would deliver.
Useful ideas born from initial quaternion work − for example the notion of scalars, vectors, div,
grad, and curl − were stripped out of their initial context, and made more "general". I use the
quotes because from my viewpoint, there is nothing more general than a number that can be added,
subtracted, multiplied, and divided. I am trying to continue the project of applying four−dimen-
sional quaternions to the four−dimensional spacetime we live in.
2

Unifying Two Views of Events
An experimentalist collects events about a physical system. A theorists builds a model to describe
what patterns of events within a system might generate the experimentalist’s data set. With hard
work and luck, the two will agree!
Events are handled mathematically as 4−vectors. They can be added or subtracted from another, or
multiplied by a scalar. Nothing else can be done. A theorist can import very powerful tools to
generate patterns, like metrics and group theory. Theorists in physics have been able to construct
the most accurate models of Nature in all of science.
I hope to bring the full power of mathematics down to the level of the events themselves. This may
be done by representing events as the mathematical field of quaternions. All the standard tools for
creating mathematical patterns − multiplication, trigonometric functions, transcendental functions,
infinite series, the special functions of physics − should be available for quaternions. Now a theo-
rist can create patterns of events with events. This may lead to a better unification between the
work of a theorist and the work of an experimentalist.
An Overview of Doing Physics with Quaternions
It has been said that one reason physics succeeds is because all the terms in an equation are tensors
of the same rank. This work challenges that assumption, proposing instead an integrated set of
equations which are all based on the same 4−dimensional mathematical field of quaternions.
Mostly this document shows in cookbook style how quaternion equations are equivalent to
approaches already in use. As Feynman pointed out, "whatever we are allowed to imagine in
science must be consistent with everything else we know." Fresh perspectives arise because, in
essence, tensors of different rank can mix within the same equation. The four Maxwell equations
become one nonhomogeneous quaternion wave equation, and the Klein−Gordon equation is part of
a quaternion simple harmonic oscillator. Even gravity may be part of a simple quaternion wave
equation, a research topic of much interest to me. Since all of the tools used are woven from the
same mathematical fabric, the interrelationships become more clear. Hope you enjoy.
3

A Brief History of Quaternions
Complex numbers were a hot subject for research in the eighteen hundreds. An obvious question
was that if a rule for multiplying two numbers together was known, what about multiplying three
numbers? For over a decade, this simple question had bothered Hamilton, the big mathematician
of his day. The pressure to find a solution was not merely from within. Hamilton wrote to his son:
"Every morning in the early part of the above−cited month [Oct. 1843] on my coming down to
breakfast, your brother William Edwin and yourself used to ask me, ’Well, Papa, can you multiply
triplets?’ Whereto I was always obliged to reply, with a sad shake of the head, ’No, I can only add
and subtract them.’"
We can guess how Hollywood would handle the Brougham Bridge scene in Dublin. Strolling
along the Royal Canal with Mrs. H−, he realizes the solution to the problem, jots it down in a
notebook. So excited, he took out a knife and carved the answer in the stone of the bridge.
Hamilton had found a long sought−after solution, but it was weird, very weird, it was 4D. One of
the first things Hamilton did was get rid of the fourth dimension, setting it equal to zero, and call-
ing the result a "proper quaternion." He spent the rest of his life trying to find a use for quater-
nions. By the end of the nineteenth century, quaternions were viewed as an oversold novelty.
In the early years of this century, Prof. Gibbs of Yale found a use for proper quaternions by reduc-
ing the extra fluid surrounding Hamilton’s work and adding key ingredients from Rodrigues con-
cerning the application to the rotation of spheres. He ended up with the vector dot product and
cross product we know today. This was a useful and potent brew. Our investment in vectors is
enormous, eclipsing their place of birth (Harvard had >1000 references under "vector", about 20
under "quaternions", most of those written before the turn of the century).
In the early years of this century, Albert Einstein found a use for four dimensions. In order to
make the speed of light constant for all inertial observers, space and time had to be united. Here
was a topic tailor−made for a 4D tool, but Albert was not a math buff, and built a machine that
worked from locally available parts. We can say now that Einstein discovered Minkowski space-
time and the Lorentz transformation, the tools required to solve problems in special relativity.
Today, quaternions are of interest to historians of mathematics. Vector analysis performs the daily
mathematical routine that could also be done with quaternions. I personally think that there may be
4D roads in physics that can be efficiently traveled only by quaternions, and that is the path which
is laid out in these web pages.
In a longer history, Gauss would get the credit for seeing quaternions first in one of his notebooks.
Rodrigues developed 3D rotations all on his own also in the 1840’s. The Pauli spin matrices and
Penrose’s spinors are reinventions of the wheel that miss out on division. Although I believe that is
a major omission and cause of subtle flaws at the foundations of modern physics, spin matrices and
spinors have many more adherents today than quaternions.
4

Mathematics
5

Multiplying Quaternions the Easy Way
Multiplying two complex numbers a + b I and c + d I is straightforward.
(a, b) (c, d) = (ac - bd, ad + bc)
For two quaternions, b I and d I become the 3−vectors B and D, where B = x I + y J + z K and
similarly for D. Multiplication of quaternions is like complex numbers, but with the addition of the
cross product.
(a, B

) (c, D

) = (ac - B

.D

, a D

+ B

c + B

x D

)
A mnemonic: (firsts − lasts, outside + inside + the cross). Note that the cross product would change
its sign if the order of multiplication were reversed, unlike the other terms. That is why quater-
nions in general do not commute.
If a is the operator d/dt, and B is the del operator, or d/dx I + d/dy J + d/dz K (all partial deriva-
tives), then these operators act on the scalar function c and the 3−vector function D in the follow-
ing manner:
¸
d
-------
dt
,

¸ (c, D

) =

dc
-------
dt
-

.D

,
d D

--------
dt
+

c +

x D
|

This one quaternion contains the time derivatives of the scalar and 3−vector functions, along with
the divergence, the gradient and the curl. Dense notation :−)
6

Scalars, Vectors, Tensors and All That
According to my math dictionary, a tensor is ...
"An abstract object having a definitely specified system of components in every coordinate system
under consideration and such that, under transformation of coordinates, the components of the
object undergoes a transformation of a certain nature."
To make this introduction less abstract, I will confine the discussion to the simplest tensors under
rotational transformations. A rank−0 tensor is known as a scalar. It does not change at all under a
rotation. It contains exactly one number, never more or less. There is a zero index for a scalar. A
rank−1 tensor is a vector. A vector does change under rotation. Vectors have one index which can
run from 1 to the number of dimensions of the field, so there is no way to know a priori how many
numbers (or operators, or ...) are in a vector. n−rank tensors have n indices. The number of num-
bers needed is the number of dimensions in the vector space raised by the rank. Symmetry can
often simplify the number of numbers actually needed to describe a tensor.
There are a variety of important spin−offs of a standard vector. Dual vectors, when multiplied by
its corresponding vector, generate a real number, by systematically multiplying each component
from the dual vector and the vector together and summing the total. If the space a vector lives in is
shrunk, a contravariant vector shrinks, but a covariant vector gets larger. A tangent vector is, well,
tangent to a vector function.
Physics equations involve tensors of the same rank. There are scalar equations, polar vector equa-
tions, axial vector equations, and equations for higher rank tensors. Since the same rank tensors
are on both sides, the identity is preserved under a rotational transformation. One could decide to
arbitrarily combine tensor equations of different rank, and they would still be valid under the
transformation.
There are ways to switch ranks. If there are two vectors and one wants a result that is a scalar, that
requires the intervention of a metric to broker the transaction. This process in known as an inner
tensor product or a contraction. The vectors in question must have the same number of dimen-
sions. The metric defines how to form a scalar as the indices are examined one−by−one. Metrics
in math can be anything, but nature imposes constraints on which ones are important in physics.
An aside: mathematicians require the distance is non−negative, but physicists do not. I will be
using the physics notion of a metric. In looking at events in spacetime (a 4−dimensional vector),
the axioms of special relativity require the Minkowski metric, which is a 4x4 real matrix which has
down the diagonal 1, −1, −1, −1 and zeros elsewhere. Some people prefer the signs to be flipped,
but to be consistent with everything else on this site, I choose this convention. Another popular
choice is the Euclidean metric, which is the same as an identity matrix. The result of general
relativity for a spherically symmetric, non−rotating mass is the Schwarzschild metric, which has
"non−one" terms down the diagonal, zeros elsewhere, and becomes the Minkowski metric in the
limit of the mass going to zero or the radius going to infinity.
An outer tensor product is a way to increase the rank of tensors. The tensor product of two vectors
will be a 2−rank tensor. A vector can be viewed as the tensor product of a set of basis vectors.
7

What Are Quaternions?
Quaternions could be viewed as the outer tensor product of a scalar and a 3−vector. Under rotation
for an event in spacetime represented by a quaternion, time is unchanged, but the 3−vector for
space would be rotated. The treatment of scalars is the same as above, but the notion of vectors is
far more restrictive, as restrictive as the notion of scalars. Quaternions can only handle 3−vectors.
To those familiar to playing with higher dimensions, this may appear too restrictive to be of inter-
est. Yet physics on both the quantum and cosmological scales is confined to 3−spatial dimensions.
Note that the infinite Hilbert spaces in quantum mechanics a function of the principle quantum
number n, not the spatial dimensions. An infinite collection of quaternions of the form (En, Pn)
could represent a quantum state. The Hilbert space is formed using the Euclidean product (q* q’).
A dual quaternion is formed by taking the conjugate, because q* q = (t^2 + X.X, 0). A tangent
quaternion is created by having an operator act on a quaternion−valued function
¸
o
-------
ot
, V

¸ (f (q), F

(q)) =

of
-------
ot
- V

.F

,
oF

-------
ot
+ V

f + V

X F
|

What would happen to these five terms if space were shrunk? The 3−vector F would get shrunk, as
would the divisors in the Del operator, making functions acted on by Del get larger. The scalar
terms are completely unaffected by shrinking space, because df/dt has nothing to shrink, and the
Del and F cancel each other. The time derivative of the 3−vector is a contravariant vector, because
F would get smaller. The gradient of the scalar field is a covariant vector, because of the work of
the Del operator in the divisor makes it larger. The curl at first glance might appear as a draw, but
it is a covariant vector capacity because of the right−angle nature of the cross product. Note that if
time where to shrink exactly as much as space, nothing in the tangent quaternion would change.
A quaternion equation must generate the same collection of tensors on both sides. Consider the
product of two events, q and q’:
(t, X

) ¡t
·
, X
·

¡ = ¡t t
·
- X

.X
·

, t X
·

+ X

t
·
+ X

x X
·

¡
scalars : t, t
·
, tt
·
- X

.X
·

polar vectors : X

, X
·

, t X
·

+ X

t
·
axial vectors : X

x X
·

Where is the axial vector for the left hand side? It is imbedded in the multiplication operation,
honest :−)
¡t
·
, X
·

¡ (t, X

) = ¡t
·
t - X
·

.X

, t
·
X

+ X
·

t + X
·

x X

¡
= ¡t t
·
- X

.X
·

, t X
·

+ X

t
·
- X

x X
·

¡
The axial vector is the one that flips signs if the order is reversed.
Terms can continue to get more complicated. In a quaternion triple product, there will be terms of
the form (XxX’).X". This is called a pseudo−scalar, because it does not change under a rotation,
but it will change signs under a reflection, due to the cross product. You can convince yourself of
this by noting that the cross product involves the sine of an angle and the dot product involves the
cosine of an angle. Neither of these will change under a rotation, and an even function times an odd
function is odd. If the order of quaternion triple product is changed, this scalar will change signs
for at each step in the permutation.
8

Terms can continue to get more complicated. In a quaternion triple product, there will be terms of
the form (XxX’).X". This is called a pseudo−scalar, because it does not change under a rotation,
but it will change signs under a reflection, due to the cross product. You can convince yourself of
this by noting that the cross product involves the sine of an angle and the dot product involves the
cosine of an angle. Neither of these will change under a rotation, and an even function times an odd
function is odd. If the order of quaternion triple product is changed, this scalar will change signs
for at each step in the permutation.
It has been my experience that any tensor in physics can be expressed using quaternions. Some-
times it takes a bit of effort, but it can be done.
Individual parts can be isolated if one chooses. Combinations of conjugation operators which flip
the sign of a vector, and symmetric and antisymmetric products can isolate any particular term.
Here are all the terms of the example from above
(t, X

) ¡t
·
, X
·

¡ = ¡t t
·
- X

.X
·

, t X
·

+ X

t
·
+ X

x X
·

¡
scalars : t =
q + q
-
--------------
2
, t
·
=
q’ + q’
-
---------------------
2
,
tt
·
- X

.X
·

=
qq’ + (qq’)
-
-------------------------------
2
polar vectors : X

=
q - q
-
--------------
2
, X
·

=
q’ - q’
-
---------------------
2
,
t X
·

+ X

t
·
=
(qq’ + (q’ q)) - (qq’ + (q’ q))
-
-----------------------------------------------------------------------------
4
axial vectors : X

x X
·

=
qq’ - (q’ q)
------------------------------
2
The metric for quaternions is imbedded in Hamilton’s rule for the field.
i

2
= j

2
= k

2
= i

j

k

= -1
This looks like a way to generate scalars from vectors, but it is more than that. It also says implic-
itly that i j = k, j k = i, and i, j, k must have inverses. This is an important observation, because it
means that inner and outer tensor products can occur in the same operation. When two quaternions
are multiplied together, a new scalar (inner tensor product) and vector (outer tensor product) are
formed.
How can the metric be generalized for arbitrary transformations? The traditional approach would
involve playing with Hamilton’s rules for the field. I think that would be a mistake, since that rule
involves the fundamental definition of a quaternion. Change the rule of what a quaternion is in one
context and it will not be possible to compare it to a quaternion in another context. Instead, con-
sider an arbitrary transformation T which takes q into q’
q - q’ = T q
T is also a quaternion, in fact it is equal to q’ q^−1. This is guaranteed to work locally, within
neighborhoods of q and q’. There is no promise that it will work globally, that one T will work for
any q. Under certain circumstances, T will work for any q. The important thing to know is that a
transformation T necessarily exists because quaternions are a field. The two most important theo-
ries in physics, general relativity and the standard model, involve local transformations (but the
technical definition of local transformation is different than the idea presented here because it
involves groups).
This quaternion definition of a transformation creates an interesting relationship between the
Minkowski and Euclidean metrics.
Let T = I, the identity matrix
9

I q I q + (I q I q)
-
---------------------------------------------
2
= (t
2
- X

.X

, 0)
(I q)
-
I q = (t
2
+ X

.X

, 0)
In order to change from wrist watch time (the interval in spacetime) to the norm of a Hilbert space
does not require any change in the transformation quaternion, only a change in the multiplication
step. Therefore a transformation which generates the Schwarzschild interval of general relativity
should be easily portable to a Hilbert space, and that might be the start of a quantum theory of
gravity.
So What Is the Difference?
I think it is subtle but significant. It goes back to something I learned in a graduate level class on
the foundations of calculus. To make calculus rigorous requires that it is defined over a mathemati-
cal field. Physicists do this be saying that the scalars, vectors and tensors they work with are
defined over the field of real or complex numbers.
What are the numbers used by nature? There are events, which consist of the scalar time and the
3−vector of space. There is mass, which is defined by the scalar energy and the 3−vector of momen-
tum. There is the electromagnetic potential, which has a scalar field phi and a 3−vector potential A.
To do calculus with only information contained in events requires that a scalar and a 3−vector form
a field. According to a theorem by Frobenius on finite dimensional fields, the only fields that fit
are isomorphic to the quaternions (isomorphic is a sophisticated notion of equality, whose subtle-
ties are appreciated only by people with a deep understanding of mathematics). To do calculus
with a mass or an electromagnetic potential has an identical requirement and an identical solution.
This is the logical foundation for doing physics with quaternions.
Can physics be done without quaternions? Of course it can! Events can be defined over the field
of real numbers, and then the Minkowski metric and the Lorentz group can be deployed to get
every result ever confirmed by experiment. Quantum mechanics can be defined using a Hilbert
space defined over the field of complex numbers and return with every result measured to date.
Doing physics with quaternions is unnecessary, unless physics runs into a compatibility issue.
Constraining general relativity and quantum mechanics to work within the same topological alge-
braic field may be the way to unite these two separately successful areas.
10

Inner and Outer Products of Quaternions
A good friend of mine has wondered what is means to multiply two quaternions together (this
question was a hot topic in the nineteenth century). I care more about what multiplying two quater-
nions together can accomplish. There are two basic ways to do this: just multiply one quaternion
by another, or first take the transpose of one then multiply it with the other. Each of these products
can be separated into two parts: a symmetric (inner product) and an antisymmetric (outer product)
components. The symmetric component will remain unchanged by exchanging the places of the
quaternions, while the antisymmetric component will change its sign. Together they add up to the
product. In this section, both types of inner and outer products will be formed and then related to
physics.
The Grassman Inner and Outer Products
There are two basic ways to multiply quaternions together. There is the direct approach.
(t, X

) ¡t
·
, X
·

¡ = ¡t t
·
- X

.X
·

, t X
·

+ X

t
·
+ X

x X
·

¡
I call this the Grassman product (I don’t know if anyone else does, but I need a label). The inner
product can also be called the symmetric product, because it does not change signs if the terms are
reversed.
even ¡(t, X

), ¡t
·
, X
·

¡¡ ÷
÷
(t, X

) (t
·
, X
·

) + (t
·
, X
·

) (t, X

)
----------------------------------------------------------------------------
2
=
¡t t
·
- X

.X
·

, t X
·

+ X

t
·
¡
I have defined the anticommutator (the bold curly braces) in a non−standard way, including a
factor of two so I do not have to keep remembering to write it. The first term would be the Lorentz
invariant interval if the two quaternions represented the same difference between two events in
spacetime (i.e. t1=t2=delta t,...). The invariant interval plays a central role in special relativity.
The vector terms are a frame−dependent, symmetric product of space with time and does not
appear on the stage of physics, but is still a valid measurement.
The Grassman outer product is antisymmetric and is formed with a commutator.
odd ¡(t, X

), ¡t
·
, X
·

¡¡ ÷
÷
(t, X

) (t
·
, X
·

) - (t
·
, X
·

) (t, X

)
----------------------------------------------------------------------------
2
= ¡0, X

x X
·

¡
This is the cross product defined for two 3−vectors. It is unchanged for quaternions.
11

The Euclidean Inner and Outer Products
Another important way to multiply a pair of quaternions involves first taking the transpose of one
of the quaternions. For a real−valued matrix representation, this is equivalent to multiplication by
the conjugate which involves flipping the sign of the 3−vector.
(t, X

)
-
¡t
·
, X
·

¡ = (t, -X

) ¡t
·
, X
·

¡
= ¡t t
·
+ X

.X
·

, t X
·

- X

t
·
- X

x X
·

¡
Form the Euclidean inner product.
(t, X

)
-
(t
·
, X
·

) + (t
·
, X
·

)
-
(t, X

)
--------------------------------------------------------------------------------
2
= ¡t t
·
+ X

.X
·

, 0

¡
The first term is the Euclidean norm if the two quaternions are the same (this was the reason for
using the adjective "Euclidean"). The Euclidean inner product is also the standard definition of a
dot product.
Form the Euclidean outer product.
(t, X

)
-
(t
·
, X
·

) - (t
·
, X
·

)
-
(t, X

)
--------------------------------------------------------------------------------
2
= ¡0, t X
·

- X

t
·
- X

x X
·

¡
The first term is zero. The vector terms are an antisymmetric product of space with time and the
negative of the cross product.
The Euclidean product is non−associative:
(a b)
-
c = (a)
-
bc
The norms of Euclidean products are associative because the norms are real valued:
| (a b)
-
c | = | (a)
-
bc |
The Euclidean product of quaternions might be a way t connect to the algebra of octonions,a non−
associative division algebra.
Implications
When multiplying vectors in physics, one normally only considers the Euclidean inner product, or
dot product, and the Grassman outer product, or cross product. Yet, the Grassman inner product,
because it naturally generates the invariant interval, appears to play a role in special relativity.
What is interesting to speculate about is the role of the Euclidean product. The Euclidean product
might be a direct connection to the algebraic structure of quantum mechanics via Hilbert spaces.
12

Quaternion Analysis
Complex numbers are a subfield of quaternions. My hypothesis is that complex analysis should be
self−evident within the structure of quaternion analysis.
The challenge is to define the derivative in a way so that a left derivative always equals a right
derivative. If quaternions would only commute... Well, the scalar part of a quaternion does com-
mute. If, in the limit, the differential element converged to a scalar, then it would commute. This
idea can be defined precisely. All that is required is that the magnitude of the 3−vector goes to
zero faster than the scalar. This might initially appears as an unreasonable constraint. However,
there is an important application in physics. Consider a set of quaternions that represent events in
spacetime. If the magnitude of the 3−space vector is less than the time scalar, events are separated
by a timelike interval. It requires a speed less than the speed of light to connect the events. This is
true no matter what coordinate system is chosen.
Defining a Quaternion
A quaternion has 4 degrees of freedom, so it needs 4 real−valued variables to be defined:
q = (a
0
, a
1
, a
2
, a
3
)
Imagine we want to do a simple binary operation such as subtraction, without having to specify the
coordinate system chosen. Subtraction will only work if the coordinate systems are the same,
whether it is Cartesian, spherical or otherwise. Let e0, e1, e2, and e3 be the shared, but unspeci-
fied, basis. Now we can define the difference between two quaternion q and q’ that is independent
of the coordinate system used for the measurement.
dq = q’ - q = ((a
0
’ - a
0
) e
0
,
(a
1
’ - a
1
) e
1
/3, (a
2
’ - a
2
) e
2
/3, (a
3
’ - a
3
) e
3
/3)
What is unusual about this definition are the factors of a third. They will be necessary in order to
define a holonomic equation later in this section. Hamilton gave each element parity with the
others, a very reasonable approach. I have found that it is important to give the scalar and the sum
of the 3−vector parity. Without this "scale" factor on the 3−vector, change in the scalar is not
given its proper weight.
If dq is squared, the scalar part of the resulting quaternion forms a metric.
dq^2 = ¸da
0
2
e
0
2
+ da
1
2

e
1
2
---------
9
+ da
2
2

e
2
2
---------
9
+ da
3
2

e
3
2
---------
9
,
2 da
0
da
1
e
0

e
1
-------
3
, 2 da
0
da
2
e
0

e
2
-------
3
, 2 da
0
da
3
e
0

e
3
-------
3
¸
What should the connection be between the squares of the basis vectors? The amount of intrinsic
curvature should be equal, so that a transformation between two basis 3−vectors does not contain a
hidden bump. Should time be treated exactly like space? The Schwarzschild metric of general
relativity suggests otherwise. Let e1, e2, and e3 form an independent, dimensionless, orthogonal
basis for the 3−vector such that:
13

-
1
---------
e
1
2
= -
1
---------
e
2
2
= -
1
---------
e
3
2
= e
0
2
This unusual relationship between the basis vectors is consistent with Hamilton’s choice of 1, i, j, k
if e
0
2
= 1. For that case, calculate the square of dq:
dq
2
=

da
0
2
e
0
2
-
da
1
2
-------------
9 e
0
2
-
da
2
2
-------------
9 e
0
2
-
da
3
2
-------------
9 e
0
2
,
2 da
0

da
1
---------
3
, 2 da
0

da
2
---------
3
, 2 da
0

da
3
---------
3
|

The scalar part is known in physics as the Minkowski interval between two events in flat space-
time. If e0^2 does not equal one, then the metric would apply to a non−flat spacetime. A metric
that has been measured experimentally is the Schwarzschild metric of general relativity. Set
e
0
2
1 2 = (1 − 2 GM/c^2 R), and calculate the square of dq:
dq
2
=

da
0
2
¸1 -
2 G M
------------
c
2
R
¸ -
dA.dA
-------------------------------
9 (1 -
2 G M
---------
c
2
R
)
,
2 da
0

da
1
---------
3
, 2 da
0

da
2
---------
3
, 2 da
0

da
3
---------
3
|

This is the Schwarzschild metric of general relativity. Notice that the 3−vector is unchanged (this
may be a defining characteristic). There are very few opportunities for freedom in basic mathemati-
cal definitions. I have chosen this unusual relationships between the squares of the basis vectors to
make a result from physics easy to express. Physics guides my choices in mathematical definitions
:−)
An Automorphic Basis for Quaternion Analysis
A quaternion has 4 degrees of freedom. To completely specify a quaternion function on the mani-
fold
1
, it must also have four degrees of freedom. Three other linearly−independent variables
involving q can be defined using conjugates combined with rotations:
q
-
= (a
0
e
0
, -a
1
e
1
/3, -a
2
e
2
/3, -a
3
e
3
/3)
q
-1
= (- a
0
e
0
, a
1
e
1
/3, -a
2
e
2
/3, -a
3
e
3
/3) = (e
1
q e
1
)
-
q
-2
÷ (- a
0
e
0
, -a
1
e
1
/3, +a
2
e
2
/3, -a
3
e
3
/3) = (e
2
q e
2
)
-
The conjugate as it is usually defined (q*) flips the sign of all but the scalar. The q*1 flips the
signs of all but the e1 term, and q*2 all but the e2 term. The set q, q*, q*1, q*2 form the basis for
quaternion analysis on the
1
manifold. The conjugate of a conjugate should give back the origi-
nal quaternion.
(q
-
)
-
= q, (q
-1
)
-1
= q, (q
-2
)
-2
= q
Something subtle but perhaps directly related to spin happens looking at how the conjugates effect
products:
(q q’)
-
= q’
-
q
-
14

(q q’)
-1
= - q’
-1
q
-1
, (q q’)
-2
= - q’
-2
q
-2
(q q’ q q’)
-1
= q’
-1
q
-1
q’
-1
q
-1
The conjugate applied to a product brings the result directly back to the reverse order of the ele-
ments. The first and second conjugates point things in exactly the opposite way. The property of
going "half way around" is reminiscent of spin. A tighter link is explored in the section on integral
and half integral spin.
Future Timelike Derivative
Instead of the standard approach to quaternion analysis which focuses on left versus right deriva-
tives, I concentrate on the ratio of scalars to 3−vectors. This is natural when thinking about the
structure of Minkowski spacetime, where the ratio of the change in time to the change in 3−space
defines five separate regions: timelike past, timelike future, lightlike past, lightlike future, and
spacelike. There are no continuous Lorentz transformations to link these regions. Each region will
require a separate definition of the derivative, and they will each have distinct properties. I will
start with the simplest case, and look at a series of examples in detail.
Definition: The future timelike derivative:
Consider a covariant quaternion function f with a domain of H and a range of H. A future timelike
derivative to be defined, the 3−vector must approach zero faster than the positive scalar. If this is
not the case, then this definition cannot be used. Implementing these requirements involves two
limit processes applied sequentially to a differential quaternion D. First the limit of the three
vector is taken as it goes to zero, (D − D*)/2 −> 0. Second, the limit of the scalar is taken, (D +
D*)/2 −> +0 (the plus zero indicates that it must be approached with a time greater than zero, in
other words, from the future). The net effect of these two limit processes is that D−>0.
o f (q, q
-
, q
-1
, q
-2
)
--------------------------------------------------
o q
=
= limit as ¡d, 0

¡ -
+0 ¡limit as (d, D

) -
¡d, 0

¡ (f (q + (d, D

), q
-
, q
-1
, q
-2
) -
f (q, q
-
, q
-1
, q
-2
)) (d, D

)
-1
¡¡
The definition is invariant under a passive transformation of the basis.
The 4 real variables a0, a1, a2, a3 can be represented by functions using the conjugates as a basis.
f (q, q
-
, q
-1
, q
-2
) = a
0
=
e
0
(q + q
-
)
--------------------------
2
f = a
1
=
e
1
(q + q
-1
)
----------------------------
(- 2/3)
=
(q + q
-1
) e
1
----------------------------
(- 2/3)
f = a
2
=
e
2
(q + q
-2
)
----------------------------
(- 2/3)
=
(q + q
-2
) e
2
----------------------------
(- 2/3)
15

f = a
3
=
e
3
(q + q
-
+ q
-1
+ q
-2
)
-------------------------------------------------
(2/3)
=
(q + q
-
+ q
-1
+ q
-2
) e
3
-------------------------------------------------
(2/3)
Begin with a simple example:
f (q, q
-
, q
-1
, q
-2
) = a
0
=
e
0
(q + q
-
)
--------------------------
2
o a
0
-----------
o q
=
o a
0
-----------
o q
-
= lim ¡lim ¡(e
0
((q + (d, D

) + q
-
) - (q + q
-
)))
(2 (d, D

))
-1
¡¡ =
e
0
-------
2
o a
0
-------------
o q
-1
=
o a
0
-------------
o q
-2
= 0
The definition gives the expected result.
A simple approach to a trickier example:
f = a
1
=
e
1
(q + q
-1
)
----------------------------
(- 2/3)
o a
1
-----------
o q
=
o a
1
-------------
o q
-1
= lim ¡lim ¡(e
1
((q + (d, D

) + q
-1
) - (q + q
-1
)))
((-2/3) (d, D

))
-1
¡¡ = -
3 e
1
-----------
2
o a
1
-----------
o q
-
=
o a
1
-------------
o q
-2
= 0
So far, the fancy double limit process has been irrelevant for these identity functions, because the
differential element has been eliminated. That changes with the following example, where the e
1

is written on the right, but the result is the same.
f (q, q
-
, q
-1
, q
-2
) = a
1
=
(q + q
-1
) e
1
----------------------------
(- 2/3)
o a
1
-----------
o q
=
o a
1
-------------
o q
-1
=
= lim ¡lim ¡((q + (d, D

) + q
-1
) - (q + q
-1
))
e
1
((-2/3) (d, D

))
-1
¡¡ =
= lim ¡lim ¡(d, D

) e
1
((-2/3) (d, D

))
-1
¡¡ =
16

= lim ¡¡d, 0

¡ e
1
¡(-2/3) ¡d, 0

¡¡
-1
¡ = -
3 e
1
-----------
2
Because the 3−vector goes to zero faster than the scalar for the differential element, after the first
limit process, the remaining differential is a scalar so it commutes with any quaternion. This is
what is required to dance around the e1 and lead to the cancellation.
The initial hypothesis was that complex analysis should be a self−evident subset of quaternion
analysis. So this quaternion derivative should match up with the complex case, which is:
z = a + b i, b = (Z - Z
-
)/2 i
o b
---------
o z
= -
i
----
2
= -
o b
-----------
o z
-
These are the same result up to a factor of three. Quaternions have three imaginary axes.
The derivative of a quaternion applies equally well to polynomials.
let f = q
2
o f
---------
o q
= lim ¡lim ¡¡(q + (d, D

))
2
- q
2
¡ (d, D

)
-1
¡¡ =
= lim ¡lim
¡¡q
2
+ q (d, D

) + (d, D

) q + (d, D

)
2
- q
2
¡ (d, D

)
-1
¡¡ =
= lim ¡lim ¡q + (d, D

) q (d, D

)
-1
+ (d, D

)¡¡ =
= lim ¡2 q + ¡d, 0

¡¡ = 2 q
This is the expected result for this polynomial. It would be straightforward to show that all polyno-
mials gave the expected results.
Mathematicians might be concerned by this result, because if the 3−vector D goes to −D nothing
will change about the quaternion derivative. This is actually consistent with principles of special
relativity. For timelike separated events, right and left depend on the inertial reference frame, so a
timelike derivative should not depend on the direction of the 3−vector.
Analytic Functions
There are 4 types of quaternion derivatives and 4 component functions. The following table
describes the 16 derivatives for this set
17

\ a
0
a
1
a
2
a
3
o
------
o q
e
0
-----
2
e
1
---------
-2/3
e
2
---------
-2/3
e
3
-------
2/3
o
--------
o q
-
e
0
-----
2
0 0
e
3
-------
2/3
o
---------
o q
-1
0
e
1
---------
-2/3
0
e
3
-------
2/3
o
---------
o q
-2
0 0
e
2
---------
-2/3
e
3
-------
2/3
This table will be used extensively to evaluate if a function is analytic using the chain rule. Let’s
see if the identity function w = q is analytic.
Let w = q = ¡a
0
e
0
, a
1

e
1
-------
3
, a
2

e
2
-------
3
, a
3

e
3
-------
3
¡
Use the chain rule to calculate the derivative will respect to each term:
o w
-----------
o a
0

o a
0
-----------
o q
= e
0
e
0
-------
2
=
1
----
2
o w
-----------
o a
1

o a
1
-----------
o q
=
e
1
-------
3
e
1
--------------------
(-2/3)
=
1
----
2
o w
-----------
o a
2

o a
2
-----------
o q
=
e
2
-------
3
e
2
--------------------
(-2/3)
=
1
----
2
o w
-----------
o a
3

o a
3
-----------
o q
=
e
3
-------
3
e
3
-----------------
(2/3)
= -
1
----
2
Use combinations of these terms to calculate the four quaternion derivatives using the chain rule.
o w
---------
o q
=
o w
-----------
o a
0

o a
0
-----------
o q
+
o w
-----------
o a
1

o a
1
-----------
o q
+
o w
-----------
o a
2

o a
2
-----------
o q
+
o w
-----------
o a
3

o a
3
-----------
o q
=
1
----
2
+
1
----
2
+
1
----
2
-
1
----
2
= 1
o w
-----------
o q
-
=
o w
-----------
o a
0

o a
0
-----------
o q
-
+
o w
-----------
o a
3

o a
3
-----------
o q
-
=
1
----
2
-
1
----
2
= 0
o w
-------------
o q
-1
=
o w
-----------
o a
1

o a
1
-------------
o q
-1
+
o w
-----------
o a
3

o a
3
-------------
o q
-1
=
1
----
2
-
1
----
2
= 0
o w
-------------
o q
-2
=
o w
-----------
o a
2

o a
2
-------------
o q
-2
+
o w
-----------
o a
3

o a
3
-------------
o q
-2
=
1
----
2
-
1
----
2
= 0
This has the derivatives expected if w=q is analytic in q.
Another test involves the Cauchy−Riemann equations. The presence of the three basis vectors
changes things slightly.
18

Let u = (a
0
e
0
, 0, 0, 0), V

= ¡0, a
1

e
1
-------
3
, a
2

e
2
-------
3
, a
3

e
3
-------
3
¡
o u
-----------
o a
0

e
1
-------
3
=
o V

-----------
o a
1
e
0
,
o u
-----------
o a
0

e
2
-------
3
=
o V

-----------
o a
2
e
0
,
o u
-----------
o a
0

e
3
-------
3
=
o V

-----------
o a
3
e
0
This also solves a holonomic equation.
Scalar

o u
-----------
o a
0
,
o V

-----------
o a
1
,
o V

-----------
o a
2
,
o V

-----------
o a
3
|

(e
0
, e
1
, e
2
, e
3
)
|

=
e
0
e
0
+
e
1
-------
3
e
1
+
e
2
-------
3
e
2
+
e
3
-------
3
e
3
= 0
There are no off diagonal terms to compare.
This exercise can be repeated for the other identity functions. One noticeable change is in the role
that the conjugate play for the basis vectors. Consider the identity function w = q*1. To show that
this is analytic in q*1 requires that one always works with basis vectors of the q*1 variety.
Let u = (- a
0
e
0
, 0, 0, 0),
V

= ¡0, a
1

e
1
-------
3
, - a
2

e
2
-------
3
, - a
3

e
3
-------
3
¡
o u
-----------
o a
0
¡-
e
1
-------
3
¡ =
o V

-----------
o a
1
e
0
,
o u
-----------
o a
0

e
2
-------
3
=
o V

-----------
o a
2
e
0
,
o u
-----------
o a
0

e
3
-------
3
=
o V

-----------
o a
3
e
0
This also solves a first conjugate holonomic equation.
Scalar

o u
-----------
o a
0
,
o V

-----------
o a
1
,
o V

-----------
o a
2
,
o V

-----------
o a
3
|

(e
0
, e
1
, e
2
, e
3
)
-1
|

=
-e
0
(-e
0
) +
e
1
-------
3
e
1
-
-e
2
----------
3
e
2
-
-e
3
----------
3
e
3
= 0
Power functions can be analyzed in exactly the same way:
Let w = q
2
= ¸a
0
2
e
0
2
+ a
1
2

e
1
2
---------
9
+ a
2
2

e
2
2
---------
9
+ a
3
2

e
3
2
---------
9
,
2 a
0
a
1
e
0

e
1
-------
3
, 2 a
0
a
2
e
0

e
2
-------
3
, 2 a
0
a
3
e
0

e
3
-------
3
¸
u = ¸a
0
2
e
0
2
+ a
1
2

e
1
2
---------
9
+ a
2
2

e
2
2
---------
9
+ a
3
2

e
3
2
---------
9
, 0, 0, 0¸
19

V

= ¡0, 2 a
0
a
1
e
0

e
1
-------
3
, 2 a
0
a
2
e
0

e
2
-------
3
, 2 a
0
a
3
e
0

e
3
-------
3
¡
o u
-----------
o a
0

e
1
-------
3
=
2 a
0
e
0
2
e
1
-------------------------
3
=
o V

-----------
o a
1
e
0
o u
-----------
o a
0

e
2
-------
3
=
2 a
0
e
0
2
e
2
-------------------------
3
=
o V

-----------
o a
2
e
0
o u
-----------
o a
0

e
3
-------
3
=
2 a
0
e
3
2
-------------------
3
=
o V

-----------
o a
3
This time there are cross terms involved.
o u
-----------
o a
1
e
0
=
2 a
1
e
0
e
1
2
-------------------------
9
=
o V

1
-----------
o a
0

e
1
-------
3
o u
-----------
o a
2
e
0
=
2 a
2
e
0
e
2
2
-------------------------
9
=
o V

2
-----------
o a
0

e
2
-------
3
o u
-----------
o a
3
e
0
=
2 a
3
e
0
e
3
2
-------------------------
9
=
o V

3
-----------
o a
0

e
3
-------
3
At first glance, one might think these are incorrect, since the signs of the derivatives are suppose to
be opposite. Actually they are, but it is hidden in an accounting trick :−) For example, the deriva-
tive of u with respect to a1 has a factor of e1^2, which makes it negative. The derivative of the
first component of V with respect to a0 is positive. Keeping all the information about signs in the
e’s makes things look non−standard, but they are not.
Note that these are three scalar equalities. The other Cauchy−Riemann equations evaluate to a
single 3−vector equation. This represents four constraints on the four degrees of freedom found in
quaternions to find out if a function happens to be analytic.
This also solves a holonomic equation.
Scalar

o u
-----------
o a
0
,
o V

-----------
o a
1
,
o V

-----------
o a
2
,
o V

-----------
o a
3
|

(e
0
, e
1
, e
2
, e
3
)
|

=
= 2 a
0
e
0
3
+
2 a
0
e
0
e
1
-----------------------
3
e
1
+
2 a
0
e
0
e
2
-----------------------
3
e
2
+
2 a
0
e
0
e
3
-----------------------
3
e
3
= 0
Since power series can be analytic, this should open the door to all forms of analysis. (I have done
the case for the cube of q, and it too is analytic in q).
4 Other Derivatives
So far, this work has only involved future timelike derivatives. There are five other regions of
spacetime to cover. The simplest next case is for past timelike derivatives. The only change is in
the limit, where the scalar approaches zero from below. This will make many derivatives look time
symmetric, which is the case for most laws of physics.
A more complicated case involves spacelike derivatives. In the spacelike region, changes in time
go to zero faster than the absolute value of the 3−vector. Therefore the order of the limit processes
is reversed. This time the scalar approaches zero, then the 3−vector. This creates a problem,

most quaternions. That will lead to the differential element not cancelling. The way around this is
to take its norm, which is a scalar.
20

A more complicated case involves spacelike derivatives. In the spacelike region, changes in time
go to zero faster than the absolute value of the 3−vector. Therefore the order of the limit processes
is reversed. This time the scalar approaches zero, then the 3−vector. This creates a problem,
because after the first limit process, the differential element is (0, D), which will not commute with
most quaternions. That will lead to the differential element not cancelling. The way around this is
to take its norm, which is a scalar.
A spacelike differential element is defined by taking the ratio of a differential quaternion element
D to its 3−vector, D − D*. Let the norm of D approach zero. To be defined, the three vector must
approach zero faster than its corresponding scalar. To make the definition non−singular every-
where, multiply by the conjugate. In the limit D D*/((D − D*)(D − D*))* approaches (1, 0), a
scalar.
o f (q, q
-
, q
-1
, q
-2
)
--------------------------------------------------
o q

o f (q, q
-
, q
-1
, q
-2
)
-
----------------------------------------------------
o q
=
= limit as (0, D

) - 0 ¡limit as (d, D

) - (0, D

)
¡(f (q + (d, D

), q
-
, q
-1
, q
-2
) - f (q, q
-
, q
-1
, q
-2
))
(d, D

)
-1
(f (q + (d, D

), q
-
, q
-1
, q
-2
) -
f (q, q
-
, q
-1
, q
-2
))
-
(d, D

)
-1 -
¡¡
To make this concrete, consider a simple example, f = q^2. Apply the definition:
Norm ¸
o q
2
-----------
o q
¸ = limit ((0, D

) - 0 (limit as (d, D

) - (0, D

)
¡¡((a, B

) + (d, D

))
2
- (a, B

)
2
¡ (d, D

)
-1

¡((a, B

) + (d, D

))
2
- (a, B

)
2
¡
-
(d, D

)
-1 -
¡¡ =
= lim ¡((a, B

) +
(0, D

) (a, B

) (0, -D

)/norm ((0, D

)) + (0, D

))
((a, B

) + (0, D

) (a, B

) (0, -D

)/norm ((0, D

)) +
(0, D

))
-
¡ =
The second and fifth terms are unitary rotations of the 3−vector B. Since the differential element D
could be pointed anywhere, this is an arbitrary rotation. Define:
(a, B

’) = (0, D

) (a, B

) (0, -D

)/norm ((0, D

))
Substitute, and continue:
= lim ¡((a, B

) + (a, B

’) + (0, D

))
((a, B

) + (a, B

’) + (0, D

))
-
¡ =
21

= lim ¡4 a^2 + 2 B

.B

+ 2 B

.B

’ + 2 D

.B

+ 2 D

.B

’ , 0

¡
= ¡4 a^2 + 2 B

.B

+ 2 B

.B

’, 0

¡ | 2 q |
2
Look at how wonderfully strange this is! The arbitrary rotation of the 3−vector B means that this
derivative is bound by an inequality. If D is in direction of B, then it will be an equality, but D
could also be in the opposite direction, leading to a destruction of a contribution from the 3−vector.
The spacelike derivative can therefore interfere with itself. This is quite a natural thing to do in
quantum mechanics. The spacelike derivative is positive definite, and could be used to define a
Banach space.
Defining the lightlike derivative, where the change in time is equal to the change in space, will
require more study. It may turn out that this derivative is singular everywhere, but it will require
some skill to find a technically viable compromise between the spacelike and timelike derivative to
synthesis the lightlike derivative.
The timelike quaternion derivative on a quaternion manifold is effectively a directional derivative
along the real axis The spacelike derivative is a normed derivative The dual limit definition estab-
lishes the link between these two well−known types of derivatives.
22

Topological Properties of Quaternions
Topological Space
If we choose to work systematically through Wald’s "General Relativity", the starting point is
"Appendix A, Topological Spaces". Roughly, topology is the structure of relationships that do not
change if a space is distorted. Some of the results of topology are required to make calculus
rigorous.
In this section, I will work consistently with the set of quaternions, H^1, or just H for short. The
difference between the real numbers R and H is that H is not a totally ordered set and multiplica-
tion is not commutative. These differences are not important for basic topological properties, so
statements and proofs involving H are often identical to those for R.
First an open ball of quaternions needs to be defined to set the stage for an open set. Define an
open ball in H of radius (r, 0) centered around a point (y, Y) [note: small letters are scalars, capital
letters are 3−vectors] consisting of points (x, X) such that
_
((x - y, X - Y)
-
(x - y, X - Y)) · (r, 0)
An open set in H is any set which can be expressed as a union of open balls.
[p. 423 translated] A quaternion topological space (H,T) consists of the set H together with a collec-
tion T of subsets of H with these properties:
1. The union of an arbitrary collection of subsets, each in T, is in T
2. The intersection of a finite number of subsets of T is in T
3. The entire set H and the empty set are in T
T is the topology on H. The subsets of H in T are open sets. Quaternions form a topology because
they are what mathematicians call a metric space, since q* q evaluates to a real positive number or
equals zero only if q is zero. Note: this is not the meaning of metric used by physicists. For exam-
ple, the Minkowski metric can be negative or zero even if a point is not zero. To keep the same
word with two meanings distinct, I will refer to one as the topological metric, the other as an inter-
val metric. These descriptive labels are not used in general since context usually determines
which one is in play.
An important component to standard approaches to general relativity is product spaces. This is
how a topology for R^n is created. Events in spacetime require R^4, one place for time, three for
space. Mathematicians get to make choices: what would change if work was done in R^2, R^3, or
R^5? The precision of this notion, together with the freedom to make choices, makes exploring
these decisions fun (for those few who can understand what is going on :−)
By working with H, product spaces are unnecessary. Events in spacetime can be members of an
open set in H. Time is the scalar, space the 3−vector. There is no choice to be made.
23

Open Sets
The edges of sets will be examined by defining boundaries, open and closed sets, and the interior
and closure of a set.
I am a practical guy who likes pragmatic definitions. Let the real numbers L and U represent
arbitrary lower and upper bounds respectively such that L < U. For the quaternion topological
space (H, T), consider an arbitrary induced topology (A, t) where x and a are elements of A. Use
inequalities to define:
an open set : (L, 0) · (x - a)
-
(x - a) · (U, 0)
a closed set : (L, 0) (x - a)
-
(x - a) (U, 0)
a half open set : (L, 0) (x - a)
-
(x - a) · (U, 0)
or (L, 0) · (x - a)
-
(x - a) (U, 0)
a boundary : (L, 0) = (x - a)
-
(x - a)
The union of an arbitrary collection of open sets is open.
The intersection of a finite number of open sets is open.
The union of a finite number of closed sets is closed.
The intersection of an arbitrary number of closed sets is closed.
Clearly there are connections between the above definitions
open set union boundary -> closed set
This creates complementary ideas. [Wald, p.424]
The interior of A is the union of all open sets contained within A.
The interior equals A if and only if A is open.
The closure of A is the intersection of all closed sets containing A.
The closure of A equals A if and only if A is closed.
Define a point set as the set where the lower bound equals the upper bound. The only open set that
is a point set is the null set. The closed point set is H. A point set for the real numbers has only
one element which is identical to the boundary. A point set for quaternions has an infinite number
of elements, one of them identical to the boundary.
What are the implications for physics?
With quaternions, the existence an open set of events has nothing to do with the causality of that
collection of events.
an open set : (L, 0) · (x - a)
-
(x - a) · (U, 0)
timelike events : scalar ((x - a)
2
) > (0, 0)
lightlike events : scalar ((x - a)
2
) = (0, 0)
24

spacelike events : scalar ((x - a)
2
) · (0, 0)
A proper time can have exactly the same absolute value as a pure spacelike separation, so these
two will be included in the same sets, whether open, closed or on a boundary.
There is no correlation the reverse way either. Take for example a collection of lightlike events.
Even though they all share exactly the same interval − namely zero − their absolute value can vary
all over the map, not staying within limits.
Although independent, these two ideas can be combined synergistically. Consider an open set S of
timelike intervals.
S = {x, a E H, a fixed; U,
L E R | (L, 0) · (x - a)
-
(x - a) · (U, 0),
and scalar ((x - a)
2
) > 0]
The set S could depict a classical world history since they are causally linked and have good topo-
logical properties. A closed set of lightlike events could be a focus of quantum electrodynamics.
Topology plus causality could be the key for subdividing different regions of physics.
Hausdorff Topology
This property is used to analyze compactness, something vital for rigorously establishing differentia-
tion and integration.
[Wald p424] The quaternion topological space (H, T) is Hausdorff because for each pair of distinct
points a, b E H, a not equal to b, one can find open sets Oa, Ob E T such that a E Oa, b I Ob and
the intersection of Oa and Ob is the null set.
For example, find the half−way point between a and b. Let that be the radius of an open ball
around the points a and b:
let (r, 0) = (a - b)
-
(a - b)/4
Oa = {a, x E H, a is fixed, r E R | (a - x)
-
(a - x) · r]
Ob = {b, x E H, b is fixed, r E R | (b - x)
-
(b - x) · r]
Neither set quite reaches the other, so their intersection is null.
Compact Sets
In this section, I will begin an investigation of compact sets of quaternions. I hope to share some
of my insights into this subtle but significant topic.
First we need the definition of a compact set of quaternions.
[Translation of Wald p. 424] Let A be a subset of the quaternions H. Set A could be opened, closed
or neither. An open cover of A is the union of open sets {Oa} that contains A. A union of open
sets is open and could have an infinite number of members. A subset of {Oa} that still covers A is
called a subcover. If the subcover has a finite number of elements it is called a finite subcover.
The set A subset of H is compact if every open cover of A has a finite subcover.
Let’s find an example of a compact set of quaternions. Consider a set S composed of points with a
finite number of absolute values:
25

S = {x1, x2, ..., xn E H; a1, a2, ...,
an E R, n is finite | (x1-x1)^.5 = (a1, 0),
(x2-x2)^.5 = (a2, 0), ...]
The set S has an infinite number of members, since for any of the equalities, specifying the abso-
lute value still leaves three degrees of freedom (if the domain had been x E R, then S would have
had a finite number of elements). The set S can be covered by an open set {O} which could have
an infinite number of members. There exists a subset {C} of {O} that is finite and still covers S.
The subset {C} would have one member for each absolute value.
C = ¸y E {O], e E R, e > 0 ¡ (a1 - e) ·
_
y
-
y · (a1 + e, 0),
(a2 - e) ·
_
y
-
y · (a2 + e, 0), ...,
one y exists for each inequality¡
Every set of quaternions composed of a finite number of absolute values like the set S is compact.
Notice that the set S is closed because it consists of a boundary without an interior. The link
between compact, closed and bound set is important, and will be examined next
A compact set is a statement about the ability to find a finite number of open sets that cover a set,
given any open cover. A closed set is the interior of a set plus the boundary of that set. A set is
bound if there exists a real number M such that the distance between a point and any member of
the set is less than M.
For quaternions with the standard topology, in order to have a finite number of open sets that cover
the set, the set must necessarily include its boundary and be bound. In other words, to be compact
is to be closed and bound, to be closed and bound is to be compact.
[Wald p. 425] Theorem 1 (Heine−Borel). A closed interval of quaternions S:
S = ¸x E H, a, b E R, a · b ¡ (a, 0)
_
x
-
x (b, 0)¡
with the standard topology on H is compact.
Wald does not provide a proof since it appears in many books on analysis. Invariably the Heine−
Borel Theorem employs the domain of the real numbers, x E R. However, nothing in that proof
changes by using quaternions as the domain.
[Wald p. 425] Theorem 2. Let the topology (H, T) be Hausdorff and let the set A subset of H be
compact. Then A is closed.
Theorem 3. Let the topology (H, T) be compact and let the set A subset of H be closed. Then A is
compact.
Combine these theorems to create a stronger statement on the compactness of subsets of quater-
nions H.
Theorem 4. A subset A of quaternions is compact if and only if it is closed and bounded.
The property of compactness is easily proved to be preserved under continuous maps.
Theorem 5. Let (H, T) and (H’, T’) be topological spaces. Suppose (H, T) is compact and the
function f: H −> H’ is continuous. The f[H] = {h’ E H’ | h’ = f(h)} is compact. This creates a
corollary by theorem 4.
26

Theorem 6. A continuous function from a compact topological space into H is bound and its
absolute value attains a maximum and minimum values.
[end translation of Wald]

1
versus
n
It is important to note that these theorems for quaternions are build directly on top of theorems for
real numbers,
1
. Only the domain needs to be changed to H
1
. Wald continues with theorems on
product spaces, specifically Tychonoff’s Theorem, so that the above theorems can be extended to

n
. In particular, the product space
4
should have the same topology as the quaternions.
Hopefully, subtlety matters in the discussion of foundations.
4
does not come equipped with a
rule for multiplication, so it is qualitatively different from H
1
, even if topologically similar to the
quaternions. A similar issue arises for
2
and the complex number manifold
1
.
27

A Quaternion Algebra Tool Set
Here is a compilation of basic algebra for quaternions. It should look very similar to complex
algebra, since it contains three sets of complex numbers, t + x i, t + y j, and t + z k. To strengthen
the link, and keep things looking simpler, all quaternions have been written as a pair of a scalar t
and a 3−vector V, as in (t, V). All these relations have been tested in a C library and a Java quater-
nion calculator.
Technical note: every tool in this set can be expressed as working with a whole quaternion q. This
is to show how to work with automorphic functions on a quaternion manifold.
Parts
scalar (q) = (q + q-)/2 = (t, 0)
vector (q) = (q - q-)/2 = (0, V)
Simple Algebra
| q | =
_
(qq-) = ¡
_
t
2
+ V.V, 0¡
norm (q) = qq -= (t
2
+ V.V, 0)
det (q) = (qq-)
2
= ¡(t
2
+ V.V)
2
, 0¡
sum (q, q’) = q + q’ = (t + t’, V + V’)
dif (q, q’) = q - q’ = (t - t’, V - V’)
conj (q) = q -= (t, -V)
inv (q) = q- /(qq-) = (t, -V)/(t
2
+ V.V)
adj (q) = q-(qq-) = (t, -V) norm (q)
Multiplication
The Grassman product as defined here uses the same rule Hamilton developed. The Euclidean
product takes the conjugate of the first of the two elements (following a tradition from quantum
mechanics).
Grassman_product (q, q’) =
qq’ = (tt’ - V.V’, tV’ + Vt’ + VxV’)
28

Grassman_even _product (q, q’) =
qq’ + q’ q
-------------------------
2
= (tt’ - V.V’, tV’ + Vt’)
Grassman_odd _product (q, q’) =
qq’ - q’ q
-------------------------
2
= (0, VxV’)
Euclidean_product (q, q’) =
q-q’ = (tt’ + V.V’, tV’ - Vt’ - VxV’)
Euclidean_even _product (q, q’) =
q-q’ + q’ q-
---------------------------------
2
= (tt’ + V.V’, 0)
Euclidean_odd _product (q, q’) =
q-q’ - q’ q-
---------------------------------
2
= (0, tV’ - Vt’ - VxV’)
Trigonometry
sin (q) =
(sin (t) cosh ( | V |), cos (t) sinh ( | V |) V/ | V |)
cos (q) =
(cos (t) cosh ( | V |), -sin (t) sinh ( | V |) V/ | V |)
tan (q) = sin (q)/cos (q)
Note: since the unit vectors of sine and cosine are the same, these two commute so the order is
irrelevant.
asin (q) = -V/ | V | asinh (q V/ | V |)
acos (q) = -V/ | V | acosh (q)
atan (q) = -V/ | V | atanh (q V/ | V |)
sinh (q) =
(sinh (t) cos ( | V |), cosh (t) sin ( | V |) V/ | V |)
cosh (q) =
(cosh (t) cos ( | V |), sinh (t) sin ( | V |) V/ | V |)
tanh (q) = sinh (q)/cosh (q)
asinh (q) = ln (q + (q^2 + 1)^.5)
acosh (q) = ln (q + /-(q^2 - 1)^.5)
atanh (q) = .5 ln ((1 + q)/(1 - q))
29

Powers
exp (q) =
(exp (t) cos ( | V |), exp (t) sin ( | V |) V/ | V |)
q^q’ = exp (ln (q) x q’)
Logs
ln (q) = (0.5 ln (t^2 + V.V), atan2 ( | V |, t) V/ | V |)
log (q) = ln (q)/ln (10)
Quaternion Exponential Multiplication
q q’ = {q, q’] + | [q, q’] | exp (pi[q, q’]/2 | [q, q’] |)
q-q’ = {q-, q’] + | [q-, q’] |
exp (pi[q-, q’]/2 | [q-, q’] |)
Andrew Millard suggested the result for the Grassman product.
30

Classical Mechanics
31

Newton’s Second Law
The form of Newton’s second law for three separate cases will be generated using quaternion
operators acting on position quaternions. In classical mechanics, time and space are decoupled.
One way that can be achieved algebraically is by having a time operator act only on space, or by
space operator only act on a scalar function. I call this the "2 zero" rule: if there are two zeros in
the generator of a law in physics, the law is classical.
Newton’s 2nd Law for an Inertial Reference Frame in Cartesian Coordinates
Define a position quaternion.
R = (t, R

)
Operate on this once with the differential operator to get the velocity quaternion.
V = ¸
d
-------
dt
, 0

¸ (t, R

) = ¡1, R

¡
Operate on the velocity to get the classical inertial acceleration quaternion.
A = ¸
d
-------
dt
, 0

¸ ¡1, R

¡ = ¸0, R

¸
This is the standard form for acceleration in Newton’s second law in an inertial reference frame.
Because the reference frame is inertial, the first term is zero.
Newton’s 2nd Law in Polar Coordinates for a Central Force in a Plane
Repeat this process, but this time start with polar coordinates.
R = (t, r Cos[Θ], r Sin[Θ], 0)
The velocity in a plane.
V = ¸
d
-------
dt
, 0

¸ (t, r Cos[Θ], r Sin[Θ], 0) =
= (1, r

Cos[Θ] - r Sin[Θ] Θ

, r

Sin[Θ] + r Cos[Θ] Θ

, 0)
Acceleration in a plane.
A = ¸
d
-------
dt
, 0

¸
(1, r

Cos[Θ] - r Sin[Θ] Θ

, r

Sin[Θ] + r Cos[Θ] Θ

, 0) =
32

= ¡0, -2 r

Sin[Θ] Θ

- r Cos[Θ] Θ
2
+ r

Cos[Θ] - r Sin[Θ] Θ

,
2 r

Cos[Θ] Θ

- r Sin[Θ] Θ
2
+ r

Sin[Θ] + r Cos[Θ] Θ

, 0¡
Not a pretty sight. For a central force, Θ

= L mr
2
, and Θ

= 0. Make these substitution and rotate
the quaternion to get rid of the theta dependence.
A = (Cos[Θ], 0, 0, -Sin[Θ])
¸
d
-------
dt
, 0

¸
2
(t, r Cos[Θ], R Sin[Θ], 0) =
= ¸0,
L
2
-------------
m
2
r
3
+ r

,
2 L r

------------
m r
2
, 0¸
The second term is the acceleration in the radial direction, the third is acceleration in the theta
direction for a central force in polar coordinates.
Newton’s 2nd Law in a Noninertial, Rotating Frame
Consider a noninertial example, with the frame rotating at an angular speed omega. The differen-
tial time operator is put into the first term of the quaternion, and the three directions for the angular
speed are put in the next terms. This quaternion is then multiplied by the position quaternion to get
the velocity in a rotating reference frame. Unlike the previous examples where the time t did not
interfere with the calculations, here the time t must be set explicitly to zero (I wonder what that
means?).
V = ¸
d
-------
dt
, Ω

¸ (0, R

) = ¡- Ω

.R

, R

+ Ω

x R

¡
Operate on the velocity quaternion with the same operator.
A = ¸
d
-------
dt
, Ω

¸ ¡- Ω

.R

, R

+ Ω

x R

¡ =
= ¸- Ω

.R

, R

+ 2 Ω

x R

+ Ω

x R

- Ω

.R

¸
The first three terms of the 3−vector are the translational, coriolis, and azimuthal alterations respec-
tively. The last term of the 3−vector may not look like the centrifugal force, but using a vector
identity it can be rewritten:
- Ω

.R

= - Ω

x (Ω

x R

) + Ω

2
R

If the angular velocity an the radius are orthogonal, then

x (Ω

x R

) = Ω

2
R

iff Ω

.R

= 0
The scalar term is not zero. What this implies is not yet clear, but it may be related to the fact that
the frame is not inertial.
33

Implications
Three forms of Newton’s second law were generated by choosing appropriate operator quaternions
acting on position quaternions. It is impressive that complicated expressions in Newtonian mechan-
ics can be encapsulated in quaternion one−line formulas. The differential time operator was decou-
pled from any differential space operators. This may be viewed as an operational definition of
"classical" physics.
34

Oscillators and Waves
A professor of mine once said that everything in physics is a simple harmonic oscillator. Therefore
it is necessary to get a handle on everything.
The Simple Harmonic Oscillator (SHO)
The differential equation for a simple harmonic oscillator in one dimension can be express with
quaternion operators.
¸
d
-------
dt
, 0

¸
2
(0, x, 0, 0) + ¸0,
k
----
m
x, 0, 0¸ =
¸0,
d
2
x
-----------
dt
2
+
k x
--------
m
, 0, 0¸ = 0
This equation can be solved directly.
x - C[2] Cos¡
_
k t
-------------
_
m
¡ + C[1] Sin¡
_
k t
-------------
_
m
¡
Find the velocity by taking the derivative with respect to time.
x

-
_
k C[1] Cos¡
_
k t
----------
_
m
¡
----------------------------------------------
_
m
-
_
k C[2] Sin¡
_
k t
----------
_
m
¡
----------------------------------------------
_
m
The Damped Simple Harmonic Oscillator
Generate the differential equation for a damped simple harmonic oscillator as done above.
¸
d
-------
dt
, 0

¸
2
(0, x, 0, 0) +
¸
d
-------
dt
, 0

¸ (0, b x, 0, 0) + ¸0,
k
----
m
x, 0, 0¸ =
= ¸0,
d
2
x
-----------
dt
2
+
b d x
------------
dt
+
k x
--------
m
, 0, 0¸ = 0
Solve the equation.
x - C[1] E
¡-b m-
_
-4 k m+b
2
m
2
¡ t
--------------------------------------------
2 m
+ C[2] E
¡-b m+
_
-4 k m+b
2
m
2
¡ t
--------------------------------------------
2 m
35

The Wave Equation
Consider a wave traveling along the x direction. The equation which governs its motion is given by
¸
d
-----------
v dt
,
d
-------
dx
, 0, 0¸
2
(0, 0, f[t v + x], 0) =
= ¸0, 0, ¸-
d
2
---------
dx
2
+
d
2
----------------
dt
2
v
2
¸ f[t v + x],
2 d
2
f[t v + x]
----------------------------------
dt dx v
¸
The third term is the one dimensional wave equation. The forth term is the instantaneous power
transmitted by the wave.
Implications
Using the appropriate combinations of quaternion operators, the classical simple harmonic oscilla-
tor and wave equation were written out and solved. The functional definition of classical physics
employed here is that the time operator is decoupled from any space operator. There is no reason
why a similar combination of operators cannot be used when time and space operators are not
decoupled. In fact, the four Maxwell equations appear to be one nonhomogeneous quaternion
wave equation, and the structure of the simple harmonic oscillator appears in the Klein−Gordon
equation.
36

Four Tests for a Conservative Force
There are four well−known, equivalent tests to determine if a force is conservative: the curl is
zero, a potential function whose gradient is the force exists, all closed path integrals are zero, and
the path integral between any two points is the same no matter what the path chosen. In this sec-
tion, quaternion operators perform these tests on quaternion−valued forces.
1. The Curl Is Zero
To make the discussion concrete, define a force quaternion F.
F = (0, -k x, -k y, 0)
The curl is the commutator of the differential operator and the force. If this is zero, the force is
conservative.
¸
d
-------
dt
, V

¸, F

= 0
Let the differential operator quaternion act on the force, and test if the vector components equal
zero.
¸
d
-------
dt
, V¸ F = (2 k, 0, 0, 0)
2. There Exists a Potential Function for the Force
Operate on force quaternion using integration. Take the negative of the gradient of the first compo-
nent. If the field quaternion is the same, the force is conservative.
F =
¸
F (dt, dx, dy, dz) =
=
¸
(k x dx + k y dy,
-k x dt + k y dz, -k y dt - k x dz, 0) =
= ¸
k x
2
----------
2
+
k y
2
----------
2
,
-k t x + k y z, -k t y - k x z, 0¸ =
37

¸
d
-------
dt
, V

¸ ¸
k x
2
----------
2
+
k y
2
----------
2
, 0

¸ =
(0, -k x, -k y, 0)
This is the same force as we started with, so the scalar inside the integral is the scalar potential of
this vector field. The vector terms inside the integral arise as constants of integration. They are
zero if t=z=0. What role these vector terms in the potential quaternion may play, if any, is
unknown to me.
3. The Line Integral of Any Closed Loop Is Zero
Use any parameterization in the line integral, making sure it comes back to go.
path = (0, r Cos (t), r Sin (t), 0)
¸
0
2 Π
F dt = 0
4. The Line Integral Along Different Paths Is the Same
Choose any two parameterizations from A to B, and test that they are the same. These paths are
from (0, r, 0, 0) to (0, −r, 2 r, 0).
path1 = ¡0, r Cos (t), 2 r Sin ¡
t
----
2
¡, 0¡
¸
0
2 Π
dt = -2 k r
2
path2 = (0, - t r + r, t r, 0)
¸
0
2
F dt = -2 k r
2
The same!
Implications
The four standard tests for a conservative force can be done with operator quaternions. One new
avenue opened up is for doing path integrals. It would be interesting to attempt four dimensional
path integrals to see where that might lead!
38

Special Relativity
39

Rotations and Dilations Create the Lorentz Group
In 1905, Einstein proposed the principles of special relativity without a deep knowledge of the
mathematical structure behind the work. He had to rely on his old math teacher Minkowski to
learn the theory of transformations (I do not know the details of Einstein’s education, but it could
make an interesting discussion :−) Eventually, Einstein understood general transformations, embod-
ied in the work of Riemann, well enough to formulate general relativity.
A. W. Conway and L. Silberstein proposed a different mathematical structure behind special relativ-
ity in 1911 and 1912 respectively (a copy of Silberstein’s work is available at quaternions.com).
Cayley had observed back in 1854 that rotations in 3D could be achieved using a pair of unit quater-
nions having a norm of one:
q’ = a q b where a
-
a = b
-
b = 1
If this works in 3D space, why not do the 4D transformations of special relativity? It turns out that
the unit quaternions must be complex−valued, or biquaternions. Is this so bad? Let me quote
P.A.M. Dirac (Proc. Royal Irish Academy A, 1945, 50, p. 261):
"Quaternions themselves occupy a unique place in mathematics in that they are the most general
quantities that satisfy the division axiom−−that the product of two factors cannot vanish without
either factor vanishing. Biquaternions do not satisfy this axiom, and do not have any fundamental
property which distinguishes them from other hyper−complex numbers. Also, they have eight
components, which is rather too many for a simple scheme for describing quantities in space−time."
Just for the record: plenty of fine work has been done with biquaternions, and I do not deny the
validity of any of it. Much effort has been directed toward "other hyper−complex numbers", such
as Clifford algebras. I am making a choice to focus on quaternions for reasons outlined by Dirac.
Dirac took a Mobius transformation from complex analysis and tried to develop a quaternion
analog. The approach is too general, and must be restricted to graft the results to the Lorentz
group. I found his approach hard to follow. I needed something simpler :−)
It was quite the wait, but De Leo finally figured out a real quaternion representation of the Lorentz
group (S. De Leo, "Quaternions and special relativity," J. Math. Phys., 37(6):2955−2968, 1996).
He defined an operator he called "bar" which multiplied a quaternion by two quaternions on either
side. He effectively did a commutator of this bar operation, which made for boosts without any
terms from the cross product. It definitely is an approach that works.
Rotation + Dilation
Multiplication of complex numbers can be thought of as a rotation and a dilation. Conway and
Silberstein’s proposals only have the rotation component albeit using a complex number. An
additional dilation term might allow quaternions to do the necessary work.
C. Möller wrote a general form for a Lorentz transformation using vectors ("The Theory of Relativ-
ity", QC6 F521, 1952, eq. 25). For fixed collinear coordinate systems:
40

X
·

= X

+ (Γ - 1) (V

.X

)
V

----------------
| V

|
2
- Γ t V

t
·
= Γ t - Γ (V

.X

)
where c = 1, Γ =
1
----------------------------------
_
1 - (v/c)
2
If V is only in the i direction, then
X
·

= (Γ X

- Γ t V

) i

+ y j

+ z k

t
·
= Γ t - Γ (V

.X

)
The additional complication to the X’ equation handles velocities in different directions than i.
This has a vector equation and a scalar equation. A quaternion equation that would generate these
terms must be devoid of any terms involving cross products. The symmetric product (anti−commu-
tator) lacks the cross product;
even (q, q’) =
q q’ + q’ q
----------------------------
2
= ¡t t
·
- X

.X
·

, t X

+ X

t
·
¡
Möller’s equation looks like it should involve two terms, one of the form AqA (a rotation), the
other Bq (a dilation).
q’ =
q + (Γ - 1)
even ¡even ¡V

-
, q¡, V

¡
--------------------------------------------------------
| V

|
2
+ Γ even ¡V

-
, q
-
¡ =
= q + (Γ - 1)
even ((V

.X

, -t V), (0, V

))
-------------------------------------------------------------------
| V

|
2
+
Γ even ((0, -V

), (t, -X

)) =
= (t, X

) + (Γ - 1)

t, (V

.X

)
V

----------------
| V

|
2
|

- Γ ((V

.X

), t V

)
This is the general form of the Lorentz transformation presented by Möller. Real quaternions are
used in a rotation and a dilation to perform the work of the Lorentz group.
Implications
Is this result at all interesting? A straight rewrite of Möller’s equation would have been dull. What
is interesting is the equation which generates the Lorentz transformation. Notice how the Lorentz
transformation depends linearly on q, but the generator depends on q and q*. That may have
interesting interpretations. The generator involves only symmetric products. There has been some
question in the literature about whether special relativity handles rotations correctly. This is proba-
bly one of the more confusing topics in physics, so I will just let the observation stand by itself.
41

Two ways exist to use quaternions to do Lorentz transformations (to be discussed in the next sec-
tion). The other technique relies on the property of a division algebra. There exists a quaternion L
such that:
q’ = L q such that
scalar (q’, q’) = scalar (q, q) = t
2
- X

.X

For a boost along the i direction,
L =
q’
---------
q
=
((Γ t - Γ v x, -Γ v t + Γ x, y, z) (t, -x, -y, -z))
----------------------------------------------------------------------------------------------------------------------------
(t
2
+ x
2
+ y
2
+ z
2
)
=
= (Γ t
2
- 2 Γ t v x + Γ x
2
+ y
2
+ z
2
, Γ v (-t
2
+ x
2
),
t y - x z - Γ t (y + v z) + Γ x (v y + z),
t z + xy + Γ t (v y - z) + Γ x (-y + v z))/
(t
2
+ x
2
+ y
2
+ z
2
)
if x = y = z = 0, then L = (Γ, -Γ v, 0, 0)
if t = y = z = 0, then L = (Γ, Γ v, 0, 0)
The quaternion L depends on the velocity and can depend on location in spacetime (85% of the
type of problems assigned undergraduates in special relativity use an L that does not depend on
location in spacetime). Some people view that as a bug, but I see it as a modern feature found in
the standard model and general relativity as the demand that all symmetry is local. The existence
of two approaches may be of interest in itself.
42

An Alternative Algebra for Lorentz Boosts: Local Transformations
Many problems in physics are expressed efficiently as differential equations whose solutions are
dictated by calculus. The foundations of calculus were shown in turn to rely on the properties of
fields (the mathematical variety, not the ones in physics). According to the theorem of Frobenius,
there are only three finite dimensional fields: the real numbers (1D), the complex numbers (2D),
and the quaternions (4D). Special relativity stresses the importance of 4−dimensional Minkowski
spaces: spacetime, energy−momentum, and the electromagnetic potential. In this section, events in
spacetime will be treated as the 4−dimensional field of quaternions. It will be shown that problems
involving boosts along an axis of a reference frame can be solved using local quaternion transforma-
tions as apposed to the global transformations of the Lorentz group.
The Tools of Special Relativity
Events are represented as 4−vectors, which can be add or subtracted, or multiplied by a scalar. To
form an inner product between two vectors requires the Minkowski metric, which can be repre-
sented by the following matrix (where c = 1).
g
µΝ
=

1 0 0 0
0 -1 0 0
0 0 -1 0
0 0 0 -1
|

{t, x, y, z].g
µΝ
. {t, x, y, z] = t
2
- x
2
- y
2
- z
2
The Lorentz group is defined as the set of matrices that preserves the inner product of two 4−vec-
tors. A member of this group is for boosts along the x axis, which can be easily defined.
Γ =
1
---------------------
_
1 - Β
2
A
x
=

Γ[Β] -Β Γ[Β] 0 0
-Β Γ[Β] Γ[Β] 0 0
0 0 1 0
0 0 0 1
|

Τhe boosted 4−vector is
A
x
.{t, x, y, z] =
¸
t
-------------------
_
1 - Β
2
-
x Β
-------------------
_
1 - Β
2
,
x
-------------------
_
1 - Β
2
-
t Β
-------------------
_
1 - Β
2
, y, z¡
To demonstrate that the interval has been preserved, calculate the inner product.
43

A
x
.{t, x, y, z]. g
µ
Ν
.A
x
.{t, x, y, z] = t
2
- x
2
- y
2
- z
2
Starting from a 4−vector, this is the only way to boost a reference frame along the x axis to another
4−vector and preserve the inner product. The transformation is classified as global because it
depends only on the velocity and not on t, x, y, or z.
Using Quaternions in Special Relativity
Events will be treated as quaternions, a skew field or division algebra that is 4 dimensional. Any
tool built to manipulate quaternions will also be a quaternion. In this way, although events play a
different role from operators, they are made of identical mathematical fabric.
A squared quaternion is:
(t, X

)
2
= (t
2
- X

.X

, 2 t X

)
The first term of squaring a quaternion is the invariant interval squared. There is implicitly, a form
of the Minkowski metric that is part of the rules of quaternion multiplication. The vector portion is
frame−dependent. If a set of quaternions can be found that do not alter the interval, then that set
would serve the same role as the Lorentz group, acting on quaternions, not on 4−vectors. If two
4−vectors x and x’ are known to have the property that their intervals are identical, then the first
term of squaring q[x] and q[x’] will be identical. Because quaternions are a division ring, there
must exist a quaternion L such that L q[x] = q[x’] since L = q[x’] q[x]^−1. The inverse of a quater-
nion is its transpose over the square of the norm (which is the first term of transpose of a quater-
nion times itself). Apply this approach to determine L for 4−vectors boosted along the x axis.
L ÷ (Γ t - Β Γ x, - Β Γ t + Γ x, y, z) (t, x, y, z)
-1
=
= (Γ t
2
+ Γ x
2
- 2 Γ Βt x + (y
2
+ z
2
), Γ Β (-t
2
+ x
2
),
t (Β Γ z + y (1 - Γ)) - x (Γ Β y + z (1 - Γ)),
t (Γ Β y + z (1 - Γ)) + x (Γ Β z + y (1 - Γ)))
/(t
2
+ x
2
+ y
2
+ z
2
)
Define the Lorentz boost quaternion L along x using this equations. L depends on the relative
velocity and position, making it a local, not global, transformation. See if L q[x] = q[x’].
L[t, x, y, z, Β] (t, x, y, z) =
(Γ t - Γ Β x , - Γ Β t + Γ x, y, z)
This is a quaternion composed of the boosted 4−vector. At this point, it can be said that _any_
problem that can be solved using 4−vectors, the Minkowski metric and a Lorentz boost along the x
axis can also be solved using the above quaternion for boosting the event quaternion. This is
because both techniques transform the same set of 4 numbers to the same new set of 4 numbers
using the same variable beta.
Confirm the interval is unchanged.
(L (t, x, y, z))
2
=
44

=

t
2
- x
2
- y
2
- z
2
,
2 (t
2
Β + x
2
Β - t x (1 + Β
2
))
--------------------------------------------------------------
-1 + Β
2
,
2 y (t - x Β)
----------------------------
_
1 - Β
2
,
2 z (t - x Β)
----------------------------
_
1 - Β
2
|

The first term is conserved as expected. The vector portion of the square is frame dependent.
Using Quaternions in Practice
The boost quaternion L is too complex for simple calculations. Mathematica does the grunge
work. A great many problems in special relativity do not involve angular momentum, which in
effect sets y = z = 0. Further, it is often the case that t = 0, or x = 0, or for Doppler shift problems,
x = t. In these cases, the boost quaternion L becomes a very simple.
If t = 0, then
L = Γ (1, Β, 0, 0)
q -> q’ = Lq
(0, x, 0, 0) -> (t
·
, x’, 0, 0) = (-Γ Β x, Γ x, 0, 0)
If x = 0, then
L = Γ (1, -Β, 0, 0)
q -> q’ = Lq
¡t, 0

¡ -> (t
·
, x’, 0, 0) = (Γ t, -Γ Β t, 0, 0)
If t = x, then
L = Γ (1 - Β, 0, 0, 0)
q -> q’ = Lq
(t, x, 0, 0) -> (t
·
, x’, 0, 0) = Γ (1 - Β) ( t, x, 0, 0)
Note: this is for blueshifts. Redshifts have a plus instead of the minus.
Over 50 problems in a sophomore−level relativistic mechanics class at MIT (8.033) have been
solved using local quaternion transformations. 90% required one of these simple forms for the
boost quaternion.
45

Implications
Problems in special relativity can be solved either using 4−vectors, the Minkowski metric and the
Lorentz group, or using quaternions. No experimental difference between the two methods has
been presented. At this point the difference is in the mathematical foundations.
An immense amount of work has gone into the study of metrics, particular in the field of general
relativity. A large effort has gone into group theory and its applications to particle physics. Yet
attempts to unite these two areas of study have failed.
There is no division between events, metrics and operators when solving problems using quater-
nions. One must be judicious in choosing quaternions that will be relevant to a particular problem
in physics and therein lies the skill. Yet this creates hope that by using quaternions, the long divi-
sion between metrics (the Grassman inner product) and groups of transformations (sets of quater-
nions that preserve the Grassman inner product) may be bridged.
46

Electromagnetism
47

Classical Electrodynamics
Maxwell speculated that someday quaternions would be useful in the analysis of electromagnetism.
Hopefully after a 130 year wait, in this section we can begin that process. The Maxwell equations
have been written with complex−valued quaternions back in Maxwell’s time. Peter Jack was the
first person to write the Maxwell equations using only real−valued quaternions. My own efforts
arose a year later, independently. The approach relies on a judicious use of commutators and
anticommutators.
The Maxwell Equations
The Maxwell equations are formed from a combinations of commutators and anticommutators of
the differential operator and the electric and magnetic fields E and B respectively (for isolated
charges in a vacuum.
even ¸¸
o
--------
ot
,

¸, (0, B

)¸ + odd ¸¸
o
--------
ot
,

¸, (0, E

)¸ =

-

B

,

X E

+
oB

--------
ot
|

= ¡0, 0

¡
odd ¸¸
o
--------
ot
,

¸, (0, B

)¸ - even ¸¸
o
--------
ot
,

¸, (0, E

)¸ =

E

,

X B

-
oE

--------
ot
|

= 4 Π (Ρ, J

)
where even (A, B) =
AB + BA
--------------------
2
, odd (A, B) =
AB - BA
--------------------
2
The first quaternion equation embodies the homogeneous Maxwell equations. The scalar term says
that there are no magnetic monopoles. The vector term is Faraday’s law. The second quaternion
equation is the source term. The scalar equation is Gauss’ law. The vector term is Ampere’s law,
with Maxwell’s correction.
The 4−Potential A
The electric and magnetic fields are often viewed as arising from the same 4−potential A. These
can also be expressed using quaternions.
E = vector ¸even ¸¸
o
--------
ot
, -

¸, (Φ, -A

)¸¸ =

0, -
oA

--------
ot
-

Φ
|

B = odd ¸¸
o
--------
ot
, -

¸, (Φ, -A

)¸ = (0,

x A

)
48

The electric field E is the vector part of the anticommutator of the conjugates of the differential
operator and the 4−potential. The magnetic field B involves the commutator.
These forms can be directly placed into the Maxwell equations.
even ¸¸
o
--------
ot
,

¸, odd ¸¸
o
--------
ot
, -

¸, (Φ, -A

)¸¸ +
odd ¸¸
o
--------
ot
,

¸, vector ¸even ¸¸
o
--------
ot
, -

¸, (Φ, -A

)¸¸¸ =
=

-

x A

,
o

x A

---------------
ot
-

X
o A

---------
ot
-

x

Φ
|

=

-

B

,
oB

--------
ot
+

x E
|

= ¡0, 0

¡
odd ¸¸
o
--------
ot
,

¸, odd ¸¸
o
--------
ot
, -

¸, (Φ, A

)¸¸ -
even ¸¸
o
--------
ot
,

¸, vector ¸even ¸¸
o
--------
ot
, -

¸, (Φ, -A

)¸¸¸ =
=

-

Φ -

oA

--------
ot
,

X

X A

+
o
2
A

----------
ot
2
+
o

Φ
-----------
ot
|

=

E

,

X B

-
oE

--------
ot
|

= 4 Π (Ρ, J

)
The homogeneous terms are formed from the sum of both orders of the commutator and anticommu-
tator. The source terms arise from the difference of two commutators and two anticommutators. It
is almost as if the even/odd operators destructively interfere to generate the homogeneous equa-
tions, while the even/even and odd/odd operators constructively interfere to describe a source.
The Lorentz Force
The Lorentz force is generated similarly to the source term of the Maxwell equations, but there a
small game required to get the signs correct for the 4−force.
odd ¡¡Γ, Γ Β

¡, (0, B

)¡ - even ¡¡-Γ, Γ Β

¡, (0, E

)¡ =
¡Γ Β

E

, Γ E

+ Γ Β

X B

¡
This is the covariant form of the Lorentz force.
Conservation Laws
The continuity equation − conservation of charge − is formed by applying the conjugate of the
differential operator to the source terms of the Maxwell equations.
49

scalar

¸
o
--------
ot
, -

¸

E

,

X B

-
oE

--------
ot
|

|

=
o
--------
ot

E

-

oE

--------
ot
+

X B

=
= scalar ¸¸
o
--------
ot
, -

¸, 4 Π (Ρ, J

)¸ = 4 Π ¸V

J

+

--------
ot
¸
The dot product of the E field and the current density plus the rate of change of the charge density
must equal zero. That means that charge is conserved.
Poynting’s theorem for energy conservation is formed in a very similar way, except that the conju-
gate of electric field is used instead of the conjugate of the differential operator.
scalar

(0, -E

)

E

,

X B

-
oE

--------
ot
|

|

= E

X B

- E

oE

--------
ot
= scalar ((0, -E

), 4 Π (Ρ, J

)) = 4 Π E

J

Additional vector identities are required before the final form is reached.
E

(

X B

) = B

(

X E

) +

(B

X E

)

X E

= -
oB

--------
ot
E

.
oE

--------
ot
=
1
----
2

oE

--------
ot
|

2
B

oB

--------
ot
=
1
----
2

oB

--------
ot
|

2
Use these equations to simplify to the following.
4 Π (E

J

, 0) =

-

(E

X B

) -
1
----
2

oE

--------
ot
|

2
-
1
----
2

oB

--------
ot
|

2
, 0
|

This is Poynting’s equation.
Implications
The foundations of classical electrodynamics are the Maxwell equations, the Lorentz force, and the
conservation laws. In this section, these basic elements have been written as quaternion equations,
exploiting the actions of commutators and anticommutators. There is an interesting link between
the E field and a differential operator for generating conservation laws. More importantly, the
means to generate these equations using quaternion operators has been displayed. This approach
looks independent from the usual method which relies on an antisymmetric 2−rank field tensor and
a U(1) connection.
50

Electromagnetic Field Gauges
A gauge is a measure of distance. Gauges are often chosen to make solving a particu-
lar problem easier. A few are well known: the Coulomb gauge for classical electro-
magnetism, the Lorenz gauge which makes electromagnetism look like a simple
harmonic oscillator, and the gauge invariant form which is used in the Maxwell
equations. In all these cases, the E and B field is the same, only the way it is mea-
sured is different. In this section, these are all generated using a differential quater-
nion operator and a quaternion electromagnetic potential.
The Field Tensor F in Different Gauges
The anti−symmetric 2−rank electromagnetic field tensor F has 3 properties: its trace
is zero, it is antisymmetric, and it contains all the components of the E and B fields.
The field used in deriving the Maxwell equations had the same information written
as a quaternion:
¸
o
--------
ot
, -

¸ (Φ, -A

) - (Φ, A

) ¸
o
--------
ot
,

¸ =

0, -
oA

--------
ot
-

Φ +

X A
|

What makes this form gauge−invariant, so no matter what the choice of gauge (involving dphi/dt
and Del.A), the resulting equation is identical? It is the work of the zero! Whatever the scalar field
is in the first term of the generator gets subtracted away in the second term.
Generating the field tensor F in the Lorenz gauge starting from the gauge−invariant from involves
swapping the fields in the following way:
¸
o
--------
ot
, -

¸

(Φ, A

) + (Φ, -A

)
---------------------------------------
2
|

-

(Φ, A

) - (Φ, -A

)
---------------------------------------
2
|

¸
o
--------
ot
,

¸ =
=

--------
ot
+

.A

, -
oA

--------
ot
-

Φ +

X A
|

The first term of the generator involves the scalar field only, (phi, 0), and the second term involves
the 3−vector field only, (0, A).
The field tensor F in the Coulomb gauge is generated by subtracting away the divergence of A,
which explains why the second and third terms involve only A, even though Del.A is zero :−)
51

¸
o
--------
ot
, -

¸ (Φ, -A

) + ¸
o
--------
ot
, -

¸

(Φ, A

) - (Φ, -A

)
---------------------------------------
4
|

+

(Φ, -A

) - (Φ, A

)
---------------------------------------
4
|

¸
o
--------
ot
,

¸ =
=

--------
ot
, -
oA

--------
ot
-

Φ +

X A
|

The field tensor F in the temporal gauge is quite similar to the Coulomb gauge, but some of the
signs have changed to target the dphi/dt term.
¸
o
--------
ot
, -

¸ (Φ, -A

) - ¸
o
--------
ot
, -

¸

(Φ, A

) + (Φ, -A

)
---------------------------------------
4
|

-

(Φ, -A

) + (Φ, A

)
---------------------------------------
4
|

¸
o
--------
ot
,

¸ =
=

-

.A

, -
oA

--------
ot
-

Φ +

X A
|

What is the simplest expression that all of these generator share? I call it the field tensor F in the
light gauge:
¸
o
--------
ot
, -

¸ (Φ, -A

) =

--------
ot
-

.A

, -
oA

--------
ot
-

Φ +

X A
|

The light gauge is one sign different from the Lorenz gauge, but its generator is a simple as it gets.
Implications
In the quaternion representation, the gauge is a scalar generated in such a way as to not alter the
3−vector. In a lists of gauges in graduate−level quantum field theory written by Kaku, the light
gauge did not make the list of the top 6 gauges. There is a reason for this. Gauges are presented as
a choice for a physicist to make. The most interesting gauges have to do with a long−running
popularity contest. The relationship between gauges is guessed, not written explicitly as was done
here. The term that did not make the cut stands out. Perhaps some of the technical issues in quan-
tum field theory might be tackled in this gauge using quaternions.
52

The Maxwell Equations in the Light Gauge: QED?
What makes a theory non−classical? Use an operational definition: a classical approach neatly
separates the scalar and vector terms of a quaternion. Recall how the electric field was defined
(where {A, B} is the even or symmetric product over 2, and [A, B] is the odd, antisymmetric
product over two or cross product).
E = vector ¸even ¸¸
o
--------
ot
,

¸, (Φ, -A

)¸¸ =

0, -

Φ -
oA

--------
ot
|

B = odd ¸¸
o
--------
ot
,

¸, (Φ, A

)¸ = (0,

X A

)
The scalar information is explicitly discarded from the E field quaternion. In this section, the
scalar field that arises will be examined and shown to be the field which gives rise to gauge
symmetry. The commutators and anticommutators of this scalar and vector field do not alter
the homogeneous terms of the Maxwell equations, but may explain why light is a quantized,
transverse wave.
The E and B Fields, and the Gauge with No Name
In the previous section, the electric field was generated differently from the magnetic field,
since the scalar field was discard. This time that will not be done.
E = even ¸¸
o
--------
ot
,

¸, (Φ, -A

)¸ =

--------
ot
-

A

, -
oA

--------
ot
-

Φ
|

B = odd ¸¸
o
--------
ot
,

¸, (Φ, A

)¸ = (0,

X A

)
What is the name of the scalar field, d phi/dt − Del.A which looks like some sort of gauge? It
is not the Lorenz or Landau gauge which has a plus sign between the two. It is none of the
popular gauges: Coulomb (Del.A = 0), axial (Az = 0), temporal (phi = 0), Feynman, unitary...
[special note: I am now testing the interpretation that this gauge constitutes the gravitational field.
See the section on Einstein’s Vision]
The standard definition of a gauge starts with an arbitrary scalar function psi. The following
substitutions do not effect the resulting equations.
Φ -> Φ’ = Φ -

--------
ot
A

-> A

’ = A

+

Ψ
This can be written as one quaternion transformation.
53

(Φ, A

) - (Φ’, A

’) = (Φ, A

) + ¸-

--------
ot
,

Ψ¸
The goal here is to find an arbitrary scalar and a 3−vector that does the same work as the scalar
function psi. Let
p = -

--------
ot
and Α

=

Ψ
Look at how the gauge symmetry changes by taking its derivative.
¸
o
--------
ot
,

¸ ¸-

--------
ot
,

Ψ¸ =
¸-

Ψ -
o
2
Ψ
----------
ot
2
,

X

Ψ -

--------
ot
+

--------
ot
¸ =
¸
op
--------
ot
-

Α

, 0¸
This is the gauge with no name! Call it the "light gauge". That name was chosen because if
the rate of change in the scalar potential phi is equal to the spatial change of the 3−vector
potential A as should be the case for a photon, the distance is zero.
The Maxwell Equations in the Light Gauge
The homogeneous terms of the Maxwell equations are formed from the sum of both orders of
the commutator and anticommutator.
even ¸¸
o
--------
ot
,

¸, odd ¸¸
o
--------
ot
,

¸, (Φ, A

)¸¸ +
odd ¸¸
o
--------
ot
,

¸, even ¸¸
o
--------
ot
, -

¸, (Φ, -A

)¸¸ =
= (-

X A

, -

X

Φ) = ¡0, 0

¡
The source terms arise from of two commutators and two anticommutators. In the classical
case discussed in the previous section, this involved a difference. Here a sum will be used
because it generates a simpler differential equation.
odd ¸¸
o
--------
ot
,

¸, odd ¸¸
o
--------
ot
,

¸, (Φ, A

)¸¸ -
even ¸¸
o
--------
ot
,

¸, even ¸¸
o
--------
ot
, -

¸, (Φ, -A

)¸¸ =
=

o
2
Φ
----------
ot
2
+

Φ , -
o
2
A

----------
ot
2
+

X (

X A

) -

A
|

=

o
2
Φ
----------
ot
2
+

2
Φ , -
o
2
A

----------
ot
2
-

2
A
|

= 4 Π (Ρ, J

)
54

Notice how the scalar and vector parts have neatly partitioned themselves. This is a wave
equation, except that a sign is flipped. Here is the equation for a longitudinal wave like sound.
o
2
w

----------
ot
2
-

2
w

= 0
The second time derivative of w must be the same as Del^2 w. This has a solution which
depends on sines and cosines (for simplicity, the details of initial and boundary conditions are
skipped, and the infinite sum has been made finite).
w

= _
n=0
·
Cos[n Π t] Sin[n Π R]
o
t
o
t
w

- o
R
o
R
w

= 0
Hit w with two time derivatives, and out comes −n^2 pi^2 w. Take Del^2, and that creates the
same results. Thus every value of n will satisfy the longitudinal wave equation.
Now to find the solution for the sum of the second time derivative and Del^2. One of the signs
must be switched by doing some operation twice. Sounds like a job for i! With quaternions,
the square of a normalized 3−vector equals (−1, 0), and it is i if y = z = 0 . The solution to
Maxwell’s equations in the light gauge is
w

= _
n=0
·
Cos[n Π t] Sin[n Π R V

]
if V

2
= -1, then o
t
o
t
w

+ o
R
o
R
w

= 0
Hit this two time derivatives yields −n^2 pi^2 w. Del^2 w has all of this and the normalized
phase factor V^2 = (−1, 0). V acts like an imaginary phase factor that rotates the spatial compo-
nent. The sum for any n is zero (the details of the solution depend on the initial and boundary
conditions).
Implications
The solution to the Maxwell equations in the light gauge is a superposition of waves − each
with a separate value of n − where the spatial part gets rotated by the 3D analogue of i. That is
a quantized, transverse wave. That’s fortunate, because light is a quantized transverse wave.
The equations were generated by taking the classical Maxwell equations, and making them
simpler.
55

The Lorentz Force
The Lorentz force acts on a moving charge. The covariant form of this law is, where W is work
and P is momentum:

d W
--------
d Τ
,
d P

--------
d Τ
|

= Γ e ¡Β

.E

, E

+ Β

X B

¡
In the classical case for a point charge, beta is zero and the E = k e/r^2, so the
Lorentz force simplifies to Coulomb’s law. Rewrite this in terms of the poten-
tials phi and A.

d W
--------
d Τ
,
d P

--------
d Τ
|

= Γ e

Β.

-
oA

--------
ot
- V

Φ
|

, -
oA

--------
ot
- V

Φ + Β

X (V

X A

)
|

In this section, I will look for a quaternion equation that can generate this covari-
ant form of the Lorentz force in the Lorenz gauge. By using potentials and
operators, it may be possible to create other laws like the Lorentz force, in
particular, one for gravity.
A Quaternion Equation for the Lorentz Force
The Lorentz force is composed of two parts. First, there is the E and B fields.
Generate those just as was done for the Maxwell equations
¸
o
--------
ot
,

¸ (Φ, A

) =

--------
ot
-

A

,
oA

--------
ot
+

Φ +

X A
|

Another component is the 4−velocity
V = ¡Γ, Γ Β

¡
Multiplying these two terms together creates thirteen terms, only 5 of whom
belong to the Lorentz force. That should not be surprising since a bit of algebra
was needed to select only the covariant terms that appear in the Maxwell equa-
tions. After some searching, I found the combination of terms required to
generate the Lorentz force.
¸
o
--------
ot
, -

¸ (Φ, -A

) ¡Γ, -Γ Β

¡ - ¡Γ, -Γ Β

¡ ¸
o
--------
ot
,

¸ (Φ, A

) =
= Γ

Β

-
oA

--------
ot
-

Φ
|

, -
oA

--------
ot
-

Φ + Β

X (

X A

)
|

=
Γ e ¡ Β

.E

, E

+ Β

x B

¡
This combination of differential quaternion operator, quaternion potential and
quaternion 4−velocity generates the covariant form of the Lorentz operator in
the Lorenz gauge, minus a factor of the charge e which operates as a scalar
multiplier.
56

This combination of differential quaternion operator, quaternion potential and
quaternion 4−velocity generates the covariant form of the Lorentz operator in
the Lorenz gauge, minus a factor of the charge e which operates as a scalar
multiplier.
Implications
By writing the covariant form of the Lorentz force as an operator acting on a
potential, it may be possible to create other laws like the Lorentz force. For
point sources in the classical limit, these new laws must have the form of Cou-
lomb’s law, F = k e e’/r^2. An obvious candidate is Newton’s law of gravity, F
= − G m m’/r^2.
57

The Stress Tensor of the Electromagnetic Field
I will outline a way to generate the terms of the symmetric 2−rank stress−momentum tensor of an
electromagnetic field using quaternions. This method may provide some insight into what informa-
tion the stress tensor contains.
Any equation written with 4−vectors can be rewritten with quaternions. A straight translation of
terms could probably be automated with a computer program. What is more interesting is when an
equation is generated by the product of operators acting on quaternion fields. I have found that
generator equations often yield useful insights.
A tensor is a bookkeeping device designed to keep together elements that transform in a similar
way. People can choose alternative bookkeeping systems, so long as the tensor behaves the same
way under transformations. Using the terms as defined in "The classical theory of fields" by Lan-
dau and Life−sized, the antisymmetric 2−rank field tensor F is used to generate the stress tensor T
T
ik
=
1
--------
4 Π
¸- F
iL
F
k
L
+
1
----
4

ik
F
LM
F
LM
¸
I have a practical sense of an E field (the stuff that makes my hair stand on end) and a B field (the
invisible hand directing a compass), but have little sense of the field tensor F, a particular combina-
tion of the other two. Therefore, express the stress tensor T in terms of the E and B fields only:
T
ik
=

W Sx Sy Sz
Sx mxx mxy myz
Sy myx myy myz
Sz mzx mzy mzz
|

W =
1
--------
8 Π
¡E

2
+ B

2
¡
Sa =
1
--------
4 Π
(E

x B

)
mab =
1
--------
4 Π
¡-Ea Eb - Ba Bb + 0.5 ∆
ab
¡E

2
+ B

2
¡¡
Together, the energy density(W), Poynting’s vector (Sa) and the Maxwell stress tensor (m_ab) are
all the components of the stress tensor of the electromagnetic field.
Generating a Symmetric 2−Tensor Using Quaternions
How should one rationally go about to find a generator equation that creates these terms instead of
using the month−long hunt−and−peck technique actually used? Everything is symmetric, so use
the symmetric product:
58

even (q, q’) =
q q’ + q’ q
-------------------------------
2
= ¡t t
·
- X

.X
·

, t X
·

+ X

t
·
¡
The fields E and B are kept separate except for the cross product in the Poynting vector. Individual
directions of a field can be selected by using a unit vector Ua:
even (E

, Ux) = (-Ex, 0) where Ux = (0, 1, 0, 0)
The following double sum generates all the terms of the stress tensor:
T
ik
=
L
a=x
y,z
L
b=x
y,z
1
--------
4 Π

¸
even (Ua, Ub)
----------------------------------
3
- 1¸
((0, E)
2
+ (0, B)
2
)
-----------------------------------------------
2
-
- even (E, Ua) even (E, Ub) - even (B, Ua) even (B, Ub) -
- even (odd (E, B), Ua) - even (odd (E, B), Ub) =
= (-Ex Ey - Ex Ez - Ey Ez - Bx By - Bx Bz - By Bz
+ Ey Bz - Ez By + Ez Bx - Ex Bz + Ex By - Ey Bx, 0)/2
Π
The first line generates the energy density W, and part of the +0.5 delta(a, b)(E^2 + B^2) term of
the Maxwell stress tensor. The rest of that tensor is generated by the second line. The third line
creates the Poynting vector. Using quaternions, the net sum of these terms ends up in the scalar.
Does the generator equation have the correct properties? Switching the order of Ua and Ub leaves
T unchanged, so it is symmetric. Check the trace, when Ua = Ub
trace (T
ik
) =
= L
a=x
y,z
1
--------
4 Π

¸
even (Ua, Ua)
----------------------------------
3
- 1¸
((0, E)
2
+ (0, B)
2
)
-----------------------------------------------
2
-
even (E, Ua)
2
- even (B, Ua)
2
|

= 0
The trace equals zero, as it should.
The generator is composed of three parts that have different dependencies on the unit vectors: those
terms that involve Ua and Ub, those that involve Ua or Ub, and those that involve neither. These
are the Maxwell stress tensor, the Poynting vector and the energy density respectively. Changing
the basis vectors Ua and Ub will effect these three components differently.
Implications
So what does the stress tensor represent? It looks like every combination of the 3−vectors E and B
that avoids quadratics (like Ex^2) and over−counting cross terms. I like what I will call the "net"
stress quaternion:
net (T
ik
) =
59

= (-Ex Ey - Ex Ez - Ey Ez - Bx By - Bx Bz - By Bz
+ Ey Bz - Ez By + Ez Bx - Ex Bz + Ex By - Ey Bx, 0)/2
Π
This has the same properties as an stress tensor. Since the vector is zero, it commutes with any
other quaternion (this may be a reason it is so useful). Switching x terms for y terms would flip the
signs of the terms produced by the Poynting vector as required, but not the others. There are no
terms of the form Ex^2, which is equivalent to the statement that the trace of the tensor is zero.
On a personal note, I never thought I would understand what a symmetric 2−rank tensor was, even
though I listen in on a discussion of the topic. Yes, I could nod along with the algebra, but without
any sense of F, it felt hollow. Now that I have a generator and a net quaternion expression, it looks
quite elegant and straightforward to me.
60

Quantum Mechanics
61

A Complete Inner Product Space with Dirac’s Bracket Notation
A mathematical connection between the bracket notation of quantum mechanics and quaternions is
detailed. It will be argued that quaternions have the properties of a complete inner−product space
(a Banach space for the field of quaternions). A central issue is the definition of the square of the
norm. In quantum mechanics:
|| ; ||
2
= · ; | ; >
In this section, the following assertion will be examined (* is the conjugate, so the vector flips
signs):
|| (t, X

) ||
2
= (t, X

)
-
(t, X

) (t, X

)
-
(t, X

)
The inner−product of two quaternions is defined here as the transpose (or conjugate) of the first
quaternion multiplied by the second. The inner product of a function with itself is the norm.
The Positive Definite Norm of a Quaternion
The square of the norm of a quaternion can only be zero if every element is zero, otherwise it must
have a positive value.
(t, X

)
-
(t, X

) = ¡t
2
+ X

.X

, 0

¡
This is the standard Euclidean norm for a real 4−dimensional vector space.
The Euclidean inner−product of two quaternions can take on any value, as is the case in quantum
mechanics for <phi|theta>. The adjective "Euclidean" is used to distinguish this product from the
Grassman inner−product which plays a central role in special relativity (see alternative algebra for
boosts).
Completeness
With the topology of a Euclidean norm for a real 4−dimensional vector space, quaternions are
complete.
Quaternions are complete in a manner required to form a Banach space if there exists a neighbor-
hood of any quaternion x such that there is a set of quaternions y
|| x - y ||
2
· Ε
4
for some fixed value of epsilon.
Construct such a neighborhood.
62

¡(t, X

) -
Ε
----
4
(t, X

-
¡(t, X

) -
Ε
----
4
(t, X

¡(t, X

) -
Ε
----
4
(t, X

-
¡(t, X

) -
Ε
----
4
(t, X

)¡ =
= ¸
Ε
4
-------
16
, 0, 0, 0¸ · (Ε
4
, 0, 0, 0)
An infinite number of quaternions exist in the neighborhood.
Any polynomial equation with quaternion coefficients has a quaternion solution in x (a proof done
by Eilenberg and Niven in 1944, cited in Birkhoff and Mac Lane’s "A Survey of Modern Algebra.")
Identities and Inequalities
The following identities and inequalities emanate from the properties of a Euclidean norm. They
are worked out for quaternions here in detail to solidify the connection between the machinery of
quantum mechanics and quaternions.
The conjugate of the square of the norm equals the square of the norm of the two terms reversed.
· Φ | ; >
-
= · ; | Φ >
For quaternions,
¡(t, X

)
-
¡t
·
, X
·

¡¡
-
= ¡t t
·
+ X

.X
·

, -t X
·

+ X

t
·
+ X

x X
·

¡
¡t
·
, X
·

¡
-
(t, X

) = ¡t
·
t + X
·

.X

, t
·
X

- X
·

t - X
·

x X

¡
These are identical, because the terms involving the cross produce will flip signs when their order
changes.
For products of squares of norms in quantum mechanics,
· ;Φ | ;Φ > = · ; | ; > · Φ | Φ >
This is also the case for quaternions.
· (t, X

) ¡t
·
, X
·

¡ ¡ (t, X

) ¡t
·
, X
·

¡ > =
= ¡(t, X

) ¡t
·
, X
·

¡¡
-
(t, X

) ¡t
·
, X
·

¡
= ¡t
·
, X
·

¡
-
(t, X

)
-
(t, X

) ¡t
·
, X
·

¡
= ¡t
·
, X
·

¡
-
(t
2
+ x
2
+ y
2
+ z
2
, 0, 0, 0) ¡t
·
, X
·

¡
= (t
2
+ x
2
+ y
2
+ z
2
, 0, 0, 0) ¡t
·
, X
·

¡
-
¡t
·
, X
·

¡
= (t, X

)
-
(t, X

) ¡t
·
, X
·

¡
-
¡t
·
, X
·

¡
63

= · (t, X

) ¡ (t, X

) > · ¡t
·
, X
·

¡ ¡ ¡t
·
, X
·

¡ >
The triangle inequality in quantum mechanics:
· ; + Φ | Φ + ; >
2
( · ; | ; > + · Φ | Φ >)
2
For quaternions,
· (t, X

) + ¡t
·
, X
·

¡ ¡ (t, X

) + ¡t
·
, X
·

¡ >
2
=
= ¡¡t + t
·
, X

+ X
·

¡
-
¡t + t
·
, X

+ X
·

¡¡
2
= ¸t
2
+ t
· 2
+ X

2
+ X
·

2
+ 2 t t
·
+ 2 X

.X
·

, 0¸
2

¸t
2
+ X

2
+ t
· 2
+ X
·

2
+
2
_
(t, X

)
-
(t, X

) (t
·
, X
·

)
-
(t
·
, X
·

), 0¸^2 =
¡ · (t, X

) ¡ (t, X

) > + · ¡t
·
, X
·

¡ ¡ ¡t
·
, X
·

¡ >¡
2
If the signs of each pair of component are the same, the two sides will be equal. If the signs are
different, then the cross terms will cancel on the left hand side of the inequality, making it smaller
than the right hand side where terms never cancel because there are only squared terms.
The Schwarz inequality in quantum mechanics is analogous to dot products and cosines in Euclid-
ean space.
| · ; | Φ > |
2
· ; | ; > · Φ | Φ >
Let a third wave function, chi, be the sum of these two with an arbitrary parameter lambda.
Χ ÷ ; + Λ Φ
The norm of chi will necessarily be greater than zero.
(; + Λ Φ)
-
(; + Λ Φ) = ;
-
; + Λ ;
-
Φ + Λ
-
Φ
-
; + Λ
-
Λ Φ
-
Φ ~ 0
Choose the value for lambda that helps combine all the terms containing lambda.
Λ - -
Φ
-
;
------------
Φ
-
Φ
;
-
; -
Φ
-
; ;
-
Φ
----------------------
Φ
-
Φ
~ 0
Multiply through by the denominator, separate the two resulting terms and do some minor
rearranging.
64

(;
-
Φ)
-
;
-
Φ ;
-
; Φ
-
Φ
This is now the Schwarz inequality.
Another inequality:
2 Re · ; | Φ > · ; | ; > + · Φ | Φ >
Examine the square of the norm of the difference between two quaternions which is necessarily
equal to or greater than zero.
0 · (t, X

) - ¡t
·
, X
·

¡ ¡ (t, X

) - ¡t
·
, X
·

¡ >
= ¡(t - t
·
)
2
+ ¡X

- X
·

¡.¡X

- X
·

¡, 0

¡
The cross terms can be put on the other side of inequality, changing the sign, and leaving the sum
of two norms behind.
¡2 ¡t t
·
+ X

.X
·

¡, 0

¡ ¸t
2
+ X

2
+ t
· 2
+ X
·

2
, 0

¸
2 Re · (t, X

) ¡ ¡t
·
, X
·

¡ > · (t, X

) ¡
(t, X

) > + · ¡t
·
, X
·

¡ ¡ ¡t
·
, X
·

¡ >
The inequality holds.
The parallelogram law:
· ; + Φ | Φ + ; > + · ; - Φ | Φ - ; > = 2 · ; | ; > + 2 · Φ | Φ >
Test the quaternion norm
· (t, X

) + ¡t
·
, X
·

¡ ¡ (t, X

) + ¡t
·
, X
·

¡ > + · (t, X

) - ¡t
·
, X
·

¡ ¡
(t, X

) - ¡t
·
, X
·

¡ > =
= ¡(t + t
·
)
2
+ ¡X

+ X
·

¡.¡X

+ X
·

¡, 0

¡ +
¡(t - t
·
)
2
+ ¡X

- X
·

¡.¡X

- X
·

¡, 0

¡ =
= 2 ¸t
2
+ X

2
+ t
· 2
+ X
·

2
, 0

¸ =
= 2 · (t, X

) ¡ (t, X

) > + 2 · ¡t
·
, X
·

¡ ¡ ¡t
·
, X
·

¡ >
This is twice the square of the norms of the two separate components.
65

Implications
In the case for special relativity, it was noticed that by simply squaring a quaternion, the resulting
first term was the Lorentz invariant interval. From that solitary observation, the power of a mathe-
matical field was harnessed to solve a wide range of problems in special relativity.
In a similar fashion, it is hoped that because the product of a transpose of a quaternion with a
quaternion has the properties of a complete inner product space, the power of the mathematical
field of quaternions can be used to solve a wide range of problems in quantum mechanics. This is
an important area for further research.
Note: this goal is different from the one Stephen Adler sets out in "Quaternionic Quantum Mechan-
ics and Quantum Fields." He tries to substitute quaternions in the place of complex numbers in the
standard Hilbert space formulation of quantum mechanics. The analytical properties of quater-
nions do not play a critical role. It is the properties of the Hilbert space over the field of quater-
nions that is harnessed to solve problems. It is my opinion that since the product of a transpose of
a quaternion with a quaternion already has the properties of a norm in a Hilbert space, there is no
need to imbed quaternions again within another Hilbert space. I like a close shave with Occam’s
razor.
66

Multiplying Quaternions in Polar Coordinate Form
Any quaternion can be written in polar coordinate form, which involves a scalar magnitude and
angle, and a 3−vector I (which in some cases can be the more familiar i).
q = || q || exp[Θ I

] = q
-
q (Cos[Θ] + I

Sin[Θ])
This representation can be useful due to the properties of the exponential function, cosines and
sines.
The absolute value of a quaternion is the square root of the norm, which is the transpose of a quater-
nion multiplied by itself.
| q | =
_
q
-
q
The angle is the arccosine of the ratio of the first component of a quaternion over the norm.
Θ = ArcCos ¸
q + q
-
-----------------
2 | q |
¸
The vector component is generated by normalizing the pure quaternion (the final three terms) to the
norm of the pure quaternion.
I =
q - q
-
----------------------------
2 | q - q
-
|
I^2 equals −1 just like i^2. Let (0, V) = (q − q*)/2.
I
2
=
(0, V) (0, V)
--------------------------------------------------
| (0, V) || (0, V) |
=
(-V.V, VxV)
--------------------------------
(V
2
, 0)
= -1
It should be possible to do Fourier analysis with quaternions, and to form a Dirac delta function (or
distribution). That is a project for the future. Those tools are necessary for solving problems in
quantum mechanics.
New Method for Multiplying Quaternion Exponentials
Multiplying two exponentials is at the heart of modern analysis, whether one works with
Fourier transforms or Lie groups. Given a Lie algebra of a Lie group in a sufficiently small
area the identity, the product of two exponentials can be defined using the Campbell−Haus-
dorff formula:
Exp[X] Exp[y] = (X + Y) +
1
----
2
[X, Y] (X + Y)
+
1
-------
12
([[X, Y], Y] - [[X, Y], X]) (X + Y) + ...
67

This formula is not easy to use, and is only applicable in a small area around unity. Quater-
nion analysis that relies on this formula would be very limited.
I have developed (perhaps for the first time) a simpler and general way to express the
product of two quaternion exponentials as the sum of two components. The product of two
quaternions splits into a commuting and an anti−commuting part. The rules for multiply-
ing commuting quaternions are identical to those for complex numbers. The anticommut-
ing part needs to be purely imaginary. The Grassman product (q q’) of two quaternion
exponentials and the Euclidean product (q* q’) should both have these properties.
Together these define the needs for the product of two quaternion exponentials.
Let q = Exp[X] q’ = Exp[Y]
q q’ =
{q, q’]
-
+ Abs[q, q’]
-
Exp¡
Π
----
2
[q, q’]
-
-------------------------------
Abs[q, q’]
-
¡
where {q, q’]
-
÷
q q’ + q’
-
q
-
-----------------------------------
2
and [q, q’]
-
÷ q q’ - q’
-
q
-
q
-
q’ = same as above
where {q, q’] =
q
-
q’ + q’
-
q and [q, q’] = q
-
q’ - q’
-
q
I call these operators "conjugators" because they involve taking the conjugate of the two
elements. Andrew Millard made the suggestion for the Grassman product that unifies these
approaches nicely. What is happening here is that both commuting and anticommuting
parts scale themselves appropriately. By using an exponential that has pi/2 multiplied by a
normalized quaternion, this always has a zero scalar, as it must to accurately represent an
anticommuting part.
68

Commutators and the Uncertainty Principle
Commutators and the uncertainty principle are central to quantum mechanics. Using quaternions
in these roles has already been established by others (Horwitz and Biedenharn, Annals of Physics,
157:432, 1984). The first proof of the uncertainty principle I saw relied solely on the properties of
complex numbers, not on physics! In this section, I will repeat that analysis, showing how commu-
tators and an uncertainty principle arise from the properties of quaternions (or their subfield the
complex numbers).
Commutators
Any quaternion can be written in a polar form.
q = (s, V) =
_
q
-
q Exp¡
s
---------------
_
q
-
q
V
---------------
_
V
-
V
¡
This is identical to Euler’s formula except that the imaginary unit vector i is replaced by the normal-
ized 3−vector. The two are equivalent if j = k = 0. Any quaternion could be the limit of the sum of
an infinite number of other quaternions expressed in a polar form. I hope to show that such a
quaternion mathematically behaves like the wave function of quantum mechanics, even if the
notation is different.
To simplify things, use a normalized quaternion, so that q* q = 1. Collect the normalized 3−vector
together with I = V/(V* V)^.5.
The angle s/(q* q)^.5 is a real number. Any real number can be viewed as the product of two other
real numbers. This seemingly irrelevant observation lends much of the flexibility seen in quantum
mechanics :−) Here is the rewrite of q.
q = Exp[a b I]
where q
-
q = 1, a b =
s
---------------
_
q
-
q
, I =
V
---------------
_
V
-
V
The unit vector "I" could also be viewed as the product of two quaternions. For classical quantum
mechanics, this additional complication is unnecessary because all "i"’s commute.
A point of clarification on notation: the same letter will be used 4 distinct ways. There are opera-
tors, A hat, which act on a quaternion wave function by multiplying by a quaternion, capital A. If
the operator A hat is an observable, then it generates a real number, (a, 0), which commutes with
all quaternions, whatever their form. There is also a variable with respect to a component of a
quaternion, a_i, that can be used to form a differential operator.
Define a linear operator A hat that multiplies q by the quaternion A.
A

q = A q
If the operator A hat is an observable, then the quaternion A is a real number, (a, 0). A real num-
ber will commute with any quaternion. This equation is functionally equivalent to an eigenvalue
equation, with A hat as an eigenvector of q and (a, 0) as the eigenvalue. However, all of the compo-
nents of this equation are quaternions, not separate structures such as an operator belonging to a
group and a vector. This might make a subtle but significant difference for the mathematical
structure of the theory, a point that will not be investigated here.
69

If the operator A hat is an observable, then the quaternion A is a real number, (a, 0). A real num-
ber will commute with any quaternion. This equation is functionally equivalent to an eigenvalue
equation, with A hat as an eigenvector of q and (a, 0) as the eigenvalue. However, all of the compo-
nents of this equation are quaternions, not separate structures such as an operator belonging to a
group and a vector. This might make a subtle but significant difference for the mathematical
structure of the theory, a point that will not be investigated here.
Define a linear operator B hat that multiplies q by the quaternion B. If B hat is an observable, then
this operator can be defined in terms of the scalar variable a.
Let B

= -I
d
--------
d a
B

q = -I
d Exp[a b I]
-----------------------------
d a
= b q
Operators A and B are linear.
(A

+ B

) q = A

q + B

q = a q + b q = (a + b) q
A

(q + q’) = A

q + A

q’ = a q + a’ q’
Calculate the commutator [A, B], which involves the scalar a and the derivative with respect to a.
[A

, B

] q = (A

B

- B

A

) q = -a I
d q
---------
d a
+ I
d a q
------------
d a
= -a I
d q
---------
d a
+ a I
d q
---------
d a
+ I q
d a
---------
d a
= I q
The commutator acting on a quaternion is equivalent to multiplying that quaternion by the normal-
ized 3−vector I.
The Uncertainty Principle
Use these operators to construct things that behave like averages (expectation values) and standard
deviations.
The scalar a−−generated by the observable operator A hat acting on the normalized q−−can be
calculated using the Euclidean product.
q
-
(A q) = q
-
a q = a q
-
q = a
It is hard to shuffle quaternions or their operators around. Real scalars commute with any quater-
nion and are their own conjugates. Operators that generate such scalars can move around. Look at
ways to express the expectation value of A.
q
-
(A

q) = q
-
a q = a q
-
q = a
-
q
-
q = (A

q)
-
q = a
Define a new operator A’ based on A whose expectation value is always zero.
Let A’ = A - q
-
(A q)
q
-
(A’ q) = q
-
(A - q
-
(A q)) q) = a - a = 0
Define the square of the operator in a way designed to link up with the standard deviation.
70

Let DA’
2
= q
-
(A
’ 2
q) - (q
-
(A

q))
2
= q
-
(A
’ 2
q)
An identical set of tools can be defined for B.
In the section on bracket notation, the Schwarz inequality for quaternions was shown.
A
’-
B

+ B
’-
A

-----------------------------------
2
] A

] ] B

]
The Schwarz inequality applies to quaternions, not quaternion operators. If the operators A’ and
B’ are surrounded on both sides by q and q*, then they will behave like scalars.
The left−hand side of the Schwarz inequality can be rearranged to form a commutator.
q
-
(A
’-
B

+ B
’-
A

) q =
q
-
A
’-
B

q + q
-
B
’-
A

q = q
-
a
’-
B

q + q
-
(-I)
-

d
--------
d a
A

q =
= q
-
a

B

q - q
-
(-I)
d
--------
d a
A

q =
q
-
(A

B

- B

A

) q = q
-
[A

, B

] q
The right−hand side of the Schwarz inequality can be rearranged to form the square of the standard
deviation operators.
q
-
| A

| | B

| q =
q
-
A
-’
A

B
-’
B

q = q
-
A
’ 2
B
’ 2
q = q
-
DA’
2
DB’
2
q
Plug both of these back into the Schwarz inequality, stripping the primes and the q’s which appear
on both sides along the way.
[A, B]
------------------
2
DA
2
DB
2
This is the uncertainty principle for complementary observable operators.
71

Connections to Standard Notation
This quaternion exercise can be mapped to the standard notation used in physics
bra : | Ψ > - q
ket : · Ψ | - q
-
operator : A - A
imaginary : i - I
commutator : [A, B] - [A, B]
norm : · Ψ | Ψ > - q
-
q
expectation of A : · Ψ | A Ψ > maps to q
-
A q
Operator A is Hermitian -
(0, A

) is anti - Hermitian,
q
-
((0, A

) q) = ((0, -A

) q)
-
q
The square of the standard
deviation : ∆A
2
= · Ψ | A
2
Ψ > - · Ψ | A Ψ >
2
- DA
2
One subtlety to note is that a quaternion operator is anti−Hermitian only if the scalar is zero. This
is probably the case for classical quantum mechanics, but quantum field theory may require full
quaternion operators. The proof of the uncertainty principle shown here is independent of this
issue. I do not yet understand the consequence of this point.
To get to the position−momentum uncertainty equation, make these specific maps
A - X
B - P = i
d
-------
dx
I = [A, B] - i [X, P]
[A, B]
------------------
2
=
I
----
2
DA
2
DB
2
-
[X, P]
------------------
2
=
i
--------
2
∆X
2
∆P
2
The product of the squares of the standard deviation for position and momentum in the x−direction
has a lower bound equal to half the expectation value of the commutator of those operators. The
proof is in the structure of quaternions.
72

Implications
There are many interpretations of the uncertainty principle. Here, the uncertainty principle is about
quaternions of the form q = Exp[a b I]. With this insight, one can see by inspection that a plane
wave Exp[((Et − P.X)/hbar I], or wave packets that are superpositions of plane waves, will have
four uncertainty relations, one for the scalar Et and another three for the three−part scalar P.X.
This perspective should be easy to generalize.
73

Unifying the Representations of Integral and Half−Integral Spin
I will show how to represent both integral and half−integral spin within the same quaternion alge-
braic field. This involves using quaternion automorphisms. First a sketch of why this might work
will be provided. Second, small rotations in a plane around two axes will be used to show how the
resulting vector points in an opposite way, depending on which involution is used to construct the
infinitesimal rotation. Finally, a general identity will be used to look at what happens under
exchange of two quaternions in a commutator.
Automorphism, Rotations, and Commutators
Quaternions are formed from the direct product of a scalar and a 3−vector. Rotational operators
that act on each of the 3 components of the 3−vector act like integral angular momentum. I will
show that a rotation operator that acts differently on two of the three components of the 3−vector
acts like half−integral spin. What happens with the scalar is irrelevant to this dimensional count-
ing. The same rotation matrix acting on the same quaternion behaves differently depending
directly on what involutions are involved.
Quaternions have 4 degrees of freedom. If we want to represent quaternions with automorphisms,
4 are required: They are the identity automorphism, the conjugate anti−automorphism, the first
conjugate anti−automorphism, and the second conjugate anti−automorphism:
I : q - q
-
: q - q
-
-1
: q - q
-1
-2
: q - q
-2
where
q
-1
÷ (e
1
q e
1
)
-
q
-2
÷ (e
2
q e
2
)
-
e1, e2, e3 are basis vectors
The most important automorphism is the identity. Life is stable around small permutations of the
identity:−) The conjugate flips the signs of the each component in the 3−vector. These two automor-
phisms, the identity and the conjugate, treat the 3−vector as a unit. The first and second conjugate
flip the signs of all terms but the first and second terms, respectively. Therefore these operators act
on only the two of the three components in the 3−vector. By acting on only two of three compo-
nents, a commutator will behave differently. This small difference in behavior inside a commuta-
tor is what creates the ability to represent integral and half−integral spins.
74

Small Rotations
Small rotations about the origin will now be calculated. These will then be expressed in terms of
the four automorphisms discussed above.
I will be following the approach used in J. J. Sakurai’s book "Modern Quantum Mechanics", chap-
ter 3, making modifications necessary to accommodate quaternions. First, consider rotations about
the origin in the z axis. Define:
R
e
3
=0
(Θ) ÷ ¡cos (Θ) e
0
, 0, 0, sin (Θ)
e
3
-------
3
¡
if q = ¡0, a
1
e
1
-------
3
, a
2
e
2
-------
3
, 0¡
R
e
3
=0
(Θ) q = q’ = ¡0, (a
1
cos (Θ) - a
2
sin (Θ)) e
0
e
1
-------
3
,
(a
2
cos (Θ) + a
1
sin (Θ)) e
0
e
2
-------
3
, 0¡
Two technical points. First, Sakurai considered rotations around any point along the z axis. This
analysis is confined to the z axis at the origin, a significant but not unreasonable constraint. Sec-
ond, these rotations are written with generalized coordinates instead of the very familiar and com-
fortable x, y, z. This extra effort will be useful when considering how rotations are effected by
curved spacetime. This machinery is also necessary to do quaternion analysis (please see that
section, it’s great :−)
There are similar rotations around the first and second axes at the origin;
R
e
1
=0
(Θ) = ¡cos (Θ) e
0
, sin (Θ)
e
1
-------
3
, 0, 0¡
R
e
2
=0
(Θ) = ¡cos (Θ) e
0
, 0, sin (Θ)
e
2
-------
3
, 0¡
Consider an infinitesimal rotation for these three rotation operators. To second order in theta,
sin (Θ) = Θ + O (Θ
3
), cos (Θ) = ¸1 -
Θ
2
-------
2
¸ + O (Θ
3
)
R
e
1
=0
(Θ ·· 1) = ¸¸1 -
Θ
2
-------
2
¸ e
0
, Θ
e
1
-------
3
, 0, 0¸ + O (Θ
3
)
R
e
2
=0
(Θ ·· 1) = ¸¸1 -
Θ
2
-------
2
¸ e
0
, 0, Θ
e
2
-------
3
, 0¸ + O (Θ
3
)
R
e
3
=0
(Θ ·· 1) = ¸¸1 -
Θ
2
-------
2
¸ e
0
, 0, 0, Θ
e
3
-------
3
¸ + O (Θ
3
)
Calculate the commutator of the first two infinitesimal rotation operators to second order in theta:
75

[R
e
1
=0
, R
e
2
=0
] =
¸¸1 -
Θ
2
-------
2
¸ e
0
, Θ
e
1
-------
3
, 0, 0¸ ¸¸1 -
Θ
2
-------
2
¸ e
0
, 0, Θ
e
2
-------
3
, 0¸ -
-¸¸1 -
Θ
2
-------
2
¸ e
0
, 0, Θ
e
2
-------
3
, 0¸ ¸¸1 -
Θ
2
-------
2
¸ e
0
, Θ
e
1
-------
3
, 0, 0¸ =
= ¡(1 - Θ
2
) e
0
2
, Θ
e
0
e
1
-------------
3
, Θ
e
0
e
2
-------------
3
, Θ
2

e
1
e
2
-------------
9
¡ -
¡(1 - Θ
2
) e
0
2
, Θ
e
0
e
1
-------------
3
, Θ
e
0
e
2
-------------
3
, - Θ
2

e
1
e
2
-------------
9
¡ =
= 2 ¡0, 0, 0, Θ
2

e
1
e
2
-------------
9
¡ = 2 (R
e
3
=0

2
) - R (0))
To second order, the commutator of infinitesimal rotations of rotations about the first two axes
equals twice one rotation about the third axis given the squared angle minus a zero rotation about
an arbitrary axis (a fancy way to say the identity). Now I want to write this result using anti−auto-
morphic involutions for the small rotation operators.
[R
-
e
1
=0
, R
-
e
2
=0
] =
¸¸1 -
Θ
2
-------
2
¸ e
0
, - Θ
e
1
-------
3
, 0, 0¸ ¸¸1 -
Θ
2
-------
2
¸ e
0
, 0, - Θ
e
2
-------
3
, 0¸ -
-¸¸1 -
Θ
2
-------
2
¸ e
0
, 0, - Θ
e
2
-------
3
, 0¸ ¸¸1 -
Θ
2
-------
2
¸ e
0
, - Θ
e
1
-------
3
, 0, 0¸ =
= ¡(1 - Θ
2
) e
0
2
, - Θ
e
0
e
1
-------------
3
, - Θ
e
0
e
2
-------------
3
, Θ
2

e
1
e
2
-------------
9
¡ -
¡(1 - Θ
2
) e
0
2
, - Θ
e
0
e
1
-------------
3
, - Θ
e
0
e
2
-------------
3
, - Θ
2

e
1
e
2
-------------
9
¡ =
= 2 ¡0, 0, 0, Θ
2

e
1
e
2
-------------
9
¡ = 2 (R
e
3
=0

2
) - R (0))
Nothing has changed. Repeat this exercise one last time for the first conjugate:
[R
-1
e
1
=0
, R
-1
e
2
=0
] = ¸-¸1 -
Θ
2
-------
2
¸ e
0
, Θ
e
1
-------
3
, 0, 0¸
¸-¸1 -
Θ
2
-------
2
¸ e
0
, 0, - Θ
e
2
-------
3
, 0¸ -
-¸-¸1 -
Θ
2
-------
2
¸ e
0
, 0, - Θ
e
2
-------
3
, 0¸
¸-¸1 -
Θ
2
-------
2
¸ e
0
, Θ
e
1
-------
3
, 0, 0¸ =
76

= ¡(1 - Θ
2
) e
0
2
, - Θ
e
0
e
1
-------------
3
, - Θ
e
0
e
2
-------------
3
, Θ
2

e
1
e
2
-------------
9
¡ -
¡(1 - Θ
2
) e
0
2
, - Θ
e
0
e
1
-------------
3
, - Θ
e
0
e
2
-------------
3
, - Θ
2

e
1
e
2
-------------
9
¡ =
= 2 ¡0, 0, 0, Θ
2

e
1
e
2
-------------
9
¡ = - 2 (R
e
3
=0

2
) - R (0))
This points exactly the opposite way, even for an infinitesimal angle!
This is the kernel required to form a unified representation of integral and half integral spin. Imag-
ine adding up a series of these small rotations, say 2 pi of these. No doubt the identity and conju-
gates will bring you back exactly where you started. The first and second conjugates in the commu-
tator will point in the opposite direction. To get back on course will require another 2 pi, because
the minus of a minus will generate a plus.
Automorphic Commutator Identities
This is a very specific example. Is there a general identity behind this work? Here it is:
[q, q’] = [q
-
, q’
-
] = [q
-1
, q’
-1
]
-1
= [q
-2
, q’
-2
]
-2
It is usually a good sign if a proposal gets more subtle by generalization :−) In this case, the nega-
tive sign seen on the z axis for the first conjugate commutator is due to the action of an additional
first conjugate. For the first conjugate, the first term will have the correct sign after a 2 pi journey,
but the scalar, third and forth terms will point the opposite way. A similar, but not identical story
applies for the second conjugate.
With the identity, we can see exactly what happens if q changes places with q’ with a commutator.
Notice, I stopped right at the commutator (not including any additional conjugator). In that case:
[q, q’] = -[q’, q] = [q
-
, q’
-
] = -[q’
-
, q
-
] =
= ¡0, a
2
a
3

e
2
e
3
-------------
9
+ a
3
a
2

e
3
e
2
-------------
9
,
a
3
a
1

e
3
e
1
-------------
9
+ a
1
a
3

e
1
e
3
-------------
9
, a
1
a
2

e
1
e
2
-------------
9
+ a
2
a
1

e
2
e
1
-------------
9
¡
[q
-1
, q’
-1
] = -[q’
-1
, q
-1
] =
= ¡0, a
2
a
3

e
2
e
3
-------------
9
+ a
3
a
2

e
3
e
2
-------------
9
,
- a
3
a
1

e
3
e
1
-------------
9
- a
1
a
3

e
1
e
3
-------------
9
, - a
1
a
2

e
1
e
2
-------------
9
- a
2
a
1

e
2
e
1
-------------
9
¡
[q
-2
, q’
-2
] = -[q’
-2
, q
-2
] =
= ¡0, - a
2
a
3

e
2
e
3
-------------
9
- a
3
a
2

e
3
e
2
-------------
9
,
a
3
a
1

e
3
e
1
-------------
9
+ a
1
a
3

e
1
e
3
-------------
9
, - a
1
a
2

e
1
e
2
-------------
9
- a
2
a
1

e
2
e
1
-------------
9
¡
Under an exchange, the identity and conjugate commutators form a distinct group from the commu-
tators formed with the first and second conjugates. The behavior in a commutator under exchange
of the identity automorphism and the anti−automorphic conjugate are identical. The first and
second conjugates are similar, but not identical.
77

Under an exchange, the identity and conjugate commutators form a distinct group from the commu-
tators formed with the first and second conjugates. The behavior in a commutator under exchange
of the identity automorphism and the anti−automorphic conjugate are identical. The first and
second conjugates are similar, but not identical.
There are also corresponding identities for the anti−commutator:
{q, q’] = {q
-
, q’
-
]
-
= -{q
-1
, q’
-1
]
-1
= -{q
-2
, q’
-2
]
-2
At this point, I don’t know how to use them, but again, the identity and first conjugates appear to
behave differently that the first and second conjugates.
Implications
Three different operators had to be blended together to perform this feat: commutators, conjugates
and rotations. These involve issue of even/oddness, mirrors, and rotations. In a commutator under
exchange of two quaternions, the identity and the conjugate behave in a united way, while the first
and second conjugates form a similar, but not identical set. Because this is a general quaternion
identity of automorphisms, this should be very widely applicable.
78

Deriving A Quaternion Analog to the Schrödinger Equation
The Schrödinger equation gives the kinetic energy plus the potential (a sum also known as the
Hamiltonian H) of the wave function psi, which contains all the dynamical information about a
system. Psi is a scalar function with complex values.
H Ψ = -i

--------
ot
=
-
2
----------
2 m
V
2
Ψ + V (0, X) Ψ
For the time−independent case, energy is written at the operator −i hbar d/dt, and kinetic energy as
the square of the momentum operator, i hbar Del, over 2m. Given the potential V(0, X) and suit-
able boundary conditions, solving this differential equation generates a wave function psi which
contains all the properties of the system.
In this section, the quaternion analog to the Schrödinger equation will be derived from first princi-
ples. What is interesting are the constraint that are required for the quaternion analog. For exam-
ple, there is a factor which might serve to damp runaway terms.
The Quaternion Wave Function
The derivation starts from a curious place :−) Write out classical angular momentum with
quaternions.
(0, L

) = (0, R

x P

) = odd ((0, R

) (0, P

))
What makes this "classical" are the zeroes in the scalars. Make these into complete quaternions by
bringing in time to go along with the space 3−vector R, and E with the 3−vector P.
(t, R

) (E, P

) = (Et - R

.P

, E R

+ P

t + R

x P

)
Define a dimensionless quaternion psi that is this product over h bar.
Ψ ÷
(t, R

) (E, P

)
---------------------------------

= (Et - R

.P

, E R

+ P

t + R

x P

)/
The scalar part of psi is also seen in plane wave solutions of quantum mechanics. The compli-
cated 3−vector is a new animal, but notice it is composed of all the parts seen in the scalar, just
different permutations that evaluate to 3−vectors. One might argue that for completeness, all
combinations of E, t, R and P should be involved in psi, as is the case here.
Any quaternion can be expressed in polar form:
q = ] q ] e
arccos (
s
--------
|q|
)
V

--------
|V

|
Express psi in polar form. To make things simpler, assume that psi is normalized, so |psi| = 1. The
3−vector of psi is quite complicated, so define one symbol to capture it:
79

I ÷
E R

+ P

t + R

x P

-------------------------------------------
| E R

+ P

t + R

x P

|
Now rewrite psi in polar form with these simplifications:
Ψ = e
(Et -R

.P

) I/
This is what I call the quaternion wave function. Unlike previous work with quaternionic quantum
mechanics (see S. Adler’s book "Quaternionic Quantum Mechanics"), I see no need to define a
vector space with right−hand operator multiplication. As was shown in the section on bracket
notation, the Euclidean product of psi (psi* psi) will have all the properties required to form a
Hilbert space. The advantage of keeping both operators and the wave function as quaternions is
that it will make sense to form an interacting field directly using a product such as psi psi’. That
will not be done here. Another advantage is that all the equations will necessarily be invertible.
Changes in the Quaternion Wave Function
We cannot derive the Schrödinger equation per se, since that involves Hermitian operators that
acting on a complex vector space. Instead, the operators here will be anti−Hermitian quaternions
acting on quaternions. Still it will look very similar, down to the last h bar :−) All that needs to be
done is to study how the quaternion wave function psi changes. Make the following assumptions.
1. Energy and Momentum are conserved.
oE
--------
ot
= 0 and
oP

--------
ot
= 0
2. Energy is evenly distributed in space
V

E = 0
3. The system is isolated
V

x P

= 0
4. The position 3−vector X is in the same direction as the momentum 3−vector P
X.P
-----------------------
| X || P |
= 1 which implies
de
I

---------
dt
= 0 and V

xe
I

= 0
The implications of this last assumption are not obvious but can be computed directly by taking the
appropriate derivative. Here is a verbal explanation. If energy and momentum are conserved, they
will not change in time. If the position 3−vector which does change is always in the same direction
as the momentum 3−vector, then I will remain constant in time. Since I is in the direction of X, its
curl will be zero.
This last constraint may initially appear too confining. Contrast this with the typical classical
quantum mechanics. In that case, there is an imaginary factor i which contains no information
about the system. It is a mathematical tool tossed in so that the equation has the correct properties.
With quaternions, I is determined directly from E, t, P and X. It must be richer in information
content. This particular constraint is a reflection of that.
Now take the time derivative of psi.
80

--------
ot
=
E I
--------

Ψ
--------------------------------------
_
1 + ¡
Et -R

.P

--------------

¡
2
The denominator must be at least 1, and can be greater than that. It can serve as a damper, a good
thing to tame runaway terms. Unfortunately, it also makes solving explicitly for energy impossible
unless Et − P.X equals zero. Since the goal is to make a direct connection to the Schrödinger
equation, make one final assumption:
5. Et − R.P = 0
Et - R

.P

= 0
There are several important cases when this will be true. In a vacuum, E and P are zero. If this is
used to study photons, then t = |R| and E = |P|. If this number happens to be constant in time, then
this equation will apply to the wave front.
if
oEt - R

.P

------------------------
ot
= 0, E =
oR

--------
ot
.P

or
oR

--------
ot
=
E
----
P

Now with these 5 assumptions in hand, energy can be defined with an operator.

--------
ot
=
E I
--------

Ψ
-I

--------
ot
= E Ψ or E = -I
o
--------
ot
The equivalence of the energy E and this operator is called the first quantization.
Take the spatial derivative of psi using the under the same assumptions:
V

Ψ = -
P

I
--------

Ψ
--------------------------------------
_
1 + ¡
Et -R

.P

--------------

¡
2
I

V

Ψ = P

Ψ or P

= I V

Square this operator.
P

2
= (mv)
2
= 2 m
mv
2
---------
2
= 2 m KE = -
2
V

2
The Hamiltonian equals the kinetic energy plus the potential energy.
H

Ψ = -I

--------
ot
= -
2
V

2
Ψ + V Ψ
Typographically, this looks very similar to the Schrödinger equation. Capital I is a normalized
3−vector, and a very complicated one at that if you review the assumptions that got us here. Phi is
not a vector, but is a quaternion. This give the equation more, not less, analytical power. With all
of the constraints in place, I expect that this equation will behave exactly like the Schrodinger
equation. As the constraints are removed, this proposal becomes richer. There is a damper to
quench runaway terms. The 3−vector I becomes quite the nightmare to deal with, but it should be
possible, given we are dealing with a topological algebraic field.
81

Typographically, this looks very similar to the Schrödinger equation. Capital I is a normalized
3−vector, and a very complicated one at that if you review the assumptions that got us here. Phi is
not a vector, but is a quaternion. This give the equation more, not less, analytical power. With all
of the constraints in place, I expect that this equation will behave exactly like the Schrodinger
equation. As the constraints are removed, this proposal becomes richer. There is a damper to
quench runaway terms. The 3−vector I becomes quite the nightmare to deal with, but it should be
possible, given we are dealing with a topological algebraic field.
Implications
Any attempt to shift the meaning of an equation as central to modern physics had first be able to
regenerate all of its results. I believe that the quaternion analog to Schrödinger equation under the
listed constraints will do the task. These is an immense amount of work needed to see as the con-
straints are relaxed, whether the quaternion differential equations will behave better. My sense at
this time is that first quaternion analysis as discussed earlier must be made as mathematically solid
as complex analysis. At that point, it will be worth pushing the envelope with this quaternion
equation. If it stands on a foundation as robust as complex analysis, the profound problems seen in
quantum field theory stand a chance of fading away into the background.
82

Introduction to Relativistic Quantum Mechanics
The relativistic quantum mechanic equation for a free particle is the Klein−Gordon equation
(h=c=1)
¸
o
2
----------
ot
2
- V
2
+ m
2
¸ I = 0
The Schrödinger equation results from the non−relativistic limit of this equation. In this section,
the machinery of the Klein−Gordon equation will be ported to quaternions.
The Wave Function
The wave function is the superposition of all possible states of a system. The product of the conju-
gate of a wave function with another wave function forms a complete inner product space. In the
energy/momentum representation, this would involve all possible energy levels and momenta.
I ÷ the sum from n = 0 to infinity of (E
n
, P

n
)
This infinite sum of quaternions should contain all the information about a system. The quaternion
wave function can be normalized.
_
n=0
·
(E
n
, P

n
)
-
(E
n
, P

n
) = _
n=0
·
¡E
n
2
+ P

n
2
, 0¡ = 1
The first quaternion is the conjugate or transpose of the second. Since the transpose of a quater-
nion wave function times a wave function creates a Euclidean norm, this representation of wave
functions as an infinite sum of quaternions can form a complete, normed product space.
The Klein−Gordon Equation
The Klein−Gordon equation can be divided into two operators that act on the wave function: the
D’Alembertian and the scalar m^2. The quaternion operator required to create the D’Alembertian,
along with vector identities, has already been worked out for the Maxwell equations in the Lorenz
gauge.
_
n=0
·

¸
o
--------
ot
,

¸
2
+ ¸
o
--------
ot
, -

¸
2
|

(E
n
, P

n
)/2 =
83

= _
n=0
·

o
2
E
n
------------
ot
2
-

E
n
-

X P

n
,
o
2
P

n
------------
ot
2
-

P

n
+

X

X P

n
+

X

E
n
|

The first term of the scalar, and the second term of the vector, are both equal to zero. What is left
is the D’Alembertian operator acting on the quaternion wave function.
To generate the scalar multiplier m^2, substitute En and Pn for the operators d/dt and del respec-
tively, and repeat. Since the structure of the operator is identical to the previous one, instead of the
D’Alembertian times the wave function, there is En^2−Pn^2. The sum of all these terms becomes
m^2.
Set the sum of these two operators equal to zero to form the Klein−Gordon equation.
_
n=0
·

¸
o
--------
ot
,

¸
2
+ ¸
o
--------
ot
, -

¸
2
+ (E
n
, P

n
)
2
+ (E
n
, -P

n
)
2
|

(E
n
, P

n
)/2 =
= _
n=0
·
¸-

(

X P

n
) -

E
n
-
P

n
(P

n
X P

n
) - (P

n
P

n
) E
n
+ E
n
3
+
o
2
E
n
------------
ot
2
,

X (

X P

n
) +

X (

E
n
) + P

n
X (P

n
X P

n
) + (P

n
X P

n
) E
n
-

(

P

n
) +P

n
E
n
2
- P

n
(P

n
P

n
) +
o
2
P

n
------------
ot
2
|

It takes some skilled staring to assure that this equation contains the Klein−Gordon equation along
with vector identities.
Connection to the Maxwell Equations
If m=0, the quaternion operators of the Klein−Gordon equation simplifies to the operators used to
generate the Maxwell equations in the Lorenz gauge. In the homogeneous case, the same operator
acting on two different quaternions equals the same result. This implies that
(;, A

) = _
n=0
·
(E
n
, P

n
)
Under this interpretation, a nonzero mass changes the wave equation into a simple harmonic oscilla-
tor. The simple relationship between the quaternion potential and the wave function may hold for
the nonhomogeneous case as well.
84

Implications
The Klein−Gordon equation is customarily viewed as a scalar equation (due to the scalar D’Alem-
bertian operator) and the Maxwell equations are a vector equation (due to the potential four vector).
In this section, the quaternion operator that generated the Maxwell equations was used to generate
the Klein−Gordon equation. This also created several vector identities which are usually not men-
tioned in this context. A quaternion differential equation is needed to perform the work of the
Dirac equation, but since quaternion operators are a field, an operator that does the task must exist.
85

Time Reversal Transformations for Intervals
The following operator R for quaternions reverses time difference between two events:
(dt, d X

) -> (-dt, d X

) = R (dt, d X

)
The quaternion R exist because quaternions are a field.
The operator R will equal (−dt, dX )(dt, dX
1
. The inverse of quaternion is the transpose over the
square of the norm, which is the scalar term of the transpose of a quaternion times itself.
R = (-dt, d X

) (dt, d X

)
-1
=
(-dt
2
+ d X

.d X

, 2 dt d X

)/(dt
2
+ d X

.d X

)
For any given event, the time−reversal operator R can be defined based on the above.
A criticism of this operator is that it is local, meaning it depends explicitly on spacetime. The two
most important theories in physics, the standard model and general relativity, are also local, so this
quality is consistent with those theories.
Classical Time Reversal
Examine the form of the quaternion which reverses time under two conditions. In the classical
region, the change in time dt is much greater than space, dX. Calculate R in this limit to one order
of magnitude in the ratio of dX/dt.
R = (-dt, d X

) (dt, d X

)
-1
=
(-dt
2
+ d X

.d X

, 2 dt d X

)/(dt
2
+ d X

.d X

, 0)
if dt >> ¡ dX ¡ then R ¡-1, 2 Β

¡
The operator R is almost the negative identity, but the vector is non−zero, so it would not com-
mute. In the classical limit, time reversal now depends on velocity, not the local position in space-
time, so classical time reversal is a global, not local operator.
Relativistic Time Reversal
For a relativistic interval involving one axis, the interval could be characterized by the following:
(T + Ε, T, 0, 0)
Find out what quaternion is required to reverse time for this relativistic interval to first order in
epsilon.
86

R =

T
2
- (T + Ε)
2
-----------------------------
T
2
+ (T + Ε)
2
,
2 T (T + Ε)
-----------------------------
T
2
+ (T + Ε)
2
, 0, 0
|

=
¡-
Ε
----
T
+ O[Ε]
2
, 1 + O[Ε]
2
, 0, 0¡
This approaches q[−e/T, 1, 0, 0], almost a pure vector, a result distinct from the classical case.
Again, in this limit, the transformation approaches a global transformation.
Implications
In special relativity, the interval between events is considered to be 4 vector are operated on by
elements of the Lorentz group. The element of this group that reverses time has along its diagonal
{−1, 1, 1, 1}, zeroes elsewhere. There is no dependence on relative velocity or any local informa-
tion. Therefore special relativity predicts the operation of time reversal should be indistinguishable
for classical and relativistic intervals. Yet classically, the boundary condition of time reversal
appears to involve entropy. For relativistic interactions, time reversal involves antiparticles.
In this section, a time reversal quaternion operator has been derived and shown to work. Time
reversal for classical and relativistic intervals have distinct limits, but these transformations have
not yet been tied explicitly to laws of physics.
87

Gravity
This is a current area of active research for me. I have two distinct approaches. The first uses
scalars, vectors, tensors, and the Christoffel symbol to characterize a dynamic metric. The work
attempts to unify gravity and EM in a rank 1 field theory using an asymmetric field strength tensor
in the Lagrange density. It has three testable predictions. Light should be bent more at second
order parameterized post−Newtonian accuracy. In a gravity wave, the mode of emission will not be
transverse (that mode of emission is done by light). There is also an new constant velocity solution
for gravity that may eliminate the need for dark matter for the velocity profile of thin disk galaxies.
The approach to gravity that uses quaternions makes the same three experimental predictions. The
difference is that the Christoffel symbol is not used. It will require the efforts of someone with
more technical math skills than I possess to determine if this approach is valid. There are aspects
of the quaternion approach I prefer, but at this time I am promoting the form that uses the Christof-
fel since the odds of it being accepted are higher.
88

Gravity by Analogy to EM
Abstract
A Lagrange density for gravity is proposed based on a strict analogy to the classical Lagrangian for
electromagnetism. The field equations are a four−dimensional wave equation. The classic field
equations contain both the Maxwell equations and Newton’s field equations under certain condi-
tions. The four−dimensional wave equation has been quantized before. The scalar and longitudinal
modes of emission are interpreted as spin 2 gravitons, so they can do the work of gravity. If gravita-
tional waves are detected, this proposal predicts scalar or longitudinal polarization. How the pro-
posal integrates with the standard model Lagrangian is worked out.
A force equation is written based on the same strict analogy to the relativistic Lorentz force of
electromagnetism. For geodesic motion, the cause of the curvature is due entirely to the gravita-
tional and electric potentials. This is a new type of statement about curvature. A specific, normal-
ized, weak−field potential is investigated. Analysis of small perturbations yields changes in the
potential that depend on an inverse distance squared. By breaking spacetime symmetry, Newton’s
law of gravity results. By using the chain rule, a stable, constant−velocity solution is apparent,
which may yield insight to the rotation profile of galaxies and early big bang cosmology, since
both require stable, constant−velocity solutions. If spacetime symmetry is preserved, the second−or-
der differential equations can be solved exactly. Eliminating the constants and rearrange terms
generates an equation that has the form of a metric equation. The Taylor series expansion of the
metric equation is identical to the Schwarzschild metric to parameterized−post−Newtonian accu-
racy. The Taylor series for the two metrics differ for higher order terms and may be tested
experimentally.
Introduction
The goal of this paper is to create one mathematical structure for gravity and electromagnetism that
can be quantized. The difference between gravity and electromagnetism is the oldest core problem
facing physics, going back to the first studies of electromagnetism in the seventeenth century.
Gravity was the first inverse square law, discovered by Isaac Newton. After twenty years of effort,
he was able to show that inside a hollow massive shell, the gravitational field would be zero. Ben
Franklin, in his studies of electricity, demonstrated a similar property using a conducting cup.
Joseph Priestly realized this meant that the electrostatic force was governed by an inverse square
law just like gravity. Coulomb got the credit for the electrostatic force law modeled on Newton’s
law of gravity.
Over a hundred years later, Einstein started from the tensor formalism of electromagnetism on the
road to general relativity. Instead of an antisymmetric field strength tensor, Einstein used a symmet-
ric tensor because the metric tensor is symmetric. There is a precedence for transforming mathemati-
cal structures between gravity and electromagnetism.
The process of transforming mathematical structures from electromagnetism to gravity will be
continued. Specifically, the gravitational analog to the classic electromagnetic Lagrange density
will be written. There are several consequence of this simple procedure. The Lagrangian contains
both terms with a connection and the Fermi Lagrangian of electromagnetism. This makes it reason-
able to suppose the Lagrangian can describe both a dynamic geometry required for gravity and the
Maxwell equations for electrodynamics. The gravitational field equations are analogues to Gauss’
and Ampere’s laws, and contain the Newton’s gravitational field equation. These field equations
are not second rank like those used in general relativity. It must be stressed that the field strength
tensor is a second order symmetric tensor, so this does not conflict with proofs that at least a sym-
metric second rank tensor is required to completely describe spacetime curvature. The Maxwell
equations result if the mass current and gravitational field are zero. The field equations have been
quantized before, but new interpretations will flow from the unification effort. A link to the
Lagrangian of the standard model will be detailed.
89

The process of transforming mathematical structures from electromagnetism to gravity will be
continued. Specifically, the gravitational analog to the classic electromagnetic Lagrange density
will be written. There are several consequence of this simple procedure. The Lagrangian contains
both terms with a connection and the Fermi Lagrangian of electromagnetism. This makes it reason-
able to suppose the Lagrangian can describe both a dynamic geometry required for gravity and the
Maxwell equations for electrodynamics. The gravitational field equations are analogues to Gauss’
and Ampere’s laws, and contain the Newton’s gravitational field equation. These field equations
are not second rank like those used in general relativity. It must be stressed that the field strength
tensor is a second order symmetric tensor, so this does not conflict with proofs that at least a sym-
metric second rank tensor is required to completely describe spacetime curvature. The Maxwell
equations result if the mass current and gravitational field are zero. The field equations have been
quantized before, but new interpretations will flow from the unification effort. A link to the
Lagrangian of the standard model will be detailed.
A weak static gravitational field in a vacuum will be studied using standard modern methods:
normalizing the potential and looking at perturbations. The potential will be plugged into a gravita-
tional force equation analogous to the Lorentz force equation of electromagnetism. The force
equation leads to a geodesic equation where the potential causes the curvature, something which is
missing from general relativity. Newton’s law of gravity is apparent if spacetime symmetry is
broken. A new class of solutions emerges for the gravitational source where velocity is constant,
but the distribution of mass varies with distance. This may provide new ways to look at problems
with the rotation profiles of disk galaxies and big bang cosmology. If spacetime symmetry is pre-
served, solving the force equation and eliminating the constants creates a metric equation similar to
the Schwarzschild metric. The metrics are equivalent to first−order parameterized post−Newtonian
accuracy. Therefore the metric will past all weak field tests. The coefficients are different to sec-
ond−order, so the proposal can be verify or rejected experimentally.
Lagrangians
The classic electromagnetic Lagrangian density has three terms: one for kinetic energy, one for a
moving change, and a third for the antisymmetric second rank field strength tensor:

EM
= -
Ρ
----
Γ
-
1
----
c
J
Μ
A
Μ
-
1
-----------
4 c
2
(o
Μ
A
Ν
- o
Ν
A
Μ
) (o
Μ
A
Ν
- o
Ν
A
Μ
)
An analogous Lagrangian for gravity would also contain these three components, but three changes
are required. First, gravity depends on mass, not electric charge, so where there is an electrical
charge −q, an inertial mass +m will be substituted. The change in sign is required so that like
charges attract for gravity. Mass does not have the same units as electric charge, so mass will have
to be multiplied by the square root of Newton’s gravitational constant G to keep the units identical.
Second, because gravity effects metrics which are symmetric, the source of gravity must also be
symmetric. The minus sign that makes the electromagnetic field strength tensor antisymmetric will
be made positive. Third, in order that symmetric object transforms like a tensor requires a replace-
ment of the exterior derivative with a covariant derivative:

G
= -
Ρ
----
Γ
+
1
----
c
J
m
Μ
A
Μ
-
1
-----------
4 c
2
(V
Μ
A
Ν
+ V
Ν
A
Μ
) (V
Μ
A
Ν
+ V
Ν
A
Μ
)
The total Lagrangian will be a merger of these two which only apply if the other force is not in
effect. The kinetic energy term is the same as either Lagrangian separately. The moving charge
term is a sum. Without loss of generality, the regular derivatives in the electromagnetic Lagrangian
(Eq. L_EM) can be written as covariant derivatives. This leads to the unified Lagrangian for grav-
ity and electromagnetism:

GEM
= -
Ρ
----
Γ
- (J
q
Μ
- J
m
Μ
) A
Μ
-
1
-----------
2 c
2
V
Μ
A
Ν
V
Μ
A
Ν
90

= -
Ρ
----
Γ
-
q -
_
G m
--------------------
c
2
V
(Φ - A

v

) -
1
-----------
2 c
2
o
Μ
A
Ν
o
Μ
A
Ν
-
1
-----------
2 c
2
F
Ν
Ω Μ
(F
ΡΜ
Ν
A
Ρ
A

+ 2 o
Μ
A
Ν
A

)
The kinetic energy term is for one particle experiencing both gravity and electromagnetism. The
Fermi Lagrangian of electromagnetism is a subset. This establishes a link to electromagnetism. The
Christoffel symbols (or connection coefficients) represent derivatives of metrics. Because a
dynamic metric is part of the Lagrangian, this Lagrangian could describe the dynamics of the
metric, which is a central accomplishment of general relativity. The potential to do both gravity
and electromagnetism is here.
In local covariant coordinates, the connect is zero, which leads to a simpler expression of the
Lagrangian:

GEM
= -
m
----
V

_
1 -

oR

--------
ot
|

2
-
q -
_
G m
--------------------
c
2
V
(Φ - A

v

) -
1
-----------
2 c
2

¸

--------
ot
¸
2
- (V

Φ)
2
-

oA

--------
ot
|

2
+ (VA

)
2
|

This is almost identical to working with the classical electromagnetic field equation by choosing
the Lorenz gauge, the difference being the inclusion of a mass term. Because the gauge was not
fixed, there is more freedom for this Lagrangian, which is required if this field equation does more
than just electromagnetism.
Classical Field Equations
The field equations can be found by applying the Euler−Lagrange equations to the Lagrange den-
sity (assuming the connection is zero for simplicity):
.
2
A
Μ
= J
q
Μ
- J
m
Μ
The fields are expressed in terms of the potential. The symmetric and antisymmetric field strength
tensors are very similar, differing only in the sign of A^v;u. The classical fields required to repre-
sent the field strength tensors should also be similar. There is a symmetric analog to the electric E
and B fields: To make a connection to the classical fields of gravity and electromagnetism, use the
following substitutions:
E

= -
oA

--------
ot
+ c V

Φ
e

=
oA

--------
ot
+ c V

Φ
B

= -c ¸
o
--------
oy
-
o
--------
oz
,
o
--------
ox
-
o
--------
oz
,
o
--------
ox
-
o
--------
oy
¸ A

= -c V

xA

b

= c ¸-
o
--------
oy
-
o
--------
oz
, -
o
--------
ox
-
o
--------
oz
, -
o
--------
ox
-
o
--------
oy
¸ A

÷ V

x A

91

The symmetric curl as defined above has all the same differential operators, but all the signs are
negative, so it is easier to remember. The symmetric field strength tensor has four more compo-
nents that lie along the diagonal. Define a field g to represent the diagonal elements:
g = ¸

--------
ot
, c
oA
x
----------
ox
, c
oA
y
----------
oy
, c
oA
z
----------
oz
¸ = V
Μ
A
Μ
The diagonal of the field strength tensor
Μ
A
n
is g. The first row and column of the asymmetric
field strength tensor is the sum of the electric field E and its symmetric analog e. The rest of the
off−diagonal terms are the sum of the magnetic field B and its symmetric analog b. If the trace of
field strength tensor is zero, then the equations are in the Lorentz gauge.
Substitute the classical fields into the field equations, starting with the scalar field equation. Impor-
tant technical note: all the operators used for the divergences and curls are contravariant, that
means they bring in an extra minus sign to these expressions for the field equations of the form

Μ

Μ
A
Ν
J
Ν
:
Ρ
q
- Ρ
m
=
1
----
c

o
2
Φ
----------
ot
2
- c V

2
Φ
=
c
----
2
(V

E

+ V

e

) +
og
0
----------
ot
This equation combines Gauss’ law and analogous equation for gravity. The two equations are
unified, but under certain physical conditions, can be isolated. A relativistic form of the Newtonian
gravitational field equation can be seen with the following constraints:

m
= -c .

2
Φ
iff Ρ
q
= 0, and
oA

--------
ot
= -c V

Φ
This equation should be consistent with special relativity without modification. The classical
Newtonian field equation arises from these physical constraints:
Ρ
m
= c
2
V

2
Φ
iff Ρ
q
= 0,
oA

--------
ot
= -c V

Φ and
og
0
----------
ot
= 0
Every aspect of classical Newtonian gravity can be represented by this proposal under these
constraints.
Gauss’ law appears under the following conditions:
Ρ
q
=
1
----
c

o
2
Φ
----------
ot
2
- c V

2
Φ
iff Ρ
m
= 0, and
oA

--------
ot
= c V

Φ
Repeat the exercise for the vector equation.
92

J

q
- J

m
=
1
----
c

o
2
A
----------
ot
2
- c V
2
A

=
1
----
2

-
oE

--------
ot
+ c V

x B

+
oe

--------
ot
- c V

x b
|

+ 2 c V

g
u
This has Ampere’s law and a symmetric analog for Ampere’s law for gravity.
This proposal for classical gravitational and electromagnetic field equations is expressed with
tensors of rank one (vectors). Einstein’s field equations are second rank. Therefore the two
approaches are fundamentally different. One must remember that although the field equations are
rank one, the field strength tensor is second rank.
With no gravitational field, the Maxwell source equations result. The homogeneous Maxwell
equations are vector identities with these choices of maps to the potentials, and are unaffected by
the proposal.
Canonical Quantization
The classical electromagnetic Lagrangian cannot be quantized. One way to realize this is to con-
sider the generalized 4−momentum:
Π
Μ
= h
_
G
o
EM
--------------------
o(
oA
Μ
--------
c ot
)
= -F
Μ0
Unfortunately, the energy component of the moment operator is zero. The commutator A
0
, Π
0
]
will equal zero, and cannot be quantized. The momentum for the unified Lagrangian of gravity and
electromagnetism does not suffer from this problem:
Π
Μ
= h
_
G
oA
Μ
-----------
c ot
When expressed with operators, the commutator A
0
, Π
0
] will not be zero, so the field can be
quantized. If the connection is zero, L_GEM generates the same field equations as the classical
electromagnetic Lagrangian with the choice of the Lorenz gauge. That field has been quantized
before, first by Gupta and Bleuler (S. N. Gupta, Proc. Phys. Soc. London, 63:681−691, 1950).
They determined that there were four modes of transmission: two transverse, one scalar, and one
transverse mode. The interpretation of these modes appears internally inconsistent to this author.
They discuss "scalar photons", but photons as the quanta of electric and magnetic fields must
transform as a vector, not a scalar. They introduce a supplemental condition solely to make the
scalar and longitudinal modes virtual. Yet there is no need to make a nonsense particle virtual.
The field in this proposal must represent both gravity and electromagnetism. The two transverse
modes are photons that do all the work of electromagnetism. The symmetric second−rank field
strength tensor cannot be represented by a photon because photons transform differently than a
symmetric tensor. Whatever particle does the work must travel at the speed of light like the trans-
verse modes of transmissions of the field. These constraints dictate that the scalar and transverse
modes of transmission for this proposal are gravitons.
There are efforts underway to detect the transverse gravitational waves predicted by general relativ-
ity. This proposal predicts the polarity of a gravitational wave will be either scalar or longitudinal,
not transverse, because those are the modes of transmission. The detection of the first gravitational
wave polarization will mark either success or failure of this unified field theory.
93

Integration with the Standard Model
The standard model does not in an obvious way deal with curved spacetime. A more explicit con-
nection will be attempted by condensing the unitary aspects of the symmetries U(1), SU(2), and
SU(3) with the 4−vectors and a curved metric. Start with the standard model Lagrangian:

SM
= I

Γ
Μ
D
Μ
I
where
D
Μ
= o
Μ
-ig
EM
YA
Μ
- ig
weak

Τ
a
-------
2
W
Μ
a
- ig
strong

Λ
b
-------
2
G
Μ
b
The electromagnetic potential A_mu is a complex−valued 4−vector. The only way to form a scalar
with a 4−vector is to use a metric. Since it is complex−valued, use the conjugate like so:
A
Μ
A
Ν-
g
ΜΝ
= | A
0
|
2
- | A
1
|
2
- | A
2
|
2
- | A
3
|
2
Use the parity operator to flip the sign of the spatial part of a 4−vector:
A
Μ
A
Ν-p
g
ΜΝ
= | A
0
|
2
+ | A
1
|
2
+ | A
2
|
2
+ | A
3
|
2
Normalize the potential:
A
Μ
-------------
| A |

A
Ν-p
-------------
| A |
g
ΜΝ
= 1
From this, it can be concluded that the normalized 4−vector is an element of the symmetry group
U(1) if the multiplication operator is the metric combined with the parity and conjugate operators.
One does not need the Y in standard model Lagrangian, so this simplifies things. The same logic
applies to the 4−vector potentials for the weak and the strong forces which happen to have internal
symmetries.
In curved spacetime, the previous equation will not equal one. Mass breaks U(1), SU(2), and SU(3)
symmetry, but does so in a precise way (meaning one can calculate what the previous equation
should equal). There is no need for the Higgs mechanism to give particles mass while preserving
U(1)xSU(2)xSU(3) symmetry, so this proposal predicts no Higgs particle will be found.
Forces
The Lorentz Force of electromagnetism involves charge, velocity and the anti−symmetric field
strength tensor:
F
EM
Μ
= q
U
Ν
-------
c
(o
Μ
A
Ν
- o
Ν
A
Μ
)
Form an analogous force for gravity using the same substitutions as before:
F
G Μ
= -
_
G m
U
Ν
-------
c
(V
Μ
A
Ν
+ V
Ν
A
Μ
)
The gravitational force and the electromagnetic force behave differently under charge inversion. If
the mass changes signs, then both side flip signs, so nothing has really changed. If electric change
changes signs, the change in momentum will not change signs. The different behavior under charge
inversion may explain why gravitational force is unidirectional, but electrical forces can attract or
repulse.
94

The gravitational force and the electromagnetic force behave differently under charge inversion. If
the mass changes signs, then both side flip signs, so nothing has really changed. If electric change
changes signs, the change in momentum will not change signs. The different behavior under charge
inversion may explain why gravitational force is unidirectional, but electrical forces can attract or
repulse.
The total force is a combination of the two:
F
GEM Μ
= (J
q Ν
- J
m Ν
) V
Μ
A
Ν
- (J
q Ν
+ J
m Ν
) V
Ν
A
Μ
If q >>

G m, the equation approaches the form of the Lorentz force law of electromagnetism. If
the force is zero, the equation has the form of a Killing’s equation, which is used to determine the
isometries of a metric. Geodesics are defined by examining the left−hand side of F_GEM:
om U
Μ
--------------

= m
oU
Μ
----------

+ U
Μ

om
--------

= 0
Assume dm/dtau = 0. Apply the chain rule, and then the definition of a covariant derivative to form
a geodesic equation:
0 = m
o
2
x
Μ
------------

2
+
m
----
c
F
ΝΩ
Μ
U
Ν
U

This equation says that if there is no force, all the acceleration seen in spacetime is due to space-
time curvature, the Christoffel symbol. The covariant derivatives on the right side of F_gEM can
also be expanded:
0 = ¡q -
_
G m¡
ox
Ν
-----------
c oΤ
o
Μ
A
Ν
-
¡q +
_
G m¡
ox
Ν
-----------
c oΤ
o
Ν
A
Μ
-
2
----
c
mF

ΜΝ
U
Ν
U

This equation says that spacetime curvature is caused by the change in the potential if there is no
external force. This is a novel statement. In general relativity, one compares two geodesics, and
based on an analysis of the tidal forces between the geodesics, determines the curvature. The uni-
fied geodesic equation asserts that the curvature can be calculated directly from the potential.
Notice that this equation contains terms linked to a mass m and a charge q, so the geodesic equa-
tion applies to electromagnetism as well as gravity.
Gravitational Force for a Weak Field
The total unified force law is relevant to physics because it contains the Lorentz force law of electro-
magnetism. It must be established that the terms coupled to the mass m are connected to what is
The next task is to find a solution to the unified field equations, and then put the solutions into the
force equation. The Poisson field equation of classical Newtonian gravity can be solved by a 1/R
potential. The potential has a point singularity where R = 0. The unified field equations are relativis-
tic, so time must also be incorporated. A 1/distance potential does not solve the field equations in
four dimensions. In local covariant coordinates where the connection is zero, the potential A_mu =
(1/sigma^2, 0) solves the field equations, where sigma squared is the Lorentz invariant distance, or
the negative of the square of the Lorentz invariant interval tau. Distance is used instead of the
interval because classical gravity depends on distance, not time. The idea is to consider the time
contribution to be very small relative to the distance. Such a potential has as a singularity that is
the entire lightcone, where sigma ^2 = 0. This singularity may not be problematic because massless
particles are described by the Maxwell equations, but that hope will required a detailed study.
Gravity is a weak effect. It is common in quantum mechanics to normalize to one and study pertur-
bations of weak fields, an approach that will be followed here. Normalizing means there are small
steps will be away from one. Only first order terms will be kept. Here is the normalized potential
with a linear perturbation:
95

Gravity is a weak effect. It is common in quantum mechanics to normalize to one and study pertur-
bations of weak fields, an approach that will be followed here. Normalizing means there are small
steps will be away from one. Only first order terms will be kept. Here is the normalized potential
with a linear perturbation:
A
Μ
=

_
G h
-------------
c
2
Σ
2
, 0
|

-

c ,

_
G

1
---------
_
2
+
k
-------
Σ
2
x
|

2
+

1
---------
_
2
+
k
-------
Σ
2
y
|

2
+

1
---------
_
2
+
k
-------
Σ
2
z
|

2
-

1
---------
_
2
+
k
-------
Σ
2
t
|

2
|

|

, 0

|

This potential solves the 4D wave equation because the shift by the one over root two factor and
the rescaling by the spring constant k over sigma square do not effect the differential equation.
One interesting aspect is the shift of units from one that depends on h − suggesting quantum
mechanics − to the normalized perturbation which appears to be classical because there is no h.
Take the derivative with respect to t, x, y, and z:

--------
ot
=
c
2
k
---------------
_
G Σ
2
+ O (k
2
)
c

--------
ox
= -
c
2
k
---------------
_
G Σ
2
+ O (k
2
)
c

--------
oy
= -
c
2
k
---------------
_
G Σ
2
+ O (k
2
)
c

--------
oz
= -
c
2
k
---------------
_
G Σ
2
+ O (k
2
)
The change in the potential is a function of a spring constant k over sigma squared. The classical
Newtonian dependence on distance is an inverse square, so this is promising. One problem is that
a potential that applies exclusively to gravity is sought, yet the non−zero gradient of Φ indicates an
electric field. The sign of the spring constant k does not effect the solution to the 4D wave field
equations but does change the derivative of the potential. A potential that only has derivatives
along the diagonal of the field strength tensor can be constructed from two potentials that differ by
spring constants that either constructively interfere to create non−zero derivatives, or destructively
interfere to eliminate derivatives.
diagonal SHO A
Μ
=
c
---------
_
G

1 ,

1
---------
_
2
+
k
-------
Σ
2
x
|

2
+

1
---------
_
2
+
k
-------
Σ
2
y
|

2
+

1
---------
_
2
+
k
-------
Σ
2
z
|

2
-

1
---------
_
2
+
k
-------
Σ
2
t
|

2
|

+
96

1 ,

1
---------
_
2
-
k
-------
Σ
2
x
|

2
+

1
---------
_
2
-
k
-------
Σ
2
y
|

2
+

1
---------
_
2
-
k
-------
Σ
2
z
|

2
-

1
---------
_
2
+
k
-------
Σ
2
t
|

2
|

,
1 ,

1
---------
_
2
+
k
-------
Σ
2
x
|

2
+

1
---------
_
2
+
k
-------
Σ
2
y
|

2
+

1
---------
_
2
+
k
-------
Σ
2
z
|

2
-

1
---------
_
2
+
k
-------
Σ
2
t
|

2
|

+
1 ,

1
---------
_
2
+
k
-------
Σ
2
x
|

2
+

1
---------
_
2
-
k
-------
Σ
2
y
|

2
+

1
---------
_
2
-
k
-------
Σ
2
z
|

2
-

1
---------
_
2
-
k
-------
Σ
2
t
|

2
|

,
1 ,

1
---------
_
2
+
k
-------
Σ
2
x
|

2
+

1
---------
_
2
+
k
-------
Σ
2
y
|

2
+

1
---------
_
2
+
k
-------
Σ
2
z
|

2
-

1
---------
_
2
+
k
-------
Σ
2
t
|

2
|

+
1 ,

1
---------
_
2
-
k
-------
Σ
2
x
|

2
+

1
---------
_
2
+
k
-------
Σ
2
y
|

2
+

1
---------
_
2
-
k
-------
Σ
2
z
|

2
-

1
---------
_
2
-
k
-------
Σ
2
t
|

2
|

,
1 ,

1
---------
_
2
+
k
-------
Σ
2
x
|

2
+

1
---------
_
2
+
k
-------
Σ
2
y
|

2
+

1
---------
_
2
+
k
-------
Σ
2
z
|

2
-

1
---------
_
2
+
k
-------
Σ
2
t
|

2
|

+
1 ,

1
---------
_
2
-
k
-------
Σ
2
x
|

2
+

1
---------
_
2
-
k
-------
Σ
2
y
|

2
+

1
---------
_
2
+
k
-------
Σ
2
z
|

2
-

1
---------
_
2
-
k
-------
Σ
2
t
|

2
|

|

97

Take the contravariant derivative of this potential, keeping only the terms to first order in the
spring constant k.
V
Μ
A
Ν
=
c
2
---------
_
G

k
-------
Σ
2

1 0 0 0
0 -1 0 0
0 0 -1 0
0 0 0 -1
|

All this work to get a multiple of the identity matrix! Plug this into the gravitational force equation:
F
g
Μ
= m c

-
k
-------
Σ
2

ot
--------

,
k
-------
Σ
2

oR

--------

|

This is a relativistic force law for a weak gravitational field for the inverse interval squared diago-
nal potential. When spacetime symmetry is broken, this equation will lead to Newton’s law of
gravity in the next section. If spacetime symmetry is maintained, then solving the force equation
and eliminating the constants yields a metric equation for gravity.
Newton’s Law of Gravity and More
Several assumptions need to be made to apply the weak gravitational force equation to a classical
gravitational system. First, assume that the spring constant is due to the source mass, k = GM.
Second, assume that the field is static, so that Σ
2
R
2
c
2
t
2
R
2
. In the approximation, it does
not depend on time.
Newtonian spacetime is different from Minkowski spacetime because the speed of light is infinite.
Spacetime symmetry must be broken. A question arises about how to do this in a formal mathemati-
cal sense. The Minkowski interval tau is a consequence of the relationship between time t and
space R. The functional relationship between time and space must be severed. By the static field
approximation, there is a distance R which is the same magnitude as the interval tau. If the interval
tau is replaced by the scalar distance R, then that will sever the functional relationship between
time and space:

ot
--------

,
oR

-----------
c oΤ
|

-

c ot
----------------
o | R |
,
oR

----------------
o | R |
|

= (0, R

)
Plug these three assumptions into force equation:
F
g
Μ
= ¡0, -
G M m
------------
R
2
R

¡
This is not quite Newton’s gravitational force law. The reason is that one must consider the left−
hand side of the force equation carefully. According to the chain rule:
om U
Μ
--------------

= m
oU
Μ
----------

+ U
Μ

om
--------

An open question is how should spacetime symmetry be broken for the derivatives with respect to
the interval tau? An interval is composed of both changes in time and space. For the acceleration
term, if the interval is only about time, then one gets back Newtonian acceleration. For logical
consistency, one might be tempted to also substitute time in the dm/dtau term. However, the sys-
tem is presumed to be static, so this would necessarily be zero. If this derivative is to have any
chance at being non−zero, it would have to be with respect to the absolute value of R as has been
done earlier in the derivation. So the classical force law should look like so:
98

An open question is how should spacetime symmetry be broken for the derivatives with respect to
the interval tau? An interval is composed of both changes in time and space. For the acceleration
term, if the interval is only about time, then one gets back Newtonian acceleration. For logical
consistency, one might be tempted to also substitute time in the dm/dtau term. However, the sys-
tem is presumed to be static, so this would necessarily be zero. If this derivative is to have any
chance at being non−zero, it would have to be with respect to the absolute value of R as has been
done earlier in the derivation. So the classical force law should look like so:
m
o
2
R

----------
ot
2
+ c
oR

--------
ot

om
----------------
o | R |
= -
G M m
------------
R
2
(R

+ V

)
This is Newton’s law of gravity working along a new direction, that for the velocity vector. This
new effect will necessarily be small due to the constant c in the term.
For a point source, the dm/d|R| term will not make a contribution, and one gets Newton’s law of
gravity. It is only if the inertial mass is distributed over space like for the big bang or galaxies will
the term come into play. If the velocity is constant, then the acceleration is zero. The equation
describes the distribution of the inertial mass m that makes up the total gravitational source mass
M. The solution to the force equation when there is no acceleration is a stable exponential. Big
bang cosmology has two problems: all matter is traveling at exactly the same speed even though it
is not possible for them to communicate (the horizon problem), and the model require high levels
of precision on initial conditions to avoid collapse (the flatness problem). [A. H. Guth, Phys. Rev.
D., 23:347−356, 1981] The force equation has a stable, constant velocity solution which may
resolve both problems of the big bang without the inflation hypothesis. Their is also a problem with
the rotation profile of thin disk galaxies.[S. M. Kent, Astron. J., 91:1301−1327, 1986; S. M. Kent,
Astron. J., 93:816−832, 1987] Once the maximum velocity is reached, the velocity stays constant.
It has been shown that galaxies should not be stable at all.[A. Toomre, Astrophys. J., 139:1217,
1964] Both problems may again be resolved with stable constant velocity solutions. Numerical
approaches on the above equation should be conducted.
A Metric Equation
The weak gravitational force equation is two second−order differential equations. The equation can
be simplified to a set of first−order differential equations by substituting (U^0, U) = (c dt/dtau,
dR/dtau)
oU
0
----------

-
k
-------
Τ
2
U
0
= 0
oU

--------

+
k
-------
Τ
2
U

= 0
The solution involves exponentials:
U
Μ
= ¡ve
-
k
----
Τ
, V

e
k
----
Τ
¡
For flat spacetime, U^mu = (v, V). The constraint on relativistic velocities in flat spacetime is:
U
Μ
U
Μ
=
c
2
dt
2
- dR
2
----------------------------

2
= c
2
= v
2
- V

V

Solve for the constants, and plug back into the constraint, multiplying through by dtau ^2.

2
= e
-2
k
----
Τ
dt
2
- e
2
k
----
Τ

dR
2
---------
c
2
Make the same two assumptions as before: the spring constant is due to the gravitational source, k
= GM/c^2, and the field is static, so Τ
2
R
2
c
2
t
2
R
2
c
2
. There is one more degree of free-
dom, because the radius R could either be positive or negative. To make the metric consistent with
experiment, choose the negative root:
99

Make the same two assumptions as before: the spring constant is due to the gravitational source, k
= GM/c^2, and the field is static, so Τ
2
R
2
c
2
t
2
R
2
c
2
. There is one more degree of free-
dom, because the radius R could either be positive or negative. To make the metric consistent with
experiment, choose the negative root:

2
= e
-2
GM
---------
c
2
R
dt
2
- e
2
GM
---------
c
2
R
dR
2
This equation has the form of a metric equation. Perform a Taylor series expansion to second order
in GM/c^2R:

2
=
¸1 - 2
GM
-----------
c
2
R
+ 2 ¡
GM
-----------
c
2
R
¡
2
¸ dt
2
- ¸1 + 2
GM
-----------
c
2
R
+ 2 ¡
GM
-----------
c
2
R
¡
2
¸ dR
2
If one compares this metric to the Schwarzschild metric in isotropic coordinates to parameterized
post−Newtonian accuracy, the coefficients are identical. For that reason, this metric is consistent
with all experimental tests weak field tests of general relativity. [C. M. Will, "Theory and experi-
ment in gravitational physics: Revised edition", Cambridge University Press, 1993.]
For higher order terms of the Taylor series expansion, the two metric will predict different coeffi-
cients. The validity of this proposal can thus be tested experimentally. It will require a great deal of
effort and skill to conduct such experiments, since many physical phenomena will have to be
accounted for (an example: the quadrupole moment of the Sun for solar tests). According to per-
sonal communication with Prof. Clifford Will − a leading authority in experimental tests of gravity
theories − no experiments are being planned to monitor second order PPN coefficients at this time.
Conclusion
Using a nineteenth century approach, an effort to unify physics from the twentieth century has been
attempted. The description of geodesics by general relativity is not complete because it does not
explicitly show how the potential source causes curvature. A dynamic metric equation is found but
it uses a simpler set of field equations (a rank one tensor instead of two). In the standard model as
elsewhere, combining two 4−vectors requires a metric. By normalizing the 4−vectors, the unitary
aspect of the standard model can be self−evident.
This theory makes three testable predictions, two subtle, one not. First, the polarity of gravitational
waves will be scalar or longitudinal, not transverse as predicted by general relativity. Second, if
gravitation effects are measured to secondary parameterized post−Newtonian accuracy, the coeffi-
cients for the metric derived here are different from the Schwarzschild metric in isotropic coordi-
nates. Such an experiment will be quite difficult to do. The third test is to see if the complete relativ-
istic force equation matches all the data for a thin spiral galaxy. It is this test which should be
investigated first.
100

Einstein’s Vision I: Classical Unified Field Equations
Abstract
The equations governing gravity and electromagnetism show both profound similarities and unam-
biguous differences. Albert Einstein worked to unify gravity and electromagnetism, mainly by
trying to generalize Riemannian geometry. Hamilton’s quaternions are a 4−dimensional topologi-
cal algebraic field related to the real and complex numbers equipped with a static Euclidean 4−ba-
sis. Riemannian quaternions as defined herein explicitly allow for dynamic changes in the basis
vectors. The equivalence principle of general relativity which applies only to mass is generalized
because for any Riemannian quaternion differential equation, the chain rule means a change could
be caused by the potential and/or the basis vectors. The Maxwell equations are generated using
quaternion potentials and operators. Unfortunately, the algebra is complicated. The unified force
field proposed is modeled on a simplification of the electromagnetic field strength tensor, being
formed by a quaternion differential operator acting on a potential, Box* A*. This generates an
even, antisymmetric−matrix field strength quaternion for electricity and an odd, antisymmetric−ma-
trix field strength quaternion for magnetism, where the even field conserves its sign if the order of
the differential and the potential are reversed unlike the odd field. Gauge symmetry is broken for
massive particles by the even, symmetric−matrix term, which is interpreted as being due to gravity.
In tensor analysis, a differential operator acting on the field strength tensor creates the Maxwell
equations. The unified field equations for an isolated source are generated by acting on the unified
force field with an additional differential operator, Box* Box* A* = 4 pi J*. This contains a quater-
nion representation of the Maxwell equations, a classical link to the quantum Aharonov−Bohm
effect, and dynamic field equations for gravity. Vacuum and zero net current solutions to the
unified field equations are discussed. The field equations conserve both electric charge density and
mass density. Under a Lorentz transformation, the gravitational and electromagnetic fields are
Lorentz invariant and Lorentz covariant respectively, but there are residual terms whose meaning is
not clear presently. An additional constraint is required for gauge transformations of a massive
field. (PACS:12.10.−g)
Einstein’s Vision Using Quaternions
Three of the four known forces in physics have been unified via the standard model: the electromag-
netic, the weak, and the strong forces. The holdout remains gravity, the first force characterized
mathematically by Isaac Newton. The parallels between gravity and electromagnetism are evident.
Newton’s law of gravity and Coulomb’s law are inverse square laws. Both forces can be attractive,
but Coulomb’s law can also be a repulsive force. A long−standing goal of modern physics is to
explain the similarities and differences between gravity and electromagnetism.
Albert Einstein had a specific idea for how to formulate an acceptable unified field theory (see Fig.
1, taken from A. Pais, "Subtle is the Lord..." the science and life of Albert Einstein", Claredon Pres,
1982).
101

One unusual aspect of Einstein’s view was that he believed the unified field would lead to a new
foundation for quantum mechanics, an idea which is not shared by most of today’s thinkers (S.
Weinberg, "Dreams of a final theory," Pantheon Books, New York, 1992). Most of Einstein’s
efforts over 40 years were directed in a search to generalize Riemannian differential geometry in
four dimensions.
To a degree which has pleasantly surprised the author, Einstein’s vision to unify gravity and electro-
magnetism has been followed. The construction of a new 4−dimensional geometry is dictated by
insights garnered from physics. Events in spacetime are composed of a scalar for time and a 3−vec-
tor for space. The four−dimensional topological algebraic field of quaternions has the same struc-
ture, so quaternions will be the starting point of this effort.
Laws of physics are expressed in a coordinate−independent way. The sum or difference of two
quaternions can only be defined if the two quaternions in question share the same 4−basis. Rieman-
nian quaternions make coordinate−independence explicit. In special relativity, regions in space-
time are delimited by the light cone, where the net change in 3−space is equal to the net change in
time. The parity between changes in 3−space and time is constructed into the definition of a Rie-
mannian quaternion. In general relativity, the field equations make the metric a dynamic variable.
The basis vectors of Riemannian quaternions can be dynamic, so the metric can be dynamic. The
dynamic nature of the basis vectors leads to the general equivalence principle, whereby any law,
even those in electromagnetism, can be the result of a change in reference frame.
Physical laws are the result of simple Riemannian quaternion differential equations. First−order
Riemannian quaternion differential equations create force fields for gravity, electricity, and magne-
tism. Second−order differential equations create dynamic field equations for gravity, the Maxwell
equations for electromagnetism, and a classical counterpart to the Aharanov−Bohm effect of quan-
tum mechanics. Third−order differential equations create conservation laws. Homogeneous solu-
tions to the second order differential equations are related to gauge symmetry.
The second paper in this series of three investigates a unified force law, with a focus on a particular
solution which may eliminate the need for dark matter to explain the mass distribution and velocity
profile for spiral galaxies. The third paper develops a new approach to quaternion analysis. The
equations of the first two papers are recast with the new definition of a quaternion derivative,
resulting in a quantum unified field and force theory.
Events in Spacetime and Quaternions
An event in spacetime is considered by the author as the fundamental form of information in phys-
ics. Events have structure. There are four degrees of freedom divided into two dissimilar parts:
time is a scalar, and space is a 3−vector. This structure should be reflected in all the mathematics
used to describe patterns of events. For this reason, this paper focuses exclusively on quaternions,
the 4−dimensional number where the terms scalar and vector where first used.
Hamilton’s quaternions, along with the far better know real and complex numbers, can be added,
subtracted, multiplied, and divided. Technically, these three numbers are the only finite−dimen-
sional, associative, topological, algebraic fields, up to an isomorphism (L. S. Pontryagin, "Topologi-
cal groups", translated from the Russian by Emma Lehmer, Princeton University Press, 1939).
Properties of these numbers are summarized in the table below by dimension, if totally ordered,
and if multiplication commutes:
102

Hamilton’s quaternions, along with the far better know real and complex numbers, can be added,
subtracted, multiplied, and divided. Technically, these three numbers are the only finite−dimen-
sional, associative, topological, algebraic fields, up to an isomorphism (L. S. Pontryagin, "Topologi-
cal groups", translated from the Russian by Emma Lehmer, Princeton University Press, 1939).
Properties of these numbers are summarized in the table below by dimension, if totally ordered,
and if multiplication commutes:
Number Dimensions Totally Ordered Commutative
Real 1 Yes Yes
Complex 2 No Yes
Quaternions 4 No No
Hamilton’s quaternions have a Euclidean 4−basis composed of 1, i, j, and k. The rules of multiplica-
tion were inspired by those for complex numbers: 1^2=1, i^2=j^2=k^2=ijk=−1. Quaternions also
have a real 4x4 matrix representation:
q (t, x, y, z) =

t -x -y -z
x t -z y
y z t -x
z -y x t
|

Although written in Cartesian coordinates, quaternions can be written in any linearly−independent
4−basis because matrix algebra provides the necessary techniques for changing the basis. There-
fore, like tensors, a quaternion equation is independent of the chosen basis. One could view quater-
nions as tensors restricted to a 4−dimensional algebraic field. For the sake of consistency, all
transformations are also constrained to the same division algebra. This constraint might first
appear too restrictive since for example it eliminates simple matrices for row permutations. Since
quaternions are an algebraic field, there necessarily exists a combination of quaternions that
achieves the action of a permutation. The need for consistency will overrule convenience.
Laws in physics are independent of coordinate systems. To make the coordinate independence
explicit, amplitudes and basis vectors will be separated using a new notation. Consider a quater-
nion 4−function, A_n=(a_0, a_1, a_2, a_3), and an arbitrary 4−basis, Ihat_n=(ihat_0, ihat_1,
ihat_2, ihat_3). In spacetime, the line that divides causality is define by the light cone. On the
light cone, the total change in 3−space over the change in time is equal to one. Physics therefore
indicates parity between the total 3−vector and the scalar, instead of weighing all four equally. A
coordinate−independent Riemannian quaternion is defined to be A_0 Ihat_n=(a_0 ihat_0, a_1
ihat_1/3, a_2 ihat_2/3, a_3 ihat_3/3).
The equivalence principle of general relativity asserts, with experiments to back it up, that the
inertial mass equals the gravitational mass. An accelerated reference frame can be indistinguish-
able from the effect of a mass density. No corresponding principle applies to electromagnetism,
which depends only on the electromagnetic field tensor built from the potential. With Riemannian
quaternions, the 4−unit vector does not have to be static, as illustrated by taking the time derivative
of the first term and using the chain rule:
oa
0
i

0
----------------
oi
0
= i

0

oa
0
----------
oi
0
+ a
0

oi

0
----------
oi
0
The unit vector for time, ihat_0, can change over an infinitely small amount of time, i_0. Any
change in a quaternion potential function could be due to contributions from a change in potential,
the ihat_0 da_0/di_0 term, and/or a change in the basis, the a_0 dihat_0/di_0 term. Is this mathemat-
ical property related to physics? Consider Gauss’ law written with Riemannian quaternions:
103

-
i

n
2
---------
9

oE
n
----------
oi
n
-
i

n
E
n
-------------
9

oi

n
----------
oi
n
= 4 ΠΡ, n = 1, 2, 3
The divergence of the electric field might equal the source, or equivalently, the divergence of the
basis vectors. The "general equivalence principle" as defined here means that any measurement
can be due to a change in the potential and/or a change in the basis vectors. The general equiva-
lence principle is applicable to both gravity and electromagnetism.
Metrics and Quaternion Products
The theories of special and general relativity dictate the distance between events in spacetime.
Although fundamentally different in their mathematical structure, inertia is a link between the two.
Special relativity dictates the transformation rules for observers who change their inertia, assuming
the system observed does not change. The field equations of general relativity detail the changes in
distance due to a system changing its inertia from the vacuum to a non−zero energy density. A
quaternion product necessarily contains information about the metric, but also has information in
the 3−vector. This additional information about quaternion products will suggest a provocative
link between metrics and inertia consistent with both special and general relativity.
Most structures in Nature do not transform like a scalar and a 3−vector. Quaternion products
multiply two 4−basis vectors, and those products will transform differently. The rules of quater-
nion multiplication mirror those of complex numbers. Instead of the imaginary number i, there is a
unit 3−vector for each quaternion playing an analogous role. The difference is that unit 3−vectors
do not all have to point in the same direction. Based on the angle between them, two different unit
3−vectors have both a dot and cross product. The dot and cross products completely characterize
the relationship between the two unit vectors. Compare the product of multiplying two complex
numbers (a, bi) and (c, di):
(a, bi) (c, di) = (ac - bd, ad + bc),
with two quaternions, (a, B I

) and (c, D I

),
(a, B I

) (c, D I

’) = (ac - BD I

I

’ + Bc I

+ BD I

xI

’)
Complex numbers commute because they do not have a cross product in the result. If the order of
quaternion multiplication is reversed, then only the cross product would change its sign. Quater-
nion multiplication does not commute due to the behavior of the cross product. If the cross product
is zero, then quaternion multiplication has all of the properties of complex numbers. If, on the
other hand, the only value of a quaternion product is equal to the cross product, then multiplication
is anti−commutative. Individually, the mathematical properties of commuting and anti−commut-
ing algebras are well known. A quaternion product is the superposition of these two types of
algebras that forms a division algebra.
Several steps are required to square of the difference of two Riemannian quaternions to form a
measure of distance. First, the basis of the two quaternions must be shared. It makes no sense to
subtract something in spherical coordinates from something in Cartesian coordinates. The basis
does not have to be constant, only shared. Every quaternion commutes with itself, so the cross
product is zero. There are seven unique pairs of basis vectors in a square:

da
0
i

0
, dA
n
I

n
-------
3
|

2
=

da
0
2
i

0
2
- dA
n
2
I

n
2
---------
9
, 2 da
0
dA
n

i

0
I

n
-------------
3
|

The signs were chosen to be consistent with Hamilton’s quaternion algebra. The four square basis
vectors i_mu^2 define the metric. If the basis vectors are not constant, then the metric is dynamic.
Define a "3−rope" to be the three other terms, which have the form i_0 I_n. Notice that the 3−rope
starts in one time−space location and will have a non−zero length if it ends up at a different loca-
tion and time. With quaternion products, the 3−rope is a natural companion to a metric for informa-
104

The signs were chosen to be consistent with Hamilton’s quaternion algebra. The four square basis
vectors i_mu^2 define the metric. If the basis vectors are not constant, then the metric is dynamic.
Define a "3−rope" to be the three other terms, which have the form i_0 I_n. Notice that the 3−rope
starts in one time−space location and will have a non−zero length if it ends up at a different loca-
tion and time. With quaternion products, the 3−rope is a natural companion to a metric for informa-
In special relativity, if the inertia of the observer but not the system is changed, the metric is invari-
ant. The 3−rope is covariant, because it is known how it changes. A complementary hypothesis to
the invariant metric of special relativity would propose the if the inertia of the system but not the
observer is changed, there exists a choice of basis vectors such that the 3−rope is invariant but the
metric changes in a known way. This could be written in algebraically using the following rule:
i

0
2
=
-1
---------
I

n
2
, ¡ i

0
I

n
¡ = 1
If the magnitude of the time and 3−space basis vectors are inversely related, the magnitude of the
product of the time basis vector with each 3−space basis vector will be constant even if the basis
vectors themselves are dynamic. This hypothesis asserts there exists such a basis, but that particu-
lar basis does not have to be used.
Hamilton had the freedom to use the rule found in the above equation, but made the more obvious
choice of i_0^2 = −I_n^2. The existence of a basis where the 3−rope is constant despite a change
in the inertia of the system will have to be treated as provisional in this paper. In the second paper
of this series, a metric with this property will be found and discussed.
Physically Relevant Differential Equations
Is there a rational way to construct physically relevant quaternion equations? The method used here
will be to mimic the tensor equations of electromagnetism. The electromagnetic field strength
tensor is formed by a differential operator acting on a potential. The Maxwell equations are
formed by acting on the field with another differential operator. The Lorentz 4−force is created by
the product of a electric charge, the electromagnetic field strength tensor, and a 4−velocity. This
pattern will be repeated starting from an asymmetric field to create the same field and force equa-
tions using quaternion differentials and potentials. The challenge in this exercise is in the interpreta-
tion, to see how every term connects to established laws of physics.
As a first step to constructing differential equations, examine how the differential operator (d/dt,
Del) acts on a potential function (phi, A):
¸
o
--------
ot
,

¸ (Φ, A

) =

--------
ot
-

.A

,
oA

--------
ot
+

Φ +

x A
|

For the sake of clarity, the notation introduced for Riemann quaternions has been suppress, so the
reader is encouraged to recognize that there are also a parallel set of terms for changes in the basis
vectors. The previous equation is a complete assessment of the change in the 4−dimensional
potential/basis, involving two time derivatives, the divergence, the gradient and the curl all in one.
A unified field theory should account for all conceivable forms of change in a 4−dimensional
potential/basis, as is the case here.
Quaternion operators and potentials have not been used to express the Maxwell equations. The
reason can be found in the previous equation, where the sign of the divergence of A is opposite of
the curl of A. In the Maxwell equations, the divergence and the curl involving the electric and
magnetic field are all positive. Many others, even in Maxwell’s time, have used complex−valued
quaternions for the task because the extra imaginary number can be used to get the signs correct.
However, complex−valued quaternions are not an algebraic field. The norm, t^2+x^2+y^2+z^2,
for a non−zero quaternion could equal zero if the values of t, x, y, and z were complex. This paper
involves the constraint of working exclusively with 4−dimensional algebraic fields. Therefore, no
matter how salutary the work with complex−valued quaternions, it is not relevant to this paper.
105

Quaternion operators and potentials have not been used to express the Maxwell equations. The
reason can be found in the previous equation, where the sign of the divergence of A is opposite of
the curl of A. In the Maxwell equations, the divergence and the curl involving the electric and
magnetic field are all positive. Many others, even in Maxwell’s time, have used complex−valued
quaternions for the task because the extra imaginary number can be used to get the signs correct.
However, complex−valued quaternions are not an algebraic field. The norm, t^2+x^2+y^2+z^2,
for a non−zero quaternion could equal zero if the values of t, x, y, and z were complex. This paper
involves the constraint of working exclusively with 4−dimensional algebraic fields. Therefore, no
matter how salutary the work with complex−valued quaternions, it is not relevant to this paper.
The reason to hope for unification using quaternions can be found in an analysis of symmetry
provided by Albert Einstein:
"The physical world is represented as a four−dimensional continuum. If in this I adopt a Rieman-
nian metric, and look for the simplest laws which such a metric can satisfy, I arrive at the relativis-
tic gravitation theory of empty space. If I adopt in this space a vector field, or the antisymmetric
tensor field derived from it, and if I look for the simplest laws which such a field can satisfy, I
arrive at the Maxwell equations for free space." [einstein1934]
The "four−dimensional continuum" could be viewed as a technical constraint involving topology.
Fortunately, quaternions do have a topological structure since they have a norm. Nature is asymmet-
ric, containing both a symmetric metric for gravity and an antisymmetric tensor for electromagne-
tism. With this in mind, rewrite out the real 4x4 matrix representation of a quaternion:
q (t, x, y, z) =

t 0 0 0
0 t 0 0
0 0 t 0
0 0 0 t
|

+

0 -x -y -z
x 0 -z y
y z 0 -x
z -y x 0
|

The scalar component (t in representation above) can be represented by a symmetric 4x4 matrix,
invariant under transposition and conjugation (these are the same operations for quaternions). The
3−vector component (x, y and z in the representation above) is off−diagonal and can be represented
by an antisymmetric 4x4 matrix, because taking the transpose will flip the signs of the 3−vector.
Quaternions are asymmetric in their matrix representation, a property which is critical to using
them for unifying gravity and electromagnetism.
Recreating the Maxwell Equations
Maxwell speculated that his set of equations might be expressed with quaternions someday (J. C.
Maxwell, "Treatise on Electricity and Magnetism," Dover reprint, third edition, 1954). The diver-
gence, gradient, and curl were initially developed by Hamilton during his investigation of quater-
nions. For the sake of logical consistency, any system of differential equations, such as the Max-
well equations, that depends on these tools must have a quaternion representation.
The Maxwell equations are gauge invariant. How can this property be built into a quaternion
expression? Consider a common gauge such as the Lorenz gauge, dphi/dt + div A = 0. In quater-
nion parlance, this is a quaternion−scalar formed from a differential quaternion acting on a poten-
tial. To be invariant under an arbitrary gauge transformation, the quaternion−scalar must be set to
zero. This can be done with the vector operator, (q−q*)/2. Search for a combination of quaternion
operators and potentials that generate the Maxwell equations:
(.
-
Vector (.
-
A
-
) - .Vector (.A))
-
-------------------------------------------------------------------------------------
2
=
=

V

(V

xA

),
o
--------
ot

oA

--------
ot
+ V

Φ
|

+ V

x(V

xA

)
|

=
=

V

B

, -
oE

--------
ot
+ V

xB
|

=
106

= (0, 4 Π J

).
This is Ampere’s law and the no monopoles vector identity (assuming a simply−connected topol-
ogy). Any choice of gauge will not make a contribution due to the vector operator. If the vector
operator was not used, then the gradient of the symmetric−matrix force field would be linked to the
electromagnetic source equation, Ampere’s law.
Generate the other two Maxwell equations:
-(.Vector (.
-
A
-
) + .
-
Vector (.A))
-
----------------------------------------------------------------------------------------
2
=
=

V

-
oA

--------
ot
- V

Φ
|

,
oV

xA

---------------
ot
+ V

x

-
oA

--------
ot
- V

Φ
|

|

=
=

V

E

,
oB

--------
ot
+ V

xE
|

=
= ¡4 ΠΡ, 0

¡.
This is Gauss’ and Faraday’s law. Again, if the vector operator had not been used, the time deriva-
tive of the symmetric−matrix force field would be associated with the electromagnetic source
equation, Gauss’ law. To specify the Maxwell equations completely, two quaternion equations are
required, just like the 4−vector approach.
Although successful, the quaternion expression is unappealing for reasons of simplicity, consis-
tency and completeness. A complicated collection of sums or differences of differential operators
acting on potentials − along with their conjugates − is required. There is no obvious reason this
combination of terms should be central to the nature of light. One motivation for the search for a
unified potential field involves simplifying the above expressions.
When a quaternion differential acts on a function, the divergence always has a sign opposite the
curl. The opposite situation applies to the Maxwell equations. Of course the signs of the Maxwell
equations cannot be changed. However, it may be worth the effort to explore equations with sign
conventions consistent with the quaternion algebra, where the operators for divergence and curl
were conceived.
Information about the change in the potential is explicitly discarded by the vector operator. Justifi-
cation comes from the plea for gauge symmetry, essential for the Maxwell equations. The Max-
well equations apply to massless particles. Gauge symmetry is broken for massive fields. More
information about the potential might be used in unification of electromagnetism with gravity. A
gauge is also matrix symmetric, so it could provide a complete picture concerning symmetry.
One Unified Force Field from One Potential Field
For massless particles, the Maxwell equations are sufficient to explain classical and quantum
electrodynamic phenomena in a gauge−invariant way. To unify electromagnetism with gravity, the
gauge symmetry must be broken, opening the door to massive particles. Because of the constraints
imposed by quaternion algebra, there is little freedom to choose the gauge with a simple quaternion
expression. In the standard approach to the electromagnetic field, a differential 4−vector acts on a
4−vector potential in such a way as to create an antisymmetric second−rank tensor. The unified
field hypothesis proposed involves a quaternion differential operator acting on a quaternion
potential:
107

.
-
A
-
=

--------
ot
-

.A

, -
oA

--------
ot
-

Φ +

x A
|

This is a natural suggestion with this algebra. The antisymmetric−matrix component of the unified
field has the same elements as the standard electromagnetic field tensor. Define the electric field E
as the even terms, the ones that will not change signs if the order of the differential operator and the
potential are reversed. The magnetic field B is the curl of A, the odd term. The justification for
proposing the unified force field hypothesis rests on the presence of the electric and magnetic fields.
In some ways, the above equation looks just like the old idea of combining a scalar gauge field
with the electromagnetic field strength tensor, as Gupta did in 1950 in order to quantize the Max-
well equations. He concluded that although useful because it is written in manifestly relativistic
form, no new results beyond the Maxwell equation are obtained. Examine just the gauge contribu-
tion to the Lagrangian for this unified field:
L = -
1
----
2
¸
oO
--------
ot
, -
1
----
3

oA
x
----------
ox
, -
1
----
3

oA
y
----------
oy
, -
1
----
3

oA
z
----------
oz
¸
2
Take the derivative of the Lagrangian with respect to the gauge variables:
oL
-------------
o
oA
Μ
-------
ox
Μ
= 0
By Noether’s theorem, this conserved current indicates a symmetry of the Lagrangian. This is why
the proposal involves new physics. The gauge is a dynamic variable constrained by the
Lagrangian.
A quaternion potential function has four degrees of freedom represented by the scalar function phi
and the 3−vector function A. Acting on this with one[or more] differential operators does not
change the degrees of freedom. Instead, the tangent spaces of the potential will offer more subtle
views on the rules for how potentials change.
The three classical force fields, g, E, and B, depend on the same quaternion potential, so there are
only four degrees of freedom. With seven components to the three classical force fields, there must
be three constraints between the fields. Two constraints are already familiar. The electric and
magnetic field form a vector identity via Faraday’s law. Assuming spacetime is simply connected,
the no monopoles equation is another identity. A new constraint arises because both the force
fields for gravity and electricity are even. It will be shown subsequently how the even force fields
can partially constructively or destructively interfere with each other.
Unified Field Equations
In the standard approach to generating the Maxwell equations, a differential operator acts on the
electromagnetic field strength tensor. A unified field hypothesis for an isolated source is proposed
which involves a differential quaternion operator acting on the unified field:
4 Π (Ρ, J

)
-
= ¸
o
--------
ot
,

¸
-
¸
o
--------
ot
,

¸
-
(Φ, A

)
-
=
=

o
2
Φ
----------
ot
2
- 2 V

oA

--------
ot
- V

V

Φ + V

(V

xA

),
108

-2 V

--------
ot
+ V

(V

A

) -
o
2
A

----------
ot
2
- V

x(V

xA

) +
+2 V

x
oA

--------
ot
+ V

xV

Φ
|

.
This second order set of four partial differential equations has four unknowns so this is a complete
set of field equations. Rewrite the equations above in terms of the classical force fields:
4 Π (Ρ, J

)
-
= ¸
og
--------
ot
+

.E

+

.B

,
-V

g +
oE

--------
ot
-

x B

+
oB

--------
ot
-

x E
|

.
The unified field equations contain three of the four Maxwell equations explicitly: Gauss’ law, the
no magnetic monopoles law, and Ampere’s law. Faraday’s law is a vector identity, so it is still true
implicitly. Therefore, a subset of the unified field equations contains a quaternion representation of
the Maxwell equations. The presence of the Maxwell equations justifies the investigation of the
unified field equations.
There is a simple relationship between Faraday’s law and the equation above. All that needs to be
done is to subtract twice the time derivative of the magnetic field from both sides. What does this
do to the 4−vector current density J? Now there is a current that transforms like a pseudo−current
density. The inclusion of a pseudo−current along with a current making the proposal more com-
plete. The volume integral of this pseudo−current density is the total magnetic flux:
k
¸ ¸ ¸
oB
--------
ot
dV =
e
-------
c
I
B
The unified field equation postulates a pseudo 3−vector current composed of the difference
between the time derivative of the magnetic field and the curl of the electric field. The Aharonov−
Bohm effect (Y Aharonov and D. Bohm, "Significance of electromagnetic potentials in the quan-
tum theory," Phys. Rev, 115:485−491, 1959) depends on the total magnetic flux to create changes
seen in the energy spectrum. The volume integral of the time derivative of the magnetic field is a
measure of the total magnetic flux. The pseudo−current density is quite unusual, transforming
differently under space inversion than the electric current density. One might imagine that a
Lorentz transformation would shift this pseudo−current density into a pseudo−charge density. This
does not happen however, because the vector identity involving the divergence of a curl still
applies. The Aharonov−Bohm phenomenon, first viewed as a purely quantum effect, may have a
classical analogue in the unified field equations.
The field equations involving the gravitational force field are dynamic and depend on four dimen-
sions. This makes them likely to be consistent with special relativity. Since they are generated
alongside the Maxwell equations, one can reasonably expect the differential equations will share
many properties, with the ones involving the symmetric−matrix gravitational force field being
more symmetric than those of the electromagnetic counterpart.
The unified source can be defined in terms of more familiar charge and current densities by sepa-
rately setting the gravity or electromagnetic field equal to zero. In these cases, the source is due
only to electricity or mass respectively. This leads to connections between the unified source,
mass, and charge:
J = J
m
iff E

= B

= 0

109

J = J
e
+ J

AB
iff g = 0.
It would be incorrect − but almost true − to say that the unified charge and current are simply the
sum of the three: mass, electric charge, and the Aharanov−Bohm pseudo−current (or total magnetic
flux over the volume). These terms constructively interfere with each other, so they may not be
viewed as being linearly independent.
Up to four linearly independent unified field equations can be formulated. A different set could be
created by using the differential operator without taking its conjugate:
4 ΠJ
-
= ..
-
A
-
=
=

o
2
Φ
----------
ot
2
+ V

V

Φ - V

(V

xA

), -
o
2
A

----------
ot
2
- V

2
A

- V

xV

Φ
|

= ¸
og
--------
ot
-

.E

-

.B

,
V

g +
oE

--------
ot
+

x B

+
oB

--------
ot
+

x E
|

.
This is an elliptic equation. Since the goal of this work is a complete system of field equations, this
may turn out to be an advantage. An elliptic equation combined with a hyperbolic one might more
fully describe gravitational and electromagnetic waves from sources. Unlike the first set of field
equations, the cross terms destructively interfere with each other.
The elliptic field equation again contains three of four Maxwell equations explicitly:Gauss’ law,
the no magnetic monopoles vector identity and Faraday’s law. This time, Ampere’s law looks
different. To be consistent with Ampere’s law, again a pseudo−current must be included. This
may be the differential form of a classical Aharonov−Bohm effect.
The only term that does not change between the two field equations is the one involving the
dynamic gravitational force. This might be a clue for why this force is only attractive.
Solutions to the Unified Field Equations
All the solutions that have been worked out for the Maxwell equations will work with the unified
field equations. For example, if the potential is static, the scalar equation for hyperbolic field
equation is the Poisson equation. The unified equations are more informative, since any potential
which is a solution to the scalar Poisson equation will also characterize the corresponding current.
The field equations of general relativity and the Maxwell equations both have vacuum solutions,
such as plane wave solutions. The unified field equations do not have such a solution, other than a
constant. Given historical tradition, this may seem like a deadly flaw. However, it may be some-
thing that is required for a final and complete theory. In a unified field theory, the gravitational
part may be zero while the electrical part is not, and visa versa. Non−zero solutions are worth
exploring
An inverse square potential plays an important role in both gravity and electromagnetism. Examine
the scalar field involving the inverse interval squared:
..
-
¸
1
------------------------------------
t
2
- x
2
- y
2
- z
2
, 0

¸ =

4 (3 t
2
+ x
2
+ y
2
+ z
2
)
-------------------------------------------------
(t
2
- x
2
- y
2
- z
2
)
3
, 0
|

110

This potential solves the Maxwell equations in the Lorentz gauge:
.
2
¸
1
------------------------------------
t
2
- x
2
- y
2
- z
2
, 0

¸ = 0
The non−zero part may have everything to do with gravity.
A plane wave solution does exist, but not for a pure vacuum. Instead, a plane wave solution exists
with the constrain that the net current is zero for the elliptical field equations
The field equations of general relativity and the Maxwell equations both have vacuum solutions. A
vacuum solution for the unified field equation is apparent for the elliptical field equations:
A = ¡Φ
0
e
K

R

-Ωt
, A

0
e
K

R

-Ωt
¡
The unified field equation will evaluate to zero if
Scalar ¸¡

----
c
, K

¡
2
¸ = 0
The dispersion relation is an inverted distance, so it will depend on the metric. The same potential
can also solve the hyperbolic field equations under different constraints and resulting dispersion
equation (not shown). There were two reasons for not including the customary imaginary number
"i" in the exponential of the potential. First, it was not necessary. Second, it would have created a
complex−valued quaternion, and therefore is outside the domain of my work. The important thing
to realize is that vacuum solutions to the unified field equations exist whose dispersion equations
depend on the metric. This is an indication that unifying gravity and electromagnetism is an appro-
priate goal.
Conservation Laws
Conservation of electric charge is implicit in the Maxwell equations. Is there also a conserved
quantity for the gravitational field? Examine how the differential operator acts on the unified field
equation:
..
-
.
-
A
-
=

o
2
g
----------
ot
2
+ V

V

g,
o
2
E

----------
ot
2
+ V

2
E

+
o
2
B

----------
ot
2
+ V

2
B
|

Notice that the gravitational force field only appears in the quaternion scalar. The electromagnetic
fields only appear in the 3−vector. This generates two types of constraints on the sources. No
change in the electric source applies to the quaternion scalar. No change in the gravitational source
applies to the 3−vector.
Scalar (.J
e
-
) =

e
----------
ot
+ V

J

e
= 0
Scalar ¡.J

AB
-
¡ = V J

AB
-
= 0
Vector (.J
m
-
) = -
oJ

m
----------
ot
+ V

Ρ
m
- V

xJ

m
= 0

The first equation is known as the continuity equation, and is the reason that electric charge is
conserved. For a different inertial observer, this will appear as a conservation of electric current
density. There is no source term for the Aharanov−Bohm current, and subsequently no conserva-
tion law. The 3−vector equation is a constraint on the mass current density, and is the reason mass
current density is conserved. For a different inertial observer, the mass density is conserved.
111

The first equation is known as the continuity equation, and is the reason that electric charge is
conserved. For a different inertial observer, this will appear as a conservation of electric current
density. There is no source term for the Aharanov−Bohm current, and subsequently no conserva-
tion law. The 3−vector equation is a constraint on the mass current density, and is the reason mass
current density is conserved. For a different inertial observer, the mass density is conserved.
Transformations of the Unified Force Field
The transformation properties of the unified field promise to be more intricate than either gravity or
electromagnetism separately. What might be expected to happen under a Lorentz transformation?
Gravity involves mass that is Lorentz invariant, so the field that generates it should be Lorentz
invariant. The electromagnetic field is Lorentz covariant. However, a transformation cannot do
both perfectly. The reason is that a Lorentz transformation mixes a quaternion scalar with a 3−vec-
tor. If a transformation left the quaternion scalar invariant and the 3−vector covariant, the two
would effectively not mix. The effect of unification must be subtle, since the transformation proper-
ties are well known experimentally.
Consider a boost along the x−axis. The gravitational force field is Lorentz invariant. All the terms
required to make the electromagnetic field covariant under a Lorentz transformation are present,
but covariance of the electromagnetic fields requires the following residual terms:
(.
’-
A
’-
)
Residual
= ¸0, (Γ
2
Β
2
- 1)
oA
x
----------
ot
+ (Γ
2
- 1)

--------
ox
,
-2 ΓΒ ¸
oA
z
----------
ot
+
oA
y
----------
ox
¸, 2 ΓΒ ¸
oA
y
----------
ot
-
oA
z
----------
ox
¸¸.
At this time, the correct interpretation of the residual term is unclear. Most importantly, it was
shown earlier that charge is conserved. These terms could be a velocity−dependent phase factor.
If so, it might provide a test for the theory.
The mechanics of the Lorentz transformation itself might require careful re−examination when so
strictly confined to quaternion algebra. For a boost along the x−axis, if only the differential transfor-
mation is in the opposite direction, then the electromagnetic field is Lorentz covariant with the
residual term residing with the gravitational field. The meaning of this observation is even less
clear. Only relatively recently has DeLeo been able to represent the Lorentz group using real
quaternions (S. De Leo, "Quaternions and special relativity," J. Math. Phys., 37(6):2955−2968,
1996). The delay appears odd since the interval of special relativity is the scalar of the square of
the difference between two events. In the real 4x4 matrix representation, the interval is a quarter of
the trace of the square. Therefore, any matrix with a trace of one that does not distort the length of
the scalar and 3−vector can multiply a quaternion without effecting the interval. One such class is
3−dimensional, spatial rotations. An operator that adds nothing to the trace but distorts the lengths
of the scalar and 3−vector with the constraint that the difference in lengths is constant will also
suffice. These are boosts in an inertial reference frame. Boosts plus rotations form the Lorentz
group.
Three types of gauge transformations will be investigated: a scalar, a 3−vector, and a quaternion
gauge field. Consider an arbitrary scalar field transformation of the potential:
A - A

= A - .
-
Λ.
The electromagnetic fields are invariant under this transformation. An additional constraint on the
gauge field is required to leave the gravitational force field invariant, namely that the scalar gauge
field solves a homogeneous elliptical equation. From the perspective of this proposal, the freedom
to choose a scalar gauge field for the Maxwell equations is due to the omission of the gravitational
force field.
Transform the potential with an arbitrary 3−vector field:
112

A - A

= A - .
-
A

.
This time the gravitational force field is invariant under a 3−vector gauge field transformation.
Additional constraints can be placed on the 3−vector gauge field to preserve a chosen electromag-
netic invariant. For example, if the difference between the two electromagnetic fields is to remain
invariant, then the 3−vector gauge field must be the solution to an elliptical equation. Other classes
of invariants could be examined.
The scalar and 3−vector gauge fields could be combined to form a quaternion gauge field. This
gauge transformation would have the same constraints as those above to leave the fields invariant.
Is there any such gauge field? The quaternion gauge field can be represented the following way:
A - A

= A - .
-
A.
If a force field is created by hitting this gauge transformation with a differential operator, then the
gauge field becomes a unified field equation. Since vacuum solutions have been found for those
equations, a quaternion gauge transformation can leave the field invariant.
Future Directions
The fields of gravity and electromagnetism were unified in a way consistent with Einstein’s vision,
not his technique. The guiding principles were simple but unusual: generate expressions familiar
from electromagnetism using quaternions, striving to interpret any extra terms as being due to
gravity. The first hypothesis about the unified field involved only a quaternion differential opera-
tor acting on a potential, no extra terms added by hand. It contained the typical potential representa-
tion of the electromagnetic field, along with a symmetric−matrix force field for gravity. The sec-
ond hypothesis concerned a unified field equation formed by acting on the unified field with one
more differential operator. All the Maxwell equations are included explicitly or implicitly. Addi-
tional terms suggested the inclusion of a classical representation of the Aharanov−Bohm effect.
Four linearly independent unified field equations exist, but only the hyperbolic and elliptic cases
were discussed. A large family of vacuum solutions exists, and will require future analysis to
appreciate.
Why did this approach work? The hypothesis that initiated this line of research was that all events
in spacetime could be represented by quaternions, no matter how the events were generated. This
is a broad hypothesis, attempting to reach all areas covered by physics. Based on the equations
presented in this paper, a logical structure can be constructed, starting from events (see figure
below). A set of events forms a pattern that can be described by a potential. The change in a poten-
tial creates a field. The change in field creates a field equation. The terms that do not change under
differentiation of a field equation form conservation laws.
113

Acknowledgements
Thanks for many productive discussions Prof. Guido Sandri.
114

Einstein’s Vision II: A Unified Force Equation with Constant Velocity
Solutions
Abstract
In quantum electrodynamics, photons have four modes of transmission, at least mathematically:
two transverse modes for electrodynamics, a longitudinal, and a scalar mode. The probabilities of
the last two modes cancel each other out for a spin 1 field, but that does not have to be the case for
a field with spin 1 and spin 2 particles. One potential solution to the field equations is found which
depend on the inverse of an interval between two events squared. The force field created by the
potential is constructed by comparison with the classical Newtonian gravitational field. The
Lagrange density L= scalar(−(J A*) − 1/2 Box* A Box A*) can contribute to a scalar mode, but
still has the field equations of Maxwell with the choice of the Lorenz gauge. A relativistic force
equation is proposed, created by the product of charge, normalized force field, and 4−velocity:
dmU/dtau = kq Box* A/|A| U*. The solution to the force equation using the inverse square interval
potential is found. Eliminating the constants generates a metric equation,

2
e
2
G M

c
2
Τ
e
2
G M

c
2
Τ
d R
2
c
2
, where tau is a lightlike interval with almost the same magnitude as
the radius R of separation between source and test masses. For a weak gravitational field, the
metric will pass the same tests as the Schwarzschild metric of general relativity. The two metrics
differ for higher order terms, which makes the proposed metric distinct and testable experimen-
tally. A constant−velocity solution exists for the gravitational force equation for a system with an
exponentially−decaying mass distribution. The dark matter hypothesis is not needed to explain the
constant−velocity profiles seen for some galaxies. Gravity is a metric theory, electromagnetism is
not. By using Riemannian quaternions which can have dynamic basis vectors, it becomes possible
to merge metric theory with the linear Maxwell equations. The proposal may also have implica-
tions for classical big bang theory.
An Opportunity for Classical Gravity?
The electrodynamic field can be quantized in a manifestly covariant form by fixing the gauge (K.
Bleuler, Helv. Phys. Acta, 23:567, 1950, and S. N. Gupta, "Theory of longitudinal photons in
quantum electrodynamics", Proc. Phys. Soc., 63:681−691). The starting point is the 4−potential
A^mu. There are four modes of transmission for photons corresponding to the four degrees of
freedom: two transverse, one scalar, and one longitudinal. Gupta calculated that "the probability of
the emission of a real longitudinal photon is canceled by the ’negative probability’ of the emission
of a corresponding scalar photon." He notes that this does not always have to be the case for the
nonhomogeneous Maxwell equations, which is the focus of this work. A scalar photon would not
change signs under a space or time reversal, so its symmetry is different from the electric 3−vector
field and the magnetic 3−pseudo−vector field, and thus does not have an obvious role to play in
electrodynamics.
My hypothesis is that the scalar and longitudinal photons for the electromagnetic field constitute
gravity. The hypothesis makes several predictions even at this preliminary stage. First, the math of
gravity and electromagnetism should be similar but not identical. The inverse square form of New-
ton’s law of gravity was a direct inspiration for Coulomb’s law. Gravity should be more symmetric
than electromagnetism because the mode is scalar, instead of transverse. The second rank field
strength tensor in general relativity is symmetric while the analogous tensor for the electromagnetic
field is antisymmetric. Since the mode of gravity is orthogonal to electromagnetism, the charges
can be likewise, so there will be no simple relationship between gravitational charge (mass) and
electric charge. Gravitational waves in general relativity are transverse, so this proposal is distinct
from general relativity. Nature exploits all the math available, so it is unreasonable to suppose that
scalar and longitudinal photons are never used for anything. Whatever phenomenon exploits the
scalar and longitudinal photons must be similar, but just as important as electromagnetism. Gravity
is a natural candidate.
115

My hypothesis is that the scalar and longitudinal photons for the electromagnetic field constitute
gravity. The hypothesis makes several predictions even at this preliminary stage. First, the math of
gravity and electromagnetism should be similar but not identical. The inverse square form of New-
ton’s law of gravity was a direct inspiration for Coulomb’s law. Gravity should be more symmetric
than electromagnetism because the mode is scalar, instead of transverse. The second rank field
strength tensor in general relativity is symmetric while the analogous tensor for the electromagnetic
field is antisymmetric. Since the mode of gravity is orthogonal to electromagnetism, the charges
can be likewise, so there will be no simple relationship between gravitational charge (mass) and
electric charge. Gravitational waves in general relativity are transverse, so this proposal is distinct
from general relativity. Nature exploits all the math available, so it is unreasonable to suppose that
scalar and longitudinal photons are never used for anything. Whatever phenomenon exploits the
scalar and longitudinal photons must be similar, but just as important as electromagnetism. Gravity
is a natural candidate.
A Gravitational Field Inside Maxwell
Newton’s classical gravitational law arises from a scalar potential. Here is the scalar field equation:
V
2
Φ = 4 ΠGΡ
For the case of a vacuum, when rho = 0, this is known as the Laplace equation. For a spherically
symmetric source, one solution is:
Φ = -
GM
-------------------------------
_
x
2
+ y
2
+ z
2
The problem with the field equation is that the Laplace operator does not have a time differential
operator. Any change in the mass density propagates at infinite speed, in conflict with special
relativity (MTW, chapter 7). One way to derive the field equations of general relativity involves
making Newton’s law of gravity consistent with the finite speed of light.
A way to repair the field equations is to use the D’Alembertian operator, which is four dimen-
sional. That expression is identical to the A^0 component of the Maxwell equations with the choice
of the Lorenz gauge. The sources are of course different. Yet the argument being made here is
that there are degrees of freedom which have yet to be exploited. For the two degrees of freedom,
we can have a different source term, mass:
.
2
A = 4 Π (J
q
- J
m
)
If one is studying scalar or longitudinal modes, the source is J_mass, the mass current density. If
one is working with transverse modes, the source is J_charge^mu , the electric charge density.
Since the modes are orthogonal, the sources can be also.
To be consistent with the classic scalar potential yet still be relativistic, the potential must have
x^2, y^2, z^2, and t^2. This suggests a particular solution to the field equations:
A = ¸
1
-------------------------------------------
c
2
t
2
- x
2
- y
2
- z
2
, 0

¸ = ¸
1
-------
Τ
2
, 0

¸
This potential is interesting for several reasons. It is the inverse of the Lorentz−invariant interval
squared. Like mass, the 4−potential will not be altered by a change in an inertial reference frame.
The interval between any two events will contribute to the potential. General relativity applies to
any form of energy, including gravitational field energy. A potential that embraces every interval
may have a broad enough scope to do the work of gravity.
The potential also has serious problems. Classical gravity depends on an inverse square force field,
not an inverse square potential. Taking the derivative of the potential puts a forth power of the
interval in the denominator. At this point, I could stop and say that this potential has nothing to do
with gravity because it has the wrong dependence on distance. An alternative is to look for an
algebraic way to repair the problem. This is the type of approach used by the early workers in
quantum mechanics like de Broglie, and will be adopted here. The equations of motion can be
normalized to the magnitude of the 4−potential:
116

The potential also has serious problems. Classical gravity depends on an inverse square force field,
not an inverse square potential. Taking the derivative of the potential puts a forth power of the
interval in the denominator. At this point, I could stop and say that this potential has nothing to do
with gravity because it has the wrong dependence on distance. An alternative is to look for an
algebraic way to repair the problem. This is the type of approach used by the early workers in
quantum mechanics like de Broglie, and will be adopted here. The equations of motion can be
normalized to the magnitude of the 4−potential:
.
2
A
-------------
| A |
= 4 Π (kJ
charge
+ GJ
mass
)
Since the magnitude of the potential is the inverse interval squared, the resulting equation has only
an interval squared in the denominator. An interval is not necessarily the same as the distance R
between the source and test mass used in the classical theory. However, I can impose a selection
rule that in the classical limit, the only events that contribute to the potential are those that are
timelike separated between the source and the test masses. It takes a timelike interval to know that
the source is a distance R away. Action−at−a−distance respects the speed of light as it must.
Search for the Source Mass
Where is the source mass in the potential? All that has been discussed so far is an interval, a dis-
tance, nothing about mass. An idea from general relativity will be borrowed, that mass can be
treated geometrically if multiplied by the constants G/c^2. The distance between the Earth and the
Sun is approximately 1.5x10^11 m, while the Sun’s mass expressed in units of distance, GM_-
Sun/c^2, is 1.5x10^3 m, eight orders of magnitude smaller. The overall length of the interval will
not be changed noticeably if the spatial separation and the Sun’s mass expressed as a distance are
summed. However, the force field is the derivative of the potential, and any change in position in
spacetime will have a far greater effect proportionally on the smaller geometric mass than the
spatial separation. Make the following change of variables:
t - t’ = A +
GM
--------------
2 c
2
A
t
R

- R

’ = B

+
GM
----------------------
2 c
2
| B

|
R

where A and B are locally constants such that tau^2 ~= A^2 − B^2. The change of variables is
valid locally, but not globally, since it breaks down for arbitrarily long time or distance away.
General relativity is also valid locally and not globally.
What is the physical interpretation of the inverse square potential and the above substitution?
Newton observed that motion in an ellipse could be caused by either a linear central force or an
inverse square law. With the above substitution, there is a linear displacement equation inside an
inverse square potential. It is like a simple harmonic oscillator inside a simple harmonic oscillator!
This oscillator works with four dimensions. Although it is confusing to confront the idea of oscilla-
tions in time, there is no need worry about it, since the equations are quite simple and their mathe-
matical consequences can be worked out. If all the terms where included, the equation would be
nonlinear.
The field is the derivative of the potential. To be correct technically, it is the contravariant deriva-
tive. This requires both a metric and a connection. In effect, all the work presented with quater-
nions uses the Minkowski metric with Cartesian coordinates. For such a choice of metric and
coordinates, the contravariant derivative equals the normal derivative. The derivative of the poten-
tial under study, a normalized interval squared with the linear displacement substitution, is
approximately:
1
-----------------
|
1
-----
Τ
2
|

o
1
-----
Τ
2
-----------
ot
= -
GM
-------------
c
2
Τ
2
117

1
-----------------
|
1
-----
Τ
2
|

o
1
-----
Τ
2
-----------
oR

=
GM
-------------
c
2
Τ
2
This should look familiar, remembering that employing the event selection rule from above, the
magnitude of tau ^2 is almost the same as R^2, differing only by the geometric mass of the source.
A Lagrangian for Four Modes
Despite its formulation using quaternions, this unification proposal is strikingly similar to earlier
work. Gupta wanted to quantize the radiation field using a form that was manifestly covariant in
its explicit treatment of time and space. He fixed the gauge with this Lagrange density:
L = -J
Μ
A
Μ
-
1
----
2
(o
Μ
A
Μ
)
2
-
1
----
4
(o
Μ
A
Ν
- o
Ν
A
Μ
) (o
Μ
A
Ν
- o
Ν
A
Μ
)
The equations of motion for this Lagrangian are the same as choosing the Lorenz gauge:
.
2
A
Μ
= J
Μ
The problem with the Lagrangian is that the field strength tensor is antisymmetric. Due to the zeros
along the diagonal, it cannot contribute directly to a scalar mode. What is needed is a Lagrange
density that could contribute directly to the scalar mode but still have the same field equations.
Here is such a Lagrangian:
L = scalar ¸-(J
q
- J
m
) A
-
-
1
----
2
.
-
A .A
-
¸
This is not as miraculous as it might first appear. It is the first of four terms generated in the contrac-
tion of the electromagnetic field strength tensor. In essence, information is not discarded, which is
what happens in making the field strength tensor antisymmetric. The one remaining modification is
to normalize both the Lagrangian and equations of motion to the size of the potential.
From a Relativistic 4−force to a Metric
A relativistic 4−force is the change in momentum with respect to the interval. The covariant force
law is similar in form to the one for electromagnetism except that the second rank tensor is asym-
metric and normalized:
F =
op
--------

= mc

--------

+ Βc
om
--------

= k q
.
-
A
-
-------------
| A |
Β
-
If this equation is to transform like the Lorentz 4−force of electromagnetism, the normalized poten-
tial must be invariant under a Lorentz transformation. That is the case of the potential under study.
In the first application of the force law, assume the derivative of the mass with respect to the inter-
val is zero. For the scalar photons, assume the charge q is the gravitational test mass. Experiments
have demonstrated that gravitational and inertial masses are equal. Assuming spherical symmetry,
the inverse interval squared potential leads to the following equations of motion:

o
2
t
----------

2
+
G M
-------------
c
2
Τ
2

ot
--------

,
o
2
R

----------

2
-
G M
-------------
c
2
Τ
2

oR

--------

|

= ¡0, 0

¡
Solve these second−order differential equations for the spacetime position:
118

t = c
1
¡Τe
GM
---------
c
2
Τ
-
GM
-------
c
2
Ei ¡
GM
-----------
c
2
Τ
¡¡ + C
2
R

= C

1
¡Τe
-
GM
---------
c
2
Τ
+
GM
-------
c
2
Ei ¡-
GM
-----------
c
2
Τ
¡¡ + C

2
where Ei is the exponential integral, Ei(t)=the integral from negative infinity to t of e^t/t dt. The
exponential integral plays other roles in quantum mechanics, so its presence is interesting.
Eight constants need to be eliminated:(c_1, C_1) and (c_2, C_2). Take the derivative of the space-
time position with respect to tau. This eliminates four constants, (c_2, C_2). The result is a
4−velocity:
ot
--------

= c
1
e
G M
------------
c
2
Τ
oR

--------

= C

1
e
-
G M
------------
c
2
Τ
In flat spacetime, beta _mu beta^mu=1, providing four more constraints. Spacetime is flat if M
goes to 0 or tau goes to infinity, leading to e^(GM/c^2|tau |) goes to 1:
¸
ot
--------

¸
2
-

oR

--------

|

oR

--------

|

= c
1
2
- C

1
C

1
= 1
Solve for c_1^2 and C_1.C_1:
c
1
2
= e
-
G M
------------
c
2
Τ

ot
--------

C

1
C

1
= e
G M
------------
c
2
Τ

oR

--------

Substitute back into the flat spacetime constraint. Rearrange into a metric:

2
= e
-2
G M
------------
c
3
Τ
ot
2
- e
2
G M
------------
c
3
Τ
oR

2
If the gravitational field is zero, this generates the Minkowski metric of flat spacetime. Con-
versely, if the gravitational field is non−zero, spacetime is curved
As expected, this become the Minkowski metric for flat spacetime if M goes to 0 or tau goes to
infinity.
No formal connection between this proposal and curvature has been established. Instead a path
between a proposed gravitational force equation and a metric function was sketched. There is a
historical precedence for the line of logic followed. Sir Isaac Newton in the Principia showed an
important link between forces linear in position and inverse square force laws. More modern
efforts have shown that the reason for the connection is due to the conformal mapping of z goes to
z^2 (T. Needham, "Newton and the transmutation of force," Amer. Math. Mon., 100:119−137,
1993). This method was adapted to a quaternion force law linear in the relativistic velocity to
generate a metric.
For a weak field, write the Taylor series expansion in terms of the total mass over the interval to
second−order in M/|tau|:
119

2
= ¸1 - 2
G M
-----------
c
2
Τ
+ 2 ¡
G M
-----------
c
2
Τ
¡
2
¸ ot
2
-
-¸1 + 2
G M
-----------
c
2
Τ
+ 2 ¡
G M
-----------
c
2
Τ
¡
2
¸ oR

2
+ O ¸¡
G M
-----------
c
2
Τ
¡
3
¸
Contrast this with the Schwarzschild solution in isotropic coordinates expanded to second order in
M/R (MTW, eq. 31.22):

2
= ¸1 - 2
G M
-----------
c
2
R
+ 2 ¡
G M
-----------
c
2
R
¡
2
¸ ot
2
-
-¸1 + 2
G M
-----------
c
2
R
+ 2.5 ¡
G M
-----------
c
2
R
¡
2
¸ oR

2
+ O ¸¡
G M
-----------
c
2
R
¡
3
¸
The magnitude of the lightlike interval tau in the unified field metric is nearly identical to the
radius R in the Schwarzschild metric, the difference being the geometric mass of the source
included in the interval tau . The metric for the scalar potential will pass the same weak field tests
of general relativity as the Schwarzschild metric to post−Newtonian accuracy, which does not use
the second order spatial term. The difference in the higher order terms can be the basis of an experi-
mental test to distinguish this proposal from general relativity. Since the effect is second order in
the field strength, such a test will challenge experimental techniques.
The two metrics are numerically very similar for weak fields, but mathematically distinct. For
example, the Schwarzschild metric is static, but the unified metric contains a dependence on time
so is dynamic. The Schwarzschild metric has a singularity at R=0. The unified gravitational force
metric becomes undefined for lightlike intervals. This might pose less of a conceptual problem,
since light has no rest mass.
The Constant Velocity Profile Solution
In the previous section, the system had a constant point−source mass with a velocity profile that
decayed with distance. Here the opposite situation is examined, where the velocity profile is a
constant, but the mass distribution decays with distance. Expand the definition of the relativistic
force using the chain rule:
c
om Β
-----------

= m c

--------

+ Β c
om
--------

The first term of the force is the one that leads to an approximation of the Schwarzschild metric,
and by extension, Newton’s law of gravity. For a region of spacetime where the velocity is con-
stant, this term is zero. In that region, gravity’s effect is on the distribution of mass over spacetime.
This new gravitational term is not due to the unified field proposal per se. It is more in keeping
with the principles underlying relativity, looking for changes in all components, in this case mass
distribution with respect to spacetime.
Start with the gravitational force in a region of spacetime with no velocity change:
Β c
om
i
----------

= k m
g
Scalar (.
-
A
-
) Β
-
Make the same assumptions as before: the gravitational mass is equal to the inertial mass and the
gravitational field employs the interval between the worldlines of the test and gravitational masses.
This generates an equation for the distribution of mass:
120

Γ
om
--------

+
Γ G M
--------------------
c
2
| Τ |
2
m, Γ Β

om
--------

-
Γ Β

G M
--------------------
c
2
| Τ |
2
m
|

= ¡0, 0

¡
Solve for the mass flow:
¡Γ m, Γ Β

m¡ = ¡c e
G M
---------------
c
2
|Τ|
, C

e
-
G M
---------------
c
2
|Τ|
¡
As in the previous example for a classical weak field, assume the magnitude of the interval is an
excellent approximation to the radius divided by the speed of light. The velocity is a constant, so it
is the mass distribution that shows an exponential decay with respect to the interval, which is
numerically no different from the radius over the speed of light. This is a stable solution. If the
mass keeps dropping of exponentially, the velocity profile will remain constant
Look at the problem in reverse. The distribution of matter has an exponential decay with distance
from the center. It must solve a differential equation with the velocity constant over that region of
spacetime like the one proposed.
The exponential decay of the mass of a disk galaxy is only one solution to this expanded gravita-
tional force equation. The behavior of larger systems, such as gravitational lensing caused by
clusters, cannot be explained by the Newtonian term (A. G. Bergmann, V. Petrosian, and R. Lynds,
"Gravitational lens images of arcs in clusters," Astrophys. J., 350:23, 1990. S. A. Grossman and R.
Narayan, "Gravitationally lensed images in abell 370," Astrophys. J., 344−637−644, 1989. J. A.
Tyson, F. Valdes, and R. A. Wenk, "Detection of systematic gravitational lens galaxy image align-
ments: Mapping dark matter in galaxy clusters," Astrophys. J. Let., 349:L1, 1990). It will remain
to be seen if this proposal is sufficient to work on that scale.
Metrics and Forces
Gravity was first described as a force by Isaac Newton. In general relativity, Albert Einstein
argued that gravity was not a force at all. Rather, gravity was Riemannian geometry, curvature of
spacetime caused by the presence of a mass−energy density. Electromagnetism was first described
as a force, modeled on gravity. That remains a valid choice today. However, electromagnetism
cannot be depicted in purely geometric terms. A conceptual gap exists between purely geometrical
and force laws.
The general equivalence principle, introduced in the first paper of this series, places geometry and
force potentials on equal footing. Riemannian quaternions, (a_0 i_0, a_1 i_1/3, a_2 i_2/3, a_3
i_3/3), has pairs of (possibly) dynamic terms for the 4−potential A and the 4−basis I. Gauss’ law
written with Riemannian quaternion potentials and operators leads to this expression:
-
i

n
2
---------
9

oE
n
----------
oi
n
-
i

n
E
n
-------------
9

oi

n
----------
oi
n
= 4 ΠΡ, n = 1, 2, 3
If the divergence of the electric field E was zero, then Gauss’ law would be due entirely to the
divergence of the basis vectors. The reverse case could also hold. Any law of electrodynamics
written with Riemannian quaternions is a combination of changes in potentials and/or basis vectors.
121

Future Directions
An algebraic path between a solution to the Maxwell equations and a classical metric gravitational
theory has been shown. No effort has been extended yet to quantize the unification proposal. Like
the early work in quantum mechanics, a collection of hunches is used to connect equations. One is
left with the question of why this might work? The action of a gauge invariant theory cannot be
inverted to generate the propagator needed for quantum mechanics. Fixing the gauge makes the
action invertible, but the additional constraint decreases the degrees of freedom. By using quater-
nions, a division algebra, the equation is necessarily invertible without imposing a constraint. If the
operation of multiplication surpasses what can be done with division, then Nature cannot harness
the most robust mathematical structure, a topological algebraic field, the foundation for doing
calculus. Nature does calculus in four dimensions, and it is this requirement that fixes the gauge. In
the future, when we understand how to do calculus with four−dimensional automorphic functions,
we may have a deep appreciation of Nature’s methods.
There is a physical explanation for gravity − it is a local, nonlinear, four−dimensional simple
harmonic oscillator. Gravity is all about oscillations. The Earth returns to approximately the same
place after one year of travel. If there were no interfering matter in the way, an apple dropped
would fall to the center of the Earth, reach the other side, and return in a little over eighty minutes.
The metric equation that results from this analysis is within the experimental constraints of current
tests of general relativity. That makes the proposal reasonable. For higher order terms of a weak
field, the proposal is different than the Schwarzschild metric of general relativity. That makes it
testable. There are very few reasonable, testable classical unified field theories in physics, so this
alone should spark interest in this line of work.
For a spiral galaxy with an exponential mass distribution, dark matter is no longer needed to
explain the flat velocity profile observed or the long term stability of the disk. Mass distributed
over large distances of space has an effect on the mass distribution itself. This raises an interesting
question: is there also an effect of mass distributed over large amounts of time? If the answer is
yes, then this might solve two analogous riddles involving large time scales, flat velocity profiles
and the stability of solutions. Classical big bang cosmology theory spans the largest time frame
possible and faces two such issues. The horizon problem involves the extremely consistent veloc-
ity profile across parts of the Universe that are not casually linked (MTW, p. 815). The flatness
problem indicates how unstable the classical big bang theory is, requiring exceptional fine tuning
to avoid collapse. Considerable effort will be required to substantiate this tenuous hypothesis.
Any insight into the origin of the unified engine driving the Universe of gravity and light is
worthwhile.
122

Strings and Quantum Gravity
In this section, a quaternion 3−string will be defined. By making this quantity dimensionless, I will
argue that it my be involved in a relativistic quantum gravity theory, at least one consistent with
current experimental tests. At the current time, this is an idea in progress, not a theory, since the
equations of motion have not been determined. It is hoped that the work in the previous section on
unified fields will provide that someday.
Strings
Let us revisit the difference between two quaternions squared, as worked out in the section of
analysis. A quaternion has 4 degrees of freedom, so it can be represented by 4 real numbers:
q = (a
0
, a
1
, a
2
, a
3
)
Taking the difference between two quaternions is only a valid operation if they share the same
basis. Work with defining the derivative with respect to a quaternion has required that a change in
the scalar be equal in magnitude to the sum of changes in the 3−vector (instead of the usual parity
with components). These concerns lead to the definition of the difference between two quaternions:
dq = ¡da
0
e
0
, da
1

e
1
-------
3
, da
2

e
2
-------
3
, da
3

e
3
-------
3
¡
What type of information must e0, e1, e2, and e3 share in order to make subtraction a valid opera-
tion? There is only one basis, so the two events that make up the difference must necessarily be
expressed in the same basis. If not, then the standard coordinate transformation needs to be done
first. A more subtle issue is that the difference must have the same amount of intrinsic curvature
for all three spatial basis vectors. If this is not the case, then it would not longer be possible to do a
coordinate transformation using the typical methods. There would be a hidden bump in an other-
wise smooth transformation! At this point, I do not yet understand the technical link between basis
vectors and intrinsic curvature. I will propose the following relationship between basis vectors
because its form suggests a link to intrinsic curvature:
-
1
---------
e
1
2
= -
1
---------
e
2
2
= -
1
---------
e
3
2
= e
0
2
If e0 = 1, this is consistent with Hamilton’s system for 1, i, j, and k. The dimensions for the spatial
part are 1/distance^2, the same as intrinsic curvature. This is a flat space, so −1/e1^2 is something
like 1 + k. In effect, I am trying to merge the basis vectors of quaternions with tools from topol-
ogy. In math, I am free to define things as I choose, and if lucky, it will prove useful later on :−)
Form the square of the difference between two quaternion events as defined above:
dq
2
= ¸da
0
2
e
0
2
+ da
1
2

e
1
2
---------
9
+ da
2
2

e
2
2
---------
9
+ da
3
2

e
3
2
---------
9
,
2 da
0
da
1
e
0

e
1
-------
3
, 2 da
0
da
2
e
0

e
2
-------
3
, 2 da
0
da
3
e
0

e
3
-------
3
¸ =
123

= (interval
2
, 3 - string)
The scalar is the Lorentz invariant interval of special relativity if e0 = 1.
Why use a work with a powerful meaning in the current physics lexicon for the vector dt dX? A
string transforms differently than a spatial 3−vector, the former flipping signs with time, the latter
inert. A string will also transform differently under a Lorentz transformation.
The units for a string are time*distance. For a string between two events that have the same spatial
location, dX = 0, so the string dt dX is zero. For a string between two events that are simultaneous,
dt = 0 so the string is again of zero length. Only if two events happen at different times in different
locations will the string be non−zero. Since a string is not invariant under a Lorentz transforma-
tion, the value of a string is
We all appreciate the critical role played by the 3−velocity, which is the ratio of dX by dt. Hope-
fully we can imagine another role as important for the product of these same two numbers.
Dimensionless Strings
Imagine some system that happens to create a periodic pattern of intervals and strings (a series of
events that when you took the difference between neighboring events and squared them, the results
had a periodic pattern). It could happen :−) One might be able to use a collection of sines and
cosines to regenerate the pattern, since sines and cosines can do that sort of work. However, the
differences would have to first be made dimensionless, since the infinite series expansion for such
transcendental functions would not make sense. The first step is to get all the units to be the same,
using c. Let a0 have units of time, and a1, a2, a3 have units of space. Make all components have
units of time:
dq
2
= ¸da
0
2
e
0
2
+ da
1
2

e
1
2
-----------
9 c
2
+ da
2
2

e
2
2
-----------
9 c
2
+ da
3
2

e
3
2
-----------
9 c
2
,
2 da
0
da
1
e
0

e
1
--------
3 c
, 2 da
0
da
2
e
0

e
2
--------
3 c
, 2 da
0
da
3
e
0

e
3
--------
3 c
¸
Now the units are time squared. Use a combination of 3 constants to do the work of making this
dimensionless.
1
----
G
-
mass time
2
--------------------
distance
3
1
----
h
-
time
---------------------------
mass distance
2
c
5
-
distance
5
-------------------
time
5
The units for the product of these three numbers are the reciprocal of time squared. This is the
same as the reciprocal of the Planck time squared, and in units of seconds is 5.5x10^85s^−2. The
symbols needed to make the difference between two events dimensionless are simple:
dq
2
=
c
5
--------
G h
¸da
0
2
e
0
2
+ da
1
2

e
1
2
-----------
9 c
2
+ da
2
2

e
2
2
-----------
9 c
2
+ da
3
2

e
3
2
-----------
9 c
2
,
2 da
0
da
1
e
0

e
1
--------
3 c
, 2 da
0
da
2
e
0

e
2
--------
3 c
, 2 da
0
da
3
e
0

e
3
--------
3 c
¸
As far as the units are concerned, this is relativistic (c) quantum (h) gravity (G). Take this con-
stants to zero or infinity, and the difference of a quaternion blows up or disappears.
124

Behaving Like a Relativistic Quantum Gravity Theory
Although the units suggest a possible relativistic quantum gravity, it is more important to see that it
behaves like one. Since this unicorn of physics has never been seen I will present 4 cases which
will show that this equation behaves like that mysterious beast!
Consider a general transformation T that brings the difference between two events dq into dq’.
There are four cases for what can happen to the interval and the string between these two events
under this general transformation.
Case 1: Constant Intervals and Strings
T : dq - dq’ such that scalar (dq
2
) =
scalar (dq’
2
) and vector (dq
2
) = vector (dq’
2
)
This looks simple, but there is no handle on the overall sign of the 4−dimensional quaternion, a
smoke signal of O(4). Quantum mechanics is constructed around dealing with phase ambiguity in
a rigorous way. This issue of ambiguous phases is true for all four of these cases.
Case 2: Constant Intervals
T : dq - dq’ such that scalar (dq
2
) =
scalar (dq’
2
) and vector (dq
2
) = vector (dq’
2
)
Case 2 involves conserving the Lorentz invariant interval, or special relativity. Strings change
under such a transformation, and this can be used as a measure of the amount of change between
inertial reference frames.
Case 3: Constant Strings
T : dq - dq’ such that scalar (dq
2
) =
scalar (dq’
2
) and vector (dq
2
) = vector (dq’
2
)
Case 3 involves conserving the quaternion string, or general relativity. Intervals change under such
a transformation, and this can be used as a measure of the amount of change between non−inertial
reference frames. All that is required to make this simple but radical proposal consistent with
experimental tests of general relativity is the following:
1 - 2
G M
-----------
c
2
R
= -
1
---------
e
1
2
= -
1
---------
e
2
2
= -
1
---------
e
3
2
= e
0
2
The string, because it is the product of e0 e1, e0 e2, and e0 e3, will not be changed by this. The
phase of the string may change here, since this involves the root of the squared basis vectors. The
interval depends directly on the squares of the basis vectors (I think of this as being 1+/− the intrin-
sic curvature, but do not know if that is an accurate technical assessment). This particular value
regenerates the Schwarzschild solution of general relativity.
125

Case 4: No Constants
T : dq - dq’ such that scalar (dq
2
) =
scalar (dq’
2
) and vector (dq
2
) = vector (dq’
2
)
In this proposal, changes in the reference frame of an inertial observer are logically independent
from changing the mass density. The two effects can be measured separately. The change in the
length−time of the string will involve the inertial reference frame, and the change in the interval
will involve changes in the mass density.
At this time I do not know how to use the proposed unified field equations discussed earlier to
generate the basis vectors shown. This will involve determining the precise relationship between
intrinsic curvature and the quaternion basis vectors.
126

Answering Prima Facie Questions in Quantum Gravity Using
Quaternions
(Note: this was a post sent to the newsgroup sci.physics.research June 28, 1998)
Chris Isham’s paper "Prima Facie Questions in Quantum Gravity" (gr−qc/9310031, October,
1993) details the structure required of any approach to quantum gravity. I will use that paper as a
template for this section, noting the highlights (but please refer to this well−written paper for
details). Wherever appropriate, I will point out how using quaternions in quantum gravity fits
within this superstructure. I will argue that all the technical parts required are all ready part of
quaternion mathematics. These tools are required to calculate the smallest norm between two
worldlines, which may form a new road to quantum gravity.
What Is Quantum Gravity?
Isham sorts the approaches to quantum gravity into four groups. First, there is the classical
approach. This begins with Einstein’s general relativity. Systematically substitute self−adjoint
operators for classical terms like energy and momentum. This gets further subdivided into the
’canonical’ scheme where spacetime is split into time and space−−Ashtekar’s work−−and a covari-
ant formulation, which is believed to be perturbatively non−renormalizable.
The second approach takes quantum mechanics and transforms it into general relativity. Much less
effort has gone in this direction, but there has been work done by Haag.
The third angle has general relativity as the low energy limit of ideas based in conventional quan-
tum mechanics. Quantum gravity dominates the world on the scale of Plank time, length, or
energy, a place where only calculations can go. This is where superstring theory lives.
The fourth possibility involves a radical new perspective, where general relativity and quantum
mechanics are only different applications of the same mathematical structure. This would require a
major "retooling". People with the patience to have read many of my posts (even if not followed
:−) know this is the task facing work with quaternions. Replace the tools for doing special relativi-
ty−−4−vectors, metrics, tensors, and groups−−with quaternions that preserve the scalar of a
squared quaternion. Replace the tools for deriving the Maxwell equations−−4−potentials, metrics,
tensors, and groups−−by quaternion operators acting on quaternion potentials using combinations
of commutators and anticommutators. It remains to be shown whether quaternions also have the
structure required for a quantum gravity theory.
Why Do We Study Quantum Gravity?
Isham gives six reasons: the inability to calculate using perturbation theory a correction for general
relativity, singularities, quantum cosmology (particularly the Big Bang), Hawking radiation, unifica-
tion of particles, and the possibility of radical change. This last reason could be a lot of fun, and it
is the reason to read this post :−)
127

What Are Prima Facie Questions?
The first question raised by Isham is the relation between classical and quantum physics. Physics
with quaternions has a general guide. Consider two arbitrary quaternions, q and q’. The classical
distance between them is the interval.
¡(t, X

) - ¡t
·
, X
·

¡¡
2
= (dt
2
- d X

.d X

, 2 dt d X

)
This involves retooling, because the distance also includes a 3−vector. There is nothing inherently
wrong with this vector, and it certainly could be computed with standard tools. To be complete,
measure the difference between two quaternions with a quaternion containing the usual invariant
scalar interval and a covariant 3−vector. To distinguishing collections of events that are lightlike
separated where the interval is zero, use the 3−vector which can be unique. Never discard useful
information!
Quantum mechanics involves a Hilbert space. Quaternions can be used to form an inner−product
space. The norm of the difference between q and q’ is
¡(t, X

) - ¡t
·
, X
·

¡¡
-
¡(t, X

) - ¡t
·
, X
·

¡¡ =
¡dt
2
+ d X

.d X

, 0

¡
The norm can be used to build all the equipment expected of a Hilbert space, including the
Schwarz and triangle inequalities. The uncertainty principle can be derived in the same way as is
done with the complex−valued wave function.
I call q q’ a Grassman product (it has the cross product in it) and q* q’ the Euclidean product (it is a
Euclidean norm if q = q’). In general, classical physics involves Grassman products and quantum
mechanics involves Euclidean products of quaternions.
Isham moves from big questions to ones focused on quantum gravity. Which classical spacetime
concepts are needed? Which standard parts of quantum mechanics are needed? Should particles
be united? With quaternions, all these concepts are required, but the tools used to build them
morph and become unified under one algebraic umbrella.
Isham points out the difficulty of clearly marking a boundary between theories and fact. He writes:
"...what we call a ’fact’ does not exist without some theoretical schema for organizing experimental
and experiential data; and, conversely, in constructing a theory we inevitably impose some prior
idea of what we mean by a fact."
My structure is this: the description of events in spacetime using the topological algebraic field of
quaternions is physics.
Current Research Programs in Quantum Gravity
There is a list of current approaches to quantum gravity. This is solid a description of the family of
approaches being used, circa 1993. See the text for details.
Prima Facie Questions in Quantum Gravity
Isham is concerned with the form of these approaches. He writes:
"I mean (by background structure) the entire conceptual and structural framework within whose
language any particular approach is couched. Different approaches to quantum gravity differ signifi-
cantly in the frameworks they adopt, which causes no harm−−indeed the selection of such a frame-
work is an essential pre−requisite for theoretical research−−provided the choice is made
consciously."
128

"I mean (by background structure) the entire conceptual and structural framework within whose
language any particular approach is couched. Different approaches to quantum gravity differ signifi-
cantly in the frameworks they adopt, which causes no harm−−indeed the selection of such a frame-
work is an essential pre−requisite for theoretical research−−provided the choice is made
consciously."
My framework was stated explicitly above, but it literally does not appear on the radar screen of
this discussion of quantum gravity. Moments later comes this comment:
"In using real or complex numbers in quantum theory we are arguably making a prior assumption
about the continuum nature of space."
This statement makes a hidden assumption, that quaternions do not belong on a list that includes
real and complex numbers. Quaternions have the same continuum properties as the real and com-
plex numbers. The important distinction is that quaternions do not commute. This property is
shared by quantum mechanics so it should not banish quaternions from the list. The omission
reflects the history of work in the field, not the logic of the mathematical statement.
General relativity may force non−linearity into quantum theory, which require a change in the
formalism. It is easy to write non−linear quaternion functions. Near the end of this section I will
do that in an attempt to find the shortest norm in spacetime which happens to be non−linear.
Now we come to the part of the paper that got me really excited! Isham described all the machin-
ery needed for classical general relativity. The properties of quaternions dovetail the needs per-
fectly. I will quote at length, since this is helpful for anyone trying to get a handle on the nature of
general relativity.
"The mathematical model of spacetime used in classical general relativity is a differentiable mani-
fold equipped with a Lorentzian metric. Some of the most important pieces of substructure underly-
ing this picture are illustrated in Figure 1.
The bottom level is a set M whose elements are to be identified with spacetime ’points’ or ’events’.
This set is formless with its only general mathematical property being the cardinal number. In
particular, there are no relations between the elements of M and no special way of labeling any
such element.
The next step is to impose a topology on M so that each point acquires a family of neighborhoods.
It now becomes possible to talk about relationships between point, albeit in a rather non−physical
way. This defect is overcome by adding the key of all standard views of spacetime: the topology
of M must be compatible with that of a differentiable manifold. A point can then be labeled
uniquely in M (at least locally) by giving the values of four real numbers. Such a coordinate sys-
tem also provides a more specific way of describing relationships between points of M, albeit not
intrinsically in so far as these depend on which coordinate systems are chosen to cover M.
In the final step a Lorentzian metric g is placed on M, thereby introducing the ideas of the length of
a path joining two spacetime points, parallel transport with respect to a Riemannian connection,
causal relations between pairs of points etc. There are also a variety of possible intermediate steps
between the manifold and Lorentzian pictures; for example, as signified in Figure 1, the idea of
causal structure is more primitive than that of a Lorentzian metric."
My hypothesis to treat events as quaternions lends more structure than is found in the set M. Specif-
ically, Pontryagin proved that quaternions are a topological algebraic field. Each point has a neigh-
borhood, and limit processes required for a differentiable manifold make sense. Label every quater-
nion event with four real numbers, using whichever coordinate system one chooses. Earlier in this
section I showed how to calculate the Lorentz interval, so the notion of length of a path joining two
events is always there. As described by Isham, spacetime structure is built up with care from four
unrelated real numbers. With quaternions as events, spacetime structure is the observed properties
of the mathematics, inherited by all quaternion functions.
Much work in quantum gravity has gone into viewing how flexible the spacetime structure might
be. The most common example involves how quantum fluctuations might effect the Lorentzian
metric. Physicists have tried to investigate how such fluctuation would effect every level of space-
time structure, from causality, to the manifold to the topology, even the set M somehow.
129

Much work in quantum gravity has gone into viewing how flexible the spacetime structure might
be. The most common example involves how quantum fluctuations might effect the Lorentzian
metric. Physicists have tried to investigate how such fluctuation would effect every level of space-
time structure, from causality, to the manifold to the topology, even the set M somehow.
None of these avenues are open for quaternion work. Every quaternion equation inherits this
wealth of spacetime structure. It is the family quaternion functions are born in. There is nothing to
stop combining Grassman and Euclidean products, which at an abstract level, is the way to merge
classical and quantum descriptions of collections of events. If a non−linear quaternion function
can be defined that is related to the shortest path through spacetime, the cast required for quantum
gravity would be complete.
According to Isham, causal structure is particularly important. With quaternions, that issue is
particularly straightforward. Could event q have caused q’? Take the difference and square it. If
the scalar is positive, then the relationship is timelike, so it is possible. Is it probable? That might
depend on the 3−vector, which could be more likely if the vector is small (I don’t understand the
details of this suggestion yet). If the scalar is zero, the two have a lightlike relationship. If the
scalar is negative, then it is spacelike, and one could not have caused the other.
This causal structure also applies to quaternion potential functions. For concreteness, let q(t) =
cos(pi t (2i + 3j + 4k)) and q’(t) = sin( pi t (5i − .1j + 2k). Calculate the square of the difference
between q and q’. Depending on the particular value of t, this will be positive, negative or zero.
The distance vectors could be anywhere on the map. Even though I don’t know what these particu-
lar potential functions represent, the causal relationship is easy to calculate, but is complex and not
trivial.
The Role of the Spacetime Diffeomorphism Group Diff(M)
Isham lets me off the hook, saying "...[for type 3 and 4 theories] there is no strong reason to sup-
pose that Diff(M) will play any fundamental role in [such] quantum theory." He is right and
wrong. My simple tool collection does not include this group. Yet the concept that requires this
idea is essential. This group is part of the machinery that makes possible causal measurements of
lengths in various topologies. Metrics change due to local conditions. The concept of a flexible,
causal metric must be preserved.
With quaternions, causality is always found in the scalar of the square of the difference. For two
events in flat spacetime, that is the interval. In curved spacetime, the scalar of the square is differ-
ent, but it still is either positive, negative or zero.
The Problem of Time
Time plays a different role in quantum theory and in general relativity. In quantum, time is treated
as a background parameter since it is not represented by an operator. Measurements are made at a
particular time. In classical general relativity in curved spacetime, there are many possible metrics
which might work, but no way to pick the appropriate one. Without a clear definition of measure-
ment, the definition is non−physical. Fixing the metric cannot be done if the metric is subject to
quantum fluctuations.
Isham raises three questions:
"How is the notion of time to be incorporated in a quantum theory of gravity?
Does it play a fundamental role in the construction of the theory or is it a ’phenomenological’
concept that applies, for example, only in some coarse−grained, semi−classical sense?
In the latter case, how reliable is the use at a basic level of techniques drawn from standard quan-
tum theory?"
Three solutions are noted: fix the background causal structure, locate events within functionals of
fields, or make no reference to time.
With quaternions, time plays a central role, and is in fact the center of the matrix representation.
Time is isomorphic to the real numbers, so it forms a totally ordered sub−field of the quaternions.
It is not time per se, but the location of time within the event quaternion (t, x i, y j, z k) that gives
time its significance. The scalar slot can be held by energy (E, px i, py j, pz k), the tangent of
spacetime, by the interval of classical physics (t^2 − x^2 − y^2 − z^2, 2 tx i, 2 ty j, 2 tz k) or the
norm of quantum mechanics (t^2 + x^2 + y^2 + z^2, 0, 0, 0). Time, energy, intervals,
norms,...they all can take the same throne isomorphic to the real numbers, taking on the properties
of a totally ordered set within a larger, unordered framework. Events are not totally ordered, but
time is. Energy/momenta are not totally ordered, but energy is. Squares of events are not totally
ordered, but intervals are. Norms are totally ordered and bounded below by zero.
130

With quaternions, time plays a central role, and is in fact the center of the matrix representation.
Time is isomorphic to the real numbers, so it forms a totally ordered sub−field of the quaternions.
It is not time per se, but the location of time within the event quaternion (t, x i, y j, z k) that gives
time its significance. The scalar slot can be held by energy (E, px i, py j, pz k), the tangent of
spacetime, by the interval of classical physics (t^2 − x^2 − y^2 − z^2, 2 tx i, 2 ty j, 2 tz k) or the
norm of quantum mechanics (t^2 + x^2 + y^2 + z^2, 0, 0, 0). Time, energy, intervals,
norms,...they all can take the same throne isomorphic to the real numbers, taking on the properties
of a totally ordered set within a larger, unordered framework. Events are not totally ordered, but
time is. Energy/momenta are not totally ordered, but energy is. Squares of events are not totally
ordered, but intervals are. Norms are totally ordered and bounded below by zero.
Time is the only element in the scalar of an event. Time appears in different guises for the scalars
of energy, intervals and norms. The richness of time is in the way it weaves through these other
scalars, sharing the center in different ways with space.
Approaches to Quantum Gravity
Isham surveys the field. At this point I think I’ll just explain my approach. It is based on a concept
from general relativity. A painter falling from a ladder travels along the shortest path through
spacetime. How does one go about finding the shortest path? In Euclidean 3−space, that involves
the triangle inequality. A proof can be done using quaternions if the scalar is set to zero. That
proof can be repeated with the scalar set free. The result is the shortest distance through spacetime,
or gravity, according to general relativity.
What is the shortest distance between two points A and B in Euclidean 3−space?
A = (0, ax, ay, az)
B = (0, bx, by, bz)
What is the shortest distance between two worldlines A(t) and B(t) in spacetime?
A (t) = (t, ax (t), ay (t), az (t))
B (t) = (t, bx (t), by (t), bz (t))
The Euclidean 3−space question is a special case of the worldline question. The same proof of the
triangle inequality answers both questions. Parameterize the norm N(k) of the sum of A(t) and B(t).
N (k) = (A + k B)
-
(A + k B)
= A
-
A + k (A
-
B + B
-
A) + k^2 B
-
B
Find the extremum of the parameterized norm.
dN
-------
dk
= 0 = A
-
B + B
-
A + 2 k B
-
B
The extremum is a minimum
d
2
N
-----------
dk
2
= 2 B
-
B ~ 0
The minimum of a quaternion norm is zero. Plug the extremum back into the first equation.
131

0 A
-
A -
(A
-
B + B
-
A)
2
----------------------------------
2 B
-
B
+
(A
-
B + B
-
A)
2
----------------------------------
4 B
-
B
Rearrange.
(A
-
B + B
-
A)
2
4 A
-
A B
-
B
Take the square root.
A
-
B + B
-
A 2
_
A
-
A B
-
B
Add the norm of A and B to both sides.
A- A + A
-
B + B
-
A + B
-
B A
-
A + 2
_
A
-
A B
-
B + B
-
B
Factor.
N (A + B) = (A + B)
-
(A + B) ¡
_
A
-
A +
_
B
-

2
The norm of the worldline of A plus B is less than the norm of A plus the norm of B.
List the mathematical structures required. To move the triangle inequality from Euclidean 3−space
to worldlines required the inclusion of the scalar time component of quaternions. The proof
required differentiation to find the minimum. The norm is a Euclidean product, which plays a
central role in quaternion quantum mechanics. Doubling A or B does not double the norm of the
sum due to cross terms, so the minimal function is not linear.
To address a question raised by general relativity with quaternions required all the structure Isham
suggested except causality using the Grassman product. The above proof could be repeated using
Grassman products. The only difference would be that the extremum would be an interval which
can be positive, negative or zero (a minimum, a maximum or an inflection point).
Certainty Is Seven for Seven
I thought I’d end this long section with a personal story. At the end of my college days, I started
drinking heavily. Not alcohol, soda. I’d buy a Mellow Yellow and suck it down in under ten
seconds. See, I was thirsty. Guzzle that much soda, and, well, I also had to go to the bathroom,
even in the middle of the night. I was trapped in a strange cycle. Then I noticed my tongue was
kind of foamy. Bizarre. I asked a friend with diabetes what the symptoms of that disease were.
She rattled off six: excessive thirst, excessive urination, foamy tongue, bad breath, weight loss, and
low energy. I concluded on the spot I had diabetes. She said that I couldn’t be certain. Six for six
is too stringent a match, and I felt very confident I had this chronic illness. I got the seventh later
when she tested my blood glucose on her meter and it was off−scale. She gave me sympathy, but I
didn’t feel at all sorry for myself. I wanted facts: how does this disease work and how do I cope?
Nothing was made official until I visited the doctor and he ran some tests. The doctor’s prescrip-
tion got me access to the insulin I could no longer produce. It was, and still is today, a lot of work
to manage the disease.
When I look at Isham’s paper, I see six constraints on the structure of any approach to quantum
gravity: events are sets of 4 numbers, events have topological neighborhoods, they live on differen-
tial manifolds, there is one of the three types of causal relationships between all events, the distance
between events is the interval whose form can vary and a Hilbert space is required for quantum
mechanics. Quaternions are six for six. The seventh match is the non−linear shortest norm of
spacetime. I have no doubt in the diagnosis that the questions in quantum gravity will be answered
with quaternions. Nothing here is official. There are many test that must be passed. I don’t know
when the doctor will show up and make it official. It will take a lot of work to manage this
solution.
132

When I look at Isham’s paper, I see six constraints on the structure of any approach to quantum
gravity: events are sets of 4 numbers, events have topological neighborhoods, they live on differen-
tial manifolds, there is one of the three types of causal relationships between all events, the distance
between events is the interval whose form can vary and a Hilbert space is required for quantum
mechanics. Quaternions are six for six. The seventh match is the non−linear shortest norm of
spacetime. I have no doubt in the diagnosis that the questions in quantum gravity will be answered
with quaternions. Nothing here is official. There are many test that must be passed. I don’t know
when the doctor will show up and make it official. It will take a lot of work to manage this
solution.
133

Length in Curved Spacetime
The Affine Parameter of General Relativity
The affine parameter is defined in Misner, Thorne and Wheeler as a multiple of the proper time
plus a displacement.
Λ = a Τ + b
The affine parameter is used to determine length in curved spacetime. In this section, the length of
a quaternion in curved spacetime will be analyzed. Under certain approximations, this length will
depend on the square of the affine parameter, but the two measures are slightly different.
Length in Flat Spacetime
Calculating the square of the interval between two events in flat spacetime was straightforward:
take the difference between two quaternions and square it.
L
flat
= (q - q
·
)
2
= ¡dt
2
- d X

2
, 2 dt d X

¡
The first term is the square of the interval. Spacetime is flat in the sense that the first term is
exactly like the Minkowski metric in spacetime. There are quaternions which preserve the interval,
and those quaternions were used to solve problems in special relativity.
Although not important in this context, it is significant that the value of the vector portion depends
upon the observer. This gives a way to distinguish between various frequencies of light for
example.
Length in Curved Spacetime
Consider if the origin is located at two different locations in spacetime. Characterize each origin
as a quaternion, calling the o and o’. In flat spacetime, the two origins would be identical. Calcu-
late the interval as done above, but account for the change in the origin.
L
curved
= ((q + o) - (q
·
+ o
·
))
2
= ¡d (t + t
o
)
2
- d (X

+ X

o
)
2
, 2 d (t + t
o
) d (X

+ X

o

Examine the first term more closely by expanding it.
¡dt
2
- d X

2
¡ + ¡dt
o
2
- d X

o
2
¡ + 2 dt dt
o
- 2 d X

d X

o
The length in curved spacetime is the square of the interval (invariant under a boost) between the
two origins, plus the square of the interval between the two events, plus a cross term, which will
not be invariant under a boost. The length is symmetric under exchange of the event with the
origin translation.
134

The length in curved spacetime is the square of the interval (invariant under a boost) between the
two origins, plus the square of the interval between the two events, plus a cross term, which will
not be invariant under a boost. The length is symmetric under exchange of the event with the
origin translation.
L curved looks similar to the square of the affine parameter:
Λ
2
= b
2
+ 2 a b Τ + a
2
Τ
2
In this case, b^2 is the origin interval squared and a = 1. There is a difference in the cross terms.
However, in the small curvature limit, delta to >> delta Xo, so tau ~ delta to. Under this approxima-
tion, the square of the affine parameter and L curved are the same.
For a strong gravitational field, L curved will be different than the square of the affine parameter.
The difference will be solely in the nature of the cross term. In general relativity, b and tau are
invariant under a boost. For L curved, the cross term should be covariant. Whether this has any
effects that can be measured needs to be explored.
There exist quaternions which preserve L curved because quaternions are a field (I haven’t found
them yet because the math is getting tough at this point!) It is my hope that those quaternions will
help solve problems in general relativity, as was the case in special relativity.
Implications
A connection to the curved geometry of general relativity was sketched. It should be possible to
solve problems with this "curved" measure. As always, all the objects employed were quaternions.
Therefore any of the previously outline techniques should be applicable. In particular, it will be
fun in the future to think about things like
((q + o) - (q
·
+ o
·
))
-
((q + o) - (q
·
+ o
·
))
= ¡d (t + t
o
)
2
+ d (X

+ X

o
)
2
, 2 d (t + t
o
) d (X

+ X

o

= ¡¡dt
2
- d X

2
¡ + ¡dt
o
2
- d X

o
2
¡ + 2 dt dt
o
+ 2 d X

d X

o
, ...¡
which could open the door to a quantum approach to curvature.
135

A New Idea for Metrics
In special relativity, the Minkowski metric is used to calculate the interval between two spacetime
intervals for inertial observers. Einstein recognized that inertial observes were "special", a unique
class. Therefore he set out to understand what was the most general notion for transformations
and metrics. This lead to his study of Riemannian geometry, and eventually to general relativity.
In this section I shall start from the Lorentz invariant interval using quaternions, then try to general-
ize this approach using a different way which might prove compatible with quantum mechanics.
For the physics of gravity, general relativity (GR) makes the right predictions of all experimental
tests conducted to date. For the physics of atoms, quantum mechanics (QM) makes the right
predictions to an even high degree of precision. The problem of building a quantum theory of
gravity (QG) hides between general relativity and quantum mechanics. General relativity deals
with the measurements of intervals in curved spacetime, special relativity (SR) being adapted to
work in flat space. Quantum mechanics is used to calculate the norms of wave functions in a flat
linear space. A quantum gravity theory will be used to calculate norms of wave functions in
curved space.
Measurement
interval norm
diff. flat SR QM
geo. curved GR QG
This chart suggests that the form of measurement (interval/norm) should be independent of
differential geometry (flat/curved). That will be the explicit goal of this section.
Quaternions come with a metric, a means of taking 4 numbers and returning a scalar. Hamilton
defined the roles like so:
i

2
= j

2
= k

2
= -1 i

j

k

= -1
The scalar result of squaring a differential quaternion in the interval of special relativity:
scalar ¡(dt, d X

)
2
¡ = dt
2
- d X

.d X

How can this be generalized? It might seem natural to explore variations on Hamilton’s rules
shown above. Riemannian geometry uses that strategy. When working with a field like quater-
nions, that approach bothers me because Hamilton’s rules are fundamental to the very definition of
a quaternion. Change these rules and it may not be valid to compare physics done with different
metrics. It may cause a compatibility problem.
Here is a different approach which generalizes the scalar of the square while being consistent with
Hamilton’s rules.
interval
2
= scalar (g dq g dq)
136

if g = ¡1, 0

¡,
then interval
2
= dt
2
- d X

.d X

If g is the identity matrix. Then then result is the flat Minkowski interval. The quaternion g could
be anything. What if g = i? (what would you guess, I was surprised :−)
scalar (((0, 1, 0, 0) (t, x, y, z))
2
) =
= ¡-t
2
+ x
2
- y
2
- z
2
, 0

¡
Now the special direction x plays the same role as time! Does this make sense physically? Here is
one interpretation. When g=1, a time−like interval is being measured with a wristwatch. When
g=i, a space−like interval along the x axis is being measured with a meter stick along the x axis.
Examine the most general case, where small letters are scalar, and capital letters are 3−vectors:
interval
2
= scalar ((g, G

) (dt, d X

) (g, G

) (dt, d X

)) =
= g
2
(dt
2
- d X

.d X

) - 4 g dt G

.d X

+
(G

.d X

)
2
- dt
2
d G

.d G

- (G

xd X

).(G

xd X

) =
In component form...
= (+ g
2
- Gx
2
- Gy
2
- Gz
2
) dt
2
+
+ (- g
2
+ Gx
2
- Gy
2
- Gz
2
) dx
2
+
+ (- g
2
- Gx
2
+ Gy
2
- Gz
2
) dy
2
+ (- g
2
- Gx
2
- Gy
2
+ Gz
2
) dz
2
+
- 4 g Gx dt dx - 4 g Gy dt dy - 4 g Gz dt dz
+ 4 Gx Gy ds dy + 4 Gx Gz dx dz + 4 Gy Gz dy dz
This has the same combination of ten differential terms found in the Riemannian approach. The
difference is that Hamilton’s rule impose an additional structure.
I have not yet figured out how to represent the stress tensor, so there are no field equations to be
solved. We can figure out some of the properties of a static, spherically−symmetric metric. Since
it is static, there will be no terms with the deferential element dt dx, dt dy, or dt dz. Since it is
spherically symmetric, there will be no terms of the form dx dy, dx dz, or dy dz. These constraints
can both be achieved if Gx = Gy = Gz = 0. This leaves four differential equations.
Here I will have to stop. In time, I should be able to figure out quaternion field equations that do
the same work as Einstein’s field equations. I bet it will contain the Schwarzschild solution too :−)
Then it will be easy to create a Hilbert space with a non−Euclidean norm, a norm that is deter-
mined by the distribution of mass−energy. What sort of calculation to do is a mystery to me, but
someone will get to that bridge...
137

The Gravitational Redshift
Gravitational redshift experiments are tests of conservation of energy in a gravitational potential.
A photon lower in a gravitational potential expends energy to climb out, and this energy cost is
seen as a redshift. In this section, the difference between weak gravitational potentials will be
calculated and shown to be consistent with experiment. Quaternions are not of much use here
because energy is a scalar, the first term of a quaternion that is a scalar multiple of the identity
matrix.
The Pound and Rebka Experiment
The Pound and Rebka experiment used the Mossbauer effect to measure a redshift between the
base and the top of a tower at Harvard University. The relevant potentials are
Φ
tower
=
G M
--------------
r + h
;
Φ
base
=
G M
--------
r
;
The equivalence principle is used to transform the gravitational potential to a speed (this only
involves dividing phi by the constant c^2).
Β
tower
=
G M
--------------------------
c
2
(r + h)
;
Β
base
=
G M
-----------
c
2
r
;
Now the problem can be viewed as a relativistic Doppler effect problem. A redshift in a frequency
is given by
Ν’ == (Γ[Β] + Β Γ[Β]) Ν
o
For small velocities, the Doppler effect is
Series[Γ[Β] + Β Γ[Β], {Β, 0, 1]]
= 1 + Β + O[Β]
2
The experiment measured the difference between the two Doppler shifts.
Series[((1 + Β
tower
) - (1 + Β
base
)) Ν
o
, {h, 0, 1]]
= -
G M Ν
o
h
------------------
c
2
r
2
+ O[h]
2
Or equivalently,
138

Ν’ = g h Ν
o
This was the measured effect.
Escape From a Gravitational Potential
A photon can escape from a star and travel to infinity ( or to us, which is a good approximation).
The only part of the previous calculation that changes is the limit in the final step.
Limit[((1 + Β
tower
) - (1 + Β
base
)) Ν
o
, h -> Infinity]
= -
G M Ν
o
--------------
c
2
r
This shift has been observed in the spectral lines of stars.
Clocks at different heights in a gravitational field
C. O. Alley conducted an experiment which involved flying an atomic clock at high altitude and
comparing it with an atomic clock on the ground. This is like integrating the redshift over the time
of the flight.
¸
0
t
-
G M h
-------------
c
2
r
2
ot = -
G h M t
----------------
c
2
r
2
This was the measured effect.
Implications
Conservation of energy involves the conservation of a scalar. Consequently, nothing new will
happen by treating it as a quaternion. The approach used here was not the standard one employed.
The equivalence principle was used to transform the problem into a relativistic Doppler shift effect.
Yet the results are no different. This is just part of the work to connect quaternions to measurable
effects of gravity.
References
For the Pound and Rebka experiment, and escape:
Misner, Thorne, and Wheeler, Gravitation, 1970.
For the clocks at different heights:
Quantum optics, experimental gravitation and measurement theory, Ed. P. Meystre, 1983 (also
mentioned in Taylor and Wheeler, Spacetime Physics, section 4.10)
139

A Brief Summary of Important Laws in
Physics Written as Quaternions
Summary
Classical Mechanics
Newton’s 2nd Law in an Inertial Reference Frame, Cartesian Coordinates
Newton’s 2nd law in an Inertial Reference Frame, Polar Coordinates, for a Central Force
Newton’s 2nd Law in a Noninertial Rotating Reference Frame
The Simple Harmonic Oscillator
The Damped SHO
The Wave Equation

Special Relativity
Rotations and Dilations Create a Representation of the Lorentz Group
An Alternative Algebra for the Lorentz Group

Electromagnetism
The Maxwell Equations
Maxwell Written With Potentials
The Lorentz Force
Conservation Laws
The Field Tensor F in Different Gauges
The Maxwell Equations in the Light Gauge (QED?)
The Stress Tensor of the Electromagnetic Field

Quantum Mechanics
Quaternions in Polar Coordinate Form
Multiplying Quaternion Exponentials
Commutators of Observable Operators
The Uncertainty Principle
Automorphic Commutator Identities
The Schrödinger Equation
The Klein−Gordon Equation
Time Reversal Transformations for Intervals

Gravity
The 3 Fields: g, E & B
Field Equations
Recreating Maxwell
Unified Field Equations
Conservation Laws
Gauge Transformations
Equations of Motion
Unified Equations of Motion
Strings
Dimensionless Strings
Behaving Like a Relativistic Quantum Gravity Theory
140

Classical Mechanics
Newton’s 2nd Law in an Inertial Reference Frame, Cartesian Coordinates
Newton’s 2nd law in an Inertial Reference Frame, Polar Coordinates, for a Central Force
Newton’s 2nd Law in a Noninertial Rotating Reference Frame
The Simple Harmonic Oscillator
The Damped SHO
The Wave Equation

Special Relativity
Rotations and Dilations Create a Representation of the Lorentz Group
An Alternative Algebra for the Lorentz Group

Electromagnetism
The Maxwell Equations
Maxwell Written With Potentials
The Lorentz Force
Conservation Laws
The Field Tensor F in Different Gauges
The Maxwell Equations in the Light Gauge (QED?)
The Stress Tensor of the Electromagnetic Field

Quantum Mechanics
Quaternions in Polar Coordinate Form
Multiplying Quaternion Exponentials
Commutators of Observable Operators
The Uncertainty Principle
Automorphic Commutator Identities
The Schrödinger Equation
The Klein−Gordon Equation
Time Reversal Transformations for Intervals

Gravity
The 3 Fields: g, E & B
Field Equations
Recreating Maxwell
Unified Field Equations
Conservation Laws
Gauge Transformations
Equations of Motion
Unified Equations of Motion
Strings
Dimensionless Strings
Behaving Like a Relativistic Quantum Gravity Theory
Each of the following laws of physics are generated by quaternion operators acting on the appropri-
ate quaternion−valued functions. The generators of these common laws often provide insight.
Classical Mechanics
Newton’s 2nd Law for an Inertial Reference Frame in Cartesian Coordinates
A = ¸
d
-------
dt
, 0

¸ ¡1, R

¡ = ¸0, R

¸
Newton’s 2nd Law in Polar Coordinates for a Central Force in a Plane
A = (Cos[Θ], 0, 0, -Sin[Θ])
¸
d
-------
dt
, 0

¸
2
(t, R Cos[Θ], R Sin[Θ], 0) =
141

=

0,
L
2
-------------
m
2
R
3
+ R

,
2 L R

------------
m R
2
, 0
|

Newton’s 2nd Law in a Noninertial, Rotating Frame
A = ¸
d
-------
dt
, Ω

¸ ¡- Ω

.R

, R

+ Ω

x R

¡ =
= ¸- Ω

.R

, R

+ 2 Ω

x R

+ Ω

x R

- Ω

.R

¸
The Simple Harmonic Oscillator (SHO)
¸
d
-------
dt
, 0

¸
2
(0, x, 0, 0) + ¸0,
k
----
m
x, 0, 0¸ =
¸0,
d
2
x
-----------
dt
2
+
k x
--------
m
, 0, 0¸ = 0
The Damped Simple Harmonic Oscillator
¸
d
-------
dt
, 0

¸
2
(0, x, 0, 0) +
¸
d
-------
dt
, 0

¸ (0, b x, 0, 0) + ¸0,
k
----
m
x, 0, 0¸ =
= ¸0,
d
2
x
-----------
dt
2
+
b d x
------------
dt
+
k x
--------
m
, 0, 0¸ = 0
The Wave Equation
¸
d
-----------
v dt
,
d
-------
dx
, 0, 0¸
2
(0, 0, f[t v + x], 0) =
= ¸0, 0, ¸-
d
2
---------
dx
2
+
d
2
----------------
dt
2
v
2
¸ f[t v + x],
2 d
2
f[t v + x]
----------------------------------
dt dx v
¸
The third term is the one dimensional wave equation. The forth term is the instantaneous power
transmitted by the wave.
A Force Is Conservative If The Curl Is Zero
odd ¸¸
d
-------
dt
, V

¸, F

¸ = 0
142

A Force Is Conservative If There Exists a Potential Function for the Force
F = ¸
d
-------
dt
, V

¸ (Φ, 0

)
A Force Is Conservative If the Line Integral of Any Closed Loop Is Zero
_
F dt = 0
A Force Is Conservative If the Line Integral Along Different Paths Is the Same
(
_
path 1
) F dt = (
_
path 2
) F dt
Special Relativity
Rotations and Dilations Create the Lorentz Group
q’ =
q + (Γ - 1)
even ¡even ¡V

-
, q¡, V

¡
--------------------------------------------------------
| V

|
2
+ Γ even ¡V

-
, q
-
¡
An Alternative Algebra for Lorentz Boosts
scalar ((t, x, y, z)
2
) = scalar ((L (t, x, y, z))
2
)
For boosts along the x axis...
If t = 0, then
L = Γ (1, Β, 0, 0)
If x = 0, then
L = Γ (1, -Β, 0, 0)
If t = x, then for blueshifts
L = Γ (1 - Β, 0, 0, 0)
For general boosts along the x axis
143

L = (Γ t
2
+ Γ x
2
- 2 Γ Βt x + (y
2
+ z
2
), Γ Β (-t
2
+ x
2
),
t (Β Γ z + y (1 - Γ)) - x (Γ Β y + z (1 - Γ)),
t (Γ Β y + z (1 - Γ)) + x (Γ Β z + y (1 - Γ)))/
(t
2
+ x
2
+ y
2
+ z
2
)
Electromagnetism
The Maxwell Equations
even ¸¸
o
--------
ot
,

¸, (0, B

)¸ + odd ¸¸
o
--------
ot
,

¸, (0, E

)¸ =

-

B

,

X E

+
oB

--------
ot
|

= ¡0, 0

¡
odd ¸¸
o
--------
ot
,

¸, (0, B

)¸ - even ¸¸
o
--------
ot
,

¸, (0, E

)¸ =

E

,

X B

-
oE

--------
ot
|

= 4 Π (Ρ, J

)
Maxwell Written with Potentials
The fields
E = vector ¸even ¸¸
o
--------
ot
, -

¸, (Φ, -A

)¸¸ =

0, -
oA

--------
ot
-

Φ
|

B = odd ¸¸
o
--------
ot
, -

¸, (Φ, -A

)¸ = (0,

x A

)
The field equations
even ¸¸
o
--------
ot
,

¸, odd ¸¸
o
--------
ot
, -

¸, (Φ, -A

)¸¸ +
odd ¸¸
o
--------
ot
,

¸, vector ¸even ¸¸
o
--------
ot
, -

¸, (Φ, -A

)¸¸¸ =
=

-

x A

,
o

x A

---------------
ot
-

X
oA

--------
ot
-

x

Φ
|

=

-

B

,
oB

--------
ot
+

x E
|

= ¡0, 0

¡
144

odd ¸¸
o
--------
ot
,

¸, odd ¸¸
o
--------
ot
, -

¸, (Φ, A

)¸¸ -
even ¸¸
o
--------
ot
,

¸, vector ¸even ¸¸
o
--------
ot
, -

¸, (Φ, -A

)¸¸¸ =
=

-

Φ -

oA

--------
ot
,

X

X A

+
o
2
A

----------
ot
2
+
o

Φ
-----------
ot
|

=

E

,

X B

-
oE

--------
ot
|

= 4 Π (Ρ, J

)
The Lorentz Force
odd ¡¡Γ, Γ Β

¡, (0, B

)¡ - even ¡¡-Γ, Γ Β

¡, (0, E

)¡ =
¡Γ Β

E

, Γ E

+ Γ Β

X B

¡
Conservation Laws
The continuity equation
scalar

¸
o
--------
ot
, -

¸

E

,

X B

-
oE

--------
ot
|

|

=

o
--------
ot

E

-

oE

--------
ot
+

X B

, 0
|

=
= scalar ¸¸
o
--------
ot
, -

¸, 4 Π (Ρ, J

)¸ = 4 Π ¸E

J

+

--------
ot
, 0¸
Poynting’s theorem for energy conservation.
scalar

(0, -E

)

E

,

X B

-
oE

--------
ot
|

|

=

E

X B

- E

oE

--------
ot
, 0
|

=

-

(E

X B

) -
1
----
2

oE

--------
ot
|

2
-
1
----
2

oB

--------
ot
|

2
, 0
|

= scalar ((0, -E

), 4 Π (Ρ, J

)) = 4 Π (E

J

, 0)
The Field Tensor F in Different Gauges
The anti−symmetric 2−rank electromagnetic field tensor F
145

¸
o
--------
ot
, -

¸ (Φ, -A

) - (Φ, A

) ¸
o
--------
ot
,

¸ =

0, -
oA

--------
ot
-

Φ +

X A
|

F in the Lorenz gauge.
¸
o
--------
ot
, -

¸

(Φ, A

) + (Φ, -A

)
---------------------------------------
2
|

-

(Φ, A

) - (Φ, -A

)
---------------------------------------
2
|

¸
o
--------
ot
,

¸ =
=

--------
ot
+

.A

, -
oA

--------
ot
-

Φ +

X A
|

F in the Coulomb gauge
¸
o
--------
ot
, -

¸ (Φ, -A

) + ¸
o
--------
ot
, -

¸

(Φ, A

) - (Φ, -A

)
---------------------------------------
4
|

+

(Φ, -A

) - (Φ, A

)
---------------------------------------
4
|

¸
o
--------
ot
,

¸ =
=

--------
ot
, -
oA

--------
ot
-

Φ +

X A
|

F in the temporal gauge.
¸
o
--------
ot
, -

¸ (Φ, -A

) - ¸
o
--------
ot
, -

¸

(Φ, A

) + (Φ, -A

)
---------------------------------------
4
|

-

(Φ, -A

) + (Φ, A

)
---------------------------------------
4
|

¸
o
--------
ot
,

¸ =
=

-

.A

, -
oA

--------
ot
-

Φ +

X A
|

F in the light gauge.
¸
o
--------
ot
, -

¸ (Φ, -A

) =

--------
ot
-

.A

, -
oA

--------
ot
-

Φ +

X A
|

The light gauge is one sign different from the Lorenz gauge, but its generator is a simple as it gets.
146

The Maxwell Equations in the Light Gauge
Note: subsequent work has suggested that the scalar in these equations is part of a unified field
theory.
even ¸¸
o
--------
ot
,

¸, odd ¸¸
o
--------
ot
,

¸, (Φ, A

)¸¸ +
odd ¸¸
o
--------
ot
,

¸, even ¸¸
o
--------
ot
, -

¸, (Φ, -A

)¸¸ =
= (-

X A

, -

X

Φ) = ¡0, 0

¡
odd ¸¸
o
--------
ot
,

¸, odd ¸¸
o
--------
ot
,

¸, (Φ, A

)¸¸ -
even ¸¸
o
--------
ot
,

¸, even ¸¸
o
--------
ot
, -

¸, (Φ, -A

)¸¸ =
=

o
2
Φ
----------
ot
2
+

Φ , -
o
2
A

----------
ot
2
+

X (

X A

) -

A
|

=

o
2
Φ
----------
ot
2
+

2
Φ , -
o
2
A

----------
ot
2
-

2
A
|

= 4 Π (Ρ, J

)
The Stress Tensor of the Electromagnetic Field
T
ik
=
L
a=x
y,z
L
b=x
y,z
1
--------
4 Π

¸
even (Ua, Ub)
----------------------------------
3
- 1¸
((0, E)
2
+ (0, B)
2
)
-----------------------------------------------
2
-
- even (E, Ua) even (E, Ub) - even (B, Ua) even (B, Ub) -
- even (odd (E, B), Ua) - even (odd (E, B), Ub) =
= (-Ex Ey - Ex Ez - Ey Ez - Bx By - Bx Bz - By Bz
+ Ey Bz - Ez By + Ez Bx - Ex Bz + Ex By - Ey Bx, 0)/2
Π
Quantum Mechanics
Quaternions in Polar Coordinate Form
q = || q || exp[Θ I

] = q
-
q (Cos[Θ] + I

Sin[Θ])
147

Multiplying Quaternion Exponentials
q q’ =
{q, q’]
-
+ Abs[q, q’]
-
Exp¡
Π
----
2
[q, q’]
-
-------------------------------
Abs[q, q’]
-
¡
Commutators of Observable Operators
[A

, B

] q = (A

B

- B

A

) q = -a I
d q
---------
d a
+ I
d a q
------------
d a
= -a I
d q
---------
d a
+ a I
d q
---------
d a
+ I q
d a
---------
d a
= I q
The Uncertainty Principle
[A, B]
------------------
2
=
I
----
2
∆A
2
∆B
2
Unifying the Representation of Spin and Angular Momentum
For small rotations:
[R
e
1
=0
, R
e
2
=0
] = 2 (R
e
3
=0

2
) - R (0))
Automorphic Commutator Identities
[q, q’] = [q
-
, q’
-
] = [q
-1
, q’
-1
]
-1
= [q
-2
, q’
-2
]
-2
{q, q’] = {q
-
, q’
-
]
-
= -{q
-1
, q’
-1
]
-1
= -{q
-2
, q’
-2
]
-2
The Schrödinger Equation
I = Exp

V

--------------
_
V

.V

(Ω t - K

.X

)
|

H Ψ = -i

--------
ot
=
-
2
----------
2 m
V
2
Ψ + V (0, X) Ψ
The Klein−Gordon Equation
_
n=0
·

¸
o
--------
ot
,

¸
2
+ ¸
o
--------
ot
, -

¸
2
+ (E
n
, P

n
)
2
+ (E
n
, -P

n
)
2
|

(E
n
, P

n
)/2 =
148

= _
n=0
·
¸-

(

X P

n
) -

E
n
-
P

n
(P

n
X P

n
) - (P

n
P

n
) E
n
+ E
n
3
+
o
2
E
n
------------
ot
2
,

X (

X P

n
) +

X (

E
n
) + P

n
X (P

n
X P

n
) + (P

n
X P

n
) E
n
-

(

P

n
) +P

n
E
n
2
- P

n
(P

n
P

n
) +
o
2
P

n
------------
ot
2
|

It takes some skilled staring to assure that this equation contains the Klein−Gordon equation along
with vector identities.
Time Reversal Transformations for Intervals
(t, X

) -> (-t, X

) = R (t, X

)
R = (-t, X

) (t, X

)
-1
= (-t
2
+ X

.X

, 2 t X

)/(t
2
+ X

.X

)
Classically
if Β ·· 1 then R ¡-1, 2 t Β

¡
R = ¡-
Ε
----
T
, 1, 0, 0¡
149

Gravity
The 3 Fields: g, E & B
¸
o
--------
ot
, -

¸ (Φ, -A

) =
¡Φ

-

.A

, - A

-

Φ +

x A

¡ = (g, E

+ B

)
Field Equations: Almost Maxwell and a Dynamic g
¸
o
--------
ot
,

¸ (g, E

+ B

) =
¡g

-

.E

-

.B

, E

+

x B

+ B

+

x E

+

g ¡ =
4 Π (Ρ
g
+ Ρ
e
, J

g
+ J

e
)
¸
o
--------
ot
, -

¸ (g, E

+ B

) =
¡g

+

.E

+

.B

, E

-

x B

+ B

-

x E

-

g ¡ =
4 Π (Ρ
g
+ Ρ
e
, J

g
+ J

e
)
Recreating Maxwell
Let U = ¡-

.E

-

.B

+ g

, E

+

x B

+ B

+

x E

+

W = ¡

.E

+

.B

+ g

, E

-

x B

+ B

-

x E

-

Mirror ((W + U)/2) + (W - U)
-
/2 =
¡

.E

+

.B

+ g

, -E

+

x B

+ B

+

x E

+

Unified Field Equations
¸
o
--------
ot
,

¸ ¸
o
--------
ot
, -

¸ (Φ, -A

) =
= ¡Φ

-

.A

+

.A

+

.

Φ -

.

x A

, - A

-

Φ

+

x A

+

Φ

-

.A

-

x A

-

x

Φ +

x

x A

¡ =
= ¡Φ

+

2
Φ , - A

-

2
A

¡ = 4 Π (Ρ
u
, J

u
)
150

¸
o
--------
ot
, -

¸ ¸
o
--------
ot
, -

¸ (Φ, -A

) =
= ¡Φ

-

.A

-

.A

-

.

Φ +

.

x A

, - A

-

Φ

+

x A

-

Φ

+

.A

+

x A

+

x

Φ -

x

x A

¡ =
= ¡Φ

-

2
Φ - 2

.A

, - A

+

2
A

- 2

Φ

+ 2

x A

¡ =
= 4 Π (Ρ
u
, J

u
)
Conservation Laws
¸
o
--------
ot
, -

¸ ¸
o
--------
ot
,

¸ (g, E

+ B

) =
= ¡g

+

2
g, E

+

x B

+

.E

-

x E

-

x

x B

¡ =
= ¸
o
--------
ot
, -

¸ 4 Π (Ρ
g
+ Ρ
e
, J

g
+ J

e
) =
= 4 Π
¡Ρ

g
+ Ρ

e
+

.J

g
+

.J

e
, J

g
+ J

e
-

Ρ
g
-

Ρ
e
-

x J

g
-

x J

e
¡
Ρ

e
+

.J

e
= 0
J

g
-

Ρ
g
-

x J

g
= 0
If the differential operator acts on the hyperbolic equation, analogous results are obtained:
¸
o
--------
ot
,

¸ ¸
o
--------
ot
, -

¸ (g, E

+ B

) =
= ¡g

+

2
g, E

+ B

+

.E

-

x

x B

-

x

x E

¡ =
= ¸
o
--------
ot
,

¸ 4 Π (Ρ
g
+ Ρ
e
, J

g
+ J

e
) =
= 4 Π
¡Ρ

g
+ Ρ

e
-

.J

g
-

.J

e
, J

g
+ J

e
+

Ρ
g
+

Ρ
e
+

x J

g
+

x J

e
¡
There are two conservation laws here, charge conservation for electromagnetism in the scalar, and
a vector conservation for gravity.
Ρ

e
-

.J

e
= 0
151

J

g
+

Ρ
g
+

x J

g
= 0
Gauge Transformations
(Φ, A

) - (Φ’, A

’) = ¡Φ - Λ

- V

.A

, A

+ V

Λ - A

+ V

x A

¡ =
(Φ, A

) + ¸-
o
--------
ot
,

¸ (Λ, A

)
Equations of Motion
¡Γ, Γ Β

¡ (g, E

+ B

) =
=
¡Γ g - Γ Β

.E

- Γ Β

.B

, Γ E

+ Γ Β

x B

+ Γ B

+ Γ Β

x E

+ Γ Β

g ¡ =
=

W

----
m
+
W

----
e
,
P

------
m
+
P

----
e

|

Unified Equations of Motion
Repeat the exercise from above, but this time, look to the potentials.
¡Γ, Γ Β

¡ ¸
o
--------
ot
, -

¸ (Φ, -A

) =
¡Γ, Γ Β

¡ ¡Φ

-

.A

, - A

-

Φ +

x A

¡ =
= ¡Γ Φ

- Γ

.A

+ Γ Β

.A

+ Γ Β

.

Φ - Γ Β

.

x A

,
-Γ A

- Γ

Φ

+ Γ

x A

+ Φ

Γ Β

-

.A

Γ Β

- Γ Β

x A

- Γ Β

x

Φ + Γ Β

x

x A

¡
That is pretty complicated! The key to simplifying this equation is to see what happens for light,
where dt/dx = dx/dt. Gamma blows up, but if the equation is over gamma, that problem becomes a
scaling factor. With beta equal to one, a number of terms cancel, which can be seen more clearly if
the terms are written out explicitly.
152

=

Φ

-
o
--------
oX

.
oA

--------
ot
+
oX

--------
ot
.
oA

--------
ot
+
oX

--------
ot
.
o
--------
oX

Φ -
oX

--------
ot
.
o
--------
oX

x A

,
- A

-
o
--------
oX

--------
ot
+
o
--------
oX

x
oA

--------
ot
+

--------
ot

oX

--------
ot
-
o
--------
oX

.A

oX

--------
ot
-
o
--------
oX

x
oA

--------
ot
-
o
--------
oX

x
o
--------
oX

Φ +
o
--------
oX

x
o
--------
oX

x A
|

It would take a real mathematician to state the proper constraints on the three pairs of cancellations
that happen when velocities get flipped. There are also a pair of vector identities, presuming sim-
ple connectedness. This leads to the following equation:
=

2 Φ

, - A

-
o
--------
oX

.A

oX

--------
ot
+
o
--------
oX

x
o
--------
oX

x A
|

The scalar change in energy depends only on the scalar potential, and the 3−vector change in
momentum only depends on the 3−vector A.
Strings
dq
2
= ¸da
0
2
e
0
2
+ da
1
2

e
1
2
---------
9
+ da
2
2

e
2
2
---------
9
+ da
3
2

e
3
2
---------
9
,
2 da
0
da
1
e
0

e
1
-------
3
, 2 da
0
da
2
e
0

e
2
-------
3
, 2 da
0
da
3
e
0

e
3
-------
3
¸ =
= (interval
2
, 3 - string)
Dimensionless Strings
dq
2
=
c
5
--------
G h
¸da
0
2
e
0
2
+ da
1
2

e
1
2
-----------
9 c
2
+ da
2
2

e
2
2
-----------
9 c
2
+ da
3
2

e
3
2
-----------
9 c
2
,
2 da
0
da
1
e
0

e
1
--------
3 c
, 2 da
0
da
2
e
0

e
2
--------
3 c
, 2 da
0
da
3
e
0

e
3
--------
3 c
¸
As far as the units are concerned, this is relativistic (c) quantum (h) gravity (G). Take this con-
stants to zero or infinity, and the difference of a quaternion blows up or disappears.
Behaving Like a Relativistic Quantum Gravity Theory
Case 1: Constant Intervals and Strings
T : dq - dq’ such that scalar (dq
2
) =
scalar (dq’
2
) and vector (dq
2
) = vector (dq’
2
)
153

Case 2: Constant Intervals
T : dq - dq’ such that scalar (dq
2
) =
scalar (dq’
2
) and vector (dq
2
) = vector (dq’
2
)
Case 3: Constant Strings
T : dq - dq’ such that scalar (dq
2
) =
scalar (dq’
2
) and vector (dq
2
) = vector (dq’
2
)
Case 4: No Constants
T : dq - dq’ such that scalar (dq
2
) =
scalar (dq’
2
) and vector (dq
2
) = vector (dq’
2
)
In this proposal, changes in the reference frame of an inertial observer are logically independent
from changing the mass density. The two effects can be measured separately. The change in the
length−time of the string will involve the inertial reference frame, and the change in the interval
will involve changes in the mass density.
154

Conclusions
What’s been done
It is an old dream (or perhaps more accurately, a recurring nightmare) to express laws of physics
using quaternions. In these web pages, quaternion operators were employed to express central
laws of physics: Newton’s second law, the Maxwell equations, and the Klein−Gordon equation
for relativistic quantum mechanics. Applications of quaternions to special relativity were done in
detail, with over 50 problems worked out explicitly. Quaternions do not make problem solving
easy. Rather, they help unite the laws themselves. Significantly, an analysis of the length of a
quaternion interval if the origin is moved establishes a connection to the machinery of general
relativity, the affine parameter.
What’s new
One might suspect that the reason for the success claimed above is that nothing proposed is new.
After all, quaternions are a linear combination of tensors of rank zero and one, and while used in
a new way here, does anything genuinely novel appear?
I believe that a new very powerful idea drives this work, namely, that events represented as quater-
nion are a topological algebraic field. This implies that any collection of events can be generated
by an appropriate quaternion function. Scalars and vectors mix under multiplication, so quater-
nions are a mixed representation.
A new view of relativistic quantum mechanics was outlined. The Klein−Gordon equation is a
scalar equation. When quaternion operators are employed, the Klein−Gordon equation is part of
a larger set, including a scalar and vector identity analogous to the Maxwell equations. These
additional identities are also valid by the conventional analysis, but they do not naturally arise, so
the parallel to the Maxwell equations is less clear.
A new link to general relativity has been proposed which is slightly different. The invariant
interval in special relativity was the first term of the difference between two events quaternions
squared. If the origin changes, then the first term of the difference between the two event quater-
nions and the origin quaternions squared is similar to the square of the affine parameter of general
relativity. The only difference lies in the cross term.
Every event, every function, every operator used was a member of the field of quaternions. This
might strike some as a comic reliance on a solitary tool. I prefer to think of it as a great demo-
cratic principle. Physics is impressively democratic, with each photon or electron obeying the
same collection of laws interchangeably. The mathematics underlying the laws of physics should
reflect this interchangeability.
155

What needs to be done
Nothing presented was proven rigorously. There were few references to the literature. This body
of work is more like a skeleton of work, a reflection of the author’s semi−formal training and
isolation from the professional physics community. There is a need to flesh these ideas out only
if there is the potential for new insights. Although I appreciate the standard approach, I feel like I
have gained new insight into why Maxwell’s equations are necessary, have a new way to view
relativistic quantum mechanics, and cling to a novel toehold on general relativity. And since each
of these is a quaternion, it becomes possible to mix and match them to create new areas of study.
I hope this work generates interest in the physics community.
More problems need to be solved. My upcoming focus will be the Dirac function and Fourier
analysis using quaternions. If I can build these functions, it should be possible to approach prob-
lems in electromagnetism and quantum mechanics. I won’t be easy, it never is, but it might be
elegant. And maybe I’ll dabble a little with curves...
156

1

INDEX
Introduction 2 What are Quaternions? 3 Unifying Two Views of Events 4 A Brief History of Quaternions Mathematics 6 Multiplying Quaternions the Easy Way 7 Scalars, Vectors, Tensors and All That 11 Inner and Outer Products of Quaternions 13 Quaternion Analysis 23 Topological Properties of Quaternions 28 Quaternion Algebra Tool Set Classical Mechanics 32 Newton’s Second Law 35 Oscillators and Waves 37 Four Tests of for a Conservative Force Special Relativity 40 Rotations and Dilations Create the Lorentz Group 43 An Alternative Algebra for Lorentz Boosts Electromagnetism 48 Classical Electrodynamics 51 Electromagnetic Field Gauges 53 The Maxwell Equations in the Light Gauge: QED? 56 The Lorentz Force 58 The Stress Tensor of the Electromagnetic Field Quantum Mechanics 62 A Complete Inner Product Space with Dirac’s Bracket Notation 67 Multiplying quaternions in Polar Coordinate Form 69 Commutators and the Uncertainty Principle 74 Unifying the Representations of Integral and Half−Integral Spin 79 Deriving A Quaternion Analog to the Schrödinger Equation 83 Introduction to Relativistic Quantum Mechanics 86 Time Reversal Transformations for Intervals Gravity 89 Unified Field Theory by Analogy 101 Einstein’s vision I: Classical unified field equations for gravity and electromagnetism using Riemannian quaternions 115 Einstein’s vision II: A unified force equation with constant velocity profile solutions 123 Strings and Quantum Gravity 127 Answering Prima Facie Questions in Quantum Gravity Using Quaternions 134 Length in Curved Spacetime 136 A New Idea for Metrics 138 The Gravitational Redshift 140 155 A Brief Summary of Important Laws in Physics Written as Quaternions Conclusions

2

What Are Quaternions?
Quaternions are numbers like the real numbers: they can be added, subtracted, multiplied, and divided. There is something odd about them, hinted at by the Latin. Quaternions are composed of four numbers that work together as one. They were discovered by several people back in the eighteen hundreds. Some enthusiasts thought quaternions would be able to express everything that could happen in our three dimensions of space and one for time because quaternions naturally had that form too. Math accidents do not happen − they revel deep things about how Nature works. Unfortunately, the big fans of quaternion mathematics claimed far more than they would deliver. Useful ideas born from initial quaternion work − for example the notion of scalars, vectors, div, grad, and curl − were stripped out of their initial context, and made more "general". I use the quotes because from my viewpoint, there is nothing more general than a number that can be added, subtracted, multiplied, and divided. I am trying to continue the project of applying four−dimensional quaternions to the four−dimensional spacetime we live in.

Now a theorist can create patterns of events with events. This work challenges that assumption. or multiplied by a scalar. Nothing else can be done. like metrics and group theory. This may be done by representing events as the mathematical field of quaternions.3 Unifying Two Views of Events An experimentalist collects events about a physical system. Even gravity may be part of a simple quaternion wave equation. proposing instead an integrated set of equations which are all based on the same 4−dimensional mathematical field of quaternions. the special functions of physics − should be available for quaternions. Mostly this document shows in cookbook style how quaternion equations are equivalent to approaches already in use. Hope you enjoy. Since all of the tools used are woven from the same mathematical fabric. . the two will agree! Events are handled mathematically as 4−vectors. All the standard tools for creating mathematical patterns − multiplication. An Overview of Doing Physics with Quaternions It has been said that one reason physics succeeds is because all the terms in an equation are tensors of the same rank. infinite series. the interrelationships become more clear. trigonometric functions. Theorists in physics have been able to construct the most accurate models of Nature in all of science. This may lead to a better unification between the work of a theorist and the work of an experimentalist. They can be added or subtracted from another. transcendental functions. a research topic of much interest to me. A theorists builds a model to describe what patterns of events within a system might generate the experimentalist’s data set. I hope to bring the full power of mathematics down to the level of the events themselves. As Feynman pointed out. A theorist can import very powerful tools to generate patterns. The four Maxwell equations become one nonhomogeneous quaternion wave equation. "whatever we are allowed to imagine in science must be consistent with everything else we know. in essence. tensors of different rank can mix within the same equation." Fresh perspectives arise because. and the Klein−Gordon equation is part of a quaternion simple harmonic oscillator. With hard work and luck.

5 Mathematics .

D dc dt . where B = x I + y J + z K and similarly for D. dt c. Multiplication of quaternions is like complex numbers. Dense notation :−) . then these operators act on the scalar function c and the 3−vector function D in the following manner: d .D. b c. and B is the del operator. b I and d I become the 3−vectors B and D. B c. That is why quaternions in general do not commute. along with the divergence. a. a. outside + inside + the cross). dD dt c xD This one quaternion contains the time derivatives of the scalar and 3−vector functions. Note that the cross product would change its sign if the order of multiplication were reversed. ad bc For two quaternions.D. If a is the operator d/dt. or d/dx I + d/dy J + d/dz K (all partial derivatives). d ac bd. but with the addition of the cross product. the gradient and the curl. a D Bc BxD A mnemonic: (firsts − lasts. unlike the other terms.6 Multiplying Quaternions the Easy Way Multiplying two complex numbers a + b I and c + d I is straightforward. D ac B.

when multiplied by its corresponding vector. polar vector equations. Symmetry can often simplify the number of numbers actually needed to describe a tensor. There is a zero index for a scalar. that requires the intervention of a metric to broker the transaction. There are a variety of important spin−offs of a standard vector. I choose this convention. The vectors in question must have the same number of dimensions. It does not change at all under a rotation. so there is no way to know a priori how many numbers (or operators. zeros elsewhere. a tensor is .. There are ways to switch ranks. Tensors and All That According to my math dictionary. Physics equations involve tensors of the same rank. The result of general relativity for a spherically symmetric. A tangent vector is. non−rotating mass is the Schwarzschild metric. An outer tensor product is a way to increase the rank of tensors. An aside: mathematicians require the distance is non−negative. generate a real number. The number of numbers needed is the number of dimensions in the vector space raised by the rank. Dual vectors. or . axial vector equations. In looking at events in spacetime (a 4−dimensional vector). Metrics in math can be anything. A rank−1 tensor is a vector.) are in a vector. Since the same rank tensors are on both sides. Vectors have one index which can run from 1 to the number of dimensions of the field. If the space a vector lives in is shrunk. and becomes the Minkowski metric in the limit of the mass going to zero or the radius going to infinity. A vector can be viewed as the tensor product of a set of basis vectors. I will confine the discussion to the simplest tensors under rotational transformations. the components of the object undergoes a transformation of a certain nature.. but physicists do not.7 Scalars. and they would still be valid under the transformation. There are scalar equations. . −1 and zeros elsewhere. never more or less. It contains exactly one number. −1. If there are two vectors and one wants a result that is a scalar. the axioms of special relativity require the Minkowski metric. Vectors. the identity is preserved under a rotational transformation. but nature imposes constraints on which ones are important in physics. tangent to a vector function.. but to be consistent with everything else on this site. The tensor product of two vectors will be a 2−rank tensor. by systematically multiplying each component from the dual vector and the vector together and summing the total. which has "non−one" terms down the diagonal. "An abstract object having a definitely specified system of components in every coordinate system under consideration and such that.. Some people prefer the signs to be flipped. −1. which is the same as an identity matrix. A rank−0 tensor is known as a scalar. a contravariant vector shrinks. This process in known as an inner tensor product or a contraction. Another popular choice is the Euclidean metric. The metric defines how to form a scalar as the indices are examined one−by−one. well. and equations for higher rank tensors. which is a 4x4 real matrix which has down the diagonal 1. under transformation of coordinates. n−rank tensors have n indices. One could decide to arbitrarily combine tensor equations of different rank. A vector does change under rotation." To make this introduction less abstract. but a covariant vector gets larger. I will be using the physics notion of a metric.

You can convince yourself of this by noting that the cross product involves the sine of an angle and the dot product involves the cosine of an angle.F. because F would get smaller. t X The axial vector is the one that flips signs if the order is reversed. F t f XF What would happen to these five terms if space were shrunk? The 3−vector F would get shrunk. nothing in the tangent quaternion would change. as would the divisors in the Del operator. An infinite collection of quaternions of the form (En. This is called a pseudo−scalar.8 What Are Quaternions? Quaternions could be viewed as the outer tensor product of a scalar and a 3−vector.X". making functions acted on by Del get larger.X. The scalar terms are completely unaffected by shrinking space. because of the work of the Del operator in the divisor makes it larger. X . To those familiar to playing with higher dimensions. and the Del and F cancel each other.X polar vectors : X. because it does not change under a rotation. t X Xt XxX X t X xX X. Quaternions can only handle 3−vectors. Pn) could represent a quantum state. Note that if time where to shrink exactly as much as space. t X axial vectors : X x X Where is the axial vector for the left hand side? It is imbedded in the multiplication operation. The treatment of scalars is the same as above. honest :−) t. A quaternion equation must generate the same collection of tensors on both sides. If the order of quaternion triple product is changed.X. The time derivative of the 3−vector is a contravariant vector. but it is a covariant vector capacity because of the right−angle nature of the cross product. q and q’: t. A tangent quaternion is created by having an operator act on a quaternion−valued function . 0). The Hilbert space is formed using the Euclidean product (q* q’). The gradient of the scalar field is a covariant vector. Consider the product of two events.X . t f t f q .F q .X tt X. t X Xt Xt XxX scalars : t. X t. and an even function times an odd function is odd. Yet physics on both the quantum and cosmological scales is confined to 3−spatial dimensions.X . A dual quaternion is formed by taking the conjugate. this scalar will change signs for at each step in the permutation. because df/dt has nothing to shrink. Neither of these will change under a rotation. but the notion of vectors is far more restrictive. The curl at first glance might appear as a draw. Note that the infinite Hilbert spaces in quantum mechanics a function of the principle quantum number n. not the spatial dimensions. t . there will be terms of the form (XxX’). . this may appear too restrictive to be of interest. but it will change signs under a reflection. Terms can continue to get more complicated. due to the cross product. X t t X . but the 3−vector for space would be rotated. In a quaternion triple product.X tt t. tt X. Under rotation for an event in spacetime represented by a quaternion. because q* q = (t^2 + X. time is unchanged. as restrictive as the notion of scalars.

involve local transformations (but the technical definition of local transformation is different than the idea presented here because it involves groups). The important thing to know is that a transformation T necessarily exists because quaternions are a field. You can convince yourself of 9 this by noting that the cross product involves the sine of an angle and the dot product involves the cosine of an angle. but it can be done.X". that one T will work for any q. This is called a pseudo−scalar. Sometimes it takes a bit of effort. X . How can the metric be generalized for arbitrary transformations? The traditional approach would involve playing with Hamilton’s rules for the field.X . t 2 qq ’ qq ’ 2 q X. k must have inverses. In a quaternion triple product. but it will change signs under a reflection. Neither of these will change under a rotation. and an even function times an odd function is odd. t X q’ Xt XxX scalars : t tt X. because it means that inner and outer tensor products can occur in the same operation. Individual parts can be isolated if one chooses. This is an important observation. This is guaranteed to work locally. 2 polar vectors : X tX Xt qq ’ q q’ q’ . T will work for any q. and symmetric and antisymmetric products can isolate any particular term. and i. because it does not change under a rotation. Let T I. in fact it is equal to q’ q^−1. but it is more than that. Change the rule of what a quaternion is in one context and it will not be possible to compare it to a quaternion in another context. there will be terms of the form (XxX’). j. When two quaternions are multiplied together. I think that would be a mistake. Instead. There is no promise that it will work globally. within neighborhoods of q and q’. the identity matrix . due to the cross product. this scalar will change signs for at each step in the permutation. 2 2 i j k 2 ijk 1 This looks like a way to generate scalars from vectors. j k = i. It also says implicitly that i j = k. a new scalar (inner tensor product) and vector (outer tensor product) are formed. If the order of quaternion triple product is changed.X q tt q . It has been my experience that any tensor in physics can be expressed using quaternions. X t. since that rule involves the fundamental definition of a quaternion.X q’ . 2 2 q’ q qq ’ q’ q 4 qq ’ 2 q’ q axial vectors : X x X The metric for quaternions is imbedded in Hamilton’s rule for the field. general relativity and the standard model. Combinations of conjugation operators which flip the sign of a vector. Here are all the terms of the example from above t.Terms can continue to get more complicated. consider an arbitrary transformation T which takes q into q’ q q’ Tq T is also a quaternion. This quaternion definition of a transformation creates an interesting relationship between the Minkowski and Euclidean metrics. Under certain circumstances. The two most important theories in physics.

10 IqIq 2 Iq Iq IqIq t2 t2 X. To make calculus rigorous requires that it is defined over a mathematical field. vectors and tensors they work with are defined over the field of real or complex numbers. which is defined by the scalar energy and the 3−vector of momentum. To do calculus with a mass or an electromagnetic potential has an identical requirement and an identical solution. which consist of the scalar time and the 3−vector of space.X. Can physics be done without quaternions? Of course it can! Events can be defined over the field of real numbers. Physicists do this be saying that the scalars. So What Is the Difference? I think it is subtle but significant. the only fields that fit are isomorphic to the quaternions (isomorphic is a sophisticated notion of equality. Therefore a transformation which generates the Schwarzschild interval of general relativity should be easily portable to a Hilbert space. and that might be the start of a quantum theory of gravity. This is the logical foundation for doing physics with quaternions. Quantum mechanics can be defined using a Hilbert space defined over the field of complex numbers and return with every result measured to date. What are the numbers used by nature? There are events. Doing physics with quaternions is unnecessary.X. Constraining general relativity and quantum mechanics to work within the same topological algebraic field may be the way to unite these two separately successful areas. which has a scalar field phi and a 3−vector potential A. There is the electromagnetic potential. . and then the Minkowski metric and the Lorentz group can be deployed to get every result ever confirmed by experiment. 0 X. 0 In order to change from wrist watch time (the interval in spacetime) to the norm of a Hilbert space does not require any change in the transformation quaternion. To do calculus with only information contained in events requires that a scalar and a 3−vector form a field. unless physics runs into a compatibility issue. whose subtleties are appreciated only by people with a deep understanding of mathematics). It goes back to something I learned in a graduate level class on the foundations of calculus. only a change in the multiplication step. There is mass. According to a theorem by Frobenius on finite dimensional fields.

both types of inner and outer products will be formed and then related to physics. X t. t1=t2=delta t.X . The Grassman Inner and Outer Products There are two basic ways to multiply quaternions together.X Xt t. because it does not change signs if the terms are reversed. In this section. Each of these products can be separated into two parts: a symmetric (inner product) and an antisymmetric (outer product) components.. . The vector terms are a frame−dependent. X t. There are two basic ways to do this: just multiply one quaternion by another. but is still a valid measurement. X t. t X Xt XxX I call this the Grassman product (I don’t know if anyone else does. or first take the transpose of one then multiply it with the other. t. The symmetric component will remain unchanged by exchanging the places of the quaternions. X .X 2 X. X tt I have defined the anticommutator (the bold curly braces) in a non−standard way. X x X This is the cross product defined for two 3−vectors. Together they add up to the product.).X t. The Grassman outer product is antisymmetric and is formed with a commutator.11 Inner and Outer Products of Quaternions A good friend of mine has wondered what is means to multiply two quaternions together (this question was a hot topic in the nineteenth century). while the antisymmetric component will change its sign. even t. The first term would be the Lorentz invariant interval if the two quaternions represented the same difference between two events in spacetime (i. X t.X 2 t. t X t. The inner product can also be called the symmetric product.. t . t . The invariant interval plays a central role in special relativity. There is the direct approach. including a factor of two so I do not have to keep remembering to write it. odd t.X tt X. I care more about what multiplying two quaternions together can accomplish.e. It is unchanged for quaternions. X .X .. X t. symmetric product of space with time and does not appear on the stage of physics. X 0. but I need a label).

X t. The Euclidean inner product is also the standard definition of a dot product. X 0.X t. t X t. this is equivalent to multiplication by the conjugate which involves flipping the sign of the 3−vector. The Euclidean product might be a direct connection to the algebraic structure of quantum mechanics via Hilbert spaces. Yet. What is interesting to speculate about is the role of the Euclidean product. t.X 2 t. the Grassman inner product.X . X t. X t. Implications When multiplying vectors in physics.X .X t.X XxX Xt Form the Euclidean inner product. or cross product. The vector terms are an antisymmetric product of space with time and the negative of the cross product. because it naturally generates the invariant interval. The Euclidean product is non−associative: ab ab c c a bc a bc The norms of Euclidean products are associative because the norms are real valued: The Euclidean product of quaternions might be a way t connect to the algebra of octonions. appears to play a role in special relativity. X tt X.12 The Euclidean Inner and Outer Products Another important way to multiply a pair of quaternions involves first taking the transpose of one of the quaternions.a non− associative division algebra. . and the Grassman outer product. For a real−valued matrix representation. Form the Euclidean outer product. t. X tt t.X 2 t. t X Xt XxX The first term is zero. one normally only considers the Euclidean inner product.X X. 0 The first term is the Euclidean norm if the two quaternions are the same (this was the reason for using the adjective "Euclidean"). t. or dot product.

Consider a set of quaternions that represent events in spacetime. orthogonal basis for the 3−vector such that: . If dq is squared. and e3 be the shared. dq q’ a1 ’ q a0 ’ a0 e0 .. a3 ’ a3 e3 3 What is unusual about this definition are the factors of a third. Without this "scale" factor on the 3−vector. This idea can be defined precisely. e1. the differential element converged to a scalar.13 Quaternion Analysis Complex numbers are a subfield of quaternions. dq ^ 2 da0 2 e0 2 da1 2 e1 2 9 da2 2 2 da0 da1 e0 e1 . They will be necessary in order to define a holonomic equation later in this section.. 9 e2 e . It requires a speed less than the speed of light to connect the events. there is an important application in physics. a1 . in the limit. change in the scalar is not given its proper weight. Let e1. e2. a3 Imagine we want to do a simple binary operation such as subtraction. without having to specify the coordinate system chosen. Well. so that a transformation between two basis 3−vectors does not contain a hidden bump. Defining a Quaternion A quaternion has 4 degrees of freedom. Hamilton gave each element parity with the others. a2 ’ a2 e2 3. If quaternions would only commute. Should time be treated exactly like space? The Schwarzschild metric of general relativity suggests otherwise. I have found that it is important to give the scalar and the sum of the 3−vector parity. Let e0. Subtraction will only work if the coordinate systems are the same. a very reasonable approach. a1 e1 3. the scalar part of a quaternion does commute. e2. and e3 form an independent. This might initially appears as an unreasonable constraint. spherical or otherwise. so it needs 4 real−valued variables to be defined: q a0 . basis. All that is required is that the magnitude of the 3−vector goes to zero faster than the scalar. 2 da0 da3 e0 3 3 3 e2 2 9 da3 2 What should the connection be between the squares of the basis vectors? The amount of intrinsic curvature should be equal. a2 . whether it is Cartesian. My hypothesis is that complex analysis should be self−evident within the structure of quaternion analysis. If the magnitude of the 3−space vector is less than the time scalar. However. This is true no matter what coordinate system is chosen. events are separated by a timelike interval. but unspecified. the scalar part of the resulting quaternion forms a metric. The challenge is to define the derivative in a way so that a left derivative always equals a right derivative. then it would commute. dimensionless. If. Now we can define the difference between two quaternion q and q’ that is independent of the coordinate system used for the measurement. 2 da0 da2 e0 3 e3 2 .

q q. q*. 2 da0 . 2 da0 3 3 3 The scalar part is known in physics as the Minkowski interval between two events in flat spacetime. To completely specify a quaternion function on the manifold 1 . calculate the square of dq: dq2 2 da0 da0 2 e0 2 da1 2 9 e0 2 da2 2 9 e0 2 da3 2 . and q*2 all but the e2 term. a3 e3 The conjugate as it is usually defined (q*) flips the sign of all but the scalar. a2 e2 a2 e2 3. The set q. Set e0 2 1 2 = (1 − 2 GM/c^2 R). Physics guides my choices in mathematical definitions :−) An Automorphic Basis for Quaternion Analysis A quaternion has 4 degrees of freedom. k if e0 2 = 1. 3. i.14 1 e1 2 1 e2 2 1 e3 2 e0 2 This unusual relationship between the basis vectors is consistent with Hamilton’s choice of 1. 3. a1 e1 3. Notice that the 3−vector is unchanged (this may be a defining characteristic). a3 e3 a3 e3 3 3 3 e1 q e1 e2 q e2 a0 e0 . it must also have four degrees of freedom. 9 e0 2 da1 da2 da3 . a1 e1 3. For that case. then the metric would apply to a non−flat spacetime. q 1 1 q. The conjugate of a conjugate should give back the original quaternion. q*1. j. q*2 form the basis for quaternion analysis on the 1 manifold. If e0^2 does not equal one. There are very few opportunities for freedom in basic mathematical definitions. A metric that has been measured experimentally is the Schwarzschild metric of general relativity. da1 da2 da3 . 2 da0 3 3 3 This is the Schwarzschild metric of general relativity. The q*1 flips the signs of all but the e1 term. a2 e2 3.dA 2GM 9 1 c2 R . q 2 2 q Something subtle but perhaps directly related to spin happens looking at how the conjugates effect products: q q’ q’ q . I have chosen this unusual relationships between the squares of the basis vectors to make a result from physics easy to express. 2 da0 . Three other linearly−independent variables involving q can be defined using conjugates combined with rotations: q q q 1 2 a0 e0 . a1 e1 a0 e0 . and calculate the square of dq: dq2 2 da0 da0 2 1 2GM c2 R dA.

I concentrate on the ratio of scalars to 3−vectors. a3 can be represented by functions using the conjugates as a basis. Implementing these requirements involves two limit processes applied sequentially to a differential quaternion D. q 2 2 1 f q. q . f q. q 1 . lightlike past. and they will each have distinct properties. q 1 . A tighter link is explored in the section on integral and half integral spin. q d. The net effect of these two limit processes is that D−>0. timelike future. Each region will require a separate definition of the derivative. The property of going "half way around" is reminiscent of spin. 1 q q’ 1 2 q’ 2 q 2 q q’ q q’ q’ q q’ 1 q 1 The conjugate applied to a product brings the result directly back to the reverse order of the elements. where the ratio of the change in time to the change in 3−space defines five separate regions: timelike past. then this definition cannot be used. (D − D*)/2 −> 0. in other words. D The definition is invariant under a passive transformation of the basis. A future timelike derivative to be defined. Definition: The future timelike derivative: Consider a covariant quaternion function f with a domain of H and a range of H. q f f a1 a2 2 a0 q q e0 q q 2 q 1 e1 2 3 q 2 e2 2 3 e1 q q 1 2 3 e2 q q 2 2 3 . (D + D*)/2 −> +0 (the plus zero indicates that it must be approached with a time greater than zero. and spacelike. Future Timelike Derivative Instead of the standard approach to quaternion analysis which focuses on left versus right derivatives. D d. This is natural when thinking about the structure of Minkowski spacetime. q . q . The 4 real variables a0. q q limit as d. The first and second conjugates point things in exactly the opposite way. lightlike future. I will start with the simplest case. D . the limit of the scalar is taken.15 q q’ 1 q’ 1 1 q 1. f q. Second. q 1 . 0 f q d. from the future). q 1 . a1. First the limit of the three vector is taken as it goes to zero. 0 2 0 limit as d. and look at a series of examples in detail. the 3−vector must approach zero faster than the positive scalar. q . a2. If this is not the case. There are no continuous Lorentz transformations to link these regions.

D 1 . D q 3 e1 2 1 q q 1 a1 q a1 q 2 0 So far. D q 1 d. D e1 2 3 d. q 1 . q . q . where the e1 is written on the right. D q q q lim lim 1 2 d. A simple approach to a trickier example: f a1 a1 q a1 q 1 e1 q q 1 2 3 lim lim 2 3 e1 d. D a0 q 1 a0 q 2 0 The definition gives the expected result. q 1 . the fancy double limit process has been irrelevant for these identity functions. q a0 q a0 q 2 a0 e0 q q 2 e0 e0 2 q d. That changes with the following example. but the result is the same. D d. because the differential element has been eliminated. D q 1 1 q q 1 2 3 d. f q. q a1 q a1 q 1 q 2 a1 q q 1 e1 2 3 lim lim e1 lim lim d.16 f a3 e3 q q q 2 3 1 q 2 q q q 1 q 2 3 2 e3 Begin with a simple example: f q.

D 1 lim lim q2 q d. The following table describes the 16 derivatives for this set . D q d. 0 e1 2 3 d. D 2q d. D 2 q2 d. because if the 3−vector D goes to −D nothing will change about the quaternion derivative. 0 d. D 2 q2 d. D q 1 d. It would be straightforward to show that all polynomials gave the expected results. D This is the expected result for this polynomial. the remaining differential is a scalar so it commutes with any quaternion. which is: z b z a b i. This is what is required to dance around the e1 and lead to the cancellation. For timelike separated events. D 1 lim lim q lim 2 q d. The derivative of a quaternion applies equally well to polynomials. so a timelike derivative should not depend on the direction of the 3−vector. So this quaternion derivative should match up with the complex case. b i 2 b z Z Z 2i These are the same result up to a factor of three. after the first limit process. let f f q q2 lim lim q d. 0 1 3 e1 2 Because the 3−vector goes to zero faster than the scalar for the differential element. right and left depend on the inertial reference frame. The initial hypothesis was that complex analysis should be a self−evident subset of quaternion analysis. Analytic Functions There are 4 types of quaternion derivatives and 4 component functions. This is actually consistent with principles of special relativity.17 lim d. D d. Mathematicians might be concerned by this result. Quaternions have three imaginary axes.

The presence of the three basis vectors changes things slightly. . Another test involves the Cauchy−Riemann equations. a1 e1 e e . Let w q a0 e0 . Let’s see if the identity function w = q is analytic. a2 2 .18 \ q q q q 1 a0 e0 2 e0 2 a1 e1 2 3 a2 e2 2 3 a3 e3 2 3 e3 2 3 e3 2 3 e3 2 3 0 e1 2 3 0 0 e2 2 3 0 0 2 0 This table will be used extensively to evaluate if a function is analytic using the chain rule. a0 q a1 q 1 a2 q 2 a3 q a3 q 1 a3 q 2 1 2 1 2 1 2 1 2 1 2 1 2 0 0 0 This has the derivatives expected if w=q is analytic in q. a3 3 3 3 3 1 2 Use the chain rule to calculate the derivative will respect to each term: w a0 w a1 w a2 w a3 w q 1 2 w q w q 1 w q 2 a0 q a1 q a2 q a3 q w a0 1 2 w a0 w a1 w a2 e0 e1 3 e2 3 e3 3 a0 q 1 2 e0 2 e1 2 3 e2 2 3 e3 2 3 w a1 1 2 1 w a3 w a3 w a3 a1 q 1 2 1 2 1 2 w a2 a2 q w a3 a3 q Use combinations of these terms to calculate the four quaternion derivatives using the chain rule.

a1 e1 e e . . V V e . 0. Consider the identity function w = q*1. 3 3 u a0 u e2 a0 3 e1 3 V e . 0. a2 0 a3 e3 3 u e3 a0 3 V e a3 0 This also solves a first conjugate holonomic equation. One noticeable change is in the role that the conjugate play for the basis vectors. e3 . 0. a1 0 V e . 9 9 9 e1 e e . 0. e2 . a3 3 3 3 3 V e a3 0 V a3 0 This also solves a holonomic equation. 0 . 2 a0 a2 e0 2 . a2 0 u . e1 . Let u a0 e0 . a1 0 V e . e3 There are no off diagonal terms to compare. This exercise can be repeated for the other identity functions. a2 2 . e1 . a2 2 . a0 e1 e 3 1 V . a1 1 . To show that this is analytic in q*1 requires that one always works with basis vectors of the q*1 variety. 0 . Scalar e0 e0 e0 . a1 e2 e 3 2 u e3 a0 3 V . a3 a2 a1 a0 e1 e2 e3 e1 e2 e3 0 3 3 3 e1 2 e 2 e 2 a2 2 2 a3 2 3 . e2 . 0. a2 e3 e 3 3 0. Scalar e0 e0 V V V u . 0. 0 9 . e0 .19 Let u u e1 a0 3 u e2 a0 3 a0 e0 . e e V 0. 2 a0 a3 e0 3 3 3 3 a0 2 e0 2 a1 2 a1 2 1 Power functions can be analyzed in exactly the same way: Let w q2 2 a0 a1 e0 u a0 e0 2 2 e1 2 9 a2 2 e2 2 9 a3 2 e3 2 .

This also solves a holonomic equation. then the 3−vector. 2 a0 a1 e0 u e1 a0 3 u e2 a0 3 u e3 a0 3 u e a1 0 u e a2 0 u e a3 0 e1 e e . The only change is in the limit. a0 V . e1 . Therefore the order of the limit processes is reversed. The simplest next case is for past timelike derivatives. e3 2 a0 e0 e3 e3 3 0 2 a0 e0 e1 e1 3 2 a0 e0 e2 e2 3 Since power series can be analytic. Scalar 2 a0 e0 3 u . Actually they are. At first glance. This time the scalar approaches zero. 2 a0 a3 e0 3 3 3 3 V e a1 0 V e a2 0 V a3 V1 e1 a0 3 V2 e2 a0 3 V3 e3 a0 3 2 a0 e0 2 e1 3 2 a0 e0 2 e2 3 2 a0 e3 2 3 2 a1 e0 e1 2 9 2 a2 e0 e2 2 9 2 a3 e0 e3 2 9 This time there are cross terms involved. This creates a problem. 2 a0 a2 e0 2 . Note that these are three scalar equalities. where the scalar approaches zero from below. most quaternions. this work has only involved future timelike derivatives. a2 V a3 e0 . (I have done the case for the cube of q. e2 . changes in time go to zero faster than the absolute value of the 3−vector. but they are not. one might think these are incorrect. This represents four constraints on the four degrees of freedom found in quaternions to find out if a function happens to be analytic. which is the case for most laws of physics. and it too is analytic in q). This will make many derivatives look time symmetric. 4 Other Derivatives So far. since the signs of the derivatives are suppose to be opposite. The way around this is to take its norm.20 V 0. this should open the door to all forms of analysis. The derivative of the first component of V with respect to a0 is positive. but it is hidden in an accounting trick :−) For example. The other Cauchy−Riemann equations evaluate to a single 3−vector equation. Keeping all the information about signs in the e’s makes things look non−standard. . which makes it negative. There are five other regions of spacetime to cover. A more complicated case involves spacelike derivatives. In the spacelike region. the derivative of u with respect to a1 has a factor of e1^2. That will lead to the differential element not cancelling. which is a scalar. a1 V .

q 1 . D . B d. B a. B ’ 0. To make the definition non−singular everywhere. D d. this is an arbitrary rotation. D 0. f q. q q limit as 0. q To make this concrete. D 2 1 1 0. In the spacelike region. D 2 0. q . Therefore the order of the limit processes is reversed. D 2 0. D 0. B norm 0. changes in time go to zero faster than the absolute value of the 3−vector. q f q 2 2 d. q . Define: a. D The second and fifth terms are unitary rotations of the 3−vector B. B a. and continue: 0. B lim a. A spacelike differential element is defined by taking the ratio of a differential quaternion element D to its 3−vector. B 2 2 0 limit as d. D 0. q 1 . To be defined. multiply by the conjugate. q 2 1 2 d. D f q. q . D a. D a. which is a scalar. B 0. D f q d. q 1 . f = q^2. Since the differential element D could be pointed anywhere. the three vector must approach zero faster than its corresponding scalar. In the limit D D*/((D − D*)(D − D*))* approaches (1. D norm 0. Let the norm of D approach zero. D 0. B ’ a. This time the scalar approaches zero. D 0. The way around this is to take its norm. B a. D a. which will not commute with most quaternions. B a. D 1 2 f q. D a. D f q. q q 0 limit as d. 0). the differential element is (0. D a. Apply the definition: Norm q2 q limit d. q . D .21 A more complicated case involves spacelike derivatives. q 1 . This creates a problem. q 1 . D 0. q . D a. D norm 0. q . B ’ Substitute. D . because after the first limit process. q d. That will lead to the differential element not cancelling. B 0. B d. then the 3−vector. D). a scalar. consider a simple example. q 1 . D lim a. D − D*.

then it will be an equality. leading to a destruction of a contribution from the 3−vector. Defining the lightlike derivative. and could be used to define a Banach space. The spacelike derivative is positive definite. The spacelike derivative can therefore interfere with itself. where the change in time is equal to the change in space.B ’ . This is quite a natural thing to do in quantum mechanics. will require more study.B 2q 2 2 D. The timelike quaternion derivative on a quaternion manifold is effectively a directional derivative along the real axis The spacelike derivative is a normed derivative The dual limit definition establishes the link between these two well−known types of derivatives.B ’.B 2 B.22 lim 4 a ^ 2 4 a^2 2 B. but it will require some skill to find a technically viable compromise between the spacelike and timelike derivative to synthesis the lightlike derivative. It may turn out that this derivative is singular everywhere.B ’ 2 D. 0 Look at how wonderfully strange this is! The arbitrary rotation of the 3−vector B means that this derivative is bound by an inequality. but D could also be in the opposite direction. 0 2 B.B 2 B. . If D is in direction of B.

The entire set H and the empty set are in T T is the topology on H. 0 An open set in H is any set which can be expressed as a union of open balls. topology is the structure of relationships that do not change if a space is distorted. . so statements and proofs involving H are often identical to those for R. makes exploring these decisions fun (for those few who can understand what is going on :−) By working with H. Y) [note: small letters are scalars. An important component to standard approaches to general relativity is product spaces. the other as an interval metric. Time is the scalar. To keep the same word with two meanings distinct. Quaternions form a topology because they are what mathematicians call a metric space.T) consists of the set H together with a collection T of subsets of H with these properties: 1. since q* q evaluates to a real positive number or equals zero only if q is zero. I will work consistently with the set of quaternions. each in T. one place for time. I will refer to one as the topological metric. First an open ball of quaternions needs to be defined to set the stage for an open set. Events in spacetime require R^4. Define an open ball in H of radius (r. 0) centered around a point (y. H^1. The subsets of H in T are open sets. Mathematicians get to make choices: what would change if work was done in R^2. X Y r. or R^5? The precision of this notion. Topological Spaces". These differences are not important for basic topological properties. In this section. space the 3−vector. The difference between the real numbers R and H is that H is not a totally ordered set and multiplication is not commutative. There is no choice to be made. product spaces are unnecessary. the Minkowski metric can be negative or zero even if a point is not zero. together with the freedom to make choices. Some of the results of topology are required to make calculus rigorous. capital letters are 3−vectors] consisting of points (x. three for space. the starting point is "Appendix A. The intersection of a finite number of subsets of T is in T 3. X) such that x y. These descriptive labels are not used in general since context usually determines which one is in play. X Y x y. Note: this is not the meaning of metric used by physicists. Roughly. The union of an arbitrary collection of subsets. or just H for short. is in T 2. For example. 423 translated] A quaternion topological space (H. Events in spacetime can be members of an open set in H.23 Topological Properties of Quaternions Topological Space If we choose to work systematically through Wald’s "General Relativity". R^3. [p. This is how a topology for R^n is created.

I am a practical guy who likes pragmatic definitions. open and closed sets. t) where x and a are elements of A. 0 a half open set : or L. 0 timelike events : scalar lightlike events : scalar a . 0 L. For the quaternion topological space (H. Define a point set as the set where the lower bound equals the upper bound. 0 0. 0 x x x a a x x x a U. What are the implications for physics? With quaternions. The closure of A equals A if and only if A is closed. A point set for the real numbers has only one element which is identical to the boundary. p. and the interior and closure of a set. 0 U. The intersection of an arbitrary number of closed sets is closed. 0 0. one of them identical to the boundary. 0 a a x a U. 0 a boundary : x a L. The union of a finite number of closed sets is closed. Use inequalities to define: an open set : a closed set : L. The only open set that is a point set is the null set. 0 U. The intersection of a finite number of open sets is open.24 Open Sets The edges of sets will be examined by defining boundaries. 0 x a x x a x 2 2 a U. Let the real numbers L and U represent arbitrary lower and upper bounds respectively such that L < U. The closure of A is the intersection of all closed sets containing A. an open set : L. A point set for quaternions has an infinite number of elements. T). consider an arbitrary induced topology (A. The interior equals A if and only if A is open. [Wald. the existence an open set of events has nothing to do with the causality of that collection of events. Clearly there are connections between the above definitions open set union boundary This creates complementary ideas.424] closed set The interior of A is the union of all open sets contained within A. The closed point set is H. 0 a x a x a The union of an arbitrary collection of open sets is open. 0 L.

0 x a x 2 and scalar x a 0 a U. 424] Let A be a subset of the quaternions H. For example. x E H. First we need the definition of a compact set of quaternions. so their intersection is null. Hausdorff Topology This property is used to analyze compactness. closed or on a boundary. 0 . A subset of {Oa} that still covers A is called a subcover. x E H. I hope to share some of my insights into this subtle but significant topic. The set S could depict a classical world history since they are causally linked and have good topological properties. Let’s find an example of a compact set of quaternions. closed or neither. T) is Hausdorff because for each pair of distinct points a. find the half−way point between a and b. The set A subset of H is compact if every open cover of A has a finite subcover. Consider a set S composed of points with a finite number of absolute values: . one can find open sets Oa. b is fixed. A closed set of lightlike events could be a focus of quantum electrodynamics. There is no correlation the reverse way either. I will begin an investigation of compact sets of quaternions. Let that be the radius of an open ball around the points a and b: let r. whether open. 0 A proper time can have exactly the same absolute value as a pure spacelike separation. Ob E T such that a E Oa.25 spacelike events : scalar x a 2 0. something vital for rigorously establishing differentiation and integration. An open cover of A is the union of open sets {Oa} that contains A. not staying within limits. so these two will be included in the same sets. [Wald p424] The quaternion topological space (H. r E R b. [Translation of Wald p. a fixed. a E H. Compact Sets In this section. Take for example a collection of lightlike events. a is fixed. b I Ob and the intersection of Oa and Ob is the null set. If the subcover has a finite number of elements it is called a finite subcover. b E H. U. S x. a not equal to b. A union of open sets is open and could have an infinite number of members. Topology plus causality could be the key for subdividing different regions of physics. Although independent. LER L. Even though they all share exactly the same interval − namely zero − their absolute value can vary all over the map. Set A could be opened. Consider an open set S of timelike intervals. these two ideas can be combined synergistically. 0 Oa Ob a b a b 4 a b x x a b x x r r a. r E R Neither set quite reaches the other.

. 0 x x b. 0 . T) be Hausdorff and let the set A subset of H be compact. [Wald p. and will be examined next A compact set is a statement about the ability to find a finite number of open sets that cover a set. xn E H. n is finite x1 x1 ^ . For quaternions with the standard topology. Notice that the set S is closed because it consists of a boundary without an interior. [Wald p. Then A is closed. T’) be topological spaces. to be closed and bound is to be compact. T) is compact and the function f: H −> H’ is continuous. closed and bound set is important. specifying the absolute value still leaves three degrees of freedom (if the domain had been x E R. The set S can be covered by an open set {O} which could have an infinite number of members. an E R.. A closed interval of quaternions S: S x E H.26 S x1. 0 . The f[H] = {h’ E H’ | h’ = f(h)} is compact. the set must necessarily include its boundary and be bound. The property of compactness is easily proved to be preserved under continuous maps. to be compact is to be closed and bound. T) and (H’.. x E R.. A subset A of quaternions is compact if and only if it is closed and bounded. 0 with the standard topology on H is compact. However. A closed set is the interior of a set plus the boundary of that set.. since for any of the equalities. Theorem 5. Let the topology (H. given any open cover..5 a2. Invariably the Heine− Borel Theorem employs the domain of the real numbers. then S would have had a finite number of elements).. a2. The subset {C} would have one member for each absolute value.5 a1. Let (H. A set is bound if there exists a real number M such that the distance between a point and any member of the set is less than M. 425] Theorem 2. . . in order to have a finite number of open sets that cover the set. This creates a corollary by theorem 4. x2 x2 ^ . Theorem 4. The link between compact.. one y exists for each inequality Every set of quaternions composed of a finite number of absolute values like the set S is compact. Theorem 3. There exists a subset {C} of {O} that is finite and still covers S. . In other words. a2 e y y a2 e. 425] Theorem 1 (Heine−Borel). Let the topology (H. .. C y E O . T) be compact and let the set A subset of H be closed. a b a. Then A is compact. 0 . The set S has an infinite number of members. a. nothing in that proof changes by using quaternions as the domain. . 0 . Suppose (H. e 0 a1 e y y a1 e. b E R.. x2. a1. Combine these theorems to create a stronger statement on the compactness of subsets of quaternions H. Wald does not provide a proof since it appears in many books on analysis.. e E R.

specifically Tychonoff’s Theorem.27 Theorem 6. . Wald continues with theorems on product spaces. 4 does not come equipped with a rule for multiplication. [end translation of Wald] 1 versus n It is important to note that these theorems for quaternions are build directly on top of theorems for real numbers. In particular. so it is qualitatively different from H 1 . the product space 4 should have the same topology as the quaternions. A continuous function from a compact topological space into H is bound and its absolute value attains a maximum and minimum values. subtlety matters in the discussion of foundations. Only the domain needs to be changed to H 1 . even if topologically similar to the quaternions. 1 . Hopefully. A similar issue arises for 2 and the complex number manifold 1 . so that the above theorems can be extended to n .

and keep things looking simpler.V 2 . q ’ qq ’ tt ’ V.V . Parts scalar q vector q Simple Algebra q q q q 2 2 t. V t. 0 0. 0 t2 V. Grassman_product q. V t2 V. qq V t t t ’.V qq V norm q The Grassman product as defined here uses the same rule Hamilton developed. Technical note: every tool in this set can be expressed as working with a whole quaternion q. V q norm q det q qq qq qq q q q q q 2 t2 t2 V. This is to show how to work with automorphic functions on a quaternion manifold. It should look very similar to complex algebra. and t + z k. All these relations have been tested in a C library and a Java quaternion calculator. as in (t. t + y j. q ’ dif q. V t ’. The Euclidean product takes the conjugate of the first of the two elements (following a tradition from quantum mechanics).V. t + x i.0 V’ V’ sum q. t.28 A Quaternion Algebra Tool Set Here is a compilation of basic algebra for quaternions.V ’. 0 V. q ’ conj q inv q adj q Multiplication q’ q’ t. To strengthen the link. V). all quaternions have been written as a pair of a scalar t and a 3−vector V. tV ’ Vt ’ V V’ . since it contains three sets of complex numbers.

q ’ Euclidean_product q. q ’ qq ’ q ’ q tt ’ V. sinh t sin cosh q q^2 1 ^ .5 ln . asin q acos q atan q V V V V V V V V asinh q V acosh q atanh q V V V V V V V V V sinh q sinh t cos cosh q cosh t cos tanh q asinh q acosh q atanh q sinh q ln q ln q . V V ’ Euclidean_odd _product q.29 Grassman_even _product q. q ’ q q’ q’ q tt ’ V.5 1 q q^2 1 q . cos t sinh . q ’ q q’ tt ’ V. 0 2 Vt ’ qq ’ 2 V V’ q’ q 0. q ’ q q’ q’ q 0. tV ’ 2 Grassman_odd _product q. cosh t sin . tV ’ Vt ’ V V ’ 2 Trigonometry sin q sin t cosh cos q cos t cosh tan q sin q V V . these two commute so the order is irrelevant.V ’. sin t sinh V V V V V V cos q Note: since the unit vectors of sine and cosine are the same.V ’.5 1 ^ . tV ’ Vt ’ Euclidean_even _product q.V ’.

q’ Andrew Millard suggested the result for the Grassman product.30 Powers exp q exp t cos q^q’ Logs V . q ’ q. q ’ q q’ q . q ’ q . .V . atan2 V . q’ q .5 ln t ^ 2 ln q V. q ’ 2 q. q’ exp pi q .t V V ln 10 Quaternion Exponential Multiplication q q’ q. exp t sin V V V exp ln q x q ’ ln q log q 0. q ’ 2 exp pi q.

31 Classical Mechanics .

0 . One way that can be achieved algebraically is by having a time operator act only on space. R 1. r Sin Θ . Newton’s 2nd Law for an Inertial Reference Frame in Cartesian Coordinates Define a position quaternion. A d . R Operate on the velocity to get the classical inertial acceleration quaternion.0 dt d . R t. 0 r Sin Θ Θ. the first term is zero. r Sin Θ . r Sin Θ r Cos Θ Θ. the law is classical. r Cos Θ The velocity in a plane. but this time start with polar coordinates. V t.32 Newton’s Second Law The form of Newton’s second law for three separate cases will be generated using quaternion operators acting on position quaternions. 0 d . V t. r Cos Θ r Sin Θ Θ. r Sin Θ r Cos Θ Θ. Newton’s 2nd Law in Polar Coordinates for a Central Force in a Plane Repeat this process. r Cos Θ . R This is the standard form for acceleration in Newton’s second law in an inertial reference frame. 0 Acceleration in a plane. R 0. r Cos Θ .0 dt 1. R t. I call this the "2 zero" rule: if there are two zeros in the generator of a law in physics.0 dt Operate on this once with the differential operator to get the velocity quaternion. A 1. In classical mechanics. time and space are decoupled. R d .0 dt 1. or by space operator only act on a scalar function. Because the reference frame is inertial.

2 r Sin Θ Θ r Cos Θ Θ 2 2 r Cos Θ r Sin Θ Θ. The differential time operator is put into the first term of the quaternion. 0 Not a pretty sight. V d . For a central force. and azimuthal alterations respectively.33 0.R Ω ΩxR The first three terms of the 3−vector are the translational. R Sin Θ . 0. What this implies is not yet clear.Ω dt Ω. 0. A Ω. then Ω R iff Ω. 2 r Cos Θ Θ r Sin Θ Θ r Sin Θ r Cos Θ Θ.R .R Ω Ωx ΩxR Ωx ΩxR 2 Ω R 2 If the angular velocity an the radius are orthogonal. Sin Θ 2 d . Make these substitution and rotate the quaternion to get rid of the theta dependence. Rotating Frame Consider a noninertial example. Newton’s 2nd Law in a Noninertial.R . and the three directions for the angular speed are put in the next terms. with the frame rotating at an angular speed omega.0 t. 2Lr .Ω dt d .R 0 The scalar term is not zero. and Θ = 0. R 2ΩxR ΩxR Ω. but it may be related to the fact that the frame is not inertial. coriolis. 0 dt 0. r Cos Θ . Unlike the previous examples where the time t did not interfere with the calculations. . R 0.R . This quaternion is then multiplied by the position quaternion to get the velocity in a rotating reference frame. here the time t must be set explicitly to zero (I wonder what that means?). R Ω. L2 m2 r3 r. but using a vector identity it can be rewritten: Ω. A Cos Θ . 0 m r2 The second term is the acceleration in the radial direction. The last term of the 3−vector may not look like the centrifugal force. Θ = L mr2 . R ΩxR Operate on the velocity quaternion with the same operator. the third is acceleration in the theta direction for a central force in polar coordinates.

34

Implications Three forms of Newton’s second law were generated by choosing appropriate operator quaternions acting on position quaternions. It is impressive that complicated expressions in Newtonian mechanics can be encapsulated in quaternion one−line formulas. The differential time operator was decoupled from any differential space operators. This may be viewed as an operational definition of "classical" physics.

35

Oscillators and Waves
A professor of mine once said that everything in physics is a simple harmonic oscillator. Therefore it is necessary to get a handle on everything. The Simple Harmonic Oscillator (SHO) The differential equation for a simple harmonic oscillator in one dimension can be express with quaternion operators.
2 d ,0 0, x, 0, 0 dt kx d2 x , 0, 0 0, 2 m dt

0, 0

k x, 0, 0 m

This equation can be solved directly.

x

C 2 Cos

k t m

C 1 Sin

k t m

Find the velocity by taking the derivative with respect to time.

k C 1 Cos x m

k t m

k C 2 Sin m

k t m

The Damped Simple Harmonic Oscillator Generate the differential equation for a damped simple harmonic oscillator as done above.
2 d ,0 0, x, 0, 0 dt d , 0 0, b x, 0, 0 dt

0,

k x, 0, 0 m 0

0,
Solve the equation.

d2 x dt2
bm

bdx dt

kx , 0, 0 m

x

C 1 E

4 k m b2 m2 2m

t

bm

C 2 E

4 k m b2 m2 2m

t

36

The Wave Equation Consider a wave traveling along the x direction. The equation which governs its motion is given by

d d , , 0, 0 v dt dx 0, 0, d2 dx2

2

0, 0, f t v d2 dt2 v2 f tv

x , 0 x , 2 d2 f t v x dt dx v

The third term is the one dimensional wave equation. The forth term is the instantaneous power transmitted by the wave. Implications Using the appropriate combinations of quaternion operators, the classical simple harmonic oscillator and wave equation were written out and solved. The functional definition of classical physics employed here is that the time operator is decoupled from any space operator. There is no reason why a similar combination of operators cannot be used when time and space operators are not decoupled. In fact, the four Maxwell equations appear to be one nonhomogeneous quaternion wave equation, and the structure of the simple harmonic oscillator appears in the Klein−Gordon equation.

and the path integral between any two points is the same no matter what the path chosen. d . all closed path integrals are zero. 2 k y z. 0 The curl is the commutator of the differential operator and the force. a potential function whose gradient is the force exists. 0.F 0 Let the differential operator quaternion act on the force. If this is zero. F F dt. dz k x dx k x dt k x2 2 ktx k y dy. dx. 0. 1. quaternion operators perform these tests on quaternion−valued forces. F 0. 0 k y2 . the force is conservative. equivalent tests to determine if a force is conservative: the curl is zero. There Exists a Potential Function for the Force Operate on force quaternion using integration. If the field quaternion is the same. and test if the vector components equal zero. dt . In this section. The Curl Is Zero To make the discussion concrete. the force is conservative. dy. k y dz. k y. k y dt k x dz. dt F 2 k. 0 2. d .37 Four Tests for a Conservative Force There are four well−known. Take the negative of the gradient of the first component. k x. 0 . kty k x z. define a force quaternion F.

0 0 F dt 0 4. .0 dt 2 2 0. 0. k x. The Line Integral Along Different Paths Is the Same Choose any two parameterizations from A to B. 0 This is the same force as we started with. so the scalar inside the integral is the scalar potential of this vector field. One new avenue opened up is for doing path integrals. making sure it comes back to go. r Cos t . They are zero if t=z=0. The Line Integral of Any Closed Loop Is Zero Use any parameterization in the line integral. r Sin t . 2 r. r. path 2Π 0. 0) to (0. if any. tr r. and test that they are the same. k y. −r. 2 r Sin 2 k r2 0. is unknown to me. t r. These paths are from (0.38 d k x2 k y2 . What role these vector terms in the potential quaternion may play. The vector terms inside the integral arise as constants of integration. 0). path1 2Π 0. 0 t . r Cos t . It would be interesting to attempt four dimensional path integrals to see where that might lead! . 3. 0 2 dt 0 path2 2 F dt 0 2 k r2 The same! Implications The four standard tests for a conservative force can be done with operator quaternions.

39 Special Relativity .

and do not have any fundamental property which distinguishes them from other hyper−complex numbers. Also. 1952. For fixed collinear coordinate systems: . 1945. but De Leo finally figured out a real quaternion representation of the Lorentz group (S. why not do the 4D transformations of special relativity? It turns out that the unit quaternions must be complex−valued." J. Cayley had observed back in 1854 that rotations in 3D could be achieved using a pair of unit quaternions having a norm of one: q’ a q b where a a b b 1 If this works in 3D space. and must be restricted to graft the results to the Lorentz group. and I do not deny the validity of any of it. The approach is too general. C.. Rotation + Dilation Multiplication of complex numbers can be thought of as a rotation and a dilation. embodied in the work of Riemann.com). Biquaternions do not satisfy this axiom.M. 50. I found his approach hard to follow. De Leo. Royal Irish Academy A. but it could make an interesting discussion :−) Eventually. Möller wrote a general form for a Lorentz transformation using vectors ("The Theory of Relativity". Math. Phys. eq. or biquaternions. A. Einstein understood general transformations. Dirac took a Mobius transformation from complex analysis and tried to develop a quaternion analog.A. such as Clifford algebras. An additional dilation term might allow quaternions to do the necessary work. 25)." Just for the record: plenty of fine work has been done with biquaternions. which made for boosts without any terms from the cross product. 1996). QC6 F521. Conway and L. which is rather too many for a simple scheme for describing quantities in space−time. 37(6):2955−2968. well enough to formulate general relativity. p. I am making a choice to focus on quaternions for reasons outlined by Dirac. Silberstein proposed a different mathematical structure behind special relativity in 1911 and 1912 respectively (a copy of Silberstein’s work is available at quaternions. He had to rely on his old math teacher Minkowski to learn the theory of transformations (I do not know the details of Einstein’s education. Dirac (Proc. Much effort has been directed toward "other hyper−complex numbers". I needed something simpler :−) It was quite the wait. He defined an operator he called "bar" which multiplied a quaternion by two quaternions on either side. It definitely is an approach that works. Einstein proposed the principles of special relativity without a deep knowledge of the mathematical structure behind the work. Is this so bad? Let me quote P. W. "Quaternions and special relativity.40 Rotations and Dilations Create the Lorentz Group In 1905. He effectively did a commutator of this bar operation. 261): "Quaternions themselves occupy a unique place in mathematics in that they are the most general quantities that satisfy the division axiom−−that the product of two factors cannot vanish without either factor vanishing. they have eight components. Conway and Silberstein’s proposals only have the rotation component albeit using a complex number.

X V V 2 ΓtV Γ V. then X t ΓX Γt ΓtV Γ V. V. . even q. Implications Is this result at all interesting? A straight rewrite of Möller’s equation would have been dull. q’ q Γ Γ 1 even even V . Notice how the Lorentz transformation depends linearly on q.X i yj zk The additional complication to the X’ equation handles velocities in different directions than i.X .X. V V 2 Γ even t.X 1. 0. There has been some question in the literature about whether special relativity handles rotations correctly. The generator involves only symmetric products. What is interesting is the equation which generates the Lorentz transformation. the other Bq (a dilation). q t V . t. t X Xt Möller’s equation looks like it should involve two terms. X V . A quaternion equation that would generate these terms must be devoid of any terms involving cross products. but the generator depends on q and q*. Real quaternions are used in a rotation and a dilation to perform the work of the Lorentz group.X Γ V.X . one of the form AqA (a rotation). This has a vector equation and a scalar equation. t V This is the general form of the Lorentz transformation presented by Möller. This is probably one of the more confusing topics in physics.41 X t X Γt Γ 1 V. V V q 1 0. q ’ q q’ 2 q’ q tt X. That may have interesting interpretations. The symmetric product (anti−commutator) lacks the cross product. 1 t. Γ 1 1 v c 2 where c If V is only in the i direction. X V V 2 2 Γ even V . so I will just let the observation stand by itself. q . Γ even V.

X scalar q ’. 0. 0. 0 The quaternion L depends on the velocity and can depend on location in spacetime (85% of the type of problems assigned undergraduates in special relativity use an L that does not depend on location in spacetime). Some people view that as a bug. z x2 y2 z2 t. y. Γ v. The other technique relies on the property of a division algebra. tz xy Γt vy z Γx y vz t2 x2 y2 z2 if x if t y y z z 0. 0 x2 . q t2 X. L q’ q Γt Γ v x. z Γ t2 2 Γ t v x Γ x2 y2 z2 . x. Γ v. then L Γ. Γvt t2 Γ x. . then L 0. y. Γ. but I see it as a modern feature found in the standard model and general relativity as the demand that all symmetry is local.42 Two ways exist to use quaternions to do Lorentz transformations (to be discussed in the next section). Γ v t2 ty xz Γt y vz Γx vy z . The existence of two approaches may be of interest in itself. q ’ For a boost along the i direction. There exists a quaternion L such that: q’ L q such that scalar q.

The Tools of Special Relativity Events are represented as 4−vectors. x. which can be represented by the following matrix (where c = 1).gµΝ . . gµΝ 1 0 0 0 0 1 0 0 0 0 1 0 0 0 0 1 t2 x2 y2 z2 t. not the ones in physics). t. z . y. x.43 An Alternative Algebra for Lorentz Boosts: Local Transformations Many problems in physics are expressed efficiently as differential equations whose solutions are dictated by calculus. Γ 1 1 Β2 ΒΓ Β Γ Β 0 0 0 0 1 0 0 0 0 1 Γ Β ΒΓ Β 0 0 x Τhe boosted 4−vector is x. there are only three finite dimensional fields: the real numbers (1D). According to the theorem of Frobenius. The foundations of calculus were shown in turn to rely on the properties of fields (the mathematical variety. the complex numbers (2D). z 1 Β2 To demonstrate that the interval has been preserved. A member of this group is for boosts along the x axis. It will be shown that problems involving boosts along an axis of a reference frame can be solved using local quaternion transformations as apposed to the global transformations of the Lorentz group. and the electromagnetic potential. z The Lorentz group is defined as the set of matrices that preserves the inner product of two 4−vectors. which can be easily defined. To form an inner product between two vectors requires the Minkowski metric. energy−momentum. z t xΒ . y. or multiplied by a scalar. Special relativity stresses the importance of 4−dimensional Minkowski spaces: spacetime. and the quaternions (4D). t. x. y. In this section. events in spacetime will be treated as the 4−dimensional field of quaternions. calculate the inner product. which can be add or subtracted. y. 1 Β2 1 Β2 x 1 Β2 tΒ .

there must exist a quaternion L such that L q[x] = q[x’] since L = q[x’] q[x]^−1. y. ΒΓt 2 Γ Βt x 1 Γ 1 Γ z2 Γ x. not on 4−vectors. The vector portion is frame−dependent. x. z . t. it can be said that _any_ problem that can be solved using 4−vectors. x. y. z t2 Γ Γ 1 Γ t2 Γ x2 t ΒΓz y t ΓΒy z 2 t x2 y2 y2 z2 . x. 2 L t. then the first term of squaring q[x] and q[x’] will be identical. not global. X t2 X. y. y. Because quaternions are a division ring. There is implicitly.44 x . The transformation is classified as global because it depends only on the velocity and not on t. If two 4−vectors x and x’ are known to have the property that their intervals are identical. L Γt Β Γ x. x. Using Quaternions in Special Relativity Events will be treated as quaternions. Γ Β x ΓΒy z 1 x ΓΒz y 1 x2 . they are made of identical mathematical fabric. y. z t2 x2 y2 z2 Starting from a 4−vector. making it a local. acting on quaternions. z Γt ΓΒx. a form of the Minkowski metric that is part of the rules of quaternion multiplication. If a set of quaternions can be found that do not alter the interval. then that set would serve the same role as the Lorentz group. t. although events play a different role from operators. z This is a quaternion composed of the boosted 4−vector. 2 t X The first term of squaring a quaternion is the invariant interval squared. x. z. z t. x. the Minkowski metric and a Lorentz boost along the x axis can also be solved using the above quaternion for boosting the event quaternion. Define the Lorentz boost quaternion L along x using this equations. x. x . In this way. Β t. Any tool built to manipulate quaternions will also be a quaternion. a skew field or division algebra that is 4 dimensional. L t. y. The inverse of a quaternion is its transpose over the square of the norm (which is the first term of transpose of a quaternion times itself). L depends on the relative velocity and position.X. This is because both techniques transform the same set of 4 numbers to the same new set of 4 numbers using the same variable beta. this is the only way to boost a reference frame along the x axis to another 4−vector and preserve the inner product. See if L q[x] = q[x’]. Apply this approach to determine L for 4−vectors boosted along the x axis. . At this point. ΓΒt Γ x. y. or z. A squared quaternion is: 2 t. y. Confirm the interval is unchanged. y. gµ Ν . transformation. z .

Redshifts have a plus instead of the minus. 0. 2 t2 Β x2 Β t x 1 Β2 1 Β2 2y t xΒ 2z t xΒ . 0. or x = 0. Β. If t = 0. 1 Β2 1 Β2 The first term is conserved as expected. Γ Β t. 0 Γ 1 Β t. 0 t.033) have been solved using local quaternion transformations. 0. then L q Γ 1 q’ t.45 t2 x2 y2 z2 . then L q Γ 1. Γ x. the boost quaternion L becomes a very simple. 0. Further. . 0 Lq t . 90% required one of these simple forms for the boost quaternion. x ’. x. 0. A great many problems in special relativity do not involve angular momentum. Using Quaternions in Practice The boost quaternion L is too complex for simple calculations. x. 0. 0 Lq t . x = t. or for Doppler shift problems. 0 If t = x. 0 Β. q’ Β. Over 50 problems in a sophomore−level relativistic mechanics class at MIT (8. The vector portion of the square is frame dependent. 0. 0 Γ Β x. x ’. it is often the case that t = 0. which in effect sets y = z = 0. 0. In these cases. 0. 0 If x = 0. Mathematica does the grunge work. . x. 0 0. 0 Note: this is for blueshifts. 0. 0. 0 q’ Lq t . then L q Γ 1. 0. x ’. 0 Γ t.

Yet this creates hope that by using quaternions. No experimental difference between the two methods has been presented. metrics and operators when solving problems using quaternions.46 Implications Problems in special relativity can be solved either using 4−vectors. One must be judicious in choosing quaternions that will be relevant to a particular problem in physics and therein lies the skill. . particular in the field of general relativity. At this point the difference is in the mathematical foundations. An immense amount of work has gone into the study of metrics. There is no division between events. or using quaternions. A large effort has gone into group theory and its applications to particle physics. Yet attempts to unite these two areas of study have failed. the Minkowski metric and the Lorentz group. the long division between metrics (the Grassman inner product) and groups of transformations (sets of quaternions that preserve the Grassman inner product) may be bridged.

47 Electromagnetism .

The vector term is Faraday’s law. 0. 0. 0. with Maxwell’s correction. 0 even 4 Π Ρ. . t . 0. The Maxwell equations have been written with complex−valued quaternions back in Maxwell’s time. E B vector even odd . A . even t B .48 Classical Electrodynamics Maxwell speculated that someday quaternions would be useful in the analysis of electromagnetism. B AB 2 BA The first quaternion equation embodies the homogeneous Maxwell equations. A t Φ t . Hopefully after a 130 year wait. . in this section we can begin that process. B . B XB E t t . The scalar equation is Gauss’ law. The approach relies on a judicious use of commutators and anticommutators. The scalar term says that there are no magnetic monopoles. The second quaternion equation is the source term. These can also be expressed using quaternions. The 4−Potential A The electric and magnetic fields are often viewed as arising from the same 4−potential A. B XE B t odd 0. Φ. . . . My own efforts arose a year later. Φ. E where even A. 0. odd A. A xA 0. independently. . The vector term is Ampere’s law. The Maxwell Equations The Maxwell equations are formed from a combinations of commutators and anticommutators of the differential operator and the electric and magnetic fields E and B respectively (for isolated charges in a vacuum. J AB 2 BA t . Peter Jack was the first person to write the Maxwell equations using only real−valued quaternions. . E odd t E .

Γ Β . but there a small game required to get the signs correct for the 4−force. Φ. vector even A . Γ Β . Φ t A A t2 The homogeneous terms are formed from the sum of both orders of the commutator and anticommutator. B ΓΒ XB even Γ. B. The source terms arise from the difference of two commutators and two anticommutators. 0. t B t 0.49 The electric field E is the vector part of the anticommutator of the conjugates of the differential operator and the 4−potential. . t E t X XA 4 Π Ρ. Φ. odd even Φ E . A t 2 t t . The magnetic field B involves the commutator. . A . Conservation Laws The continuity equation − conservation of charge − is formed by applying the conjugate of the differential operator to the source terms of the Maxwell equations. . 0 . Φ A t . vector even xA t xE . It is almost as if the even/odd operators destructively interfere to generate the homogeneous equations. t x . odd t . E ΓΒ E . . J . 0. odd Γ. even odd t . odd X A t xA. while the even/even and odd/odd operators constructively interfere to describe a source. ΓE This is the covariant form of the Lorentz force. . . . The Lorentz Force The Lorentz force is generated similarly to the source term of the Maxwell equations. XB . These forms can be directly placed into the Maxwell equations. Φ. Φ.

except that the conjugate of electric field is used instead of the conjugate of the differential operator. 4 Π E J. 4 Π Ρ. E E. E t t . Poynting’s theorem for energy conservation is formed in a very similar way. That means that charge is conserved. 0 The foundations of classical electrodynamics are the Maxwell equations. This approach looks independent from the usual method which relies on an antisymmetric 2−rank field tensor and a U(1) connection. J Additional vector identities are required before the final form is reached. these basic elements have been written as quaternion equations. XB E t t XB . E. There is an interesting link between the E field and a differential operator for generating conservation laws. In this section. E . 4 Π Ρ. the means to generate these equations using quaternion operators has been displayed. .50 scalar E t . XB E t E 4ΠE J XB E E t 0. E XE E. J 4Π J Ρ t scalar The dot product of the E field and the current density plus the rate of change of the charge density must equal zero. the Lorentz force. 0 This is Poynting’s equation. Implications EXB 1 2 E t 2 1 2 B t 2 . and the conservation laws. scalar scalar 0. exploiting the actions of commutators and anticommutators. E t XB B t 1 2 1 2 B XE BXE E t B t 2 B B t 2 Use these equations to simplify to the following. More importantly.

Gauges are often chosen to make solving a particular problem easier. even though Del. A 2 Φ t Φ. (phi. only the way it is measured is different. The field tensor F in the Coulomb gauge is generated by subtracting away the divergence of A. .51 Electromagnetic Field Gauges A gauge is a measure of distance. and the second term involves the 3−vector field only. In all these cases. Generating the field tensor F in the Lorenz gauge starting from the gauge−invariant from involves swapping the fields in the following way: t . A). XA . A t Φ. A few are well known: the Coulomb gauge for classical electromagnetism. A Φ. What makes this form gauge−invariant. In this section. 0). the resulting equation is identical? It is the work of the zero! Whatever the scalar field is in the first term of the generator gets subtracted away in the second term. A 2 Φ.A. The first term of the generator involves the scalar field only. The Field Tensor F in Different Gauges The anti−symmetric 2−rank electromagnetic field tensor F has 3 properties: its trace is zero. and it contains all the components of the E and B fields. these are all generated using a differential quaternion operator and a quaternion electromagnetic potential. it is antisymmetric.A).A is zero :−) . so no matter what the choice of gauge (involving dphi/dt and Del. the E and B field is the same. and the gauge invariant form which is used in the Maxwell equations. (0. A t A t Φ . Φ. the Lorenz gauge which makes electromagnetism look like a simple harmonic oscillator. which explains why the second and third terms involve only A. A t . The field used in deriving the Maxwell equations had the same information written as a quaternion: t 0. Φ A XA Φ.

A Φ. A 4 . A t . Φ. There is a reason for this. A Φ. Φ. A 4 A t Φ XA t The field tensor F in the temporal gauge is quite similar to the Coulomb gauge. but its generator is a simple as it gets.A. In a lists of gauges in graduate−level quantum field theory written by Kaku. but some of the signs have changed to target the dphi/dt term.A. Φ . A 4 Φ. Φ. The most interesting gauges have to do with a long−running popularity contest. A Φ t . Perhaps some of the technical issues in quantum field theory might be tackled in this gauge using quaternions. A t Φ XA The light gauge is one sign different from the Lorenz gauge. A t . . . t . The relationship between gauges is guessed. Φ. A Φ.52 t . The term that did not make the cut stands out. A t Φ t XA What is the simplest expression that all of these generator share? I call it the field tensor F in the light gauge: t . the light gauge did not make the list of the top 6 gauges. Φ. . the gauge is a scalar generated in such a way as to not alter the 3−vector. A 4 Φ. Gauges are presented as a choice for a physicist to make. t A Φ. Implications In the quaternion representation. not written explicitly as was done here.

The following substitutions do not effect the resulting equations. It is none of the popular gauges: Coulomb (Del. E B vector even odd . A t Φ . See the section on Einstein’s Vision] The standard definition of a gauge starts with an arbitrary scalar function psi. . Φ A t t . temporal (phi = 0). transverse wave. E B even odd t t . What is the name of the scalar field.A which looks like some sort of gauge? It is not the Lorenz or Landau gauge which has a plus sign between the two. In this section. antisymmetric product over two or cross product). Φ t XA A. and [A. Φ. A XA 0. . A The scalar information is explicitly discarded from the E field quaternion.. . but may explain why light is a quantized. the electric field was generated differently from the magnetic field. Φ. unitary.53 The Maxwell Equations in the Light Gauge: QED? What makes a theory non−classical? Use an operational definition: a classical approach neatly separates the scalar and vector terms of a quaternion. the scalar field that arises will be examined and shown to be the field which gives rise to gauge symmetry. The commutators and anticommutators of this scalar and vector field do not alter the homogeneous terms of the Maxwell equations. B} is the even or symmetric product over 2. and the Gauge with No Name In the previous section. t . Φ A Φ’ A’ Φ A Ψ t Ψ This can be written as one quaternion transformation. Feynman. This time that will not be done. Φ. Recall how the electric field was defined (where {A. [special note: I am now testing the interpretation that this gauge constitutes the gravitational field. B] is the odd. axial (Az = 0).A = 0). . The E and B Fields. A A 0. 0. since the scalar field was discard. d phi/dt − Del.. Φ.

54

Φ, A

Φ ’, A ’

Φ, A

Ψ , t

Ψ

The goal here is to find an arbitrary scalar and a 3−vector that does the same work as the scalar function psi. Let

p

Ψ t

and Α

Ψ

Look at how the gauge symmetry changes by taking its derivative.

t

, Ψ p t

Ψ , Ψ t 2Ψ , X t2 Α, 0

Ψ

Ψ t

Ψ t

This is the gauge with no name! Call it the "light gauge". That name was chosen because if the rate of change in the scalar potential phi is equal to the spatial change of the 3−vector potential A as should be the case for a photon, the distance is zero. The Maxwell Equations in the Light Gauge The homogeneous terms of the Maxwell equations are formed from the sum of both orders of the commutator and anticommutator.

even odd

t

, ,

, odd , even X Φ

t

, t ,

, Φ, A , Φ, A

t

XA,

0, 0

The source terms arise from of two commutators and two anticommutators. In the classical case discussed in the previous section, this involved a difference. Here a sum will be used because it generates a simpler differential equation.

odd even
2

t

, t

, odd , , even Φ,
2 2

t

, t X
2

, Φ, A , , Φ, XA 4 Π Ρ, J A A

Φ t2
2

A t2

Φ t2

Φ,

2

A t2

A

55

Notice how the scalar and vector parts have neatly partitioned themselves. This is a wave equation, except that a sign is flipped. Here is the equation for a longitudinal wave like sound.
2

w t2

2

w

0

The second time derivative of w must be the same as Del^2 w. This has a solution which depends on sines and cosines (for simplicity, the details of initial and boundary conditions are skipped, and the infinite sum has been made finite).

w
n 0 t t

Cos n Π t Sin n Π R w
R R

w

0

Hit w with two time derivatives, and out comes −n^2 pi^2 w. Take Del^2, and that creates the same results. Thus every value of n will satisfy the longitudinal wave equation. Now to find the solution for the sum of the second time derivative and Del^2. One of the signs must be switched by doing some operation twice. Sounds like a job for i! With quaternions, the square of a normalized 3−vector equals (−1, 0), and it is i if y = z = 0 . The solution to Maxwell’s equations in the light gauge is

w
n 0

Cos n Π t Sin n Π R V
2

if V

1, then

t

t

w

R

R

w

0

Hit this two time derivatives yields −n^2 pi^2 w. Del^2 w has all of this and the normalized phase factor V^2 = (−1, 0). V acts like an imaginary phase factor that rotates the spatial component. The sum for any n is zero (the details of the solution depend on the initial and boundary conditions).

Implications
The solution to the Maxwell equations in the light gauge is a superposition of waves − each with a separate value of n − where the spatial part gets rotated by the 3D analogue of i. That is a quantized, transverse wave. That’s fortunate, because light is a quantized transverse wave. The equations were generated by taking the classical Maxwell equations, and making them simpler.

56

The Lorentz Force
The Lorentz force acts on a moving charge. The covariant form of this law is, where W is work and P is momentum:

dW dP , dΤ dΤ

Γ e Β.E, E

ΒXB

In the classical case for a point charge, beta is zero and the E = k e/r^2, so the Lorentz force simplifies to Coulomb’s law. Rewrite this in terms of the potentials phi and A.

dW dP , dΤ dΤ

Γ e Β.

A t

Φ ,

A t

Φ

ΒX

XA

In this section, I will look for a quaternion equation that can generate this covariant form of the Lorentz force in the Lorenz gauge. By using potentials and operators, it may be possible to create other laws like the Lorentz force, in particular, one for gravity. A Quaternion Equation for the Lorentz Force The Lorentz force is composed of two parts. First, there is the E and B fields. Generate those just as was done for the Maxwell equations

t

,

Φ, A

Φ t

A,

A t

Φ

XA

Another component is the 4−velocity

V

Γ, Γ Β

Multiplying these two terms together creates thirteen terms, only 5 of whom belong to the Lorentz force. That should not be surprising since a bit of algebra was needed to select only the covariant terms that appear in the Maxwell equations. After some searching, I found the combination of terms required to generate the Lorentz force.

t

,

Φ, A t Β.E, E

A

Γ, Φ ,

ΓΒ A t

Γ, Φ

ΓΒ ΒX XA

t

,

Φ, A

Γ Β Γe

Β

B

This combination of differential quaternion operator, quaternion potential and quaternion 4−velocity generates the covariant form of the Lorentz operator in the Lorenz gauge, minus a factor of the charge e which operates as a scalar multiplier.

quaternion potential and quaternion 4−velocity generates the covariant form of the Lorentz operator in the Lorenz gauge. F = − G m m’/r^2. it may be possible to create other laws like the Lorentz force. An obvious candidate is Newton’s law of gravity. For point sources in the classical limit. Implications By writing the covariant form of the Lorentz force as an operator acting on a potential. these new laws must have the form of Coulomb’s law. . F = k e e’/r^2.57 This combination of differential quaternion operator. minus a factor of the charge e which operates as a scalar multiplier.

express the stress tensor T in terms of the E and B fields only: Tik W Sx Sy Sz Sx mxx mxy myz Sy myx myy myz Sz mzx mzy mzz 1 8Π 1 4Π 1 4Π E 2 W Sa mab B 2 ExB Ea Eb Ba Bb 0. Poynting’s vector (Sa) and the Maxwell stress tensor (m_ab) are all the components of the stress tensor of the electromagnetic field. What is more interesting is when an equation is generated by the product of operators acting on quaternion fields. Generating a Symmetric 2−Tensor Using Quaternions How should one rationally go about to find a generator equation that creates these terms instead of using the month−long hunt−and−peck technique actually used? Everything is symmetric. People can choose alternative bookkeeping systems. A tensor is a bookkeeping device designed to keep together elements that transform in a similar way. A straight translation of terms could probably be automated with a computer program. Using the terms as defined in "The classical theory of fields" by Landau and Life−sized. Therefore. This method may provide some insight into what information the stress tensor contains. the antisymmetric 2−rank field tensor F is used to generate the stress tensor T Tik 1 4Π FiL Fk L 1 ik ∆ FLM FLM 4 I have a practical sense of an E field (the stuff that makes my hair stand on end) and a B field (the invisible hand directing a compass). a particular combination of the other two. I have found that generator equations often yield useful insights. so long as the tensor behaves the same way under transformations. so use the symmetric product: . Any equation written with 4−vectors can be rewritten with quaternions.58 The Stress Tensor of the Electromagnetic Field I will outline a way to generate the terms of the symmetric 2−rank stress−momentum tensor of an electromagnetic field using quaternions. the energy density(W). but have little sense of the field tensor F.5 ∆ab E 2 B 2 Together.

1. Implications So what does the stress tensor represent? It looks like every combination of the 3−vectors E and B that avoids quadratics (like Ex^2) and over−counting cross terms. 0. The third line creates the Poynting vector. Using quaternions. B . Ub even odd E. The rest of that tensor is generated by the second line. 0 where Ux 0. E 2 2 0. 0 2 Π The first line generates the energy density W.z y. B 2 2 even E.z a x 1 4Π even Ua. Changing the basis vectors Ua and Ub will effect these three components differently. Ux Tik y. I like what I will call the "net" stress quaternion: net Tik . B . Ua even B. Ua even E. These are the Maxwell stress tensor. Ub Bx By Ex Bz Bx Bz Ex By By Bz Ey Bx. Ua 3 2 1 0. Individual directions of a field can be selected by using a unit vector Ua: even E. Ub even odd E. Does the generator equation have the correct properties? Switching the order of Ua and Ub leaves T unchanged.z a x b x Ex. t X Xt The fields E and B are kept separate except for the cross product in the Poynting vector. those that involve Ua or Ub. q ’ q q’ 2 q’ q tt X. Ub 3 1 0. b)(E^2 + B^2) term of the Maxwell stress tensor. and those that involve neither. Check the trace. so it is symmetric. when Ua = Ub trace Tik y. the Poynting vector and the energy density respectively.X . the net sum of these terms ends up in the scalar. B 2 even E. Ua The trace equals zero. and part of the +0. as it should. Ua 0 The generator is composed of three parts that have different dependencies on the unit vectors: those terms that involve Ua and Ub.59 even q. E 0. Ua Ex Ey Ey Bz Ex Ez Ez By Ey Ez Ez Bx even B.5 delta(a. 0 The following double sum generates all the terms of the stress tensor: 2 2 1 4Π even Ua. even B.

. Now that I have a generator and a net quaternion expression. it looks quite elegant and straightforward to me. 0 2 This has the same properties as an stress tensor. even though I listen in on a discussion of the topic. but not the others. I never thought I would understand what a symmetric 2−rank tensor was. Yes. which is equivalent to the statement that the trace of the tensor is zero. it felt hollow. but without any sense of F. Switching x terms for y terms would flip the signs of the terms produced by the Poynting vector as required.60 Π Ex Ey Ey Bz Ex Ez Ez By Ey Ez Ez Bx Bx By Ex Bz Bx Bz Ex By By Bz Ey Bx. I could nod along with the algebra. On a personal note. Since the vector is zero. it commutes with any other quaternion (this may be a reason it is so useful). There are no terms of the form Ex^2.

61 Quantum Mechanics .

otherwise it must have a positive value. Completeness With the topology of a Euclidean norm for a real 4−dimensional vector space. Quaternions are complete in a manner required to form a Banach space if there exists a neighborhood of any quaternion x such that there is a set of quaternions y x y 2 Ε4 for some fixed value of epsilon. X t2 X. X t. Construct such a neighborhood. The inner product of a function with itself is the norm.62 A Complete Inner Product Space with Dirac’s Bracket Notation A mathematical connection between the bracket notation of quantum mechanics and quaternions is detailed. The adjective "Euclidean" is used to distinguish this product from the Grassman inner−product which plays a central role in special relativity (see alternative algebra for boosts). The Positive Definite Norm of a Quaternion The square of the norm of a quaternion can only be zero if every element is zero. . X The inner−product of two quaternions is defined here as the transpose (or conjugate) of the first quaternion multiplied by the second. The Euclidean inner−product of two quaternions can take on any value. as is the case in quantum mechanics for <phi|theta>. quaternions are complete. t. so the vector flips signs): t. the following assertion will be examined (* is the conjugate. X t. It will be argued that quaternions have the properties of a complete inner−product space (a Banach space for the field of quaternions). X t.X. X 2 t. A central issue is the definition of the square of the norm. X t. In quantum mechanics: 2 In this section. 0 This is the standard Euclidean norm for a real 4−dimensional vector space.

X t.X . Φ t. X t. 0. 0. X t. They are worked out for quaternions here in detail to solidify the connection between the machinery of quantum mechanics and quaternions. t. X Ε 4 t.X t.") Identities and Inequalities The following identities and inequalities emanate from the properties of a Euclidean norm. X tt t t X. cited in Birkhoff and Mac Lane’s "A Survey of Modern Algebra. X y2 t. X . X t. 0 t. because the terms involving the cross produce will flip signs when their order changes. X x2 t. X t. t X tX X t Xt X xX XxX These are identical.63 t.X t.X t. X t.X. 0. 0 t.X t. 0.X . X t. Φ For quaternions.X t2 t.X z2 .X t. X t. X t2 y2 t. 0. X Ε 4 t. 0 16 Ε4 . Φ Φ Φ Φ This is also the case for quaternions. Any polynomial equation with quaternion coefficients has a quaternion solution in x (a proof done by Eilenberg and Niven in 1944. X x2 t. X Ε 4 t.X z2 . 0.X t. 0. The conjugate of the square of the norm equals the square of the norm of the two terms reversed. For products of squares of norms in quantum mechanics. 0. X Ε 4 t. 0 An infinite number of quaternions exist in the neighborhood.X t. X Ε4 .X t.X t.X t.

64 t. X t.X t.X t. Φ Χ ΛΦ Φ Φ Φ Φ ΛΦ 2 Φ Φ Let a third wave function. The norm of chi will necessarily be greater than zero. The Schwarz inequality in quantum mechanics is analogous to dot products and cosines in Euclidean space.X . X t. X t t2 t. then the cross terms will cancel on the left hand side of the inequality. . X t X 2 t. Λ Φ Φ Φ 0 Multiply through by the denominator.X The triangle inequality in quantum mechanics: Φ For quaternions. 0 ^2 2 t. making it smaller than the right hand side where terms never cancel because there are only squared terms. 0 t2 2 X 2 t 2 X 2 t.X .X 2 X X 2 t. If the signs are different. ΛΦ Λ Φ Λ Φ Λ ΛΦ Φ 0 Choose the value for lambda that helps combine all the terms containing lambda. chi. separate the two resulting terms and do some minor rearranging. X If the signs of each pair of component are the same. be the sum of these two with an arbitrary parameter lambda. X t. X t.X t 2 t.X t. the two sides will be equal. X t.X 2tt X 2 2 X.X t. Φ 2 Φ Φ 2 2 t.X t.

X t. X t. X t. X t 2 t.65 Φ Another inequality: Φ Φ Φ This is now the Schwarz inequality. 0 t. X X . .0 2 2 X .X t.0 The cross terms can be put on the other side of inequality. X t.X t. 2 Re Φ Φ Φ Examine the square of the norm of the difference between two quaternions which is necessarily equal to or greater than zero.X . X t. 2 t t 2 Re t.X t2 X 2 t 2 X . X t.X This is twice the square of the norms of the two separate components.0 X . X t.X 2 2 t. and leaving the sum of two norms behind.X t.0 t. X t t 2 t2 2 t t X 2 t. 0 t t. X t.X Φ Φ Φ Φ 2 2 Φ Φ Test the quaternion norm t.X X X t 2 X .0 2 t.X X t. The parallelogram law: X.X t.X X . X X . X The inequality holds. changing the sign. X X .

66 Implications In the case for special relativity. Note: this goal is different from the one Stephen Adler sets out in "Quaternionic Quantum Mechanics and Quantum Fields. . The analytical properties of quaternions do not play a critical role. it was noticed that by simply squaring a quaternion. the power of a mathematical field was harnessed to solve a wide range of problems in special relativity. it is hoped that because the product of a transpose of a quaternion with a quaternion has the properties of a complete inner product space. there is no need to imbed quaternions again within another Hilbert space. the power of the mathematical field of quaternions can be used to solve a wide range of problems in quantum mechanics. I like a close shave with Occam’s razor. It is my opinion that since the product of a transpose of a quaternion with a quaternion already has the properties of a norm in a Hilbert space." He tries to substitute quaternions in the place of complex numbers in the standard Hilbert space formulation of quantum mechanics. It is the properties of the Hilbert space over the field of quaternions that is harnessed to solve problems. From that solitary observation. This is an important area for further research. In a similar fashion. the resulting first term was the Lorentz invariant interval.

Y . That is a project for the future. V 0. Y 2 X. X X X Y Y . 0 I^2 equals −1 just like i^2. q q exp Θ I q q Cos Θ I Sin Θ This representation can be useful due to the properties of the exponential function. cosines and sines.. V V. I q 2 q q q 0. whether one works with Fourier transforms or Lie groups. which is the transpose of a quaternion multiplied by itself. I2 1 It should be possible to do Fourier analysis with quaternions. Let (0. Y . Those tools are necessary for solving problems in quantum mechanics. and a 3−vector I (which in some cases can be the more familiar i).67 Multiplying Quaternions in Polar Coordinate Form Any quaternion can be written in polar coordinate form. Y X Y 1 X. V 0. q q q q 2 q q The angle is the arccosine of the ratio of the first component of a quaternion over the norm.V. V) = (q − q*)/2. VxV V2 . Given a Lie algebra of a Lie group in a sufficiently small area the identity. and to form a Dirac delta function (or distribution). New Method for Multiplying Quaternion Exponentials Multiplying two exponentials is at the heart of modern analysis. The absolute value of a quaternion is the square root of the norm. . which involves a scalar magnitude and angle. V 0. Θ ArcCos The vector component is generated by normalizing the pure quaternion (the final three terms) to the norm of the pure quaternion. the product of two exponentials can be defined using the Campbell−Hausdorff formula: Exp X Exp y 1 12 X..

68 This formula is not easy to use. By using an exponential that has pi/2 multiplied by a normalized quaternion. What is happening here is that both commuting and anticommuting parts scale themselves appropriately. The anticommuting part needs to be purely imaginary. q ’ q. q ’ Π 2 Abs q. q ’ q q’ q’ q same as above q. Andrew Millard made the suggestion for the Grassman product that unifies these approaches nicely. Let q q q’ Exp X q’ Exp Y q. this always has a zero scalar. . and is only applicable in a small area around unity. as it must to accurately represent an anticommuting part. The product of two quaternions splits into a commuting and an anti−commuting part. q ’ q q’ q’ q I call these operators "conjugators" because they involve taking the conjugate of the two elements. I have developed (perhaps for the first time) a simpler and general way to express the product of two quaternion exponentials as the sum of two components. q ’ q ’ q and q. q ’ q. q ’ where q q’ q q’ where q q’ Abs q. q ’ q’ q 2 Exp and q. Together these define the needs for the product of two quaternion exponentials. The rules for multiplying commuting quaternions are identical to those for complex numbers. Quaternion analysis that relies on this formula would be very limited. The Grassman product (q q’) of two quaternion exponentials and the Euclidean product (q* q’) should both have these properties.

69

Commutators and the Uncertainty Principle
Commutators and the uncertainty principle are central to quantum mechanics. Using quaternions in these roles has already been established by others (Horwitz and Biedenharn, Annals of Physics, 157:432, 1984). The first proof of the uncertainty principle I saw relied solely on the properties of complex numbers, not on physics! In this section, I will repeat that analysis, showing how commutators and an uncertainty principle arise from the properties of quaternions (or their subfield the complex numbers). Commutators Any quaternion can be written in a polar form.

q

s, V

q q Exp

s q q

V V V

This is identical to Euler’s formula except that the imaginary unit vector i is replaced by the normalized 3−vector. The two are equivalent if j = k = 0. Any quaternion could be the limit of the sum of an infinite number of other quaternions expressed in a polar form. I hope to show that such a quaternion mathematically behaves like the wave function of quantum mechanics, even if the notation is different. To simplify things, use a normalized quaternion, so that q* q = 1. Collect the normalized 3−vector together with I = V/(V* V)^.5. The angle s/(q* q)^.5 is a real number. Any real number can be viewed as the product of two other real numbers. This seemingly irrelevant observation lends much of the flexibility seen in quantum mechanics :−) Here is the rewrite of q.

q

Exp a b I q q 1, ab s , q q I V V V

where

The unit vector "I" could also be viewed as the product of two quaternions. For classical quantum mechanics, this additional complication is unnecessary because all "i"’s commute. A point of clarification on notation: the same letter will be used 4 distinct ways. There are operators, A hat, which act on a quaternion wave function by multiplying by a quaternion, capital A. If the operator A hat is an observable, then it generates a real number, (a, 0), which commutes with all quaternions, whatever their form. There is also a variable with respect to a component of a quaternion, a_i, that can be used to form a differential operator. Define a linear operator A hat that multiplies q by the quaternion A.

Aq

Aq

If the operator A hat is an observable, then the quaternion A is a real number, (a, 0). A real number will commute with any quaternion. This equation is functionally equivalent to an eigenvalue equation, with A hat as an eigenvector of q and (a, 0) as the eigenvalue. However, all of the components of this equation are quaternions, not separate structures such as an operator belonging to a group and a vector. This might make a subtle but significant difference for the mathematical structure of the theory, a point that will not be investigated here.

If the operator A hat is an observable, then the quaternion A is a real number, (a, 0). A real number will commute with any quaternion. This equation is functionally equivalent to an eigenvalue equation, with A hat as an eigenvector of q and (a, 0) as the eigenvalue. However, all of the components of this equation are quaternions, not separate structures such as an operator belonging to a group and a vector. This might make a subtle but significant difference for the mathematical structure of the theory, a point that will not be investigated here. Define a linear operator B hat that multiplies q by the quaternion B. If B hat is an observable, then this operator can be defined in terms of the scalar variable a.

70

Let B Bq I

I

d da bq

d Exp a b I da Aq Aq Bq A q’

Operators A and B are linear.

A A q

B q q’

aq aq

bq

a

b q

a’ q’ dq da daq da

Calculate the commutator [A, B], which involves the scalar a and the derivative with respect to a.

A, B q aI dq da

AB aI

BA q dq da

aI Iq

I Iq

da da

The commutator acting on a quaternion is equivalent to multiplying that quaternion by the normalized 3−vector I. The Uncertainty Principle Use these operators to construct things that behave like averages (expectation values) and standard deviations. The scalar a−−generated by the observable operator A hat acting on the normalized q−−can be calculated using the Euclidean product.

q

Aq

q aq

aq q

a

It is hard to shuffle quaternions or their operators around. Real scalars commute with any quaternion and are their own conjugates. Operators that generate such scalars can move around. Look at ways to express the expectation value of A.

q

Aq

q aq A q q A

aq q Aq q Aq

a q q

Aq

q

a

Define a new operator A’ based on A whose expectation value is always zero.

Let A ’ q A’ q

q

a

a

0

Define the square of the operator in a way designed to link up with the standard deviation.

71

Let DA ’2

q

A’ 2 q

q

A’ q

2

q

A’ 2 q

An identical set of tools can be defined for B. In the section on bracket notation, the Schwarz inequality for quaternions was shown.

A’ B’ 2

B’ A’

A’

B’

The Schwarz inequality applies to quaternions, not quaternion operators. If the operators A’ and B’ are surrounded on both sides by q and q*, then they will behave like scalars. The left−hand side of the Schwarz inequality can be rearranged to form a commutator.

q

A’ B’

B’ A’ q q B’ A’ q q I q a’ B’ q q I d A’ q da

q A’ B’ q q a’ B’ q q A’ B’

B’ A’ q

d A’ q da q A’ , B’ q

The right−hand side of the Schwarz inequality can be rearranged to form the square of the standard deviation operators.

q A’ B’ q q A ’ A’ B ’ B’ q

q A’ 2 B’ 2 q

q DA ’2 DB ’2 q

Plug both of these back into the Schwarz inequality, stripping the primes and the q’s which appear on both sides along the way.

A, B 2

DA2 DB2

This is the uncertainty principle for complementary observable operators.

B 2 I 2 i d dx i X. B A. The proof is in the structure of quaternions. The proof of the uncertainty principle shown here is independent of this issue. A q q The square of the standard deviation : ∆A2 Ψ A2 Ψ Ψ AΨ 2 DA2 One subtlety to note is that a quaternion operator is anti−Hermitian only if the scalar is zero. . P 2 i 2 ∆X2 ∆P2 The product of the squares of the standard deviation for position and momentum in the x−direction has a lower bound equal to half the expectation value of the commutator of those operators. but quantum field theory may require full quaternion operators. P DA2 DB2 X.72 Connections to Standard Notation This quaternion exercise can be mapped to the standard notation used in physics bra : ket : Ψ Ψ q q A I A. B operator : A imaginary : i commutator : norm : Ψ Ψ expectation of A : Operator A is Hermitian 0. q 0. A q 0. make these specific maps A B I X P A. This is probably the case for classical quantum mechanics. A is anti Hermitian. B q q Ψ AΨ maps to q A q A. To get to the position−momentum uncertainty equation. I do not yet understand the consequence of this point.

or wave packets that are superpositions of plane waves. This perspective should be easy to generalize. With this insight.73 Implications There are many interpretations of the uncertainty principle. Here.X. the uncertainty principle is about quaternions of the form q = Exp[a b I]. will have four uncertainty relations.X)/hbar I]. . one can see by inspection that a plane wave Exp[((Et − P. one for the scalar Et and another three for the three−part scalar P.

Life is stable around small permutations of the identity:−) The conjugate flips the signs of the each component in the 3−vector. The same rotation matrix acting on the same quaternion behaves differently depending directly on what involutions are involved. Second. the identity and the conjugate. Rotations. This small difference in behavior inside a commutator is what creates the ability to represent integral and half−integral spins.74 Unifying the Representations of Integral and Half−Integral Spin I will show how to represent both integral and half−integral spin within the same quaternion algebraic field. What happens with the scalar is irrelevant to this dimensional counting. respectively. and the second conjugate anti−automorphism: I: q : q 1 2 q q q q 1 2 :q :q 1 2 where q q e1 q e1 e2 q e2 e1. I will show that a rotation operator that acts differently on two of the three components of the 3−vector acts like half−integral spin. These two automorphisms. First a sketch of why this might work will be provided. the first conjugate anti−automorphism. By acting on only two of three components. 4 are required: They are the identity automorphism. and Commutators Quaternions are formed from the direct product of a scalar and a 3−vector. Finally. depending on which involution is used to construct the infinitesimal rotation. Quaternions have 4 degrees of freedom. This involves using quaternion automorphisms. Rotational operators that act on each of the 3 components of the 3−vector act like integral angular momentum. a general identity will be used to look at what happens under exchange of two quaternions in a commutator. If we want to represent quaternions with automorphisms. e2. . treat the 3−vector as a unit. a commutator will behave differently. Therefore these operators act on only the two of the three components in the 3−vector. small rotations in a plane around two axes will be used to show how the resulting vector points in an opposite way. The first and second conjugate flip the signs of all terms but the first and second terms. the conjugate anti−automorphism. Automorphism. e3 are basis vectors The most important automorphism is the identity.

sin Θ Re1 Re2 Re3 0 Θ 1 1 1 O Θ 3 . Sakurai considered rotations around any point along the z axis. cos Θ Θ2 2 Θ2 2 Θ2 2 e0 . This extra effort will be useful when considering how rotations are effected by curved spacetime. it’s great :−) There are similar rotations around the first and second axes at the origin. J. a1 cos Θ a1 sin Θ e0 e3 3 if q Re3 0 Θ q a2 sin Θ e2 . Re1 Re2 0 Θ Θ 0 e1 .75 Small Rotations Small rotations about the origin will now be calculated. Sakurai’s book "Modern Quantum Mechanics". 3 a2 cos Θ Two technical points. 0. z. This analysis is confined to the z axis at the origin. y. a2 2 . To second order in theta. Θ 1 O Θ3 O Θ3 O Θ3 O Θ3 Θ Θ Θ 1 1 1 e1 . 0 3 3 q’ 0. sin Θ Θ2 2 Consider an infinitesimal rotation for these three rotation operators. First. 0. 0. Θ 0 e0 . a1 e1 e . 0. 0 3 e2 cos Θ e0 . making modifications necessary to accommodate quaternions. Θ Calculate the commutator of the first two infinitesimal rotation operators to second order in theta: .0 3 e3 3 0 e0 . chapter 3. Second. sin Θ . a significant but not unreasonable constraint. This machinery is also necessary to do quaternion analysis (please see that section. these rotations are written with generalized coordinates instead of the very familiar and comfortable x.0 3 e0 e1 . 0 3 e2 . sin Θ 0. 0. 0. consider rotations about the origin in the z axis. 0. I will be following the approach used in J. These will then be expressed in terms of the four automorphisms discussed above. First. Define: Re3 0 Θ cos Θ e0 .0 3 cos Θ e0 . 0.

Repeat this exercise one last time for the first conjugate: R 1 e1 0. R 1 e2 0 1 1 1 Θ2 2 e0 . R e1 0 . 0 3 2 3 e e e e e e 1 Θ2 e0 2 . 0. Θ2 1 2 3 3 9 e0 e1 e0 e2 e e 1 Θ2 e0 2 . Θ 0 2 . 0. 0 3 1 Θ2 2 e0 . Θ 0 1 . Θ 1 . 0. Θ e2 . Θ 1 . Θ . Θ 2 . 0.0 3 Θ2 e e0 . Now I want to write this result using anti−automorphic involutions for the small rotation operators. 0 2 3 . 0. 0. Θ .76 Re1 1 0. Θ2 1 2 2 Re3 0 Θ2 R 0 9 Nothing has changed. 0 2 3 e Θ 2 . the commutator of infinitesimal rotations of rotations about the first two axes equals twice one rotation about the third axis given the squared angle minus a zero rotation about an arbitrary axis (a fancy way to say the identity). Θ2 1 2 2 Re3 0 Θ2 R 0 9 1 To second order. Θ 1 Θ2 2 e0 . Θ 1 . Θ e1 .Θ .0 3 e2 Θ2 e . e Θ2 e0 . Θ . 0.0 1 e0 . Θ 0 2 . Θ 3 3 9 e e 2 0. 0. Θ2 1 2 3 3 9 e0 e1 e0 e2 2 2 2 e1 e2 1 Θ e0 . Θ2 1 2 3 3 9 e e 2 0. Θ 1 . 0. 0.0 3 e2 Θ2 e 1 e0 . 0. 0 2 3 Θ2 e 1 e0 . Θ 1 . 0. Re2 0 Θ2 e e0 . 0. R e2 0 1 Θ2 2 Θ2 2 e0 . Θ . Θ e2 . 0. 0 2 3 Θ2 2 e0 . 0. 0 3 2 3 e e e e e e 1 Θ2 e0 2 . Θ 0 1 .0 1 e0 . 0.

a1 a2 1 2 9 9 9 q 1. The first and second conjugates in the commutator will point in the opposite direction. 9 9 e1 e3 e e a1 a3 . Θ . A similar. but not identical story applies for the second conjugate. a2 a3 2 3 a3 a2 3 2 . 0. q’ 1 1 q 2. The behavior in a commutator under exchange of the identity automorphism and the anti−automorphic conjugate are identical. Automorphic Commutator Identities This is a very specific example. 9 9 e3 e1 e1 e3 e e a3 a1 a1 a3 . Is there a general identity behind this work? Here it is: q. q 1 e e e e 0. the negative sign seen on the z axis for the first conjugate commutator is due to the action of an additional first conjugate. Imagine adding up a series of these small rotations. say 2 pi of these. q q . the identity and conjugate commutators form a distinct group from the commutators formed with the first and second conjugates. I stopped right at the commutator (not including any additional conjugator). q 2 a2 a3 e3 e1 9 e2 e3 e e a3 a2 3 2 . With the identity. Θ This points exactly the opposite way. but the scalar. q e e e e 0. q’ q’ . . Notice.77 e0 e1 e e e e . q’ 1 a2 a1 e2 e1 9 q’ 1. 0. but not identical. The first and second conjugates are similar. a3 a1 2 a2 a1 e2 e1 9 q’ 2. Θ2 1 2 2 Re3 0 Θ2 R 0 9 1 Θ2 e0 2 . In that case: q. the first term will have the correct sign after a 2 pi journey. q’ 0. q’ q 1. No doubt the identity and conjugates will bring you back exactly where you started. a1 a2 1 2 9 9 9 q 2. Θ 0 2 . because the minus of a minus will generate a plus. a1 a2 1 2 9 9 a2 a1 e2 e1 9 Under an exchange. Θ . even for an infinitesimal angle! This is the kernel required to form a unified representation of integral and half integral spin. q ’ q ’. q’ 2 2 It is usually a good sign if a proposal gets more subtle by generalization :−) In this case. Θ 3 3 9 e e 2 0. we can see exactly what happens if q changes places with q’ with a commutator. third and forth terms will point the opposite way. 9 9 e3 e1 e1 e3 e e a3 a1 a1 a3 . Θ2 1 2 3 3 9 e0 e1 e0 e2 2 2 2 e1 e2 1 Θ e0 . To get back on course will require another 2 pi. For the first conjugate. a2 a3 2 3 a3 a2 3 2 . q ’ q .

The first and second conjugates are similar. . the identity and conjugate commutators form a distinct group from the commutators formed with the first and second conjugates. but not identical set.78 Under an exchange. q’ q 1. q’ 1 q 2. while the first and second conjugates form a similar. Because this is a general quaternion identity of automorphisms. I don’t know how to use them. the identity and first conjugates appear to behave differently that the first and second conjugates. In a commutator under exchange of two quaternions. the identity and the conjugate behave in a united way. but not identical. Implications Three different operators had to be blended together to perform this feat: commutators. and rotations. but again. q’ 2 At this point. conjugates and rotations. this should be very widely applicable. mirrors. These involve issue of even/oddness. q ’ q . The behavior in a commutator under exchange of the identity automorphism and the anti−automorphic conjugate are identical. There are also corresponding identities for the anti−commutator: 1 2 q.

Given the potential V(0. In this section. Make these into complete quaternions by bringing in time to go along with the space 3−vector R. the quaternion analog to the Schrödinger equation will be derived from first principles. R E. so define one symbol to capture it: . One might argue that for completeness. X) and suitable boundary conditions. Ψ t. What is interesting are the constraint that are required for the quaternion analog. Psi is a scalar function with complex values.79 Deriving A Quaternion Analog to the Schrödinger Equation The Schrödinger equation gives the kinetic energy plus the potential (a sum also known as the Hamiltonian H) of the wave function psi. assume that psi is normalized. P Et R. as is the case here.P. The Quaternion Wave Function The derivation starts from a curious place :−) Write out classical angular momentum with quaternions. but notice it is composed of all the parts seen in the scalar. all combinations of E. X Ψ For the time−independent case. L 0. HΨ i Ψ t 2 2 2m Ψ V 0. P What makes this "classical" are the zeroes in the scalars. R 0. t. which contains all the dynamical information about a system. For example. R and P should be involved in psi. just different permutations that evaluate to 3−vectors. and E with the 3−vector P. there is a factor which might serve to damp runaway terms. The complicated 3−vector is a new animal. and kinetic energy as the square of the momentum operator. energy is written at the operator −i hbar d/dt. The 3−vector of psi is quite complicated. so |psi| = 1. Any quaternion can be expressed in polar form: q q e arccos s q V V Express psi in polar form. R x P odd 0. i hbar Del. solving this differential equation generates a wave function psi which contains all the properties of the system. 0. P Et R. E R Pt RxP The scalar part of psi is also seen in plane wave solutions of quantum mechanics.P. t. over 2m. R E. To make things simpler. E R Pt RxP Define a dimensionless quaternion psi that is this product over h bar.

Now take the time derivative of psi. they will not change in time. Instead. The position 3−vector X is in the same direction as the momentum 3−vector P X. If energy and momentum are conserved. Another advantage is that all the equations will necessarily be invertible.P I This is what I call the quaternion wave function. It is a mathematical tool tossed in so that the equation has the correct properties. Energy is evenly distributed in space 3. t. P and X. Still it will look very similar. 1. down to the last h bar :−) All that needs to be done is to study how the quaternion wave function psi changes. there is an imaginary factor i which contains no information about the system. Here is a verbal explanation. E t E 0 0 and P t 0 2. since that involves Hermitian operators that acting on a complex vector space. Changes in the Quaternion Wave Function We cannot derive the Schrödinger equation per se. The system is isolated xP 0 4. In that case. Unlike previous work with quaternionic quantum mechanics (see S. Energy and Momentum are conserved. Since I is in the direction of X. . Make the following assumptions.80 I ER Pt RxP ER Pt RxP Now rewrite psi in polar form with these simplifications: Ψ e Et R. its curl will be zero.P X P 1 which implies deI dt 0 and xeI 0 The implications of this last assumption are not obvious but can be computed directly by taking the appropriate derivative. the operators here will be anti−Hermitian quaternions acting on quaternions. the Euclidean product of psi (psi* psi) will have all the properties required to form a Hilbert space. The advantage of keeping both operators and the wave function as quaternions is that it will make sense to form an interacting field directly using a product such as psi psi’. This last constraint may initially appear too confining. then I will remain constant in time. With quaternions. That will not be done here. If the position 3−vector which does change is always in the same direction as the momentum 3−vector. As was shown in the section on bracket notation. Contrast this with the typical classical quantum mechanics. This particular constraint is a reflection of that. It must be richer in information content. Adler’s book "Quaternionic Quantum Mechanics"). I is determined directly from E. I see no need to define a vector space with right−hand operator multiplication.

P 2 I Ψ P Ψ or P mv2 2 I Square this operator. If this is used to study photons. analytical power. E and P are zero. Take the spatial derivative of psi using the under the same assumptions: Ψ PI 1 Ψ Et R. not less. energy can be defined with an operator. E R . it also makes solving explicitly for energy impossible unless Et − P.P 0 There are several important cases when this will be true. make one final assumption: 5. Et − R. P 2 mv 2 2m 2 m KE 2 2 The Hamiltonian equals the kinetic energy plus the potential energy. then this equation will apply to the wave front. but it should be possible.P 2 The denominator must be at least 1. HΨ I Ψ t 2 2 Ψ VΨ Typographically.P or t R t E P Now with these 5 assumptions in hand. given we are dealing with a topological algebraic field. if Et t EI Ψ t R. It can serve as a damper. With all of the constraints in place. Phi is not a vector. This give the equation more. this looks very similar to the Schrödinger equation. Since the goal is to make a direct connection to the Schrödinger equation. Ψ t I Ψ E Ψ or E I t The equivalence of the energy E and this operator is called the first quantization. this proposal becomes richer. In a vacuum. then t = |R| and E = |P|. .P 0. As the constraints are removed.81 Ψ t EI 1 Ψ Et R. Unfortunately. There is a damper to quench runaway terms. and a very complicated one at that if you review the assumptions that got us here. a good thing to tame runaway terms. The 3−vector I becomes quite the nightmare to deal with. I expect that this equation will behave exactly like the Schrodinger equation. and can be greater than that.P = 0 Et R. Capital I is a normalized 3−vector.X equals zero. If this number happens to be constant in time. but is a quaternion.

analytical power. I believe that the quaternion analog to Schrödinger equation under the listed constraints will do the task. Implications Any attempt to shift the meaning of an equation as central to modern physics had first be able to regenerate all of its results.Typographically. Capital I is a normalized 82 3−vector. this proposal becomes richer. This give the equation more. not less. given we are dealing with a topological algebraic field. Phi is not a vector. As the constraints are removed. whether the quaternion differential equations will behave better. There is a damper to quench runaway terms. and a very complicated one at that if you review the assumptions that got us here. My sense at this time is that first quaternion analysis as discussed earlier must be made as mathematically solid as complex analysis. but is a quaternion. If it stands on a foundation as robust as complex analysis. At that point. this looks very similar to the Schrödinger equation. it will be worth pushing the envelope with this quaternion equation. I expect that this equation will behave exactly like the Schrodinger equation. These is an immense amount of work needed to see as the constraints are relaxed. the profound problems seen in quantum field theory stand a chance of fading away into the background. With all of the constraints in place. . The 3−vector I becomes quite the nightmare to deal with. but it should be possible.

Pn This infinite sum of quaternions should contain all the information about a system. along with vector identities. 0 2 1 The first quaternion is the conjugate or transpose of the second. Pn n 0 E2 n Pn . In the energy/momentum representation. normed product space. The Wave Function The wave function is the superposition of all possible states of a system. has already been worked out for the Maxwell equations in the Lorenz gauge. The Klein−Gordon Equation The Klein−Gordon equation can be divided into two operators that act on the wave function: the D’Alembertian and the scalar m^2. In this section. this representation of wave functions as an infinite sum of quaternions can form a complete. The quaternion wave function can be normalized. the sum from n 0 to infinity of En . this would involve all possible energy levels and momenta. Pn n 0 En . The product of the conjugate of a wave function with another wave function forms a complete inner product space. Pn 2 . En .83 Introduction to Relativistic Quantum Mechanics The relativistic quantum mechanic equation for a free particle is the Klein−Gordon equation (h=c=1) 2 2 t2 m2 0 The Schrödinger equation results from the non−relativistic limit of this equation. En . the machinery of the Klein−Gordon equation will be ported to quaternions. t . Since the transpose of a quaternion wave function times a wave function creates a Euclidean norm. 2 n 0 2 t . The quaternion operator required to create the D’Alembertian.

and the second term of the vector. and repeat. Pn 2 En . To generate the scalar multiplier m^2. the quaternion operators of the Klein−Gordon equation simplifies to the operators used to generate the Maxwell equations in the Lorenz gauge. En . Set the sum of these two operators equal to zero to form the Klein−Gordon equation. 2 X Pn t . are both equal to zero. Pn En Pn En Pn Pn En En 3 2 n 0 Pn X X Pn Pn Pn X Pn X Pn En 2 En . Pn Under this interpretation. substitute En and Pn for the operators d/dt and del respectively. t2 Pn X Pn En Pn X Pn X Pn 2P n Pn Pn t2 It takes some skilled staring to assure that this equation contains the Klein−Gordon equation along with vector identities.84 2 n 0 2 En t2 Pn En X X Pn . Since the structure of the operator is identical to the previous one. Pn 2 En . there is En^2−Pn^2. instead of the D’Alembertian times the wave function. A n 0 En . a nonzero mass changes the wave equation into a simple harmonic oscillator. The sum of all these terms becomes m^2. What is left is the D’Alembertian operator acting on the quaternion wave function. The simple relationship between the quaternion potential and the wave function may hold for the nonhomogeneous case as well. the same operator acting on two different quaternions equals the same result. In the homogeneous case. . X Pn X En Pn t2 The first term of the scalar. Connection to the Maxwell Equations If m=0. 2 n 0 2 t . This implies that .

This also created several vector identities which are usually not mentioned in this context. the quaternion operator that generated the Maxwell equations was used to generate the Klein−Gordon equation. A quaternion differential equation is needed to perform the work of the Dirac equation. but since quaternion operators are a field. In this section.85 Implications The Klein−Gordon equation is customarily viewed as a scalar equation (due to the scalar D’Alembertian operator) and the Maxwell equations are a vector equation (due to the potential four vector). . an operator that does the task must exist.

Calculate R in this limit to one order of magnitude in the ratio of dX/dt. R dt. so it would not commute. time reversal now depends on velocity. R dt.d X For any given event. 0 if dt The operator R is almost the negative identity. d X R dt. The two most important theories in physics. 2 Β d X. Classical Time Reversal Examine the form of the quaternion which reverses time under two conditions.d X.d X. not local operator.d X.86 Time Reversal Transformations for Intervals The following operator R for quaternions reverses time difference between two events: dt. d X 2 dt d X. dX )(dt. T. the time−reversal operator R can be defined based on the above. so classical time reversal is a global. 2 dt d X dX then R 1 dt2 1. are also local. 0. 2 dt d X 1 dt2 d X. d X 2 dt d X. dX 1 . the change in time dt is much greater than space. d X dt. . d X dt. In the classical limit. so this quality is consistent with those theories. the interval could be characterized by the following: T Ε. d X dt. The operator R will equal (−dt. dX. which is the scalar term of the transpose of a quaternion times itself. 0 Find out what quaternion is required to reverse time for this relativistic interval to first order in epsilon. but the vector is non−zero. A criticism of this operator is that it is local. d X The quaternion R exist because quaternions are a field. the standard model and general relativity. In the classical region. meaning it depends explicitly on spacetime. not the local position in spacetime. Relativistic Time Reversal For a relativistic interval involving one axis. The inverse of quaternion is the transpose over the square of the norm.

zeroes elsewhere. 0. For relativistic interactions. There is no dependence on relative velocity or any local information. but these transformations have not yet been tied explicitly to laws of physics.87 R T2 T2 Ε T T T O Ε Ε Ε 2 2 2 . 1. a result distinct from the classical case. in this limit. The element of this group that reverses time has along its diagonal {−1. . Again. 1. In this section. time reversal involves antiparticles. Therefore special relativity predicts the operation of time reversal should be indistinguishable for classical and relativistic intervals. Time reversal for classical and relativistic intervals have distinct limits. 1}. Implications In special relativity. almost a pure vector. 0]. the boundary condition of time reversal appears to involve entropy. the interval between events is considered to be 4 vector are operated on by elements of the Lorentz group. a time reversal quaternion operator has been derived and shown to work. 0. 2T T T2 T O Ε 2 Ε Ε 2 . 0 This approaches q[−e/T. Yet classically. 0 .1 . 0. 1. the transformation approaches a global transformation.

In a gravity wave. It has three testable predictions. tensors. but at this time I am promoting the form that uses the Christoffel since the odds of it being accepted are higher. vectors.88 Gravity This is a current area of active research for me. There is also an new constant velocity solution for gravity that may eliminate the need for dark matter for the velocity profile of thin disk galaxies. The difference is that the Christoffel symbol is not used. I have two distinct approaches. Light should be bent more at second order parameterized post−Newtonian accuracy. There are aspects of the quaternion approach I prefer. The approach to gravity that uses quaternions makes the same three experimental predictions. the mode of emission will not be transverse (that mode of emission is done by light). The work attempts to unify gravity and EM in a rank 1 field theory using an asymmetric field strength tensor in the Lagrange density. . and the Christoffel symbol to characterize a dynamic metric. It will require the efforts of someone with more technical math skills than I possess to determine if this approach is valid. The first uses scalars.

It must be stressed that the field strength tensor is a second order symmetric tensor. Einstein used a symmetric tensor because the metric tensor is symmetric. The gravitational field equations are analogues to Gauss’ and Ampere’s laws. constant−velocity solutions. since both require stable. the gravitational field would be zero. For geodesic motion. Coulomb got the credit for the electrostatic force law modeled on Newton’s law of gravity. weak−field potential is investigated. so this does not conflict with proofs that at least a symmetric second rank tensor is required to completely describe spacetime curvature. a stable. The Maxwell equations result if the mass current and gravitational field are zero. The four−dimensional wave equation has been quantized before. he was able to show that inside a hollow massive shell. going back to the first studies of electromagnetism in the seventeenth century. Einstein started from the tensor formalism of electromagnetism on the road to general relativity. demonstrated a similar property using a conducting cup. this proposal predicts scalar or longitudinal polarization. How the proposal integrates with the standard model Lagrangian is worked out. The difference between gravity and electromagnetism is the oldest core problem facing physics. Analysis of small perturbations yields changes in the potential that depend on an inverse distance squared. The classic field equations contain both the Maxwell equations and Newton’s field equations under certain conditions.89 Gravity by Analogy to EM Abstract A Lagrange density for gravity is proposed based on a strict analogy to the classical Lagrangian for electromagnetism. There is a precedence for transforming mathematical structures between gravity and electromagnetism. Ben Franklin. Specifically. so they can do the work of gravity. The Lagrangian contains both terms with a connection and the Fermi Lagrangian of electromagnetism. Joseph Priestly realized this meant that the electrostatic force was governed by an inverse square law just like gravity. which may yield insight to the rotation profile of galaxies and early big bang cosmology. After twenty years of effort. and contain the Newton’s gravitational field equation. The scalar and longitudinal modes of emission are interpreted as spin 2 gravitons. A force equation is written based on the same strict analogy to the relativistic Lorentz force of electromagnetism. If gravitational waves are detected. The Taylor series expansion of the metric equation is identical to the Schwarzschild metric to parameterized−post−Newtonian accuracy. If spacetime symmetry is preserved. Instead of an antisymmetric field strength tensor. in his studies of electricity. The field equations are a four−dimensional wave equation. Over a hundred years later. By breaking spacetime symmetry. Introduction The goal of this paper is to create one mathematical structure for gravity and electromagnetism that can be quantized. the cause of the curvature is due entirely to the gravitational and electric potentials. A specific. There are several consequence of this simple procedure. The Taylor series for the two metrics differ for higher order terms and may be tested experimentally. This is a new type of statement about curvature. These field equations are not second rank like those used in general relativity. Newton’s law of gravity results. the gravitational analog to the classic electromagnetic Lagrange density will be written. The process of transforming mathematical structures from electromagnetism to gravity will be continued. Gravity was the first inverse square law. A link to the . By using the chain rule. constant−velocity solution is apparent. normalized. the second−order differential equations can be solved exactly. discovered by Isaac Newton. Eliminating the constants and rearrange terms generates an equation that has the form of a metric equation. The field equations have been quantized before. This makes it reasonable to suppose the Lagrangian can describe both a dynamic geometry required for gravity and the Maxwell equations for electrodynamics. but new interpretations will flow from the unification effort.

so the proposal can be verify or rejected experimentally. so where there is an electrical charge −q. Mass does not have the same units as electric charge. These field equations are not second rank like those used in general relativity. The kinetic energy term is the same as either Lagrangian separately. something which is missing from general relativity. The force equation leads to a geodesic equation where the potential causes the curvature. the source of gravity must also be symmetric. one for a moving change. It must be stressed that the field strength tensor is a second order symmetric tensor. The Maxwell equations result if the mass current and gravitational field are zero. Newton’s law of gravity is apparent if spacetime symmetry is broken. The metrics are equivalent to first−order parameterized post−Newtonian accuracy. L_EM) can be written as covariant derivatives. The minus sign that makes the electromagnetic field strength tensor antisymmetric will be made positive. the regular derivatives in the electromagnetic Lagrangian (Eq. A new class of solutions emerges for the gravitational source where velocity is constant. but the distribution of mass varies with distance. not electric charge. The potential will be plugged into a gravitational force equation analogous to the Lorentz force equation of electromagnetism. The gravitational field equations are analogues to Gauss’ and Ampere’s laws. but new interpretations will flow from the unification effort. If spacetime symmetry is preserved. First. Therefore the metric will past all weak field tests. and contain the Newton’s gravitational field equation. The Lagrangian contains 90 both terms with a connection and the Fermi Lagrangian of electromagnetism. Second. Specifically. solving the force equation and eliminating the constants creates a metric equation similar to the Schwarzschild metric. The moving charge term is a sum. and a third for the antisymmetric second rank field strength tensor: EM Ρ Γ 1 Μ J AΜ c 1 4 c2 Μ AΝ Ν AΜ Μ AΝ Ν AΜ An analogous Lagrangian for gravity would also contain these three components. gravity depends on mass. A link to the Lagrangian of the standard model will be detailed. in order that symmetric object transforms like a tensor requires a replacement of the exterior derivative with a covariant derivative: G Ρ Γ 1 Jm Μ AΜ c 1 4 c2 Μ AΝ Ν AΜ Μ AΝ Ν AΜ The total Lagrangian will be a merger of these two which only apply if the other force is not in effect. The field equations have been quantized before. the gravitational analog to the classic electromagnetic Lagrange density will be written. This makes it reasonable to suppose the Lagrangian can describe both a dynamic geometry required for gravity and the Maxwell equations for electrodynamics. A weak static gravitational field in a vacuum will be studied using standard modern methods: normalizing the potential and looking at perturbations. Third. but three changes are required. Lagrangians The classic electromagnetic Lagrangian density has three terms: one for kinetic energy. because gravity effects metrics which are symmetric. The change in sign is required so that like charges attract for gravity. so this does not conflict with proofs that at least a symmetric second rank tensor is required to completely describe spacetime curvature. This may provide new ways to look at problems with the rotation profiles of disk galaxies and big bang cosmology. This leads to the unified Lagrangian for gravity and electromagnetism: GEM Ρ Γ Jq Μ Jm Μ AΜ 1 2 c2 Μ AΝ Μ AΝ . an inertial mass +m will be substituted.The process of transforming mathematical structures from electromagnetism to gravity will be continued. There are several consequence of this simple procedure. so mass will have to be multiplied by the square root of Newton’s gravitational constant G to keep the units identical. Without loss of generality. The coefficients are different to second−order.

there is more freedom for this Lagrangian. the difference being the inclusion of a mass term. The Christoffel symbols (or connection coefficients) represent derivatives of metrics. x x z . . which is a central accomplishment of general relativity. There is a symmetric analog to the electric E and B fields: To make a connection to the classical fields of gravity and electromagnetism. this Lagrangian could describe the dynamics of the metric. The potential to do both gravity and electromagnetism is here. This establishes a link to electromagnetism. use the following substitutions: E e B b c A t c A t c Φ c Φ y y z z . differing only in the sign of A^v. Because the gauge was not fixed.91 Ρ Γ 1 2 c2 q c2 Μ G m V Μ Φ A v 1 2 c2 ΩΜ Ν Ν ΡΜ AΝ AΝ AΡ AΩ 2 Μ AΝ AΩ The kinetic energy term is for one particle experiencing both gravity and electromagnetism. . which leads to a simpler expression of the Lagrangian: GEM m V 1 Φ t 2 R t 2 q c2 G m V 2 Φ A A v 2 1 2 c2 Φ 2 A t This is almost identical to working with the classical electromagnetic field equation by choosing the Lorenz gauge. the connect is zero. x x y A A c A A z y . The Fermi Lagrangian of electromagnetism is a subset.u. which is required if this field equation does more than just electromagnetism. The classical fields required to represent the field strength tensors should also be similar. Because a dynamic metric is part of the Lagrangian. In local covariant coordinates. Classical Field Equations The field equations can be found by applying the Euler−Lagrange equations to the Lagrange density (assuming the connection is zero for simplicity): 2 AΜ Jq Μ Jm Μ The fields are expressed in terms of the potential. The symmetric and antisymmetric field strength tensors are very similar.

so it is easier to remember. Gauss’ law appears under the following conditions: 2 Ρq 1 c Φ t2 c 2 Φ A t c Φ iff Ρm 0. Define a field g to represent the diagonal elements: g Φ . but all the signs are negative. Substitute the classical fields into the field equations. . and Repeat the exercise for the vector equation.c y Az z Μ AΜ The diagonal of the field strength tensor Μ An is g.c x Ay . that means they bring in an extra minus sign to these expressions for the field equations of the form Μ Ν JΝ : ΜA Ρq c 2 Ρm 1 c E 2 Φ t2 e c 2 Φ g0 t This equation combines Gauss’ law and analogous equation for gravity. starting with the scalar field equation. The rest of the off−diagonal terms are the sum of the magnetic field B and its symmetric analog b.92 The symmetric curl as defined above has all the same differential operators. Every aspect of classical Newtonian gravity can be represented by this proposal under these constraints.c t Ax . The classical Newtonian field equation arises from these physical constraints: Ρm c2 2 Φ A t c Φ and g0 t 0 iff Ρq 0. The symmetric field strength tensor has four more components that lie along the diagonal. If the trace of field strength tensor is zero. can be isolated. The two equations are unified. then the equations are in the Lorentz gauge. The first row and column of the asymmetric field strength tensor is the sum of the electric field E and its symmetric analog e. but under certain physical conditions. This equation should be consistent with special relativity without modification. A relativistic form of the Newtonian gravitational field equation can be seen with the following constraints: Ρm iff Ρq c 2 Φ and A t c Φ 0. Important technical note: all the operators used for the divergences and curls are contravariant.

There are efforts underway to detect the transverse gravitational waves predicted by general relativity. The momentum for the unified Lagrangian of gravity and electromagnetism does not suffer from this problem: ΠΜ h G AΜ c t When expressed with operators. The commutator A0 . With no gravitational field. the Maxwell source equations result.93 Jq 1 2 Jm 1 c E t 2 A t2 c B 2 A e t c b 2 c gu c This has Ampere’s law and a symmetric analog for Ampere’s law for gravity. The homogeneous Maxwell equations are vector identities with these choices of maps to the potentials. This proposal for classical gravitational and electromagnetic field equations is expressed with tensors of rank one (vectors). They discuss "scalar photons". The symmetric second−rank field strength tensor cannot be represented by a photon because photons transform differently than a symmetric tensor. . One must remember that although the field equations are rank one. That field has been quantized before. The field in this proposal must represent both gravity and electromagnetism. the commutator A0 . because those are the modes of transmission. Proc. not a scalar. This proposal predicts the polarity of a gravitational wave will be either scalar or longitudinal. Yet there is no need to make a nonsense particle virtual. One way to realize this is to consider the generalized 4−momentum: ΠΜ h G EM AΜ c t FΜ0 Unfortunately. They introduce a supplemental condition solely to make the scalar and longitudinal modes virtual. The interpretation of these modes appears internally inconsistent to this author. the energy component of the moment operator is zero. Whatever particle does the work must travel at the speed of light like the transverse modes of transmissions of the field. L_GEM generates the same field equations as the classical electromagnetic Lagrangian with the choice of the Lorenz gauge. but photons as the quanta of electric and magnetic fields must transform as a vector. The two transverse modes are photons that do all the work of electromagnetism. 1950). Gupta. N. Π0 ] will equal zero. Soc. first by Gupta and Bleuler (S. and one transverse mode. and are unaffected by the proposal. Canonical Quantization The classical electromagnetic Lagrangian cannot be quantized. If the connection is zero. London. These constraints dictate that the scalar and transverse modes of transmission for this proposal are gravitons. so the field can be quantized. 63:681−691. They determined that there were four modes of transmission: two transverse. The detection of the first gravitational wave polarization will mark either success or failure of this unified field theory. Therefore the two approaches are fundamentally different. the field strength tensor is second rank. Einstein’s field equations are second rank. Π0 ] will not be zero. one scalar. and cannot be quantized. not transverse. Phys.

If electric change changes signs. The different behavior under charge inversion may explain why gravitational force is unidirectional. so nothing has really changed. In curved spacetime. . A more explicit connection will be attempted by condensing the unitary aspects of the symmetries U(1). it can be concluded that the normalized 4−vector is an element of the symmetry group U(1) if the multiplication operator is the metric combined with the parity and conjugate operators. and SU(3) symmetry. and SU(3) with the 4−vectors and a curved metric. One does not need the Y in standard model Lagrangian. velocity and the anti−symmetric field strength tensor: FΜ EM q UΝ c Μ AΝ Ν AΜ Form an analogous force for gravity using the same substitutions as before: FG Μ G m UΝ c Μ AΝ Ν AΜ The gravitational force and the electromagnetic force behave differently under charge inversion. but does so in a precise way (meaning one can calculate what the previous equation should equal). SU(2). so this proposal predicts no Higgs particle will be found. use the conjugate like so: AΜ AΝ gΜΝ AΜ AΝ AΜ A p A0 2 A1 2 A2 2 A3 2 Use the parity operator to flip the sign of the spatial part of a 4−vector: gΜΝ A0 2 A1 2 A2 2 A3 2 Normalize the potential: AΝ p gΜΝ A 1 From this. Start with the standard model Lagrangian: SM ΓΜ DΜ Τa a W 2 Μ Λb b G 2 Μ where DΜ Μ igEM YAΜ igweak igstrong The electromagnetic potential A_mu is a complex−valued 4−vector. then both side flip signs. Since it is complex−valued. There is no need for the Higgs mechanism to give particles mass while preserving U(1)xSU(2)xSU(3) symmetry. the previous equation will not equal one. but electrical forces can attract or repulse. so this simplifies things. Forces The Lorentz Force of electromagnetism involves charge.94 Integration with the Standard Model The standard model does not in an obvious way deal with curved spacetime. If the mass changes signs. the change in momentum will not change signs. The only way to form a scalar with a 4−vector is to use a metric. Mass breaks U(1). SU(2). The same logic applies to the 4−vector potentials for the weak and the strong forces which happen to have internal symmetries.

The covariant derivatives on the right side of F_gEM can also be expanded: 0 q q G m G m xΝ c Τ xΝ c Τ Μ Ν AΝ 2 m c ΜΝ Ω AΜ UΝ UΩ This equation says that spacetime curvature is caused by the change in the potential if there is no external force. The different behavior under charge inversion may explain why gravitational force is unidirectional. This is a novel statement. all the acceleration seen in spacetime is due to spacetime curvature. so time must also be incorporated. Only first order terms will be kept. In general relativity. the Christoffel symbol. the change in momentum will not change signs. one compares two geodesics.The gravitational force and the electromagnetic force behave differently under charge inversion. or the negative of the square of the Lorentz invariant interval tau. and then the definition of a covariant derivative to form a geodesic equation: 2 0 m xΜ Τ2 m c Μ ΝΩ UΝ UΩ This equation says that if there is no force. This singularity may not be problematic because massless particles are described by the Maxwell equations. The unified geodesic equation asserts that the curvature can be calculated directly from the potential. and based on an analysis of the tidal forces between the geodesics. The idea is to consider the time contribution to be very small relative to the distance. an approach that will be followed here. 0) solves the field equations. then both side flip signs. It must be established that the terms coupled to the mass m are connected to what is known about gravity. so nothing has really changed. If electric change changes signs. A 1/distance potential does not solve the field equations in four dimensions. the potential A_mu = (1/sigma^2. The unified field equations are relativistic. but electrical forces can attract or repulse. and then put the solutions into the force equation. not time. The next task is to find a solution to the unified field equations. the equation has the form of a Killing’s equation. but that hope will required a detailed study. The potential has a point singularity where R = 0. In local covariant coordinates where the connection is zero. the equation approaches the form of the Lorentz force law of electromagnetism. The Poisson field equation of classical Newtonian gravity can be solved by a 1/R potential. Distance is used instead of the interval because classical gravity depends on distance. Such a potential has as a singularity that is the entire lightcone. Notice that this equation contains terms linked to a mass m and a charge q. determines the curvature. The total force is a combination of the two: FGEM Μ Jq Ν Jm Ν Μ AΝ Jq Ν Jm Ν Ν AΜ If q >> G m. where sigma squared is the Lorentz invariant distance. Gravity is a weak effect. If the force is zero. If 95 the mass changes signs. which is used to determine the isometries of a metric. Gravitational Force for a Weak Field The total unified force law is relevant to physics because it contains the Lorentz force law of electromagnetism. Geodesics are defined by examining the left−hand side of F_GEM: m UΜ Τ m UΜ Τ UΜ m Τ 0 Assume dm/dtau = 0. Here is the normalized potential with a linear perturbation: . Normalizing means there are small steps will be away from one. Apply the chain rule. so the geodesic equation applies to electromagnetism as well as gravity. where sigma ^2 = 0. It is common in quantum mechanics to normalize to one and study perturbations of weak fields.

x. Here is the normalized potential with a linear perturbation: 96 AΜ c G h . Normalizing means there are small steps will be away from one.Gravity is a weak effect. One interesting aspect is the shift of units from one that depends on h − suggesting quantum mechanics − to the normalized perturbation which appears to be classical because there is no h. It is common in quantum mechanics to normalize to one and study perturbations of weak fields. so this is promising. The classical Newtonian dependence on distance is an inverse square. The sign of the spring constant k does not effect the solution to the 4D wave field equations but does change the derivative of the potential. One problem is that a potential that applies exclusively to gravity is sought. yet the non−zero gradient of Φ indicates an electric field.0 c2 Σ2 G 1 2 1 2 k x Σ2 2 2 1 2 1 2 k y Σ2 2 2 k z Σ2 k t Σ2 . diagonal SHO AΜ c G 1 1 2 1 2 k x Σ2 2 2 1 2 1 2 k y Σ2 2 2 k z Σ2 k t Σ2 .0 This potential solves the 4D wave equation because the shift by the one over root two factor and the rescaling by the spring constant k over sigma square do not effect the differential equation. an approach that will be followed here. or destructively interfere to eliminate derivatives. A potential that only has derivatives along the diagonal of the field strength tensor can be constructed from two potentials that differ by spring constants that either constructively interfere to create non−zero derivatives. Only first order terms will be kept. and z: Φ t c c c Φ x Φ y Φ z c2 k G Σ2 O k2 O k2 O k2 O k2 c2 k G Σ2 c2 k G Σ2 c2 k G Σ2 The change in the potential is a function of a spring constant k over sigma squared. Take the derivative with respect to t. y.

1 1 2 1 2 2 2 1 1 2 1 2 k x Σ2 k z Σ2 k x Σ2 k z Σ2 2 2 1 2 1 2 1 2 1 2 k y Σ2 k t Σ2 k y Σ2 k t Σ2 2 2 2 2 .97 2 2 1 1 2 1 2 k x Σ2 k z Σ2 k x Σ2 k z Σ2 2 1 2 1 2 1 2 1 2 k y Σ2 k t Σ2 k y Σ2 k t Σ2 2 2 2 . 1 1 2 1 2 2 2 1 1 2 1 2 k x Σ2 k z Σ2 k x Σ2 k z Σ2 2 2 1 2 1 2 1 2 1 2 k y Σ2 k t Σ2 k y Σ2 k t Σ2 2 2 2 2 . 1 1 2 1 2 2 2 1 1 2 1 2 k x Σ2 k z Σ2 2 2 1 2 1 2 k y Σ2 k t Σ2 2 2 .

This is not quite Newton’s gravitational force law. then one gets back Newtonian acceleration. assume that the spring constant is due to the source mass. so that Σ2 R2 c2 t2 R2 . Newtonian spacetime is different from Minkowski spacetime because the speed of light is infinite. so this would necessarily be zero. 2 Τ Σ R Τ This is a relativistic force law for a weak gravitational field for the inverse interval squared diagonal potential. Τ c Τ GMm R R2 c t . By the static field approximation. Spacetime symmetry must be broken. In the approximation. R Plug these three assumptions into force equation: FΜ g 0. The functional relationship between time and space must be severed. The reason is that one must consider the left− hand side of the force equation carefully. it would have to be with respect to the absolute value of R as has been done earlier in the derivation. For logical consistency. A question arises about how to do this in a formal mathematical sense. this equation will lead to Newton’s law of gravity in the next section. one might be tempted to also substitute time in the dm/dtau term. First. Newton’s Law of Gravity and More Several assumptions need to be made to apply the weak gravitational force equation to a classical gravitational system. However. then that will sever the functional relationship between time and space: t R . Μ AΝ c2 G k Σ2 1 0 0 0 0 1 0 0 0 0 1 0 0 0 0 1 All this work to get a multiple of the identity matrix! Plug this into the gravitational force equation: FΜ g mc k Σ2 t k . For the acceleration term. If spacetime symmetry is maintained. k = GM. So the classical force law should look like so: . then solving the force equation and eliminating the constants yields a metric equation for gravity. it does not depend on time. if the interval is only about time. If the interval tau is replaced by the scalar distance R. there is a distance R which is the same magnitude as the interval tau. When spacetime symmetry is broken. the system is presumed to be static. If this derivative is to have any chance at being non−zero. assume that the field is static. According to the chain rule: m UΜ Τ m UΜ Τ UΜ m Τ An open question is how should spacetime symmetry be broken for the derivatives with respect to the interval tau? An interval is composed of both changes in time and space. keeping only the terms to first order in the spring constant k. The Minkowski interval tau is a consequence of the relationship between time t and space R.98 Take the contravariant derivative of this potential. R R R 0. Second.

It is only if the inertial mass is distributed over space like for the big bang or galaxies will the term come into play. Numerical approaches on the above equation should be conducted.An open question is how should spacetime symmetry be broken for the derivatives with respect to the interval tau? An interval is composed of both changes in time and space.[S. This new effect will necessarily be small due to the constant c in the term. Their is also a problem with the rotation profile of thin disk galaxies. M. M. so this would necessarily be zero. D. so Τ2 R2 c2 t2 R2 c2 . it would have to be with respect to the absolute value of R as has been done earlier in the derivation.. one might be tempted to also substitute time in the dm/dtau term. However. J. 91:1301−1327. and plug back into the constraint. V). There is one more degree of freedom. A Metric Equation The weak gravitational force equation is two second−order differential equations. 93:816−832. that for the velocity vector. if the interval is only about time. Astron. 139:1217. 1987] Once the maximum velocity is reached. 1964] Both problems may again be resolved with stable constant velocity solutions. Big bang cosmology has two problems: all matter is traveling at exactly the same speed even though it is not possible for them to communicate (the horizon problem). For a point source.[A. For logical 99 consistency. VeΤ k For flat spacetime. To make the metric consistent with experiment. J. S. Phys. the dm/d|R| term will not make a contribution. Toomre. then the acceleration is zero. If this derivative is to have any chance at being non−zero. and the model require high levels of precision on initial conditions to avoid collapse (the flatness problem). Guth. H. If the velocity is constant. the system is presumed to be static. k = GM/c^2. The solution to the force equation when there is no acceleration is a stable exponential. The equation can be simplified to a set of first−order differential equations by substituting (U^0. choose the negative root: . Kent. Kent. Astrophys. Astron. J. 1981] The force equation has a stable. Rev. So the classical force law should look like so: 2 m R t2 c R t m R GMm R2 R V This is Newton’s law of gravity working along a new direction. and one gets Newton’s law of gravity.. The constraint on relativistic velocities in flat spacetime is: UΜ UΜ c2 dt2 dR2 dΤ2 c2 v2 V V Solve for the constants. constant velocity solution which may resolve both problems of the big bang without the inflation hypothesis. the velocity stays constant. multiplying through by dtau ^2.. and the field is static. U) = (c dt/dtau. dΤ2 e 2 k Τ dt2 e2 k Τ dR2 c2 Make the same two assumptions as before: the spring constant is due to the gravitational source. because the radius R could either be positive or negative. The equation describes the distribution of the inertial mass m that makes up the total gravitational source mass M. dR/dtau) U0 Τ U Τ UΜ k 0 U Τ2 k U Τ2 k Τ 0 0 The solution involves exponentials: ve . It has been shown that galaxies should not be stable at all. then one gets back Newtonian acceleration. For the acceleration term.. 23:347−356. [A. 1986. U^mu = (v.

Conclusion Using a nineteenth century approach. the unitary aspect of the standard model can be self−evident. . The description of geodesics by general relativity is not complete because it does not explicitly show how the potential source causes curvature. This theory makes three testable predictions. In the standard model as elsewhere. not transverse as predicted by general relativity. M. the two metric will predict different coefficients. A dynamic metric equation is found but it uses a simpler set of field equations (a rank one tensor instead of two). because the radius R could either be positive or negative. There is one more degree of freedom. "Theory and experiment in gravitational physics: Revised edition". First. To make the metric consistent with experiment. k = GM/c^2. the coefficients for the metric derived here are different from the Schwarzschild metric in isotropic coordinates. It will require a great deal of effort and skill to conduct such experiments. since many physical phenomena will have to be accounted for (an example: the quadrupole moment of the Sun for solar tests). It is this test which should be investigated first. Such an experiment will be quite difficult to do. an effort to unify physics from the twentieth century has been attempted. the coefficients are identical. Perform a Taylor series expansion to second order in GM/c^2R: dΤ2 1 2 GM c2 R 2 GM c2 R 2 dt2 1 2 GM c2 R 2 GM c2 R 2 dR2 If one compares this metric to the Schwarzschild metric in isotropic coordinates to parameterized post−Newtonian accuracy. this metric is consistent with all experimental tests weak field tests of general relativity. one not. 1993. two subtle. Clifford Will − a leading authority in experimental tests of gravity theories − no experiments are being planned to monitor second order PPN coefficients at this time. Second. so Τ2 R2 c2 t2 R2 c2 . the polarity of gravitational waves will be scalar or longitudinal. Cambridge University Press. According to personal communication with Prof. if gravitation effects are measured to secondary parameterized post−Newtonian accuracy. The validity of this proposal can thus be tested experimentally. For that reason. Will. [C. By normalizing the 4−vectors. The third test is to see if the complete relativistic force equation matches all the data for a thin spiral galaxy. and the field is static.] For higher order terms of the Taylor series expansion.Make the same two assumptions as before: the spring constant is due to the gravitational source. choose the negative root: 100 dΤ2 e 2 GM c2 R dt2 e2 GM c2 R dR2 This equation has the form of a metric equation. combining two 4−vectors requires a metric.

antisymmetric−matrix field strength quaternion for magnetism. a classical link to the quantum Aharonov−Bohm effect. Both forces can be attractive. Albert Einstein worked to unify gravity and electromagnetism. Albert Einstein had a specific idea for how to formulate an acceptable unified field theory (see Fig." the science and life of Albert Einstein". "Subtle is the Lord. the first force characterized mathematically by Isaac Newton. The parallels between gravity and electromagnetism are evident. the algebra is complicated. Box* A*.−g) Einstein’s Vision Using Quaternions Three of the four known forces in physics have been unified via the standard model: the electromagnetic. The holdout remains gravity. but there are residual terms whose meaning is not clear presently. A long−standing goal of modern physics is to explain the similarities and differences between gravity and electromagnetism. The unified field equations for an isolated source are generated by acting on the unified force field with an additional differential operator. This contains a quaternion representation of the Maxwell equations. Box* Box* A* = 4 pi J*. antisymmetric−matrix field strength quaternion for electricity and an odd. 1982). An additional constraint is required for gauge transformations of a massive field. The equivalence principle of general relativity which applies only to mass is generalized because for any Riemannian quaternion differential equation. Claredon Pres. Hamilton’s quaternions are a 4−dimensional topological algebraic field related to the real and complex numbers equipped with a static Euclidean 4−basis.101 Einstein’s Vision I: Classical Unified Field Equations Abstract The equations governing gravity and electromagnetism show both profound similarities and unambiguous differences. a differential operator acting on the field strength tensor creates the Maxwell equations. The field equations conserve both electric charge density and mass density. Pais.. Unfortunately. Vacuum and zero net current solutions to the unified field equations are discussed. the weak. This generates an even. Newton’s law of gravity and Coulomb’s law are inverse square laws. 1. Riemannian quaternions as defined herein explicitly allow for dynamic changes in the basis vectors. The Maxwell equations are generated using quaternion potentials and operators. being formed by a quaternion differential operator acting on a potential. Gauge symmetry is broken for massive particles by the even. and dynamic field equations for gravity. the chain rule means a change could be caused by the potential and/or the basis vectors.10. and the strong forces.. (PACS:12. where the even field conserves its sign if the order of the differential and the potential are reversed unlike the odd field. In tensor analysis. The unified force field proposed is modeled on a simplification of the electromagnetic field strength tensor. the gravitational and electromagnetic fields are Lorentz invariant and Lorentz covariant respectively. which is interpreted as being due to gravity. taken from A. but Coulomb’s law can also be a repulsive force. Under a Lorentz transformation. symmetric−matrix term. . mainly by trying to generalize Riemannian geometry.

The sum or difference of two quaternions can only be defined if the two quaternions in question share the same 4−basis. S. Physical laws are the result of simple Riemannian quaternion differential equations. topological. 1992). can be added. and space is a 3−vector. The equations of the first two papers are recast with the new definition of a quaternion derivative. if totally ordered. Weinberg. even those in electromagnetism. First−order Riemannian quaternion differential equations create force fields for gravity. The construction of a new 4−dimensional geometry is dictated by insights garnered from physics. and a classical counterpart to the Aharanov−Bohm effect of quantum mechanics. subtracted. resulting in a quantum unified field and force theory. the Maxwell equations for electromagnetism. associative. For this reason. an idea which is not shared by most of today’s thinkers (S. these three numbers are the only finite−dimensional. this paper focuses exclusively on quaternions. New York. electricity. There are four degrees of freedom divided into two dissimilar parts: time is a scalar. Laws of physics are expressed in a coordinate−independent way. and if multiplication commutes: . Third−order differential equations create conservation laws. the field equations make the metric a dynamic variable. the 4−dimensional number where the terms scalar and vector where first used. and magnetism. The basis vectors of Riemannian quaternions can be dynamic. The dynamic nature of the basis vectors leads to the general equivalence principle. Homogeneous solutions to the second order differential equations are related to gauge symmetry. To a degree which has pleasantly surprised the author. In general relativity. Second−order differential equations create dynamic field equations for gravity. Properties of these numbers are summarized in the table below by dimension.102 One unusual aspect of Einstein’s view was that he believed the unified field would lead to a new foundation for quantum mechanics. The third paper develops a new approach to quaternion analysis. Events in spacetime are composed of a scalar for time and a 3−vector for space. The four−dimensional topological algebraic field of quaternions has the same structure. translated from the Russian by Emma Lehmer. Riemannian quaternions make coordinate−independence explicit. algebraic fields. so the metric can be dynamic. This structure should be reflected in all the mathematics used to describe patterns of events. whereby any law. up to an isomorphism (L. Princeton University Press." Pantheon Books. where the net change in 3−space is equal to the net change in time. Hamilton’s quaternions. Events in Spacetime and Quaternions An event in spacetime is considered by the author as the fundamental form of information in physics. with a focus on a particular solution which may eliminate the need for dark matter to explain the mass distribution and velocity profile for spiral galaxies. Technically. In special relativity. Events have structure. can be the result of a change in reference frame. along with the far better know real and complex numbers. and divided. 1939). "Dreams of a final theory. multiplied. "Topological groups". so quaternions will be the starting point of this effort. regions in spacetime are delimited by the light cone. Einstein’s vision to unify gravity and electromagnetism has been followed. The second paper in this series of three investigates a unified force law. Most of Einstein’s efforts over 40 years were directed in a search to generalize Riemannian differential geometry in four dimensions. The parity between changes in 3−space and time is constructed into the definition of a Riemannian quaternion. Pontryagin.

all transformations are also constrained to the same division algebra. For the sake of consistency. No corresponding principle applies to electromagnetism. y. ihat_0. "Topological groups". i. i_0. a_2. a_3 ihat_3/3). the a_0 dihat_0/di_0 term. and k.103 Hamilton’s quaternions. This constraint might first appear too restrictive since for example it eliminates simple matrices for row permutations. instead of weighing all four equally. multiplied. a quaternion equation is independent of the chosen basis. and an arbitrary 4−basis. there necessarily exists a combination of quaternions that achieves the action of a permutation. Laws in physics are independent of coordinate systems. can change over an infinitely small amount of time. amplitudes and basis vectors will be separated using a new notation. translated from the Russian by Emma Lehmer. can be added. i^2=j^2=k^2=ijk=−1. Therefore. j. algebraic fields. 1939). subtracted. Any change in a quaternion potential function could be due to contributions from a change in potential. x. that the inertial mass equals the gravitational mass. the line that divides causality is define by the light cone. z t x y z x t z y y z t x z y x t Although written in Cartesian coordinates. and divided. and if multiplication commutes: Number Dimensions Totally Ordered Commutative Real 1 Yes Yes Complex 2 No Yes Quaternions 4 No No Hamilton’s quaternions have a Euclidean 4−basis composed of 1. ihat_3). ihat_1. if totally ordered. a_1. A_n=(a_0. up to an isomorphism (L. and/or a change in the basis. Is this mathematical property related to physics? Consider Gauss’ law written with Riemannian quaternions: . with experiments to back it up. associative. S. The rules of multiplication were inspired by those for complex numbers: 1^2=1. which depends only on the electromagnetic field tensor built from the potential. In spacetime. ihat_2. a_2 ihat_2/3. On the light cone. Ihat_n=(ihat_0. Pontryagin. A coordinate−independent Riemannian quaternion is defined to be A_0 Ihat_n=(a_0 ihat_0. Princeton University Press. An accelerated reference frame can be indistinguishable from the effect of a mass density. a_1 ihat_1/3. Technically. the total change in 3−space over the change in time is equal to one. Properties of these numbers are summarized in the table below by dimension. the 4−unit vector does not have to be static. quaternions can be written in any linearly−independent 4−basis because matrix algebra provides the necessary techniques for changing the basis. Quaternions also have a real 4x4 matrix representation: q t. like tensors. The equivalence principle of general relativity asserts. Physics therefore indicates parity between the total 3−vector and the scalar. To make the coordinate independence explicit. Consider a quaternion 4−function. Since quaternions are an algebraic field. With Riemannian quaternions. The need for consistency will overrule convenience. these three numbers are the only finite−dimensional. a_3). One could view quaternions as tensors restricted to a 4−dimensional algebraic field. the ihat_0 da_0/di_0 term. topological. as illustrated by taking the time derivative of the first term and using the chain rule: a0 i0 i0 i0 a0 i0 a0 i0 i0 The unit vector for time. along with the far better know real and complex numbers.

This additional information about quaternion products will suggest a provocative link between metrics and inertia consistent with both special and general relativity. but also has information in the 3−vector. the mathematical properties of commuting and anti−commuting algebras are well known. only shared. then multiplication is anti−commutative. and those products will transform differently. The four square basis vectors i_mu^2 define the metric. Metrics and Quaternion Products The theories of special and general relativity dictate the distance between events in spacetime. inertia is a link between the two. The basis does not have to be constant.104 2 in 9 En in in En 9 in in 4 ΠΡ. If the basis vectors are not constant. Instead of the imaginary number i. n 1. Quaternion products multiply two 4−basis vectors. If. If the order of quaternion multiplication is reversed. First. D I ’ ac BD I I ’. 2 da0 dAn 0 n 9 3 2 The signs were chosen to be consistent with Hamilton’s quaternion algebra. two different unit 3−vectors have both a dot and cross product. A quaternion product is the superposition of these two types of algebras that forms a division algebra. aD I ’ Bc I BD I I ’ Complex numbers commute because they do not have a cross product in the result. on the other hand. the only value of a quaternion product is equal to the cross product. It makes no sense to subtract something in spherical coordinates from something in Cartesian coordinates. Notice that the 3−rope starts in one time−space location and will have a non−zero length if it ends up at a different location and time. the divergence of the basis vectors. (a. which have the form i_0 I_n. then quaternion multiplication has all of the properties of complex numbers. D I ). Every quaternion commutes with itself. Special relativity dictates the transformation rules for observers who change their inertia. assuming the system observed does not change. Compare the product of multiplying two complex numbers (a. di ac bd. There are seven unique pairs of basis vectors in a square: da0 i0 . The field equations of general relativity detail the changes in distance due to a system changing its inertia from the vacuum to a non−zero energy density. The general equivalence principle is applicable to both gravity and electromagnetism. so the cross product is zero. With quaternion products. Define a "3−rope" to be the three other terms. bi) and (c. Most structures in Nature do not transform like a scalar and a 3−vector. there is a unit 3−vector for each quaternion playing an analogous role. bi c. dAn In 3 2 da0 i0 2 2 dAn 2 In i I . Several steps are required to square of the difference of two Riemannian quaternions to form a measure of distance. the basis of the two quaternions must be shared. the 3−rope is a natural companion to a metric for information about distance. Although fundamentally different in their mathematical structure. The difference is that unit 3−vectors do not all have to point in the same direction. . with two quaternions. The rules of quaternion multiplication mirror those of complex numbers. A quaternion product necessarily contains information about the metric. then the metric is dynamic. then only the cross product would change its sign. B I ) and (c. a. di): a. Quaternion multiplication does not commute due to the behavior of the cross product. B I c. The "general equivalence principle" as defined here means that any measurement can be due to a change in the potential and/or a change in the basis vectors. ad bc . 3 The divergence of the electric field might equal the source. The dot and cross products completely characterize the relationship between the two unit vectors. or equivalently. If the cross product is zero. Individually. Based on the angle between them. 2.

the gradient and the curl all in one. a metric with this property will be found and discussed. However. so the reader is encouraged to recognize that there are also a parallel set of terms for changes in the basis vectors. but that particular basis does not have to be used. y. Quaternion operators and potentials have not been used to express the Maxwell equations. but made the more obvious choice of i_0^2 = −I_n^2. the notation introduced for Riemann quaternions has been suppress. complex−valued quaternions are not an algebraic field. because it is known how it changes. have used complex−valued quaternions for the task because the extra imaginary number can be used to get the signs correct.105 The signs were chosen to be consistent with Hamilton’s quaternion algebra. A complementary hypothesis to the invariant metric of special relativity would propose the if the inertia of the system but not the observer is changed. The existence of a basis where the 3−rope is constant despite a change in the inertia of the system will have to be treated as provisional in this paper. The previous equation is a complete assessment of the change in the 4−dimensional potential/basis. A): t . Therefore. the metric is invariant. t^2+x^2+y^2+z^2. This paper involves the constraint of working exclusively with 4−dimensional algebraic fields. the divergence. In the second paper of this series. i0 In 1 If the magnitude of the time and 3−space basis vectors are inversely related. This pattern will be repeated starting from an asymmetric field to create the same field and force equations using quaternion differentials and potentials. there exists a choice of basis vectors such that the 3−rope is invariant but the metric changes in a known way. Define a "3−rope" to be the three other terms. Φ. If the basis vectors are not constant. the 3−rope is a natural companion to a metric for information about distance. if the inertia of the observer but not the system is changed. A Φ t . Hamilton had the freedom to use the rule found in the above equation. The Maxwell equations are formed by acting on the field with another differential operator. as is the case here. In the Maxwell equations. x. The four square basis vectors i_mu^2 define the metric. and a 4−velocity. A t Φ xA For the sake of clarity. Notice that the 3−rope starts in one time−space location and will have a non−zero length if it ends up at a different location and time. Del) acts on a potential function (phi. the divergence and the curl involving the electric and magnetic field are all positive. As a first step to constructing differential equations. then the metric is dynamic. and z were complex. With quaternion products. The 3−rope is covariant. . Physically Relevant Differential Equations Is there a rational way to construct physically relevant quaternion equations? The method used here will be to mimic the tensor equations of electromagnetism. the electromagnetic field strength tensor. even in Maxwell’s time. This could be written in algebraically using the following rule: 2 i0 1 In 2 . examine how the differential operator (d/dt. involving two time derivatives. it is not relevant to this paper. The reason can be found in the previous equation. Many others. to see how every term connects to established laws of physics. for a non−zero quaternion could equal zero if the values of t. which have the form i_0 I_n. The electromagnetic field strength tensor is formed by a differential operator acting on a potential. A unified field theory should account for all conceivable forms of change in a 4−dimensional potential/basis. the magnitude of the product of the time basis vector with each 3−space basis vector will be constant even if the basis vectors themselves are dynamic.A. The Lorentz 4−force is created by the product of a electric charge. The challenge in this exercise is in the interpretation. where the sign of the divergence of A is opposite of the curl of A. In special relativity. This hypothesis asserts there exists such a basis. no matter how salutary the work with complex−valued quaternions. The norm.

For the sake of logical consistency. where the sign of the divergence of A is opposite of the curl of A. 1954). In the Maxwell equations. I arrive at the Maxwell equations for free space. In quaternion parlance." Dover reprint. any system of differential equations. The norm. such as the Maxwell equations. because taking the transpose will flip the signs of the 3−vector. y and z in the representation above) is off−diagonal and can be represented by an antisymmetric 4x4 matrix. This can be done with the vector operator. Nature is asymmetric. invariant under transposition and conjugation (these are the same operations for quaternions). this is a quaternion−scalar formed from a differential quaternion acting on a potential. y. t^2+x^2+y^2+z^2. The Maxwell equations are gauge invariant. How can this property be built into a quaternion expression? Consider a common gauge such as the Lorenz gauge. x. gradient. and z were complex. Therefore." [einstein1934] The "four−dimensional continuum" could be viewed as a technical constraint involving topology. If in this I adopt a Riemannian metric. The 3−vector component (x. containing both a symmetric metric for gravity and an antisymmetric tensor for electromagnetism. complex−valued quaternions are not an algebraic field. However. third edition. and look for the simplest laws which such a metric can satisfy. the quaternion−scalar must be set to zero. (q−q*)/2. If I adopt in this space a vector field. Quaternions are asymmetric in their matrix representation. have used complex−valued quaternions for the task because the extra imaginary number can be used to get the signs correct. quaternions do have a topological structure since they have a norm. x. Fortunately. dphi/dt + div A = 0. I arrive at the relativistic gravitation theory of empty space. Maxwell. The reason can be found in the previous equation. B. The reason to hope for unification using quaternions can be found in an analysis of symmetry provided by Albert Einstein: "The physical world is represented as a four−dimensional continuum. The divergence. or the antisymmetric tensor field derived from it. Recreating the Maxwell Equations Maxwell speculated that his set of equations might be expressed with quaternions someday (J. even in Maxwell’s time. y. C. E t t A . To be invariant under an arbitrary gauge transformation. Search for a combination of quaternion operators and potentials that generate the Maxwell equations: Vector A 2 Vector A t B Φ A A . With this in mind.Quaternion operators and potentials have not been used to express the Maxwell equations. "Treatise on Electricity and Magnetism. z t 0 0 0 0 t 0 0 0 0 t 0 0 0 0 t 0 x y z x 0 z y y z 0 x z y x 0 The scalar component (t in representation above) can be represented by a symmetric 4x4 matrix. and if I look for the simplest laws which such a field can satisfy. it is not relevant to this paper. no matter how salutary the work with complex−valued quaternions. Many others. for a non−zero quaternion could equal zero if the values of t. This paper involves the constraint of working exclusively with 4−dimensional algebraic fields. and curl were initially developed by Hamilton during his investigation of quaternions. a property which is critical to using them for unifying gravity and electromagnetism. the divergence and the curl involving the electric and 106 magnetic field are all positive. that depends on these tools must have a quaternion representation. rewrite out the real 4x4 matrix representation of a quaternion: q t.

Gauss’ law. The Maxwell equations apply to massless particles. Of course the signs of the Maxwell equations cannot be changed. More information about the potential might be used in unification of electromagnetism with gravity. Justification comes from the plea for gauge symmetry. the Maxwell equations are sufficient to explain classical and quantum electrodynamic phenomena in a gauge−invariant way. Gauge symmetry is broken for massive fields. However. E Vector A t A A t Φ 4 ΠΡ. The unified field hypothesis proposed involves a quaternion differential operator acting on a quaternion potential: . essential for the Maxwell equations. 0 . consistency and completeness. This is Gauss’ and Faraday’s law. if the vector operator had not been used. two quaternion equations are required. Generate the other two Maxwell equations: Vector A t E. the divergence always has a sign opposite the curl. A complicated collection of sums or differences of differential operators acting on potentials − along with their conjugates − is required. To unify electromagnetism with gravity. the gauge symmetry must be broken. To specify the Maxwell equations completely. B t A 2 Φ . When a quaternion differential acts on a function. so it could provide a complete picture concerning symmetry. Ampere’s law. the quaternion expression is unappealing for reasons of simplicity. One motivation for the search for a unified potential field involves simplifying the above expressions. The opposite situation applies to the Maxwell equations. There is no obvious reason this combination of terms should be central to the nature of light. In the standard approach to the electromagnetic field. opening the door to massive particles. Because of the constraints imposed by quaternion algebra. it may be worth the effort to explore equations with sign conventions consistent with the quaternion algebra. just like the 4−vector approach. 4 Π J . Again. Although successful. This is Ampere’s law and the no monopoles vector identity (assuming a simply−connected topology). the time derivative of the symmetric−matrix force field would be associated with the electromagnetic source equation. then the gradient of the symmetric−matrix force field would be linked to the electromagnetic source equation. A gauge is also matrix symmetric.107 0. there is little freedom to choose the gauge with a simple quaternion expression. One Unified Force Field from One Potential Field For massless particles. Information about the change in the potential is explicitly discarded by the vector operator. Any choice of gauge will not make a contribution due to the vector operator. If the vector operator was not used. where the operators for divergence and curl were conceived. a differential 4−vector acts on a 4−vector potential in such a way as to create an antisymmetric second−rank tensor.

It will be shown subsequently how the even force fields can partially constructively or destructively interfere with each other. no new results beyond the Maxwell equation are obtained. In some ways. Assuming spacetime is simply connected. A quaternion potential function has four degrees of freedom represented by the scalar function phi and the 3−vector function A. the ones that will not change signs if the order of the differential operator and the potential are reversed. E. g. The antisymmetric−matrix component of the unified field has the same elements as the standard electromagnetic field tensor. so there are only four degrees of freedom. A t Φ xA This is a natural suggestion with this algebra. y 1 3 Az z 2 Take the derivative of the Lagrangian with respect to the gauge variables: 0 By Noether’s theorem. This is why the proposal involves new physics. Φ t2 . as Gupta did in 1950 in order to quantize the Maxwell equations. Examine just the gauge contribution to the Lagrangian for this unified field: L 1 2 L AΜ xΜ t . A new constraint arises because both the force fields for gravity and electricity are even. the no monopoles equation is another identity. With seven components to the three classical force fields. Φ. and B. The electric and magnetic field form a vector identity via Faraday’s law. this conserved current indicates a symmetry of the Lagrangian. depend on the same quaternion potential. The magnetic field B is the curl of A. x 1 3 Ay . He concluded that although useful because it is written in manifestly relativistic form. 1 3 Ax . there must be three constraints between the fields. Unified Field Equations In the standard approach to generating the Maxwell equations. A unified field hypothesis for an isolated source is proposed which involves a differential quaternion operator acting on the unified field: 4 Π Ρ. Define the electric field E as the even terms. a differential operator acts on the electromagnetic field strength tensor. Two constraints are already familiar. A A . Acting on this with one[or more] differential operators does not change the degrees of freedom.108 A Φ t . The three classical force fields. Φ t .A. the odd term. the above equation looks just like the old idea of combining a scalar gauge field with the electromagnetic field strength tensor. Instead. The gauge is a dynamic variable constrained by the Lagrangian. The justification for proposing the unified force field hypothesis rests on the presence of the electric and magnetic fields. the tangent spaces of the potential will offer more subtle views on the rules for how potentials change. J 2 t 2 A t .

Rev. The presence of the Maxwell equations justifies the investigation of the unified field equations." Phys.B. one can reasonably expect the differential equations will share many properties.109 2 2 Φ t A t 2 A Φ . The unified source can be defined in terms of more familiar charge and current densities by separately setting the gravity or electromagnetic field equal to zero. All that needs to be done is to subtract twice the time derivative of the magnetic field from both sides. This leads to connections between the unified source. One might imagine that a Lorentz transformation would shift this pseudo−current density into a pseudo−charge density. Since they are generated alongside the Maxwell equations. with the ones involving the symmetric−matrix gravitational force field being more symmetric than those of the electromagnetic counterpart. 1959) depends on the total magnetic flux to create changes seen in the energy spectrum.E B t . and Ampere’s law. The Aharonov− Bohm effect (Y Aharonov and D. "Significance of electromagnetic potentials in the quantum theory. and charge: J Jm iff E B 0 . may have a classical analogue in the unified field equations. first viewed as a purely quantum effect. The inclusion of a pseudo−current along with a current making the proposal more complete. transforming differently under space inversion than the electric current density. the source is due only to electricity or mass respectively. There is a simple relationship between Faraday’s law and the equation above. The field equations involving the gravitational force field are dynamic and depend on four dimensions. In these cases. the no magnetic monopoles law. a subset of the unified field equations contains a quaternion representation of the Maxwell equations. because the vector identity involving the divergence of a curl still applies. Rewrite the equations above in terms of the classical force fields: 4 Π Ρ. What does this do to the 4−vector current density J? Now there is a current that transforms like a pseudo−current density. Faraday’s law is a vector identity. J g E t xB g t . This does not happen however. 115:485−491. The Aharonov−Bohm phenomenon. xE . A t2 A This second order set of four partial differential equations has four unknowns so this is a complete set of field equations. The volume integral of this pseudo−current density is the total magnetic flux: k B dV t e c B The unified field equation postulates a pseudo 3−vector current composed of the difference between the time derivative of the magnetic field and the curl of the electric field. Bohm. so it is still true implicitly. Therefore. mass. This makes them likely to be consistent with special relativity. The unified field equations contain three of the four Maxwell equations explicitly: Gauss’ law. The pseudo−current density is quite unusual. The volume integral of the time derivative of the magnetic field is a measure of the total magnetic flux.

it may be something that is required for a final and complete theory. other than a constant. this may seem like a deadly flaw. Ampere’s law looks different. Solutions to the Unified Field Equations All the solutions that have been worked out for the Maxwell equations will work with the unified field equations. To be consistent with Ampere’s law. so they may not be viewed as being linearly independent. and visa versa. the cross terms destructively interfere with each other. the no magnetic monopoles vector identity and Faraday’s law. These terms constructively interfere with each other. An elliptic equation combined with a hyperbolic one might more fully describe gravitational and electromagnetic waves from sources.110 J Je JAB iff g 0. the gravitational part may be zero while the electrical part is not. However. B t xE . In a unified field theory. this may turn out to be an advantage. such as plane wave solutions.E E t xB Φ . Given historical tradition. A different set could be created by using the differential operator without taking its conjugate: 4 ΠJ 2 A Φ t2 g t . electric charge. Since the goal of this work is a complete system of field equations. The field equations of general relativity and the Maxwell equations both have vacuum solutions. The elliptic field equation again contains three of four Maxwell equations explicitly:Gauss’ law.B. The unified field equations do not have such a solution. and the Aharanov−Bohm pseudo−current (or total magnetic flux over the volume). 2 A .0 . For example. Examine the scalar field involving the inverse interval squared: 1 t2 x2 y2 z2 .0 4 3 t2 x2 y2 z2 t2 x2 y2 z2 3 . It would be incorrect − but almost true − to say that the unified charge and current are simply the sum of the three: mass. Up to four linearly independent unified field equations can be formulated. This time. if the potential is static. This may be the differential form of a classical Aharonov−Bohm effect. The unified equations are more informative. again a pseudo−current must be included. Unlike the first set of field equations. A t2 2 A Φ g This is an elliptic equation. The only term that does not change between the two field equations is the one involving the dynamic gravitational force. Non−zero solutions are worth exploring An inverse square potential plays an important role in both gravity and electromagnetism. since any potential which is a solution to the scalar Poisson equation will also characterize the corresponding current. the scalar equation for hyperbolic field equation is the Poisson equation. This might be a clue for why this force is only attractive.

so it will depend on the metric. the mass density is conserved.K c 2 0 The dispersion relation is an inverted distance. There is no source term for the Aharanov−Bohm current. This generates two types of constraints on the sources. Instead.0 0 The non−zero part may have everything to do with gravity. The 3−vector equation is a constraint on the mass current density. The same potential can also solve the hyperbolic field equations under different constraints and resulting dispersion equation (not shown). The electromagnetic fields only appear in the 3−vector. and is the reason that electric charge is conserved. There were two reasons for not including the customary imaginary number "i" in the exponential of the potential. A0 eK R Ωt The unified field equation will evaluate to zero if Scalar Ω . This is an indication that unifying gravity and electromagnetism is an appropriate goal. and therefore is outside the domain of my work. No change in the gravitational source applies to the 3−vector. A vacuum solution for the unified field equation is apparent for the elliptical field equations: A Φ0 eK R Ωt . and subsequently no conservation law. Conservation Laws Conservation of electric charge is implicit in the Maxwell equations. For a different inertial observer. The important thing to realize is that vacuum solutions to the unified field equations exist whose dispersion equations depend on the metric. but not for a pure vacuum. it would have created a complex−valued quaternion. Second. E t2 2 2 E B t2 2 B Notice that the gravitational force field only appears in the quaternion scalar. it was not necessary.111 This potential solves the Maxwell equations in the Lorentz gauge: 2 1 t2 x2 y2 z2 . . this will appear as a conservation of electric current density. a plane wave solution exists with the constrain that the net current is zero for the elliptical field equations The field equations of general relativity and the Maxwell equations both have vacuum solutions. Scalar Scalar Vector Je JAB Jm Ρe t JAB Jm t Je 0 Ρm 0 Jm 0 The first equation is known as the continuity equation. For a different inertial observer. A plane wave solution does exist. Is there also a conserved quantity for the gravitational field? Examine how the differential operator acts on the unified field equation: 2 A g t2 2 g. and is the reason mass current density is conserved. No change in the electric source applies to the quaternion scalar. First.

An additional constraint on the gauge field is required to leave the gravitational force field invariant. "Quaternions and special relativity. Three types of gauge transformations will be investigated: a scalar. Transform the potential with an arbitrary 3−vector field: . If so. Most importantly. In the real 4x4 matrix representation. it might provide a test for the theory. For a boost along the x−axis. 1 Φ . but covariance of the electromagnetic fields requires the following residual terms: ’ A’ Residual 0. Only relatively recently has DeLeo been able to represent the Lorentz group using real quaternions (S. since the transformation properties are well known experimentally. The 3−vector equation is a constraint on the mass current density. The meaning of this observation is even less clear. a 3−vector. All the terms required to make the electromagnetic field covariant under a Lorentz transformation are present. The gravitational force field is Lorentz invariant. If a transformation left the quaternion scalar invariant and the 3−vector covariant. The delay appears odd since the interval of special relativity is the scalar of the square of the difference between two events. An operator that adds nothing to the trace but distorts the lengths of the scalar and 3−vector with the constraint that the difference in lengths is constant will also suffice. One such class is 3−dimensional. Consider an arbitrary scalar field transformation of the potential: A A’ A Λ. What might be expected to happen under a Lorentz transformation? Gravity involves mass that is Lorentz invariant.The first equation is known as the continuity equation. the correct interpretation of the residual term is unclear. if only the differential transformation is in the opposite direction. For a different inertial observer." J. spatial rotations. For a different inertial observer. Γ2 Β2 Ay x . These terms could be a velocity−dependent phase factor. and is the reason that electric charge is conserved. and subsequently no conservation law. The effect of unification must be subtle. The mechanics of the Lorentz transformation itself might require careful re−examination when so strictly confined to quaternion algebra. Consider a boost along the x−axis. Therefore. x 2 ΓΒ Az t At this time. The electromagnetic fields are invariant under this transformation. then the electromagnetic field is Lorentz covariant with the residual term residing with the gravitational field. However. any matrix with a trace of one that does not distort the length of the scalar and 3−vector can multiply a quaternion without effecting the interval. Phys. the interval is a quarter of the trace of the square. These are boosts in an inertial reference frame. this will appear as a conservation of electric current density. 1996). De Leo. a transformation cannot do both perfectly. and is the reason mass current density is conserved. There is no source term for the Aharanov−Bohm current. From the perspective of this proposal. and a quaternion gauge field. so the field that generates it should be Lorentz invariant. Math. Transformations of the Unified Force Field 112 The transformation properties of the unified field promise to be more intricate than either gravity or electromagnetism separately. Boosts plus rotations form the Lorentz group. The electromagnetic field is Lorentz covariant. namely that the scalar gauge field solves a homogeneous elliptical equation. The reason is that a Lorentz transformation mixes a quaternion scalar with a 3−vector. the mass density is conserved. 37(6):2955−2968. it was shown earlier that charge is conserved.. 2 ΓΒ 1 Ay t Ax t Az x Γ2 . the two would effectively not mix. the freedom to choose a scalar gauge field for the Maxwell equations is due to the omission of the gravitational force field.

a logical structure can be constructed. For example. The terms that do not change under differentiation of a field equation form conservation laws. The second hypothesis concerned a unified field equation formed by acting on the unified field with one more differential operator. A set of events forms a pattern that can be described by a potential. The first hypothesis about the unified field involved only a quaternion differential operator acting on a potential. no extra terms added by hand. If a force field is created by hitting this gauge transformation with a differential operator. Since vacuum solutions have been found for those equations. starting from events (see figure below). All the Maxwell equations are included explicitly or implicitly. Why did this approach work? The hypothesis that initiated this line of research was that all events in spacetime could be represented by quaternions. but only the hyperbolic and elliptic cases were discussed. The change in field creates a field equation. Future Directions The fields of gravity and electromagnetism were unified in a way consistent with Einstein’s vision. Other classes of invariants could be examined. Four linearly independent unified field equations exist. A large family of vacuum solutions exists. The guiding principles were simple but unusual: generate expressions familiar from electromagnetism using quaternions.113 A A’ A . no matter how the events were generated. Based on the equations presented in this paper. then the 3−vector gauge field must be the solution to an elliptical equation. Is there any such gauge field? The quaternion gauge field can be represented the following way: A A’ A . This gauge transformation would have the same constraints as those above to leave the fields invariant. if the difference between the two electromagnetic fields is to remain invariant. Additional constraints can be placed on the 3−vector gauge field to preserve a chosen electromagnetic invariant. The change in a potential creates a field. It contained the typical potential representation of the electromagnetic field. and will require future analysis to appreciate. . along with a symmetric−matrix force field for gravity. then the gauge field becomes a unified field equation. This is a broad hypothesis. The scalar and 3−vector gauge fields could be combined to form a quaternion gauge field. not his technique. striving to interpret any extra terms as being due to gravity. Additional terms suggested the inclusion of a classical representation of the Aharanov−Bohm effect. attempting to reach all areas covered by physics. This time the gravitational force field is invariant under a 3−vector gauge field transformation. a quaternion gauge transformation can leave the field invariant.

. Guido Sandri.114 Acknowledgements Thanks for many productive discussions Prof.

"Theory of longitudinal photons in quantum electrodynamics". and one longitudinal. the math of gravity and electromagnetism should be similar but not identical. For a weak gravitational field." He notes that this does not always have to be the case for the nonhomogeneous Maxwell equations. so its symmetry is different from the electric 3−vector field and the magnetic 3−pseudo−vector field. The hypothesis makes several predictions even at this preliminary stage. so this proposal is distinct from general relativity. instead of transverse. so there will be no simple relationship between gravitational charge (mass) and electric charge. which makes the proposed metric distinct and testable experimentally. created by the product of charge. which is the focus of this work. electromagnetism is not. The starting point is the 4−potential A^mu. Phys. The inverse square form of Newton’s law of gravity was a direct inspiration for Coulomb’s law. Gravity is a metric theory. but still has the field equations of Maxwell with the choice of the Lorenz gauge. Gupta calculated that "the probability of the emission of a real longitudinal photon is canceled by the ’negative probability’ of the emission of a corresponding scalar photon. The second rank field strength tensor in general relativity is symmetric while the analogous tensor for the electromagnetic field is antisymmetric. Whatever phenomenon exploits the . and S. The probabilities of the last two modes cancel each other out for a spin 1 field. photons have four modes of transmission. but that does not have to be the case for a field with spin 1 and spin 2 particles. 23:567. By using Riemannian quaternions which can have dynamic basis vectors. Nature exploits all the math available. Since the mode of gravity is orthogonal to electromagnetism. Acta. The solution to the force equation using the inverse square interval potential is found. and thus does not have an obvious role to play in electrodynamics. Bleuler.. 2 2 An Opportunity for Classical Gravity? The electrodynamic field can be quantized in a manifestly covariant form by fixing the gauge (K. One potential solution to the field equations is found which depend on the inverse of an interval between two events squared. The dark matter hypothesis is not needed to explain the constant−velocity profiles seen for some galaxies. so it is unreasonable to suppose that scalar and longitudinal photons are never used for anything. Eliminating the constants generates a metric equation. A relativistic force equation is proposed. at least mathematically: two transverse modes for electrodynamics. Gravitational waves in general relativity are transverse. A scalar photon would not change signs under a space or time reversal. The proposal may also have implications for classical big bang theory. normalized force field. The force field created by the potential is constructed by comparison with the classical Newtonian gravitational field. the charges can be likewise. Phys. Proc. 63:681−691). where tau is a lightlike interval with almost the same magnitude as the radius R of separation between source and test masses. N. Gravity should be more symmetric than electromagnetism because the mode is scalar. and 4−velocity: dmU/dtau = kq Box* A/|A| U*. A constant−velocity solution exists for the gravitational force equation for a system with an exponentially−decaying mass distribution. a longitudinal. and a scalar mode. Helv. it becomes possible to merge metric theory with the linear Maxwell equations. First. My hypothesis is that the scalar and longitudinal photons for the electromagnetic field constitute gravity. The Lagrange density L= scalar(−(J A*) − 1/2 Box* A Box A*) can contribute to a scalar mode.115 Einstein’s Vision II: A Unified Force Equation with Constant Velocity Solutions Abstract In quantum electrodynamics. one scalar. Gupta. There are four modes of transmission for photons corresponding to the four degrees of freedom: two transverse. Soc. GM GM dΤ2 e 2 c Τ e2 c Τ d R2 c2 . the metric will pass the same tests as the Schwarzschild metric of general relativity. The two metrics differ for higher order terms. 1950.

My hypothesis is that the scalar and longitudinal photons for the electromagnetic field constitute gravity. Classical gravity depends on an inverse square force field. The inverse square form of Newton’s law of gravity was a direct inspiration for Coulomb’s law. so this proposal is distinct from general relativity. The interval between any two events will contribute to the potential. Whatever phenomenon exploits the scalar and longitudinal photons must be similar. A potential that embraces every interval may have a broad enough scope to do the work of gravity. the mass current density. and will be adopted here. Since the mode of gravity is orthogonal to electromagnetism. so there will be no simple relationship between gravitational charge (mass) and electric charge. including gravitational field energy. Any change in the mass density propagates at infinite speed. Gravity is a natural candidate. Like mass. For a spherically symmetric source. in conflict with special relativity (MTW. instead of transverse. The hypothesis makes several predictions even at this preliminary stage. General relativity applies to any form of energy. the sources can be also. If one is working with transverse modes. Here is the scalar field equation: 2 Φ 4 ΠGΡ For the case of a vacuum. z^2. this is known as the Laplace equation.0 Τ2 This potential is interesting for several reasons. The potential also has serious problems. mass: 2 A 4 Π Jq Jm If one is studying scalar or longitudinal modes.0 1 . not an inverse square potential. the potential must have x^2. Gravity should be more symmetric than electromagnetism because the mode is scalar. That expression is identical to the A^0 component of the Maxwell equations with the choice of the Lorenz gauge. A way to repair the field equations is to use the D’Alembertian operator. First. Nature exploits all the math available. but just as important as electromagnetism. the charges can be likewise. the source is J_charge^mu . the electric charge density. Since the modes are orthogonal. The second rank field strength tensor in general relativity is symmetric while the analogous tensor for the electromagnetic field is antisymmetric. the source is J_mass. The sources are of course different. Taking the derivative of the potential puts a forth power of the interval in the denominator. It is the inverse of the Lorentz−invariant interval squared. Gravitational waves in general relativity are transverse. An alternative is to look for an algebraic way to repair the problem. we can have a different source term. so it is unreasonable to suppose that scalar and longitudinal photons are never used for anything. For the two degrees of freedom. One way to derive the field equations of general relativity involves making Newton’s law of gravity consistent with the finite speed of light. The equations of motion can be normalized to the magnitude of the 4−potential: . y^2. the math of116 gravity and electromagnetism should be similar but not identical. and t^2. I could stop and say that this potential has nothing to do with gravity because it has the wrong dependence on distance. To be consistent with the classic scalar potential yet still be relativistic. which is four dimensional. chapter 7). A Gravitational Field Inside Maxwell Newton’s classical gravitational law arises from a scalar potential. Yet the argument being made here is that there are degrees of freedom which have yet to be exploited. one solution is: Φ x2 GM y2 z2 The problem with the field equation is that the Laplace operator does not have a time differential operator. This is the type of approach used by the early workers in quantum mechanics like de Broglie. the 4−potential will not be altered by a change in an inertial reference frame. This suggests a particular solution to the field equations: A 1 c2 t2 x2 y2 z2 . when rho = 0. At this point.

This requires both a metric and a connection. Action−at−a−distance respects the speed of light as it must. there is no need worry about it. all the work presented with quaternions uses the Minkowski metric with Cartesian coordinates. the force field is the derivative of the potential. The change of variables is valid locally. The field is the derivative of the potential. is approximately: 1 1 Τ2 1 Τ2 t c2 GM Τ2 . The equations of motion can be normalized to the magnitude of the 4−potential: 2 A A 4 Π kJcharge GJmass Since the magnitude of the potential is the inverse interval squared. and will be adopted here. is 1. In effect. Search for the Source Mass Where is the source mass in the potential? All that has been discussed so far is an interval. GM_Sun/c^2. I can impose a selection rule that in the classical limit. At this point. Taking the derivative of the potential puts a forth power of the interval in the denominator. and any change in position in spacetime will have a far greater effect proportionally on the smaller geometric mass than the spatial separation. a distance. a normalized interval squared with the linear displacement substitution. To be correct technically. The overall length of the interval will not be changed noticeably if the spatial separation and the Sun’s mass expressed as a distance are summed. since it breaks down for arbitrarily long time or distance away. However.5x10^3 m. it is the contravariant derivative. nothing about mass. Make the following change of variables: t R t’ R’ A B GM t 2 c2 A GM 2 c2 B R where A and B are locally constants such that tau^2 ~= A^2 − B^2. there is a linear displacement equation inside an inverse square potential. However. since the equations are quite simple and their mathematical consequences can be worked out. What is the physical interpretation of the inverse square potential and the above substitution? Newton observed that motion in an ellipse could be caused by either a linear central force or an inverse square law. Although it is confusing to confront the idea of oscillations in time. This is the type of approach used by the early workers in quantum mechanics like de Broglie. the equation would be nonlinear. An idea from general relativity will be borrowed. An alternative is to look for an algebraic way to repair the problem. I could stop and say that this potential has nothing to do117 with gravity because it has the wrong dependence on distance. while the Sun’s mass expressed in units of distance. With the above substitution. that mass can be treated geometrically if multiplied by the constants G/c^2. It is like a simple harmonic oscillator inside a simple harmonic oscillator! This oscillator works with four dimensions.The potential also has serious problems. An interval is not necessarily the same as the distance R between the source and test mass used in the classical theory. eight orders of magnitude smaller. Classical gravity depends on an inverse square force field. the only events that contribute to the potential are those that are timelike separated between the source and the test masses. the contravariant derivative equals the normal derivative. the resulting equation has only an interval squared in the denominator. not an inverse square potential. It takes a timelike interval to know that the source is a distance R away. For such a choice of metric and coordinates. If all the terms where included.5x10^11 m. The distance between the Earth and the Sun is approximately 1. The derivative of the potential under study. but not globally. General relativity is also valid locally and not globally.

It is the first of four terms generated in the contraction of the electromagnetic field strength tensor. which is what happens in making the field strength tensor antisymmetric. He fixed the gauge with this Lagrange density: L JΜ AΜ 1 2 Μ AΜ 2 1 4 Μ AΝ Ν AΜ Μ AΝ Ν AΜ The equations of motion for this Lagrangian are the same as choosing the Lorenz gauge: 2 AΜ JΜ The problem with the Lagrangian is that the field strength tensor is antisymmetric. assume the charge q is the gravitational test mass. The covariant force law is similar in form to the one for electromagnetism except that the second rank tensor is asymmetric and normalized: F p Τ mc Β Τ Βc m Τ kq A A Β If this equation is to transform like the Lorentz 4−force of electromagnetism. Experiments have demonstrated that gravitational and inertial masses are equal. Assuming spherical symmetry. remembering that employing the event selection rule from above. What is needed is a Lagrange density that could contribute directly to the scalar mode but still have the same field equations. it cannot contribute directly to a scalar mode. The one remaining modification is to normalize both the Lagrangian and equations of motion to the size of the potential. For the scalar photons. Here is such a Lagrangian: L scalar Jq Jm A 1 2 A A This is not as miraculous as it might first appear. the inverse interval squared potential leads to the following equations of motion: 2 t Τ2 GM c2 Τ2 t . A Lagrangian for Four Modes Despite its formulation using quaternions. Gupta wanted to quantize the radiation field using a form that was manifestly covariant in its explicit treatment of time and space. the normalized potential must be invariant under a Lorentz transformation. this unification proposal is strikingly similar to earlier work. differing only by the geometric mass of the source.118 1 1 Τ2 1 Τ2 R c2 GM Τ2 This should look familiar. assume the derivative of the mass with respect to the interval is zero. Τ 2 R Τ2 GM c2 Τ2 R Τ 0. From a Relativistic 4−force to a Metric A relativistic 4−force is the change in momentum with respect to the interval. the magnitude of tau ^2 is almost the same as R^2. information is not discarded. That is the case of the potential under study. In essence. Due to the zeros along the diagonal. In the first application of the force law. 0 Solve these second−order differential equations for the spacetime position: .

More modern efforts have shown that the reason for the connection is due to the conformal mapping of z goes to z^2 (T. so its presence is interesting. this become the Minkowski metric for flat spacetime if M goes to 0 or tau goes to infinity. spacetime is curved As expected. Sir Isaac Newton in the Principia showed an important link between forces linear in position and inverse square force laws. C_1) and (c_2. leading to e^(GM/c^2|tau |) goes to 1: t Τ 2 R Τ GM c2 Τ R Τ t Τ c1 2 C1 C1 1 Solve for c_1^2 and C_1. Mon.. For a weak field. Spacetime is flat if M goes to 0 or tau goes to infinity. Conversely. (c_2.119 t R c1 Τe c C1 Τe GM 2 Τ GM Ei c2 GM Ei c2 GM c2 Τ GM c2 Τ C2 C2 GM c2 Τ where Ei is the exponential integral. C_2). write the Taylor series expansion in terms of the total mass over the interval to second−order in M/|tau|: . 100:119−137. Needham. "Newton and the transmutation of force. if the gravitational field is non−zero. The result is a 4−velocity: t Τ R Τ c1 e c C1 e 2 GM Τ GM c2 Τ In flat spacetime. 1993). Instead a path between a proposed gravitational force equation and a metric function was sketched. This eliminates four constants. Take the derivative of the spacetime position with respect to tau. beta _mu beta^mu=1. Math. Rearrange into a metric: Τ2 e 2 GM c3 Τ R 2 If the gravitational field is zero. There is a historical precedence for the line of logic followed. This method was adapted to a quaternion force law linear in the relativistic velocity to generate a metric. Eight constants need to be eliminated:(c_1.C_1: c1 2 C1 e C1 ec 2 GM Τ R Τ t2 e2 GM c3 Τ Substitute back into the flat spacetime constraint. No formal connection between this proposal and curvature has been established." Amer. The exponential integral plays other roles in quantum mechanics. Ei(t)=the integral from negative infinity to t of e^t/t dt. providing four more constraints. this generates the Minkowski metric of flat spacetime. C_2).

gravity’s effect is on the distribution of mass over spacetime. Here the opposite situation is examined. such a test will challenge experimental techniques. Start with the gravitational force in a region of spacetime with no velocity change: Βc mi Τ k mg Scalar A Β Make the same assumptions as before: the gravitational mass is equal to the inertial mass and the gravitational field employs the interval between the worldlines of the test and gravitational masses. For example. which does not use the second order spatial term. It is more in keeping with the principles underlying relativity. but the unified metric contains a dependence on time so is dynamic. This new gravitational term is not due to the unified field proposal per se. since light has no rest mass. the difference being the geometric mass of the source included in the interval tau . eq. In that region. This generates an equation for the distribution of mass: . The Constant Velocity Profile Solution In the previous section. 31. the Schwarzschild metric is static. and by extension. This might pose less of a conceptual problem.5 2 GM c2 R 2 2 t2 R 2 GM c2 R GM c2 R O GM c2 R 3 The magnitude of the lightlike interval tau in the unified field metric is nearly identical to the radius R in the Schwarzschild metric. but mathematically distinct. the system had a constant point−source mass with a velocity profile that decayed with distance. The unified gravitational force metric becomes undefined for lightlike intervals. in this case mass distribution with respect to spacetime. where the velocity profile is a constant. Expand the definition of the relativistic force using the chain rule: c mΒ Τ mc Β Τ Βc m Τ The first term of the force is the one that leads to an approximation of the Schwarzschild metric. The difference in the higher order terms can be the basis of an experimental test to distinguish this proposal from general relativity. For a region of spacetime where the velocity is constant. The Schwarzschild metric has a singularity at R=0. Since the effect is second order in the field strength. this term is zero. Newton’s law of gravity. but the mass distribution decays with distance.22): Τ2 1 2 1 2 GM c2 R 2. The metric for the scalar potential will pass the same weak field tests of general relativity as the Schwarzschild metric to post−Newtonian accuracy. looking for changes in all components.120 Τ2 1 2 1 2 GM c2 Τ 2 2 GM c2 Τ GM c2 Τ 2 2 t2 O GM c2 Τ 3 GM c2 Τ R 2 Contrast this with the Schwarzschild solution in isotropic coordinates expanded to second order in M/R (MTW. The two metrics are numerically very similar for weak fields.

That remains a valid choice today. cannot be explained by the Newtonian term (A. Ce As in the previous example for a classical weak field. A. 1990). Γ Β m cec 2 GM Τ . A. gravity was Riemannian geometry.121 Γ m Τ c2 ΓGM Τ 2 m. curvature of spacetime caused by the presence of a mass−energy density. Gauss’ law written with Riemannian quaternion potentials and operators leads to this expression: 2 in 9 En in in En 9 in in 4 ΠΡ. so it is the mass distribution that shows an exponential decay with respect to the interval. G. The exponential decay of the mass of a disk galaxy is only one solution to this expanded gravitational force equation. F. It will remain to be seen if this proposal is sufficient to work on that scale. (a_0 i_0. J. Γ Β m Τ GM c2 Τ ΓΒGM m c2 Τ 2 0. Let. and R. If the mass keeps dropping of exponentially. has pairs of (possibly) dynamic terms for the 4−potential A and the 4−basis I. A.. However. Valdes. Petrosian. assume the magnitude of the interval is an excellent approximation to the radius divided by the speed of light. electromagnetism cannot be depicted in purely geometric terms. Narayan. Wenk. Lynds. Albert Einstein argued that gravity was not a force at all." Astrophys. Electromagnetism was first described as a force. 1989. In general relativity. 2. "Detection of systematic gravitational lens galaxy image alignments: Mapping dark matter in galaxy clusters. n 1. 349:L1. V. such as gravitational lensing caused by clusters. 350:23. The general equivalence principle." Astrophys. "Gravitationally lensed images in abell 370. Any law of electrodynamics written with Riemannian quaternions is a combination of changes in potentials and/or basis vectors. a_3 i_3/3). J. Tyson. and R. S. 3 If the divergence of the electric field E was zero. Bergmann. Metrics and Forces Gravity was first described as a force by Isaac Newton. The velocity is a constant. then Gauss’ law would be due entirely to the divergence of the basis vectors. introduced in the first paper of this series. 0 Solve for the mass flow: Γ m. J. J. The distribution of matter has an exponential decay with distance from the center. a_2 i_2/3. It must solve a differential equation with the velocity constant over that region of spacetime like the one proposed. .." Astrophys. The behavior of larger systems. 344−637−644. the velocity profile will remain constant Look at the problem in reverse. The reverse case could also hold. Rather.. This is a stable solution. places geometry and force potentials on equal footing. Riemannian quaternions. a_1 i_1/3. Grossman and R. "Gravitational lens images of arcs in clusters. 1990. modeled on gravity. A conceptual gap exists between purely geometrical and force laws. which is numerically no different from the radius over the speed of light.

a division algebra. Gravity is all about oscillations. If there were no interfering matter in the way.122 Future Directions An algebraic path between a solution to the Maxwell equations and a classical metric gravitational theory has been shown. For a spiral galaxy with an exponential mass distribution. For higher order terms of a weak field. Considerable effort will be required to substantiate this tenuous hypothesis. and it is this requirement that fixes the gauge. the foundation for doing calculus. Nature does calculus in four dimensions. That makes the proposal reasonable. we may have a deep appreciation of Nature’s methods. the proposal is different than the Schwarzschild metric of general relativity. p. when we understand how to do calculus with four−dimensional automorphic functions. This raises an interesting question: is there also an effect of mass distributed over large amounts of time? If the answer is yes. and return in a little over eighty minutes. By using quaternions. Classical big bang cosmology theory spans the largest time frame possible and faces two such issues. There is a physical explanation for gravity − it is a local. In the future. The Earth returns to approximately the same place after one year of travel. then Nature cannot harness the most robust mathematical structure. flat velocity profiles and the stability of solutions. Mass distributed over large distances of space has an effect on the mass distribution itself. but the additional constraint decreases the degrees of freedom. Any insight into the origin of the unified engine driving the Universe of gravity and light is worthwhile. The flatness problem indicates how unstable the classical big bang theory is. Like the early work in quantum mechanics. There are very few reasonable. so this alone should spark interest in this line of work. a topological algebraic field. requiring exceptional fine tuning to avoid collapse. No effort has been extended yet to quantize the unification proposal. 815). One is left with the question of why this might work? The action of a gauge invariant theory cannot be inverted to generate the propagator needed for quantum mechanics. The metric equation that results from this analysis is within the experimental constraints of current tests of general relativity. a collection of hunches is used to connect equations. The horizon problem involves the extremely consistent velocity profile across parts of the Universe that are not casually linked (MTW. nonlinear. reach the other side. an apple dropped would fall to the center of the Earth. four−dimensional simple harmonic oscillator. If the operation of multiplication surpasses what can be done with division. That makes it testable. Fixing the gauge makes the action invertible. dark matter is no longer needed to explain the flat velocity profile observed or the long term stability of the disk. then this might solve two analogous riddles involving large time scales. the equation is necessarily invertible without imposing a constraint. . testable classical unified field theories in physics.

a1 . a3 Taking the difference between two quaternions is only a valid operation if they share the same basis. A quaternion has 4 degrees of freedom. the same as intrinsic curvature. Strings Let us revisit the difference between two quaternions squared. and e3 share in order to make subtraction a valid operation? There is only one basis. If this is not the case. so it can be represented by 4 real numbers: q a0 . and if lucky. 2 da0 da3 e0 3 3 3 . 2 da0 da2 e0 3 e2 2 e 2 da3 2 3 . and k. as worked out in the section of analysis.123 Strings and Quantum Gravity In this section. I am trying to merge the basis vectors of quaternions with tools from topology. j. i. at least one consistent with current experimental tests. da1 e1 e e . There would be a hidden bump in an otherwise smooth transformation! At this point. Work with defining the derivative with respect to a quaternion has required that a change in the scalar be equal in magnitude to the sum of changes in the 3−vector (instead of the usual parity with components). At the current time. This is a flat space. I am free to define things as I choose. These concerns lead to the definition of the difference between two quaternions: dq da0 e0 . a2 . da3 3 3 3 3 What type of information must e0. this is consistent with Hamilton’s system for 1. If not. e1. this is an idea in progress. so the two events that make up the difference must necessarily be expressed in the same basis. so −1/e1^2 is something like 1 + k. I will propose the following relationship between basis vectors because its form suggests a link to intrinsic curvature: 1 e1 2 1 e2 2 1 e3 2 e0 2 If e0 = 1. it will prove useful later on :−) Form the square of the difference between two quaternion events as defined above: dq2 da0 2 e0 2 2 da0 da1 e0 da1 2 e1 2 9 da2 2 e1 . I will argue that it my be involved in a relativistic quantum gravity theory. not a theory. since the equations of motion have not been determined. 9 9 e2 e . In effect. A more subtle issue is that the difference must have the same amount of intrinsic curvature for all three spatial basis vectors. By making this quantity dimensionless. I do not yet understand the technical link between basis vectors and intrinsic curvature. a quaternion 3−string will be defined. e2. then the standard coordinate transformation needs to be done first. The dimensions for the spatial part are 1/distance^2. da2 2 . It is hoped that the work in the previous section on unified fields will provide that someday. In math. then it would not longer be possible to do a coordinate transformation using the typical methods.

the differences would have to first be made dimensionless. dX = 0. the former flipping signs with time. and the difference of a quaternion blows up or disappears. Use a combination of 3 constants to do the work of making this dimensionless. Take this constants to zero or infinity. Let a0 have units of time. However. the results had a periodic pattern). which is the ratio of dX by dt. 2 da0 da2 e0 3c e3 2 . The symbols needed to make the difference between two events dimensionless are simple: dq2 c5 Gh da0 2 e0 2 da1 2 e1 2 9 c2 da2 2 2 da0 da1 e0 e1 . since the infinite series expansion for such transcendental functions would not make sense. 2 da0 da2 e0 3c e3 2 . 9 c2 e2 e . . the value of a string is We all appreciate the critical role played by the 3−velocity. For a string between two events that have the same spatial location. so the string dt dX is zero. A string will also transform differently under a Lorentz transformation. and a1. a2. the latter inert. It could happen :−) One might be able to use a collection of sines and cosines to regenerate the pattern. a3 have units of space. Dimensionless Strings Imagine some system that happens to create a periodic pattern of intervals and strings (a series of events that when you took the difference between neighboring events and squared them. Hopefully we can imagine another role as important for the product of these same two numbers. For a string between two events that are simultaneous. Why use a work with a powerful meaning in the current physics lexicon for the vector dt dX? A string transforms differently than a spatial 3−vector. 3 string The scalar is the Lorentz invariant interval of special relativity if e0 = 1. Only if two events happen at different times in different locations will the string be non−zero. The first step is to get all the units to be the same.124 interval2 . This is the same as the reciprocal of the Planck time squared.5x10^85s^−2. Since a string is not invariant under a Lorentz transformation. and in units of seconds is 5. Make all components have units of time: dq2 da0 2 e0 2 da1 2 e1 2 9 c2 da2 2 2 da0 da1 e0 e1 . 2 da0 da3 e0 3 3c 3c e2 2 9 c2 da3 2 Now the units are time squared. this is relativistic (c) quantum (h) gravity (G). dt = 0 so the string is again of zero length. The units for a string are time*distance. 2 da0 da3 e0 3 3c 3c e2 2 9 c2 da3 2 As far as the units are concerned. using c. 1 G mass time2 distance3 1 h time mass distance2 c5 distance5 time5 The units for the product of these three numbers are the reciprocal of time squared. since sines and cosines can do that sort of work. 9 c2 e2 e .

but there is no handle on the overall sign of the 4−dimensional quaternion. There are four cases for what can happen to the interval and the string between these two events under this general transformation. Case 1: Constant Intervals and Strings T : dq dq ’ such that scalar dq2 scalar dq ’2 and vector dq2 vector dq ’2 This looks simple. because it is the product of e0 e1. Intervals change under such a transformation. This issue of ambiguous phases is true for all four of these cases. or special relativity. All that is required to make this simple but radical proposal consistent with experimental tests of general relativity is the following: 1 2 GM c2 R 1 e1 2 1 e2 2 1 e3 2 e0 2 The string. a smoke signal of O(4). . and this can be used as a measure of the amount of change between non−inertial reference frames. Strings change under such a transformation. Quantum mechanics is constructed around dealing with phase ambiguity in a rigorous way. and e0 e3. The interval depends directly on the squares of the basis vectors (I think of this as being 1+/− the intrinsic curvature. or general relativity. since this involves the root of the squared basis vectors. This particular value regenerates the Schwarzschild solution of general relativity.125 Behaving Like a Relativistic Quantum Gravity Theory Although the units suggest a possible relativistic quantum gravity. e0 e2. Case 3: Constant Strings T : dq dq ’ such that scalar dq2 scalar dq ’2 and vector dq2 vector dq ’2 Case 3 involves conserving the quaternion string. Case 2: Constant Intervals T : dq dq ’ such that scalar dq2 scalar dq ’2 and vector dq2 vector dq ’2 Case 2 involves conserving the Lorentz invariant interval. will not be changed by this. it is more important to see that it behaves like one. Since this unicorn of physics has never been seen I will present 4 cases which will show that this equation behaves like that mysterious beast! Consider a general transformation T that brings the difference between two events dq into dq’. The phase of the string may change here. but do not know if that is an accurate technical assessment). and this can be used as a measure of the amount of change between inertial reference frames.

This will involve determining the precise relationship between intrinsic curvature and the quaternion basis vectors. The change in the length−time of the string will involve the inertial reference frame. and the change in the interval will involve changes in the mass density. changes in the reference frame of an inertial observer are logically independent from changing the mass density. The Missing Link At this time I do not know how to use the proposed unified field equations discussed earlier to generate the basis vectors shown.126 Case 4: No Constants T : dq dq ’ such that scalar dq2 scalar dq ’2 and vector dq2 vector dq ’2 In this proposal. . The two effects can be measured separately.

circa 1993. The classical distance between them is the interval. This is solid a description of the family of approaches being used.128 What Are Prima Facie Questions? The first question raised by Isham is the relation between classical and quantum physics. q and q’. To be complete. Physics with quaternions has a general guide. Quaternions can be used to form an inner−product space. and it certainly could be computed with standard tools." .what we call a ’fact’ does not exist without some theoretical schema for organizing experimental and experiential data. which causes no harm−−indeed the selection of such a framework is an essential pre−requisite for theoretical research−−provided the choice is made consciously. He writes: ". 0 t." My structure is this: the description of events in spacetime using the topological algebraic field of quaternions is physics. See the text for details. The norm of the difference between q and q’ is t. X t. but the tools used to build them morph and become unified under one algebraic umbrella. conversely. Current Research Programs in Quantum Gravity There is a list of current approaches to quantum gravity. all these concepts are required.d X. Different approaches to quantum gravity differ significantly in the frameworks they adopt. X 2 dt2 d X. Isham points out the difficulty of clearly marking a boundary between theories and fact. t. Consider two arbitrary quaternions. There is nothing inherently wrong with this vector. because the distance also includes a 3−vector. In general.d X. X t. in constructing a theory we inevitably impose some prior idea of what we mean by a fact. The uncertainty principle can be derived in the same way as is done with the complex−valued wave function.. Which classical spacetime concepts are needed? Which standard parts of quantum mechanics are needed? Should particles be united? With quaternions. He writes: "I mean (by background structure) the entire conceptual and structural framework within whose language any particular approach is couched. classical physics involves Grassman products and quantum mechanics involves Euclidean products of quaternions. X d X. measure the difference between two quaternions with a quaternion containing the usual invariant scalar interval and a covariant 3−vector. use the 3−vector which can be unique.. including the Schwarz and triangle inequalities. X dt2 t. To distinguishing collections of events that are lightlike separated where the interval is zero. I call q q’ a Grassman product (it has the cross product in it) and q* q’ the Euclidean product (it is a Euclidean norm if q = q’). 2 dt d X This involves retooling. X The norm can be used to build all the equipment expected of a Hilbert space. Never discard useful information! Quantum mechanics involves a Hilbert space. and. Isham moves from big questions to ones focused on quantum gravity. Prima Facie Questions in Quantum Gravity Isham is concerned with the form of these approaches.

spacetime structure is built up with care from four unrelated real numbers. Specifically. Some of the most important pieces of substructure underlying this picture are illustrated in Figure 1. which causes no harm−−indeed the selection of such a framework is an essential pre−requisite for theoretical research−−provided the choice is made consciously. Different approaches to quantum gravity differ significantly in the frameworks they adopt. since this is helpful for anyone trying to get a handle on the nature of general relativity. for example. Quaternions have the same continuum properties as the real and complex numbers. A point can then be labeled uniquely in M (at least locally) by giving the values of four real numbers. causal relations between pairs of points etc. spacetime structure is the observed properties of the mathematics. even the set M somehow."I mean (by background structure) the entire conceptual and structural framework within whose language any particular approach is couched. Near the end of this section I will do that in an attempt to find the shortest norm in spacetime which happens to be non−linear. General relativity may force non−linearity into quantum theory. With quaternions as events. the idea of causal structure is more primitive than that of a Lorentzian metric. This property is shared by quantum mechanics so it should not banish quaternions from the list." My hypothesis to treat events as quaternions lends more structure than is found in the set M. that quaternions do not belong on a list that includes real and complex numbers. and limit processes required for a differentiable manifold make sense. I will quote at length. This set is formless with its only general mathematical property being the cardinal number. as signified in Figure 1. to the manifold to the topology. there are no relations between the elements of M and no special way of labeling any such element. so the notion of length of a path joining two events is always there. In the final step a Lorentzian metric g is placed on M. Pontryagin proved that quaternions are a topological algebraic field. The omission reflects the history of work in the field. Moments later comes this comment: "In using real or complex numbers in quantum theory we are arguably making a prior assumption about the continuum nature of space. It now becomes possible to talk about relationships between point. The properties of quaternions dovetail the needs perfectly. The most common example involves how quantum fluctuations might effect the Lorentzian metric. The bottom level is a set M whose elements are to be identified with spacetime ’points’ or ’events’. The important distinction is that quaternions do not commute. but it literally does not appear on the radar screen of this discussion of quantum gravity." My framework was stated explicitly above. Earlier in this section I showed how to calculate the Lorentz interval. 129 "The mathematical model of spacetime used in classical general relativity is a differentiable manifold equipped with a Lorentzian metric. It is easy to write non−linear quaternion functions. Label every quaternion event with four real numbers. Each point has a neighborhood. This defect is overcome by adding the key of all standard views of spacetime: the topology of M must be compatible with that of a differentiable manifold. thereby introducing the ideas of the length of a path joining two spacetime points. Physicists have tried to investigate how such fluctuation would effect every level of spacetime structure. albeit in a rather non−physical way. In particular. Much work in quantum gravity has gone into viewing how flexible the spacetime structure might be. which require a change in the formalism. albeit not intrinsically in so far as these depend on which coordinate systems are chosen to cover M. from causality. inherited by all quaternion functions. not the logic of the mathematical statement." This statement makes a hidden assumption. Now we come to the part of the paper that got me really excited! Isham described all the machinery needed for classical general relativity. using whichever coordinate system one chooses. As described by Isham. Such a coordinate system also provides a more specific way of describing relationships between points of M. There are also a variety of possible intermediate steps between the manifold and Lorentzian pictures. The next step is to impose a topology on M so that each point acquires a family of neighborhoods. . parallel transport with respect to a Riemannian connection.

by the interval of classical physics (t^2 − x^2 − y^2 − z^2. so it is possible. Every quaternion equation inherits this wealth of spacetime structure. intervals. With quaternions. only in some coarse−grained. the tangent of spacetime. but it still is either positive. For two events in flat spacetime. negative or zero. With quaternions. so it forms a totally ordered sub−field of the quaternions.Much work in quantum gravity has gone into viewing how flexible the spacetime structure might be. let q(t) = cos(pi t (2i + 3j + 4k)) and q’(t) = sin( pi t (5i − . semi−classical sense? In the latter case. . locate events within functionals of fields. time plays a central role. Fixing the metric cannot be done if the metric is subject to quantum fluctuations. If a non−linear quaternion function can be defined that is related to the shortest path through spacetime. Depending on the particular value of t.. The most common example involves how quantum fluctuations might effect the Lorentzian metric. from causality. Yet the concept that requires this idea is essential. and is in fact the center of the matrix representation. The Role of the Spacetime Diffeomorphism Group Diff(M) Isham lets me off the hook. that is the interval. to the manifold to the topology. According to Isham. the causal relationship is easy to calculate. The concept of a flexible. Measurements are made at a particular time. the definition is non−physical. then the relationship is timelike. Is it probable? That might depend on the 3−vector. which could be more likely if the vector is small (I don’t understand the details of this suggestion yet). In classical general relativity in curved spacetime. x i. energy. Isham raises three questions: "How is the notion of time to be incorporated in a quantum theory of gravity? Does it play a fundamental role in the construction of the theory or is it a ’phenomenological’ concept that applies.[for type 3 and 4 theories] there is no strong reason to suppose that Diff(M) will play any fundamental role in [such] quantum theory. The scalar slot can be held by energy (E. Could event q have caused q’? Take the difference and square it. then it is spacelike. 2 tz k) or the norm of quantum mechanics (t^2 + x^2 + y^2 + z^2. Time. In quantum. y j. causal structure is particularly important." He is right and wrong. The distance vectors could be anywhere on the map. Calculate the square of the difference between q and q’. The Problem of Time Time plays a different role in quantum theory and in general relativity. is the way to merge classical and quantum descriptions of collections of events. Without a clear definition of measurement. py j. how reliable is the use at a basic level of techniques drawn from standard quantum theory?" Three solutions are noted: fix the background causal structure. 2 tx i. Physicists have tried to investigate how such fluctuation would effect every level of spacetime structure. the cast required for quantum gravity would be complete. This causal structure also applies to quaternion potential functions. If the scalar is negative. causal metric must be preserved. 0. which at an abstract level. z k) that gives time its significance. It is not time per se. px i. time is treated as a background parameter since it is not represented by an operator. There is nothing to stop combining Grassman and Euclidean products. This group is part of the machinery that makes possible causal measurements of lengths in various topologies. but no way to pick the appropriate one. 2 ty j. or make no reference to time. Time is isomorphic to the real numbers. For concreteness. the two have a lightlike relationship. that issue is particularly straightforward. for example. even the set M somehow. pz k). but is complex and not trivial.1j + 2k). Metrics change due to local conditions. but the location of time within the event quaternion (t. this will be positive. It is the family quaternion functions are born in. saying ". My simple tool collection does not include this group.. If the scalar is positive. With quaternions. and one could not have caused the other. 130 None of these avenues are open for quaternion work. If the scalar is zero. causality is always found in the scalar of the square of the difference. 0. In curved spacetime. Even though I don’t know what these particular potential functions represent. 0). negative or zero. there are many possible metrics which might work. the scalar of the square is different.

Plug the extremum back into the first equation. 2 tx i. Time appears in different guises for the scalars of energy. At this point I think I’ll just explain my approach. az 0. px i. by t . N k A A dN dk d2 N dk2 A kB A B A kB k^2 B B k A B Find the extremum of the parameterized norm. A proof can be done using quaternions if the scalar is set to zero.. Events are not totally ordered. y j. ay t . 2 ty j. but intervals are. How does one go about finding the shortest path? In Euclidean 3−space. It is based on a concept from general relativity. Approaches to Quantum Gravity Isham surveys the field. bz t What is the shortest distance between two worldlines A(t) and B(t) in spacetime? The Euclidean 3−space question is a special case of the worldline question. the tangent of spacetime. that involves the triangle inequality. Parameterize the norm N(k) of the sum of A(t) and B(t). pz k).. so it forms a totally ordered sub−field of the quaternions. norms. time plays a central role. az t t. Time is the only element in the scalar of an event. bz t. or gravity. The result is the shortest distance through spacetime. ax. energy. The same proof of the triangle inequality answers both questions. ay. by. according to general relativity. intervals. 0 A B B A 2kB B The extremum is a minimum 2B B 0 The minimum of a quaternion norm is zero. That proof can be repeated with the scalar set free. by the interval of classical physics (t^2 − x^2 − y^2 − z^2. z k) that gives time its significance. The scalar slot can be held by energy (E. and is in fact the center of the matrix representation. The richness of time is in the way it weaves through these other scalars. 0. but energy is. Time is isomorphic to the real numbers. 2 tz k) or the norm of quantum mechanics (t^2 + x^2 + y^2 + z^2. x i. but the location of time within the event quaternion (t. Squares of events are not totally ordered. taking on the properties of a totally ordered set within a larger. It is not time per se.they all can take the same throne isomorphic to the real numbers. py j. bx t . What is the shortest distance between two points A and B in Euclidean 3−space? A B A t B t 0. Energy/momenta are not totally ordered. 0). bx. sharing the center in different ways with space. A painter falling from a ladder travels along the shortest path through spacetime. 0.. but time is. ax t . Time. . Norms are totally ordered and bounded below by zero.131 With quaternions. unordered framework. intervals and norms.

gravity: events are sets of 4 numbers. It will take a lot of work to manage this solution. the distance between events is the interval whose form can vary and a Hilbert space is required for quantum 133 mechanics. The seventh match is the non−linear shortest norm of spacetime. I don’t know when the doctor will show up and make it official. . there is one of the three types of causal relationships between all events. I have no doubt in the diagnosis that the questions in quantum gravity will be answered with quaternions. events have topological neighborhoods. Quaternions are six for six. Nothing here is official. There are many test that must be passed. they live on differential manifolds.

calling the o and o’. Thorne and Wheeler as a multiple of the proper time plus a displacement. . this length will depend on the square of the affine parameter. which will not be invariant under a boost. Lflat q q 2 dt2 d X .134 Length in Curved Spacetime The Affine Parameter of General Relativity The affine parameter is defined in Misner. Spacetime is flat in the sense that the first term is exactly like the Minkowski metric in spacetime. and those quaternions were used to solve problems in special relativity. the length of a quaternion in curved spacetime will be analyzed. The length is symmetric under exchange of the event with the origin translation. plus a cross term. the two origins would be identical. Length in Curved Spacetime Consider if the origin is located at two different locations in spacetime. In this section. Lcurved d t q to 2 o q d X o Xo 2 2 . but the two measures are slightly different. There are quaternions which preserve the interval. 2 dt d X 2 The first term is the square of the interval. Calculate the interval as done above. plus the square of the interval between the two events. it is significant that the value of the vector portion depends upon the observer. Characterize each origin as a quaternion. but account for the change in the origin. In flat spacetime. This gives a way to distinguish between various frequencies of light for example. dt2 dX 2 dto 2 d Xo 2 2 dt dto 2 d X d Xo The length in curved spacetime is the square of the interval (invariant under a boost) between the two origins. 2d t to d X Xo Examine the first term more closely by expanding it. Under certain approximations. Although not important in this context. Length in Flat Spacetime Calculating the square of the interval between two events in flat spacetime was straightforward: take the difference between two quaternions and square it. Λ aΤ b The affine parameter is used to determine length in curved spacetime.

as was the case in special relativity. which could open the door to a quantum approach to curvature. in the small curvature limit. L curved will be different than the square of the affine parameter. It should be possible to solve problems with this "curved" measure. 2d t 2 dX dto 2 d Xo 2 dt dto 2 d X d Xo .The length in curved spacetime is the square of the interval (invariant under a boost) between the two origins. L curved looks similar to the square of the affine parameter: 135 Λ2 b2 2abΤ a2 Τ2 In this case. There exist quaternions which preserve L curved because quaternions are a field (I haven’t found them yet because the math is getting tough at this point!) It is my hope that those quaternions will help solve problems in general relativity. The length is symmetric under exchange of the event with the origin translation. Under this approximation. so tau ~ delta to. As always. The difference will be solely in the nature of the cross term.. the cross term should be covariant. However. . . plus a cross term. it will be fun in the future to think about things like q o to q 2 2 o d X q Xo 2 o q o to d X Xo d t dt2 . plus the square of the interval between the two events. which will not be invariant under a boost. delta to >> delta Xo.. Implications A connection to the curved geometry of general relativity was sketched. For L curved. b^2 is the origin interval squared and a = 1. Whether this has any effects that can be measured needs to be explored. all the objects employed were quaternions. There is a difference in the cross terms. In general relativity. the square of the affine parameter and L curved are the same. b and tau are invariant under a boost. For a strong gravitational field. In particular. Therefore any of the previously outline techniques should be applicable.

Hamilton defined the roles like so: 2 2 i j k 2 1 2 ijk 1 The scalar result of squaring a differential quaternion in the interval of special relativity: scalar dt. Quaternions come with a metric. a means of taking 4 numbers and returning a scalar. quantum mechanics (QM) makes the right predictions to an even high degree of precision. Here is a different approach which generalizes the scalar of the square while being consistent with Hamilton’s rules. and eventually to general relativity. then try to generalize this approach using a different way which might prove compatible with quantum mechanics. When working with a field like quaternions. curved SR GR QM QG This chart suggests that the form of measurement (interval/norm) should be independent of differential geometry (flat/curved). a unique class. that approach bothers me because Hamilton’s rules are fundamental to the very definition of a quaternion. Change these rules and it may not be valid to compare physics done with different metrics. In this section I shall start from the Lorentz invariant interval using quaternions. This lead to his study of Riemannian geometry. It may cause a compatibility problem.136 A New Idea for Metrics In special relativity. Einstein recognized that inertial observes were "special". Measurement interval norm diff. Therefore he set out to understand what was the most general notion for transformations and metrics. For the physics of gravity. For the physics of atoms. Riemannian geometry uses that strategy. the Minkowski metric is used to calculate the interval between two spacetime intervals for inertial observers. flat geo. interval2 scalar g dq g dq . A quantum gravity theory will be used to calculate norms of wave functions in curved space. d X dt2 d X. general relativity (GR) makes the right predictions of all experimental tests conducted to date. special relativity (SR) being adapted to work in flat space. The problem of building a quantum theory of gravity (QG) hides between general relativity and quantum mechanics. General relativity deals with the measurements of intervals in curved spacetime. Quantum mechanics is used to calculate the norms of wave functions in a flat linear space. That will be the explicit goal of this section.d X How can this be generalized? It might seem natural to explore variations on Hamilton’s rules shown above.

Here I will have to stop. What if g = i? (what would you guess. y. dt2 d X. or dt dz. 0 . 2 scalar d X. 0 t. z 2 Now the special direction x plays the same role as time! Does this make sense physically? Here is one interpretation. G dt. In time. Then then result is the flat Minkowski interval. Since it is static. These constraints can both be achieved if Gx = Gy = Gz = 0. or dy dz. . I should be able to figure out quaternion field equations that do the same work as Einstein’s field equations. This leaves four differential equations. but someone will get to that bridge.d X G xd X . G dt.137 if g 1. d X 4 g dt G. d X g. What sort of calculation to do is a mystery to me. dx dz. When g=i.. where small letters are scalar.d G Gy2 Gy2 Gy2 Gy2 g2 g2 g2 g2 Gx2 Gx2 Gx2 Gx2 Gz2 dt2 Gz2 dx2 Gz2 dy2 Gz2 dz2 4 g Gz dt dz 4 Gy Gz dy dz 4 g Gx dt dx 4 Gx Gy ds dy 4 g Gy dt dy 4 Gx Gz dx dz This has the same combination of ten differential terms found in the Riemannian approach. Since it is spherically symmetric. a norm that is determined by the distribution of mass−energy. We can figure out some of the properties of a static. there will be no terms of the form dx dy. spherically−symmetric metric. The quaternion g could be anything. dt dy. Examine the most general case. I have not yet figured out how to represent the stress tensor.d X In component form. I bet it will contain the Schwarzschild solution too :−) Then it will be easy to create a Hilbert space with a non−Euclidean norm. there will be no terms with the deferential element dt dx.d X then interval2 If g is the identity matrix.. and capital letters are 3−vectors: interval2 g2 dt2 G. so there are no field equations to be solved. 0 x2 y2 z2 . 1.. I was surprised :−) scalar t2 0. G xd X dt2 d G. When g=1.d X g.. x. a space−like interval along the x axis is being measured with a meter stick along the x axis. 0. The difference is that Hamilton’s rule impose an additional structure. a time−like interval is being measured with a wristwatch.

c2 r Now the problem can be viewed as a relativistic Doppler effect problem. 0. the difference between weak gravitational potentials will be calculated and shown to be consistent with experiment. The equivalence principle is used to transform the gravitational potential to a speed (this only involves dividing phi by the constant c^2). and this energy cost is seen as a redshift. A photon lower in a gravitational potential expends energy to climb out. A redshift in a frequency is given by Ν’ Γ Β ΒΓ Β Νo For small velocities. In this section. r h GM . The relevant potentials are Φtower Φbase GM r h GM . h. the first term of a quaternion that is a scalar multiple of the identity matrix. r . Β. 0. The Pound and Rebka Experiment The Pound and Rebka experiment used the Mossbauer effect to measure a redshift between the base and the top of a tower at Harvard University. . Series Βtower O h 2 1 Βbase Νo . Βtower Βbase c2 GM . 1 G M Νo h c2 r2 Or equivalently. 1 The experiment measured the difference between the two Doppler shifts. Quaternions are not of much use here because energy is a scalar.138 The Gravitational Redshift Gravitational redshift experiments are tests of conservation of energy in a gravitational potential. the Doppler effect is Series Γ Β 1 Β O Β 1 2 Β Γ Β .

Alley conducted an experiment which involved flying an atomic clock at high altitude and comparing it with an atomic clock on the ground. and Wheeler. Implications Conservation of energy involves the conservation of a scalar. h Infinity G M Νo c2 r This shift has been observed in the spectral lines of stars. For the clocks at different heights: Quantum optics. nothing new will happen by treating it as a quaternion. O. P. Limit 1 Βtower 1 Βbase Νo . section 4.139 Ν’ g h Νo This was the measured effect. 1983 (also mentioned in Taylor and Wheeler. Gravitation. Yet the results are no different. Meystre. experimental gravitation and measurement theory. which is a good approximation). t 0 GMh c2 r2 t GhMt c2 r2 This was the measured effect. Ed. Clocks at different heights in a gravitational field C. This is like integrating the redshift over the time of the flight. and escape: Misner. This is just part of the work to connect quaternions to measurable effects of gravity. The equivalence principle was used to transform the problem into a relativistic Doppler shift effect. The only part of the previous calculation that changes is the limit in the final step. 1970. The approach used here was not the standard one employed. Thorne. Spacetime Physics. Consequently. References For the Pound and Rebka experiment. Escape From a Gravitational Potential A photon can escape from a star and travel to infinity ( or to us.10) .

for a Central Force Newton’s 2nd Law in a Noninertial Rotating Reference Frame The Simple Harmonic Oscillator The Damped SHO The Wave Equation Special Relativity Rotations and Dilations Create a Representation of the Lorentz Group An Alternative Algebra for the Lorentz Group Electromagnetism The Maxwell Equations Maxwell Written With Potentials The Lorentz Force Conservation Laws The Field Tensor F in Different Gauges The Maxwell Equations in the Light Gauge (QED?) The Stress Tensor of the Electromagnetic Field Quantum Mechanics Quaternions in Polar Coordinate Form Multiplying Quaternion Exponentials Commutators of Observable Operators The Uncertainty Principle . Polar Coordinates.140 A Brief Summary of Important Laws in Physics Written as Quaternions Summary Classical Mechanics Newton’s 2nd Law in an Inertial Reference Frame. Cartesian Coordinates Newton’s 2nd law in an Inertial Reference Frame.

0 dt 1. R Cos Θ . R Sin Θ . E & B Field Equations Recreating Maxwell Unified Field Equations Conservation Laws Gauge Transformations Equations of Motion Unified Equations of Motion Strings Dimensionless Strings Behaving Like a Relativistic Quantum Gravity Theory Each of the following laws of physics are generated by quaternion operators acting on the appropriate quaternion−valued functions. 0. Sin Θ 2 d . The generators of these common laws often provide insight.The Stress Tensor of the Electromagnetic Field 141 Quantum Mechanics Quaternions in Polar Coordinate Form Multiplying Quaternion Exponentials Commutators of Observable Operators The Uncertainty Principle Automorphic Commutator Identities The Schrödinger Equation The Klein−Gordon Equation Time Reversal Transformations for Intervals Gravity The 3 Fields: g. 0. R Newton’s 2nd Law in Polar Coordinates for a Central Force in a Plane A Cos Θ .0 t. 0 dt . Classical Mechanics Newton’s 2nd Law for an Inertial Reference Frame in Cartesian Coordinates A d . R 0.

d2 dx2 2 0.0 0. 0. R Ω. . 0 m d d . f t v d2 dt2 v2 f tv x .0 0. The Wave Equation d2 x dt2 bdx dt kx . 0 m 0 0. 0. 0.R Ω ΩxR The Simple Harmonic Oscillator (SHO) 2 d . 2LR .F 0 .R . k x. 0. 0 dt d . 0 m R2 Newton’s 2nd Law in a Noninertial. 0 k x.R . 0 v dt dx 0. A Force Is Conservative If The Curl Is Zero odd d . 0. 0. b x. . R 2ΩxR ΩxR Ω.142 0. L2 m2 R3 R. 0. 0 m The Damped Simple Harmonic Oscillator 2 d . 0. The forth term is the instantaneous power transmitted by the wave. 0. 0 m dt2 0. 0 dt d2 x kx 0. 0 dt 0. 0 0. Rotating Frame A d . 0 x . x. dt .Ω dt Ω. 0. x. 2 d2 f t v x dt dx v The third term is the one dimensional wave equation.

. Β. z 2 For boosts along the x axis. q . 0. then Γ 1.143 A Force Is Conservative If There Exists a Potential Function for the Force F d . x. then t.. 0. z 2 scalar L t. 0 Β. y. 0 A Force Is Conservative If the Line Integral of Any Closed Loop Is Zero F dt 0 A Force Is Conservative If the Line Integral Along Different Paths Is the Same F dt path 1 path 2 F dt Special Relativity Rotations and Dilations Create the Lorentz Group q’ q Γ 1 even even V . dt Φ. 0. 0 Γ 1. 0. y. x. V V An Alternative Algebra for Lorentz Boosts 2 Γ even V . Γ 1 Β. q scalar If t = 0. 0 L L If t = x. then for blueshifts For general boosts along the x axis . L If x = 0.

B XE B t odd 0. E odd t E . A t Φ t . Φ. A . . . . Γ Electromagnetism The Maxwell Equations even t B . 0. E Maxwell Written with Potentials The fields E B vector even odd . A xA 0. . . 0 even 4 Π Ρ. odd t . t x . 0. . B. B XB E t t . J t . t . 0 . 0. Φ. . Φ A t . . Φ. A . vector even xA t xE X A t xA. Γ Β 1 Γ x ΓΒy z 1 1 Γ x ΓΒz y 1 2 z t2 x2 . B t 0. 0. The field equations even odd t . .144 L Γ t2 Γ x2 t ΒΓz y t ΓΒy z t2 x2 y2 2 Γ Βt x y2 z2 . Γ . 0. Φ.

0 The Field Tensor F in Different Gauges The anti−symmetric 2−rank electromagnetic field tensor F . The Lorentz Force A t2 XB odd Γ. t . 0 t Poynting’s theorem for energy conservation. t . scalar E 0. J 4Π E J Ρ . 0 E t t scalar . Φ. Γ Β . Φ t A Φ E . B ΓΒ XB even Γ. A t 2 . E ΓΒ E . E . . 0. 0 E . J . t X E t XA 4 Π Ρ. 0. E t t . E. Φ. . J 4 Π E J. odd . Γ Β . 0 t 1 2 E t XB E t 2 EXB scalar 0. ΓE Conservation Laws The continuity equation scalar E t . XB E E E. XB X B. vector even A .145 odd even t . 4 Π Ρ. 1 2 B t 2 . 4 Π Ρ.

A 4 Φ. Φ A XA Φ. A Φ. Φ. Φ. A Φ t . A 4 A t Φ XA t F in the temporal gauge. Φ. .A. . A t Φ. F in the light gauge. . A Φ.146 t 0. A Φ. A t . A Φ. A 4 Φ. Φ. A t A t Φ . t A Φ. A t Φ XA The light gauge is one sign different from the Lorenz gauge. t .A. but its generator is a simple as it gets. F in the Coulomb gauge t . A t . t A t Φ XA t . A 2 Φ. Φ. t . A 2 Φ t Φ. Φ.A. . XA . A t . F in the Lorenz gauge. A 4 . Φ .

E 2 0.z y. odd Φ t . . XA 4 Π Ρ.147 The Maxwell Equations in the Light Gauge Note: subsequent work has suggested that the scalar in these equations is part of a unified field theory. Ub 3 1 0. 2 A t2 A The Stress Tensor of the Electromagnetic Field Tik y. even odd t . Ub even odd E. A t XA. odd . even X . B . . . J A A . B . t . A . t X 2 t . odd even 2 0. Ua even B. B 2 2 even E. Φ. . 0 t . Φ.z a x b x 1 4Π even Ua. 0 2 Π Quantum Mechanics Quaternions in Polar Coordinate Form q q exp Θ I q q Cos Θ I Sin Θ . Ub Bx By Ex Bz Bx Bz Ex By By Bz Ey Bx. even Φ. t . Φ. Φ. 2 Φ t2 2 A t2 Φ t2 2 Φ. Ub even odd E. . Ua Ex Ey Ey Bz Ex Ez Ez By Ey Ez Ez Bx even B. Ua even E. A .

Pn En . En . Re2 0 2 Re3 0 Θ2 R 0 Automorphic Commutator Identities q. q ’ Commutators of Observable Operators A. q’ q . q’ 2 2 2 2 q 1. Pn . B 2 I 2 ∆A2 ∆B2 Unifying the Representation of Spin and Angular Momentum For small rotations: Re1 0. q’ 1 1 1 1 q 2. q ’ Exp Π q. q ’ The Schrödinger Equation q . q ’ Abs q. Pn En . B q aI dq da AB aI BA q dq da aI Iq dq da I Iq daq da da da The Uncertainty Principle A. X Ψ The Klein−Gordon Equation 2 n 0 2 2 2 t . q’ q 2.148 Multiplying Quaternion Exponentials q q’ q. 2 t . q ’ q. q ’ 2 Abs q. q’ Exp HΨ i V V. q’ q 1.V Ψ t Ωt 2 K.X 2 2m Ψ V 0.

X t.X if Β R 1 then R Ε . t2 Pn X Pn En Pn X Pn X Pn 2P n Pn Pn t2 It takes some skilled staring to assure that this equation contains the Klein−Gordon equation along with vector identities. X 1 R t. 2 t Β . X t. 1.149 X Pn n 0 En Pn En Pn Pn En En 3 2 Pn X X Pn Pn Pn X Pn X Pn En 2 En .X. Time Reversal Transformations for Intervals t. X t2 X. X R Classically t. 0. 2 t X t2 X. 0 T 1.

A.A xA t . . xA x Φ A x Φ xA Φ Φ.E . E xB W xB U B xB B 2 B xE xE g g Mirror .B. . Φ 2 A 4 Π Ρu . Jg Je xB B xE g 4 Π Ρg Recreating Maxwell Let U W .B g. Φ.150 Gravity The 3 Fields: g. Jg g.E . . A A Φ xA g. . . E Ρe . .E W .B U g. E 2 E g. E B xB B xE g .B . x A. E B Field Equations: Almost Maxwell and a Dynamic g t g .E g. Φ . Ju .B.A A 2 A .E Unified Field Equations xE g t Φ . E B Je 4 Π Ρg t g .A Φ. E & B t Φ . . E Ρe .

151

t Φ

, .A xA

t

,

Φ, . Φ .A A

A . x A, xA
2

.A Φ

A Φ Φ 2 x

Φ xA xA

x A 2

Φ

2

Φ

2 .A ,

4 Π Ρu , Ju
Conservation Laws

t g

,
2

t g, E

,

g, E xB

B .E Je xE x xB

t 4Π Ρg Ρe Jg

,

4 Π Ρg

Ρe , Jg

Ρe 0

.Jg

.Je , Jg

Je

Ρg

Ρe

x Jg

x Je

.Je Ρg

x Jg

0

If the differential operator acts on the hyperbolic equation, analogous results are obtained:

t g

,
2

t

, B

g, E .E Ρe , Jg

B x Je xB x xE

g, E

t 4Π Ρg

,

4 Π Ρg

Ρe

.Jg

.Je , Jg

Je

Ρg

Ρe

x Jg

x Je

There are two conservation laws here, charge conservation for electromagnetism in the scalar, and a vector conservation for gravity.

Ρe

.Je

0

152

Jg

Ρg

x Jg

0

Gauge Transformations

Φ, A Φ, A
Equations of Motion

Φ ’, A ’ t ,

Φ Λ,

Λ

. , A

Λ

x

Γ, Γ Β Γg W m

g, E Γ Β.E P W , m e

B Γ Β.B, Γ E P e ΓΒxB ΓB ΓΒxE ΓΒg

Unified Equations of Motion Repeat the exercise from above, but this time, look to the potentials.

Γ, Γ Β Γ, Γ Β ΓΦ ΓA

t Φ

, .A,

Φ, A

A Φ Γ Β. Φ ΦΓΒ Φ ΓΒx xA xA Γ Β. x A,

Γ .A Γ Φ

Γ Β.A Γ xA

.A Γ Β

ΓΒxA

ΓΒx

That is pretty complicated! The key to simplifying this equation is to see what happens for light, where dt/dx = dx/dt. Gamma blows up, but if the equation is over gamma, that problem becomes a scaling factor. With beta equal to one, a number of terms cancel, which can be seen more clearly if the terms are written out explicitly.

153

Φ A .A

X

.

A t Φ t

X . t X X x x A t

A t A t X

X . t Φ t x

X

Φ X t

X . t

X

x A,

X X X t

X

Φ

X

x

X

xA

It would take a real mathematician to state the proper constraints on the three pairs of cancellations that happen when velocities get flipped. There are also a pair of vector identities, presuming simple connectedness. This leads to the following equation:

2 Φ,

A

X

.A

X t

X

x

X

xA

The scalar change in energy depends only on the scalar potential, and the 3−vector change in momentum only depends on the 3−vector A.
Strings

dq2

da0 2 e0 2 2 da0 da1 e0

da1 2

e1 2 9

da2 2

e1 , 2 da0 da2 e0 3 string

e2 2 e 2 da3 2 3 , 9 9 e2 e , 2 da0 da3 e0 3 3 3

interval2 , 3
Dimensionless Strings

dq2

c5 Gh

da0 2 e0 2

da1 2

e1 2 9 c2

da2 2

2 da0 da1 e0

e1 , 2 da0 da2 e0 3c

e3 2 , 9 c2 e2 e , 2 da0 da3 e0 3 3c 3c e2 2 9 c2 da3 2

As far as the units are concerned, this is relativistic (c) quantum (h) gravity (G). Take this constants to zero or infinity, and the difference of a quaternion blows up or disappears.
Behaving Like a Relativistic Quantum Gravity Theory Case 1: Constant Intervals and Strings

T : dq dq ’ such that scalar dq2 scalar dq ’2 and vector dq2 vector dq ’2

changes in the reference frame of an inertial observer are logically independent from changing the mass density. and the change in the interval will involve changes in the mass density.154 Case 2: Constant Intervals T : dq dq ’ such that scalar dq2 scalar dq ’2 and vector dq2 vector dq ’2 Case 3: Constant Strings T : dq dq ’ such that scalar dq2 scalar dq ’2 and vector dq2 Case 4: No Constants vector dq ’2 T : dq dq ’ such that scalar dq2 scalar dq ’2 and vector dq2 vector dq ’2 In this proposal. The change in the length−time of the string will involve the inertial reference frame. . The two effects can be measured separately.

Quaternions do not make problem solving easy. and the Klein−Gordon equation for relativistic quantum mechanics. If the origin changes. The only difference lies in the cross term. including a scalar and vector identity analogous to the Maxwell equations. so the parallel to the Maxwell equations is less clear.155 Conclusions What’s been done It is an old dream (or perhaps more accurately. These additional identities are also valid by the conventional analysis. Physics is impressively democratic. This implies that any collection of events can be generated by an appropriate quaternion function. This might strike some as a comic reliance on a solitary tool. In these web pages. so quaternions are a mixed representation. the affine parameter. the Klein−Gordon equation is part of a larger set. they help unite the laws themselves. Applications of quaternions to special relativity were done in detail. does anything genuinely novel appear? I believe that a new very powerful idea drives this work. quaternions are a linear combination of tensors of rank zero and one. but they do not naturally arise. with over 50 problems worked out explicitly. quaternion operators were employed to express central laws of physics: Newton’s second law. an analysis of the length of a quaternion interval if the origin is moved establishes a connection to the machinery of general relativity. Every event. that events represented as quaternion are a topological algebraic field. A new link to general relativity has been proposed which is slightly different. When quaternion operators are employed. the Maxwell equations. I prefer to think of it as a great democratic principle. The mathematics underlying the laws of physics should reflect this interchangeability. every function. The invariant interval in special relativity was the first term of the difference between two events quaternions squared. . Rather. Significantly. The Klein−Gordon equation is a scalar equation. with each photon or electron obeying the same collection of laws interchangeably. a recurring nightmare) to express laws of physics using quaternions. and while used in a new way here. then the first term of the difference between the two event quaternions and the origin quaternions squared is similar to the square of the affine parameter of general relativity. namely. A new view of relativistic quantum mechanics was outlined. every operator used was a member of the field of quaternions. After all. Scalars and vectors mix under multiplication. What’s new One might suspect that the reason for the success claimed above is that nothing proposed is new.

And since each of these is a quaternion. and cling to a novel toehold on general relativity. This body of work is more like a skeleton of work.. I hope this work generates interest in the physics community. but it might be elegant. I won’t be easy. There were few references to the literature. My upcoming focus will be the Dirac function and Fourier analysis using quaternions. it never is. And maybe I’ll dabble a little with curves. it becomes possible to mix and match them to create new areas of study. Although I appreciate the standard approach.. If I can build these functions. I feel like I have gained new insight into why Maxwell’s equations are necessary. . have a new way to view relativistic quantum mechanics. There is a need to flesh these ideas out only if there is the potential for new insights. More problems need to be solved. it should be possible to approach problems in electromagnetism and quantum mechanics.156 What needs to be done Nothing presented was proven rigorously. a reflection of the author’s semi−formal training and isolation from the professional physics community.

Sign up to vote on this title