1

Where did the laws of physics come from?

Victor J. Stenger
Department of Philosophy, University of Colorado at Boulder
Boulder, Colorado
Department of Physics and Astronomy, University of Hawaii at Manoa
Honolulu, Hawaii
E-mail: vstenger@[Link]

Abstract
The laws of physics are constrained so that they select out no preferred coordinate
system or reference frame. This is called the principle of covariance. This principle can be
further generalized to include the coordinates in the abstract space of the functions used
to formulate those laws. This is called global gauge invariance. When this symmetry
applies independently at every point in space-time, it is called local gauge invariance.
These symmetries are almost all that are needed to derive most of the familiar laws the
law of physics, including classical mechanics, the great conservation laws, quantum
mechanics, special and general relativity, and electromagnetism. Those structures that
do not follow directly from coordinate invariance result from spontaneously broken
symmetries.

PACS numbers: 01.55, 01.70, 11.30, 11.30E, 14.80.C

1.0 Introduction
Most laypeople think of the laws of physics as something like the Ten
Commandments—rules governing the behavior of matter imposed by some great
lawgiver in the sky. However, no stone tablet has ever been found upon which such
laws were either naturally or supernaturally inscribed. On the contrary, the laws of
 2
physics are human inventions—mathematical formulas that quantitatively describe the
results of observations and measurements. These formulas are first inferred from and
then tested against observations. If they hold up, they are eventually reformulated as
part of general and universal theories that are derived from a minimum number of
assumed fundamental principles. Very often, a "law" will turn out to be nothing more
than a circular definition, such as Ohm's law which says that the voltage is proportional
to the current in a resistor, where a resistor is defined as a device that obeys Ohm's law.
Since the time of Copernicus and Galileo it has been realized that the laws of
physics should not single out any particular space-time reference frame, although a
distinction between inertial and noninertial frames was maintained in Newtonian
physics. That distinction was removed in 1916 by Einstein who formulated his general
theory of relativity in a covariant way. That is, the form of Einstein's equations is the
same in all reference frames, inertial or noninertial.
As this experience showed, physicists are highly constrained in the way they
may formulate the laws of physics. Not only must they agree with the data, the
equations that are used to describe that data should not be written in such a way as to
specify a privileged coordinate system or reference frame. This principle of covariance
generalizes other notions such as the Copernican and cosmological principles and the
principle of Galilean relativity. The application of this principle is not a matter of choice;
centuries of observations have shown that to do otherwise produces calculations that
disagree with the data in some reference frames.
In 1918, Noether showed that coordinate independence was more than just a
constraint on the mathematical form of physical laws.[1] She proved that some of the
most important physics principles are, in fact, nothing more than tautologies that follow
from space-time coordinate independence: energy conservation arises from time
translation invariance, linear momentum conservation comes from space translation
invariance, and angular momentum conservation is a consequence of space rotation
invariance. These conserved quantities were simply the mathematical generators of the
 3
corresponding symmetry transformation.
As the twentieth century progressed, invariance or symmetry principles became
an increasingly dominant idea in physics. Not only were space-time coordinate
symmetries built into theories, the notion of coordinate independence was extended to
the abstract spaces physicists use to represent the other degrees of freedom of systems.
Rotational symmetry was also applied to the space of quantum state vectors, resulting
in derived properties of spin, isospin, charge, baryon number, and other observables
that agreed with measurements.
Charge conservation, for example, was found to follow from the invariance of
the Schrödinger equation to changes in the phase of the complex wave function. And
then, a remarkable discovery was made. It was found that the Schrödinger equation
could be made invariant to a local phase change in the wave function, that is, a change in
phase that varies from point to point in space-time, provided that vector and scalar
potentials were added. The potentials turned out to be exactly those that give the
classical electric and magnetic fields. This local quantum phase symmetry was precisely
related to the local classical gauge symmetry of electrodynamics. Maxwell's equations
were derived from a single principle—local phase invariance.
If we think of the Schrödinger wave function as a "vector" in 2-dimensional
complex space, then changing phase is equivalent to a rotation in that space and phase
invariance, or gauge invariance, is equivalent to rotational invariance. Indeed, the
generator of that transformation is the electric charge whose conservation follows from
global gauge invariance.
In the standard model, the fields associated with the weak and strong nuclear
forces are obtained by extending the idea of gauge symmetry to higher dimensions of
abstract space. There the situation is complicated by the fact that all the symmetries are
not exact at the "low temperatures" of current experimentation. Good thing. The
diversity and complexity of the universe is a result of broken symmetries, without
which we would not be here to do the experiments.
 4
Twentieth century physics was also marked by the discovery that symmetries
are often broken. In the 1950s, it was found that weak interactions maximally violated
space reflection symmetry; that is, they were not invariant under the parity operation P
that changes the handedness or chirality of a system. In the 1960s certain rare decays
were found to be noninvariant under the combined operation CP, where C changes a
particle to its antiparticle. The study of the origin of CP violation remains a subject of
considerable experimental and theoretical effort to this date.
In this paper, it will be shown that much of familiar physics can be derived from
the generalized notion of coordinate invariance applied not only in space-time but in
the spaces of other observables and the spaces of the functions that are used to
mathematically describe physical phenomena. In order to make this result accessible to
the greatest number of people, the mathematical level will be limited to that of an
advanced undergraduate student in physics or mathematics. The equations will appear
very familiar—just those found in physics textbooks, and it may appear that the author
is using hindsight to make things come out the way they already are. However, the
reader is asked to look carefully at how those equations are obtained. Certain familiar
principles normally taken as axioms, such as the quantization of angular momentum
and the invariance of the speed of light will be derived from the hypothesized
symmetry principles without additional assumptions.

2.0 Gauge Symmetry

Let q = (qo, q1 , q2 , q3 , . . . qn ) be the set of observables of a physical system such as a

particle or group of particles and take them to be the coordinates of an n-dimensional

vector q in q-space. Spatial coordinates and time are included and placed on the same
footing as the other observables. Thus a point in q-space, designated by the vector q,
represents a particular set of measurements on a system. The generalized principle of
covariance says that the laws of physics must be the same for any origin or orientation
 5
of q, that us, any choice of coordinate system.
Let us define a vector y in another multidimensional space we will call y-space.

Assume y = y(q). The state vectors of quantum mechanics are familiar examples of y-

space vectors. We can imagine a set of coordinate axes in y-space. Extending the notion

of covariance to this space we will assume that the following principle holds: the laws of
physics cannot depend on the orientation of the vector y in y-space. This principle is called

gauge symmetry.

3. Gauge Transformations and their Generators

To get started as simply as possible, let us take y(q) to be a complex function, that is, a

2-dimensional vector with coordinates (Re{y }, Im{y }). Let us perform a unitary

transformation on y:

y' = Uy (3.1)

where U† U = 1, so

y† y' = y† U† Uy= y† y (3.2)

This transformation does not change the magnitude of y,

|y'| = (y'† y' )1/2 = (y† y )1/2 = |y| (3.3)

That is, |y| is invariant to the transformation, as required by gauge symmetry. We can

write the operator U

 6
U = exp(iq) (3.4)

where q† = q, that is, q is a hermitian operator. Then,

y' = exp(iq) y (3.5)

So, U simply changes the complex phase of y. It could be called a "phase

transformation," or just simply a unitary transformation. However, in the

amplifications of this idea that we will discussing, the designation "gauge
transformation" has become conventional. When q is a constant we have a global gauge

[Link] q is a not a constant but a function of position and time it is called

a local gauge transformation.

Note also that the operation U corresponds to a rotation in the complex space of
y. Later we will generalize these ideas to where y is a vector in higher dimensions and q
will be represented by a matrix. But this basic idea of a gauge transformation as
analogous to a rotation in an abstract function space will be maintained and gauge
invariance viewed as an invariance under such rotations.
Let us write

q = eG (3.6)

where e is an infinitesimal number and G is another operator. Then

U ≈ 1 + ieG (3.7)

.
where G† = G is hermitian and is called the generator of the transformation. Then,
 7
y' ≈ y + ie G y (3.8)

Suppose we have a transformation that takes the variable qm to q'm = qm + em.

Then

y'(q'm ) = y(qm + em ) ≈ y(qm ) + em ∂y/∂qm (3.9)

It follows that the generator can be written

Gm = -i∂/∂qm (3.10)

Define

∂
Pm ≡ hG = -ih (3.11)
∂q
m

† introduced only if you want the units of Pm to be

where h is an arbitrary constant

different from the reciprocal of the units of qm. The transformation operator can then be
† written

∂
U =1+ (3.12)
∂qm

For example, suppose that q1 †

= x, the x-coordinate of a particle. Then
 8
∂
P1 ≡ Px = -ih (3.13)
∂x

†
which we recognize as the quantum mechanical operator for the x-component of
momentum. Note that this association was not assumed but derived and no connection
with mass and velocity has yet been made. This just happens to be the form of the
generator of a space translation. Similarly, we can take q2 = y, q3 = z and obtain the

generators Py and Pz .

It may also be noted that q might contain operationally defined momenta, in

which case spatial coordinates would then be introduced in the manner of (3.13).
Of course, h will turn out to be the familiar quantum of action of quantum
mechanics, h = h/2π where h is Planck's constant. Physicists often take h = 1 in "natural
units." We
† will leave h in our equations at this point to maintain familiarity, however it
should
† be recognized that this constant, when expressed in non-dimensionless
† units,
will turn out†to be an arbitrary number determined only by that choice of units. No
additional physical assumption about the "quantization of action" need be made and
Planck's constant should not be viewed as a metric constant of nature. In particular, h
cannot be zero. Once we have made the connection of (Px , Py , Pz) with the 3-

momentum, quantization of action will already be in place. †

We can also associate one of the variables, say qo with the time t. In order to

provide a connection with the fully relativistic treatment we will make later, let qo ≡ ict,

where c is, like h, another arbitrary conversion factor. Later we will associate it with the
speed of light in a vacuum and find (not assume) that it is a Lorentz invariant. For now,

†
∂ h ∂
Po = ih =- (3.14)
∂q o c ∂t

†
 9
We can then define

∂
H ≡ -iPoc = ih (3.15)
∂t

†
which we recognize as the quantum mechanical Hamiltonian (energy) operator. Note,
again, that this familiar result was not assumed but derived. No connection with the
physical quantity energy has yet been made. This just happens to be the form of the
generator of a time translation.

4. Quantum Mechanics from Gauge Transformations

Suppose we have a complex function y(x, y, z, t) that describes, in some unspecified

way, the state of a system. It will evolve with time according to

∂
Hy = ih (4.1)
∂t

†
This is the time-dependent Schrödinger equation of quantum mechanics, where y is

interpreted as the wave function. If H is independent of time, we have the solution

Ê i ˆ
y (t) = y (0)expÁ - Ht ˜ (4.2)
Ë h ¯

so †

Ê i ˆ
U(t) = expÁ - Ht ˜ (4.3)
Ë h ¯

†
 10

is the time evolution operator.

At this point, then, we have the makings of quantum mechanics with no physical
assumptions whatsoever. That is, we have a mathematical theory that looks like
quantum mechanics although we have not yet identified the operators H and P with the
physics quantities energy and momentum. We have simply noted that these are
generators of time and space translations respectively, which are themselves gauge
transformations.
Let us proceed along these same lines, considering only the mathematics of
gauge transformations and leaving the physics to later. This does not stop us from
using the Dirac bra and ket notation for linear vectors and operators. Again, no physical
assumption is being made. We are simply using a convenient mathematical formalism.

So, let y be a linear vector and y be its dual. For simplicity, we take our linear

vectors to have unit norm,

† †
y y =1 (4.4)

A unitary transformation on y† will preserve the norm.

† y¢ = U y > (4.5)

† y U †U y = 1 (4.6)

Let A be a linear operator that

† gives another vector of unit norm

f = Ay (4.7)

†
 11

Then,

f Ay = 1 (4.8)

and †

f U † AU y = f ¢ A y = f ¢ f ¢ = 1 (4.9)

We can define †

A' = U† A U (4.10)

and write

f A¢ y = 1 (4.11)

Let us consider the specific case

† where U is the time evolution operator,

y(t) = U y(0) (4.12)

In that case, †

f (0) U † AU y (0) = f(t ) A y (t) = f(t ) f (t) = 1 (4.13)

Alternatively,
† define
 12

A(t) = U † A(0)U (4.14)

Then, †

f (0) A(t) y (0) = 1 (4.15)

This illustrates the two

† approaches to time evolution in quantum mechanics. In
the Schrödinger picture, the state vector varies with time while the operators stay fixed.
In the Heisenberg picture, the state vectors remain fixed while the operators evolve
with time.
If we now interpret, in usual quantum mechanical fashion, the state vectors in
terms of probabilities and the operators in terms of observables, the expectation value
for the observable A, the mean value expected for an ensemble of measurements of A

when the system is in the state y(0) is, in the Schrödinger picture,

† A(0) = y (0) A y (0) (4.16)

It evolves with time according

† to

A(t) = y(t) A y (t) (4.17)

where †

y(t) = U y(0) (4.18)

†
 13
In the Heisenberg picture we have

A(0) = y(0) A(0)y(0) (4.19)

and †

A(t) = y (0) A(t) y(0) (4.20)

where †

Êi ˆ Ê i ˆ
A(t) = U † A(0)U = expÁ Ht ˜ A(0)expÁ - Ht ˜ (4.21)
Ëh ¯ Ë h ¯

Let us †
look further at the time evolution of operators. Suppose we make an
infinitesimal transformation in time t Æ t + dt. Then

i
U(t) = 1 - Ht (4.22)
h

†Ê i ˆ Ê i ˆ i
A(t + dt) = Á 1+ Ht ˜ A(t)Á 1- Ht˜ = A(t) - [ A, H ] (4.23)
Ë h ¯ Ë h ¯ h

Since †

∂A
A(t + dt) = a(t) + dt (4.24)
∂t

†
 14

it follows that

∂A i
= - [ A,H ] (4.25)
∂t h

† observable then is
The time rate of change of an

dA ∂A ∂A dq k
= +Â (4.26)
dt ∂t k ≠0 ∂q k dt

or, †

n
dA i ∂A dq k
= - [A, H ] + Â (4.27)
dt h k =1
∂q k dt

†
where the sum excludes the time variable.
Now we move to gauge transformations involving the non-temporal variables
of a system. Consider the case where A = Pj . Then,

dPj n ∂P
i j dq k
dt
=-
h
[ ]
Pj ,H + Â
∂q k dt
(4.28)
k=1

†
Next, let us look at the transformation of these non-temporal variables. Let the variable

q k Æ q k + e k , where ek is infinitesimal. Then, as we saw above, the transformation

†
 15
operator is

i
U =1+ Pe (4.29)
h k k

Thus, †

i
y(q k + e k ) = y (q k ) + P e y (q k ) (4.30)
h k k

Suppose we have an
† operator A defined by

A y (q k ) = f (q k ) (4.31)

In our previous consideration

† of the time variable we derived the time evolution
equation for an operator in the Heisenberg picture where the time dependence is
carried by the operator rather than the state vector. Let us continue to work in that
picture. The state vectors will then not depend explicitly on time, but they still can
depend on the other variables. So,

f(q k + e k ) = U f (q k ) = UAU +U y (q k ) = UAU + y( q k + ek ) = A¢ y (q k + e k ) (4.32)

†
where

A' = UAU† (4.33)

 16
We can write this

Ê i ˆ Ê i ˆ i
A¢ = Á 1+ Pke k ˜ AÁ 1- Pkek ˜ = 1- [ A, Pk ] (4.34)
Ë h ¯ Ë h ¯ h

so, †

∂A i
= - [ A, Pk ] (4.35)
∂q k h

From the differential form †

of the operators Pk,

[Pj , Pk ] = 0 (4.36)

for j ≠ k, and so †

∂Pj
=0 (4.37)
∂q k

†
Recall (4.28),

dPk i ∂P dq j
= - [Pk ,H ] + Â k (4.38)
dt h j
∂q j dt

† all zero, so
The summed terms are
 17

dPk i
= - [Pk ,H ] (4.39)
dt h

We can also think of†qk as an operator, so

∂q k i
= 1 = - [ q k ,Pk ] (4.40)
∂q k h

or, †

[q k ,Pk ] = ih (4.41)

This can also be seen from †

È h ∂ ˘ h ∂y h ∂
[q k ,Pk ]y = Íqk , i ∂q
˙y = q k
i
- q y = ihy
∂q k i ∂q k k
(4.42)
Î k˚

†
For example,

[x,Px ] = ih (4.43)

the familiar quantum mechanical

† commutation relation.
Now, we can also write
 18
∂H h
= [H, Pk ] (4.44)
∂q k i

Thus, †

dPk ∂H
=- (4.45)
dt ∂q k

which is the operator version †

of one of Hamilton's classical equations of motion and
another way of writing Newton's second law of motion. Here we see that we have
developed another profound concept, from gauge invariance alone. When the
Hamiltonian of a system does not depend on a particular variable, then the observable
corresponding to the generator of the gauge transformation of that variable is
conserved. This is a version of Noether's theorem mentioned in the Introduction.
From this point, the rest of quantum mechanics can be developed. Observables
A are represented as hermitian operators and the expectation value of A is

A = y Ay (4.46)

The possible results of a measurement

† of A is determined by the solutions of the
eigenvalue equation

Aa =a a (4.47)

where a is the eigenstate of A

† corresponding to eigenvalue a. When the state of a

system is an eigenstate of an observable, the measurement of that observable will

†
 19
always yield the eigenvalue corresponding to that state.

The symbol a a stands for an operator that projects y onto the a axis.

When the eigenvectors a form a complete set,

† † †

† Âa a =1 (4.48)
a

In that case, the state vector of†a system will be the linear combination

y =Â a ay (4.49)
a

2 † for y to be found in the eigenstate a . The wave

where a y is the probability

function is defined as the inner product

† † †

y (q) = q y (4.50)

where q are the eigenstates †

of the spatial coordinates (or space-time coordinates

relativistically) of the particles of the system. Momentum-space wave functions are also

† often used.
More generally, the eigenstates q are the basis states of a particular, arbitrary

representation, like the unit vectors i, j, and k of the Cartesian coordinate axes x, y, z.

y(q) is the projection of y on†q .

We can represent y and q as column matrices. Then

† †
 20
y (q) = Â y †i q i (4.51)
i

where yi† is a row matrix. †

In this representation, the observable A is a square matrix,

A = Âyi† Aijy j (4.52)

i, j

or simply given by the matrix

† equation,

A = y † Ay (4.53)

Thus, the gauge transformation

† can be written

y ¢ = Uy = exp(iq )y (4.54)

where U and q, and the corresponding

† generators, are square matrices.

5.0 Rotation and Angular Momentum

The variables (q1 , q2 , q3 ) can be identified with the coordinates (x, y, z) of a particle and

the corresponding momentum components are the generators of translations of these

coordinates. (In this formulation, nothing prevents other particles being included with
their space-time variables associated with other sets of four q's; note that by having
each particle carry its own time coordinate we can maintain a fully relativistic scheme.)
These coordinates can equally well be angular variables and the conjugate
 21
momenta the corresponding angular momenta. These angular momenta will be
conserved when the Hamiltonian is invariant to the gauge transformations that
correspond to displacements of the corresponding angles. In this case, the
displacements will be rotations about the spatial axes. For example, if we take (q1 ,$q2 ,$q3 )

= (fx ,$fy ,$fz ), where fx is the angle of rotation about the x-axis, etc., then the generators

of the rotations about these axes will be the angular momentum components
(Lx ,Ly ,$Lz). Rotational invariance about any of these axes will lead to conservation of

angular momentum about that axis.

Let us look for a moment at rotations in familiar 3-dimensional space. Suppose
we have a vector V = (Vx , Vy ) in the x-y plane. Let is rotate it counter clockwise about

the z-axis by an angle f. We can write the transformation as a matrix equation:

ÊVx¢ ˆ Ê cos f - sin f ˆÊVx ˆ

ÁÁ ˜˜ = Á ˜ÁÁ ˜˜ (5.1)
ËVy¢ ¯ Ë sin f cosf ¯ËVy ¯

†
Specifically, let us consider an infinitesimal rotation of the position vector r = (x, y) by df

about the z-axis. From above,

Ê x¢ˆ Ê 1 -df ˆÊ x ˆ Ê x - ydf ˆ

Á ˜ =Á ˜Á ˜ = Á ˜ (5.2)
Ë y¢¯ Ë df 1 ¯Ë y ¯ Ë y + xdf ¯

And so,
†

dx = -ydf (5.3)

and
 22
dy = x df (5.4)

For any function f(x, y),

∂f ∂f
f (x + dx, y + dy) = f (x,y) + dx + dy (5.5)
∂x ∂y

† we can write (reusing the function symbol f ),

to first order. Or,

∂f ∂f
f (f + df) = f (f ) - ydf + xdf (5.6)
∂x ∂y

= f (f ) + idfGf
†
where

†
Ê ∂ ∂ˆ
G = -i Á x - y ˜ = xPy - yPx = Lz (5.7)
Ë ∂y ∂x ¯

†
the angular momentum about z. Similarly,

Lx = yPz - zPy (5.8)

and

Ly = zPx - xPz (5.9)

This result can be generalized as follows. If you have function that depends on a spatial
 23
position vector r = (x, y, z), and you rotate that position vector by an angle q about an

arbitrary axis, then that function transforms as

f'(r) = exp(iL•q )f(r) (5.10)

where the direction of q is the direction of the axis of rotation. Once again this has the

form of a gauge transformation, or phase transformation in f, where

U = exp (i L•q ) (5.11)

From the previous commutation rules one can show that the generators Lx , Ly ,

and Lz do not mutually commute. Rather,

[Lx ,Ly ] = ihLz (5.12)

and cyclic permutations. Thus

† the order of successive rotations is important.
Most quantum mechanics textbooks will contain the proof of the following
result, although it is not always stated so generally: Any vector operator J whose
components obey the angular momentum commutation rules,

[ J x , J y ] = ihJ z (5.13)

will have the following eigenvalue

† equations:

J 2 j ,m = j(j + 1)h2 j ,m (5.14)

†
 24
2
where J = J 2x + J 2y + J 2z is the square of the magnitude of J.

† J z j ,m = mh j ,m (5.15)

where m goes from -j to + j †

in steps of one: m = -j, -j+1, . . . ,j-1, j. This implies that j is an
integer (including zero) or a half-integer.

6. Rotation and Gauge Transformations

We have already noted that the gauge transformation is like a rotation in the complex
space of a function. Let us now generalize that concept.
Again, it is important not to confuse the two different spaces involved in our
discussion. First we have the space spanned by the variables {q} of a system. We have
generally taken the first four of these to be the subspace of 4-dimensional space-time in
which we describe events. If our "system" contains more than one event, then
additional groups of 4-dimensional subspaces can be reserved for these. Other
subspaces are left available for other variables.
Besides q-space, an additional abstract space have alread introudced as y-space is

used to describe the quantum state of a system. That space has coordinate axes that are

defined by an arbitrary choice of basis vectors of the system q , where if Q is the

operator corresponding to an observable,

†
Qq =qq (6.1)

Thus, †
 25
y = Â ci qi (6.2)
i

The basis states are usually taken

† to be orthonormal, that is,

q i q j = dij (6.3)

2 2†
so y y = 1 and c i = qi y is the probability for a measurement of Q giving the

value qi when the system is in the state y .

† †
For example, the basis states are frequently chosen to be x , y , and z , where
† coordinates of a particle, that is, all the eigenstates of
the observables are all the possible
the eigenvalue equation † † †

Xx = xx (6.5)

The quantity y (x) = x y is, for

† historical reasons, called the wave function, although it

often has nothng to do with waves. Since x is usually regarded as a continuous variable,

y-space
† is infinite dimensional. That is, x is not one axis but an infinite number of

axes, one for every real number x. Even if we assume that x is discrete in units of the
Planck length, and space is finite,
† we still have an awfully large number of dimensions.

If the particle is an electron, then y-space may also include the basis states + 12

and - 21 that are the eigenstates of the z-component of spin of the electron. Even
†
though spatial coordinates are more familiar than spins, 2-dimensional spin subspace is

† a lot easier to visualize than the subspace of spatial coordinate eigenstates.

 26
In the 2-dimensional subspace spanned by the spin state vector of an electron,

the basis states + 12 and - 21 can be thought of as analogous to the unit vectors i and

j in the more familiar 2-dimensional subspace (x, y). The spin state y is in general a 2-
† † oriented at some arbitrary angle. The basis vectors define two
dimensional vector
possible orientations of the spin angular momentum vector†S in familiar 3-dimensional
space, one along the z-axis and the other opposite. (The choice of z-axis here is arbitrary

conventio). Thus, for example, if S points originally along the z-axis, a rotation of 180o

will take it to point along -z.

However, note that a rotation in y-space of only 90o takes the spin state from

+ 12 to - 21 . This implies that the unitary transformation matrix in this case is

† † Ê qˆ
U = expÁ i ˜ I (6.6)
Ë 2¯

where I is the unit 2x2 matrix.†

More generally,

Ê s • qˆ
U = expÁ i ˜ (6.7)
Ë 2 ¯

† in the direction around which we rotate, and s is the

where the axial vector q points

Pauli spin vector whose components are conventionally written

 27
Ê0 1ˆ Ê 0 -i ˆ Ê1 0 ˆ
sx = Á ˜ sy = Á ˜ sz = Á ˜ (6.8)
Ë 1 0¯ Ëi 0 ¯ Ë 0 -1¯

† †
We see that U again has the form of a gauge†transformation. The generator of the
gauge transformation in the spin vector subspace of a spin 1/2 particle is the spin
angular momentum operator (in units of h ), S = s/2. We could also have obtained this

result from our previous proof that the gauge transformation for a rotation in 3-space
is †

U = exp(iL• q ) (6.9)

where L is the angular momentum. Here L = S = s/2.

7. Special Relativity
Now we are ready to inject some familiar physics into the mix. It turns out to be most
elegant to do this within the framework of special relativity. But note that, as was the
case for quantum mechanics, the usual starting axioms will not be asserted. Rather they
will be derived from the assumption of gauge invariance.
Let us consider the first four variables (q0 , q1 , q2 ,q3 of our set {q} which we

have arbitrarily set to (x0 , x1 , x2 ,x3 ) = (ict, x, y, z), where t is the time and (x, y, z) are the

spatial coordinates of an event. The constant c is simply a factor that converts units of
time to units of distance. It will turn out to be the invariant speed of light in a vacuum,
but that is not being assumed at this point. Also, the assumption that q0 is an imaginary

number is not necessary; it just makes things easier to work out at this level of
sophistication.
 28
Let x' = (x'0 , x'1 , x'2 , x'3 ) be the position of the event in reference frame moving

at a speed v = bc along the z-axis with respect to the reference frame x, where

x¢m = Lnm xn (7.1)

and the convention is used in which

† repeated Greek indices are summed from 0 to 3. As

is shown in many textbooks, the proper distance will be invariant if Lnm is the Lorentz

transformation operator

†
1 0 0 0
0 1 0 0
L nm = (7.2)
0 0 cosy siny
0 0 - siny cosy

where cosy = g, siny = ibg, and g = (1 - b2 )1/2. By writing it this way, we see that the

Lorentz transformation between reference frames moving at constant velocity with

respect to one another along their respective z-axes is equivalent to a rotation by an
angle y in the (x3 , x0 )-plane. That is, Lorentz invariance is analogous to rotational

invariance in 3-space.
The complex angle y is a mathematical artifact of taking the zeroth component

of the 4-vector to be imaginary number and time a real number. We can make y real

by using a non-Euclidean metric.

We have seen that the generators of space-time translation form a 4-component
set:
 29
Ê H ˆ
P = (P0 , P1 ,P2 , P3 ) = Á i , Px ,Py ,Pz ˜ (7.3)
Ë c ¯

† c is just a units-conversion constant. Quantum mechanically,

where we recall that

Pk p k = pk pk (7.4)

where pk is the eigenvalue of†Pk when the system is in a state given by the eigenvector

pk . Similarly,

† HE =EE (7.5)

Let us work with these eigenvalues—which

† still have not been identified with familiar
physical energy and momentum! But, that's coming up fast now. Write

Ê E ˆ
p = (p0 , p1 , p2 , p3 ) = Á i , px , py , p z ˜ (7.6)
Ë c ¯

The squared length†of the 4-vector

pm pm = p¢m p¢m ≡ -m2 c2 (7.7)

is invariant to rotations in†4-space. The invariant quantity m is called the mass of the
particle. Note that the length of the 4-momentum vector is (in the metric we have
chosen to use)
 30

1/2
( pm pm ) = imc (7.8)

† in this way guarantees the invariance in the important

Defining 4-momentum
result of an earlier section, namely the classical Hamilton equation of motion (4.45),

dPk ∂H
=- (7.9)
dt ∂q k

†
This definition allows us to connect the operator Pk with the operationally defined

momentum pk and the operator H with the operationally defined energy E.

Working with the operationally defined quantities, we can write (using boldface
type for familiar 3-dimensional spatial vectors)

dp • dr = -dEdt (7.10)

Or, in terms of 4-vectors, †

dpm dxm = 0 (7.11)

which is Lorentz invariant. †

Suppose we have a particle of mass m. Let (x', y', z') be the coordinate axes in the
reference frame in which the particle is at rest, |p'| = 0. Then its energy in that
reference frame is

E' = mc2 (7.12)

 31

which is the rest energy. Next let us look at the particle in another reference frame
(x,$y,$z) in which the particle is moving along the z-axis at a constant speed v. Then, from
the Lorentz transformation, the 3-momentum of the particle in that reference frame
will be

Ê b ˆ
pz = gÁ p¢z + E¢˜ = g (0 + bmc) (7.13)
Ë c ¯

†
We can write this in vector form

p = g mv (7.14)

We note that p Æ mv when v << c, So, we have (finally) derived the well-known the

relationship between momentum and velocity. Nowhere previously was it assumed

that p = mv.
The energy of the particle in the same reference frame is

E = g ((E¢ + bp¢z) = gmc 2 (7.15)

Note that, in general, the†velocity of a particle is

pc2 p
v= Æ (7.16)
E m

†
when v << c since, in that case, E = mc2 . We can also show that, for all v,
 32

Ê 2 ˆ1/2
E = Á pc + m2 c 4 ˜ (7.17)
Ë ¯

†
This is a "free particle" since

dp
F= = -—E = 0 (7.18)
dt

†
More generally we can write

E = mc2 + T + V(r) (7.19)

where mc2 is the rest energy.

† The quantity

Ê 2 ˆ 1/2 1
T = Á pc + m2 c4 ˜ - mc2 Æ mv2 (7.20)
Ë ¯ 2

†
when v << c, is the kinetic energy, or energy of motion, and V(r) is the potential energy.
The force on the particle is then

F = -—V (7.21)

We are now in a position to interpret the meaning of c, which was introduced

originally as a simple conversion factor. Suppose we have a particle of zero mass and 3-
momentum of magnitude |p|. Then, the energy of that particle will be
 33

E = |p|c (7.22)

and the speed

pc 2 pc 2
v= Æ =c (7.23)
E pc

†
Thus c is the speed of a zero mass particle, sometimes called "the speed of light." Since c
is the same constant in all references frames, the invariance of the speed of light, one of
the axioms of special relativity, is thus seen to follow from 4-space rotational symmetry.
So we have now shown that the generators of translations along the four axes of
space-time are the components of the 4-momentum, which includes energy in the
zeroth component and 3-momentum in the other components. These have their
familiar connections with the quantities of classical physics. Mass is introduced as a
Lorentz invariant quantity that is proportional to the length of the 4-momentum
vector. The conversion factor c is shown to be, as expected, the Lorentz-invariant speed
of light in a vacuum.

8. Classical Mechanics
Except for specific laws of force for gravity and electromagnetism, all of classical
mechanics can now be inferred from the above discussion. Conservation of energy,
linear momentum, and angular momentum follow from global gauge invariance in
space-time. Newton's first and third laws of motion follow from momentum
conservation. Newton's second law basically defines the force on a body as the time
rate of change of momentum,
 34
dp
F= (8.1)
dt

†
Above we saw that, for the operators P and H,

dP
= -—H (8.2)
dt

†
The classical observables will correspond to the eigenvalues of these and so

dp
= -—E (8.3)
dt

† explicitly on spatial position,

If E = T + V and T does not depend

dp
F- = -—V (8.4)
dt

† more generalized and advanced formulations of classical

as in the previous section. The
mechanics, such as Lagrange's and Hamilton's equations of motion, can be now
developed in the usual way.

9. Electromagnetism
In the following sections we will switch to the conventions used in slightly more
advanced physics so that the resulting equations agree with the textbooks at that level.
We have already seen that h and c are arbitrary conversion factors, so we will work in
units where h = c = 1. Furthermore, we will use a non-Euclidean (but still geometrically

†
†
 35
flat) metric in defining our 4-vectors:

Ê1 0 0 0 ˆ
Á ˜
n Á 0 -1 0 0˜
hm = (9.1)
Á 0 0 -1 0 ˜
Á ˜
Ë 0 0 0 -1¯

where the space-time position

† 4-vector is x = (t, x, y, z), where we re-use x, the
momentum 4-vector is p$=$(E,$px ,$py ,$pz), and

2
pm hnm pn = E 2 - p = m2 (9.2)

This choice of metric has†the advantages of enabling us to directly identify the mass
with the invariant length of the 4-momentum vector and eliminating the need for
imaginary zeroth components.
In quantum mechanics, the state of a free particle is an eigenstate of energy and
momentum. Consider the 4-momentum eigenvalue equation for a spinless particle
(spin can be included, but this is sufficient for present purposes)

i∂m f = pm f (9.3)

† ∂f
where we now use the convention ∂m f ≡ . The quantity f is the eigenfunction
∂xm

f (x) = x pm and can be thought of as having two abstract dimensions, its real and
†
imaginary parts. If we rotate the axis in this space by an angle q we have the gauge

† transformation,
 36

f' =exp(iq)f (9.4)

The eigenvalue equation is unchanged, provided that q is independent of the space-time

position x. This is the type of gauge invariance we have already considered, what we
call global gauge invariance. The generator of the transformation, q, is conserved.

Below we will identify q as the negative of the charge of the particle.

Now suppose that q depends on x. In this case, we do a local gauge

transformation and

[ ( )]
∂m f ¢ = exp[iq (x)] ∂m + i ∂m q f (9.5)

and the eigenvalue equation

† is not invariant to this operation. Let us define a new
operator, the covariant derivative,

Dm = ∂m + iq Am (9.6)

where q is a constant and Am transforms as

A'm = Am + ∂m x(x) (9.7)

where q(x) = -q x(x). Then,

D'mf' = (∂m + iq A'm) f' = [∂m + iq Am + iq( ∂m x)] exp(iq)f (9.8)

= exp(iq) [∂m + iq Am]f + exp(iq) f [i(∂m q) - i(∂m q)]

 37

= exp(iq) [∂m + iq Am]f

= exp(iq)Dmf

Recall the the operator Pm associated with the relativistic 4-momentum is

Pm = -i∂m (9.9)

Let us define, analogously,

Pm = -iDm (9.10)

Writing

Pm = Pm + qAm (9.11)

we see that this operator Pm is precisely the canonical 4-momentum in classical

mechanics for a particle of charge q interacting with an electromagnetic field described

by the 4-vector potential Am = (Ao, A), where Ao = V/c in terms of the scalar potential V

and A is the 3-vector potential. We will further justify this connection below. As already
mentioned, q(x) = -q x(x) and thus q is conserved when x(x) is a constant. Also, note that

for neutral particles q = 0 and no fields need to be introduced.

In quantum mechanics, the canonical momentum must be used in place of the
mechanical momentum in the presence of an electromagnetic field. For example, the
Schrödinger equation for a non-relativistic particle of mass m and charge q in an
electromagnetic field described by the 3-vector potential A and scalar potential V is
 38

Ê 2 ˆ Ê 2 ˆ ∂y
Á P - qA + qV)˜y = Á -ih— - qA + qV)˜y = ih (9.12)
Ë ¯ Ë ¯ ∂t

†
In quantum field theory, the basic quantity from which calculations proceed is
the Lagrangian density. The Klein-Gordon Lagrangian density for a spinless particle of
mass m is

L = - 12 ∂m ∂m f + 21 m2 f 2 (9.13)

This is not locally gauge invariant.

† However, it becomes so if we write it

L = - 12 DmDm f + 21 m2 f 2 (9.14)

The corresponding Klein-Gordon

† equation, the relativistic analogue of the Schrödinger
equation for spinless particles, becomes

DmDm f + m2 f = 0 (9.15)

Spin 1/2 particles of†mass m are described by the Dirac Lagrangian which
similarly can be made gauge invariant by writing it, using conventional notation,

L = iy g m Dmy - my y (9.16)

The corresponding Dirac †

equation
 39

ig mDmy - my = 0 (9.17)

also is gauge invariant. (Note:

† while I have not derived these equations, no additional
physical assumptions are required in their derivation).
A spin 1 particle of mass mA is described by the Procca Lagrangian

1 mn
L=- F Fmn + m2A Am Am (9.18)
16p

where †

Fmn = ∂mAn - ∂nAm (9.19)

The first term in L is gauge invariant while the second is not unless we set mA = 0. This

leads to the deeply important result that particles with spin 1 whose Lagrangians are
locally gauge invariant are necessarily massless. The photon is one such particle.
However, other spin 1 fundamental particles exist with nonzero masses. These masses
result from broken gauge symmetry, as we will briefly discuss below.
In any case, the existence of a vector field Am associated with a massless spin 1

particle is implied by the assumption of local gauge invariance. It is a field introduced to

maintain local gauge invariance. That field can be identified with the classical
electromagnetic fields E and B and the particle with the photon. That is, the photon is
the quantum of the field Am, which itself is associated with the classical 4-vector

electromagnetic potential.
To see the classical connection, note that
 40

A'k = Ak + ∂k x(x) (9.20)

where k = 1, 2, 3, or, in 3-vector notation

A' = A + —x (9.21)

It follows that the 3-vector

B' = — x A' = — x A - — x —x = — x A = B (9.22)

is locally gauge invariant. Furthermore,

—•B = —•(— x A) = 0 (9.23)

Thus, B may be interpreted as the familiar classical magnetic field 3-vector; the above
equation is Gauss's law of magnetism, one of Maxwell's equations. The zeroth
component of the 4-vector potential,

∂x
A¢o = Ao + (9.24)
∂xo

can be written †

∂x
V¢ = V - (9.25)
∂t

†
 41
which implies that the 3-vector

∂A ¢
E ¢ = -—V¢ -
∂t

∂x ∂A ∂—x
= -—V + — - - (9.26)
† ∂t ∂t ∂t

∂A
= -—V - =E
† ∂t

† Furthermore,
is also locally gauge invariant.

Ê ∂A ˆ
— ¥ ÁE + ˜ = -— ¥ —V = 0 (9.27)
Ë ∂t ¯

so †

∂(— ¥ A) ∂B
—¥E=- =- (9.28)
∂t ∂t

† induction, another of Maxwell's equations, with E interpreted

which is Faraday's law of
as the classical electric field.
Summarizing, we have found that the motion of a free charged particle is not
invariant under a local gauge transformation. However, we can make it invariant by
adding a term to the canonical momentum that corresponds to the 4-vector potential of
the electromagnetic field. Thus the electromagnetic force can be thought of as a
fictitious force that is introduced to preserve local gauge symmetry. Conservation of
charge follows from global gauge symmetry.
 42

10. The Subnuclear Forces

The gauge transformation just described corresponds to a rotation in the abstract space
of the 4-momentum eigenstate, which is the state of any particle of constant
momentum. Here the transformation operator

U = exp(iq) (10.1)

can be trivially thought of as a1x1 matrix. The set of all such unitary matrices comprises
the transformation group U(1). The generators of the transformation, q, form a set of

1x1 matrices that, clearly, mutually commute. Whenever the generators of a

transformation group commute, that group is termed abelian. Electromagnetism is thus
an abelian gauge theory.
Recall from our discussion of angular momentum that the unitary operator

Ê1 ˆ
U = expÁ is • q˜ (10.2)
Ë2 ¯

† state vectors. In this case U is represented by a 2x2 matrix.

operates in the space of spin
The set of all such matrices comprises the transformation group SU(2), where the prefix
S specifies that the matrices of the group are unimodular, that is, have unit determinant.
This follows from the fact that, for any matrix U,

U = exp(iA) (10.3)

we have
 43
detU = exp(TrA) (10.4)

Since the Pauli matrices are traceless, detU = 1.

Following a procedure similar to what was done above for U(1), let us write

È 1 ˘
U = expÍ- igt ¥ q˙ (10.5)
Î 2 ˚

† of t form a set of matrices identical to the Pauli spin

where the three components

matrices and we use a different symbol just to avoid confusion with spin. While the spin
S = s/2 is a vector in familiar 3-dimensional space, t is a 3-vector in some more abstract

space we will call isospin space. The 3-vector T = t/2 is called the isospin or isotopic spin.

Global gauge invariance under SU(2) implies conservation of isospin. The quarks and
leptons of the standard model have T = 1/2. The quantity g is a constant analogous to
the electric charge that measures the strength of the interaction.
Once again it is important not to confuse isospin space with the 2-dimensional
subspace of the state vectors on which U operates. When the isospin space 3-vector x(x)

depends on the space-time (yet another space) position 4-vector x we once more have a
local gauge transformation. The generators being like angular momenta do not
mutually commute, so the transformation group is non-abelian. This type of non-abelian
gauge theory is called a Yang-Mills theory.
Let us attempt to make this clearer by rewriting U with indices rather than
boldface vector notion:

È 1 ˘
U = expÍ- igt kx k (x)˙ (10.6)
Î 2 ˚

†
 44
where the repeated Latin index k is understood as summed from 1 to 3.
Encouraged by our success in obtaining the electromagnetic force from local
U(1) gauge symmetry, let us see what we can get from local SU(2) symmetry.
Following the U(1) lead, we define a covariant derivative

1
D m = ∂m + igt k Wmk (10.7)
2

†
where W mk are three 4-vector potentials analogous to to the electromagnetic 4-vector

potential Am. As before, the introduction of the fields W mk maintains local gauge
†
invariance. Or, we can say that local gauge invariance implies the presence of three 4-

vector potentials W mk . In the standard model,†these are interpreted as the fields of the

weak interaction.
In†quantum field theory, a particle is associated with every field, the so-called
quantum of the field. The spin and parity of the particle, JP , is determined by the
transformation properties of the field. The quantum of a scalar field has JP = 0+ ; a
vector field has JP = 1. For the electromagnetic field described by the potential Am, the

quantum is the photon. Since Am is a vector field, the photon has spin 1. It is a vector

gauge boson.

Similarly, the weak fields W mk will have three spin 1 particles as their

quanta—three vector gauge bosons W-, Wo, and W+ , where the superscripts specify the
electric charges of the particles.
† These can also be viewed as the three eigenstates of a
particle with isospin T = 1.
If the U(1) symmetry of electromagnetism and the SU(2) symmetry of the weak
interaction were perfect, we would see the photon and three W bosons above.
 45
However, these symmetries are broken at the "low" energies at which most physical
observations are made, including those at the current highest energy particle
accelerators. This symmetry breaking leads to a mixing of the electromagnetic and
weak forces. Here, briefly, is how this comes about in what is called unified electroweak
theory.
The covariant derivative for electroweak theory (assumed, not derived) is
written

Y t
Dm = ∂m + ig1 Bm + ig2 k W mk (10.8)
2 2

†called Bm and the constant g1 replaces the electric charge in that

where the U(1) field is

term. The quantity Y is a constant called the hypercharge generator that can take on
different values in different applications, a detail that need not concern us here. The
SU(2) term includes a constant g2 , the vector T = t/2, or isospin, and the vector field

W mk ,$k$= 1,"2,"3.

Neither B nor Wo , the quanta of the fields Bm and W mo , appear in experiments at

†
current accelerator energies. Instead, the particles that do appear are the photon and Z,
† o . These together with the W+ and W-,
whose fields Am and Zm are mixtures of Bm and W m

the quanta of the fields W mk ,"k$=$1,"2, constitute the vector gauge bosons of the
†
electroweak sector of the standard model. Their mixing is also like a rotation, gauge
symmetry being
† broken in this case,
 46
Ê Am ˆ Ê cosq W sin qW ˆÊ Bm ˆ
Á ˜ =Á ˜Á ˜ (10.9)
Ë Zm ¯ Ë - sin q W cosq W ¯ÁË Wmo ˜¯

where the rotation angle qw is called the Weinberg (or weak) mixing angle. This parameter
†
is not determined by the theory and must be found from experiment. The current value
of sin2 qw = 0.23115. The constants that determine the strength of the interaction are

g1 = e/cosqw g2 = e/sinqw (10.10)

where e is the unit electric charge.

As we have seen, the masses of gauge bosons are fundamentally zero. While the

photon is massless, the W± and Z bosons have large masses. These masses are shown to

arise from another symmetry-breaking process called the Higgs mechanism. The
symmetry-breaking is apparently spontaneous, that is, not determined by any known
deeper physical principle. Spontaneous symmetry breaking describes a situation, like
the ferromagnet, where the fundamental laws are symmetric and obeyed at higher
energy, but the lowest energy state of the system breaks the symmetry.
Moving beyond the weak interactions and SU(2), we have the strong
interactions and SU(3). In general, for SU(n) there are n2 - 1 dimensions in the subspace.
Let us add the new term to the previous ones that included the electroweak forces

Y t l
Dm = ∂m + ig1 Bm + ig2 k W mk + ig 3 a Gam (10.11)
2 2 2

† 3 for the four dimensions of space-time, the repeated index k is

where m = 0, 1, 2,

summed from 1 to 3 in the SU(2) term and the repeated index a is summed from 1 to 8
in the SU(3) term. The la are eight traceless 3x3 matrices analogous to the three Pauli
 47
2x2 isospin matrices tk, and the Gam are eight spin 1 fields analogous to the singlet field

Bm and the triplet field W mk . of the electroweak interaction. The gauge bosons in this case
†
are eight gluons. The symmetry is not broken, so they are massless. Global gauge
invariance under
† SU(3) implies the conservation of another quantity called color charge.
While there is much more to the standard model, this should suffice to illustrate
its basis in gauge symmetry and the importance of spontaneous broken symmetry.

11. General Relativity

General relativity can also be cast in the form of a gauge theory,[2] but to do so would
take us well beyond the mathematical scope of this paper which we have tried to limit
to the undergraduate level. However, we can outline how the principle of general
covariance leads to general relativity. Typical treatments emphasize the role of two
other principles, Mach's principle and the principle of equivalence. However, while Einstein
acknowledged Mach's influence on his thinking, it is not clear that Mach's principle
plays any significant part in deriving general relativity, especially since the principle
itself is ill-defined. The principle of equivalence between gravity and acceleration is
usually given great prominence, but that can also be seen as a consequence of the
principle of covariance.
Consider the equation of motion for a freely falling body in terms of a
coordinate system y = (y 0 , y1, y 2 ,y 3 ), where y o = ict , falling along with the body. Since

dy = 0,

† †

† d 2y
=0 (11.1)
dt 2

Also, †
 48

d 2y o
=0 (11.2)
dt 2

Furthermore, †

1 2
(dt )2 = (dt)2 - 2
2 dy = (dt) (11.3)
c

†
Let us work in units where c = 1. We can write the above in 4-vector form,

d 2y a
=0 (11.4)
dt 2

† of motion for a body not acted on by any [Link]

This the 4-dimensional equation
expresses the fact that a freely falling body experiences no external force.

Next, let us consider a coordinate system xa fixed to a second body such as the

earth, or any system that may be in relative acceleration. The equation of motion can be
transformed to that coordinate system as follows:

d Ê ∂y r dx m ˆ
Á ˜=0 (11.5)
dt ÁË ∂x m dt ˜¯

from which, after some algebra,[3] we find

†

d 2x l dx m dxn
2 + Glmn =0 (11.6)
dt dt dt

†
 49

where

2
∂x l ∂ y r
Glmn = (11.7)
∂y r ∂x m ∂xn

is called the affine connection. An observer on earth witnesses a body accelerating

†
toward the earth and interprets it as the action of a "gravitational force." The principle of
equivalence thus merely defines gravity as the invisible force that produces the
observed acceleration. Glmn is a field that describes that force. Although Glmn has three

indices, it is not a tensor since it is not Lorentz invariant.

We can obtain the Newtonian picture in the limit of low speeds, dxk/dt ≈ 0. In

this case, dt = dt, d2 xo/dt2 = 0, and

d 2x k
= Gkoo = gk (11.8)
dt 2

for k = 1,2,3, where g = (g1 , g2 , g3 ) is the Newtonian field vector ("acceleration due to
†
gravity"). Thus, the Gkoo elements of the affine connection are just the Newtonian

gravitational field components in the limit of low speeds. Additional elements then are
needed to describe gravity at speeds near the speed of light.
The Newtonian field vector for any distribution of mass can be obtained from
the gravitational potential f which is in general a solution of Poisson's equation

— 2f = 4pGr (11.9)

where r is the mass density and

† g = -—f. For example, suppose that we have a point
 50
mass m so that r(r) = md(r). Then,

Gm
f (r) = - (11.11)
r

the familiar Newtonian result.†

While the modified equation of motion (11.6) contains relativistic effects of
gravity, it is not covariant. It has a different form in the two reference frames. Einstein
sought to find equations to describe gravity that were covariant. He started with the
Poisson equation (11.9) above, which is noncovariant since r is the mass or energy

density and energy is the zeroth component of a 4-vector. Let us search for a covariant
quantity to replace density.
Suppose we have a dust cloud in which all the dust particles are moving
slowly, that is, with v << c, in some reference fame. Let the energy density in that
frame be ro . Let Eo be the rest energy of each particle (c = 1) and no be the number per

unit volume. Then,

r o = Eo no (11.12)

In some other reference frame the energy density will be

r = E n = g Eo g no = g2 ro (11.13)

where g is the Lorentz factor, E = g Eo, and n = g no. To see the latter, note that n =

dN/dV, where dN is the number of particles in the volume element dV, dNo = dN, and

dV = dVo/g from Fitzgerald-Lorentz contraction.

Note that r is not simply the component of a 4-vector because of the factor g2 .
 51
Rather it must be made part of a second-rank tensor. We can write

Tmn = r ov m v n (11.14)

where vm is the 4-velocity of the

† cloud. Then, since vo = dt/dt = g,

Too = r ov o v o = g 2 r o = r (11.15)

Tmn is the energy-momentum

† tensor, or stress-energy tensor. The other components

comprise energy and momentum flows in various directions: Toi is the energy flow per

unit area in the i-direction, that is, a heat flow; Ti i is the flow of momentum component i

per unit area in their direction, the pressure across the i plane; Ti j is the flow of

momentum component i per unit area in the j-direction, the viscous drag across the j-
plane; Tio is the density of the i component of momentum.

Einstein thus wrote, as the covariant form of Poisson's equation,

Gmn = -8pGTmn (11.16)

where G is Newton's constant

† and the factor is chosen so that we get Poisson's equation
in the Newtonian limit. Since the energy-momentum tensor Tmn is covariant, the

quantity Gmn is also a covariant tensor field.

By associating Gmn with the energy-momentum tensor, Einstein was using, at

most, a very weak form of Mach's principle that was much earlier proposed by Leibniz.
Leibniz had objected to Newton's notion of an absolute space with respect to which
bodies accelerate and argued that at least another body must be present for space and
time concepts to be useful.[4]
 52
In what has become the standard model of general relativity, Einstein related
Gmn to the curvature of space-time in a non-Euclidean geometry. In non-Euclidian

geometry, the proper distance between two points in space-time is

(Ds)2 = Dx m gmn Dxn (11.17)

where gmn is the metric tensor.

†

Einstein assumed that Gmn is a function of gmn and its first and second derivatives.

In its usual form, Einstein's field equation is given as

1
Rmn - gmn R + Lg mn = -8pGTmn (11.18)
2

where Rmn and R are contractions of the rank four Riemann curvature tensor. To see the
†
explicit forms of these quantities, consult any textbook on general relativity.
The quantity L is the infamous cosmological constant. It is often reported in the

media and in many books on cosmology that the cosmological constant was a "fudge
factor" Einstein introduced to make things come out the way he wanted. Perhaps that
was his motivation, but the fact is that unless one makes further assumptions, a
cosmological constant is required by Einstein's equations of general relativity and
should be kept in the equations until some principle is found that shows it to be zero.[5]
For many years the measurements of the cosmological constant gave zero within
measuring errors, but in the past two decades Einstein's fudge factor has resurfaced
again in cosmology.

12. Conclusions
The sophisticated reader who at least has glanced at the equations in this paper will
 53
recognize them as very familiar. Almost every one can be found in standard textbooks.
What has been attempted here is to show that those equations do not follow from very
unique or very surprising physical properties of the universe. Rather, they arise from
the very simple notion that whatever mathematical "laws" you write down to describe
measurements, your equations cannot depend on the origin or direction of the
coordinate systems you define in the space of those measurements or the space of the
functions used to describe those laws. That is, they cannot reflect any privileged point of
view. Except for the complexities that result from spontaneously broken symmetries,
the laws of physics may be the way they are because they cannot be any other way. Or,
at least they may have come about the simplest way possible. Table 11.1 summarizes
these conclusions.

Acknowledgements
Special thanks to Brent Meeker and other members of the avoid-L discussion group for
their help with this manuscript.

Table 1. The laws and other basic ideas of physics and their origin.
Law/idea of Physics Origin
Conservation of momentum Space translation symmetry
Conservation of angular momentum Space rotation symmetry
Conservation of energy
(First law of thermodynamics) Time translation symmetry
Newton's 1st Law of Motion Conservation of momentum
(space translation symmetry)
Newton's 2nd Law of Motion Definition of force
Newton's 3rd Law of Motion Conservation of momentum
(space translation symmetry)
Second law of thermodynamics Statistical definition of the arrow of time
Special relativity Space-time rotation symmetry
Invariance of speed of light Space-time rotation symmetry
 54
General relativity Principle of covariance
Quantum time evolution
(time-dependent Schrödinger equation) Global gauge invariance
Quantum operator differential forms Global gauge invariance
Quantum operator commutation rules Global gauge invariance
Quantization of action Global gauge invariance
Quantization rules for angular momenta Global gauge invariance
Maxwell's equations of electromagnetism Local gauge invariance under U(1)
Quantum Lagrangians for particles in
presence of electromagnetic field Local gauge invariance under U(1)
Conservation of electric charge Global gauge invariance under U(1)
Masslessness of photon Local gauge invariance under U(1)
Conservation of weak isospin Global gauge invariance under SU(2)
Electroweak Lagrangian Mixing of U(1) and S(2) local gauge
symmetries (spontaneous symmetry
breaking)
Conservation of color charge Global gauge invariance under SU(3)
Strong interaction Lagrangian Local gauge invariance under SU(3)
Masslessness of gluon Local gauge invariance under SU(3)
Structure of the vacuum (Higgs particles) Spontaneous symmetry breaking
Doublet structure of quarks and leptons Conservation of weak isospin
(global gauge invariance under SU(2))
Masses of particles Higgs mechanism
(spontaneous symmetry breaking)
 55
References
1. E. Noether, Nachr. d. König. Gesellsch. d. Wiss. zu Göttingen, Math-phys. Klasse,
235 (1918), p. 7; English translation M. A. Travel, Transport Theory and Statistical
Physics 1(3), 183 (1971). See also Nina Byers, in Proceedings of the Israel
Mathematical Conference 12 (1999). Online at
[Link] This
contains links to Noether's original paper including an English translation.

2. F. W. Hehl, J. D. McCrea, E. W. Mielke and Y. Ne'eman, Phys. Rep. 258, 1 (1995).

3. Steven Weinberg, Graviation and Cosmology: Principles and Applications of the

General Theory of Relativity (New York London Sydney Toronto: John Wiley &
Son, 1972), p. 102.

4. Jennifer Trusted, Physics and Metaphysics: Theories of Space and Time, Paperback
edition (London and New York: Routledge, 1994), p. 90.

5. Weinberg, Gravitation and Cosmology, pp. 613-6.