GAUSS - RIEMANN - EINSTEIN

or
What I Did on My Summer Vacation
Michael P. Windham
February 25, 2008
Preface
For the last forty years most every spring my fancy turned to dierential
geometry, among other things. I have always wanted to understand it and
always struggled. I also wanted to know something about general relativity.
Well, I think I have gotten a grip on the basics and that is what these notes
are about.
I have organized the presentation around the key gures in the devel-
opment of the subject, at least in the development of my understanding of
the subject. They are Karl Friedrich Gauss, Bernhard Riemann, and Albert
Einstein.
In 1827 Gauss produced a paper that fundamentally explained the ge-
ometry of surfaces in space. In 1854, Riemann generalized Gauss ideas to
something he called manifolds. In 1912 Einstein used manifolds to describe
gravity. These contributions will provide the organization for these notes.
There is a lot to say here and I will not say it all. I hope only to follow a
thread that will get me where I want to go with as few digressions as possible.
Where I want to go is to general relativity. I will not be very complete in
deriving or proving things. My plan is to give you a feel for the subject, not
a rigorous development.
But, rst a modest beginning with curves in space.
Michael P. Windham
2
Chapter 1
Curves
This little journey begins with a discussion of curves in space and how to
investigate their geometry. The rst step is to describe the curve in a quan-
titative way that allows you to apply the massive machinery of calculus to
study it. The appropriate quantitative description is a parameterization. The
curve is the image of a function : R R
3
. In what follows I will implicitly
assume any functions are suciently nice for what I say to be true. Usu-
ally this means the functions have as many derivatives as needed and the
derivatives are not zero when it would be embarrassing for them to be so.
A parameterization (t) = (x(t), y(t), z(t)), describes the points on the
curve as a function of the parameter t. Think of as drawing the curve. The
study of the curve begins with the derivative of ,

(t) = x

(t), y

(t), z

(t),
where I am using x, y, z to denote a vector as opposed to (x, y, z) which
denotes a point. A point describes a location and a vector, which describes a
magnitude and direction, is the natural descriptor for change. In particular,
as the curve is drawn, the speed with which it is drawn, how fast the pen
is moving, and the direction it is being drawn are naturally described by a
vector, and these are exactly what

(t) quanties. I may appear to be saying

that the parameter t is time. That is not necessarily true, but it is seldom
harmful to think so.
3
4 CHAPTER 1. CURVES
t
x
y
z
(t)

"
(t)
(t)
Perhaps the most basic geometric characteristic of a curve is its length.
The derivative vector is the key to calculating length since, according to just
about any rst year calculus book
s(t) =
_
t
a
|

(u)| du =
_
t
a

_
_
dx
du
_
2
+
_
dy
du
_
2
+
_
dz
du
_
2
du (1.1)
is the distance along the curve from (a) to (t). The length of the derivative
vector is just the derivative of s with respect to t, so it measures how quickly
the curve is being drawn relative to t.
If my goal was to study in detail the geometry of curves, the next step
would be to look at the second derivative

(t), since it describes how the

drawing process changes. To do so, would lead us to curvature and torsion
and to a complete characterization of curves in terms of them. As fascinating
as this is, we will not go there. Let me just point out that if

(t) = 0 for
all t, then the curve is a straight line. The converse is true for so called
constant speed curves, namely ds/dt is constant.
If we lived in the 17th century we would have no problem thinking of
an innitesimal piece of length along the curve, the length of a point.
Moreover, it would be natural to think of a point as an innitesimal box
with sides parallel to the axes of length dx, dy, and dz. The length of the
point itself would simply be the innitesimal length of the diagonal of the
box, namely ds =

dx
2
+ dy
2
+ dz
2
. This heuristic approach is frowned
upon these days, but is wonderful to use to think about geometry and even
physics, at least in the privacy of your own home.
It is also comfortable and even useful to think of a particle traveling
through space and (t) is its location at time t. The derivative

(t) would
then be the velocity of the particle, and

(t) its acceleration. This interpre-

tation will become more important when we leap into physics.
Chapter 2
GAUSS - Surfaces
Gauss takes the stage now because of his 1827 paper Disquisitiones generales
circa supercies curvas (General investigations of curved surfaces). Michael
Spivak has called this paper the single most important work in the history
of dierential geometry.
The problem the paper addresses is to study the geometry of surfaces in
space. As with curves, a good way to represent them quantitatively is with
a parameterization, a function A : R
2
R
3
whose image is the surface, or
if you like A(x
1
, x
2
) = (x(x
1
, x
2
), y(x
1
, x
2
), z(x
1
, x
2
)) draws the surface using
two parameters x
1
and x
2
.
The geometric properties of the surface are quantied by the motion of its
tangent planes and their normal vectors as you move about the surface. At
a point (x, y, z) = A(x
1
, x
2
) the tangent plane is the plane through (x, y, z)
parallel to the vectors
A
x
i
=
_
x
x
i
,
y
x
i
,
z
x
i
_
for i = 1 and 2. These two vectors are tangent to the surface because they are
the tangent vectors of curves in the surface through (x, y, z) drawn by letting
one of the parameters vary while the other is constant. The normal to the
surface at the point is the cross product of the two tangent vectors. I will be
assuming that the parameterization is suciently nice that the two tangent
vectors are linearly independent and therefore the normal vector does not
vanish. I will not go into details, but, for example, curvature of the surface
can be quantied by derivatives of the normal vector, just as curvature of
curves is quantied in terms of

(t).
5
6 CHAPTER 2. GAUSS - SURFACES
I do want to discuss the notions of intrinsic and extrinsic properties
of the surface. To put it quaintly, intrinsic properties are those that little
two dimensional people that lived in the surface could deduce about their
world and extrinsic properties are those that you have to be outside the
surface in space to see. The normal vector is extrinsic, but amazingly the
tangent plane is intrinsic. To see how this is possible, you should think of the
domain of the parameterization as a map of the surface. Each point on the
map has coordinates (x
1
, x
2
) that correspond to the point A(x
1
, x
2
) in the
surface. Presumably the little people living in the surface would eventually
produce such a map of their world. They would also eventually recognize
the importance of drawing vectors on their map to describe directions and
develop linear algebra to entertain themselves. Each vector v = v
1
, v
2

drawn with its tail at a point (x

1
, x
2
) on the map corresponds to a vector in
the tangent plane at the point A(x
1
, x
2
) in the surface, namely v
1
A
x
1
+v
2
A
x
2
.
Therefore, the little people could not see the tangent plane to their world,
but they could do the same calculations with vectors on their map that we
would with the tangent vectors to their surface in space. They would get the
same information from their results, just as we get information about space
itself with our vector calculations in it. More on this shortly, but rst let us
look at something that is easier to believe is intrinsic, length.
A curve in the surface is a curve in space. The length of a curve param-
eterized by : R R
3
whose image is in the surface can be calculated in
the usual way by (1.1). However, the little people would see this curve and
draw a corresponding curve on their map. Moreover the curve on the map
would be parameterized by a : R R
2
satisfying (t) = A(a(t)). The tan-
gent vectors to the curve on the surface would be related to corresponding
tangent vectors a

(t) = a

1
(t), a

2
(t) to the curve on the map by

(t) = a

1
(t)A
x
1
(a(t)) + a

2
(t)A
x
2
(a(t))
using the chain rule. Moreover, letting denote the usual euclidean inner
product, we have
1
|

|
2
=

A
x
i
A
x
j
a

i
a

j
The little people can then calculate the length of their tangent vector by
|a

|
2
=

g
ij
a

i
a

j
1
I will not put any indices on summations. You can assume that any indices that do
not appear outside the sum are summed over.
7
with g
ij
(x
1
, x
2
) = A
x
i
(x
1
, x
2
) A
x
j
(x
1
, x
2
) and integrate the length of their
tangent vector to get the length of the curve. Of course, they are not aware
of the A
x
i
s but being clever they would no doubt eventually derive the g
ij
s.
Their linear algebra would have an inner product of vectors given by
v w =

g
ij
v
i
w
j
In fact, they could have innitely many inner products, one for each point
on their map, since the g
ij
s are functions of the location on the map. They
could measure length and angles between vectors, but the measurements
would depend on where they drew the vectors. This situation may appear
disturbing, but we are faced with it all the time when we measure distances
between points on a at map of our world. After all, Greenland appears on
a map of the world to be much larger than it really is on the world itself.
The 17th century folk would simply say
ds
2
=

g
ij
dx
i
dx
j
is the (innitesimal) line element for the surface.
The next question to consider is what curves do the little people think
are straight lines. I already mentioned that for a curve in space, if

is
zero along the curve then it is a straight line, that is, if the curve is drawn
at a constant rate, without turning it is straight. If the curve is drawn in a
surface, a tangent vector to the curve is tangent to the surface, so that the
little people can perceive its eect. The vector

(t) may not be tangent to

the surface, so that the little people cannot see it - well, not all of it anyway.
They can see the tangential component of

(t). In particular, again using

the chain rule,

i
A
x
i
+ a

i
a

j
A
x
i
x
j
=

(a

k
+

k
ij
a

i
a

j
)A
x
k
+ the normal component
where the
k
ij
s are obtained from the tangential components of the A
x
i
x
j
s
and like the g
ij
s are functions of (x
1
, x
2
). Therefore, the little people would
learn that for the curves on their maps
a

=
_
a

1
+

1
ij
a

i
a

j
, a

2
+

2
ij
a

i
a

j
_
for some functions
k
ij
they would deduce by experience. They would think
a line was straight, if a

were zero, that is, the components would be

8 CHAPTER 2. GAUSS - SURFACES
solutions to the dierential equations
a

k
+

k
ij
a

i
a

j
= 0 (2.1)
Such curves are called geodesics, they may curve in space, but not in the
surface.
2
I have just been and probably will continue to be a little sloppy by
calling curves on the map, a(t), and curves on the surface, (t) = A(a(t)),
geodesics, but you will get used to it.
I should point out that the shortest curve between two points is a geodesic.
This fact takes a little calculus of variations to verify, so I will say no more. I
should also point out that geodesics on the map may not look straight, that
is, you may not be able to draw them with a ruler.
How about an example, one we can relate to, the unit sphere. The unit
sphere centered at the origin can be parameterized by longitude and lat-
itude , namely A(, ) = (cos cos , cos sin , sin) draws the sphere.

A
/2
-
/2
We have the following
ds
2
= cos
2
d
2
+ d
2
and

1
11
=
1
22
=
2
12
=
2
21
=
2
22
= 0

1
12
=
1
21
= tan

2
11
= cos sin
Okay, these coecients are not easy to compute. You can use the formula
that appears below or you can do it the way I did, using mathematica (see
Gray, 1999).
2
With a little work and trickery one can show that any curve satisfying these equations
is a constant speed curve, that is

g
ij
a

i
a

j
is constant. By rescaling the parameter you
can, without loss of generality, assume the constant is one.
9
The geodesics can be obtained by solving a

= 0 or

2 tan

= 0

+ cos sin
2
= 0
to obtain great circles, curves on the sphere obtained by intersecting the
sphere with a plane through the origin. Actually, the equations are quite
dicult to solve, but at least it is clear that a(t) = (t, 0) (the equator) and
a(t) = (c, t) for any constant c (lines of longitude) are solutions. You might
nd it interesting to show that these are the only geodesics that can be drawn
on the map with a ruler.
3
On the other hand, there are many others. For
example, a(t) =
_
sin
1
(sint/

1 + cos
2
t), sin
1
(sint/

2)
_
draws a geodesic
going through (0, 0) in the direction 1, 1/

2. Check it out! In fact, pick any

point and any initial direction, then standard existence results in dierential
equations say that there is a geodesic that goes through the point you picked
in the direction you picked. Finally, that the equator is a geodesic makes it
reasonable to believe that any great circle would be, since it could be moved
around to coincide with the equator without distorting distances.
Before leaving surfaces, I want make one technical point that will have
philosophical implications later on. The
k
ij
s are called Christoel symbols
and the g
ij
s are the metric coecients. Since the Christoel symbols arose
from orthogonally projecting onto the tangent plane it would seem reasonable
that they would be related to the metric coecients, which, in fact they are.

k
ij
=
1
2

g
lk
_
g
lj
x
i

g
ij
x
l
+
g
il
x
j
_
where g
ij
s are the entries of the inverse of the matrix whose entries are the
g
ij
s. The relationship is not pretty, but all you need to observe is that the
Christoel symbols can be calculated from the metric coecients and their
rst derivatives.
3
If we were the little people and the sphere were the earth, this mapping would be the
plate carree projection. There is a map called the gnomonic projection which allows you
draw geodesics with a ruler.
10 CHAPTER 2. GAUSS - SURFACES
Chapter 3
RIEMANN - Manifolds
By 1854 Bernhard Riemann had nished his doctoral dissertation in complex
variables and his Habilitationsschrift on trigonometric series at the University
of Gottingen and needed a job. He applied for the position of Privatdocent at
the University (a lecturer who received no salary, but was merely forwarded
fees paid by those students who chose to attend his classes - scary!). He was
required to give a probationary inaugural lecture on a topic chosen by the
faculty from a list of three that he was to propose. His rst two choices were
the topics in his dissertation and Habilitationsschrift, his third was something
he called the foundations of geometry. Tradition had it that one of the two
he was known for would be chosen. Gauss chose the third. The lecture,
titled

Uber die Hypothesen welche der Geometrie zu Grunde liegen (On the
Hypotheses that lie at the Foundations of Geometry) became the foundation
of modern dierential geometry and the tool that Einstein needed to deal
with gravity.
Now, revolutionizing mathematics and physics was not what Riemann
intended. He merely wanted to deal with a problem that had been around a
while, but was a hot topic at the time, namely - is there something besides
Euclid? Euclidean geometry was almost sacred. By the middle of the 19th
century people began to wonder about its exalted position. The problem
was the parallel postulate: given a line and a point not on the line there
is exactly one line through the given point parallel to the given line. This
axiom was the most complicated of Euclids postulates and people began
to wonder if it was really necessary to assume it. Proofs of the parallel
postulate abounded. Gauss knew that the postulate was necessary, because
the geometry of the sphere did not satisfy the parallel postulate. There are
11
12 CHAPTER 3. RIEMANN - MANIFOLDS
no parallel lines on a sphere. The analogs of straight lines are great circles,
which always intersect. In 1829, Bolyai and Lobachevsky had independently
constructed apparently consistent geometrical theories that assumed there
could be more than one line through a given point parallel to a given line.
Such a geometry can be realized on a surface that looks like a saddle. Gauss
had this result even earlier, but told only a few people. He may have directly
or indirectly inspired the work of Bolyai and Lobachevsky.
OK, so the Euclidean plane was not the only two-dimensional geometry,
it was at but there were others that were not. Of course, it was clear to
everyone that space is at, so Euclid was still good where it counts. Isaac
Newton, for one said so. But, how can you be sure? Riemann in his lecture
suggested a procedure for being sure and in a later paper worked out the
details. I will describe essentially what Riemann proposed.
Riemann says we can view any space as an n-fold extended manifold
where n is the number of independent directions we can go. We can lay out a
coordinate system (x
1
, . . . , x
n
) in our space to quantify location. This notion
has been made more precise with the modern denition of a dierentiable
manifold. The parameterizations I have described do the job for curves and
surfaces. Riemann also suggests that we should be able to measure distances
using an innitesimal displacement or line element of the form
ds =
_

g
ij
dx
i
dx
j
Looks familiar !
So, according to Riemann you have a set with a coordinate system (a
manifold) and a metric (nowadays the two together are a Riemannian man-
ifold) and away you go. Everything is intrinsic. A manifold does not sit in
some larger place like surfaces sitting in space. It just is. He is now ready
to show that with just this beginning you can tell if your manifold is at -
intrinsically.
The coecients of the metric line element should tell you when your
geometry is at, but how? Certainly, if your metric coecients are constant,
such as ds
2
= dx
2
+ dy
2
for R
2
, then your world is at. Unfortunately,
the coecients may not be constant and your world is still at. For R
2
in
polar coordinates we have ds
2
= dr
2
+ r
2
d
2
. So, you can describe the same
geometrical structure with dierent coordinate systems, and have dierent
looking metrics.
13
On the other hand, if you have two dierent coordinate systems, then
the coordinates themselves are related by a transformation, for example x =
r cos and y = r sin relate rectangular and polar cordinates in R
2
. The
metric coecients should also be related. For coordinate systems (x
1
, . . . , x
n
)
and ( x
1
, . . . , x
n
), the metric coecients satisfy (more chain rule stu)
g
ij
=

x
k
x
i
x
l
x
j
g
kl
(3.1)
Therefore the question of whether your world is at or not, becomes: can you
change your coordinate system (x
1
, . . . , x
n
) into a new system ( x
1
, . . . , x
n
)
so that your new metric coecients are the euclidean coecients, g
ij
= 1, if
i = j and zero otherwise. Which boils down to can the system of dierential
equations
g
ij
=

x
k
x
i
x
k
x
j
(3.2)
be solved for x
1
, . . . , x
n
?
The system in (3.2) looks harmless enough, but in fact it is not in any
standard form for which partial dierential equations methods can be used.
So, when in doubt dierentiate and do a lot of algebra and maybe a miracle
happens and indeed it does - you can get

x
i
_
x
k
x
j
_
=

l
ij
x
k
x
l
a system of linear equations with coecients precisely our old friends the
Christoel symbols. There are a lot of equations, but each contains only
one of the new coordinates, so that for each x
k
you have a linear system for
its partial derivatives. Solutions to these equations could then be integrated
to get x
k
, with suitable initial conditions to ensure a non-trivial coordinate
system. For solutions to exist certain well known integrability conditions
on the coecients must be satised
1
, namely

l
ik
x
j

l
ij
x
k
+

p
ik

l
pj

p
ij

l
pk
_
= 0
1
Integrability conditions are common for partial dierential equations. Here is one you
may know. Given P and Q dened on R
2
, for there to be a function f so that f
x
= P
and f
y
= Q, the functions P and Q must satisfy the integrability condition P
y
= Q
x
.
14 CHAPTER 3. RIEMANN - MANIFOLDS
(see Spivak, vol I p 187). Therefore, if your world is at, then these combi-
nations of the Christoel symbols for your coordinate system are all zero. In
fact, the condition is sucient for local solutions, which in this case means
that, if the condition is satised, then your world is at.
The left side of this equation is denoted R
l
ijk
and these numbers are the
coecients of the Riemann curvature tensor.
2
The miracle continues because
the Riemann curvature tensor coecients for two dierent coordinate systems
are related by

x
l
x
i
x

x
j
x

x
k
x

R
l
ijk
(3.3)
which means that if the Riemann curvature tensor vanishes in one coordinate
system, it vanishes in all. It does not matter what coordinate system you
are using, if the Riemann curvature tensor is zero, your world is at!
I mention in passing that since the s are determined by the metric
coecients and their rst derivatives, the Rs are functions of the metric
coecients and their rst and second derivatives. You can read the footnote
again now, if you want to.
A couple of interesting simplications of the Riemann curvature tensor
have appeared. The rst is the Ricci curvature tensor
R
ij
=

R
l
ilj
which happens to be symmetric, R
ij
= R
ji
. The other is the scalar curvature
R =

g
ij
R
ij
2
I do not want to get into a big discussion of tensors, but will call something a tensor
when it is one. So, what is it? Tensors are natural generalizations of vectors and matrices.
They are fundamental objects for describing geometrical and even physical concepts. If
you can pose a fact or law of nature in terms of a tensors coecients in one coordinate
system the law will look the same in any other.
Let me just say that you could even dene a tensor as a collection of numbers for each
point in an n-dimensional coordinate system with subscripts and superscripts taking values
from 1 to n that transform linearly when the coordinates are changed using the partial
derivatives of one set of coordinates with respect to the other. Subscripts transform one
way and superscripts another. Look carefully at (3.3), which comes up shortly and you
will see what I mean. In fact, (3.1) says that the metric coecients are the coecients
of a tensor. People, including me, often simply write the symbol for the coecients of a
tensor and refer to it as the tensor, for example I might say the metric tensor g
ij
.
This vague denition of tensor may be annoying or at least unmotivated, but it will
do for our purposes and give you something to look into. Well, I guess this footnote has
become a digression itself, but after all it is a footnote, you did not have to read it.
15
The two will become a good deal more interesting when we get to general
relativity.
If you think about it, the Riemann curvature tensor can have a lot of coef-
cients, actually, n
4
. Fortunately, there are symmetries and antisymmetries
in the subscripts and superscripts, such as R
l
ijj
= 0, so that there are not
so many independent coecients, n
2
(n
2
1)/12 to be exact. So, for n = 2
there is only one and all you need to know is the scalar curvature, R. In three
dimensions there are six, and all you need to know is the Ricci curvature ten-
sor, R
ij
. In dimension four, the interesting one for general relativity, there
are twenty, ten of which can be summarized in the Ricci curvature tensor
and the other ten, well, that is a long story.
I have used the word curvature a lot, but only to determine if things
do not curve. Needless to say, there is more that could be said about how
Riemann curvature is really curvature, but I will just leave it as more for you
to look into. To stimulate your interest, let me make a couple of unsupported
observations about curvature in surfaces. If you lived in a plane, you could
determine it was at because you would nd that R = 0 at every point. On
a sphere you would nd that R was always positive and on a torus you would
nd that R was positive at some points, negative at some points, and even
zero at some points. You might be surprised to know that little people living
in the surface of a cylinder would think their world was at. They would
nd that R = 0 at every point and, they would be correct. A cylinder is just
a rolled up plane, rolled without stretching or folding. The length of curves,
the geodesics, and all intrinsic aspects of the geometry are not changed by
rolling up the plane. It only appears to be a cylinder to those looking at it
from the outside - extrinsically.
16 CHAPTER 3. RIEMANN - MANIFOLDS
Chapter 4
EINSTEIN - Gravity
Before looking at Einsteins gravity, I want to look at Newtons. According
to Newton an object moves in a straight line (geodesic!) at a constant speed
in the absence of forces. If you asked Newton what a force was he would say
it was anything that causes you to deviate from travelling in a straight line
at a constant speed.
Suppose we have an object with mass M sitting at the origin of a co-
ordinate system in space, then, according to Newton the gravitational force
exerted by the object on a passing particle will accelerate the particle by an
amount inversely proportional to the square of the distance between them,
that is, for a particle at a location (x, y, z) = (t) at time t,

(t) =
MG
r
2
u
where where G is Newtons gravitational constant of proportionality, r =

x
2
+ y
2
+ z
2
= |(t)| is the distance between the object and the particle
and u = (t)/r is a unit vector in the direction from the object to the
particle.
Gravitational force is conservative, that is, it is the gradient of a potential
function, (x, y, z) = MG/r in this situation, so that

(t) =
Furthermore, the divergence of , that is, the Laplacian,
2
, satises

2
= 0 (4.1)
17
18 CHAPTER 4. EINSTEIN - GRAVITY
This equation is the Newtonian gravitational eld equation and characterizes
gravity in our situation. By characterize I mean that solutions to the
equation tell you how gravity acts. That may seem a little far fetched, since
a lot of functions satisfy the equation. Our situation has a single mass. If you
add the physically reasonable assumption that gravity due to that mass is
radially symmetric, then is a function of r alone and (4.1) becomes a simple
ordinary dierential equation whose solutions are of the form = a/r + b.
If you also add the physically reasonable assumption that the gravitational
eect of the mass goes to zero as you get further away, then b must be zero
and you are back to where we started.
In general, there is other mass about and the eld equation becomes

2
= 4G
where is the density of the mass at each point in the universe. You tell me
and I will tell you , how gravity works for that distribution of mass.
That is gravity according to Newton.
Now, to Einstein, but not directly to gravity, rst we need to talk about
spacetime.
We live in spacetime, a four dimensional manifold where one coordinate
measures time and the other three distance. The points in spacetime
are called events. To Newton time and space were absolute, the same for
everyone. To Einstein, the speed of light c was absolute. Suppose I think
I am standing still, you are moving ahead with a velocity v, and a photon
whizzes by both of us. According to Newton, if I measure the speed of the
photon, I will get c, but if you measure it you will get c v, assuming you
were following the same line as the photon. According to Einstein we would
both measure the speed of the photon to be c.
Many surprising conclusions follow from Einsteins statement that the
speed of light is the same for all observers. Perhaps the most fundamental
conclusion is that two dierent observers could measure dierent times and
distances between the same two events. Time and space are relative, not
absolute.
1
1
Actually, distance between events was not absolute, even for Newton. If a person
riding a train going 120 km/hr drops a ball from a height of one meter, she would say that
the ball traveled one meter straight down before it hit the oor of the train. Her mother-
in-law, standing on the platform watching the train go by would say that the ball traveled
about 3.5 meters along a diagonal path. The distance between events in Newtonian space
depends on the observer. Who is correct? The mother-in-law, she always is.
19
It can be shown (dont you just hate that phrase) that there is one mea-
surement that is the same for all observers, the so called proper time, ,
given by

2
= t
2
(x
2
+ y
2
+ z
2
)/c
2
where t is the time and x, y, and z the distances between two events
measured by an observer, any observer. Proper time is the arclength of
spacetime. The Minkowski metric for spacetime is
d
2
= dt
2
(dx
2
+ dy
2
+ dz
2
)/c
2
Actually that is the metric for at spacetime, where there is no gravity
and no forces.
According to Einstein there are forces, but gravity is not one of them. In
the absence of forces, such as electromagnetism, free falling particles travel
along geodesics, just like Newtons particles in the absence of forces except
that for Einstein there is one less force. In other words, the path of a particle
through spacetime () = (t, x, y, z) = (
0
(),
1
(),
2
(),
3
()), satises
the geodesic equations (2.1), namely

k
=

k
ij

j
where the derivatives are with respect to proper time. But where do the s
come from? A metric tensor, of course. But where does the metric tensor
come from? GRAVITY, of course. Mass changes the shape of the universe
and the eect of the change in shape we call gravity.
The metric tensor g
ij
that denes the geometry is essentially Einsteins
gravitational potential function analogous to Newtons . The s are built
from the metric coecients and their rst derivatives (I said that fact would
become interesting) analogous to . All that remains is the Einstein
gravitational eld equation to characterize gravity analogous to Newtons

2
= 4G, that is, to describe how the distribution of mass in the uni-
verse determines the metric tensor, hence the geometry of the universe.
At this point I have the impression that Einstein just started guessing
and decided to come up with an equation that would be reasonable, not
necessarily be based on some fundamental principle, but would survive if it
agreed with experiment. Of course, his guesses were educated. In particular,
the eect of gravity was due to mass, or mass-energy, really, since E = mc
2
said mass and energy were the same thing. There was already available
20 CHAPTER 4. EINSTEIN - GRAVITY
a general description of mass-energy in the energy-momentum tensor, T
ij
,
a symmetric tensor that could describe the distribution of mass-energy in
spacetime analogous to Newtons . So, he wanted a symmetric tensor that
was related to the geometry and should be built from the metric coecients
and their rst and second derivatives, as was Newtons
2
. Well folks, the
Ricci curvature tensor, R
ij
ts the bill. So, his rst choice was simply to say
that R
ij
is proportional to T
ij
, that is,
R
ij
= kT
ij
for some constant k. Unfortunately, this equation only works when there is
no mass except at a single point as in the example we used for Newtonian
gravity where the equation becomes R
ij
= 0. The reason it does not work in
general is that there is a way of calculating something called the divergence
of a tensor and the divergence of T
ij
is zero, but the divergence of R
ij
is not.
To x the problem Einstein added more geometry. It can be shown that
the divergence of
R
ij

1
2
R g
ij
+ g
ij
is zero where R is the scalar curvature and is any constant. In fact, this
tensor is the only tensor with zero divergence and the correct indices that
can be built linearly from the Riemann curvature tensor. Einstein was able
to determine that the proportionality constant k should be 8G and could
see no reason why should be there at all, so he presented to the world the
Einstein eld equation for gravitation
R
ij

1
2
Rg
ij
= 8GT
ij
and that is where the journey almost ends.
The equation looks very simple and elegant, but it is, in fact, a nasty
system of ten second order, non-linear dierential equations. Solutions are
not easy to come by, but several have been obtained in special circumstances.
One of the rst was presented to Einstein in 1916 by Karl Schwarzschild
who was also calculating artillery trajectories on the Russian front at the
time. Schwarzschild obtained the metric for the same situation I used for my
Newtonian example, one mass M.
d
2
=
_
1
2GM
r
_
dt
2

_
1
2GM
r
_
1
dr
2
r
2
(cos
2
d
2
+ d
2
)
21
Note the presence of Newtons potential function . I also mention it because
it predicts black holes! But, I will let you take that from here.
Not too long after the equation was published, Alexander Friedman told
Einstein that his equation implied that the universe could be expanding.
Einstein refused to believe that was possible. The sacred universe should
be static. Friedman nally convinced Einstein, who then decided to change
the equation by putting the g
ij
term back in. Carefully choosing would
rule out an expanding or contracting universe. became known as the
cosmological constant. About ve years later Edwin Hubble saw his red
shift that showed that the universe was expanding. Einstein removed the
cosmological constant once again and stated that putting it in was the worst
mistake of his life. In recent years there have been indications that should
be put back in. I refer you to Brian Greenes The Elegant Universe for that
story.
Some people have felt that at this point Einstein made a real mistake. He
launched his ill-fated campaign for a unied eld theory that incorporated
all forces into the geometry of spacetime. He never found it, but I bet he
could almost taste it, so near, yet so far. To be fair to him, you can almost
taste it. At the beginning of the twentieth century there really was only one
other force, electromagnetism. Quantum theory was in its infancy and its
strong and weak nuclear forces had yet to come to prominence. Einstein did
not think too much of quantum theory anyway. It was he who said about
quantum theory that I am convinced that He does not play dice. In his
mind all he needed to do was to somehow incorporate electromagnetism into
geometry. There was room. Gravity was covered by the Ricci tensor, which
used up half of the available information in the curvature tensor, there were
still ten independent coecients left. There is also a tensor that characterizes
the electromagnetic eld, the Maxwell tensor, F
ij
. It is antisymmetric, that
is F
ji
= F
ij
, so it had six degrees of freedom, plenty of room for it. There is
also the current density tensor J
i
with four degrees of freedom. It all seem to
add up, at least, it seemed that it should. I can see Einstein saying to himself
that it just has to t together somehow, but he never could make it happen.
In fact, Theodor Kaluza sent a paper to Einstein in 1919 showing how it
might be possible to unify electromagnetism and gravity, but it required
a ve dimensional space, one time dimension and four spatial dimensions.
Einstein and no one else, at least for a while, could conceive of such a thing
- look around, there are obviously only three spatial dimensions.
It may yet be that both Einstein and Kaluza were on the right track. It
22 CHAPTER 4. EINSTEIN - GRAVITY
may, in fact, be possible to nd a unied Theory of Everything and it may
live in a ten or more dimensional universe. But, that is where string theory
comes in and Brian Greene tells that story very well. So, I think it is time
for me to come to the end of my story.
Einsteins eld equation must stand as one of the great intellectual achieve-
ments. I enjoyed very much trying to understand it.
23
References
Carroll, Sean M. (2004), Spacetime and Geometry, San Francisco: Addison
Wesley.
Faber, Richard (1983), Dierential Geometry and Relativity Theory, New
York: Marcel Dekker.
Flanders, Harley (1989), Dierential Forms with Applications to the Physical
Sciences, New York: Dover.
Gray, Alfred (1999), Modern Dierential Geometry of Curves and Surfaces
with Mathematica, 2nd Edition, Boca Raton: CRC Press.
Greene, Brian (2003), The Elegant Universe, New York: Random House.
Greene, Brian (2004), The Fabric of the Cosmos, NewYork: Alfred A. Knopf.
Hartle, James B. (2003), Gravity: an Introduction to Einsteins General Rel-
ativity, San Francisco: Addison Wesley.
Lawden, Derek F. (1982), Introduction to Tensor Calculus, Relativity and
Cosmology, New York: Dover.
Ludvigsen, Malcolm (1999), General Relativity, Cambridge: Cambridge Press.
Millman, Richard S. and Parker, George D. (1977), Elements of Dierential
Geometry, Englewood Clis: Prentice Hall.
Mlodinow, Leonard (2001), Euclids Window, New York: The Free Press.
Spivak, Michael (1999), A Comprehensive Introduction to Dierential Geom-
etry, 3rd Edition, Houston: Publish or Perish.
Finally, I refer you to Wolframs Mathematica Information Center, a veritable
wealth of information and resources on many subjects, to be found at
http://library.wolfram.com/infocenter/