Physics 161: Black Holes

Physics 161: Black Holes
Kim Griest
Department of Physics, University of California, San Diego, CA 92093
ABSTRACT
Introduction to Einstein’s General Theory of Relativity as applied especially to black

holes. Aimed at upper division Physics Majors. Taught as Physics 161 at UCSD. Last
taught Spring 2014.
Contents
1 Introduction 2
1.1 Tour of the Universe . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.2 Black Holes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.3 Curved Spacetime . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.4 Example of two surveyors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.5 Tidal force as curvature . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
2 The Metric 6
2.1 Using a metric to find distances, areas, etc. . . . . . . . . . . . . . . . . . . . . . . . 7
2.2 Metrics Continued . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
2.3 Expanding Universe and FRW metric . . . . . . . . . . . . . . . . . . . . . . . . . . 8
2.4 Spacetime metrics and nomenclature . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
2.5 Schwarzschild metric . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
3 General and Special Relativity 13
3.1 Spacetime diagrams . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
3.2 More special relativity: how time and space appears to other observers: Lorentz
transformation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
–2–
3.3 Calibrating the axes in a spacetime diagram: hyperbolas=circles! . . . . . . . . . . . 15
3.4 Time dilation and length contraction by spacetime diagram . . . . . . . . . . . . . . 17
3.5 Other special relativity you need to know . . . . . . . . . . . . . . . . . . . . . . . . 18
3.6 Time Dilation in a gravitational field . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
3.7 Old Idea of Black Holes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
4 Geodesics: Moving in “Straight Lines” Through Curved Spacetime 21
4.1 Geodesics and Calculus of Variations . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
4.2 Geodesics as equations of motion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
4.3 Euler-Lagrange Equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
4.4 First Example of Euler-Lagrange equations: classical mechanics . . . . . . . . . . . . 24
4.5 Second example of Euler-Lagrange equations: Flat space geodesics . . . . . . . . . . 24
4.6 Geodesics in Minkowski spacetime . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
4.7 Conserved quanties in the Euler-Lagrange formalism: Energy and Momentum . . . . 26
5 Equivalence Principle, Gravitational Redshift and Geodesics of the Schwarzschild

Metric 28
5.1 Gravitational Redshift from the Schwarzschild metric . . . . . . . . . . . . . . . . . . 28
5.2 Light bending and the Equivalence Principle . . . . . . . . . . . . . . . . . . . . . . 29
5.3 Gravitational Redshift again . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
5.4 Geodesics of Schwarzschild metric from Euler-Langrange . . . . . . . . . . . . . . . . 30
6 Distances and Times Around a Black Hole 34
6.1 Can you fall into a black hole? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
6.2 Time to fall into a black hole . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
7 Shooting Light Rays into Black Holes, Inside a Black Hole, Orbits in the Schwarzschild
Metric, Effective potentials 38
7.1 Shooting Light into a Black Hole . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
7.2 Inside the Black Hole . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39

–3–
7.3 Orbits in the Schwarzschild metric . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
7.4 Effective Potential for Newtonian Orbits . . . . . . . . . . . . . . . . . . . . . . . . . 41
7.5 Effective Potential for Schwarzschild Orbits . . . . . . . . . . . . . . . . . . . . . . . 42
8 Extracting Energy from a Black Hole; Light Orbits 46
8.1 Extracting Energy from a Black Hole . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
8.2 Geodesics and motion of light around a black hole . . . . . . . . . . . . . . . . . . . 47
9 Light Bending and Gravitational Lenses 51
9.1 Formulas . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
9.2 Powerpoint presentation on Gravitational Lensing . . . . . . . . . . . . . . . . . . . 54
10 Some Questions and Puzzles Around Black Holes 55
10.1 How fast is an object going when it enters a black hole? . . . . . . . . . . . . . . . . 55
10.2 What does it look like standing near or falling into a black hole? . . . . . . . . . . . 56
11 Death by Black Hole 59
11.1 The final plunge . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
11.2 How do you die when you go into a black hole? . . . . . . . . . . . . . . . . . . . . . 59
12 Where Do Metrics Come From? 62
13 Inside the Black Hole: Kruskal-Szerkeres Coordinates; General Black Holes 68
13.1 Coordinate problem at r = rS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
13.2 Kruskal-Szerkeres coorindates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
13.3 General Black Holes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73
14 Rotating Black Holes: the Kerr Metric 76
14.1 Kerr Metric . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
14.2 Horizon of Kerr metric: maximum rotation and charge of black holes . . . . . . . . . 76
14.3 Singularity of Kerr metric . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77

–4–
14.4 The Ergosphere . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
14.5 Light orbits, inmost stable orbits for Kerr metric . . . . . . . . . . . . . . . . . . . . 79
14.6 Energy extraction from rotating black holes; ergosphere and Penrose process . . . . 80
14.7 Inside the Kerr black hole . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81
15 Can Anything Escape a Black Hole? Hawking Radiation 87
16 Entropy and Black Holes; Observing Real Black Holes 91
16.1 Observations of Black Holes; powerpoint slide presentation . . . . . . . . . . . . . . . 91
16.2 Entropy and Black Holes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91
17 Gravitational Waves 94
17.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94
17.2 Linearized Weak Field GR . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94
17.3 Connection with Newton . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95
17.4 Gravitational Waves . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95
17.5 Detecting Gravitational Waves . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98

–5–
1. Introduction
This is Physics 161, Black Holes. This course is not a prerequisite for anything, so I am
assuming everyone is taking it for interest. This also means that we can modify the content to
what you want to learn about. So be sure to communicate to me what you are liking and not
liking. I hope this course will be a fun application of your physics, math, and engineering skills,
applied to bizarre situations that really occur in nature. It’s a big Universe out there and the kind
of things going on are pretty amazing. I also find it amazing that we humans – dots on a tiny
planet orbiting one of 100 billion stars in one of 100 billion galaxies – can actually work out what
the Universe is like. You will hopefully find that the tools needed to do this are within your grasp.
Astrophysics cuts across all disciplines, so you will be using much of what you learned previously.
The prerequisites for this course are the entire Physics 2 or Physics 4 sequence. Since we will be
studying Einstein’s Theory of General Relativity (GR) and this is based upon the simpler Special
Relativity, it is crucial that you have studied Special Relativity previously. For example, if you
only tool Physics 2A,B,C and missed out 2D, then you shouldn’t take this course.
GR was Einstein’s crowning achievement and in my opinion is one of the greatest achievements
humans have yet produced. It is almost never taught at the undergraduate level, so there is really
no adequate book. I’ve developed a way to get to the essential physics without using graduate level
math, but we will stretch your math skills a little. I will teach the extra math you need, so don’t
worry.
My plan for the course is to do a mixture of hand waving explanations, analogies, examples,
and mathematical calculations, some of which will be optional for you to learn. So the level
of presentation will vary. For this reason interaction, questions and feedback from you will be
essential. If you aren’t following something, please just raise your hand and say so. GR and Black
holes (BH) will definitely blow your mind, and you will discover that many “dumb” questions have
very subtle answers. So be sure to ask all the dumb questions you come up with. You can and will
understand black holes, curved spacetime, etc. by the end of the quarter.
The grading of this course will be a little different. There will not be any tests. There will be
graded homeworks which will account for 60% of the grade. The other 40% will be from a final
paper or talk given during the final. Everyone must attend the final and hear the talks. There is
a handout that tells about the grading. Note that I am very concerned about academic integrity.
You may not copy anything from anybody else, or from any source whatsoever. You must do all
the work yourself. You may talk to others on how to do homework problems, but you must work
out 100% of everything yourself; otherwise you are cheating and if caught you will suffer very severe
consequences. Please read and reread the handout for details of the grading what is expected.
Finally, note that I have not been able to find an appropriate book for this class. I’m on
my 3rd book this quarter. The one I used last time was too easy; I’ve used the current required
textbook before, but it may be somewhat too hard. So it is essential that you come to class. That
is where you will learn at the level needed to do the homework.
–6–
1.1. Tour of the Universe
We will start with a slide show of the Universe. This is so that you get oriented to what is out
there in the Universe, and so when I say the word “galaxy” or “supernova” you have a feeling for
what I am talking about.
1.2. Black Holes
Black Holes and the expansion of the Universe (covered in Physics 162) are two subjects that
rest completely on Einstein’s General Relativity. We will not be able to cover GR in depth, but
we will understand the essential concepts at a level even most PhD physicists do not. And we will
do actual calculations of what happens around and even inside black holes. Most physicists don’t
study GR because it only differs from Newton’s gravity theory and from Special Relativity in a
few cases. But GR is Nature’s choice – whenever GR differs from Newton, GR has shown to be
right. It is how Nature actually works, and requires a radical rethinking of physical reality. GR
and Quantum Mechanics are the two subjects I know that are most likely to surprise you.
1.3. Curved Spacetime
GR says that gravity is not really a “force”; but instead is curved spacetime. What does that
mean?
Galileo and Newton view motion with respect to a rigid Euclidean reference frame that extends
throughout all space and endures forever. Within this ideal frame, there exists the mysterious force
of gravity – a foreign influence. Einstein says, “there is no such thing”. Climb into a spaceship and
see for yourself – no gravity there! Suppose you are floating in a space ship with no windows. Can
you tell you whether you are out in the middle of free space or orbiting the Earth? Not really!
This is the starting point of GR, Physics is locally gravity free.All free particles move in straight
lines and constant speed. In an inertial frame, physics looks simple. But such frames are inertial
only in a limited region, i.e. local. Complications arise when motion is described in nearby local
frames. Any difference between direction in one local frame and a nearby frame is described in terms
of “curvature of spacetime”. Curvature implies it is impossible to use a single Euclidean frame for
all space. In a small region, curvature is small, that is it looks flat. Einstein adds together many
local regions and has a theory with no gravity force. Newton has a single flat space and an extra
force. These are radically different views. Einstein is right, but usually Newton’s view is good
enough for calculation.
–7–
1.4. Example of two surveyors

Fig 1: Surveyors on Earth going north.
Let me give an example that is extremely helpful in understanding what I just said. Consider
two surveyors standing 100 meters apart on the equator. They both decide to start out perfectly
parallel towards the north by rolling a big ball directly north. Some time later as they roll their
balls, one notices that the distance between the two balls is less than the initial 100m. “Hey” one
surveyor calls, you aren’t going straight, you are coming towards me. The other says “I’m going
perfectly straight, it’s you that’s moving.” After a lot of checking they decide they both are rolling
the balls straight, but that there must be some mysterious force that is pulling the balls toward
each other. (What is happening of course, is that both balls are approaching the north pole, and
would hit each other there.) They try the experiment with bigger balls and discover that the big
balls come closer as the go north by the same amount. Since F = ma, the bigger balls require a
bigger force and thus they decide this force is proportional to the mass of the object. In fact, it
seems all objects moving north attract all other objects with a force proportional to their mass.
“We have made a great discovery; let’s call this force gravity”, the surveyors decide.
The surveyors think they have a new force because they think they are moving on a flat surface,
but in reality are on the large curved surface of the Earth. They don’t realize the reason for the
balls coming together is the curvature of the Earth’s surface. In fact, you can do the math for the
radius of the Earth and even find the value of the effective “Newton’s constant G” (not the same
of course as our normal G, and this “gravity” does not fall-off as r−2 .)
From Einstein’s view, there is no force. The movement together of the balls is proof that the
Earth’s surface is curved. Einstein says the same thing with regard to actual gravity that pulls
the falling apple toward the Earth. No force, but curved spacetime. Note in the example of the
surveyors only space (Earth surface) was curved; in GR both space and time are curved. This view
in fact explains a major mystery of Newton’s law. Newton had two types of mass: m = F/a is
“inertial mass”, telling how hard it is to accelerate things, while the m in F = GM m/r2 is the
gravitational mass, telling how much gravity comes off the object. Why are these masses the same?
In Coulomb’s law, the source of the force is the charge, and it is not the same as the mass. This
is a mystery, but it has been tested carefully many times and the two masses are always equal.
Einstein’s answer is that there is only the inertial mass, which curves spacetime. Gravity as a force,
doesn’t exist.
1.5. Tidal force as curvature
The principle of relativity you learned in Special Relativity says physics is the same in all
inertial frames. Consider traveling in a moving train or plane. Drop a ball; it falls just like when
standing on the ground. You can play catch or pour wine on a plane, even though for someone
watching from outside the plane the ball or wine would travel in a parabola. The principle of
–8–
relativity says one cannot tell whether or not one is moving in a frame with constant velocity
(except by looking outside at someone else). So consider a mass floating in an orbiting rocket ship;
not touching anything, just floating. Where does it get it marching orders from? Newton says
both the mass and ship get their orders from the distant Earth. Einstein says the mass gets its
orders locally. A free falling frame is a “local inertial frame” so since there is nothing inside the
spaceship pushing on the mass, it stays still with respect to the spaceship. In fact, according to
Einstein both the space ship and mass are sampling the local curvature of spacetime which is what
is causing them to orbit. Things move in “straight lines” in inertial frames; the mass can veer, but
only responding to structure of spacetime right there. Newton says the mass would go “straight”
in his ideal all pervading reference frame but the Earth deflects it.
How do you tell if a frame is inertial? Easy, just check every particle, light ray, etc. to see
if they move in straight lines at constant speed. So inside the space ship it is an inertial frame
and everything moves simply. Simple? Too simple! Where is gravity at all? How do we see the
curvature?
Fig. 2: Balls in a space ship
Consider two balls in a space ship. We put them side by side 25 m apart. If the space ship is in
orbit, the balls just float there. They don’t move apart or together, and if there were no windows,
there would be no way to tell they were in orbit above the Earth or in the middle of space far from
any star or planet. Now, instead of in orbit, drop the entire space ship from a height of 250 meters
above the earth. The ship and balls both fall straight down, and will hit the ground 7 seconds
p
later (t = 2d/g). While falling, the balls still seem to be floating in deep space away from all
forces. However, if you check carefully there is a small effect. Going straight down towards the
Earth’s center, the balls are about 1 × 10−3 m closer together when they hit. l = θr, dl/l = dr/r,
dl = ldr/r = (25)(250)/6.4 × 106 = 1 × 10−3 m. Watching this from the ground it is clear what
is happening, but inside it seems as if the balls are attracting each other. After 7 seconds they
have moved about 1 mm closer. This is not actually the gravity attraction between the two balls,
but is the “tidal” force and in fact proves that the space is curved. Note that if your measuring
instruments had an accuracy of worse than 1 mm, then this attraction could not be detected. We
say that to an accuracy of 1 mm and a time under 7 seconds this 25 m wide space is a local inertial
frame, but for longer times, or better accuracy, it is not. Smaller size ships and shorter times give
more approximately inertial frames. However, if you add enough small frames together you can
detect the curvature. Consider a ring of balls above the Earth’s surface each separated by 25 m
and drop them all together. After 7 seconds they are all 1 mm closer. In each frame you can’t see
it, but by adding up all the frames you see that entire circle around the Earth has shrunk. The
factor is 1 mm/25000 mm = 1/25000, and the distance from the center of the Earth shrinks by
same factor (1/25000)(6.4 × 106 )m = 250m. Note this is just like the distance around a line of
latitude shrinks for surveyors rolling balls toward north pole. The smallness of this effect in a single
spaceship actually shows the smallness of the curvature of spacetime, which is part of the reason
GR is not easy to experimentally distinguish from Newton.
–9–
2. The Metric
In GR the key concept is the metric. GR replaces gravity with curvature of spacetime. The
metric tells how to measure distances in space and time. The metric contains all the info about
curvature in a simple formula. It is the key to understanding GR and to be able to calculate
anything.
Examples of metrics1
• 3-D flat space metric: ds = dx2 + dy 2 + dz 2 , (or ds2 = dx2 + dy 2 + dz 2 ). This is just
p
the Pythagorian theorem! (We use dx rather than x because we want to talk about “local”
curvature. We find x from dx by integrating.)
• 2-D flat space metric: ds2 = dx2 + dy 2 .
• 4-D flat space: ds2 = dx2 + dy 2 + dz 2 + dw2 . w is “4th” dimension; note how easy this is to
write mathematically, but it is hard to visualize! Note the number of dimensions described
by a metric is the number of different differentials.
• 2-D curved space metric (surface of sphere):

Fig. 3: Coordinates for surface of sphere
ds2 = r02 dθ2 + r02 sin2 θdφ2 , where r0 is a constant. Note if you define dx0 = r0 dθ and
dy 0 = r0 sin θdφ then locally ds2 = dx02 + dy 02 , and it looks like flat space in small enough
areas. However, if you move far then θ changes, and the distance between points on a sphere is
p
not given by a flat space formula: s 6= (x0 − x0 )2 + (y0 − y 0 )2 . You need to do the integral:
Z θ1 ,φ1 Z θ1 ,φ1 q
s= ds = r02 dθ2 + r02 sin2 θdφ2 .
θ0 ,φ0 θ0 ,φ0
This is a longer distance than the flat space distance between these points.
• 3-D curved metric. Above we had a curvature in 2-D space. How aboutpcurved 3-D space?
Easy to write mathematically, but hard to visualize. For example, ds2 = x2 + y 2 + z 2 dx2 +
dy 2 + dz 2 is a curved 3-D space. Just add almost any mathematical function to the flat 3-D
metric and
it will be curved. We can try to visualize using “embedding”. In the curved 2-D metric example
we can visualize the curvature by drawing a 3-D sphere, that is by curving the 2-D surface through
3-D space. Similarly we could visualize curved 3-D space as curved through a 4-D space! But note
1
Technically the metric is a rank two tensor, e.g. a matrix, and what we are showing here is called the “line
element”. We will use the terms “metric” and “line elemement” interchangeably in this course, since they contain
the same information, but in a more advanced course the distinction would be made.
– 10 –
the curvature of 3-D space does not need the 4th dimension. That sounds hard, but if we can
ignore one of the 3-D spatial dimensions, we again have curved 2-D embedded in 3-D. We will show
examples later. Note that the space around the Earth is curved, but we can’t see that curvature!
We can measure its effects however and prove it is there! It becomes an experimental question.
2.1. Using a metric to find distances, areas, etc.
The left hand side of a metric (aka line element) gives the real distance measured in meters.
The right hand side gives the coordinates, specified by the differentials, as well as functions and
constants. The coordinates may or may not be in units of distance, so to find an actual distance,
area, or volume, the left hand side must be used. For example, consider the 3-D metric
ds2 = f (x, y, z)dx2 + g(x, y, z)dy 2 + h(x, y, z)dz 2 ,
where f (x, y, z), g(x, y, z), and h(x, y, z) are functions that specify the curvature of the space. To
find a distance in the x direction, one would hold y and z constant. That is one would set dy = 0
√
and dz = 0, to get ds ≡ dlx = f dx. The actual distance from x0 to x1 would therefore be
R x1 √
lx = x0 f dx. Likewise to measure a distance in the z direction you would set dx = dy = 0
√
and integrate dlz = hdz. The distance along an arbitrary curve can be found by factoring out,
R Rp
say dx, and integrating: l = dl = f + g(dy/dx)2 + h(dz/dx)2 dx, where dy/dx and dz/dx are
calculated along the integration path.
Finally, while in flat space a small area is given by dA = dxdy, in curved space one must
√
use dA = dlx dly = f gdxdy (or dly dlz , etc.). The volume element is thus dV = dlx dly dlz =
√
f ghdxdydz.
2.2. Metrics Continued
It was mathematicians who first wondered whether a consistent system could be created with
different (non-Euclidean) metrics. Gauss, Bolyai, Lobachevski, Reimann, etc. worked it all out.
When Einstein came along with his great intution that this might be relevant to the real world he
applied this math, “differential geometry” in his GR.
GR takes the matter and energy in a system and predicts the metric. This is done
by solving Einstein’s field equations, something we will not do in this course. Once you have the
metric you can calculate distances, times, and motions of particles. We will do this. GR gives
some surprising results, and so far every prediction that has been tested has been experimentally
verified.
– 11 –
Fig. 1.— Figure for Chapter 1
2.3. Expanding Universe and FRW metric
For example, consider the entire Universe filled uniformly with matter. One can solve the field
equations for the metric in this case and find (for the right amount of matter):
p
ds = R(t) dx2 + dy 2 + dz 2 ,
– 12 –
where the scale factor, R(r) = (t/t0 )1/2 at early times and R(t) = (t/t0 )2/3 or R(t) = exp (Ht/t0 )
at late times depending on the cosmology. Here t0 = 13.7 Gyear is the age of the Universe today.
This is called the Friedmann, Robertson, Walker (FRW) metric and was quite shocking to Einstein
when he first realized this is what his GR theory predicted for uniform matter. The shocking thing
is that R(t) changes with time. Thus the distance between objects changes with time, even when
they are “at rest”! Consider two galaxies a distance D apart today (one at (x0 , y0 , z0 ) and the
p
other at (x, y, z). Distance between them is s = (x − x0 )2 + (y − y0 )2 + (z − z0 )2 , with t = t0 . A
few years from now they are farther apart by a factor (t/t0 )2/3 even though their positions, x, x0 ,
y, y0 etc. haven’t changed!
This is the expansion of the Universe. Note it is not an explosion that happened long ago and
caused everything to blast apart. Note also what happens when t = 0 in this metric. R(0) = 0,
implying s = 0 no matter what the values of x, y, etc. are. That is, at t = 0, everything in
the Universe is touching everything else! This is the big bang. This metric predicts the big bang
happened at t = 0, and that the Universe expanded since then because the metric is changing with
time.
2.4. Spacetime metrics and nomenclature
Spacetime metrics combine the 3-D space metric with the time metric. The flat spacetime
metric, also known as the Minkowski metric, is written in Cartesian coordinates as:
ds2 = −c2 dt2 + dx2 + dy 2 + dz 2 .
This combines the flat space 3-D metric with the flat time metric: c2 dt2 . The two are combined
together so that ds is a Lorentz invariant,pthat is if the time between two events is dt, and the
distance between those same two events is dx2 + dy 2 + dz 2 , then the spacetime interval between
these two events has the same value (ds), no matter which inertial reference frame is used to make
the measurement. It is crucial for this to work that the “time part” of the metric have a different
sign that the “space part”.
One uses the spacetime metric to meausure the “invariant spacetime interval”, ds, between two
events. One also uses the space part of the metric to measure real distances (distances measured
by a meter stick). The time part of the metric is used to measure real (that is measured by a
clock) times between events. The distance between two events is best measured when the time of
the events is the same, that is measured at the same√ time, or by setting dt = 0. Thus the proper
distance between two objects is defined as dl = ds2 , with dt = 0. To measure times between
events, the events should be at the same position, i.e. one clock at different times. So we set
√
dx = dy = dz = 0. The proper time between events is defined as dτ = −ds2 /c, Note, that
because the time part of the metric has a minus sign, you have to add a minus sign to cancel it
out. While ds2 can be positive, negative, or zero, the invariant interval, ds is always positive. You
should never get an imaginary number out of a metric!
– 13 –
Actually the sign of ds2 is very important:
• ds2 < 0 implies a time-like interval, meaning the two events can be causally related, that
is, it is possible for the event with the earlier time to have caused the event with the later
time.
• ds2 > 0 implies a space-like interval. It is impossible for either of these events to have
caused the other. (Like the ends of the same ruler.)
• ds2 = 0 implies a null or light-like interval. These two events can be connected by light
rays only. Note in this case the definitions of proper time and proper distances above don’t
apply because you can’t set dt = 0 or dx = 0. The proper time between lightlike separated
events is defined as ds = 0.
Note that with the Minkowski metric, the variable t measures time and the variable x measures
spatial distance. However, in more general metrics, the variables used in the metric do not represent
time or space directly. To get actual time or space measurements you have to use the left-hand side
of the metric, i.e. ds, as described above. The variables in the metric are called the “coordinate
time” or “coordinate distances”, or even just the coordinates, and may or may not correspond to
clock time, or meter-stick distance.
Also, note that many books (and I myself) set the speed of light, c, equal to one. This makes
the units easier to work throughout all of Special and General Relativity. It means if you are
measuring time in years, you are measuring distance in light-years, or if you are measuring distance
in meters, you are measuring time in units of the time for light to travel one meter. For example,
one can say this 50 minute class session lasts about 900 billion meters. If we are halfway through
it we have about 450 billion meters to go. A useful thing to know is that light travels about 1 foot
in one nanosecond.
When using these units, you have to add back in powers of c in order to get to useful units.
There is always a unique way to do this. In these units velocities are dimensionless, so if you want
a velocity in m/s just multiple by c = 3 × 108 m/s. Energies and masses are related by E = mc2 ,
so if you know the mass and want an energy just multiply by c2 . [e.g. v = .001 means, v = .001c,
or v = 300km/s.]
Finally note that in many books the space part of the metric has the minus signs, and the
time part is positive. As long as there is a relative minus sign between space and time, it doesn’t
matter which has the minus sign, but it is very important to pick one convention and keep it. This
is called the “signature of the metric”. We will use the “East coast” signature (- + + +), while
others use the “West coast” signature (+ - - - ). With the √
opposite signature metric, the definitions
√
of proper distance and proper time change: e.g. dτ = ds2 /c, dl = −ds2 , and several other
formulas change as well.
– 14 –
2.5. Schwarzschild metric
The FRW metric is valid on very large scales where matter is distributed approximately uni-
formly. On scales the size of the Earth, Solar System, or even galaxies, matter is concentrated
in a central source, so FRW is not the proper solution of Einstein’s field equation and is not the
correct metric. The simplest case is for a spherical object, such as the Earth, Sun, (or black hole!).
Solving Einstein’s field equations for the region outside of a spherical object of mass, M gives the
Schwarzschild metric:
2GM −1 2

2 2GM 2
ds = − 1 − dt + 1 − dr + r2 dθ2 + r2 sin2 θdφ2 ,
rc2 rc2
where this is in spherical coordinates with r2 = x2 + y 2 + z 2 . There are several new features here
and we will spend a lot of time on this metric.
• First note that flat 3-D spatial metric in spherical coordinates is ds2 = dr2 + r2 dθ2 +
r2 sin2 θdφ2 = dr2 + r2 dΩ2 , Note dΩ2 = dθ2 + sin2 θdφ2 is shorthand for the angu-
lar part. This looks similar to, but is different that the 2-D curved metric on the surface
of the sphere because here r is a variable not a constant. You can tell the variables because
they are differential (i.e. dr not only r).
• Second, I included the time part of the metric dt. This makes it a spacetime metric, and not
just space. We will talk more about this.
• Third, the time part of the metric has a minus sign! We will come back to what that means
and how to deal with it.
• Fourth, the metric does not change with time. Unlike the FRW metric, the variable t does
not appear explicitly anywhere in the Schwarzschild metric. The dt doesn’t count since that
is just tells how to measure time. Thus this metric is not expanding or contracting and just
sits there. (Good thing).
• Fifth, the metric is spherically symmetric, which is why we switched to spherical coordinates.
• Sixth, there are factors in front of both dr and dt which mean that both the space and the
time are curved in this metric. If those factors were equal to 1, then this would be the 3-D
flat spacetime Minkowski metric in spherical coordinates.
• Seventh, note that something weird happens when 2GM/rc2 = 1. The dt2 term goes to zero
and the dr2 term goes to infinity. This happens when a mass M is squeezed into a ball of
radius
2GM
rSC = ,
c2
where rSC is called the Schwarzschild radius. This is the event horizon radius of a black hole,
and we expect some weird things to happen at that radius. We will discover that whenever
– 15 –
the size of an object is smaller than its Schwarzschild radius it is a black hole. For the mass
of the Sun this radius is
(2)(6.67 × 10−11 m3 /kg s2 )(2 × 1030 kg M/M ) M
rSC = 8 2
= 3km ,
(3 × 10 m/s) M
where M is the mass of the Sun. Thus if you could jam the entire mass of the Sun into a
sphere of radius 3 km, it would be a black hole.
• Finally, note that if the key factor 2GM/rc2 is small, then the Schwarzschild metric is very
close to the 3-D flat space metric. For the Sun, 2GM/rc2 = 3 km/7 × 105 km = 4 × 10−6 .
So even at the surface of the Sun, spacetime is only 4 parts in a million away from being
flat. The smallness of this number is one reason why it is hard to measure the effects of GR
that differ from Newton’s law. Around the Earth spacetime is even closer to flat. Note that
far away from the mass, r → ∞, the Schwarzschild metric turns into exactly the flat space
metric. This is what we expect of course; there is no gravity infinitely far away from a mass.
– 16 –
3. General and Special Relativity
GR is based upon the equivalence principle: All local inerital frames are equivalent
for performance of all physical experiments. Remember that a local inertial reference
frame is one that is free falling and non-rotating, where particles move at constant velocities
in straight lines (or stay still). A very important point is that Special Relativity is how spacetime
behaves in any inertial frame. This is equally true for someone out in deep space far from anything,
for someone orbiting the Earth in a spaceship, or for someone falling in an elevator with broken
cables. All three someones are in inertial reference frames and would find the same result for
any experiment they did; that is they could not tell which of the situations they were in, without
looking outside (looking outside makes the frame bigger, i.e. non-local). The influence of the
Earth’s gravity on the guy falling in the elevator is not measurable by local experiments.
The metric of Special Relativity (SR) is the Minkowski metric ds2 = −c2 dt2 + dx2 + dy 2 + dz 2 ,
and GR is an extension of SR. So we start by understanding the SR metric. In SR the relations
between any two local inertial frames are the Lorentz transformations which you have seen before.
These, and all of SR, can be derived from two postulates:
1. The principle of relativity: results of experiments don’t depend on velocity of lab.
2. All inertial frames measure the same speed of light which is exactly c = 299792458 m/s! This
postulate is the same as requiring the invariant interval ds to be the same in every frame.
These postulates can be used to derive the Lorentz transformations.
3.1. Spacetime diagrams
A very nice tool i sthe spacetime diagram. These will be indespensible when we try and go
inside a black hole, so it is worthwhile trying to understand them.
Figure: Intro to Spacetime diagram
Points:
• Axes are x vs. t. Points are events, lines represent observers moving through spacetime
(worldlines) or extended objects (if spacelike).
• The slope of an observer’s worldline is dt/dx = 1/v.
• With c = 1, an observer moving at the speed of light is a 450 line. Nothing can move at a
shallower slope, since that would mean v > c. Thus only events within the “lightcone” can
be casually connected. [We suppress the y and z directions; if we include y then it is a cone.
– 17 –
• Straight lines represent inertial observers (each related by a Lorentz transformation.) A line
straight up is someone standing still in the lab frame.
• Curved lines represent accelerated observers.
3.2. More special relativity: how time and space appears to other observers:
Lorentz transformation
Now let’s use spacetime diagrams to understand some of the weirdness of special relativity.
Suppose observer O uses (t, x) and is in the lab, while observer O0 uses (t0 , x0 ) and is moving in the
x-direction with velocity v.
Figure: O0 moving at v in O coordinates
Where do the t0 and x0 coordinate axes go in this diagram? Well, the worldline of O0 is their t0
axis, since this is x0 = 0 to them (and also y 0 = z 0 = 0). Where is the x0 axis? To find this consider
O0 spacetime diagram from O0 ’s point of view.
Figure: O0 spacetime diagram in O0 coordinates: reflecting light
From O0 view t0 axis goes straight up and x0 axis is at right angles. Consider a photon emitted
at (t0 , x0 ) = (−a, 0) in the +x’ direction. It reaches the x0 axis at t0 = 0, and then is reflected back
(by a mirror). It returns to the t0 axis (x0 = 0) at (t0 , x0 ) = (a, 0). Consider other photons emitted
at different times, −b, etc. and we see that we could define the x0 axis as the locus of points that
will reflect photons emitted at t0 = −a, to reach back at t0 = a.
This should be a frame-independent definition, so we can find the x0 axis in the O frame using
it.
Figure: t’ and x’ axes in O coordinates: reflecting light
To do this we use postulate two: light travels at c = 1 in all frames, that is, the light pulses
O0 uses to find the x’ axis travel at 450 also in the O spacetime diagram. We see that the x0 axis is
not the same as the x axis! So events which are simultaneous to O0 (i.e. happen at the same values
of t0 , e.g. t0 = 0 on the x0 axis), are not simultaneous to O (don’t happen at same values of t, i.e.
are not parallel to the x axis.) This is the famous failure of simultaneity in SR, and is made very
obvious in spacetime diagrams.
With some work, from the above postulates one can derive the famous Lorentz tranforma-
tions:
v
t = γ(t0 + 2 x0 )
c
x = γ(vt0 + x0 )
y = y0
– 18 –
z = z0,
where the Lorentz factor is the greek letter gamma:
p
γ = 1/ 1 − v 2 /c2 .
Note that with c = 1 these simplify:
t = γ(t0 + vx0 )
x = γ(vt0 + x0 )
y = y0; z = z0,
√
with γ = 1/ 1 − v 2 .
One can easily find the inverse transforms, just by setting v → −v in the above: Inverse
Lorentz transformation:
t0 = γ(t − vx)
x0 = γ(−vt + x)
In the homework you will use these to prove that ∆s between two events is the same in every
frame.
3.3. Calibrating the axes in a spacetime diagram: hyperbolas=circles!
Figure: Hyperbolas on space time diagram
How does one calibrate the axes in a spacetime diagram? Consider the equation: t2 − x2 = a2 .
All the points on this hyperbola have the same invariant “distance” from (0, 0). ∆s2 = −∆t2 +
∆x2 = −a2 ; e.g. x = 0 and t = a, etc. Thus this hyperbola is the equivalent of a circle in normal
Euclidian geometry. The proper time is the time that someone traveling along the line would
measure on their wristwatch. So no matter what the speed, all travelers from the origin to the
hyperbola take the same time to get there (i.e. they age the same amount, a). Note that −a2 is
the invariant interval, and since ds2 = −a2 < 0 this hyperbola is for timelike separations.
Similarly one can plot the equation: t2 − x2 = −b2 , at get the sideways hyperbola which
represents spacelike separations. These points all have the same proper length from the origin.
Figure: Events A and B on space time diagram
To calibrate the axes, we consider two events, A and B. A lies at (t, x) = (1, 0) on the O t
axis, while event B lies at (t0 , x0 ) = (t0 , 0) on the O0 t’ axis. We pick B to lie along the hyperbola
so the invariant interval (proper time) from the origin is the same to both events A and B. Since
the interval ∆s is invariant (same in all frames), and event A has t = 1, x = 0, we have ∆s2 =
– 19 –
Fig. 2.— Figure for Chapter 3 A
−t2 + x2 = −1. Therefore in moving frame, -t02 + x02 = −1 also. Event B occurs at x0 = 0, which
implies t0 = 1. But on the spacetime diagram, the distance from the (0,0) to event B looks looks
farther than the distance from (0,0) to event A! The point here is that one can’t use Euclidian
intuition about distances on spacetime diagrams: It is Minkowski geometry. The proper distance
is ∆s2 = −∆t2 + ∆x2 , not ∆t2 + ∆x2 as we are used to. Now you see the importance of the minus
– 20 –
signs in the spacetime metric. They change the geometry from Euclidian to Minkowski.
In the above, the moving frame observer, O0 , says t0 = 1, but the lab frame says the clock tick
happened at a longer time. (Just read off the t time from the t axis!)
This says that the moving person’s time runs slowly, when observed from the lab frame.
Finally, note that on a space time diagram we can easily see which events are separated by
time-like intervals and which are separated by space-like intervals. Thus we can see which events
can and cannot be causally connected. The 450 lines are called the lightcone. If we add say
the y dimension you see why it is called a cone, but if we add the y and z dimensions it is actually
a cone in 4-D! The cone opening towards larger time is called the future lightcone and contains
all future events that can be influenced by it. The cone opening towards the past is call the past
lightcone and contains all the events that could have influenced the event at the apex of the cones.
You can draw a lightcone around any event and see which events can influence which other events.
3.4. Time dilation and length contraction by spacetime diagram
Last time we saw how to calibrate spacetime diagrames. We can use that result to derive the
famous time dilation: moving clocks run slow to stationary observers.
Numerically we can calculate for time dilation for event B, considering starting a clock at the
origin (event A). ∆s2 = −∆t02 + ∆x02 = −∆t2 + ∆x2 . For a clock tick at event B, ∆t = t, ∆x = x,
∆t0 = t0 . In the O, lab frame: Event B occurs x = ∆x = v∆t, and time occurs at t = ∆t. In O0 ,
the moving frame, ∆x0 = x0 = 0 since events B and A occur at the some place (O0 is not moving
√
in its own frame). So −∆t02 + 0 = −∆t2 + ∆t2 v 2 = −(1 − v 2 )∆t2 , or ∆t = ∆t0 / 1 − v 2 = γ∆t0 .
This is the time dilation you learned before in special relativity.
One can similarly find out the length contraction you learned about in SR. If you remember
the derivation, you will remember that it is more tricky than the time dilation calculation. The
reason for the trickiness is easy to see in the spacetime diagram. Consider the origin as point A
and a ruler of length a laid out on a spacecraft with its far end at point B. How does that look? It
lies along the x0 axis, which we have just seen how to find (the set of points which all have t0 = 0.
Now draw the sideway hyberbola of all proper distances a. If we just project the point B onto the
x axis we find the wrong answer, that the ruler looks longer in the lab frame. What have we done
wrong? By length of ruler in the lab frame we mean the distance between the two ends measured
at the same lab frame time t. Thus the distance between points A and B is not the distance we
are after. We need to follow point A up in time until it reaches the same time t as point B has.
We know the t0 axis has the same angle from the 450 line as the A-B segment, so we can just go
along the t0 axis until we get as high as point B. Now the length of the ruler as measured in the lab
frame is the difference in x of the two ends measured at the same time. We see the ruler is shorter
as measured in the lab frame as expected.
– 21 –
Fig. 3.— Figure for Chapter 3 B
3.5. Other special relativity you need to know
This class has special relativity as a prerequisite, so I’m not going to cover much more. However,
you should look over the book to refamiliarize yourself with several topics. For example, you should
understand E = mc2 . In fact, your need to know that the real equation is E 2 = (mc2 )2 + (pc)2 ,
– 22 –
where p is the momentum. With c = 1 this is more suggestively written: m2 = E 2 − p2 , and

we see that the rest mass is an invariant in much the same way that the interval ds is. You also
should be familiar with relativistic energy and momentum: E = mγc2 , and p = mvγ. Try Taylor
expanding the relativistic expression for kinetic energy T = E − mc2 ≈ 12 mv 2 , and find the next
order correction to the famous Newtonian formula.
3.6. Time Dilation in a gravitational field
Remember the metric of flat spacetime is
ds2 = −c2 dt2 + dx2 ,
where we have supressed the y and z dimensions. In this metric, we find the proper time by setting
√
dx = 0 and using dτ = −ds2 /c = dt, and see that coordinate t is actually the proper time between
events. Likewise the proper distance is dl = dx, so that the coordinate x is the proper distance.
Compare this to the the Schwarzschild metric
2GM −1 2

2 2 2GM 2
ds = −c 1 − dt + 1 − dr + r2 dθ2 + r2 sin2 θdφ2 .
rc2 rc2
Here the proper time is found by setting dr = dθ = dφ = 0. Note then ds2 ≤ 0, that is a time-like
separation. The proper time is again
r
p 2GM
dτ = −ds2 /c = 1 − dt.
rc2
q
So if dt = 1 sec, dτ = 1 − 2GM rc2
sec, which is a smaller (i.e. shorter) time. This proper time is
the clock time measured at a distance r from the Earth (or a black hole). Thus we see that clocks
run slower in a gravitational field than in deep space (where r → ∞ and dt = dτ ).
Let’s see how big effect this is for us living on the surface of the Earth. MEarth = 6 × 1027 gm,
and rEarth = 6.38×108 cm, and a useful number to calculate is the Schwarzchild radius of the Earth
rSC = 2GMEarth /c2 = 0.886 cm. Then 2GMEarth /rEarth c2 = .886/6.38 × 108 = 1.4 × 10−9 . Using a
√
Taylor expansion, this says dτ = 1 − 1.4 × 10−9 ≈ 1 − 12 1.4 × 10−9 . Or dτ = (1 − 6.9 × 10−10 )dt.
So time runs slower here on Earth (as measured from deep space by about 1 sec every 45 years!
Not a big effect, but measurable. Gravity curves time.
However in other environments this factor can be bigger. Especially wierd is what it does
p
when r = rSC , Then 1 − 2GM/rc2 → 0, and the measured time seems to stop! This occurs at
the horizon of a black hole.
– 23 –
3.7. Old Idea of Black Holes
The idea of black holes has been around since the 1700’s when Laplace, Mitchell, and others
thought about giant stars. One can even derive the Schwarzschild radius using Newton’s laws and
get the right answer (by coincidence I think?) Consider a big spherical ball of mass M and radius
r. The gravitational potential energy at the surface of the ball is V = −GM m/r, where m is
just some small test mass. Since V goes to zero at infinity, the escape velocity from the surface
is found by setting this potential energy to the the kinetic energy E = 12 mv 2 , and solving for v.
p
This gives vesc = 2GM/r. This escape velocity increases as the mass increases, or the radius
decreases. And one could ask what happens when the escape velocity is equal to or larger than
c. Then according to Newton’s corpuscular theory of light (which is incorrect!) light would get
trapped in such an object, and the giant star would be dark. Setting vesc = c one finds the radius
of such an object is r = 2GM/c2 , precisely the Schwarzschild radius. This Newtonian calculation
gives the right answer, but is wrong. For example, light particles are massless so E 6= 12 mv 2 . There
are several errors in this calculation which just happen to cancel. The correct description requires
the Schwarzschild metric.
– 24 –
4. Geodesics: Moving in “Straight Lines” Through Curved Spacetime
We saw that gravity curves time and space. A very important result is how things move in
curved space. The basic principle is that things move on paths that are as “straight” as possible.
These can be defined as the paths which extremize the distance between two points, and are called
geodesics. In 3-D Euclidian geometry a straight line is defined as the shortest distance between
two points, and Newton’s law say in the absense of outside forces particles move along such lines.
It is similar in GR, but sometimes it is the “maximum” invariant interval which is relevant, not
the minimum. That is why we say “extremizes” the distance instead of minimizes.
To get an idea of this, consider the 2-D analogy of an ant crawling on the surface of an apple.
The ant is forced to follow the curved 2-D space of the apple skin, but suppose it walks as “straight”
as possible, i.e. not veering to the left or right. As it walks from one side of the apple to the other
coming near the stem, the ant will be “deflected” and come away from the stem at a different angle
than it approached the stem. If the ant is close to the stem it could even circle the stem continually
while still following the “geodesic”. This should seem similar to the analogy of the two surveyors.
Now consider two points on opposite sides of the stem and ask which path is the shortest
between them (the geodesic). Note that without the curved surface the shortest path would be
quite different. However, the metric requires us to stay on the surface.
If we want to calculate the motion of objects in GR, we need to be able to find the geodesics.
By finding them we will be able to derive Newton’s laws from the Schwarzschild metric, as well as
Einsteinian corrections to Newton’s laws. We can find out the real answer obeyed by actual objects
in the solar system. We will also be able to find out how objects move around black holes.
4.1. Geodesics and Calculus of Variations
GR says that the motion of a particle that experience no external forces is a geodesic of the
spacetime metric. One can summarize GR in two statements: 1. Matter and Energy tell spacetime
how to curve. 2. Curved spacetime tells matter and energy how to move. In solving for the
geodesics we are finding how matter and energy (light) move.
In 3-D Euclidian space the definition of a geodesic, aka “straight line” is the shortest distance
bewteen two points. Mathematically this can p be found from calculus of variations on the metric
distance. The differential distance ds = dx2 + dy 2 + dz 2 can be integrated between two end
points to find
Z b Z bp
s= ds = 1 + (dy/dx)2 + (dz/dx)2 dx,
a a
where we factored out a dx so the integral is done over the x-axis. Note now the integrand contains
the derivatives of the the functions y(x) and z(x) which define the path. How do find the path
(functions y(x) and z(x)) that minimize s? We need a general method. This method is called the
– 25 –
“calculus of variations” and you probably learned it in calculus. However, I will remind you of how
it is done. If you haven’t seen it before, that’s ok, since I’ll show you all you need. This is beautiful
and fun math that is used throughout advanced physics.
The basic idea is to minimize ds the same way you minimize any function is calculus: take its
derivative, set it to zero, and solve. This solution will be the geodesic, that is the extremal path.
This method works on curved surfaces since ds measures the actual distance following the curved
surface. It works the same way in GR; you just add the time part of the metric.
4.2. Geodesics as equations of motion
A geodesic on pure spatial manifold (e.g. curved surface) is a line. For example, straight lines
on a plane, or great circles on a sphere surface. In GR the geodesics include time and so are actually
the equations of motion! You can understand this by remembering the distinction between time
and space. One has choice in spatial motion, but one is forced by nature to move forward in the
time direction: 1 sec per sec. You will hit Janurary 1 of next year no matter what you do. Now
in flat space a possible geodesic is one in which you don’t move in space at all: dx = dy = dz = 0.
Then ds = dt is an extremal path and you just sit there getting older. However, if spacetime is
curved, then motion in dt can require motion in dx! Think of the surveyor analogy, where motion
north (z) required motion in the x and y direction to stay on the surface of the Earth. Thus since
motion in t direction is forced, motion in the x (or other spatial) direction will also be forced. Thus
you have the geodesic requiring dx/dt 6= 0. dx/dt is a velocity, so you can’t stay at rest and be on
the geodesics. This is how gravity attracts things and why geodesics near a mass require bodies to
fall towards the mass center. Let’s do some math.
4.3. Euler-Lagrange Equations
To get used to calculus of variations, let’s do a calculation some of you have already done:
Lagrange’s equations in classical mechanics. Then we will do it on metrics. Doing it several times
is necessary and I highly recommend you go over all these calculations at home several times.
R R
We want to extremize the integral s = ds, which can be written s = (ds/dt)dt, where we
have factored out a t. We can factor out any variable we want so as to make the integrating easier.
As an example, consider the general “least action” integral
Z
S = Ldt,
where L is called the Lagrangian. In our example L = ds/dt, but in regular classical mechanics
the Lagrangian is L = T − V = 12 mẋ2 − V (x), where ẋ = dx/dt = vx , T is the kinetic energy, and
V (x) is the potential energy. (Here we used only the x and t dimensions.) Thus for 1-D classical
– 26 –
Rb
mechanics S = a ( 12 mẋ2 − V (x))dt is called the action, and the equations of motion, F = ma, with
F = −dV /dx are found minimizing this action (principle of “least action”). In our problem we are
trying to find the path that gives the shortest (or longest) distance along the path between two
fixed points.
Let’s do it first in general and get the equations.

Z
S = L(x, y, z, ẋ, ẏ, ż, λ)dλ,
where we explicitly put in the velocities vx , vy , and vz as variables that have to be solved for, and
have factored out the general variable λ ( usually just time). This lambda is called the “affine
parameter” and is the variable you use to trace along the geodesic path. In the above, the dot
means differentiating with respect to λ, i.e. ẋ = dx/dλ. Now we take the “variation” of this integral
δS using the chain rule and set it to zero: δS = 0, and solve.
Z
∂L ∂L ∂L
δS = δx(λ) + δ ẋ + δy + · · · dλ = 0.
∂x ∂ ẋ ∂y
dx d
Next note that δ(ẋ) = δ dλ = dλ δx, where δx(λ) is a small deviation from the extremal path. Now
integrate by parts the term with the δ ẋ, by using:
Z b Z b
udv = uv|ba − vdu.
a a
Z b Z b Z b
∂L ∂L d ∂L b d ∂L
δ ẋdλ = (δx)dλ = δx|a − δxdλ.
a ∂ ẋ a ∂ ẋ dλ dẋ a dλ dẋ
We want a path defined by x(λ), y(λ), etc. that goes from point a to point b. Thus we want δx(λ),
the variation of the path from the geodesic to be zero at the end points. That is δx(λ = a) = 0 and
δx(λ = b) = 0 (and similarly for δy and δz). Thus the first term on the right hand side of above
equation vanishes, and the equation becomes:
Z
∂L d ∂L
0= − δxdλ + y and z parts.
∂x dλ ∂ ẋ
The final step in getting our Euler-Lagrange equations is to note that the variation in the path
δx(λ), δy(λ), and δz(λ) is completely arbitrary. Thus for the integral as a whole to vanish, the
integrand itself must vanish everywhere. That is, to be true for every possible function δx(λ), the
parts of the intregrand multiplying δx, δy, and δz must be zero. Thus we have the Euler-Lagrange
equations:
∂L d ∂L
− = 0.
∂x dλ ∂ ẋ
∂L d ∂L
− = 0.
∂y dλ ∂ ẏ
∂L d ∂L
− = 0.
∂z dλ ∂ ż
– 27 –
4.4. First Example of Euler-Lagrange equations: classical mechanics
As a first example of the use of The Euler-Lagrange equations, let the Lagrangian be L =
T − V = 12 mẋ2 − V (x) and let λ = t. Then ∂L ∂V
∂x = − ∂x = F , where we used the normal definition
of force as the derivative of the potential energy. Since v = ẋ, ∂L d ∂L d
∂ ẋ = mẋ, and dλ ∂ ẋ = dt mẋ = ma,
where the acceleration a = dv/dt. Thus the Euler-Lagrange equations found by extremizing the
action is just F = ma. This might seem like a lot of work to get something you already know, but
the beauty of the method is that it works in difficult situations and in difficult coordinate systems.
It is usually a lot easier to write down the kinetic and potential energy than it is to use the vector
form of F = ma in complicated situations.
4.5. Second example of Euler-Lagrange equations: Flat space geodesics
Now let’s extremize the 3-D flat space metric to see if the shortest distance between two points
R
is indeed a straight line! So s = ds, with
p p
ds = dx2 + dy 2 + dz 2 = 1 + ẏ 2 + ż 2 dx,
p
where we have chosen λ = x, and L = 1 + ẏ 2 + ẋ2 . The y Lagrangian equation thus reads:
d ∂L ∂L
− = 0,
dx ∂ ẏ ∂y
or
d 1 2 2 −1/2
(1 + ẏ + ż ) 2ẏ − 0 = 0,
dx 2
or
ẏ
p = constant.
1 + ẏ 2 + ż 2
For the z equation we find similarly,
ż
p = constant.
1 + ẏ 2 + ż 2
Dividing the y equation by pthe z equation we get ẏ/ż = constant, so ż = c1 ẏ. Substituting this into
the y equation, we get ẏ/ 1 + ẏ 2 + c21 ẏ 2 = constant. Since the only variable in this entire equation
is ẏ, solving this equation for ẏ will give a constant. Thus we find the Euler-Lagrange equations
for the extremal distance between two points are dy/dx = my , and dz/dx = mz , where my and
mz are some constants found by the boundary condition. Thus y = my x + by and z = mz x + bz ,
the equation for a straight line in 3-D. This again might seem like a lot of work to prove that the
shortest distance between two points is a straight line, but the method is general.
Before going on to extremize the invariant interval in a spacetime metric, I want to do the
last problem again and show you a useful trick. In the above we used x as the affine parameter.
– 28 –
R
Instead we could have used s itself. Thus we write our integral as s = ds, with the Lagrangian
L = 1! However we write this ‘one’ in a special way:
r
p dx dy dx p
ds = dx + dy + dz = ( )2 + ( )2 + ( )2 ds = ẋ2 + ẏ 2 + ż 2 ds,
2 2 2
ds ds ds
p
where now L = 1 = ẋ2 + ẏ 2 + ż 2 . As before in the Euler-Lagrange equations the term ∂L/∂x = 0,
and similarly for the y and z equations, so they reduce to:
!
d ẋ
p = 0,
ds ẋ2 + ẏ 2 + ż 2
and similarly for the y and z equations. But since L = 1 is a constant it comes out of the differential,
becoming
d ẋ
= dẋ/ds = 0.
ds L
This says that ẋ = dx/ds = constant, or x = mx s+bx , and similarly, y = my s+by , and z = mz s+bz .
Again the equation for a straight line, with s being the distance traveled along the line, but with
less algebra.
4.6. Geodesics in Minkowski spacetime
Next we use the Euler-Lagrange equations to extremize the invariant interval, ds2 = −dt2 +
dx2 + dy 2 + dz 2 , between two events in Minkowski spacetime. This will give us the equations
√
of motion of Special Relativity! Let’s use proper time pτ as the affine parameter: dτ = −ds2 .
R Rp R
We write s = dτ = dt2 − dx2 − dy 2 − dz 2 = ṫ2 − ẋ2 − ẏ 2 − żp
2 dτ , where ṫ = dt/dτ ,
ẋ = dx/dτ , etc. Since basically τ = s, here again the Lagrangian L = 1 = ṫ2 − ẋ2 − ẏ 2 − ż 2 .
Noting that ∂L/∂t = 0, and similarly for ∂L/∂x = 0, etc. the Euler-Lagrange equations
become
d ∂L
= 0,
dτ ∂ ṫ
and
d ∂L
= 0,
dτ ∂ ẋ
and similarly for y and z. Thus as we move along the path (varying τ to move along the geodesic)
we have 4 conserved quantities, ∂L ∂ ṫ
= constant, ∂L
∂ ẋ = constant, etc. Explicitly, these give ∂L/∂ ṫ =
ṫ/L = ṫ = constant, and ∂L/∂ ẋ = ẋ = constant, so our equations are ṫ = ct , ẋ = cx , ẏ = cy , and
ż = cz , or t = ct τ + t0 , x = cx τ + x0 , etc., where the ct ,cx , t0 , x0 , etc. are constants.
To find the values of the constant let’s simplify to only the x direction and let time t start
with τ = 0, so t = ct τ . Then x = cx τ + x0 = (cx /ct )t + x0 , and we recognize the combination
cx /ct = vx the velocity in the x direction. Thus cx = ct vx . Now evaluate the Lagrangian L = 1 =
– 29 –
p p p p p
ṫ2 − ẋ2 = c2t − c2x = c2t − c2t vx2 , or 1 = ct 1 − vx2 . Thus ct = 1/ 1 − vx2 = γ, the Lorentz
factor, and we have as our geodesics the equations of special relativity: motion in a straight line
with time dilation included: t = γτ , x = vx t + x0 , etc.
Note that it is possible to have all the constants vx = vy = vz = 0, so just t = τ (standing still
aging) is a geodesic.
In summary, what did here is extremize (in fact maximize) the proper time between two
events to find the geodesics. Thus the geodesic is that path for which the maximum time passes
on the wrist watch of the observer traveling that path. [It is maximum, rather than minium due
to the minus sign in the metric.] Note that this is basically the answer to twin paradox. The
twin that went out and then back did not travel a geodesic, they accelerated three times, while
the stay at home twin did not accelerate and therefore followed a geodesic. Thus we understand
why the stay at home twin ages more (in fact ages maximally!). Whenever someone accelerates,
they leave their geodesic and therefore are aging less! The solution to the twin paradox is also
easily understood using a spacetime diagram. Both twins start together at the origin. One twin
stays on Earth (worldline is geodesic going straight up). The other accelerates close to the speed
of light (worldline close to 450 ). The proper time for the speedy twin is very small; remember
the hyperbola of constant proper time. Halfway out, the speedy twin accelerates again (leaves the
geodesic again), and speeds home, meeting up with the first twin. From this it is clear that if one
minimized the proper time, it would require accelerating to the speed of light for half the time and
then coming back with a total proper time approaching 0. We see that actual geodesics maximize
the proper time.
4.7. Conserved quanties in the Euler-Lagrange formalism: Energy and Momentum
Note that as we did the derivations of the geodesics above we came across quantities that did
not change as we traced along the affine parameter, that is we found conserved quantities. This
is a general and important thing to watch for in using the Euler-Lagrange formalism. In general
there will be a conserved quantity whenever the Lagrangian L does not depend explicitly on one
of the variables. Actually there will be one conserved quantity per variable. This is easy to see.
To be general, let’s use the 4-vector notation xµ , where x0 = t, x1 = x, x2 = y, and x3 = z, so
µ runs from 0 to 3 and xµ can represent any of the spacetime variables. In this notation, all the
Euler-Lagrange equations can be written in one line:
∂L d ∂L
− = 0.
∂xµ dλ ∂dẋµ
∂L
So if L does not depend on one variable, call it xµ , we have ∂dxµ = 0, and the Euler-Lagrange
d ∂L
equation reads: dλ ∂ ẋµ = 0. Thus along the geodesic (running over values of the affine parameter
∂L
λ), the quantity ∂ ẋµ is a constant.
– 30 –
In general we define pµ , the conjugate momentum, of a variable xµ , as
∂L
pµ = .
∂ ẋµ
Thus we see that if the Lagrangian does not depend explicitly on a variable, then that variable’s
conjugate momentum is conserved. This pµ may be an actual momentum, but it could be some
other conserved quantity. If one takes L = 12 mẋ2 , then in fact px = mvx , which is conserved since
L does not depend upon x. If instead we took L = 12 mẋ2 − V (x), then the conjugate momentum
is still px = mvx , but it is not conserved along the geodesic as the affine parameter (t) changes. Of
course this is because there is a force due to the potential energy.
For the Minkowski metric, let’s see what these conjugate momenta are. We saw above that the
metric did not depend explicitly on any of the xµ ; t,x, y, or z, thus we expect to have 4 conserved
quantities. Consider the t equation with affine parameter proper time τ . We have pt = ∂L ∂ ṫ
= ṫ.
Likewise for the x equation we find px = ẋ, py = ẏ, etc. During the above calculation we found
that the constant for the Euler-Lagrange equations (which we called ct ) was equal to the Lorentz
factor γ. Thus we see that
p0 = pt = γ = E/m,
where we noticed that in special relativity the energy of a particle is E = mγ, where m is the rest
mass. Thus the conserved quantity associated with the t variable is nothing other than the energy.
(It is energy per unit mass actually, but since pt is a constant we can multiply by another constant
m and still have a constant.)
We also previously found the momentum conjugate to the x variable, px = ∂L ∂ ẋ = ẋ = cx .

Using cx /ct = v from before, we have px = vx ct = vx pt = vx γ. But in special relativity the
momentum is just mvγ, so see that px = Px /m, is the momentum per unit mass. (We call the
actual momentum P to distinguish it from the momentum per unit mass.) Thus the 4 conserved
quantities we discovered are just the energy and 3 components of momentum.
We can find the relation between these quantities from our definition of the Lagrangian: L =
p
1 = ṫ2 − ẋ2 − ẏ 2 − ż 2 . Substituting in pt = ṫ = E/m, px = ẋ = Px /m, etc. we find
q
1 = (E/m)2 − (Px /m)2 − (Py /m)2 − (Py /m)2 .
Squaring, multiplying through by m2 , and using P 2 = Px2 + Py2 + Pz2 , we find the well known result
E 2 = P 2 + m2 . Finally using units to put back the c’s, E 2 = P 2 c2 + mc4 , which reduces to the
famous E = mc2 in the limit of zero velocity (zero momentum).
– 31 –
5. Equivalence Principle, Gravitational Redshift and Geodesics of the

Schwarzschild Metric
5.1. Gravitational Redshift from the Schwarzschild metric
We already discussed the curvature of time in the Schwarzschild metric. Clocks run slower
in a gravitation field (at smaller values of r). This is one of the more measurable effects of GR.
Measurements are possible because atomic transitions act like small clocks which send messengers
(photons) to us here at Earth. Recall the proper time in the Schwarzschild metric:
r
2GM
dτ = dt 1 − ,
rc2
where dτ is tick of a clock measured at r, and dt is the length of that same tick measured far away
(r → ∞).
Now consider some atomic transition, for example, the n = 3 to n = 2 transition of atomic
Hydrogen. This is called the H-alpha transition and results in emission of a photon with energy
Eγ = 13.6eV( 312 − 212 ) = 1.89 eV. Using λ = c/ν, and Eγ = hν, we find the corresponding wavelength
Hα : 6563.5 Angstroms. Now in the frame at r, the photon has a fixed frequency ν0 , and each
oscillation is like a little clock ticking with period dτ0 = 1/ν0 . Thus the time between crests in
the outgoing photon measures dτ . That same photon travels out to us at r = ∞. Then that same
p
period is observed as dt. Thus the dτ∞ ≡ dτobs = dt = dτ0 / 1 − 2GM/rc2 is longer. Using the
definitions above of ν and λ we see that the frequency decreases by the same factor and the more
easily observed wavelength is longer at when observed at infinity than the emitted wavelength at r
by
λ0 λ0
λobs = q =p ,
1 − 2GM 1 − rrS
rc2
where we used the definition of the Schwarzschild radius rS .
The redshift of an emission line is defined as

∆λ λobs − λ0 rS
z= = = (1 − )−1/2 − 1,
λ0 λ0 r
where λ0 is the emitted wavelength in the rest frame of the atom. Note for small values of
2GM/rc2 = rS /r, this expression can be Taylor expanded as z ≈ GM/rc2 = 12 rS /r.
This gravitational redshift is totally different from either the normal Doppler shift redshift
caused by an atom moving away from the observer, or the cosmological redshift caused by the
expanding Universe. Let’s see how big the effect is by calculating it for some typical astronomical
objects. First how about emission from the surface of the Sun? We expect the wavelengths recieved
by us on Earth to be longer than the emitted wavelengths since the emitted photons have to climb
out of the potential well of the Sun where clocks run slower. Using the approximation above
– 32 –
z ≈ 12 3km/7 × 105 km ≈ 2 × 10−6 . This is not a very big effect; a clock on the surface of the Sun
loses about 30 seconds/year. For H-alpha the line redshift would be only 0.014 Angstroms.
How about for a white dwarf star? The mass of a white dwarf is about 0.6M , and the radius
is about 5500 km, near to the Earth radius. Thus z ≈ 12 (.6)(3km)/5500km ≈ 1.6 × 10−4 , and the
wavelength of H-alpha emission recieved on Earth would be 6563.5(1 + z) = 6564.6 Angstroms,
quite easily distinguishable from local H-alpha lines. This result agrees well with the measurements
from the nearby white dwarf star Sirius B. A clock on the surface of Sirius B would lose about 0.6
second/hour.
Next consider the case of a neutron star, with M = 1.4M and R = 10 km. Now rS =
(1.4)(3km) = 4.2 km, so really should not use the approximation, but calculate z = [1−(4.2km/10km)]−1/2 −
1 = 0.313. This is big shift: the H-alpha line would appear at 8618.3 Angstroms and clocks would
run 31% slower.
Finally what about light emitted from the Schwarzschild radius of a black hole? With r = rS ,
z → ∞! Light would be invisible since it would be redshifted to infinitely low energies. Clocks
sitting at that location would appear to have stopped!
5.2. Light bending and the Equivalence Principle
To find proper answers in GR we calculate using the metric, in our case the Schwarzschild
metric, but there is an interesting, more intuitive way that uses the Equivalence Principle directly.
Let’s look briefly at two examples. Imagine someone in a falling elevator. The equivalence principle
says that experiments done in this falling frame will give the same results as experiments done
floating in free space far from any massive object. Suppose the guy in the falling elevator shines a
laser beam horizonally from one side of the elevator to the other. That light must follow geodesics
and therefore go “straight” across the elevator; if the beam leaves from a height of 3 feet above the
elevator floor, it must arrive on the other side also three feet above the floor. From the point of view
of the elevator man, the light must travel exactly horizontally. But now think of someone watching
through the (glass) walls of the elevator. The man shoots the laser from a height of 3 feet above the
floor, but then while the light is traveling across, the elevator continues to fall. Thus when the light
hits the opposite wall 3 feet above the floor, the wall has moved down. Therefore the light cannot
go straight in the Earth frame. The light must also fall. This gedanken experiment shows the light
must fall in a gravitational field in the same way as everything else. Of course, from the point of
view of GR, light is just following the geodesic through a curved spacetime; we will calculate those
soon, but here we forsee what that calculation will give. In particular, light from distant stars
travelling near the limb of the Sun must be “deflected” towards the Sun. Sir Author Eddington
did this experiment in 1919, and it garnered world wide fame for Albert Einstein, because it was
one of the first proofs of Einstein’s GR theory. In the homework, I am asking you to calculate how
far the light falls in a gravitational field, that is how much it bends.
– 33 –
5.3. Gravitational Redshift again
The equivalence principle can also be used to show that light must gravitationally redshift.
Again consider our man in a falling elevator. This time he shoots his laser from the floor to the
ceiling.
First think about it from the point of view of someone outside watching through the glass
walls. In the Earth frame the freely falling elevator picks up speed while the light is in transit,
and so there is Doppler shift caused by the speeding elevator ceiling. The time the photons spend
in transit is t = h/c, where h is the height of the elevator, and the speed increase of the elevator
is ∆v = gt = gh/c, where g = GMEarth /r2 = 9.8m/s2 . We could find the expected Doppler
shift from the relativistically correct formula νobs = ν0 (1 + ∆v)/(1 − ∆v), or since λ ∝ ν −1 ,
p
p
λobs = λ0 (1 − ∆v)/(1 + ∆v), but since a Taylor expansion of this is good enough for the slow
speeds involved here we have: λobs ≈ λ0 (1 − 12 ∆v)(1 − 12 ∆v) ≈ λ0 (1 − ∆v), or λobs − λ0 = −∆vλ0 ,
or z ≈ −∆v. Thus in the Earth frame we expect the light beam to be blue shifted by ∆λ/λ0 =
−∆v/c = −gh/c2 , where I put the c0 s back in. However, the equivalence principle says that this
cannot be! The man in the elevator can’t tell whether he is falling or floating freely in deep space,
so he expects correctly that the frequency of light recieved at the elevator’s ceiling is the same as
the frequency he sent from the floor. Thus there must be another effect to cancel this one! The
light traveling up must be gravitationally redshifted to counterbalance this effect. Thus there must
be a factor of ∆λ/λ0 = v/c = gh/c2 = GM h/r2 c2 . This is not quite the result we got before from
the metric, since there is a factor of r−2 rather than r−1 . Before we found the redshift between
r = r0 and far away (r = ∞) and here we found it only between r0 and r. Thus to compare to
this equivalent principle result we should use our previous calculation twice. We had z ≈ 12 rs /r
as the gravitational redshift between r and ∞, so to find the redshift betweent r = rEarth and
r = rEarth + h we should find ∆z ≈ 12 rs [1/rEarth − 1/(rEarth + h)], or saying that h is much less
than rEarth this gives ∆z ≈ 12 rs h/rEarth 2 = GM h/r2 c2 , precisely what we found above from the
equivalence principle.
5.4. Geodesics of Schwarzschild metric from Euler-Langrange
Let’s return to our work on geodesics and apply the Euler-Lagrange equations to find the
geodesics of the Schwarzschild metric. Remember that the Schwarzschild metric is the unique
metric around stationary, spherically symmetric, uncharged objects, so what these geodesics do is
tell us how thing move around the Earth, around the Sun, and around uncharged, non-spinning
black holes. These are the General Relativistic extension of Newton’s and Kepler’s laws of motion.
Recall the Schwarzschild metric is
2GM −1 2

2 2GM 2
ds = − 1 − dt + 1 − dr + r2 dθ2 + r2 sin2 θdφ2 .
rc2 rc2
– 34 –
R
As above, we set c = 1, and take affine paramter λ = s = τ , and extremize s = Ldτ , with
" #1/2
ṙ2

2GM
L=1= 1− ṫ2 − − r2 θ̇2 − r2 sin2 θφ̇2
r 1 − 2GMr
The Euler-Lagrange equation for t is then

d ∂L ∂L
− = 0,
dτ ∂ ṫ ∂t
which since ∂L/∂t = 0, implies there is a conserved quantity we will call energy per unit mass:
∂L E
≡ .
∂ ṫ m
Calculating,
∂L 1 2GM
= [· · · ]−1/2 (1 − )2ṫ,
∂ ṫ 2 r
where we abbreviated L = [· · · ]1/2 . Using L = 1, we find our t equation

2GM E
1− ṫ = .
r m
We don’t yet know that the constant has anything to with energy, but we call it E/m, because
of our experience with the Minkowski metric. For r → ∞, the Schwarzschild metric goes to the
Minkowski metric and for the Minkowski metric ṫ = E/m.
Next we find the φ equation:

d ∂L ∂L
= = 0,
dτ ∂ φ̇ ∂φ
so again we have a conserved quantity pφ = ∂L/∂ φ̇ ≡ −l/m, where we will call this conserved
quantity the angular momentum per unit mass. Recall Noether’s theorem which says if physics is
unchanged by a rotation then angular momentum is conserved. Since the metric does not depend
explicitly on the angle φ, we get that result here. Doing the differentiation we find
∂L 1
= [· · · ]−1/2 (−r2 sin2 θ2φ̇) = −r2 sin2 θφ̇.
∂ φ̇ 2
Thus our φ equation reads

l
= r2 sin2 θφ̇.
m
Note that it makes sense that we called the constant of motion l/m, since this matches the normal
definition of angular momentum, l = ~r × P~ , with v = r sin θφ̇.
Next, we consider the θ equation. Here we find that

d ∂L ∂L
= 6= 0,
dτ ∂ θ̇ ∂θ
– 35 –
thus we do not have a conserved quantity for this equation. We find

∂L 1
= [· · · ]−1/2 (−r2 φ̇2 2 sin θ cos θ) = −r2 φ̇2 sin θ cos θ,
∂θ 2
and
∂L 1
= [· · · ]−1/2 (−r2 2θ̇) = −r2 θ̇.
∂ θ̇ 2
Thus our θ equation reads:
d 2
(r θ̇) = r2 φ̇2 sin θ cos θ.
dτ
Finally we come do the r equation, which is kind of messy because of all the explicit r depen-
dence in the metric. However, we don’t have to do it, because we can get the fourth equation we
need to specify the equations of motion from our definition of the Lagrangian, L2 = 1:
2GM 2 ṙ2
1 = (1 − )ṫ − 2GM
− r2 θ̇2 − r2 sin2 θφ̇2 .
r (1 − r )
Since our object and metric are spherically symmetric we can simplify things by only considering
motion in the equatorial plane (θ = π/2, and θ̇ = 0). Of course we have to remember that we made
this assumption later when we use our equations! If we try to consider motion that has a changing
θ, or which is not in this plane we would need to come back to the equation above. In this case
then, the equation L = 1 becomes:
2GM 2 ṙ2
1 = (1 − )ṫ − 2GM
− r2 φ̇2 .
r (1 − r )
Now eliminate variables other than r, using the constants of motion we have from the above
equations: l/m = r2 φ̇, and E/m = (1 − 2GM
r )ṫ, to get:
E2 ṙ2 l2
1= − − .
m2 (1 − 2GM
r ) (1 − 2GMr )
m2 r2
We can write this in a nicer form by solving for mṙ2 ,
E2 l2

2GM
mṙ2 = − m+ (1 − ),
m mr2 r
Remembering the definition of ṙ and using dimensional analysis to put back the c’s, we can write
this as: 2
E2 l2

dr 2GM 2
m = − 1− mc + .
dτ mc2 rc2 mr2
Notice that, as I mentioned in my handwaving description, a step in proper time dτ forces a step
dr in the r direction. Thus this geodesic equation shows that things fall due to the spacetime
curvature of the metric. We will look at these geodesic equations in some detail, but for now let’s
just take one limit of this last equation.
– 36 –
Suppose the angular momentum l = 0, which we might expect for radial infall towards a
spherical mass. Our equation then is:
E2

dr 2 2GM
m( ) = − 1− mc2 = 0.
dτ mc2 rc2
Next consider a case where you start at rest far from the object so that at proper time τ = 0,
m(dr/dτ )2 = 0, and r → ∞. Our equation becomes:
E2 2GM
− (1 − )mc2 = 0,
mc2 ∞
or E 2 /mc2 = mc2 , or simply E = mc2 ! So at τ = 0 the total energy is just E = mc2 . In Newtonian
mechanics the energy at infinity is usually defined as E = 0. Isn’t it nice how these important
results are just built right into the general relativistic view of spacetime. Also since energy E is
conserved along geodesics we know that E = m always. This E is not the Newtonian energy; it is
the conserved quantity, which is better than the sum of 12 mv 2 + V . Finally note that if we would
have started with some velocity at r = ∞ then E > mc2 but it still would have been conserved.
At later times during this radial infall from rest, our equation becomes:
dr 2 E2 2GM 2GM m
m( ) = − mc2 + mc2 = .
dτ mc2 rc2 r
That is simply
1 dr GM m
m( )2 − = 0,
2 dτ r
which looks remarkably like the Newtonian case 12 mv 2 − GM m/r = 0! But this is not the same
equation. It is fully relativistic and the time derivative is τ not t (remember t = γτ in SR). In
general you integrate these 4 equations to get the complete picture of motion near the Earth, Sun,
or Black Hole. We will come back to these soon.
– 37 –
6. Distances and Times Around a Black Hole
Suppose you have a very powerful spaceship and fly near to a black hole. Can you notice
anything different from deep space? (Actually the calculations and results we get here are also true
around the Earth on a very small scale.)
Suppose we fly close to a small black hole of mass M = 3M . (Actually, the smallest black hole
we expect to find in nature is around 3M , so we pick this number.) We know the Schwarzschild
radius for this black hole is about 3 × 2.95 km = 8.85km, so we are careful to stay farther away
from the hole than this. Let’s suppose we fly all the way around the hole and measure a distance
around (circumference) of C = 2π30km = 188.5 km. How far are we from the hole? Naively, we
expect we are 30 km from the hole, but we should check this by using the Schwarzschild metric:
2GM −1 2

2 2GM 2
rc2 rc2
To find the proper distance in the θ direction, we set dt = dr = dφ = 0, and the find the proper
length √
dlθ = ds2 = rdθ.
Note that this is the same as the proper distance in flat space! So there is no curvature in the θ
(or φ) direction! To find the distance around the black hole we integrate
Z 2π Z π
C= dlθ = 2 rdθ = 2πr,
0 0
just as in flat space. Thus if we have measured the distance around as 2π 30 km, we know we are
at radial coordinate r = 30 km.
But does that mean we are 30 km from the center of the hole? We have to use the metric
again to find out. This time we want the radial direction, so we set dt = dθ = dφ = 0, and we get
√ dr
dlr = ds2 = .
2GM 1/2

1− r
To find the distance from radial coordinate r1 to radial r2 we integrate from r1 to r2 . If we set
r1 = 8.85 km and r2 = 30 km, we can find our distance to the black hole horizon itself.
2GM −1/2
Z r2
∆lr = dr 1 − .
r1 r
Defining r
2GM p
Ai ≡ 1− = 1 − rS /ri ,
ri c2
we evaluate the integral as
r2
r2 A2 + r2 − rS /2
Z
rS
∆lr = dlr = r2 A2 − r1 A1 + ln ,
r1 2 r1 A1 + r1 − rS /2
– 38 –
where we used the Schwarzschild radius rS = 2GM/c2 = 2.95326M/M km.
Using this formula we can find the distances. For example starting at r=30 km, and moving
in to r = 20 km, we we naively expect to move 10 km, but we actually move 12.51 km. This is
very weird, since after moving 12.51 km inward from r = 30km, the distance around the black hole
would be 2π20 km not 2π17.49 km. Thus we are seeing directly the curving of space. It is just like
the example of the two surveyors, but now the curvature is not in any direction we can experience!
In the above, the way we tell what value of r we are at is to travel around the hole and use the
circumference. Using the above formula we also find the distance from r1 = 10 km to r2 = 30 is
29.50 km, and the distance from the horizon at rS = 8.86 km to r = 30 km is 35.98 km.
Fig. 4.— Figure for Chapter 6: Embedding diagram of curved space near a black hole
We can visualize the spatial curvature around a black hole by an embedding diagram. The
key in this diagram is that the radial coordinate is just the straight 3-D distance, while the proper
distance is measured along the curved surface. We continue the embedding diagram even inside
the black hole horizon, which turns out to be correct, though we will have to think carefully before
understanding why.
Finally, note that this metric and embedding diagram work not only for black holes, but also
for the Earth! If the distance to the center of the Earth is rEarth , then the distance around the
Earth is really not equal to 2πrEarth ! Can you figure the difference?
6.1. Can you fall into a black hole?
We found above that the space is flat in the θ and φ directions, but curved in the radial
p
direction with proper distance dlr = dr/ 1 − rS /r. Looking at this again we see that moving a
p
small proper distance dlr implies moving a smaller radial coordinate distance dr = dlr 1 − rS /r.
This seems fine, but look at what happens right near the Schwarzschild radius rS . When r → rS ,
then dr → 0, that is you are not moving at all in the radial coordinate! Does this mean that you
can’t get into the black hole? No, because it is a square-root singularity, which implies that it is
– 39 –
integrable.
√
Consider the area under the curve y = 1/ x. As x → 0, y → ∞, but still the area is
R x0 −1/2 √
A= 0 x dx = 2x1/2 |x0 0 = 2 x0 which is finite. Likewise dl/dr = (1 − rs /r)−1/2 is integrable
and we gave the formula for the integration above. We therefore find that the proper distance from
r = 30 km to rS is 35.98 km. So just because a function goes to infinity in a region of interest
doesn’t always mean there is a problem. Of course, sometimes it does mean there is a problem as
we shall see.
6.2. Time to fall into a black hole
Next let’s calculate the time to fall into a black hole. There are several ways that we will do
this. First, let’s return to our geodesics for radial infall. Remember we considered the case with
angular momentum l = 0, and found that starting at rest from r = ∞ meant the conserved energy
was E = mc2 , and the equation of motion is
1 dr GM m
m( )2 − = 0.
2 dτ r
Let’s find the proper time τ it takes for this freely falling object to go from r0 = 30km to r =
rS = 8.85 km, the Schwarzchild radius of a 3 M black hole. This is the time as measured by the
wristwatch of the falling guy. We take the negative square root of this equation since we want to
fall in (r decreasing):
dr √
= − 2GM r−1/2 ,
dτ
or Z r Z τ √
1/2
drr = − 2GM dτ,
30 0
or √
2 3/2 r
r |r0 = −τ 2GM ,
3
or
2 p 3/2
τ= (1/ 2GM/c2 )(r0 − r3/2 ).
3c
(Note I put the c back in to get the units right and make calculation easier.) So to go from
r0 = 30 km to r = rS = 8.85 km in our 3M black hole takes τ = 23 (1/ (2.95)(3)km)((30km)3/2 −
p
(8.85km)3/2 )/3 × 105 km/s = 1.03 × 10−4 s. Thus it takes about 0.1 millisecond! It is not clear yet,
but this same equation works inside the black hole, so we can also find how long the falling guy has
to live before hitting the singularity at the center. Just taking r = 0 in the above equation gives
3/2
2 r0
τ= √ = 0.124ms.
3 2GM
Thus our guy gets only an extra 0.021 ms to live inside the black hole!
– 40 –
Now this is the time freely falling starting at infinity. We could also find the time to fall in
if we started from rest at r = 30 km. For this we go back to the more general geodesic equation
before plugging in E = mc2 ,
E2

dr 2 2GM
m( ) = − 1− mc2 = 0.
dτ mc2 rc2
Now we want the boundary conditions starting at τ = 0: dr/dτ = 0 at r = r0 . Plugging this

p
in, we find the conserved energy is E = ±mc2 1 − 2GM/r0 c2 . Note this is the quantity that is
conserved along the geodesic, and is NOT E = mc2 . For the case of falling in we take the negative
square root, then integrate the above equation as before. This can be done and we find a more
complicated formula. The time to fall into r = 0 is a little simplier:
r
π r0
τ = r0 ,
2 rS
for a time of 0.29 ms, about three times longer than when you start falling from far away.
Now what does this look like to someone watching from far away. They don’t use the proper
time τ , but use the coordinate for “far away” time t. We can convert the equations above to
coordinate time t by using our time geodesic equation: 1 − 2GM E

r ṫ = m , or dt/dτ = (E/m)/(1 −
rS /r). Then
m rS
dr/dt = (dr/dτ )/(dt/dτ ) = − (rS /r)1/2 (1 − ).
E r
This equation can be solved as before, but we will find some trouble in doing it as we get close to
rS . Consider that limit, r → rS . Then 1 − rS /r → 0, and a tiny step in dr means an infinite step
in dt. This is different than the case before with proper distance because this is not an integrable
square root singularity. This is real infinity that cannot be integrated over. In fact, if you try to do
the integral you will find you never get to r = rS ! Time slows down and motion ceases. Everything
hangs up at the horizon.
– 41 –
7. Shooting Light Rays into Black Holes, Inside a Black Hole, Orbits in the
Schwarzschild Metric, Effective potentials
7.1. Shooting Light into a Black Hole
Let’s calculate radial motion into black holes another way, a way that is often very useful,
because it bypasses the geodesic equations. Let’s calculate how long it takes to shoot a ray of light
into black hole.
For light we know the invarient interval ds2 = 0, that is the metric distance, aka proper time,
is 0. So for light itself how long does it take to get into a black hole? It takes the same time it
take light to go anywhere, zero! From light’s point of view no time ever passes since dτ = 0 always.
OK, that means we set the metric equal to zero for light. This is the very useful trick. Considering
radial infall, we can also set dφ = dθ = 0, and get
dr
dt = .
(1 − rS /r)
Integrating both sides from t = 0 at r = r0 to t = t at r = r, we find

r0 − rS
t = r0 − r + rS ln .
r − rS
From our previous work we expect the time for light go from r = 30 km to r = rS to be less than a
millisecond. But plugging into the above equation we get a factor ln((30 − 8.85)/(8.85 − 8.85)) →
ln(∞) → ∞. There is a logarithmic divergence and again we see we never get into the black hole!
Of course, from the point of view of the light ray, or the falling guy, they get in and are crushed
within a millisecond. It is just that time viewed from far away runs differently at the horizon of a
black hole. The problem is with the use of this far away time. Also, while it is true that everything
that ever falls into a black hole seems to “hang up” at the horizon, it is not that case that someone
looking closely at a black hole sees all that junk. Remember the light coming to you from the
falling objects redshifts to infinity and therefore those objects become invisible very quickly.
As an interesting aside consider what would happen if you tried to lower yourself slowly into a
black hole on a very strong rope. I’m not going to do the calculation but the effective acceleration
of gravity you would feel would increase without limit. The effective g, which is GM/r2 = 9.8 m/s2
here on Earth becomes
g = (GM/r2 )(1 − rS /r)−1/2 .
Thus the force becomes infinitely strong at r = rS , the rope will break and you will fall to your
death.
– 42 –
7.2. Inside the Black Hole
The Schwarzchild metric is the solution to Einstein’s GR equations in the vacuum around a
spherically symmetric object. Thus we expect these to work even if the object is smaller than rS .
What happens at and inside r = rS = 2GM/c2 ? The metric written in the r and t coordinates is
bad at r = rS , but actually it is OK inside.
2GM −1 2

2 2GM 2
rc2 rc2
First consider lightcones, the causally connected past and future. How did we find these in Special
Relativity? Considering just one space and one time dimension: ds2 = −dt2 + dx2 = 0, for light.
This defines the null geodesics. Solving we find dt = ±dr, or dt/dr = ±1, which implies the
lightcones are lines of 450 in a spacetime diagram. Remember also that above the 450 lines is
the timelike ds2 < 0 future, while outside the lines is the ds2 > 0 spacelike elsewhere. In Special
Relativity all lightcones have these 450 lines.
How about in General Relativity. Again we set ds2 = 0 to find the null geodesics, giving
−dt2 (1
− rS /r) + dr2 /(1 − rS /r) = 0, or
dt 1
=± .
dr 1 − rrS
Thus the lightcones are not 450 lines. As r → ∞, rS /r → 0, so dt/dr → ±1, and the lightcones are
at 450 , but closer to the spherical mass the angles are smaller, dt/dr > ±1. Thus the lightcones
squeeze-up. Thus the path of light as it travels towards a spherical mass in a spacetime diagram
is not on a 450 line, but on a curved line that gets more steep as it approaches the Schwarzschild
radius.
We can calculate the angle of the lightcone at any value of r very simply. Just use
dt
= tan θ.
dr
Thus at 30 km from a 3M mass hole the angle of the lightcone is dt/dr = 1/(1 − 8.85/30) = 1.42,
or θ = 550 . This angle goes to 900 at r = rS . So we see that right at r = rS , dt/dr → ∞, and no
progress in r is possible! Not even light can make it into the black when far-away time is used as
the coordinate. How does this jive with the fact that we calculated things can fall into black holes
in just milliseconds? The point is that t is time measured by someone far away, not the traveler.
Now look what happens to the metric when r < rS . Notice that since rs /r > 1, the terms
(1 − rS /r) become less than zero.
ds2 = −dt2 (1 − rS /r) + dr2 /(1 − rS /r) + r2 dθ2 + r2 sin2 θdφ2 .
Thus we see that the t coordinate and the r coordinate terms switch signs! Remember that in GR
r is not a distance, it is a coordinate. You have to use the metric to find distances. Same with
– 43 –
Fig. 5.— Figures for Chapter 7: (a) Path of light as it approaches a black hole and (b) Lightcones
around and inside a black hole
t. The minus sign is how we recognize the time coordinate in GR, so this means that r becomes
timelike and t becomes spacelike! If you try to find the lightcones inside the black hole you see
they flip over sideways!
– 44 –
The same force that moves eveyone towards the future now moves things towards r = 0, which
is the future lightcone! This one way of understanding the reason that once inside the black hole
you can’t get out. There is no force in nature that can move you backward in time. Thus inside a
hole there is no force in nature that can move something towards larger r! Note that while proper
time increases, t actually decreases. This is not a problem since t is no longer a time dimension. It
is a space dimension; r is the time. We can see this explicitly by considering two points inside, say
near r = 2 km and r = 2.1 km, so dr = 0.1 km. Take dt = 0 and calculate the invariant interval:
ds2 = +dr2 /(1 − 8.85/2) = −0.0029 km < 0, that is the interval is timelike, and thus these points
can be causally connected! Similarly, two events with dr = 0 and dt 6= 0 have ds2 > 0 and therefore
are spacelike separated and cannot be causally connected. Moving straight up in the spacetimem
diagram is not allowed.
The angle of the lightcone just inside rS is still 900 , but rather than being squeezed, it is now
wide open. As you move in towards r = 0, the lightcone closes up, but continues to point toward
r = 0. Inside dr/dt = ±(1 − rS /r). As r → 0, dr/dt → ±∞, or dt/dr → 0, and the cone squeezes
in pointing towards the center of the hole. So while at first when you are inside the hole there some
freedom of movement, in the end you are directed directly at the center.
7.3. Orbits in the Schwarzschild metric
Let’s go back to our geodesic equations and find the orbits around a black hole or other spherical
object. So far we only considered radial orbits. To extend this to circular and eliptical orbits we will
use the method of effective potentials, which allows one discover the main types of orbits and the
main points of interest without doing the hard work of actually solving the differential equations.
This is very general technique and worth learning.
7.4. Effective Potential for Newtonian Orbits
First let’s do it in Newtonian mechanics, by writing the total energy E = T +V = 12 mv 2 −GM/r

in spherical coordinates. Recall that in 3-D non-relativistic mechanics, v 2 = ẋ2 + ẏ 2 + ż 2 , where
ẋ = dx/dt, etc., and z = r cos θ, x = r sin θ cos φ, y = r sin θ sin φ. Let’s simplify by considering
motion in the x-y plane so z = 0, θ = π/2, and θ̇ = 0. Then ż = 0, ẋ = ṙ cos φ − r sin φφ̇, and ẏ =
ṙ sin φ + r cos φφ̇. Substituting this into the formula for kinetic energy and using sin2 φ + cos2 φ = 1,
we find:
1
T = m(ṙ2 + r2 φ̇2 ).
2
Next defining the angular momentum l = mvφ r = mφ̇r2 , so rφ̇ = l/(mr), we can write the total
energy as
l2

1 2 GM m
E = m ṙ + 2 2 − ,
2 m r r
– 45 –
or
1 dr
E = m( )2 + Vef f ,
2 dt
with
GM m l2
Vef f = − + 2 .
r 2r m
Thus we have written the equation of motion as a one dimensional equation in the radial coordinate,
r, with energy E = T + Vef f . The value of doing this is that we are very familar with the the
solutions of a one dimensional equation for a particle moving in a potential well with the shape
Vef f . Depending on the total energy, the particle may escape the well and travel off to infinity, or
it might be trapped in the well and oscillate back and forth in the bottom of the well. For small
enough energy it might be at rest at the bottom. So we now think of the radial coordinate r as the
one dimension and can easily understand the possible Newtonian orbits around a spherical object.
To discover the possible orbits we draw a plot of r vs. the effective potential. We see that Vef f
drops from infinity at r = 0, reaching a minimum at a value of r = rmin , and then rises slowly to
Vef f = 0 at r = ∞. The exact shape depends upon the constants (conserved quantities) l, and E,
as well as M .
The easiest solution to this one dimensional problem is the particle as rest at the minimum.
The particle just sits there, which means r = constant. This is the circular orbit of Kepler’s laws.
The radius of the circular orbit is found by finding the minimum of the effective potential by solving
dVef f /dr = 0. This gives 0 = GM m/r2 − l2 /(mr3 ), or r = GM/v 2 , where we substituted back in
l = mvr before solving for r. Note that we did not substitute in for l before taking the derivative
with respect to r, since l is a conserved quantity and therefore constant along the orbit. This is
the result you get from F = ma, aka GM m/r2 = mv 2 /r. In this case, the total energy is at its
minimum which is less than zero. Emin = −mv 2 + 12 mv 2 = − 12 mv 2 . Note if you had a different
value of v, that would imply a different value of l, and a different rcirc .
Another obvious solution is the particle having E < 0, so the particle oscillates around the
bottom between turning points r1 and r2 where E = 12 mṙ2 + Vef f . These orbits are the elliptical
Kepler orbits, and the position and speed of the orbiting object at turning points can be found from
the equation above. There are also orbits that have E > 0 which can come in from infinity, reach
a point of closest approach and then return to infinity. These are the hyperbolic Kepler orbits.
Finally, a particle with E = 0 will do a similar similar thing, but the orbit will be parabolic.
7.5. Effective Potential for Schwarzschild Orbits
Let’s apply the same effective potential technique to the full r geodesic equation we derived
earlier. 2
E2 l2

dr rS 2
m = − 1− mc + ,
dτ mc2 r mr2
– 46 –
Fig. 6.— Figures for Chapter 7: (a) Effective potential for Newtonian Potential and (b) Effective
potential for Schwarzschild Geodesics
where rS = 2GM/c2 . Here the effective potential is
l2 l2 rS l2

rS 2 rS
Vef f = 1 − mc + = m − m + − ,
r mr2 r mr2 mr3
– 47 –
and keep in mind that we are using the proper time τ , not t as the time variable, and l is the
angular momentum and E is the total energy. We can spot the different types of orbits as we did
for the Newtonian effective potential. Now we see there are five different types of orbits.
1. At the minimum we have a bound circular orbit. The radius is found from dVef f /dr = 0 as
before.
2. We also have the bound orbits with turning points r1 and r2 like in the Newtonian case, but
here these turn out to not be elliptical! The orbits don’t close. If we do the perhelion advance
of Mercury we will show this. The time to between r1 and r2 and back is more than the time
to go around φ = 3600 , so the perhelion advances.
3. Next, as for Newtonian hyperbolic orbits, we have orbits that start at r = ∞, come in, reach
a distance of closest approach, and go out again. Close examination shows these are close to,
but not exactly the same as, the Newtonian case.
4. Now look at maximum point on top of the hill. We find this point also when we set dVef f /dr =
0, since the slope is zero here. If you set a particle there and carefully balanced it, it would
stay there. Thus this is another circular orbit, but it is unstable. A tiny perturbation outward
and the particle will escape to infinity. A tiny perturbation inward and it will fall in the hole.
5. Finally, there is another new type of orbit not found in the Newtonian case, the capture orbit.
If the energy is high enough, the particle goes over the high point and then plunges down
into the hole, never to return. In the Newtonian case a particle can never be captured in this
way due to gravity alone, even though this is how we think when, for example, a comet hits
the Sun. Hitting the Sun’s surface invokes non-gravitational forces; a point mass object could
never capture anything in Newtonian mechanics.
Let’s find the radii of the circular orbits. We can write

l2 rS m rS l2
Vef f = m + − − ,
mr2 r mr3
and differentiate and set to zero.
dVef f 2l2 rS m 3rS l2
=− 3 + 2 + = 0.
dr mr r mr4
We can simplify this to a quadratic equation in r by multiplying through by mr4 :
rS m2 c2 r2 − 2l2 r + 3rs l2 = 0,
which has two solutions: r !

l2 3r2 m2 c2
r± = 1± 1− S 2 ,
rS m2 c2 l
– 48 –
where I’ve put the c’s back in for fun and reference.
Note that if the quantity in the square root is negative we don’t have a real solution. Thus
there is no circular orbit in that case. This happens when the angular momentum is too small,
i.e. there are only capture orbits. If the quantity in the square root in positive, then we have two
solutions, that is, two circular orbits at different radii, as we saw in the plot. Thus we have two
circular orbit solutions if l2 > 3rS2 m2 c2 , and none if not.
When there are two orbits we can tell if the orbits are stable or unstable by taking the 2nd
derivative. If d2 Vef f /dr2 > 0, the the curvature of the effective potential is positive and it is a
minimum. This means the orbit is stable. On the other hand if if d2 Vef f /dr2 < 0, it means there is
a maximum in the effective potential at that point, and the orbit is unstable. We can do the math,
but we saw before from the picture that the smaller solution (the one with the minus square root)
was unstable, and the larger solution with the plus square is a stable orbit.
From the above discussion we see that the smallest possible stable circular orbit will happen
when the square root term vanishes. This happens when 1 = 3rS m2 c2 /l2 = 0, or l2 = 3rS2 m2 c2 .
The value of the radius for this value of the angular momentum can be found just by plugging this
value of l into the formula for r± , and we find that the minimum stable circular radius is
l2
rmin = = 3rS ,
rS m2 c2
that is just exactly three times the Schwarzschild radius. Actually we should check that this orbit
is stable, since for this value of l2 , the stable and unstable orbits are the same. Plugging in rmin ,
and the value of l2 above into
d2 Vef f 6l2 2rS m 12rS l2
= − − ,
dr2 mr4 r3 mr5
we find that d2 Vef f /dr = 0 at this combined minimum. Thus this orbit is just really neither stable
nor unstable, but neutral. It is called, however, the minimum or last stable orbit, since an orbit
just a tiny bit larger is in fact stable.
This last stable orbit value is a very important result. When things fall into black holes they
have trouble getting in because of angular momentum. The tend to get ripped apart and form
what are called accretion disks. The material in the disk gradually spirals inward. However, when
the radius reaches 3rS , there is no longer a stable circular orbit, so the material all just flows into
the hole. Thus we expect that when we look at real black holes in space we will see material
down to 3rS but not any closer. When we detect X-rays from black holes, we hope we are looking
at material radiating from the distance 3rS . And when we find periodic signals from black hole
candidates, we assume this is the radius at which the material is orbiting.
– 49 –
8. Extracting Energy from a Black Hole; Light Orbits
8.1. Extracting Energy from a Black Hole
Now let me tell you why black holes are the most efficient energy generation devices we have
yet concieved. Much better than nuclear fission or even nuclear fusion. Of course, there is the small
problem of not having a black hole near by and available for use.
First consider throwing some junk into a black hole from far away. We learned before that the
total energy of the stuff far from the black hole is just E = mc2 , where m is the mass of the stuff.
Suppose this stuff doesn’t go straight into the hole, but is slowed down in a series of orbits, finally
settling into the last stable orbit we just discussed. What is the energy of the stuff now? We can
use the radial geodesic equation to find out. On a circular orbit dr/dτ = 0, so we have
E2 l2

rS
= 1− m+ .
m r mr2
Next we plug in l2 = 3rS2 m2 , and r = 3rS ,
E2 1 3r2 m2 1 1 8
= (1 − )(m + S 2 ) = (1 − )(1 + )m = m.
m 3 m9rS 3 3 9
Thus r
8 2
E= mc .
9
Somehow during the transition from far away to the last stable orbit the energy has decreased.
Note that in doing this we did not follow a geodesic. E is conserved along geodesics, so starting
at r = ∞ means E = mc2 still at r = 3rS . Thus the stuff must accelerate to get to the smaller
stable circular orbit. Typically falling stuff hits other falling stuff and the collisions are what cause
the acceleration. Thus, the original energy must have radiated away, and therefore be q available for
8
other use. The fractional amount of energy radiated away is (mc2 − E)/mc2 = 1 − 9 = 5.7%.
Let’s compare this with other forms of energy generation. When burned, a gallon of gasoline
produces 1.32 × 108 Joules of energy and has a mass of 2.8 kg. Thus the fractional mass energy
lost when a gallon of gasoline is burned is ∆E/E = (1.32 × 108 )/((2.8)(3 × 108 )2 ) = 5.2 × 10−10 .
This is thus also the fraction of the rest mass of the gasoline that is turned into useful energy, a far
cry from the 5.7% we get from the black hole.
Next, consider nuclear energy, say the complete fission of 1 kg of pure Uranium 235. This
produces 8.28 × 1013 Joules, so ∆E/E = (8.28 × 1013 )/((1)(3 × 108 )2 ) = 0.0092%, still below the
black hole. Finally one usually thinks of the main source of energy in the Universe as nuclear fusion
of Hydrogen into Helium (as in the Sun). Actually in large stars, the Helium is then burned to
Carbon, Oxygen, etc. all the way up to Iron, which is the most stable element. This whole chain
of nuclear fusion releases about 0.9% of the rest mass into usable energy, still more than a factor
of six below what you get from throwing your garbage into a black hole. Black holes are the best
– 50 –
devices we know of for producing energy. In fact, using a rotating black hole we can turn even
more of the rest mass of stuff into useful energy. The energy generation method discussed here is
thought to be the power source of quasars, the most luminous and energetic steady objects in the
Universe.
8.2. Geodesics and motion of light around a black hole
In one sense light is just the same as a particle with it’s rest mass m → 0. To find the geodesics
for light, then, we should be able to just take the limit m → 0 in the geodesic equations we already
found. As a reminder I reproduce those equations below, in the special case with θ = π/2 relevant
for motion in the x-y plane.
2
E2 l2

dr rS
m = − 1− m+
dτ m r mr2
dφ l
=
dτ mr2
dθ
=0
dτ
dt E rS −1
= 1− .
dτ m r
But we immediately see that if we try to take the limit m → 0 in these equations we get into
trouble. We get nonsense like 0 = ∞, etc. Actually we talked about this trouble before when
discussing light. For light the proper time dτ is always zero. Thus when we chose to use τ for our
affine parameter, we excluded the use of these equations for light! What we should do is go back to
the Schwarzschild metric, choose a different affine paramter, and use the Euler-Lagrange equations
to rederive the geodesics for light. However, we can instead use a trick. This is another of those
useful tricks worth learning.
We note that in the limit m → 0, the speed of the particle goes to c and relativistic factor
γ → ∞. That is why the proper time τ → 0. However, the combination mγ is the energy of the
particle and stays fine as m → 0. Thus if we use λ = τ /m as the affine parameter everything may
work out fine. We do that by substituting τ = mλ everywhere in the geodesic equations above,
and then multiplying through by m:
2
l2

dr 2 rS

2
=E − 1− m + 2
dλ r r
dφ l
= 2
dλ r
dθ
=0
dλ
– 51 –
dt rS −1
=E 1− .
dλ r
Now we can take the limit m → 0 and get equations that make sense. Only the r equation changes
to 2
dr rS l2
= E2 − 1 − .
dλ r r2
Note that the trick we did was to try to rescale our original equation that had a bad affine parameter
τ to a different affine parameter that behaved ok. We scaled by a constant, such that we could
take the limit we wanted. If the equations we got didn’t make sense we might have tried another
rescaling. The basic idea of rescaling differential equations is a trick that is often useful when you
want to take some limit of the equations.
Note one problem with the above is that we can’t use the old definitions of E and l which
involve m. We know from the Euler-Lagrange treatment that these quantities are conserved along
geodesics because there is no explicit t or φ dependence in the metric, so we can still call them
energy and angular momentum, but we don’t know yet what the formulas for them are when
considering light geodesics.
OK, we are now ready to do the effective potential treatment for the photon geodesic equation.
We count (dr/dλ)2 as the kinetic energy term, and therefore find that
rS l2 l2 rS l2
Vef f = 1 − = − .
r r2 r2 r3
Plotting this we can discover the light orbits around a black hole.
Using what we learned from the massive case we see that there are three types of orbits: an
unstable circular orbit, a coming in, then out orbit, and a capture orbit. The really new thing here
is the circular light orbit. In Newtonian mechanics light cannot orbit anything, but here in GR it
can! Let’s find the radius of the light orbit:
dVef f 2l2 3rS l2
= − 3 + 4 = 0,
dr r r
or
3
rlightorbit = rS .
2
Thus at 1.5rS , just half the last stable orbit for massive particles, light will orbit the hole.
Since it is an unstable orbit, a black hole can trap light for some amount of time, and then release
it back out to infinity.
Light orbits can produce some pretty wierd effects. Suppose an advanced race built a tube
around a black hole just at the light orbit radius. This is outside the trapped surface, so people
could go in that tube and still get away from the black hole. If you were standing in such a tunnel,
what would you see? The light would go round and round, and you would see the back of your
head. Actually further down the tube you would see yourself again. The tube would in fact look
– 52 –
Fig. 7.— Figure for Chapter 8: Effective potential for light orbits in Schwarzschild metric
infinitely long, and perfectly straight. If you tried to measure its length with a laser you would
never find the end. So by looking you would not know that you were in a circular tube! It would
look perfectly straight to you!
Even weirder would be if you built a such a tube a tiny bit inside the light orbit radius
– 53 –
r = 1.5rS . In this case if you shone your laser beam, it would curve inward towards the hole and
hit the inner wall of the tube. Thus to you it would seem that the inner wall was further out
than the center. The appearance would be that the entire tube was curving outward, not inward!
However, if you got on your skate board and gave yourself a push, you it would not take any energy
to coast around the hole, it is all at the same r. It would look like you were going uphill, but feel
like you were going level! Finally, what would it be like if you built the tube just outside the light
orbit radius? Check out the effective potential; light trying to circle just outside the peak in the
potential (light orbit radius) will “fall” out to infinity. Thus the light would go for a way down the
tube but eventually hit the ceiling. Thus it would look like the tube case curing down towards the
hole (which it is). However, again a skateboard ride would feel like you are going straight, even
though it looks like you are going downhill.
– 54 –
9. Light Bending and Gravitational Lenses
9.1. Formulas
Let’s calculate the result of the most famous GR experiment. The simple formula we are going
to derive for light bending is also the basis of all gravitational lensing, which has become a major
industry in modern astrophysics. We start with the geodesics we found previously for light.
r
dr rS l2
= ± E 2 − (1 − ) 2 ,
dλ r r
and
dφ l
= 2.
dλ r
Here we haven’t really discussed what λ is, so let’s get rid of it by dividing the second equation by
the first: dφ/dr = (dφ/dλ)/(dr/dλ):
−1/2
rS l2

dφ l 2 ±1
=± 2 E − (1 − ) 2 = q .
dr r r r 2
r2 El2 − (1 − rS 1
r ) r2
This is an equation for the angular deviation of the light ray φ, as a function of the radial
coordinate, r. The good thing is that only the combination l/E appears. We can define this
constant as
l
b= ,
E
and call it the impact parameter, or distance of closest approach. This makes sense since for
light we expect the angular momentum to be l = pr sin φ, and E = p, so b = l/E = r sin φ, which
is just the normal definition of the impact parameter for something being fired at an object from
far away.
This equation tells us how the angle of the light φ changes as we approach and then recede
from a spherical mass. There are several possible trajectories, including circular, and inspiralling,
but for now let’s consider trajectories that start far away, come close, and then leave to far away
on the other side. We would like to calculate by angle the light is bent because of its passage
near to the massive object. The first step is a change of variables from r to u = r−1 . Then
r = 1/u and dr = −du/u2 . The differential equation is changed using dφ/du = (dφ/dr)(dr/du),
with dr/du = −1/u2 . Thus we get
−1/2
dφ 1
=± − u2 + rS u3 .
du b2
To find the angle we integrate this starting at r = ∞ (u = 0) with φ = π, going to r = b (u = 1/b),

and ending at r = ∞ (u = 0) with φ ≈ 0. We see a problem, in that the integration variable goes
from 0 to 0. We solve this problem by noting that by symmetry the angle of light bending from
– 55 –
Fig. 8.— Figures for Chapter 9 and 10: (a) Definition of Impact parameter,light bending trajectory,
and (b) Angles of light in Schwarzschild and shell frames
infinity to b, must be the same as the angle of light bending from b to infinity. So we can just go
– 56 –
from u = 0 to u = 1/b, and then multiply our final answer by two. So half our angle is:
Z 1/b −1/2
1
φ= − u2 + rS u3 du.
0 b2
This is the complete answer. We can do this integral numerically for any values of b and get all the
possible light orbits. However, for almost all cases, the angle of light bending is small and we can
find a very important analytic formula. This formula is the one almost always used in graviational
lensing. To find it we note that usually the quantity rS /r 1, or equivalently rS u3 u2 . This
will be true if the impact parameter b is much larger tha rS which is usually true. In this case we
can turn this integral into one we can do analytically. We change variables again to y, where
y 2 = u2 − rS u3 ,
so y = (u2 − rS u3 )1/2 = u(1 − rS u)1/2 ≈ u(1 − rS u/2). Then dy ≈ du − rS udu = (1 − rS u)du, and
du/dy ≈ (1 − rS u)−1 ≈ 1 + rS u ≈ 1 + rS y. We do all this so that we can make the integration
variable y so we can do the integral, but also keep the first order term that has an rS in it. The
integral becomes
Z 1/b −1/2 Z 1/b
1 2 du 1 + rS y
φ= 2
−y dy ≈ 1 dy.
0 b dy 0 ( b2 − y 2 )1/2
This integral can be easily evaluated or looked up in a table to give
r
−1
1 1 1
2| b ,
φ ≈ sin (by)|0 − rS
b
− y 0
b2
π rS
or φ ≈ 2 + b . The total change in φ is twice this or π + 2rS /b, for a total light bending angle of
2rS 4GM
∆φ = = .
b bc2
We can compare this answer with the experiment Eddington first did. During a total solar
eclipse one can see stars near the Sun. For a light ray going near the limb of the Sun the impact
parameter would be nearly the Sun’s radius, b = 7 × 105 km, and we know rS = 3 km for the Sun,
so ∆φ = 8.5 × 10−6 radians, or 1.75 arcsec. This is close to the value found by Eddingon, and was
the first real test of GR. This formula is the basis of all gravitational lensing!
How is this test done? Thing of a bunch of stars about 1/2 degree apart on the sky. At night
you carefully measure the angular distance between them. Now suppose the Sun, during the day,
comes through that area. Well you can’t see the stars because the Sun lights up the sky. But
during a total eclipse of the Sun, the Moon blocks out the Sun and you can see the stars. Thus
you can measure the effect on the starlight of passing by the Sun. How do the stars appear? They
appear farther apart than when the Sun is not present, because the light bends around the Sun.
We always think in our minds that light goes straight, so we extrapolate back to the wrong position
when the light bends. We see from this that the 1.75 arcsec is not really right. The stars farther
– 57 –
from the limb of the Sun bend less than 1.75 arcsec. In fact, what would a single circularly blob of
light (e.g. a galaxy) look like when it is gravitationally lensed like this? The light from the part of
the blob closest to the Sun will bend more than the light from the far edge of the blob. Thus the
circular blob of light will be squashed radially. It will also be stretched tangentially. This is the
hallmark of gravitationally lensed objects and can be used to tell whether or not something has
been lensed. The limit of this is when the blob of light is directly behind a very small lens which
is powerful enough. Then the circular blob of light is stretched into a ring of light around the lens.
This is called the Einstein ring, and has been observed. What determines the radius of the ring?
It is easily derived from the light bending formula above. It depends on rS which depends only on
the mass of the lens. Thus if we see an Einstein ring, we can know the mass of the lens, even if the
lens mass itself is not shining and is invisible. This is one was to search for the dark matter. There
is one complication, in that the Einstein radius also depends on the distance of the lens. We may
come back to this.
9.2. Powerpoint presentation on Gravitational Lensing
Gravitational Lensing is divided into divided into two basic catagories: strong lensing when
two or more images of the source appears and weak lensing, where the image is squashed and
magnified, but no additional image is made. These are used for various purposes in astrophysics.
• Strong lensing of galaxies: Einstein rings
• Strong lensing of quasars, Einstein crosses, time delays, dark matter substructure
• weak lensing and dark matter
• weak lensing and the bullet cluster merger proof of dark matter
• Microlensing search for MACHOs
• Microlensing search for extra-solar planets

– 58 –
10. Some Questions and Puzzles Around Black Holes
10.1. How fast is an object going when it enters a black hole?
Suppose we take the case where someone starts from rest at ∞ and falls into the hole. We
have already calculated this case. Of course, as viewed by the person falling, there is no speed
at all. But as viewed from far away the person’s time slows down and then stops as it enters
the Schwarzschild
q radius. The calculation is done starting from our radial and time equations:
2GM
= ± rrS , and dt/dτ = (E/m)(1 − rrS )−1 = (1 − rrS )−1 , where we used conserved
p
dr/dτ = ± r
energy E = m which is valid starting at rest from infinity. This dt/dτ equation might seem funny
since we have often seen instead dt/dτ = (1 − rrS )−1/2 coming from the metric proper time at a
fixed value r, but the τ in these cases differ. One is for a stationary observer at fixed r, and the
other is the proper time for someone falling into the hole from infinity. Dividing these equation we
find the relation between r and t, that is the speed as seen from from away:
r
dr rS rS
vfar away = =− 1− .
dt r r
We see again that as r → rS , v → 0. The far away observer never sees the person fall in. Following
Wheeler, we define the shell frame as the frame that is static at a given value of r, that is someone
who is just hovering over the hole at certain radius. We on Earth are in a shell frame, since the
Earth’s surface holds us here.
How does it look from someone on a shell near the horizon at r = rS ?
Distances are measured in this frame by poking a meter stick in the radial direction, and thus
are just the proper distances dlr from the Schwarzschild metric:
rS −1/2
dlr = drshell = dr 1 − ,
r
where rS = 2GM/c2 as usual. How does time run in the shell frame? Well, r is not changing
between clock ticks, so compared to far-away time t, the metric value of dτ is the shell frame time:
rS 1/2
dtshell = dt 1 − .
r
Using the two equations above, we can relate times and distances in the shell frame to those in
the far away frame (e.g. Schwarzschild coordinates r and t). We can also find the speed measured
in the shell frame:
drshell rS dr
vshell = = (1 − )−1 .
dtshell r dt
For the person falling in from far away, we put in the result for dr/dt above to find:
r
drshell rs
vshell = =− .
dtshell r
– 59 –
This gives the remarkable result that to a shell observer, sitting at r = rS (well, just above that
so they can tell us what happened!), the falling objects goes by at vshell = 1, that is the speed of
light! Isn’t it strange that the same object doing the same thing can be moving at 0 speed or c
from different vantage points.
We have to ask now what is the energy of that falling object. From far away it is E = m and this
stays the same; energy is conserved along geodesics according to our Euler-Lagrange calculation.
According to the shell observer, they see something going be at nearly the speed of light, that is
γ → ∞, so E = mγ → ∞. This is actually consistent with what we said before that the redshift of
something escaping from the Schwarzschild radius was infinite. That is, it would take an infinite
amount of energy to get out, and the escape velocity is c. We once again that in special and general
relativity, the velocity and energy of an object are frame dependent, and there is no such thing as
the “real velocity” or “real energy” of an object. Also note that in General Relativity it is usually
impossible to divide up the energy into kinetic and potential energy in a meaningful way. These
are frame dependent also (because velocity is).
There is yet one other velocity to think about. The mixed coordinate speed dr/dτ q which we
used above. This is how fast the falling person sees the outside world go by: dr/dτ = ± 2GM r =
p rS
± r . Wheeler calls this the rain frame. Here we see again the speed goes to c at the horizon.
Also we see that inside the horizon the speed becomes faster than light! Is this really true? Well
the formula is correct, but the interpretation of r changes inside. So there is not really motion
faster than c.
10.2. What does it look like standing near or falling into a black hole?
Using our light bending formulas we can describe how things look to someone falling into a
black hole, or to someone sitting on a shell at a distance r. We have to be careful what variable we
use. So far we mostly discussed things from the point of view of the “far-away” or Schwarzschild
coordinate frame. We also discussed the “free-fall” frame. What is the metric in the free fall frame?
Minkowski of course; that’s called the equivalence principle. There is one other frame we’ve kind of
discussed which is useful. The frame we have here on Earth, sitting stationary at a fixed distance
from the center.
So let’s figure out what a black hole “looks” like when we are standing still above it at a fixed
value of r, i.e. in the shell frame. For someone a coordinate distance r from the hole, looking
toward the hole, what will they see? Let’s define the angle of looking from directly overhead, that
is θ = 0 is looking straight away from the hole and θ = π is looking directly at its center. Looking
straight at the center of the hole one sees only blackness, since no light can come out of it (assuming
that there is no one else between you and the hole). At a certain angle one would see the “edge” of
the black hole, that is a separation between the blackness of the hole and the sky around it. Now
because light travels on curved geodesics, one cannot use normal Euclidian geometry to find that
– 60 –
angle. One must use the light geodesic equation we derived last time. To find the angular size of
the hole we want to find those geodesics that just barely make it into the hole, and those that just
barely miss the hole. When we look out at a certain angle, we are specifying a light geodesic. We
want to map these angles into the various geodesics and see which go into the hole and which don’t.
We found earlier that the light geodesics are specified only by the “impact parameter” b = l/E,
and rS , so there should be some mapping between angle of looking and b.
We can tell geodesics which go into the hole and which don’t using the effective potential for
light, which we found from the radial geodesic equation. Remember that for given values of E and
l (actually only the combination b = l/E occurs) there are light paths that come in towards the
black hole, reach a point of closest approach (which we call the turnaround radius rta ), and then
return to r = ∞. There are also geodesics that go over the hump and plunge into the black hole,
and one geodesic at the light orbit radius, rlo = 32 rS that circles the black hole.
It seems clear that if the turnaround radius is outside the light orbit radius, then the light will
escape the hole, and thus the critical geodesic, the light path that just goes in and therefore defines
the visible “edge” of the black hole, is the one that has rta = rlo , or rta = 32 rS .
We find this critical geodesic by first finding the turnaround radius from our radial geodesic
equation,
1/2
rS l2

dr 2

= E − 1− ,
dλ r r2
by setting dr/dλ = 0. This gives E 2 = (1 − rS /rta )l2 /rta
2 , or
rta
b= p ,
1 − rS /rta
where we used the definition b = l/E. We find the critical geodesic, which is specified by bcrit , the
critical value of b, by setting the turnaround radius to its smallest possible value, the light orbit
√
radius. Thus we find bcrit = 12 27rS = 2.598rS .
Now we need a way to map this value bcrit into a critical viewing angle θcrit . First we want
to map angles in coordinate space, r and φ, to angles in the shell frame. Consider someone sitting
still on a shell at r. Consider a light ray sent out from event A at r to event B at r + dr at some
angle θ.
What does this do in the shell frame? Well since the Schwarzschild metric is flat in the angular
directions, know that that dlφ = rdφ will be the same to both the far away and shell observers.
p
We also know from the metric and definition of proper length that drshell ≡ dlp = dr/ 1 − rS /r is
longer than dr. Thus the angles will be different. Thus locally we can write tan θshell = rdφ/drshell .
Next, we can use our light bending formula, the combination of radial and φ geodesics we
found last time, to relate these quantites to b,
dφ ±1
= q .
dr r2 b12 − (1 − rS 1
r ) r2
– 61 –
Combining the above two equations we find:

p p
rdφ 1 − rs /r 1 − rs /r
tan θshell = = ±p .
dr (r2 /b2 ) + (rs /r) − 1
This is good for any value of r and b, but we can specialize to the critical value of b to find the
critical angle p
1 − rs /r
tan θcrit = ± p .
(4/27)(r2 /rs2 ) + rs /r − 1
We can now use these formulas to find the angular size of a black hole as we hover over it.
First plug in the light orbit radius r = 32 rS , to find tan θcrit = ∞, or θcrit = 900 . This is just as
expected, since looking out at 900 means looking straight out over the horizon which is the light
path. We know light with θshell = 900 or more will never get out, but that light sent at slightly
less angle will get out. Thus below you, like an infinite black plane, you see blackness (the hole).
Looking out in any direction you see yourself at a distance of 3πrS , and then another ring of you’s
at twice this distance, and then again another ring, and so on forever! Above you the entire sky
is displayed. Even stars behind the black hole can be seen, since curved light paths bring them to
you. The entire celestial sphere is projected onto 1/2 the sky.
Going further out than r = 1.5rS we will see stars in more than one-half the sky, and the black
hole in less. Light launched at angle from the vertical less than the θshell (critical) will escape the
black hole, while other light rays will be captured. You just reverse these light rays to know what
the shell observer sees.
For example, using our equation above, (or integrating the light bending equation shows) that
at r = 2.5rS , tan θcrit = ±1.356. Now there are actually several angles that satisfy this value of
tan θcrit , since arctan is a multivalued function and there is the ± from when we took the square
root. The easiest thing is to use common sense. Naively taking the inverse tangent gives θ = ±540 ,
neither of which makes sense, since at this radius we expect the black hole to be smaller than half
the sky. But adding 1800 to −540 is also a solution and gives θcrit = 1260 which makes sense and
is the correct answer. Plug in other values for yourself: for example at r = 1.08rS the angle is
400 , while for r = 1.001rS the angle is 4.70 . Thus as you move closer to the black hole than the
light orbit radius, it seems to cover more and more of the sky, almost completely enveloping you
as you approach the Schwarzschild radius. The entire sky is still visible, but mapped into a small
cone directly above your head. Finally as you cross the horizon to inside the black hole that small
opening closes up; at that point there is no shell observer possible and so our equations no longer
apply.
– 62 –
11. Death by Black Hole
11.1. The final plunge
We discussed what it looks like for someone standing still in the shell frame as they approach
a black hole. Next we ask, what does the observer who free falls into the black hole see. This is
more complicated because you have to take into account special relativistic forward beaming and
also abberation, that is the effect of the motion of the observer relative to the stars has on viewing
angles. The basic result is that as one moves faster and closer to the black hole, everything the
shell observer sees gets shifted forward in the direction of motion; that is, what the shell observer
sees at 900 , the free falling observer sees at a larger angle. Thus while the shell observer who
approaches close to the horizon sees black everywhere except directly away from the hole, the free
falling observer still sees a little more than half the sky. However, almost all the stars are shifted
forward in the direction of motion and form a ring very near the edge of the black hole.
For the free falling observer, the stars are visible from inside the black hole and the final view
just before death is all the stars moved into a ring at 900 , with the black hole filling exactly half
the forward sky.
This all is described in more detail in a book called “Exploring Black Holes”, by Taylor and
Wheeler, pages 5-24 to 5-30. Andrew Hamilton of the University of Colorado has a great web page
with explanations and simulated movies of orbiting and plunging into a black hole.
11.2. How do you die when you go into a black hole?
We saw before that near a black hole the gravitational acceleration can be very large. Is that
true near all black holes? If someone is in free fall, then locally there is no force of gravity and so
it should be very comfortable, just floating in space. That is the equivalence principle. However,
this depends on the size of the spacecraft, and of the person being small enough so that locally the
metric is Minkowski. We measure the amount of deviation from perfect Minkowski space using the
concept of tidal acceleration. For someone falling feet first into a black hole this can be defined as
the difference in acceleration between their feet and their head. The fact that the Moon’s attraction
of the oceans differ from the side of the Earth near the Moon to the middle of the Earth gives rise
to the tides; which is why this is called the tidal force (or tidal acceleration in GR).
This tidal force is what will eventually kill you. If you are falling in feet first, you can see
that since your feet are closer to the hole, they will be pulled in faster than your head. You will
experience this as a stretching feeling. In addition since everything falls towards the center of the
hole, the angle of falling on your left side will differ slightly from the angle of fall on your right
side. Thus you will be squeezed side-to-side. The net result of lengthwise stretching and sideways
squashing is called spaghettification. Things falling into small black holes get stretched and
squeezed into long skinny strands, like a spaghetti noodle!
– 63 –
How do we calculate the tidal acceleration? First consider the Newtonian case here on Earth.
The acceleration of gravity is given by
GM 1 rS
gNewton = − 2
= − 2,
r 2r
where we use here the Schwarzschild radius of the Earth rS = 3km (MEarth /M ) = 0.90 cm!
This seems to have units of meter−1 , so to get it into meters/second2 , we multiply by c2 . Thus
gEarth = −0.5(0.009/(6.37×106 )2 )(3×108 )2 = 9.8 m/s2 , as expected. Note that if you want you can
just remember rS (Earth) = 0.90 cm, instead of gEarth . The tidal force on Earth is the difference of
this acceleration between your head and feet, which we will say are seperated by a distance ∆r ≈ 2
meter. For small differences we can use calculus to find ∆g = g(r+dr)−g(r) = (dg/dr)dr ≈ atide ∆r.
Thus
dg rS
atide = = 3,
dr r
a very simple formula. The difference in acceleration is then ∆g = atide ∆r. For our 2 meter tall
person and for the surface of the Earth, this is just ∆g = (0.009)(6.37×106 )−3 (2)c2 = 3×10−6 m/s2 .
This is only about 3 × 10−7 gee’s, and you can’t even feel the stretching force standing here on the
Earth’s surface.
Now, let’s do the same thing for the Schwarzschild metric. We want to use τ rather than
t since we want the experience of the free-faller. We will use r for simplicity, but perhaps we
should use some other measure of distance, such as the proper length. We could convert using the
formulas above, but for now we only want a rough estimate. Start from the geodesic equation for
p
free falling from infinity, dr/dτ = − rS /r, and take the derivative with respect to τ , to find the
√
bulk acceleration, g = d2 r/dτ 2 = − rS (−1/2)r−3/2 dr/dτ = − 12 rS /r2 . Then the tidal acceleration
is the derivative of this with respect to r.
rS 2
∆g = c ∆r.
r3
This is the same as the Newtonian formula, except we must be careful to use coordinate variable r
and proper time τ .
We can now see when someone dies when they fall into a black hole. I’m not sure what
acceleration between the head and feet would kill someone, but probably 100 gee’s would do it.
I’ve heard that jet pilots black out around 7 gee’s. So we can ask how close to a hole before
1/3
reaching that tidal acceleration. rkill = (rS ∆rc2 /∆g)1/3 = rS [(2 meter)c2 /((100)(9.8m/s2 ))]1/3 =
560km(rS /km)1/3 . The kill radius in units of the Schwarzschild radius can be found by dividing
both sides of the above equation by rS , giving:
rkill
= 560 km (rS /km)−2/3 .
rS
Thus for a 3M black hole, the kill radius is about 1200 km, or about 130 times the horizon radius.
We see that all the discussion we had before about exploration of a small black hole and coming
within 30 km of the center would have to be done by robots, if at all.
– 64 –
Notice that in the above equation, the kill radius in units of rS goes like an inverse power of
rS . Thus for big enough black holes the tidal field should be managable even at the horizon. For
example, consider a 3 billion solar mass black hole (such things are thought to power quasars).
This would have rS = 3(3 × 109 )km = 9 × 109 km, about twice the size of the solar system (Neptune
orbits 4.5 × 109 km from the Sun). Then rkill /rS = 1.3 × 10−4 , or rkill = 1.1 × 106 km, well inside
Mercury’s orbit of 58 × 106 km, so you are deep within the black hole before being killed. In fact,
the tidal acceleration as someone enters such a black holel is only 2.2 × 10−10 gee’s, a thousand
time less than we experience here on Earth. Thus someone could be falling into a very large black
hole and never even feel it! They would certainly be killed at the kill radius above which would
happen a few hours later.
– 65 –
12. Where Do Metrics Come From?
We pulled our Schwarzschild metric out of a hat. But to really find a metric you have to solve
Einstein’s general relativistic field equations. For a simple spherical situation in vacuum like the
Schwarzschild case, this is not too hard, but in general it is nearly impossible. Einstein’s equations
are a set of 10 coupled partial differential equations that are made even more difficult because there
are some symmetries, called general coordinate invariance that have to be obeyed. There are really
only a very few analytic solutions known for realistic situations.
I want to write down these equations for you so you can see what they are and why it is difficult
to solve them. We are not going to try to solve them. In order to understand these equations we
will have to learn a little tensor notation. Tensor notation is a way to write complicated equations
in a short hand. This is similar to writing Maxwell’s equations using the div and curl. Maxwell
himself did not have that convienent notation and the equations he wrote were much longer and
more complicated to look at.
In tensor notation Einstein’s field equations are very simple:

8πG
Gµν = Tµν + Ληµν .
c4
We need to go over the terms in this equation slowly and step by step. First remember the idea of
4-vectors, like
xµ = (t, x, y, z),
or
pµ = (E, px , py , pz ).
Here the index µ is equal to 0,1,2, or 3, with x0 = t, x1 = x, x2 = y, x3 = z, p0 = E, p1 = px , etc.
The zeroth component is the timelike component, and the first through third are spacelike.
Note that we can have indices both above and below the letter. The indices above the letter
must not be mistaken for taking a power. When index is above the letter it indicates a 4-vector,
when it is below the letter it indicates a 1-form.
In differential geometry one would learn exactly what differental forms are, and how they differ
from 4-vectors, but here we will skip all that. One can create a 1-form corresponding to a 4-vector
using a metric. I will show that in a minute, but in flat space, the 1-forms corresponding to the
4-vectors above are:
xµ = (−t, x, y, z),
and
pµ = (−E, px , py , pz ).
The only difference in this case is that the index is lower rather than upper, and the timelike
component got a minus sign. If one used the other signature metric, then the time-like component
got a minus sign.
– 66 –
The way you create these 1-forms is using the flat space Minkowski metric of special relativity:
 
−1 0 0 0
 0 1 0 0 
ηµν =  .
 
 0 0 1 0 
0 0 0 1
We wrote this earlier as ds2 = −dt2 + dx2 + dy 2 + dz 2 , which technically is called the line element.
The metric is actually a rank 2 tensor, that is the 4 by 4 matrix given above. We form the line
element (what we have been calling the metric) by muliplying the 4 by 4 matrix by the two copies
of four vector dxµ : X
ds2 = dxµ ηµν dxν .
µν
Summing first over ν we get

X
ηµν dxν = (−dx0 , dx1 , dx2 , dx3 ) = (−dt, dx, dy, dz) = dxµ .
ν
Then the sum over µ is

X
ds2 = dxµ dxµ = −(dx0 )2 + (dx1 )2 + (dx3 )2 + (dx3 )2 = −dt2 + dx2 + dy 2 + dz 2 ,
µ
that is, the line element above that we have been calling the Minkowski metric. Note that in order
to get the 1-form (lower the index), we just multiply by ηµν and sum over one of the indices. Thus
in general in flat space we can use ηµν to lower indices:
X
vµ = ηµν v ν ,
ν
where v µ is any 4-vector. We can also raise indices using η µν which is the same as ηµν . Raising and
lowering works for all tensors. As another example try working out µ pµ pµ = −E 2 +p2x +p2y +p2z =
P
−m2 . The beauty of the notation is that one can tell what an object is just from the unsummed
over indices. If there are none, then the object is a “scalar”, and is an invariant like the mass or
proper distance; it won’t change by a boost, translation, or rotation. If there is one upper index it
is a 4-vector, and will tranform like xµ (a simple Lorentz transformation). If there is one free lower
index it means means a 1-form. Two lower indices means a 2-form, etc. and the transformation
properties of these are well determined as well.
Now, most people use the Einstein summation convention which means to just leave out
all the summation symbols! Any index that is repeated is summed over. For example then one
writes the equation above as
vµ = ηµν v ν ,
and then ds2 = ηµν dxµ dxν . We have been calling ds2 the metric because it contains the same
information as ηµν , the Minkowski, or flat space metric. We go from flat space to full curved space
– 67 –
in GR by changing the elements of the 4 by 4 matrix. We use the symbol gµν for the full curved
space metric and write the line element
ds2 = gµν dxµ dxν .
In spherical coorindates we don’t use (t, x, y, z), but instead use (t, r, θ, φ) as coodinates for (x0 , x1 , x2 , x3 ).
Thus remembering the Schwarzschild metric we have g00 = −(1−rS /r), g11 = (1−rS /r)−1 , g22 = r2 ,
and g3 = r2 sin2 θ, with all gµν = 0 when µ 6= ν. In curved space gµν 6= g µν , but we won’t get into
that.
Now let’s look again at Einstein’s equation:

8πG
Gµν = Tµν + Ληµν .
c4
We see that since µ and ν can each take any value from 0 to 3, this is actually 16 equations. There
are some symmetries that mean that not all 16 equations are independent: basically Gµν , Tµν , and
ηµν are all symmetric, meaning T02 = T20 , etc. which makes 6 of the equations redundant, thereby
reducing the number to 10. We also see the ηµν flat space metric. The constant term Λ multiplying
ηµν is called the cosmological constant and was introduced by Einstein under the false belief
it would stop the Universe from expanding. We will come back to this term after looking at the
term with the rank 2 Tµν tensor. The other factors in this term are just Newton’s constant, G,
and the speed of light. The two-index Tµν is called the Stress-Energy tensor. This is where you
specify what the gravitational sources are: the masses, energies, stresses, etc. that curve spacetime.
We have said that in GR, mass and energy curves spacetime, and we specify exactly what those
source terms are in this tensor. Note for a black hole the source term would be a delta-function
at the origin with a mass, m, while for the expanding Universe case it would be a uniform density
of energy throughout the Universe. A fairly general case often considered is the perfect isotropic
fluid. In this case T00 = −ρ(t, r, θ, φ), Tii = p(t, r, θ, φ), and for µ 6= ν, Tµν = 0, where ρ is the
energy density and p is the pressure. Note that we are using the convention that greek indices
such as µ and ν run from 0 to 3, while latin indices such as i, and j run only from 1
to 3, that is, latin indicate the space-like dimensions.
It would take quite a bit of work to understand why this is the form of the stress energy
tensor, but some idea can be had by thinking in terms of 4-vectors. Time and space are combined
into a 4-vector because as one speeds up, they can be transformed in a well known way (by the
Lorentz transformation). What is constant during such a “boost” is neither time nor the space
but the contraction of the 4-vector, which is called the invariant interval s (or τ ). Likewise it is
the contraction pµ pµ = m2 which is constant in the momentum 4-vector. So somehow the source
of the gravitational field (curvature) must be 4-vector like. We know that mass, that is mass per
unit volume, or density, causes gravity, but mass is really made of energy and momentum. Thus
both energy and momentum are sources of gravity. But we need the densities, that is the mass or
momentum per unit space (or time). This is because GR is a local theory described by differential
equations, so it is only the amount of something at a local point (the density) that comes into the
– 68 –
equations. The energy per unit volume is the density and is written ρ above. It can be a function
of space and time, but it is an initial condition that is specified before solving the field equations.
Since Energy is in the zero position in a 4-vector, the density goes into the zero-zero position of
Tµν . In the momentum position we put some kind of density of momentum what ever that is. In
fact, in GR it is the fluxes of momentum that go in the Tii positions. The flux of momentum is
another way of saying the pressure, p̃ (For pressure I put a tilde over the p̃ to distinguish it from
momentum, which I just call p). This can be understood by thinking of what pressure is. Pressure
is force per unit area: p̃ = F/A, and F = dp/dt, is the change in momentum, so p̃ = dp/(dtdA),
or pressure equals the flux of momentum through a given small surface element. This is a more
general definition of pressure than you may be used to, but is great because it works even when the
particles causing the pressure are not bouncing off walls. In the case where particles are bouncing
off walls, it gives the same number as the normal defintion of pressure.
In general a component of the stress-energy tensor Tµν is the flux of the µ component of
momentum (pµ ) across a surface of constant xν . So again T 00 is the flux of p0 = E across a surface
of constant x0 = t, that is the energy density ρ(x, y, z), and T 11 is the flux of momentum p1 = px
across a surface of constant x1 = x, or the x-momentum per time in the y-z direction, that is, the
pressure.
For the isotropic fluid case (which covers many cases of interest) the stress-energy Tensor then
reads:  
−ρ 0 0 0
 0 p 0 0 
Tµν =  .
 
 0 0 p 0 
0 0 0 p
For non-isotropic cases, we have to specify the off-diagonal terms also. Keeping with the 4-vector
idea, these involve energy and momentum fluxes in space and time. Thus these are stresses, e.g.
momentum fluxes in the y-direction transfered across the x-direction, i.e. y-forces across x-surfaces.
It is surprising, but inevitable that just squeezing a ball gives rise to gravitational fields!
As a little aside, we note that it is energy, not just rest mass, that goes in the stress-energy
tensor, so electric and magnetic fields also count. Thus electromagetic fields and radiation give rise
to curved spacetime and “gravitational fields”. It is a little complicated, but for a pure electromag-
netic field one can write out the stress energy tensor in terms of electric field E and magnetic field
B as follows:
1
T αβ = Fγα F βγ − g αβ Fγδ F γδ ,
4
where we raise and lower using the metric gµν , and
 
0 −Ex −Ey −Ez
µν
 E
 x 0 −Bz By 
F = .

 Ey Bz 0 Bx 
Ez −By Bx 0
– 69 –
Throughout we use the Einstein summation convention, so there are many sums in the above
equations. In electromagnetic theory, the tensor F µν is called the field strength tensor, and is
related to the 4-vector form of the electric and magnetic fields, Aµ . The above tensor notation is
quite fun in that it allows Maxwells’ equations to be written in a very simple form, which we don’t
reproduce here.
Now look again at the 3rd term in the Einstein equations. We now see that since Λ is a
constant it represents a uniform stress-energy. In fact, it represents an energy density of empty
space; that is, remove all matter, radiation, etc. of every sort from the space and Λ represents the
energy density then. This seems like a weird concept, and when the expansion of the Universe was
discovered, Einstein decided that the cosmological constant term was not needed. In fact, this term
is mathematically allowed, and actually is completely acceptable in modern quantum field theory,
and so should be included. It is an experimental question what the value of Λ is. If there is no
energy density of the vacuum, then one can just set Λ = 0. When we write Tµν for a black hole
we typically don’t include Λ and so set Λ = 0. If we didn’t set it to zero then we would not get
the Schwarzschild metric, but a metric that actually has two horizons in it, one near the normal
Schwarzschild radius, and another at a very great distance corresponding to the eventual “edge” of
the visible Universe. Maybe more on that next quarter in Phyiscs 162!
Next look at the first term in Einstein’s field equations: Gµν which is called the Einstein tensor.
We to define need one more notational convention: We will write derivatives with respect to xµ , as
∂µ . That is ∂α f (x) = ∂f /∂xα , and ∂ α f (x) = ∂f /∂xα , where as usual if α = 0, this means a time
derivative, and if α = 2, is means a derivative with respect to y, etc.
Now the Einstein tensor is defined in terms of the metric and contractions of the Reimann
curvature tensor Rµνκ λ :
1
Gµν = Rµν − gµν R,
2
where R is called the Ricci scalar curvature
R = g λν g µκ Rλµνκ .
Don’t forget that we are using the Einstein summation convention so there are actually 256 terms
in the equation for R! We raise and lower indices of the Reimann and other tensors by contracting
σ . The Ricci tensor is formed by
with the metric, for example: Rλµνκ = gλσ Rµνκ
Rµκ = g λν Rλµνκ .
Note that the rank four Reimann tensor is a shorthand way of writing 256 numbers! It is formed
from the metric and the rank three affine connection Γλµν which contains 64 numbers:
λ
Rµνκ = ∂κ Γλµν − ∂ν Γλµκ − Γηµν Γλκη − Γηµκ Γλνη .
The final definition is the affine connection also known as the Christoffel symbol, which
involves derivatives of the metric:
1
Γσλµ = g νσ (∂λ gµν − ∂µ gλν − ∂ν gµλ ) .
2
– 70 –
Whew! So starting with the the stress energy tensor and the cosmological constant, one solves
this set of coupled differential equations for the metric gµν . That is, we need to find what sixteen
quantities contained in gµν can be differentiated and summed over as specified above so as to satify
the Field equations. Given how complicated the above formulas are you might guess that this is
not an easy task! In fact it is made even more difficult because the solution of the above equations
is not unique. There are certain contraints call the Bianchi Indenties that are simultaneously
satisified
Rαβγδ;µ + Rαβδµ;γ + Rαβµγ;δ = 0,
where for any tensor V , the co-variant derivative V;νµ is defined
V;νµ = ∂ν V µ + Γµνκ V κ .
The Bianchi identities are where convervation of momentum and energy are encoded into General
Relativity. These identities plus the symmetric nature of the tensors, imply extra symmetries in
the field equations that reduce the 16 differential equations to effectively only 6, but one must be
very careful in consistently picking values of some seemingly arbitrary functions caused by these
coordinate invariances. This is called choosing a gauge. The results you get may depend on the value
of these arbitrary functions (the gauge you choose), and thus workers who picked different values
might get different answers. Luckily, any quantity that can be actually measured in an experiment
must be independent of the gauge and therefore the same in every gauge. It is important in GR to
calculate things that can be measured.
One can thus see why solving these equations is not easy. For realistic conditions (realistic
values of the stress-energy tensor) there are only a few known analytic solutions. Even on a
computer these equations are extremely difficult and only recently has progress been made solving
them for realistic situations. And when you solve them what you get is the metric. Thus, while
there is some value in going through the steps of finding the analytic solution for the black hole
stress-energy, in the end, to find the physics you still just start from the metric as we did in this
class. Now you see why in this class we decided to skip the field equations and just pull metrics
out of a hat.
Hopefully this lecture helped you see some of the full complexity of GR.
– 71 –
13. Inside the Black Hole: Kruskal-Szerkeres Coordinates; General Black Holes
13.1. Coordinate problem at r = rS
Let’s look at the point r = rS = 2GM/c2 in the Schwarzschild metric. The metric blows up
here, but we found that for things falling in, there is no problem. Forces don’t become large, in
fact nothing too bad happens. The problem here is not the physics, it is the coordinates. There is
a coordinate singularity at r = rS , but no physical singularity. Besides, coordinates where t runs
backward, and r and t switch roles are just confusing.
As a first example of how coordinates can be bad consider polar coordinates on a plane:
ds2 = dr2 + r2 dθ2 .
In the angular direction this says the distance is dsθ = rdθ. But what look what happens as r → 0.
Then ds/r = dθ → ∞, and the metric doesn’t make any sense.
Fig: Grid of coordinates on surface of sphere
As a second example of coordinate problems, consider the surface of the sphere.
ds2 = R2 dθ2 + R2 sin2 θdφ2 ,
where R is a constant (the radius of the sphere). We can plot out these coordinates on a grid φ vs.
θ and everything looks fine. However, at θ = 0, all the values of φ are the same point, that is φ is
not defined. However, there is nothing special about the actual point θ = 0 on a sphere. It is only
the coordinates that are bad at the point θ = 0. In fact it is very easy to fix this problem by just
moving the origin to another point on the sphere!
The problem with the Schwarzschild metric at r = rS is similar; it is not a physical singularity,
but only a coordinate singularity. Remember how light cones squeeze up and then flip sideways as
one crosses rS ?
Fig: Coordinate singularity of Schwarzschild coordinates
13.2. Kruskal-Szerkeres coorindates
Knowing the problem above for the Schwarzschild coordinates, there was a long hard search
for better coordinates to describe the inside and horizon of black holes. These were found in 1960
by Kruskal and Szekeres. These coordinates show the complete geometry inside the black hole and
hold some surprizes. Changing coordinates is just changing variables. Here are the definitions: For
r > rS (outside black hole):
1/2
r ct
u= −1 er/2rS cosh ,
rS 2rS
– 72 –
1/2
r ct
v= −1 er/2rS sinh ,
rS 2rS
while for r < rS (inside black hole):
r 1/2 r/2rS

ct
u= 1− e sinh ,
rS 2rS
r 1/2 r/2rS

ct
v = 1− e cosh .
rS 2rS
With this change of variable the Schwarzschild metric becomes:
4rS3 −r/rS
ds2 = e (−dv 2 + du2 ) + r2 dθ2 + r2 sin2 θdφ2 .
r
It it easiest to see this by calculating du and dv in terms of dt and dr and plugging these into the
equation above to recover the Schwarzschild metric. Recall the definitions of the hyperbolic sine
and cosine: cosh x = (ex + e−x )/2, and sinh x = (ex − e−x )/2, with identity: cosh2 x − sinh2 x = 1.
One sees immediately that this metric does not contain any singularity or problem at r = rS . It
does however have a problem at r = 0. This is true physical singularity. The density and curvature
become infinite there, so no change of coordinates can remove that problem. This is a problem for
quantum gravity or string theory.
For use later we note the following identity (valid both inside and outside the black hole):

2 2 r
u −v = − 1 er/rS .
rS
Let’s plot out and explore this form of the Schwarzschild metric. We plot u vs. v, and try to
understand how particles move in these new coordinates. Everything we did before is still valid, it
is just the coordinates that are different; like switching from Cartesian to Polar coordinates on a
plane.
Recall that we find the time and space coordinates by the signs of the terms in the metric.
We see that θ, φ, and u are space-like coordinates, and that v is timelike. So we plot u on the
horizontal axis and v on the vertical axis.
• First note that light travels at 450 everywhere in these coordinates. Remember we find the
null (lightlike) geodesics by setting ds = 0. We can also set dθ = dφ = 0 to consider only
radial moving light. The KS metric then reads: 0 = −dv 2 + du2 , or dv/du = ±1. Thus
lightcones look exactly like they do in flat Minkowski space. This is the best property of
these coordinates. There is no squeezing up as one approaches the horizon, or flipping over
inside the black hole.
• Regions of constant r are hyperbolas (i.e. u2 − v 2 =constant). Thus particles that don’t move
at all in the Schwarzschild metric actually move along hyperbolas in these coordinates. This
– 73 –
Fig. 9.— Figure for Chapter 13: Kruskal-Szekeres coordinates from Misner,Thorne, & Wheeler
can be seen from the above equation for u2 − v 2 which depends only on r and not on t. So
different values of constant r give different hyperbolas. The old variable t advances as you
move along one of these hyperbolas. A particle at rest does not move straight up in u and v
coordinates, but moves along a hyperbola.
– 74 –
Fig. 10.— Figure for Chapter 13: Kruskal-Szekeres coordinates
• Note that the sigularity point r = 0 is also a hyperbola. Plug in r = 0 and find u =
sinh ct/2rS and v = cosh ct/2rS . Noting that cosh2 x − sinh2 x = 1, we see that v 2 − u2 = 1,
is the hyperbola of the singularity at the center of a black hole. Thus in these coordinates
the singularity is spread out over u and v. In fact, there are two hyperbolas here. These
coordinates show a better picture of the singularity than the Schwarzschild coordiates. We
– 75 –
draw singularities as jagged lines; this indicates nothing can pass by and anything that touches
will be destroyed. In our plot we have suppressed the theta and phi directions. Really then
the singularity is a set of two surfaces! (Rotate our coordinate around the axis to represent
say the phi direction.
• To find where lines of constant t are, divide the definitions of u and v. We see that v/u =
tanh ct/2rS outside the black hole and v/u = coth ct/2rS inside. Thus v/u depends only
t and not on r, so surfaces that have constant t have constant values of v/u, and actually
are straight lines. We can draw these on our coordinate plot. Where these lines cross the
hyperbola shows us how time evolves along the hyperbolas. We see that lines with higher
slopes represent later times. Actually we knew this already because light travels up at a 450
angle toward the future.
• The line t = ∞ is the line u = v, that is, tanh (ct/2rS ) = ±1 has as its solution, t = ±∞.
Remembering that lightcones are still 450 angles one can go from a finite time with v < u
past t = ∞ to where v > u. However, after t = ∞, we see that all future lightcones will lead
to the hyperbola that represents the singularity. Thus the inside of the black hole in these
coordinates is the region above the diagonal lines, and one still has to hit the singularity and
get killed once you are inside the black hole.
• We see that the horizon of the black hole is thus the null line that goes through the origin.
These 450 lines divide up the entire spacetime into four regions. The outside to the right is
called region I. It is where we start when we fall into the black hole.
The region on the top is called region II. It contains the inside of the black hole. The
singularity towards the future is called the black hole.
The region to the left is called region III. It is another asymptotically flat region. This region
is connected to region I by the wormhole.
The region on the bottom is called region IV. The singularity on the bottom (towards the
past) is called the white hole. Note that the future lightcones show that it is impossible to
ever reach the white hole.
• Using the identity above for u2 − v 2 , we see that inside the hole lines of constant r are side-to-
side hyperbolas like the singularity, but outside the hole lines of constant r are up-and-down
hyperbolas. A shell observer outside the hole therefore moves along one of these up-and-down
hyperbolas. The world lines of objects at rest are not straight up and the slope of the world
lines does not give the velocity. However, moving at c is a 450 angle. So if you drop an object
into the hole from a shell it would start to accelerate towards the hole, that is, the slope of its
world line would shift towards the hole and eventually approach 450 as it crossed the horizon.
Once inside it would eventually hit the singularity on the top.
So we see that a black hole geometry is more complicated than Schwarzschild imagined. We
recognize regions I and II as the normal outside and inside of a black hole, but what are these extra
– 76 –
regions? Region III is pretty interesting. If you go far along the -u axis you get to a place where
spacetime is flat. This is called asymptotically flat which is like the far-away coordinates we are
used to, except this region III is not just the same outside as region I. It seems to be a completely
new area. In theory it could be a completely different Universe, or a part of this Universe far from
region I. Thus black holes seem to connect different regions. This can be shown using an embedding
diagram.
The area of the diagram between regions I and III is called the Einstein-Rosen bridge, or
Schwarzschild throat or Schwarzschild wormhole. If you could go through it you could enter
the black hole from region I, and come out of the black hole into region III, which could be millions
of lightyears away, or even part of another Universe!
However, can you get through the wormhole from region I to region III? Try and do it. Start
from anywhere in region I and stay inside the future lightcone. You see that because of the 450
angle of the lightcone, you can’t go from region I to region III. The closest you can come is to
start at the t = −∞ null line, but even in that case you can’t get out of region II and will hit the
singularity. So Schwarzschild wormholes can’t be used to get anywhere. We will talk about other
wormholes later.
How about region IV, the white hole. Well, things can come out of white holes, but nothing
can go in. They only exist in the past lightcones. In fact, the same theorems that say that black
holes can’t be destroyed also say that white holes can’t be created. So unless they were formed at
the beginning of the Universe, we don’t expect to find any.
Finally, we note that in real black holes we don’t expect regions III or IV to exist. We see this
by imagining the creation of a black hole due to the collapse of a star. Remember the Schwarzschild
metric is valid only outside of spherical objects. Inside objects, one would need a different solution
to Einstein’s equations. Thus follow the collapse of a star into a black hole. The surface of the star
follows a path just like when we dropped a particle into the hole. But only outside this surface
does the Schwarzschild metric hold. As the surface collapses it moves up, finally entering region II,
which is when the star turns into a black hole. At that point the trapped surface forms, and all
the material in the star gets crushed into the singularity. But all this time regions III and IV were
not really there. So in the final black hole we find no reason to suppose that these regions exist.
They are mathematical solutions to Einstein’s equations, but don’t exist because they were never
formed.
13.3. General Black Holes
So far we’ve talked only about the simplest black holes: spherical ones, described by the Schwarzschild
metric. There are other more general solutions to Einstein equations that are known, but still much
is not known. For example, it is conjectured by Kip Thorne that black holes are very general, that
– 77 –
Fig. 11.— Figures for Chapter 13b: Schwarzschild wormhole
whenever a mass M is concentrated inside a region with circumference in any direction smaller than
2πrS , (2π(2GM/c2 )), then a horizon forms, i.e. that a region from which light cannot escape comes
into existance. This conjecture has not been proved so is more like a physicists rule of thumb. The
big problem is configurations that change with time, static situations are more well understood. I
list below some general known things about black holes.
– 78 –
• When a black hole forms, the horizon grows from r = 0 outward to r = rS . It is not an
instaneous thing. This can be seen by following photons out of a star as it collapses. The
ones near the center are the first to get trapped.
• Stationary black holes have no hair! This is the common way of stating a famous theorem
about black holes: The metric of a black hole is completely specified by three things and only
three things: it’s mass, it’s angular momentum, and its charge. Thus any information about
stuff that fell into a black hole is completely lost. The Schwarzschild solution was found in
1916, and soon after Reissner (1916) and Nordstrom (1918) found the metric for a charged
black hole. Kerr round the rotating black hole solution in 1963, and the charged rotating
solution was found by Newman et al. in 1965. The proof of this theorem is due to Hawking
(1972), Carter (1973), and Robinson (1975). This also implies that any intial quadrapole,
octopole moments will be radiated away by gravitational waves and you are left with a Kerr
hole or Schwarzschild hole.
• Hawking area theorem. In any process involving classical horizons, the area of the horizon
cannot decrease. This implies that black holes can never bifurcate. The proof assumes that
local energy density is greater than zero; quantum mechanical processes can can violate this
assumption.
• Cosmic censorship: naked singularies can’t exist. This is a conjecture. It says that whenever
there is a singularity (place where curvature goes to infinity) there will be a horizon sur-
rounding it. All known static solutions of GR obey this, but there have been recent claims
by people doing numerical GR that there can be exceptions.
– 79 –
14. Rotating Black Holes: the Kerr Metric
14.1. Kerr Metric
Almost all stars rotate. Thus when massive stars collapse into black holes at the end of their
lives the resulting black holes should also be rotating. Thus the Schwarzschild metric we have been
using is probably not the metric of most real black holes! This was known for a long time, but not
until 1963 was the metric for a rotating black hole found by Kerr. For a black hole rotating in the
φ direction the metric in Boyer-Lindquist coordinates is:
∆ − a2 sin2 θ 2 4aGM r sin2 θ Σ (r2 + a2 )2 + a2 ∆ sin2 θ
ds2 = − dt − dtdφ + dr2 + Σdθ2 + sin2 θdφ2 ,
Σ Σ ∆ Σ
where
a = J/M
is the angular momentum per unit mass of the spinning hole,
Σ = r2 + a2 cos2 θ,
and
∆ = r2 − 2GM r + a2 .
Actually the charged rotating black hole metric is nearly the same. Just replace ∆ by
∆ = r2 − 2GM r + a2 + Q2 /G,
where Q is the charge of the hole (in the proper units (esu)), and one has to add the c’s back in to
use this equation.
Let’s analyze this metric some. First note what happens if we take a = 0. Then Σ = r2 , and
∆ = r2 − rrS , so the metric becomes
rs 2 rS −1 2
ds2 = − 1 − dt + 1 − dr + r2 dθ2 + r2 sin2 θdφ2 ,
r r
which is just the Schwarzschild metric.
Next, notice the cross term, dtdφ. In the tensor formulation this would correspond to a
non-diagonal term in gµν . It means that there is a blending of the the coordinates φ and t; the
forced motion we all must take in time will somehow be blended with motion in the φ direction.
We will come back to this in a minute. Finally, note one needs caution using these coordinates
since for a 6= 0 surfaces of constant t and constant r do not have the metric of a 2-sphere!
14.2. Horizon of Kerr metric: maximum rotation and charge of black holes
Next, how do we find the horizon, that is, the trapped surface or distance from which light can’t
escape? The proper way is to find the regions of the metric where all future lightcones converge.
– 80 –
This requires more differential geometry than we have, but earlier for the Schwazschild metric we
found the horizon by when the sign of the dr term changed. Looking at the metric above, this will
happen when the sign of ∆ changes. Thus the horizon is at ∆ = 0. Solving ∆ = r2 −2GM r+a2 = 0,
we find two solutions for the horizon of a Kerr hole:
r
p rS rs 2
r± = GM ± (GM )2 − a2 = ± − a2 .
2 2
These are called the inner and outer horizons, and the larger one r+ , turns out to be the trapped
surface. The smaller one is called the inner or Cauchy horizon. Note that if we take a = 0,
r+ = rS as expected, and that for a 6= 0, r+ is always smaller than the Schwarzschild radius. Thus
rotating black holes are smaller! Since there is no θ dependence in this equation we see that the
horizons are both spherical.
Also note that if a2 > (GM )2 , then there is no solution. A more careful analysis shows that
when a2 > (GM )2 a black hole (trapped surface) cannot form. Thus there is a maximum angular
momentum that a black hole can have:
rS GM 2
amax = GM = , or Jmax =
2 c
This is called the maximal black hole. Also note that from the brief description of the charged
rotating black hole above, there is also a maximal charge, which can be found from the condition
that a horizon exists.
GM 2
2
J 2

Q
≥ G + ,
c2 c2 Mc
where I put the c’s back in to help with the units, each term is in units of meter2 , and cgs units must
be used (with charge in esu). If this inequality is violated no horizon can form. The maximally
charged black hole is therefore non-rotating, and the maximally rotating black hole is non-charged.
In practice most of the Universe is not heavily charged, and for real black holes we expect the
Kerr metric to obtain, and the maximally rotating black hole to be given by Jmax above. Note
for maximal black holes the horizon is a sphere of radius r+ = GM = rS /2, half the size of a
non-rotating black hole of the same mass.
14.3. Singularity of Kerr metric
We would like to find the singularity in this metric, that is, the point that corresponds to r = 0
in the Schwarzschild metric. The proper way to do this is to calculate the curvature, as I discussed
in a previous lecture, and then look for the places where it goes to infinity. The answer, which makes
2
sense from inspection of the metric is that the singularity occurs at Σ = 0, or rsing + a2 cos2 θ = 0.
If a = 0, we have rsing = 0 as we should, but if a 6= 0, we have the singularity only at θ = π/2
and r = 0. This is peculiar, since why should it matter what θ is when r = 0? Actually there is
– 81 –
a problem with the coordinates inside the black hole, which means that this is not really the right
answer. Consider the region Σ = 0 in limit M → 0, and a 6= 0. Then since there is no mass we know
the metric must be equivalent to the Minkowski flat space metric; which it is in fact in spheroidal
coordinates. Then Σ = 0 is just a coordinate singularity of the type we have seen before. So it
takes some coordinate transformations and fairly subtle analysis to show that the actual singularity
at Σ = 0 is actually a ring of radius a in the x-y plane. For maximal black holes both the horizon
and a = rS /2. Thus the ring singularity is just inside the horizon. If one entered the black hole at
the equator one would hit the singularity immediately, but one entered from the north pole, one
could all the way to the center without hitting the singularity.
14.4. The Ergosphere
Next, let’s find a very interesting new property of spinning black holes. Consider a photon
traveling in φ direction, so ds2 = 0, and dθ = dr = 0. The Kerr metric gives us the equation:
∆ − a2 sin2 θ 2 4aGM r sin2 θ (r2 + a2 )2 − a2 ∆ sin2 θ

0=− dt − dtdφ + sin2 θdφ2 .
Σ Σ Σ
Divide this equation through by dt2 , and we get a quadratic equation for φ̇ = dφ/dt:
Aφ̇2 + B φ̇ + C = 0,
where
(r − a2 )2 − a2 ∆ sin2 θ
2
A= sin2 θ,
Σ
4GM ar sin2 θ
B=− ,
Σ
and
∆ − a2 sin2 θ
C=− .
Σ
This has solution: √
dφ −B ± B 2 − 4AC
= .
dt 2A
Notice now that if C = 0, we have φ̇ = (−B ± B)/(2A), which has φ̇ = 0 as a solution. In
this case we have that light itself, traveling in the φ direction stands still! (The other solution,
φ̇ = −B/A is for light traveling in the same direction as the hole is spinning, while the solution
φ̇ = 0 is for light traveling opposite to the spin. If even light cannot move against the direction
of the hole spin, then massive particles that must move slower than light obviously can’t stay still
no matter what they do. This “stationary light” solution occurs when C = 0, or ∆ = a2 sin2 θ, or
r02 − 2GM R + a2 − a2 sin2 θ = 0, or r02 − 2GM R + a2 cos2 θ = 0, or
p
r0 = GM + G2 M 2 − a2 cos2 θ.
– 82 –
√
This is an ellipoidally shaped region region that has a minimum radius of r0 = GM + G2 M 2 − a2
at θ = 0 (north and south poles), and a maximum radius of r0 = 2GM = rS at θ = π/2 (equator).
The region inside this ellipsoid is called the ergosphere. Inside this region nothing can stand still.
Everything rotates with the hole. The spinning hole actually drags the spacetime around it with
it! Light itself going against the rotation direction will be carried backward around the hole. This
is called “frame dragging”. If you drop something straight down into a spinning black hole, it will
start orbiting the hole even though there is nothing but empty space outside the hole. Actually,
this frame dragging occurs even outside the ergosphere, but it is possible to overcome it by using
rockets. It actually occurs not just around black holes, but around any spinning object, since the
Kerr metric is the description of spacetime in the space around any spinning object. It falls off as
r−3 . Thus if you put a very precise gyroscope in orbit around the Earth, the frame dragging will
cause it to precess. This is called the Lense-Thirring effect. The size of the precession is very small,
only about 0.042 arcsec per year, but there was a NASA satellite called Gravity Probe-B, that was
in orbit trying to measure this effect. It seemed to have failed due to experimental errors.
We can now picture a spinning hole from the top and from the side.
Fig. 12.— Figure for Chapter 14a: Spinning Black hole: Kerr metric: Ergosphere, and horizon
Considering a maximally spinning hole with a = GM = rS /2, from the top there is the horizon
at r = GM = rS /2, just half the normal Schwarzschild radius. Then there is the ergosphere which
from the top is a circle of radius 2GM = rS , but from the side goes from just touching the horizon
at the north and south poles, to being out at rS at the equator. We will discuss in a minute, why
the ergosphere is called by that name.
14.5. Light orbits, inmost stable orbits for Kerr metric
In the same way we used the Euler-Langrange formalism to find the geodesics (orbit equations)
for the Schwarzschild metric, we can find the geodesics for the Kerr metric. We can then go through
the effective potential treatment and find the types of orbits. We won’t spend the time to do this,
– 83 –
but the result is that there are two closest stable orbits for massive particles and also two for light
rays. Recall for the Schwarzschild metric the light orbit was at r = 1.5rS = 3GM , and the closest
stable orbit was at r = 3rS = 6GM .
For the Kerr hole these orbits depend upon the value of angular momentum a = J/M , but
for maximally rotating holes (a = GM ) the light orbit for light rays going in the same direction
as the spin is at r = GM = rS /2 (prograde light orbit), This is the same r as the horizon and
the singularity! For light rays going opposite to the spin the light orbit is at r = 4GM = 2rS
(retrograde light orbit).
For massive particles around a maximally spinning hole, the innermost stable orbit is at R =
GM = rS /2 (prograde), and at r = 9GM = 4.5rS for retrograde orbits. Again the prograde
(innermost stable circular orbit (ISCO) is at the same r as the horizon.
For reference, the effective potential for the Kerr metric is

κGM L2 1 a2 GM
Vef f = − + 2 + (κ − E 2 )(1 + 2 ) − 3 (L − aE)2 ,
r 2r 2 r r
where L is the conserved angular momentum per unit mass of the test particle (not J the angular
momentum of the hole!), and E is the conserved energy per unit mass of the test particle, and
where κ = 0 for light ray geodesicis and κ = 1 for massive particles.
Recall that we calculated earlier the amount of energy that could be extracted by throwing
garbage into a Schwarzschild black hole. We used the geodesic equations to find that a fraction of
the rest mass could be converted to useful energy by gradually bringing the garbage down to the
innermost stable orbit and collecting the energy that had to be radiated away to get there. Finally
the garbage goes down the hole, but we get out the difference between the energy E = mc2 far away
and the energy at that last stable circular orbit. For the Schwarzschild case, we found a rest energy
p
fraction fSC = 1 − 8/9 = 5.7% could be obtained. Doing a similar calculation for a maximally
spinning rotating black hole we find the maxmimal energy that can be extracted is a rest energy
p p
fraction of f = 1 − 1/3 = 42.2% for the prograde inmost stable orbit and f = 1 − 25/27 = 3.7%
for the inmost retrograde stable orbit. Thus we have discovered an even more efficient way of
producing energy. Throw garbage into a spinning black hole making sure to do it in the same
direction as the hole is spinning. This is the most efficient energy producing device I have heard
of.
14.6. Energy extraction from rotating black holes; ergosphere and Penrose process
A very interesting property of rotating black holes is that there exist negative energy particle
trajectories. If one finds the geodesics and effective potential as described above one finds the
energy to be: (Shapiro and Teukolsky equation 12.7.26)
2aGM L + (L2 r2 ∆ + m2 r∆ + r3 ṙ2 )1/2
E= ,
r3 + a2 r + 2GM a2
– 84 –
where E and L are the conserved quantities found from the Euler-Lagrange treatment of the Kerr
metric. If one solves for the regions where E < 0, one finds that it corresponds to what we called
the ergosphere. This property is why it is called the ergosphere, since ergo is the Greek word for
energy. The ergosphere is also called the static limit, since particles cannot stay at rest inside this
surface. The particles which are retrograde inside the ergosphere can have negative energy, but
only particles inside the ergosphere can be scattered into geodesics with negative energies. Such
negative energy particles always have trajectories that carry them into the black hole, so we can
never see them. However, this gives rise to an interesting possible process called the Penrose process
that allows one, in principle, to extract energy from a rotating black hole. Suppose, one shoots
a massive particle inside the ergosphere, and then at a preprogrammed time it splits into to two
particles, one of which has negative energy and one of which has positive energy. Penrose showed
that the negative energy particle would go down the hole, but the positive energy particle could
escape, carrying with it more energy that it came in with!
Ein = Eout + Edown ,
or Eout = Ein − Edown , which implies Eout > Ein since Edown < 0. This does not violate energy
conservation, since it turns out the negative energy particle will slow down the spinning of the hole
and reduce its energy. Thus this is a way to extract energy from a spinning hole. In fact, this
particular process is probably not too useful astrophysically, because in order to get it to work the
relative velocity of the pieces of the splitting particle must be greater than c/2.
14.7. Inside the Kerr black hole
Recall that the singularity at Σ = 0 was a ring of radius a in the equatorial plane. In order to
properly investigate the properties inside the Kerr black hole we need some coordinates analogous
to the Kruskal-Szekeres coordinates for the Schwarzschild metric. These were found by Kerr and
Schild in 1965. They proved that singularity is a ring as described above and also showed that
there were other regions inside just like in the Schwarzschild case.
They found that the ring singularity has some very weird properties. As one passes through
it one goes through r = 0 to negative values of r! And since r becomes timelike near the ring, for
small values of r and θ near π/2, they found orbits in the the φ direction that were timelike. Since
φ a periodic function, this means looping around in φ near the ring means looping around in time;
time becomes periodic inside the ring! That means that closed time-like loops must exist near the
singularity. This is thus a time machine!
Even weirder would be the case the case discussed above where Q2 /G + a2 > G2 M 2 , and
there would be no solution for a horizon. This would of course violate the cosmic censorship “no
naked singularity” conjecture, but this is only a conjecture at this point. So in the pure Kerr
solution, there would be no horizon, but there might still be a singularity! In this case, this would
seem to be a “naked singularity”; that is a singularity not protected by a horizon. Because of the
– 85 –
Fig. 13.— Figure for Chapter 14b: Penrose process of extracting energy from rotating black hole
closed timelike loops, one can in principle make use of the causality violation occuring near the
ring singularity to go “backwards in time” by an arbitrary amount as measured by t coordinate.
However, people think that in real stars this can’t happen for reasons we will discuss below.
The full geometry of the Kerr black hole is more complicated and rich than the Schwarzschild
– 86 –
geometry, so we want to simply the drawing of the Kruskal-Szekeres coordinates, by using what
are called Penrose diagrams. Here we stretch time and space some and draw infinity as an edge.
Lightcones stay the same, but the edges represent going all the out to infinity in time or space.
Fig. 14.— Figure for Chapter 14c: Penrose diagram of Schwarzschild Black hole
Recall region I is where we start and region II is inside the black hole; once here you have to
hit the squiggly line which represents the singularity. Region III is the white hole, and region IV
– 87 –
is another asymptotically flat region of space time, which you can’t get to without traveling faster
than c.
Fig. 15.— Figure for Chapter 14d: Penrose diagram of Kerr Black hole
The Penrose diagram of the Kerr geometry starts out the same as the Schwarzschild diagram,
but the inner horizon and the ring singularity makes a difference. The inside of the inner horizon
(r < r− ) is split into two regions V and VI. Once you pass through the inner horizon you have
– 88 –
Fig. 16.— Figure for Chapter 14e: Penrose diagrams of Black holes
have a choice. You can either hit the singularity and die, or go through the ring and come out into
other regions labeled V’ and VI’. Regions V and VI are still inside the black hole, but regions V’
or VI’ are other asymptotically flat spacetimes, similar to region IV. In these new asymptotically
flat regions, the ring singularity is completely “naked” and has negative mass. With respect to the
original region II, the singularity of course is inside the black hole, but for someone in region V’ or
– 89 –
VI’ there is no black hole and it is possible to escape to any distance. At this point one can actually
go through the ring again and get a region VII identical in structure to region III, and continue
the process ad infinitum! In theory one can extend the structure downward as well through what
used to be the white hole.
Note that in the Schwarzschild case one could not travel from region I to region III because
the future lightcones all contained the singularity. In the Kerr metric however, this is not true!
By going through the ring singularity one can reach another asymptotically flat spacetime without
hitting any singularity. So is this a real wormhole? Can a rotating black hole be used as both a
wormhole and a time machine? The answer is not yet proved one way or the other! The conjecture
is that all the inner structure we have been discussing is unstable. An actual collapse of the spinning
star is not perfectly symmetric or smooth. The “no hair” theorems say that one gets to the pure
Kerr geometry only after the spacetime settles down. In particular, it is conjectured that the ring
singularity is unstable and it is the inner horizon at r = r− that becomes the actual singularity.
In this case, anything going into a Kerr black hole would have to be destroyed. There are several
reasons for this and Wald’s book has a long description on page 318. One reason is that any small
perturbations (light rays or particles) entering the black hole would get an infinite redshift by the
time they get to the inner horizon. These would then have to be put into the stress-energy tensor
and would destroy the perfect symmetry needed to maintain the ring singularity. The same thing
would be true of anyone trying to go through the singularity into another spacetime. Their mass
would change the metric so much as to destroy it.
– 90 –
15. Can Anything Escape a Black Hole? Hawking Radiation
In 1974 Stephen Hawking startled the Physics community by proving that black holes are not
black; they radiate and lose mass. We have always heard and seen proofs that nothing can get
out of a black hole, but here was a proof that radiation must escape! This was not any mistake
in previous work – General Relativity is a classical theory, and Hawking used quantum mechanics
in his proof. Hawking was considered to be a very smart guy before this result, but this result is
what made him famous.
We have considered photons classically following geodesics, but in quantum mechanics the
uncertainty principle means that rays cannot be localized to arbitrary precesion. Near the horizon
of a black hole this changes their behavior.
To do the Hawking calculation properly one needs to use Quantum Field Theory (QFT), the
relativistically correct version of quantum mechanics, and this theory is beyond the scope of this
course. So we will just describe things heuristically to give you a conceptual idea of what is going
on here.
The uncertainty principle has several forms, one of which is
∆E∆t ≥ ~,
where ∆E is the uncertainty in a particle’s energy in a quantum state for a measurment lasting a
time ∆t. Thus on a very short time scale one can’t know precisely the energy of any quantum state.
This principle is preserved in Quantum Field Theory, and it means that the vacuum is filled with
energy fluctuations that violate energy conservation. As long as the violation lasts for a time less
than ∆t = ~/∆E, everything is OK. Short times implies small distances since particles travel at
speeds less than c, so on small scales energy does not have to be conserved. Only over larger times
and larger distances is energy conservation valid. This is how particles tunnel under barriers that
seem to not allow particles through. Thus in QFT we visualize the vacuum, that is empty space,
as being a boiling caldron of particles and anti-particles being continually created and destroyed.
These particles are living on borrowed energy and so have to disappear again in time to satisfy
the uncertainty relation, unless they can get enough energy somehow to “go on mass shell” and
survive. One can actually use this idea to calculate the probability of creating particles in particle
accelerators. The particles in the accelerator transfer their energy to the virtual particles living
on borrowed energy and allow them to become real.
Now consider such fluctuations near the horizon of a black hole. Suppose two photons are
spontaneously created, one with energy E and the other with energy −E. The particle with −E
cannot propagate freely through space and would normally have to pay back its borrowed energy in
a short time ∆t = ~/E. However the particle can gain energy by falling into the black hole. If the
negative energy particle falls in, then that negative energy will be forced down to the singularity
and be added to the black hole. Then the positive energy photon can escape to infinity. Recall
that inside the black hole forward in time means decreasing r, and that the energies of particles
– 91 –
can be either negative or positive depending on whether t is going forward or backward. Over long
times the energy has to be conserved, so if −E goes down the hole, +E has to escape to infinity.
Another way to say this is that the virtual particle can gain enough energy to live by falling into
the hole. If it was originally a negative energy photon, then the hole mass will decrease and the
excess energy radiated away.
I realize this is not a very satisfactory explanation. It is not easy to explain without QFT, and
even then it is tricky, but careful consideration by many physicists have convinced most everyone
that this effect in fact happens. However, it has not been observed experimentally, as we shall see.
Also, since all types of particles are created by quantum fluctuations near the horizon, a black hole
must emit all types of particles: photons, electrons, quarks, etc.
Their are other ways to get at Hawking radiation that are more general. I like a method that
considers what space time looks like in a permanently accelerating frame (like the shell frame we
discussed earlier, which is the frame of us sitting still on Earth.) If one uses Rindler coordinates
one can show that someone moving with a constant acceleration through empty space, actually
sees a thermal bath of particles and also an event horizon. This is called the Unruh effect and
the radiation is called Unruh radiation. One can calculate the temperature that is seen using this
method, say in the shell frame just above the horizon of a black hole and get Hawking’s answer.
In any case, Hawking showed that the radiation coming out of a black hole is a nearly perfect
blackbody, characterized by the Hawking Temperature
~c3

−7 M
T = = 10 K,
8πkGM M
where k is Boltzmann’s constant and the temperature is given in degrees Kelvin. Recall that the
typical energy of a photon radiated from a blackbody is given by
~c3
E = kT = .
8πGM
Also the total energy coming off of a spherical object (like a star or black hole) is proportional to
its surface area A and the temperature to the fourth power,
L = σAT 4 ,
where A = 4πR2 , and the Stefan-Boltzmann constant is σ = 5.67 × 10−8 W m−2 K−4 .
We can use these formulas to find out the temperature of a black hole, and also how much
energy is coming off it. But first we note above that the temperature of the smallest black hole
expected to exist in nature, (3M ) is a few times 10−7 K. This is a very low temperature!
Of course the Hawking radiation causes the black hole to decrease in mass. One can ask how
long will it take to reduce the mass of the black hole to zero, that is to completely evaporate a
black hole? Well, currently with a cosmic microwave background radiation filling the Universe with
– 92 –
a 2.7K radiation field, black holes are not losing any mass, but are gaining! So black holes would
never evaporate in the current environment. However, the Universe is cooling down and eventually
may get to below 10−7 K, and then black holes could start losing mass. Also if there were black
holes with masses less than about the Moon’s mass, then their Hawking temperature would be
greater than 2.7K and they would be evaporating (as long as other stuff wasn’t falling it).
Now the area of a black hole is A = 4πrS2 = 16πG2 M 2 , and the energy lost through Hawking
radiation is equal to the change in the black hole mass:
dM
= −Lhawking = −σAT 4 = −σ16πG2 M 2 T 4 = −16πσG2 (10−7 M K)4 M −2 .
dt
We integrate this from the hole’s current mass to zero mass over the black hole’s lifetime tlife ,
Z tlife Z 0 2
M dM
dt = − ,
0 M B
where B = 16πσG2 (10−7 M K)4 . Thus we find

3 3
M3

M M
tlife = ≈ 1010 years 12
66
≈ 10 years .
3B 10 kg M
Since the Universe is only 1.37×1010 years old we see that for normal black holes, their lifetimes
are way longer than the age of the Universe. These holes are going to be around for a very long
time.
But since the temperature goes inversely with the mass, lighter holes evaporate faster and also
have less mass to evaporate. Thus their lifetimes are much shorter. We can ask how light would a
black hole have to be in order to just be evaporating today?
I give you this question as a homework problem! It is a value quite a bit smaller than black
holes expected to form from normal stellar processes. I think the answer comes out to be around the
mass of a mountain. While no known stellar processes could form such small black holes, Stephen
Hawking proposed that the Big Bang itself could create a large number of these primordial black
holes. These then could be floating around in space. Could you detect them? The Schwarzschild
radius of the Earth is only 0.88 centimeter and these would be much smaller. Their gravity would
be much less than a planet or moon or even an asteroid. So could you see it? Well, look again at
the temperature formula. What happens as M → 0? The temperature of the black hole increases;
in fact it increases without limit! If you plug in a guess at the mass of such a primordial black
hole into the Hawking Temperature formula you find that they are radiating at a temperature of
billions of degrees K! Thus the shine very bright! However, they are small and can’t be seen very
far away until the very end when they explode in a burst of very high energy gamma rays.
We can be more precise. Consider the last second of the primordial black hole’s life. Plugging
τ = 1 second into the lifetime formula you get a mass of about 106 kg which (using E = M c2 ) is
– 93 –
about 1023 Joules. Thus the power output during the last second is about 1023 Joules/sec which can
be compared to the solar luminosity of 4 × 1026 J/s. Thus this is a few thousand times dimmer than
the Sun, about the energy of a very small star. This could be seen if it was fairly near. The energy
spectrum would not be that of a star but would be peaked in the gamma rays. Thus the signature
of these small primoridal black holes is an intense burst of gamma rays that last a very short time.
When gamma ray bursts were discovered people considered this as a possible source, but the energy
spectrum was wrong. For evaporating black holes we expect the energy of the emitted radiation to
become more energic near the end since the temperature continually rises. The non-detection of
these bursts mean that primordial black holes that have lifetimes equal to the age of the Universe
can’t exist. However, primordial black holes that are more massive and therefore live longer are still
a possibility and could even be the dark matter. There has been much theoretical work done on
how such holes could be created in the early universe and much experimental work done on trying
to detect them. So far, there have been no detections, but many mass ranges have been ruled out.
Note the final fate of an evaporating black hole is still not really known. All of Hawking’s
calculations break down when the quantum mechanical wavelength of the particles near the size of
the black hole horizon λ ∼ rS . As the black holes evaporate and get smaller we have to eventually
reach this limit. At that point we would need a real theory of quantum gravity, which does not
yet exist. String theory is a candidate, but so far they can’t calculate anything using string theory.
We can estimate when the breakdown occurs using the above condition
λ ∼ rS .
Using
E = hν = hc/λ = M c2 ,
we find
h
λ= ,
Mc
and the above condition becomes
h GM
= 2 ,
Mc c
or M 2 = hc/G, or
hc/G ≈ 5.5 × 10−5 gm,
p
Mplanck =
where we introduced the standard name of this mass: the Planck mass. At this mass, and at the
scales corresponding to this mass, we expect quantum effects to become so important that standard
General Relativity breaks down. This is the scale string theory and any theory of quantum gravity
works at.
– 94 –
16. Entropy and Black Holes; Observing Real Black Holes
16.1. Observations of Black Holes; powerpoint slide presentation
16.2. Entropy and Black Holes
It turns out that because black holes have a temperature, they also have an entropy. In fact the
entropy of the black hole is enormous even though a black hole is described by only three numbers
and therefore seems to be a very simple state! Remember entropy is a measure of the disorder in
a system; it is equal to k times the log of the number of quantum states of the system.
A clue to this is the Hawking area theorem: the area of the horizon of a black hole cannot
decrease. That is
dA/dt ≥ 0.
It turns out this is the same statement as the second law of thermodynamics, that the entropy
in an isolated system cannot decrease. We won’t prove this, but can get a flavor of it by fooling
around with equations we already know.
Consider the area of a black hole horizon A = 4πrS2 = 16πG2 M 2 /c4 , where I put the c’s back
in. Taking the derivative of the above equation we get
32πM G2
dA = dM,
c4
or
c4 dA
dM = .
32πG2 M
The total energy of the black hole is E = M c2 , so dE = c2 dM , and the Hawking temperature of
the black hole is T = hc3 /(8πkGM ), so we can write
hc3
3
kc dA
dE = ,
8πkGM 4hG
or
kAc3

dE = T d .
4hG
This all follows simply from the Hawking temperature formula, but Hawking and Bekenstein noticed
that this equation is the same as a famous equation from thermodynamics.
dE = T dS,
where S is the thermodynamic entropy of the system. Thus they identified the entropy of the black
hole as
A kc3

S= .
4 hG
– 95 –
Thus in this identification the entropy is proportional to the surface area of the black hole.
What is totally cool about this identification, is the Hawking area theorem that says the area of
a black hole can never decrease becomes the second law of thermodynamics which says that
the entropy of a system can never decrease. This thermodynamic identification also implies that
black holes should radiate as blackbodies, which they do. In fact, the surface area of a black hole
does decrease as it radiates, but then one needs to take into account the entropy of the radiation.
Overall the entropy of the entire system, black hole plus Hawking radiation does not decrease. The
thermodynamics works out great. It is a beautiful connection that Hawking and Bekenstein and
others made, and fits together perfectly.
String theorists recently had one of their few successes, when they managed to calculate the
entropy of a black hole using string theory and got the Hawking/Bekenstein answer. It was a post-
diction not a prediction, but it gave great encouragement to the string theory community and lead
to several new and deep understandings of the relation between black holes and thermodynamics,
including information loss in black holes.
The string theorists were able to count the quantum states of a black hole by finding a dual
system, that is, a set of equations that described a black hole but using a specific quantum field
theory rather than general relativity. The black hole was dual to a gas of hot gluon-like particles,
in which the quantum states could be counted. It came out to be just the Hawking/Bekenstein
answer. The spectrum of radiation coming from the gas of hot gluons could also be counted using
the QFT and it was Hawking answer. One should note that this calculation was not for a normal
spherical black hole, but for a certainly limiting case of a very particular model that does not exist
in the real world, that is a maximally charged black hole with 4 supersymmetries), but still this
was a great success for string theory and helps greatly in our understanding of what black holes
really are.
Finally, there are interesting new developments that are very related to these issues. They go
by the name “holographic” principle in string theory. The basic puzzle is why the entropy of a black
hole is proportional to its surface area. Normally we think of entropy as an extensive quantity;
if you double the volume of a box of gas, you double the entropy of that system. Here however,
the entropy does not go as the volume. This is weird since you really think that the number of
quantum states that a system has should depend upon its volume size, not the area of the bounding
surface. The holographic idea is to generalize this principle to all systems: the number of degrees
of freedom of any system should be limited by a bounding surface area, and scale as surface area,
not volume. For a black hole it might make some kind of sense, since we saw everything kind of
gets hung up at the horizon and never makes it in (as observed from far away). But why would
this be true in general? There are several deep string theory reasons for these speculations and
they are the topics of current research.
An interesting way to see some of this is consider paving the horizon of a black hole with
little “Planck areas”, little squares of size Planck length. How many such squares are there on the
– 96 –
p
horizon of a black hole? Well, the Planck length is lpl = hG/c3 , where of course this is just an
order of magnitude quantity (there could easily be factor of π or 2, etc.). The Planck area then is
Apl = hG/c3 = 2.6 × 10−70 m2 . The number of Planck areas on the horizon is then just the ratio of
the horizon area, A = 4πrS2 to the Planck area. N = A/(hG/c3 ) = Ac3 /(hG). This is remarkably
close to the entropy of the black hole S = A4 (kc3 /(hG)). If one decided that each Planck area was
capable of holding one bit of information (e.g. 1 or 0) and redefined the Planck size to be twice as
big as above, then the number of bits that a BH horizon could hold would be Nbits = A4 (c3 /(hG))
and the number of states would Nstates = 2Nbits . Finally using information theory to calculate the
entropy of an N bit system one finds S = k log2 (Nstates ) = A4 (kc3 /(hG)), exactly the entropy of the
black hole. What does this mean? I’m not sure, but it could be saying something deep about the
nature of spacetime and information. It is the basic observation that gave rise to the idea of the
holographic principle. We will have to see how this holographic principle plays out over the next
few years.
– 97 –
17. Gravitational Waves
17.1. Introduction
Newton’s law of gravity has a problem. Consider two masses M1 and M2 separated by a long
distance. According to Newton, the force between them is F = GM1 M2 /r2 . Now suppose you
move the first mass closer. Newton’s law predicts that the force on the 2nd mass changes, and in
fact changes instantly. But that can’t happen. Things cannot communicate faster than the speed
of light without violating causality. Newton’s thus law violates causality. This is a problem, which
luckily Einstein’s theory of GR fixes. When the first mass is moved, the metric of spacetime is
changed, and the solution to Einstein’s field equations shows that this change in metric propagates
outward at the speed of light. It is actually very similar to electricity and magnetism, where if
one moves a charge it causes a change electric field that propagates outward at c. This is called
electromagnetic radiation. Thus one expects that GR has something like gravitational radiation,
aka gravity waves in it.
In order find these, we can proceed in the same way as you would in E&M. In E&M you write
down Maxwell’s equations and set all the sources, e.g. charges and currents to zero. This gives the
~ = 0, which has solution Ei = Ai exp(iωt − i~k · ~x); that is a sine
wave equation (∇2 − c12 ∂ 2 /∂t2 )E
or cosine traveling wave moving in direction ~k, with polarization Ai . Plugging this solution into
the equation gives ω = ±ck. Since the speed of a wave is v = ω/k, this implies electromagnetic
radiation travels at the speed of v = c, the speed of light.
17.2. Linearized Weak Field GR
We would like to do the same thing with Einstein’s field equations, but as we have seen these
are substantially more complicated:
Gµν = 8πGTµν ,
where we set the cosmological constant to zero. For gravitational radiation, we want to set all
16 terms in the stress-energy tensor Tµν to zero, and then untangle the complicated differential
equations implied in Gµν to solve for the metric gµν , where the line element we have been using
is ds2 = g00 dt2 + g11 dx2 + g22 dy 2 + g33 dz 2 . This is hard, but can be made easier if we employ
the “weak field” limit. This is an example of perturbation theory, where we assume that whatever
answer we get for gµν will be very close to the flat space Minkowski metric ηµν , where remember
η00 = −1, η11 = η22 = η33 = 1, and all other elements of ηµν are zero. Thus we start by writing
gµν = ηµν + hµν ,
where hµν 1. We expand out the field equations, dropping any terms of order h2 or higher.
Remarkably this gives (in Lorentz gauge)
∂2

− 2 + ∇2 h̄µν = −16πGT µν ,
∂t
– 98 –
where h̄µν = hµν − 12 η µν h, is the “reverse trace weak field metric”, and the trace is h = hµµ =
hµν ηµν = −h00 + h11 + h22 + h33 .
17.3. Connection with Newton
Before using this equation for gravitational waves we can make a connection with Newton by
noting that for normal slowly moving matter the only significant element of the stress-energy tensor
is T 00 = ρ, the mass density. Thus we can solve this equation using only the 00 component and
setting T 11 = T 22 = T 33 = 0. If we set up a mass density and look for a static solution, we can
also set the time derivative to zero, giving
∇2 h̄00 = −16πGρ.
This is almost exactly the equation for the Newtonian potential ∇2 φ = 4πGρ, where in the New-
tonian case for a spherical mass M , the gravitational potential would be φ = −GM/r. Thus we
can think of h as the gravitational potential with h̄00 = −4φ, h̄11 = h̄22 = h̄33 = 0. Using the
definition of the trace-reverse metric allows us to solve for the components of hµν . For example,
h̄00 = h00 − 12 η 00 h = h00 + h/2. Note trace of the tensor h̄µν is just h̄ = 4φ, implying h = −4φ.
Thus h00 = h̄00 − h/2 = −4φ + 2φ = −2φ. Likewise, h̄11 = h11 − h/2 = 0, implying h11 = −2φ.
Likewise h22 = h33 = −2φ. Plugging these values for hµν into the line element (metric) we get
ds2 = −(1 + 2φ)dt2 + (1 − 2φ)(dx2 + dy 2 + dz 2 ), where we used our perturbation expansion,
g 00 = −1 + h00 , g 11 = 1 + h11 , etc.
Finally, compare this to the Schwarzschild metric, for simplicity paying attention only to the
time and radial components, dr2 ∼ dx2 + dy 2 + dz 2 . In the weak field limit one can Taylor expand
the (1 − 2GM/r)−1 term in the Schwarzschild metric to find to find ds2 ≈ (−1 + 2GM/r)dt2 +
(1 + 2GM/r)dr2 + ..., precisely the weak field metric found above with φ substituted. Thus the
connection between the Newtonian potential and the metric becomes more clear.
17.4. Gravitational Waves
Now set T µν = 0 in the weak field Einstein equations. This gives
∂2

− 2 + ∇ h̄µν = 0,
2
∂t
again a wave equation for the 16 components of h̄µν . The solution is again sine and cosine traveling
waves
h̄µν = Aµν exp(iωt − i~k · ~x),
where now the “polarization”, Aµν , of the gravity wave has more components. Again we see
ω = ±k, or the speed of gravity waves is 1 (that is c). It is worth discussing the polarization since
– 99 –
it differs from electromagnetism. Recall that an electromagnetic wave traveling in the z direction
is transverse, meaning it can have an “x” component or a “y” component, but no “z” component.
This is set by the gauge condition in electromagnetism. A gravity wave has more components and
it’s gauge condition means that it is quadrapolar. Traveling in the z direction it has two possible
polarizations, one called + polarization with Axx = −Ayy and all other components zero, and the
other, called ×, with Axy = Ayx and all other components zero (note I am calling A11 as Axx , etc.)
This polarization has a big effect on what gravity waves do.
For example, consider a gravity wave traveling in the z direction with + polarization. The
spatial part of the metric will be something like: ds2 = (1 + h11 )dx2 + (1 + h22 )dy 2 + (1 + h33 )dz 2 ,
where we should substitute h11 = A11 sin(ωt), h22 = hyy = Ayy sin(ωt), h33 = Azz sin(ωt), and
we assume we are at z = 0, so we can set ~k · ~x = 0. This gives: ds2 = (1 + h sin(ωt))dx2 + (1 −
h sin(ωt))dy 2 +dz 2 −...dt2 . This means that there will be no change in the z direction, and opposite
sinusoidal motion in the x and y directions. What will this do?
Imagine a circular ring of test particles in the x-y plane. As the gravity wave comes through
what will happen? What will the forces on the particles be? Zero of course! In GR gravity is not a
force! In fact, the coordinate positions (x, y, z) will not change at all. But using the metric above
we see that the distances between the particles will move! This is the effect of the gravity wave
passing by. When the sin(ωt) in the x-direction is maximum (ωt = π/2), it will be at a minimum
in the y-direction. Thus the circle of test particles will alternatively squeeze and stretch in the x
and y directions. You can work out the × polarization also, it is similar except the squeezing and
stretching at a 450 angle.
So to detect gravity waves all you have to do is measure the distances between test masses.
They should move in the odd pattern just described when a gravity wave comes by. Do we expect
gravity waves to exist? Yes of course. As we said at the beginning there has to be gravity waves
whenever masses move, just as in electromagnetism there has to be electromagnetic waves whenever
charges move. But be careful. A charge moving at uniform velocity does cause a change in the
electric field to propagate outward, but does not radiate electromagnetic wave waves that carry
energy. It takes an accelerating charge to create electromagnetic waves. Likewise it takes an
accelerating movement of mass to create a gravitational wave. In fact, due to the quadrapolar
nature of the polarization, it takes a quadrapolar motion to do it. For example, an expanding
spherical shell of mass has no quadrapole moment and therefore does not emit gravitational waves.
However, if two masses move back and forth (say on a spring) or orbit around each other (like two
stars) then there is a quadrapole moment and there is gravitational radiation.
The formula for the strain (i.e. h, i.e. change in the metric) when two object of mass M orbit
each other is roughly:
GM l02 ω 2
A∼ ,
rc4
where l0 is the rough size of the orbit, ω is the angular velocity of the orbit, and r is your distance
from the system. We can make a good rule of thumb by noting that Schwarzschild radius GM/c2 ≈
– 100 –
Fig. 17.— Figure for Chapter 17a: Effect of passing gravitational wave on ring of test particles
rS , and the speed of orbit is roughly v ≈ l0 ω. so

r v 2
S
A≈ .
r c
Consider then a barbell with two 100 kg masses separated by 1 m, spinning at 10 m/s. What is the
– 101 –
size of the metric distortion caused by the resulting gravitational wave? Using rS = 3km(M/M ),
3000m 100kg 10m/s
h∼A∼( )( )( )2 ∼ 10−40 .
1m 2 × 10 kg 3 × 108 m/s
30
Recall this A is the fractional change in distance between two test particles sitting 1 m from the
barbells. This is an absolutely tiny distortion, much much less than the radius of an atomic nucleus.
So this is not measurable.
17.5. Detecting Gravitational Waves
Can one think of anything that would give rise to to a measurable gravitational wave? Well,
in the equation above, we need to increase both the masses and speeds involved. How about
two neutron stars (masses 1.4 M ) orbiting very close to each other? In 1974, Hulse and Taylor
discovered a binary pulsar, PSR 1913+16. This is system of two neutrons stars orbiting other
with a period of 7.75 hours. The system is about 6.4 kpc (21,000 light years) away from us, so
plugging these numbers in you find A ∼ 10−23 , a much bigger value. But over one meter this
is a change in distance of less than 100 millionth of the radius of a proton, unmeasurable with
current equipment. Still the 1993 Nobel prize was given to Hulse and Taylor for their discovery of
gravitational radiation using this system. Why?
Well the beauty of this system is that one of the neutron stars is a pulsar. It is rapidly
spinning and beaming us with a period of P = 0.059029997929613(7) seconds. This is an incredibly
accurately measured period, and for a long time this system was more stable and accurate than
any atomic clock here on earth. These pulses were modified every 7.75 hour by the Doppler effect
indicating that the pulsar orbited another neutron star, and in addition a periastron advance was
measured at 4.2 degrees per year (a GR effect larger than Mercury’s perihelion advances in century).
Using all this information almost everything about this system could be understood. It all checked
out using Einstein gravity, Over time is was also possible to detect that the orbital period was
slowing down at a rate of about 76.5 microseconds/year. Why does the orbital period change?
Something must be bleeding energy from the system, at a rate of about 7.4 × 1024 Watts. What
could this be? If one plots the orbital period over a period of years it falls in exactly the manner
predicted by GR, but only if one includes the energy lost from gravitational radiation. Thus this
was an indirect measurement of gravitation wave emission. To this day, this system (and others
like it) are the only “detections” we have of gravitational radiation.
Is there hope to ever directly detect a gravitational wave? Well we need to be able to measure
strains of around 10−23 . Of course if objects were closer than PSR 1913+16 or involved heavier
objects (e.g. super massive black holes weighing billions of solar masses) then larger values of
h might be produced. The current leading experiment is called LIGO, the Laser Interferometer
Gravitational-Wave Observatory. It is two large Michelson interferometers each about 4 km long.
One interferometer is in Lousiana and the other is in Washington state. Each interferometer has
– 102 –
Fig. 18.— Figure for Chapter 17b: Orbital decay of PSR B1913+16.The data points indicate the
observed change in the epoch of periastron with date while the parabola illustrates the theoretically
expected change in epoch according to general relativity
large test masses carefully hung in vacuum pipes 4 km apart in an L shape. There are mirrors
attached to the test masses and laser systems that carefully measure the distances between these test
masses. If a gravity wave comes from outer space onto the detector, one arm of the interferometer
– 103 –
will stretch while the other shrinks. This shrinking/stretching shape changing will reoccur with
the frequency of the gravity wave. For instance, for the binary pulsar, the period will be 7.5 hours.
However, LIGO cannot detect this kind of period; its sensitivity is only to waves with frequencies
between 30 Hz and 7000 Hz. There are very few astronomical sources of strong gravity waves that
emit at these frequencies, so in fact, there really is almost nothing that LIGO can detect. LIGO
started in 2002 and still has not detected anything except noise. In fact, when trying to measure
distances to a tiny fraction of the radius of the proton, there are innumerable sources of noise:
thermal motion, ground motion, traffic, logging, ocean waves, etc, etc. etc. This is why there are
two interferometers separated by about 3000 km. A gravity wave from space will hit both detectors,
while noise sources will affect one but not the other. Only signals that appear in both detectors
are being considered as real gravity waves. However, there is some hope for LIGO since they are
upgrading it. This will make it sensitive to strains about 10 times smaller, and there are actual
astronomical sources of gravity waves in the frequency band that might be detected.
Finally, there is an exciting proposed space mission called LISA, the Laser Interferometer Space
Antenna, that if launched, would give huge numbers of detections. Here the interferometer arms
would be around 5 million km! This long distance moves the range of frequency sensitivity down
to between 10−5 Hz and 0.1 Hz, just in the range of binary pulsars, orbiting black holes and other
likely astronomical phenomena. Here the interferometer is made of 3 free flying spacecraft orbiting
the Sun (not the Earth!) and the lasers will be measuring the distances between the test masses
made of pure gold and platinum to about 20 picometers, for a strain sensitivity of around 10−23 as
required.

Physics 161: Black Holes

Uploaded by

Copyright:

Available Formats

Physics 161: Black Holes

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Physics 161: Black Holes

Uploaded by

Copyright:

Available Formats

Physics 161: Black Holes

Department of Physics, University of California, San Diego, CA 92093

Introduction to Einstein’s General Theory of Relativity as applied especially to black

1.1 Tour of the Universe . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3

1.2 Black Holes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3

1.3 Curved Spacetime . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3

1.4 Example of two surveyors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4

1.5 Tidal force as curvature . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4

2.1 Using a metric to find distances, areas, etc. . . . . . . . . . . . . . . . . . . . . . . . 7

2.2 Metrics Continued . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7

2.3 Expanding Universe and FRW metric . . . . . . . . . . . . . . . . . . . . . . . . . . 8

2.4 Spacetime metrics and nomenclature . . . . . . . . . . . . . . . . . . . . . . . . . . . 9

2.5 Schwarzschild metric . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11

3 General and Special Relativity 13

3.1 Spacetime diagrams . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13

3.3 Calibrating the axes in a spacetime diagram: hyperbolas=circles! . . . . . . . . . . . 15

3.4 Time dilation and length contraction by spacetime diagram . . . . . . . . . . . . . . 17

3.5 Other special relativity you need to know . . . . . . . . . . . . . . . . . . . . . . . . 18

3.6 Time Dilation in a gravitational field . . . . . . . . . . . . . . . . . . . . . . . . . . . 19

3.7 Old Idea of Black Holes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20

4 Geodesics: Moving in “Straight Lines” Through Curved Spacetime 21

4.1 Geodesics and Calculus of Variations . . . . . . . . . . . . . . . . . . . . . . . . . . . 21

4.2 Geodesics as equations of motion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22

4.3 Euler-Lagrange Equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22

4.4 First Example of Euler-Lagrange equations: classical mechanics . . . . . . . . . . . . 24

4.5 Second example of Euler-Lagrange equations: Flat space geodesics . . . . . . . . . . 24

4.6 Geodesics in Minkowski spacetime . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25

4.7 Conserved quanties in the Euler-Lagrange formalism: Energy and Momentum . . . . 26

5 Equivalence Principle, Gravitational Redshift and Geodesics of the Schwarzschild

5.1 Gravitational Redshift from the Schwarzschild metric . . . . . . . . . . . . . . . . . . 28

5.2 Light bending and the Equivalence Principle . . . . . . . . . . . . . . . . . . . . . . 29

5.3 Gravitational Redshift again . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30

5.4 Geodesics of Schwarzschild metric from Euler-Langrange . . . . . . . . . . . . . . . . 30

6 Distances and Times Around a Black Hole 34

6.1 Can you fall into a black hole? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35

6.2 Time to fall into a black hole . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36

7.1 Shooting Light into a Black Hole . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38

7.2 Inside the Black Hole . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39

7.3 Orbits in the Schwarzschild metric . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41

7.4 Effective Potential for Newtonian Orbits . . . . . . . . . . . . . . . . . . . . . . . . . 41

7.5 Effective Potential for Schwarzschild Orbits . . . . . . . . . . . . . . . . . . . . . . . 42

8 Extracting Energy from a Black Hole; Light Orbits 46

8.1 Extracting Energy from a Black Hole . . . . . . . . . . . . . . . . . . . . . . . . . . . 46

8.2 Geodesics and motion of light around a black hole . . . . . . . . . . . . . . . . . . . 47

9 Light Bending and Gravitational Lenses 51

9.2 Powerpoint presentation on Gravitational Lensing . . . . . . . . . . . . . . . . . . . 54

10 Some Questions and Puzzles Around Black Holes 55

10.1 How fast is an object going when it enters a black hole? . . . . . . . . . . . . . . . . 55

11 Death by Black Hole 59

11.1 The final plunge . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59

11.2 How do you die when you go into a black hole? . . . . . . . . . . . . . . . . . . . . . 59

12 Where Do Metrics Come From? 62

13 Inside the Black Hole: Kruskal-Szerkeres Coordinates; General Black Holes 68

13.1 Coordinate problem at r = rS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68

13.2 Kruskal-Szerkeres coorindates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68

13.3 General Black Holes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73

14 Rotating Black Holes: the Kerr Metric 76

14.1 Kerr Metric . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76

14.3 Singularity of Kerr metric . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77