1

PH30101 General Relativity

Prof Tim Birks
“General relativity without tensors”
General relativity (GR) states that spacetime is deformed by
gravitating masses. Freely-moving objects follow straight lines
(or their nearest equivalents) in this curved spacetime, even in
the presence of gravity. GR is our current theory of gravity and,
alongside the "standard model" of quantum/particle physics,
forms our best account yet of how the Universe fundamentally
works.
A complete treatment of GR relies on tensor analysis, a chunk of
advanced mathematics we'd need to spend a lot of time learning
before even starting the actual physics. But, knowing the
curvature of spacetime, we can deduce the motion of particles
and light without knowing about tensors. (It's still mathematical,
but it's maths you already know.) In this unit we will use this
approach to examine what curved spacetime means, compare the
predictions of GR and Newtonian gravity, and explore the
properties of the quintessential GR object: the black hole.
Spacetime curvature is described by metrics, which we won't be
able to derive without tensors. We'll just take them as given. But,
when you learned quantum mechanics, did it bother you (did
you even notice) that you never saw where Schrödinger's
equation came from?

TA Birks, University of Bath, 15 August 2018

0. Course admin
2
Revision: You will need material from previous units, including:
special relativity (SR): PH10004 for Physics students, and
PH20076 for Maths and Physics students.
Unfortunately PH20076 is taught at the same time as this unit. M&P
students should find that the results of SR that I use early on will make
more sense later. Or, since PH20076 is Self-Directed Learning, you
could always direct yourself to learn some SR ahead of time... At least
it'll be fresh on your minds; the Physics students did it in YR1 and may
have forgotten it all by now. In any case I will cover the key points of SR
in a revision lecture early on, so that should keep you going.

Newtonian mechanics and gravity: including gravitational

potential, angular momentum, orbits, planetary motion, impact
parameter
calculus: differentiation and integration (including line and
multiple integrals), polar plots, ordinary differential equations
(separable, forced s.h.m.) and especially coordinate systems like
spherical polars. But, no vector analysis or complex numbers.
geometry: basic stuff (triangles, circles, parallel lines), curves
(ellipses, hyperbolae) and spheres (surface area, latitude,
longitude, great circles).
calculus of variations: not necessary, but it would enable you to
understand where the geodesic equation of motion comes from.
thermal physics and QM: familiarity with entropy, microstates,
black-body radiation, the uncertainty principle(s).

To contact me outside timetabled contact time, use email

([email protected]) or try my office (3W 3.17B) - I can
usually find some time if I'm not in the middle of something.
0. Course admin
3
The Moodle page for the unit contains:
Organisational information (when different lectures and
problems classes will be held, etc). I will assume you have read
this and will keep checking for changes from time to time,
especially if you miss any announcements in lectures.
The problem set as a pdf file.
Model answers to problem set. These will become available after
the corresponding problems classes. The idea is to attempt the
problems before being given the answers ...
A sheet of useful equations, most of which you do not need to
memorise. Be sure to bring this sheet to all the lectures and
classes. I'll even hand out paper copies in an early lecture.
These notes as a pdf file. On each page of the notes, a header
summarises what's on it and indicates the expected (no
promises) lecture number when it's covered.

I will not be handing out paper copies of the notes, problem set,
model answers or anything else! (Except the sheet of useful
equations.)
If you decide to print the notes I recommend printing them
double-sided, black-and-white, and 2 pages on each side*. In
August 2018 this costs £1.86 using University printers. The
notes are designed to be comfortable to read at this size.
It costs 22p to print the problem set in the same way.

* If you're using Adobe Reader and University printers: select printer UPS1 or UPS2
select "Print in grayscale (black and white)"
click "Multiple" and select 2 Pages per sheet, "Print on both sides of paper" and "Flip on short edge"
Remember to deselect these options next time you print something!
0. Course admin
Structure of the unit
4

non-Euclidean geometry
L6
equivalence
principle
L1,2 special
relativity L4

acceleration
GR: curved L5
Newton's spacetime
gravity L1-6
L3

Schwarzschild Kerr metric

metric L7,8 L17-18
GR: the geodesic
equation
3 classical
L9
tests orbits of particles
L3-12 and light
L10-12

black holes L13-16

thermodynamics and optional

quantum gravity lecture
L19-20 L21
0. Course admin
Table of contents
5
section title page lecture
0 Course admin 1
1 Introduction: deforming time and space 6 L1
We needed a new theory of gravity! 6
The principle of equivalence 8
Consequences of the principle of equivalence 12 L2
General relativity 15
2 Newtonian gravity 18 L3
Time 18
The effective potential 19
Shapes of orbits 22
3 Special relativity 26 L4
Relativistic units 26
Spacetime 28
The interval ds2 and the Minkowski metric 30
Acceleration in SR 33 L5
4 Geometry 41 L6
Flat space 41
Curved space 43
Flat spacetime 47
Curved spacetime 48
5 The Schwarzschild metric 49 L7
What the Sch. coordinates mean 49
Gravitational time dilation 55 L8
6 The geodesic equation of motion 59 L9
7 Orbits in Schwarzschild spacetime 64 L10
Equations of motion 64
The effective potential 66
Bound orbits 70 L11
Photon orbits 76 L12
8 Schwarzschild black holes 82 L13
The singularity at the Schwarzschild radius 82
Painlevé-Gullstrand (PG) coordinates 87 L14
The central singularity (r = 0) 92 L15
Kruskal-Szekeres (KS) coords 97 L16
9 Kerr (rotating) black holes 104 L17
The Kerr metric 104
Orbits around Kerr black holes 109 L18
10 GR and quantum mechanics 115 L19
The thermodynamics of black holes 115
Quantum gravity 119 L20
1. Introduction / A new theory of gravity
Lecture 1 6
1. Introduction: deforming time and space
We needed a new theory of gravity!
General Relativity (GR) is Einstein's theory of gravity. So what's
wrong with Newton's?
Mm
F G
r2

F is the force now, when M and m are r apart. But special

relativity (SR) says no influence can travel faster than light, so
gravitational influences too must take time to get from M to m.
Indeed, r is the difference between the positions of M and m
measured simultaneously, but SR says that simultaneity depends
on your frame of reference.
If that criticism is too vague and picky for you, try the following
thought experiment. In frame S, test mass m lies initially at rest
half-way between two identical streams of masses, spaced by l,
moving at the same speed v in opposite directions:

The net gravitational force on m is zero, by rotation symmetry:

the attraction of the upper stream is balanced by the attraction of
the lower stream.
 m does not accelerate
1. Introduction / A new theory of gravity
Lecture 1 7
Then look at things in frame S' moving at speed v to the right
relative to S. Now the lower stream is at rest, m moves to the left
at speed v, and the upper stream moves to the left at a speed
greater than v.

According to SR, moving objects undergo length contraction.

The upper stream moves faster in S' than in S and experiences
more length contraction, so its spacing is <l. The lower stream
moves slower in S' than in S and experiences less length
contraction, so its spacing is >l.
There's therefore more mass (per unit length) above m than
below, and a net upward gravitational force F on m.
 m accelerates upwards

These two viewpoints contradict one another: m can't both move

upwards (and eventually collide with the upper stream) and not
do so. This exemplifies the fact that:

Newtonian gravity is not consistent with SR.

In our thought experiment, F is balanced by a new repulsive

force between co-moving masses called "gravito-magnetism", cf
the well-known velocity-dependent relativistic force between
electric charges. But rather than patching up Newton's theory in
this way, Einstein preferred to start from scratch with:
1. Introduction / Principle of equivalence
Lecture 1 8
The principle of equivalence
Einstein's thinking on gravity was based on a familiar result
from Newton's theory - the acceleration g of test mass m due to
mass M is independent on m:
Mm
mg  G [ma = F]
r2

The fact that all free-falling masses accelerate equally was well
known before Newton (Galileo etc) and has been experimentally
verified to within one part in 1012. Yet in Newton's theory it is an
astonishing coincidence, because the m's on both sides of the
above equation represent logically-distinct concepts:

inertial mass m gravitational mass m

(resistance to (source of
acceleration) gravitational force)

Why are they identical?

Actually there are other forces that accelerate independently of
mass. For example, "g-forces" that push you backwards in an
accelerating rollercoaster, centrifugal forces that pull outward on
a curved path and Coriolis forces that spin weather systems.
What all these forces have in common is that they don't exist...
They are pseudo-forces that appear only in accelerating (ie, non-
inertial) frames of reference. The acceleration that all masses
seem to have in common is merely the acceleration of the frame
itself.
1. Introduction / Principle of equivalence
Lecture 1 9
Einstein's big idea was that the gross "9.8 ms-2" effect of gravity
is also a pseudo-force. It appears only because, standing on the
ground, we're really using an upwardly-accelerating frame of
reference. He realised this by imagining that a man falling from
a building feels no gravity as he falls, which Einstein called the
"happiest thought" of his life.

Free-falling frames  inertial frames

It follows that being at rest (in the conventional sense) in a

gravitational field is the same as being accelerated relative to a
freely-falling (and hence inertial) frame. Einstein generalised
this mechanical result to all physical phenomena, leading to

The Principle of Equivalence: no local experiment can

distinguish between gravity and accelerated motion.

According to the principle, experiments at rest on the ground

yield the same results as they would in a rocket accelerating
smoothly at 9.8 ms-2 far from gravitating masses. Without
external interactions (for example, no looking out of windows),
you can't tell the difference.

at rest in a in an accelerating
gravitational field rocket, no gravity

in both cases, the ball accelerates downwards relative to the inhabitant

1. Introduction / Principle of equivalence
Lecture 1 10
The word "local" is important. Unlike uniform acceleration,
gravity has a centre. Free-falling frames have vector
accelerations that vary from
place to place, in magnitude
(g is slightly bigger at your
feet than your head) and
direction (the vertical in Bath is
~1½º away from the vertical in
London). A big-enough
experiment can use this spatial
variation to distinguish gravity
and accelerated motion, and we
can't find a common frame of
reference that eliminates gravity
everywhere:

There are no universal inertial frames, only local ones.

Stitching together neighbouring local inertial frames across an

extended region turns the
flat Minkowski spacetime
of SR into the curved
spacetime of GR. (Like
the way stitching
together lots of flat city-
scale maps produces a
spherical surface on a
continental scale.)

Cut out the U-shape of stitched-together small-scale maps and lay the two maps for the
north pole over each other (with the right orientation). The paper strip forms part of a
globe, even though each individual map is approximately flat.
1. Introduction / Principle of equivalence
Lecture 1 11
The essence of gravity in GR - what can't be eliminated by
moving to a new frame of reference - is the spatial variation left
over when the gross effect of gravity is subtracted by moving to
a freely-falling local inertial frame.
Imagine a free-falling sphere of loose gravel, ignoring air
resistance etc. The bottom of the sphere has bigger g than the
top, and at the sides the directions of g converge slightly to point
to the centre of the gravitating mass*. Shifting to the (inertial)
frame of the centre of the sphere means subtracting gaverage :

What's left over is a vertical tension and horizontal compression,

tending to deform the sphere into an ellipsoid. If the gravel was
water, with a rocky ball inside rotating once per day, the whole
lot in free fall towards the Moon, you might recognise these left-
over forces as the tides.

Tidal forces are the essence of gravity.

They encapsulate the spatial variations discussed on the previous

page. In our unit we won't study tidal forces much, but this
concept is central to the tensor formulation of GR as a whole.

* The gravel doesn't have to actually hit the gravitating mass. Free fall is just motion
without forces other than gravity, and can be upwards or sideways (like an orbit) as
well as the classic vertical drop.
1. Introduction / Gravitational time dilation
Lecture 2 12
Consequences of the principle of equivalence
The principle of equivalence quickly leads to two surprising
consequences about space and time in the presence of gravity.
• Consequence #1: gravitational time dilation
Rocket R undergoes constant acceleration g far from gravitating
masses. Time t is measured by inertial observer O. Pulses of
light leave the ceiling A every period DtA. What is the period DtB
of the pulses reaching the floor B, if h is the height of the room?

Let pulse 1 leave A when R is instantaneously at rest relative to

O. It reaches B after time* t1 = h/c, the time of flight for light to
travel distance h.
Pulse 2 is emitted time DtA later. By then, R has accelerated to
speed* u = g DtA. As the pulse travels downwards, the floor
travels upwards through distance* l  u t1 = g DtA × h/c (using t1
as our first approximation for the time of flight of pulse 2).
Pulse 2 therefore travels a shorter distance h - l to B, with a
shorter time of flight (our second approximation) of*
(h - l ) gh
t2   t1 - 2 Dt A
c c

* From the definitions of velocity and acceleration - or "suvat" if you prefer.

1. Introduction / Gravitational time dilation
Lecture 2 13
Here's a time-line for these various events:

The time DtB between the two pulses reaching B is therefore

gh  gh 
Dt B  Dt A  t2 - t1  Dt A  t1 - D t A - t1   1 - 2  Dt A
c2  c 
So DtB < DtA: the pulses arrive at B more often than they leave A.
Nothing paradoxical so far: the times of flight of the pulses are
clearly different, just giving a fancy kind of Doppler shift.
But the equivalence principle says we get the same result in a
room R' at rest in a gravitational field g.

Now the room doesn't move. The times of flight are now the
same. But the equivalence principle demands DtB < DtA still. An
inhabitant of the room, measuring these times, concludes:

Gravitational time dilation: time passes more slowly

lower down in a gravitational field*.

* The difference is tiny on Earth: ~1 ms per century between a typical ceiling and floor.
1. Introduction / Curved spacetime
Lecture 2 14
• Consequence #2: curved spacetime
In a (t versus x) spacetime diagram of the events in R', A and B
are at rest so their worldlines are vertical, with constant values of
x separated by h. This means lines CD and EF are parallel.
The two pulses of light travel at the same speed c, so their
worldlines make the same angle to the axes. This means lines
CE and DF are also parallel. By definition, the quadrilateral
CDFE is therefore a parallelogram.

But CD = DtA and EF = DtB, so the time dilation result means

CD  EF
CDFE is a parallelogram with a pair of unequal opposite sides!
Obviously this contradicts a basic theorem of plane geometry, as
deduced by the ancient Greeks such as Euclid. In fact such a
shape cannot be accurately represented on a flat sheet like the
above diagram. The sheet would need to be curved.

Gravity causes spacetime to be curved.

The geometry of spacetime is "non-Euclidean".

1. Introduction / General relativity
Lecture 2 15
General Relativity
J Wheeler summarised GR in two parts like this:

“ Spacetime tells matter how to move, 

matter tells spacetime how to curve. ” 

Part  (it logically comes first...) means that sources of gravity

cause the geometry of spacetime to depart from the flatness of
the Minkowski spacetime of SR. This is described by the
fundamental equation of GR, the Einstein field equation:

G m  8 GT m m,  = 0, 1, 2, 3
4-D components,
usually 0 is time
Einstein tensor describes
the curvature of spacetime stress-energy tensor
Newton's represents sources of
(in a differentiated form) constant gravity

This simple-looking equation plays the role of a force law, like

F = GMm/r2, telling us how a gravitational field (represented by
the curvature of spacetime) is produced by sources of gravity.
However, it is not as simple as it looks! It's written in the
language of tensor analysis. The two tensors Gm and Tm can
each be thought of as a 4×4 symmetric matrix, with 10
independent components*. Just for the sake of curiosity - you are
not expected to remember this, or even the Einstein field
equation itself, for the exam - take a look at Tm on the RHS:

* m and  act as matrix row and column indices - not exponents / powers
1. Introduction / General relativity
Lecture 2 16

Ordinary mass (represented by energy density) is merely T00, the

time-time component of Tm. Other sources of gravity in GR are
momentum density (space-time components) and pressure and
stress (space-space components). Meanwhile, on the LHS, Gm is
a set of complicated derivatives of something called the metric,
which describes the spacetime curvature geometrically.
The use of tensors is an elegant way to express the principle of
general covariance, which states that the laws of physics should
be valid in all frames of reference, not just inertial ones. But
written out in full, the Einstein equation becomes 10 coupled
nonlinear partial differential equations in non-cartesian
coordinates. Actually solving these equations, to get the metric,
is hideously complicated. Indeed, it has only ever been done
exactly in a handful of very simple cases. We'd need to spend
most of the semester learning tensor analysis before even
beginning any physics. Therefore, I regret to inform you that

In this unit*, we will not learn about how "matter

tells spacetime how to curve".

* To study GR with tensors, and learn about matter telling spacetime how to curve, take
PH40112 Relativistic Cosmology
1. Introduction / General relativity
Lecture 2 17
Part  of Wheeler's comment means that the motion of particles
(and light) is determined by the metric found in part . The rule
is that SR remains valid locally, and that free-falling particles
follow geodesic world-lines. (In a curved space or spacetime, a
geodesic is the nearest thing to a straight line.) This rule plays
the role of an equation of motion, like F = ma, telling us how the
particle moves in a given spacetime.
Although the metric is really a tensor, it is actually possible to
write it as a so-called "line element" without knowing anything
about tensors. Here's an important example, the Schwarzschild
metric for a spherically-symmetric mass M:
-1
 2GM  2 2  2GM 
ds 2  - 1 - 2  c dt   1 - 2  dr  r dq  r sin q df
2 2 2 2 2 2

 cr   cr 

It relates a small spacetime interval ds2 (as in SR) to small

changes in four coordinates. The coordinates resemble spherical
polars (r, q, f) with time t tacked on. The metric generalises the
Pythagoras theorem, and is a geometrical description of how the
mass warps spacetime in its vicinity.
Note there are no tensors to be seen, just ordinary calculus. You
don't need tensors to perform calculations with it either. Tensors
are elegant but for part  they are optional - and we will opt
out*! I am therefore pleased to inform you that

In this unit, we will take spacetime metrics derived

elsewhere to learn about how "spacetime tells matter
how to move".

* Other omitted GR topics: for cosmology, gravitational waves and other astrophysics,
see PH40112 Relativistic Cosmology and PH40113 High Energy Astrophysics; for
exotic matter and warp drives - get back to me when they've become science
2. Newtonian gravity / Time
Lecture 3 18
In the next few lectures, we'll do some background and revision
work using Newtonian gravity, special relativity and the
geometry of curved spaces and spacetimes. Then we'll be ready
for some actual GR.

2. Newtonian gravity
We will look at Newtonian gravity, partly to introduce some
useful tools and partly to see where it differs from GR. Einstein
originally proposed 3 observations, which were realistic to make
using the technology of his day, where GR's predictions differ
from the Newtonian ones. They are called the three classical
tests of GR. To contrast the two theories, we need to know what
the Newtonian predictions are.

Time
The prediction for one test can be written down right away:

Classical test #1, gravitational time dilation

Newton: There is no time dilation - time is absolute.

Newton wrote, “Absolute, true, and mathematical time, of itself,

and from its own nature flows equably without regard to
anything external.” In contrast, we have already seen how the
equivalence principle predicts an inequable flow of time
between the ceiling and the floor.
For the other two tests, we need to study the "planetary" motion
of particles around a gravitating mass - ie, orbits. Most of this
should be revision for you!
2. Newtonian gravity / Effective potential
Lecture 3 19
The effective potential
A test particle m moves near a gravitating point-mass M at O.

• The angular momentum of m, defined by L = r × p, has a

magnitude L that can be thought of in two ways [revision!]:
df
L  mr 2  mvb
dt 2. moment of
1. angular version momentum:
of p = mv : I p ×  distance

For central forces like gravity, angular momentum is conserved.

The mass m is rarely important, so we'll work with angular
momentum per unit mass or specific angular momentum l:
L df
l  r2  vb (i)
m dt
1 2 GMm
• The energy of m = KE + PE  2
mv -
r
Likewise introduce energy per unit mass or specific energy EN. If
we write speed v in its radial and angular components vr and vf:

1  dr  1  df  GM
2 2

EN     r 2   - [v2 = (vr)2 + (vf)2]

2  dt  2  dt  r

(N for vr = dr/dt vf = r 
"Newtonian")
= radial KE/m + angular KE/m + grav. potential
2. Newtonian gravity / Effective potential
Lecture 3 20
This is the energy equation for a particle undergoing 2-D orbital
motion. But, if we substitute df/dt = l/r2 using the angular
momentum formula (i) we can pretend that the particle is
undergoing 1-D radial motion, by combining the angular KE
with the true potential to make an effective potential VN(r). The
key feature is that the form of VN depends only on position r not
velocity, which is what we expect from a potential function:

EN  K N  VN
2
1  dr  GM l2
   -  2 (ii)
2  dt  r 2r

KN = radial KE/m

GM l2
effective potential VN (r )  -  2 (iii)
r 2r

true gravitational potential "centrifugal term"

(from angular KE)

Now we can analyse the radial part of the particle's motion as if

it was just moving along r subject to the effective potential VN.
[When we need the angular part of the motion, we can solve the
angular momentum formula (i).]
Note that KN can never be negative - it's something squared. So
the particle can only be where EN  VN(r), and the gap between
EN and VN relates to the particle's speed (in the r direction) there.
A plot of VN(r) therefore tells us a lot about the possible orbits.
2. Newtonian gravity / Effective potential
Lecture 3 21
Plot VN(r) for a fixed l, and possible orbits for different EN:

EN = Vmin  stable circular bound orbit at A. This is the only

value of r allowed for this energy.
Vmin < EN < 0  elliptical bound orbit between perihelion
(closest point) B and aphelion (furthest point) C.
EN  0  hyperbolic or parabolic escape orbit with perihelion D
and no aphelion.

what the orbits might look like in space:

Only if l = 0 can we get plunge orbits that reach r = 0.

2. Newtonian gravity / Shape of orbit
Lecture 3 22
Shapes of orbits (the Kepler problem)
Differential equations (i) and (ii) can in principle be solved to
find r(t) and f(t). But if we're only interested in the shape of the
orbit not the timing, then finding the "polar plot" r(f) instead is
enough. (If you think the following mathematical trickery is not
obvious - you're right!)
Work with u = 1/r instead of r, and write dr/dt as
dr dr du df
 [chain rule, twice]
dt du df dt
du
 -l [subst (i) and du/dr]
df
Substitute into (ii):
2
EN 1  du  GM 1 2
   - u  u
l 2
2  df  l 2
2
Differentiate this w.r.t. f, EN and l being constants. [NB chain
rule when differentiating (du/df)2, then du/df cancels]:
d 2u GM
 u 
df 2 l2
Our trickery has given us something easy: the equation for
forced s.h.m.. The general solution follows using YR1 methods:
1 GM
u (f )   2 1   sin f  (iv)
r (f ) l
where  is one constant of integration, and the other went to
define f = /2 to be the perihelion (the maximum of u(f) if we
take   0). You may recall  being the eccentricity of the orbit.
2. Newtonian gravity / Shape of orbit
Lecture 3 23
• Bound orbits
If  < 1, u is greater than 0 for all f so r = 1/u is always finite.
This corresponds to a bound orbit. The curve turns out to be an
ellipse of course (or a circle if  = 0), but the important thing for
us is that u(f) in (iv) has a period of exactly 2. This means that
the particle returns to where it started in r after each complete
revolution in f, so a bound orbit is a closed curve. If nothing else
disturbs the orbit, each perihelion is at the same place in space.

And this is the second classical test:

Classical test #2, perihelion shift

Newton: In the absence of other influences, a bound
orbit is closed and its perihelion does not shift.

Incidentally, while we're here, let's capture some results from

(iv) for future reference:
l2
perihelion distance, 1/u at f = /2, is rP 
GM (1   )
l2
aphelion distance, 1/u at f = -/2, is rA 
GM (1 -  )
rP  rA l2
 semi-major axis is a  (v)
2 GM (1 -  2 )
2. Newtonian gravity / Shape of orbit
Lecture 3 24
• Escape orbits
If  > 1, r in (iv) becomes infinite (u = 0) for some f,
corresponding to an escape orbit with a hyperbolic shape. Let's
look at the extreme case where the particle approaches at high
speed v with a large impact parameter b (see p. 19), so that it
will be deflected only slightly by gravity and its speed is roughly
constant. Our aim is to find the small deflection angle q.

Big scale view:

perihelion (and line of
symmetry) at f = /2
r   when f = -q/2

Small scale view:

r  b when f = /2
(a very good approx for
small q)

At perihelion (f = /2), substitute r = b and l = vb (from (i) on

p. 19) into (iv) to get  :
1 GM
 1   
b v 2b 2
v 2b v 2b
  -1  [large v and b]
GM GM
2. Newtonian gravity / Shape of orbit
Lecture 3 25
As the particle approaches from a long way away (f = -q/2),
substitute r   into (iv) to relate q to  :

0  1   sin f
q
 1 -  sin(q / 2)  1 - [small q]
2
Then use our value for  to getq :
2 2GM
q 
 v 2b

The final classical test is about the deflection of light by gravity.

In Newtonian gravity we'll treat light as a particle with a speed
of v = c:

Classical test #3, deflection of light

Newton: Light approaching mass M with (large) impact
parameter b is deflected by an angle of
2GM
q
c 2b

We now have the Newtonian results for all three of Einstein's

classical tests, ready to compare with the predictions of GR.
3. Special relativity / Units
Lecture 4 26
3. Special relativity
My first lecture on SR will revise Dr Sloan's teaching, and the
second lecture will be new material on accelerated motion.
I will adopt two conventions that differ from Dr Sloan's. Here's
the first one:
Relativistic units
Relativity is about spacetime, a unified 4-D continuum, so we
really should measure all four dimensions using the same units.
We therefore adopt SI but with time measured in metres.
1 metre of time is the time it takes light to travel one metre:
1
1 metre = s  3 ns (ie, very small)
2.988  108
distance 1 metre
 speed of light c    1 metre per metre
time 1 metre
These units are therefore sometimes called "c = 1 units". In fact
all velocities v are expressed as unit-less fractions of c.
To convert a physical quantity with ordinary SI units into c = 1
units, multiply or divide by whatever power of (conventional) c
eliminates seconds from the units. For example, G:

6.672  10-11 m3 s -2 kg -1
G
(2.988 108 ms -1 ) 2
no "s"
 7.423  10-28 mkg -1
3. Special relativity / Units
Lecture 4 27
NB sometimes the seconds are hidden inside derived SI units, in
which case use the dimensions of the unit to find them.
eg [joules] = M L2 T-2  kg m2 s-2 so the unit of energy
becomes kg - does that
surprise you?

To convert the other way, from a value in c = 1 units into

ordinary units, do the opposite: multiply or divide the value by
whatever power of (conventional) c restores the right SI unit.
These units simplify relativity (eg, no more "ct" on spacetime
diagrams) but conversion is undeniably awkward. When
evaluating formulae that use c = 1, any values you plug in must
have the right units. You need to get used to this, and may be
tested on it in the exam. (However, it is not my intention to set
exam questions to deliberately catch you out.)
Some experts go one step further and adopt the metre as the unit
of mass as well. For them, 1 metre of mass is the mass that has a
Schwarzschild radius of 2 metres, or G = c = 1. The motivations
for G = 1 are much less compelling than for c = 1, however,
since mass is not another dimension of spacetime. I will keep
kilograms and G  1, but be aware that various combinations of
these conventions are in use "out there" - be careful when
pulling equations from books, papers and online sources. (Don't
get me started on what these people do to electromagnetic units!)

From this point on, unless otherwise noted, all our

derivations, formulae, problems and exam questions
will be in c = 1 units.
3. Special relativity / SR revision
Lecture 4 28
Spacetime
• Event: a particular place at a particular time, specified by a set
of four spacetime coordinates like t, x, y and z.
• Proper distance s : that which is measured by rulers. The ruler
measures the distance between two events at the same time (in
the ruler's frame of reference).
• Proper time t : that which is measured by clocks. The clock
measures the time between two events at the same place (in the
clock's frame of reference).
• Inertial frame of reference: a frame of reference in which
Newton's first law of motion holds; an unaccelerated frame.
• Spacetime diagram: a map of t versus x, y and z. (The
limitations of 2-D paper usually restrict us to x.)

events A and B, seen in two different frames of reference:

frame S, coords t and x frame S', coords t' and x', moving
at velocity v relative to S

• Special relativity: the laws of physics (including the speed of

light) are the same in all inertial frames. This causes the space
(dx and dx') and time (dt and dt') increments between two events
to be different in different frames.
3. Special relativity / SR revision
Lecture 4 29
• Lorentz transformation: relates coordinate increments in
different frames
d x   (d x - vd t ) d x   (d x ' vd t ')
d t    (d t - vd x ) d t   (d t ' vd x ') c = 1,
where 1 remember...
 1
1- v 2

• Simultaneity: observers in different frames may disagree about

which events are simultaneous (or, indeed, co-located).
• Length contraction: moving rulers shorten by  along the
direction of motion.
• Time dilation: moving clocks slow by .
• Rest frame (of a particle): the inertial frame in which the
particle is instantaneously at rest, even if it is accelerating.
• Proper time t : time measured in the rest frame of the particle,
ie the particle's "personal", "experienced" or "wrist-watch" time.
• Worldline: a path in spacetime. Along a non-inertial worldline,
t is the integral of dt 's of infinitesimal inertial segments.

• Principle of maximal proper time: the inertial worldline

between two events maximises proper time.
3. Special relativity / SR revision
Lecture 4 space dimensions
30
The interval ds2 and the Minkowski metric
time dimensions

d s 2  -d t 2  d x 2 [1+1 D]
• d s 2  -d t 2  d x 2  d y 2  d z 2 [3+1 D]

Almost Pythagoras in 4-D, but minus signs distinguish time and

space dimensions. Whether the minus signs are attached to the
time or the space increments
d s 2  -d t 2  d x 2 versus d s2  d t 2 - d x2
is a convention: the "metric signature". It's a free choice (both
are in use) but, once chosen, we must stick to it. I have chosen
the "space-like" or (-, +, +, +) signature in which time has the
minus sign, so that concepts of curved space link more directly
to curved spacetime. This is the other way (besides units) in
which my conventions differ from Dr Sloan's!

• The interval between two events is invariant or absolute -

the same for all inertial frames
d s 2  -d t 2  d x 2  -d t 2  d x2

• Recall (p. 28) that proper distance ds is the distance between

two events in a frame where they are simultaneous. That is, ds is
dx in a frame where dt = 0. From ds2 = -dt2 + dx2 :
ds 2 = ds2
• Similarly, recall that proper time dt is the time between two
events in a frame where they are co-located. That is, dt is dt in a
frame where dx = 0. From ds2 = -dt2 + dx2 :
dt 2 = -ds2
3. Special relativity / SR revision
Lecture 4 31
• Though written as a square, ds2 can be positive, negative or
zero. This suggests there's a thing called ds that can sometimes
be imaginary. Resist that suggestion! There are no complex
numbers in GR*. We don't ever square-root ds2 without first
replacing it with whichever of ds 2 or -dt 2 makes the square
root real (see below). Otherwise, just leave it squared and treat it
as an object that can have either sign.
What the sign of the interval means:

d s 2  -d t 2  d x 2

• Space-like interval AB: ds2 > 0 (more space than time)

The proper distance between A and B is given by ds 2 = ds2. The
proper time between A and B is not defined (no complex
numbers!) since there is no frame in which they are co-located.
• Time-like interval AC: ds2 < 0 (more time than space)
The proper time between A and C is given by dt 2 = -ds2. The
proper distance between A and C is not defined, since there is no
frame in which they are simultaneous.
• Light-like (or null) interval AD: ds2 = 0 even if A  D
The proper distance and proper time between A and D are both
zero.

* Unless they're grafted in from other branches of physics, like wave equations
3. Special relativity / SR revision
Lecture 4 32
• Because ds2 is invariant, so is its sign  an interval that is
time-like (etc) in one frame is time-like (etc) in all frames.
• Matter always travels slower than light, so dt2 > dx2, ds2 < 0,
and its worldline is always time-like. (We all travel more
through time than we do through space.)
• Causality is the idea that a cause must precede its effect
according to all observers. It is an essential part of the concept of
"time". Therefore, no influences can pass between events
separated by a space-like interval because different observers
will disagree about the time order of the events, and some will
say the events are simultaneous. Causally-connected events must
be separated by a time-like or light-like interval.
• Light cone: the set of all light-like worldlines through a given
event A. All events causally connected to A, or reachable from A
(if in the future), or from which A can be reached (if in the past),
lie within its light cone.

Did you notice this vertical bar on the last 3 pages? It

marks the most conceptually-important revision material
for this unit, which you must know AND THOROUGHLY
UNDERSTAND to do well at GR!!
3. Special relativity / SR revision; Acceleration in SR
Lecture 4, 5 33
• The expression for the interval in the flat spacetime of SR
d s 2  -d t 2  d x 2  d y 2  d z 2
is an example of a spacetime metric and is called the Minkowski
metric. Like the more-general metrics we'll meet in GR, it
relates a real physical measurement ds2 (giving a proper distance
or a proper time, as if measured by a physical ruler or clock) to
mathematical coordinates like dx and dt. The coordinate
increments change if you transform to other coordinates (eg a
Lorentz transformation, or cartesian-to-polar), but the physical
interval is absolute and independent of the mathematics.

• Relativistic energy: E = m [E = mc2 in c = 1 units]

 specific energy e  E/m =  [energy per unit mass]

Acceleration in SR
This is a bit of a side-track from GR. But acceleration mimics
the gross effects of gravity, so studying accelerated motion in SR
will give us some hints about gravity before starting GR.
• Proper acceleration a: the acceleration of a particle (or rocket)
R measured in its own (instantaneous) rest frame S' :
d 2s d 2 x dx
a 2  2 while 0
dt dt  dt 
This is the acceleration that R itself feels. If a = 9.8 ms-2, it will
feel like gravity on the surface of the Earth.
In a different inertial frame in which R is not at rest, the same
expression gives a mere "coordinate acceleration", whose value
does not relate simply to what it feels like to R.
3. Special relativity / Acceleration in SR
Lecture 5 34
• Uniform acceleration
Say R's worldline in inertial frame S is the positive-x branch of
x2 - t 2  X 2
where X is a constant. Spend a moment to see that x = X at t = 0
and x  t as t  . The worldline is a hyperbola with 45º
asymptotes through O.

It can be shown (see the problem set, Q7) that R has an

instantaneous velocity (measured in S) of
t
v (vi)
x
(|v| < 1 always) and a constant proper acceleration of
1
a
X
R is like a rocket that accelerates uniformly for all time
3. Special relativity / Acceleration in SR
Lecture 5 35
• Causality of eternal acceleration
Consider event B above the line t = x. No part of R's worldline is
inside the future light cone of B, and B is never inside the past
light cone of any event on R. This means that no objects or
signals originating at B can ever reach, communicate with or
influence R. Objects or signals from R can travel across the line
t = x, but then can never return to R.

Therefore

The line t = x is an event horizon* for R.

A probe P is released from R at t = 0, when R is instantaneously

at rest. P is unpowered, so remains at rest with a vertical
worldline. Meanwhile R continues its acceleration and sees P
drop "downwards" in the negative-x direction, as if off a cliff on
a world with acceleration due to gravity a = 1/X.
Like Sputnik, all P does is emit periodic light (or radio) pulses,
to be detected by R. Plot the worldlines of R, P and the pulses on
a spacetime diagram:

* Likewise, ponder the causal status of the line t = -x and the events below it for R ...
3. Special relativity / Acceleration in SR
Lecture 5 36

The pulses (blue lines at 45º) are emitted regularly by P, but the
time between the arrival of the pulses at R lengthens, and pulses
emitted beyond the horizon t = x never arrive at all. R sees P
approach, but never reach, the horizon, and interprets this as
time running ever more slowly for P.
Yet P just sits at rest in inertial frame S. Nothing special or even
noticeable happens to P, at the horizon or afterwards.
So, the accelerating rocket experiences a gravity-like field with
many of the properties we associate with black holes, yet
described entirely by SR.

• The Rindler frame - an accelerating frame of reference

Say S' is the rest frame of R at a particular event A on its
worldline, when its velocity is v relative to S. As seen from R,
the events that are simultaneous with A define an isochrone - a
line of constant time. Simultaneous means dt' = 0, so the
(reverse) Lorentz transformations give
d x   (d x  vd t )  d x
d t   (d t   vd x)   vd x
3. Special relativity / Acceleration in SR
Lecture 5 37
Dividing gives dt/dx = v, which is not the definition of velocity -
it's upside-down! - but the slope of the isochrone through A.
However, the line joining O to A also has a slope of v because
(vi) on p. 34 tells us that t/x = v. Therefore straight lines t = vx
through the origin are isochrones for R when its velocity is v.

This isochrone result does not depend on the parameter X, which

determines the acceleration of R and the value of x when t = 0. A
whole family of accelerating rockets R1, R2, R3 etc following
similar worldlines but with different X = X1, X2, X3 etc therefore
shares the same isochrones, through events like A1, A2, A3 etc at
which the rockets share the same velocity v.
3. Special relativity / Acceleration in SR
Lecture 5 38
This isochrone exercise has important consequences:
1. The Ri are always at rest with each other, since at a given time
(ie on an isochrone) they have the same velocity v. Instead of a
family of rockets, the Ri could be parts of one (tall) rocket.
2. Their relative positions are Xi at t = 0, and therefore always.
3. X therefore defines a position co-ordinate in an accelerating
frame of reference (the Rindler frame) in which the Ri are at rest.
The hyperbolae x2 - t2 = X2 are lines* of constant X, and the
event horizon t = x is a degenerate hyperbola with X = 0.
4. The straight lines t = vx are lines of constant time in this
frame, and so conceptually define the frame's time coordinate T.
(We'd need to do more work to get T quantitatively, eg equating
it to the proper time of one of the Ri, but we won't bother.)
5. How can the Ri always be at rest relative to each other while
having different proper accelerations a = 1/X? Between adjacent
isochrones (ie for a given increment dT of time coordinate), all
Ri change velocity by the same dv. But from the definition of
acceleration (and using proper variables):
d v  a dt
Therefore, in the time coordinate increment dT, the proper time
increments dt for different X values must differ
1
dt   X
a
It's "gravitational" time dilation again: time passes more slowly
"lower down" at smaller X, but bigger proper acceleration acting
for smaller proper time yields the same change of velocity.

* Dunno what the fancy word for that is: iso-X'es?

3. Special relativity / Acceleration in SR
Lecture 5 39
6. X = 0 means dt = 0: time for the accelerating frame stands
still at the event horizon where v  1, the speed of light. This is
an example of a coordinate singularity in that the Rindler
coordinates (T, X) fail at the horizon, but spacetime itself is well-
behaved there physically (the probe P doesn't even notice it).

In summary, GR and black holes versus SR and accelerating

frames:
effect GR (gravity) SR (Rindler)
non-inertial frames  
more weight (ma) lower down  
more time dilation lower down  
an event horizon  
coordinate singularity at horizon  
uneventful fall through horizon  
a centre of attraction  
curved spacetime  
tidal forces  
physical singularity at centre  
peace, joy and long life beyond the horizon  
Hawking radiation  *
* It's called Unruh radiation instead, but it's analogous
3. Special relativity / Acceleration in SR
Lecture 5 40
• Coordinate time and proper time
The Rindler frame illustrates the distinction between coordinate
time T and proper time t. In everyday life we're used to t being
T2 - T1 (it's t = 1 hour between T1 = 8:00 pm and T2 = 9:00 pm)
but in relativity that ain't necessarily so. In GR we use the metric
to relate coordinate time to proper time.
Coordinate time T is time as in "what's the time?" It's a label
which, with 3 space coordinates, uniquely specifies an event, a
point in spacetime. Like a runner's position in a race, it should
put events in the right order (eg 2nd before 3rd) and identify
simultaneous events (eg joint 4th) for a given frame, but may not
relate simply to the passage of time. Its value is mathematically-
defined, may be different in different coordinate systems, may
not be in time units and may not even have physical meaning.
A spatial analogy would be using house numbers on a street to
identify the location of a bus stop or a particularly interesting
dead squirrel. You expect house numbers to get things in the
right order (the squirrel at no. 17 is further along than the bus
stop at no. 9) but they don't tell you distances. A coordinate
singularity would be like the length of a house tending to zero so
that an infinity of house numbers is crammed into zero distance -
without doing anything at all to the actual street.
Proper time t is time as in "how much time?" It's the physical
time between events, as measured by a clock on a particular
worldline. A different worldline may yield a different t between
the same events, but t is physical not mathematical; it's the same
for a given worldline in all coordinate systems (because it's
given by the interval ds2, and ds2 is invariant); and it's always
measured in time units (for us, metres).
4. Geometry / Flat space
Lecture 6 41
4. Geometry
Here's another Wheeler quote: “Gravity is geometry.”
We're familiar with flat space; what about curved spacetime?
Two routes:
flat space
curved space flat spacetime
curved spacetime

This webpage, by a philosopher of science:

http://www.pitt.edu/~jdnorton/teaching/HPS_0410/chapters/index.html
contains a very readable introduction to curved spacetime. At least read
ch. 24 "General Relativity" just over half-way down. Ch. 18-23, on non-
Euclidean geometry, provide a fuller background.

Flat* space
This is the "Euclidean geometry" of the ancient Greeks (like
Euclid) that you learned at school. Of course you know that the
following propositions are true:

parallel lines never meet circumference: C = 2  r

interior angles: A + B + C = 180º Pythagoras: ds2 = dx2 + dy2

* NB "flat" here doesn't mean two-dimensional; it means not curved

4. Geometry / Flat space
Lecture 6 42
• Straight lines - two equivalent definitions:
1. Keep moving forward, don't deviate:

B
2. Shortest distance: minimise  A
ds

• Pythagoras in 3-D / "solid" geometry

ds2 = dx2 + dy2 + dz2
the metric of flat 3-D space in cartesian coords.

• The metric of flat space in other coordinate systems*

2-D plane polars

ds 2  dr 2  r 2 df 2

3-D cylindrical polars

ds 2  dr 2  r 2 df 2  dz 2 [plane polars with dz]
3-D spherical polars [problem set, Q10]
ds 2  dr 2  r 2 dq 2  r 2 sin 2 q df 2 (vii)

"metric coefficients" are what multiply the infinitesimals

(so the metric coeff of dr2 is 1)

* Because coordinates change direction in general, we'll need to use infinitesimals dx

etc instead of dx etc from now on
4. Geometry / Curved space
Lecture 6 43
Curved space
First explored as differential geometry in the 19th century, but
familiar to cartographers mapping the Earth's spherical surface:
some parallel lines meet
(positive, or spherical, curvature)

or some non-parallel lines

don't meet (negative, or
hyperbolic, curvature)
circumference: C  2  r
interior angles: A + B + C  180º
not the geometry of Euclid & Co: "non-Euclidean geometry"

• Straight lines  geodesics in the language of curved space,

but still defined by "don't deviate" and "shortest distance*".
eg, on a spherical surface, the geodesics are arcs of great circles

The lines of longitude at A and B are "parallel" (both  equator)

and "straight", but they meet at the north pole N.
* Strictly-speaking, that all neighbouring paths are longer - it's a local-minimum thing
4. Geometry / Curved space
Lecture 6 44
• Intrinsic curvature: 2-D cylindrical and spherical surfaces are
extrinsically curved in 3-D space. But in GR we only care about
intrinsic curvature, measured within spacetime without reference
to hypothetical higher "embedding" dimensions*.
The "curved" surface of a cylinder is actually intrinsically flat,
as measured by a 2-D inhabitant. It can be unwrapped to a flat
sheet without tearing, crumpling, or distorting 2-D geometric
figures:
A + B + C = 180º
an intellig-ant
inhabit-ant

unwrap

In contrast, the surface of a sphere really is intrinsically curved.

It can't be flattened without distortion (eg there are no perfect
map projections for the whole world).

A + B + C = 270º

Knowing nothing about the

3rd dimension, the ant can
still tell it lives in a curved
space just by doing 2-D
geometry on (big) circles,
triangles, etc.

* Why? Because spacetime doesn't have an outside to look at it from!

4. Geometry / Curved space
Lecture 6 45
• Spheres are positively curved (the circumferences of circles are
less than 2r, see Q8). So are cones (a circle round the apex has
a short circumference), but all the curvature is at the apex. Our
ant knows it's flat everywhere else, and we can make a cone
from a flat sheet:

On a cone made from the diagram, points A

and A' coincide. There are two straight
lines joining A and B so ABA' is the
edge of a digon: a polygon with 2 straight
sides and 2 corners. Even better, the
straight line AA' joins A to itself and
is the edge of a monogon.
I mention polygons just to show how weird non-Euclidean
geometry gets, even on simple familiar surfaces. They're not
otherwise very important in GR ...

• Negatively curved surfaces are like saddles, or Pringles. The

circumferences of circles are more than 2r (Q8), and the
interior angles of triangles add up to less than 180º. (For weird
polygons in hyperbolic space, look up apeirogon.)
4. Geometry / Curved space
Lecture 6 46
• Local flatness: Even in a curved space*, a small-enough region
is approximately flat and Euclidean (eg city maps don't need
complicated projections).

A + B + C  180º

• Curved metrics: eg for the 2-D surface of a sphere, take the

metric of flat 3-D space in spherical polars

ds 2  dr 2  r 2 dq 2  r 2 sin 2 q df 2 [(vii) on p. 42]

and set r = R (constant) to yield the metric of a curved 2-D space
mapped by coordinates q and f:
ds 2  R 2 dq 2  R 2 sin 2 q df 2
The metric relates the (maths) coordinates dq and df to the
(physics) distance ds. In doing so, it defines the geometry of the
space. However, you can't usually tell just by looking - a
complicated metric could just be flat space in weird coordinates.
The definitive test is to compute the Riemann curvature tensor
from the metric but, even without tensor analysis, we can still
explore curvature by measuring geometric shapes. (See problem
set Q8 for an example using the above metric.)

* If it's well-enough behaved, unlike the tip of a cone!

4. Geometry / Flat spacetime
Lecture 6 47
Flat spacetime space dimensions
time dimensions

This is the 4-D, or 3+1-D, spacetime of SR, originated by

Einstein and Minkowski.
• Straight lines  the geodesics are inertial worldlines. The
"don't deviate" definition means move inertially, without forces.
The "shortest distance" definition means we want the spacetime
interval ds2 as small as possible. But ds2 is negative for the time-
like worldlines of matter and its physical meaning is proper time
dt 2 = -ds2. So the minus sign means that geodesics in relativity
are defined by inertial motion and maximal proper time.
• Geometry: Parallel worldlines stay parallel and never meet
(they are objects with the same velocity) so spacetime in SR is
flat. However, the minus sign in the metric makes it
"Minkowskian" not "Euclidean" (eg the ancient Greeks would
not recognise a 45º line as having zero length). Still, the spatial
part (a "slice" at constant t) is ordinary 3-D Euclidean space.
• The Minkowski metric of flat spacetime (see p. 33) is:
ds 2  -dt 2  dx 2  dy 2  dz 2 [cartesian coords]
or
ds 2  -dt 2  dr 2  r 2 dq 2  r 2 sin 2 q df 2 [spherical polars]
by substituting for the spatial part of the metric. See problem set
Q10.
4. Geometry / Curved spacetime
Lecture 6 48
Curved spacetime
This is the spacetime of GR, generalising features of both curved
space and flat spacetime.
• Straight lines  the geodesics are inertial worldlines, defined
by "inertial motion" and "maximal proper time" (as in SR).
• Geometry: Parallel worldlines don't stay parallel (due to tidal
gravity), so spacetime with gravity is curved: non-Euclidean and
non-Minkowskian.
• Metric: "ds2 = ..." with 3 space coords and 1 time coord (it
may not be easy to tell which is which). It relates mathematical
coordinates to the physical interval ds2, and defines the geometry
of the spacetime.
• Local flatness: A small-enough region is approximately flat
and Minkowskian  SR is always valid locally. Therefore:
• Space-like: ds2 > 0, proper distance ds 2 = ds2
(as in SR)
• Time-like: ds < 0, proper time dt = -ds
2 2 2

• Mechanics: The principle of maximal proper time, together

with dt 2 = -ds2 from the appropriate metric, leads to the
geodesic equation of motion (see L9), which can be solved to
find the worldlines of free-falling (including orbiting) particles
and light.

We will be using this relation, with a negative ds2

(certainly not an ordinary square) all the time.
Recall my advice at the bottom of p. 32!
5. Schwarzschild metric / Meanings of coords
Lecture 7 49
5. The Schwarzschild metric
The solution of Einstein's field equation outside an isolated, non-
rotating, spherically-symmetric source of gravity of mass M (eg
outside a planet or star, or everywhere around a black hole):
-1
 2GM  2  2GM 
ds 2  - 1 -  dt  1 -  dr  r dq  r sin q df
2 2 2 2 2 2

 r   r 
• My shorthand: "Sch." = Schwarzschild.
• Singularities: Infinite ds2 for certain coordinate values, at
which we can't use the metric to do geometry. There are two:
at r = 2GM  rs (the Schwarzschild radius)
and r=0 (the centre)
Both are well inside planets and ordinary stars, so for such
objects r > rs everywhere the metric is valid. This means we can
forget the singularities until we study black holes.

What the Sch. coordinates mean

Actually they don't have to mean anything. They're just labels to
specify events. It so happens that the coords of the Sch. metric
do have meanings - but they're not quite what they appear.
• r >> rs: Far from M (or when M  0 since rs = 2GM), the Sch.
metric becomes flat spacetime in recognisable coords:
ds 2  -dt 2  dr 2  r 2 dq 2  r 2 sin 2 q df 2 [see p. 47]
 (t, r, q, f) are spherical polar coords and time in this limit.
Sch. spacetime is asymptotically flat: SR is valid far from M,
and hence Newton's laws as well for speeds slow compared to
light.
5. Schwarzschild metric / Meanings of coords
Lecture 7 50
• Meaning of t:
The square of the proper time dt between events in the same
place is -ds2 with dr = dq = df = 0 (p. 48):
1/2
 r 
rs  2GM dt  1 - s  dt [from the Sch. metric]
 r
For large r >> rs (where gravity is weak), dt = dt
 coordinate time t is the proper time experienced by an
observer at rest at large r (the observer at infinity).
However, for smaller r, proper time t does not match coordinate
time t  gravitational time dilation. The time t experienced by
an observer at rest at finite r is less than the time t experienced
by the observer at infinity*.

• Meanings of q and f:
The square of the proper distance ds between neighbouring
events is ds2 when the time displacement dt = 0. On a shell of
constant r (ie, dr = 0):

ds 2  r 2 dq 2  r 2 sin 2 q df 2 [from the Sch. metric]

which is the same as the metric of a spherical surface of radius r
(see p. 46).
 q and f are the angles of ordinary spherical polars for any
value of r, not just r >> rs. q is the angle down from an
arbitrarily-defined pole (co-latitude) and f is the angle around
from an arbitrarily-defined prime meridian (longitude).

* The expression on p. 13 from the equivalence principle used approximations. The one
here is exact.
5. Schwarzschild metric / Meanings of coords
Lecture 7 51

Because the constant-r shell has the same metric as a spherical

surface of radius r, it has the same geometry: its circumference
is C = 2r, its surface area is A = 4r2, etc.
There will be times when we don't care much about q and f,
especially since they behave so normally. Then the q and f parts
of the metric can be abbreviated to
d 2  dq 2  sin 2 q df 2 ["solid angle form"]
and the Sch. metric looks like
-1
 r  2  rs 
ds 2  - 1 - s  dt  1 -  dr  r d 
2 2 2

 r   r

• Meaning of r:
This is the trickiest. It is not the distance from the origin - we
can't measure distances through the singularities! But we can
measure the circumference C of a constant-r shell*.
 r is defined as the reduced circumference or circumferential
radius of the surface of constant r
C
r
2

* Corollary: a black hole can have a surface area but not a volume
5. Schwarzschild metric / Meanings of coords
Lecture 7 52
The preceding point is not merely pedantic. The square of the
radial proper distance ds between neighbouring values of r is
ds2 when dt = dq = df = 0, so
dr
ds  1/2
 rs 
1 - 
 r
For finite r, ds > dr: there's more distance between concentric
shells than you'd expect from their circumferences. This is a
non-Euclidean result: an example of gravity warping spacetime.
We can still loosely call r "the radius", but don't forget it isn't!
Here's an (oblique) view of some concentric circles on a flat
Euclidean plane, with their separations Ds = r2 - r1 marked:

And here are constant-r shells in Sch. spacetime (in units of rs),
with their proper-distance separations Ds marked*:

* You'll derive the formula to calculate extended values of Ds in Q13

5. Schwarzschild metric / Meanings of coords
Lecture 7 53
How can we visualise this excess space between shells of given
circumferences?
(a) Don't try: it's non-Euclidean curved spacetime!
(b) OK, if you must: imagine a (Euclidean) hyperspace, with an
artificial fake not-really-there extra z dimension into which we
"push" the shells until their separations are right:

This is an example of an
• Embedding diagram: An imagined surface z(r, f) in cylindrical
polars, with the same relationship between the arc length s and
coords r and f as the Sch. metric at fixed time t. (Spherical
symmetry, so fix q = /2 without loss of generality.)
For a given f, our Sch. proper distance ds should match the arc
length ds for increments dr and dz:

arc length: ds dz
proper distance:

dr  ds 
dr
 rs 
1/2

ds 2  dr 2  dz 2 1 - 
 r
5. Schwarzschild metric / Meanings of coords
Lecture 7 54
Equate ds 's:  rs 
-1

 1 -  dr 2
 dr 2
 dz 2

 r
2
 dz  rs
    [ both sides by dr2]
 dr  r - rs

This is a separable differential equation, which is easy to

integrate to yield the embedding diagram z(r, f):

dr
z (r , f )   dz   rs1/2   2rs1/2 r - rs  k
r - rs

from the  (arc lengths are the same constant of integration (just moves
with the surface either way up) the whole surface up or down)

This surface (shown on the previous page) is known as Flamm's

paraboloid, and I'm sure you've seen it before. Indeed there was
probably a lump of something in the middle, pulling down a
rubber sheet across which you roll a marble to illustrate curved
spacetime acting like a gravitational force blah blah.
So, what exactly is pulling the lump down so that it deforms the
sheet??
In fact an embedding diagram is just a picture of how proper
distances relate to changes in r. Our z axis has no physical
existence - there is no conceptual need for a hyperspace in which
to embed the surface. And there are several reasons (see Q14)
why rolling a marble across the surface does not relate to
particle motion in GR!
5. Schwarzschild metric / Time dilation
Lecture 8 55
Gravitational time dilation
We saw in the last lecture that the Sch. metric predicts time
dilation. This is of course one of the classical tests of GR. Now
we can compare Einstein's answer to Newton's:

Classical test #1, gravitational time dilation

Newton: There is no time dilation - time is absolute.
1/2
Einstein:  2GM 
dt  1 -  dt
 r 
Time slows down in a gravitational field.

The "slow down" factor is close to 1 in most cases:

1/2
 2GM 
1 -   1 - 10 -9 [Earth's surface due to Earth]
 r 
 1 - 10-8 [Earth's orbit due to the Sun]
 1 - 10-5 [Sun's surface due to the Sun]
 1 - 10-3 [white dwarf's surface]
 1 - 0.5 [neutron star's surface]
 1-1 [black hole event horizon]

• Approximate expression: if r >> rs, use a binomial approx

 GM 
dt  1 -  dt [(1 + x)n 1 + nx]
 r 
Mm
 (1 - gr )dt [Newton: F  mg  G ]
r2
(matches the result on p. 13 from the equivalence principle)
5. Schwarzschild metric / Time dilation
Lecture 8 56
• Global Positioning System (GPS)

Radio signals from 3 (out of 24) satellites specify their positions

and times. Comparison with the receiver's own clock  time
travelled by each signal  the distance di to the satellite  the
position of the receiver is at the intersection of 3 spheres of radii
di centred on the satellites. (The satellites carry precise atomic
clocks, but obviously a cheap receiver doesn't. A fourth satellite
is therefore required to provide accurate time information.)
The timing calculations must include the effect of gravitational
time dilation between the satellite's orbit and the ground. There
will also be some SR time dilation due to the motion. Use the
Sch. metric to find the time dilation for the satellite (in orbit "up
there") and the receiver (on a rotating Earth "down here").
Sch. metric for equatorial motion (q = /2) at fixed r (dr = 0):

 2GM  2 2 2
dt 2  -ds 2  1 -  dt - r df
 r 
 2GM  df
 dt 2 1 - - r 2 2  [ df  dt   dt ]
 r  dt

Divide dt for the satellite by dt for the receiver:

5. Schwarzschild metric / Time dilation
Lecture 8 57
1/2
 2GM 2 
 1 - - r 2

sat sat 
dt sat rsat
 
dt rec  1- 2GM
- rrec
2 2 
rec
 
 rrec 
Use a binomial approx to dispose of the square root (both
numerator and denominator are close to 1), and subtract 1 to
write the answer as a relative discrepancy in time:
Dt dt sat - dt rec GM rrec 2
rec
2
GM rsat2
sat
2
   - -
t dt rec rrec 2 rsat 2
For GPS satellites in 12-hour orbits, and a receiver at the equator
(for simplicity), the discrepancy is 38.6 µs / day, of which 45.7
µs / day is due to gravitational time dilation and the rest is due to
the satellite's relative velocity and SR.

• Gravitational redshift
Use the period T of an e.m. wave as a clock.
2
1/2

  2GM 
   1 -  r
T  r 
frequency at  frequency at r

Light of frequency r emitted at r is redshifted to  as it travels

to . (Or, light emitted at  is blueshifted as it travels to r.)
WS Adams' 1925 redshift measurement in light from Sirius B (a
white dwarf) seemed to confirm Einstein's classical test, but was
later shown to be contaminated by light from Sirius A. The first
reliable astronomical redshift was measured in 1962 by JW
Brault in light from the Sun, confirming GR to 5% accuracy.
5. Schwarzschild metric / Time dilation
Lecture 8 58
• Pound and Rebka experiment
The first successful measurement of gravitational time dilation
(confirming the classical test) was not astronomical but an
experiment on the Earth by R Pound and GA Rebka in 1959,
confirming GR to 10% accuracy.  rays from a source at the
bottom of a tower travelled to the top, 22.5 m above. Their
frequency was measured using the Mössbauer effect, a very
precise solid-state method, tuned using the Doppler effect by
oscillating the source.
• Shapiro delay
Light appears to slow down in gravitational fields, eg consider
radial (dq = df = 0) motion of light (ds2 = 0), Sch. metric:
-1
 2GM  2  2GM 
ds 2  0  - 1 -  dt  1 -  dr
2

 r   r 
dr  2GM 
   1 -  ["coordinate velocity" dr/dt]
dt  r 

|dr/dt| < 1 doesn't mean light has actually slowed down. r and t
are just coordinates, not real distance and time except at .
However, it does represent a time delay as seen from large r.
This was measured by II Shapiro by bouncing radar signals off
Venus and Mercury in 1966-7, confirming GR to 20% accuracy.
• Atomic clocks
State-of-the-art strontium clocks are so accurate that the effects
of gravitational time dilation over height differences of 2 cm are
noticeable in the lab.
6. Geodesic equation / General metric
Lecture 9 59
6. The geodesic equation of motion
We've seen the effect of the Sch. metric on time and space, but
not yet the motion of free-falling particles. We need an equation
of motion, to replace F = ma. This is derived from the principle
of maximal proper time (p. 48). The proper time is specified by
the metric, with ds2 = - dt 2 for time-like worldlines.
• A general metric
Use the most-general spacetime coords xm = x0, x1, x2, x3 to write
an arbitrary metric. It is conventional in GR and tensor analysis
to write the indices m = 0, 1, 2, 3 as superscripts (despite not
being exponents or powers, see p. 63), where m = 0 represents
time where possible. Although we don't use tensors in this unit,
we'll still follow these conventions.
The general metric is written in "line-element" form as a sum:
3 3
ds   g m dx m dx
2

m 0  0

The gm are the metric coefficients (and form the metric tensor
when collected together, perhaps as a 44 matrix). Writing it out
in full, the coeff multiplying a diagonal term like (dx3)2 is g33.
But, mixed or off-diagonal or cross terms like dx2dx3 appear
twice, as g23dx2dx3 and g32dx3dx2, even though dx2dx3 and dx3dx2
are the same. We could subtract an amount from g23 and add it to
g32 without changing the overall sum. To remove this unwanted
freedom we define gm to be symmetric (gm = gm), so the term
containing dx2dx3 in the overall sum is 2g23dx2dx3. This reduces
the 16 independent components of gm to 10.
6. Geodesic equation / General metric
Lecture 9 60
A diagonal metric is one where the mixed terms are all zero.
g m  0 if m   (diagonal metric)

In a diagonal metric, the time coord can be recognised

by having a negative metric coefficient g00 < 0,

because we need a time-like ds2 when the other coords are fixed.

• Example: the Sch. metric

-1
 2GM  2  2GM 
ds  - 1 -
2
 dt   1 -  dr 2
 r 2
dq 2
 r 2
sin 2
q d f 2

 r   r 
In the above notation: x0 = t, x1 = r, x2 = q, x3 = f (though we
could shuffle the numbers around for the spatial coords).
-1
 2GM   2GM 
gtt  g 00  - 1 -  grr  g11  1 - 
 r   r 
gqq  g 22  r 2 gff  g33  r 2 sin 2 q
There are no mixed terms like drdq so the metric is diagonal,
and we identify t as the time coord because gtt < 0 (if r > 2GM).
• The geodesic equation
The calculus of variations (which most of you won't know) is
used to maximise the proper time between two given events.
There's a derivation on the Moodle page if you're interested, but
it won't be in the exam. The result is a differential equation for
each coord x(t) as t (acting as a parameter) is varied, for the
inertial / free-fall / geodesic worldline that connects the events.
Here it is for an arbitrary metric, for which we can substitute the
metric of whatever spacetime we're studying:
6. Geodesic equation / Simplifications
Lecture 9 61

d  3 dx b  1 3 3 g m dx m dx
 gab  - 
0 [geodesic
dt  b 0 dt  2 m 0  0 xa dt dt equation]

These are 4 equations, one for each coord xa (a = 0, 1, 2, 3), like

F = ma is really a vector equation with 3 components. (In
contrast, b, m and  are dummy indices that get summed over.)
This is a fundamentally-important equation for GR, but as a
replacement for F = ma it's not pretty! Further tensor notation
improves its appearance a bit (p. 63) but not much. However,
often some useful simplifications will apply.

• Simplifying the geodesic equation:

(#1) If there's a coord xa which none of the gm depends on, then
gm / xa = 0 and the double sum in that xa 's equation vanishes
3
dx b
  gab
b 0 dt
is a constant of the motion, ie it is conserved (because its t
derivative is zero).

(#2) Many metrics (including the Sch. metric) are diagonal:

gm  0 if m  
If there's a coord xa which none of the gm depends on (as in #1),
only the term with b = a survives:
dxa
 gaa (viii)
dt
is conserved.
6. Geodesic equation / Simplifications
Lecture 9 62
(#3) If all the gm only depend on one of the coords xa, the
equations of motion for the other xm can be found as in #1 and #2
above. Then we can get the remaining xa equation of motion
directly from the metric:
3 3
- ds  dt  - g m dx m dx
2 2

m 0  0

3 3
dx m dx
 1  - g m [ dt 2]
m 0  0 dt dt

This can be solved for the dxa/dt that we seek, given that the
other dxm/dt are already known.
If the metric is diagonal, the equation further simplifies to
2
3
 dx m 
1  - g mm   [diagonal metric]
m 0  dt 

All three of these simplifications can be used when considering

motion under the Sch. metric.

• Trivial example: free-fall motion in the absence of gravity, use

the Minkowski metric in cartesian coords
ds 2  -dt 2  dx 2  dy 2  dz 2 [p. 47]
A diagonal metric with coeffs: gtt = -1, gxx = gyy = gzz = 1
None of the gm depends on any of the coords, so (viii) 
dt dx dy dz
 k0  k1  k2  k3
dt dt dt dt
constant time dilation constant velocity (magnitude and direction)
6. Geodesic equation / Simplifications
Lecture 9 63
• Aside: the Einstein summation convention (not examined)
Why write indices as superscripts, risking confusion with
exponents? It's the Einstein summation convention of tensor
analysis: if an index is repeated in a product, once as a subscript
and once as a superscript, summation over that index is implied
and the S signs are dropped. Thus the general metric on p. 59
becomes
ds 2  g m dx m dx
m and  appear as super/sub pairs, so each is a dummy index
summed from 0 to 3. The geodesic equation becomes
d  dx b  1 g m dx m dx
0  gab -
dt  dt  2 xa dt dt
Again m,  and b are summed, but a is not repeated: it is not a
dummy index to be summed over, but simply takes the four
values 0, 1, 2 or 3 in turn, giving four separate equations.
This upstairs-downstairs feature represents an important
distinction within tensor analysis, but it won't be important for
us in PH30101. Because we don't use tensors in this unit, we'll
just keep the S signs. However, we will use the superscripts in xa
etc for consistency, to help those who are going to study GR
further.
7. Orbits / Equations of motion
Lecture 10 64
7. Orbits in Schwarzschild spacetime
Now we can use the Sch. metric and the geodesic equation to
study the motion of free-falling particles - ie, their orbits.
Equations of motion
• Symmetry
The orbit must lie in a plane through the origin. (What could
determine whether it leaves the plane to the right or the left?)
Since the origins of q and f (pole and prime meridian) are
arbitrary, without loss of generality we can orient the coords so
that the q = /2 plane coincides with the orbit. This means
sinq = 1 and dq = 0 and yields the equatorial Sch. metric for the
remaining 3 coords:
-1
 2GM  2  2GM 
ds  - 1 -  dt  1 -  dr  r df
2 2 2 2

 r   r 
• f equation
The metric is diagonal and none of its coefficients depends on f,
so
df df
gff  r2 [(viii) on p. 61]
dt dt
is constant. Since this matches specific angular momentum l in
the Newtonian limit [(i) on p. 19] and is conserved, we will call
it the relativistic specific angular momentum:
df
l  r2
dt
and our "angular velocity" f equation of motion is
df l
 2 [l constant]
dt r
7. Orbits / Equations of motion
Lecture 10 65
• t equation
Motion through spacetime  an equation of "motion" for t(t).
None of the metric coefficients depends on t, so
dt  2GM  dt
gtt  - 1 -  [(viii) on p. 61]
dt  r  dt
is constant. For large r, only SR time dilation makes the
particle's t differ from the static observer's t, so dt/dt =  (p. 29).
Now, in SR, energy per unit mass is e = E/m =  (p. 33). Since
minus our constant is specific energy at r   but conserved
everywhere, we will call it the relativistic specific energy:
 2GM  dt
e  1 - 
 r  dt
and our "time dilation" t equation of motion is
dt e

dt  2GM  [e constant]
1 - 
 r 
• r equation
The metric coeffs do depend on r so there's no constant for r.
But since r is the only such coord, we can use simplification #3
of the geodesic equation (p. 62) and get "radial velocity" dr/dt
directly from the metric and the other equations of motion:
m 2
3
 dx 
1  - g mm  
[diagonal metric]
m 0  dt 
2  df 
2 2 2
 2GM   dt  1  dr 
 1 -   -   - r  
 r   dt   2GM   dt   dt 
1 - 
 r 
7. Orbits / Equations of motion
Lecture 10 66
2
 2GM  e2  dr 
2
1 2 l
 1 -  -   -r 4
 r   2GM 2  2GM   dt  r
 1 - r
2
1 - 

 r 
Solve to obtain:
1/2
dr 
  2GM   l 
2

  e 2 -  1 -   1  2 

dt 
  r   r  

These 3 equations of motion can in principle be solved for the

spacetime coords t, r, q (= /2) and f as parametric functions of
t, tracing out an orbital path.

The effective potential

For now, though, we get more insight by defining an effective
(specific) energy E:
e2 - 1
E [then subst e2 from here]
2
The reason is that, far from M (where e = ) and at non-
relativistic speeds (where  2 = (1 - v2)-1  1 + v2), E tends to the
Newtonian kinetic energy per unit mass ½v2. We can therefore
compare E to the energy EN we used to study Newtonian orbits
in L3, and see that the expressions are almost the same.
The only difference is an extra Einstein term in the effective
potential V(r), compared to (iii) on p. 20. It's negligible for
r >> rs, where GR and Newtonian gravity agree. But it
completely changes the behaviour of V(r) for small r, going to
- instead of + as r  0.
7. Orbits / Effective potential
Lecture 10 67
2
1  dr  GM l 2 GMl 2
GR: E   -  2- 3
2  dt  r 2r r
2
1  dr  GM l2
cf Newton: EN    -  2 [(ii) on p. 20]
2  dt  r 2r

effective effective effective

energy / m KE / m potential V(r)

GM l 2 GMl 2
 V (r )  -  2- 3
r 2r r

gravitational centrifugal term extra

potential (from angular KE) "Einstein" term

Plot V(r) for big-enough* fixed l: the centrifugal barrier rises to a

peak before falling to - as r  0, allowing new types of orbits:

* If l is too small, the peak subsides and V(r) becomes monotonic

7. Orbits / Effective potential
Lecture 10 68
Recall that allowed orbits are where E  V(r):
E = Vmin and r > point F  stable circular bound orbit at A. This
is the only allowed value of r (beyond F) for this energy.
Vmin < E < 0 and r > point F  ellipse-like bound orbit between
perihelion B and aphelion C.
0  E < Vmax and r > point F  hyperbola- or parabola-like
escape orbit with perihelion D and no aphelion.
The above are analogous to the Newtonian cases, but now we
also have:
E = Vmax  unstable circular "knife-edge" orbit at F.
E > Vmax  plunge orbit, the particle falls all the way to r = 0
(if it's moving inwards) even though l  0, or it escapes (if it's
moving outwards).
E < Vmax and r < point F  trapped orbit with aphelion at G
before falling to r = 0.
7. Orbits / Effective potential
Lecture 10 69
We'll need more space to draw possible orbits than we did in the
Newtonian case! Qualitatively, while f (angular) motion
continues, r (radial) motion slows down more (and has more
proper distance to travel) at perihelion than in the Newtonian
case, so the orbit "wraps around" more before r increases again.
The V(r) curves were schematic sketches, but here are some
actual computer-generated orbital paths. The Newtonian curves
are of course also GR curves in the appropriate limit:

bound: escape: plunge:

Newtonian

the only Newtonian

ellipse hyperbola plunge orbit: l = 0

precessing
"ellipse"
relativistic

trapped:

almost the unstable

circular orbit
7. Orbits / Bound orbits
Lecture 11 70
Bound orbits
Now we will study bound orbits in a little more detail. Follow
the procedure from L3 on p. 22, but with GR's l and E:
df
l  r2
dt like (i) and (ii) on
2 pp. 19-20 but with
1  dr  GM l 2 GMl 2
E   -  2- 3 the Einstein term
2  dt  r 2r r
Work with u = 1/r and the chain rule as before to get
2
E 1  du  GM 1 2
   - u  u - GMu 3

l 2 2  df  l2 2
then differentiate w.r.t. f to get an orbit equation:
d 2u GM
 u   3GMu 2
df 2
l 2

just like the Newtonian one but with an extra term at the end.

• Circular orbits
The orbit equation has exact solutions for u = u0, constant:
GM
u0  2
 3GMu02
l
Solve this quadratic* for u0 and use r = 1/u to get the radius:
6GM
r0  1/2
  GM  
2

1  1 - 12   
  l  

* Or, derive more directly by seeking the maximum and minimum of V(r) - see Q20
7. Orbits / Bound orbits
Lecture 11 71
There are real solutions r if the square root is non-negative:

l  12 GM
which are a stable (minimum V) circular orbit with r0 > 6GM and
an unstable (maximum V) circular orbit with r0 < 6GM.

V(r)

As l is reduced, the peak in V(r) declines until at

l  12 GM
the two orbits coincide (at a point of inflection in V) to become
the innermost stable circular orbit or ISCO with r0 = 6GM:

l > 12 GM

l < 12 GM

For smaller l, no circular orbits are possible. This contrasts with

the Newtonian case, where circular orbits exist for any l or r0.
7. Orbits / Bound orbits
Lecture 11 72
• Perihelion shift
The extra term in the orbit equation prevents an exact solution
for non-circular orbits, but we can find an approximation for
orbits that are well outside the Sch. radius. First consider again a
circular orbit (p. 70), for which
GM
u0   3GMu 2
0
l2
Now consider an almost-circular orbit that differs from the
circular orbit by a small f-dependent perturbation f(f):
u (f )  u0  u0 f (f )
Substitute into the orbit equation on p. 70:
d2 f GM
u0  u  u f   3GMu 2
 6GMu 2
f  3GMu 2 2
f
df 2 0 0 2 0 0 0
l

equal  cancel (small)2  neglect

d2 f
  1 - 6GMu0  f  0
df 2
This is the s.h.m. equation, and has the solution


f (f )  A cos 1 - 6GMu0  f
1/2

where the origin of f is chosen so that there is a perihelion
(max f  max u  min r) at f = 0. The period of the
perturbation (ie f at the next perihelion) is
2 ["2/"]
f
1 - 6GMu0 
1/2
7. Orbits / Bound orbits
Lecture 11 73
Since we're assuming r >> rs, GMu0 = GM/r0 is small compared
to 1 and we can safely use a binomial approx:

f  2 1 - 6GMu0   2 1  3GMu0 
-1/2

2
 GM  [u0 = GM/l2 + 3GMu02  GM/l2]
 2  6  
 l 
This period exceeds 2 (one revolution) by
2
 GM 
Df  6  
 l  Df

Subst. for l using the (approximately valid) Newtonian result (v)

from p. 23 for the semi-major axis a in terms of l:
l2 6 GM
a  D f 
GM (1 -  2 ) (1 -  2 )a
Unlike Newtonian orbits, GR orbits don't quite close, which is
one of the classical tests:

Classical test #2, perihelion shift

Newton: In the absence of other influences, a bound
orbit is closed and its perihelion does not shift, Df = 0.
Einstein: In the absence of other influences, the
perihelion of a bound orbit precesses by angle
6 GM
Df  per orbit.
(1 -  )a
2
7. Orbits / Bound orbits
Lecture 11 74
• Mercury
Careful astronomical observations in the late 19th century found
that the perihelion of the planet Mercury shifts at the rate of
574" (arc seconds) per century. Newtonian theory successfully
explained most of it (due to known "other influences" like the
effects of Venus, Jupiter etc), but a residual shift of 43 "/century
could not be accounted for. This value was large compared with
measurement uncertainty, so could not be ignored.
Previously, a 300"/century irregularity in the orbit of Uranus led
UJJ Le Verrier to propose the existence of an unknown planet
further out - Neptune, discovered in 1846. Making the most of a
good idea, he then proposed another unknown planet to explain
the discrepancy in Mercury's orbit. He called it Vulcan, and
expected it to orbit closer to the Sun than Mercury. However, it
was never found.
Then Einstein calculated the perihelion shift for Mercury due to
GR (using the preceding equation) and got the answer:
43 "/century! Hence the anomaly was explained by a new theory
of gravity, rather than a new planet.
This was perhaps not as impressive as the other two classical
tests. Here Einstein explained an old observation, rather than
predicting the outcomes of new ones. (If it had failed the test,
maybe Einstein wouldn't have published his theory and none of
us would ever have known of the failure.)
A much bigger perihelion shift >4º/year has since been measured
in the binary neutron stars PSR 1913+16, also matching the
predictions of GR.
7. Orbits / Radial motion
Lecture 11 75
• Radial motion
For a vertical drop, f does not change at all, so
df
l  r2 0 [f equation of motion]
dt
We therefore find r as a function of time from:
1/2
dr   r   l  
2
  e 2 -  1 - s  1  2   [r equation of motion]
dt   r   r  
1/2
 r 
   e2 - 1  s  (ix) [l = 0]
 r 
For r in terms of coordinate time t rather than proper time t, use
dt e
 [t equation of motion]
dt  rs 
1 - 
 r
1/2
 dr dr dt 1  r   r 
    e - 1  s  1 - s 
2
[chain rule]
dt dt dt e r  r

An important special case is free fall from rest at large r. Rest at

large r means dr/dt = 0 as r  , which means e2 = 1:
1/2
dr r 
  s  (x)
dt r [free fall from
dr
1/2
r   r  rest at large r]
   s  1 - s  (xi)
dt r  r
All these equations can be integrated to give r(t) or r(t), but in
most cases the exercise is no fun at all. We are going to need
them later on though.
7. Orbits / Photon orbits
Lecture 12 76
Photon orbits
We'd like to use our equations of motion from p. 64-65

df l dt e
 2 
dt r dt  rs 
1 - 
 r
to study light paths near gravitating masses. But, for light-like
worldlines ds2 = -dt 2 = 0 so we can't use proper time t! (Recall
from SR that light does not experience proper time.) Instead we
divide the equations and use t instead of t as the parameter:
df b  2GM 
 1 - 
dt r 2  r 
l
where b
e
As we did with particle motion, we get the r equation directly
from the Sch. metric. Set ds2 = 0 (light-like) and divide by dt2:

 df 
2 2
1  dr  r2
1 2     
 2GM   dt  1 - 2GM   dt 
1 -  



 r  r
2
 dr  b  2GM 
2
1
 2    2 1 -  [subst. for df/dt]
 2GM    r 
dt r 
1 - 
 r 

We now have equations of motion for f(t) and r(t).

7. Orbits / Photon orbits
Lecture 12 77
• The meaning of b
Consider a light beam approaching M with impact parameter x.
Its path is straight while it is far away:

x
f  sin f  [r >> x  small angle f]
r
df x
 - 2 [differentiate]
dr r
Also the light is heading almost directly towards M, so
dr
 -1 [speed of light, towards M]
dt
df df dr x
   2
[chain rule]
dt dr dt r
Compare with our f equation of motion for r  :
df b  2GM  b
 2 1 -  
dt r  r  r2
 b is the impact parameter of the light path far from M.

• Effective potential
Our r equation (previous page) can be written as
2
1  1  2GM  dr 
-1
1  2GM 
   1 -    2 
1- 
 b  r  dt  r  r 
2
b
7. Orbits / Photon orbits
Lecture 12 78
which is of the form:
constant = (function of velocity)2 + function of r
just like an energy equation! The light is only allowed where the
"pretend energy" 1/b2 exceeds a photon effective potential
1 2GM
V p (r )  - 3
r2 r
Unlike V(r) for particles there's no local minimum so no bound
orbits, but there is a maximum where Vp(r) = 1/b2crit:

Allowed photon orbits are where Vp(r)  1/b2:

b > bcrit and r > point F  hyperbola-like escape orbit with
perihelion D and no aphelion.
b = bcrit  unstable circular orbit at F.
b < bcrit  plunge orbit, the light falls all the way to r = 0 (if it's
moving inwards), or it escapes (if it's moving outwards).
b > bcrit and r < point F  trapped orbit with aphelion at G
before falling to r = 0.
7. Orbits / Photon orbits
Lecture 12 79
• Orbit equation for light
To find the photon orbits, follow the steps we did on p. 22 and
p. 70 but with the f and energy-like equations:
df b  2GM 
 1 - 
dt r 2  r 
2
1  1  2GM  dr 
-1
1  2GM 
  1 -    2 
1- 
b 2  b  r  dt  r  r 

Use u = 1/r and the chain rule (as before) to write the 2nd
equation with du/df instead of dr/dt, then differentiate w.r.t. f:
d 2u
 u  3GMu 2
df 2

an orbit equation like the one for particles but with a bit missing.
• Deflection of light
We'll use the equation to find out how much light is deflected by
a massive object. The geometry is like that on p. 24:

First consider the straight path followed when M = 0:

1 sin f
b  r sin f  u (f )  
r b
is the equation of an undeflected path with impact parameter b.
7. Orbits / Photon orbits
Lecture 12 80
For weak deflection, where the impact parameter is large
compared to the Sch. radius:
2GM
1
b
consider an almost-straight path that differs from the straight
path by a small f-dependent perturbation f(f):
sin f f (f )
u (f )  
b b
Subst. into the orbit equation, neglect terms with f  2GM/b
(which are second-order in smallness) and use a trig identity to
write sin2f in terms of cos2f:
d2 f 3GM
 f  1 - cos 2f 
df 2
2b
This is the equation for forced s.h.m., which can be solved using
YR1 methods:
GM
f (f )   3  cos 2f 
2b
1 GM 
 u (f )  sin f   3  cos 2f 
b  2b 

To get the overall deflection angle q, use the fact that as r  

(u = 0), incoming rays approach from f = -q/2:
GM
0  - sin q / 2    3  cos q 
2b
Since the deflection angle is small:
sin q / 2   q / 2 cosq  1
4GM
 q
b
7. Orbits / Photon orbits
Lecture 12 81
So this is our final classical test for GR:

Classical test #3, deflection of light

Newton: Light approaching mass M with (large) impact
parameter b is deflected by an angle* of
2GM
q
b
Einstein: Light approaching mass M with (large) impact
parameter b is deflected by twice that angle
4GM
q
b

• Observing deflection of light

The deflection of starlight by the Sun was measured in 1919 by
A Eddington during a total eclipse, when stars near the Sun in
the sky could be observed. His expedition measured a deflection
of 1.98" for a certain star, compared with the GR value of 1.74"
and a Newtonian value of 0.87".
This was the first successful prediction of a new observation by
GR (the perihelion shift of Mercury was an explanation of an old
observation), and it made Einstein world-famous.
The deflection of light has since been observed on a much
bigger scale by the imaging of distant galaxies by closer ones:
gravitational lensing. See Q23 in the problem set.

* I've put c = 1 into the expression from p. 25.

8. Black holes / Singularity at the Sch. radius
Lecture 13 82
8. Schwarzschild black holes
black hole = massive object with an event horizon
event horizon = surface that can only be crossed in one direction
Sch. black hole = spherically-symmetric black hole, without spin
Recall that the Sch. metric has singularities (infinities) at r = 0
and r = rs. Because the metric is only valid outside an object, we
ignored them for planets and ordinary stars that are bigger than
their own Sch. radius. For black holes, where the mass is entirely
within rs, the r = rs singularity lies in a valid part of the metric
and can't be ignored any more.

The singularity at the Schwarzschild radius

-1
 r  2  rs 
ds  - 1 - s
2
 dt   1 -  dr 2
 r 2
d  2

 r   r

• Sch. metric   at r = rs  apparently weird things:

1/2
Time dilation:  r 
dt  1 - s  dt [from p. 55]
 r
 time stops (and infinite redshift) at rs?

Vertical drop, eg from rest at large r:

1/2
dr r   r 
"coord velocity"  -  s  1 - s  [(xi) on p. 75]
dt r  r
 falling objects stop at rs?
8. Black holes / Singularity at the Sch. radius
Lecture 13 83
dr
Inside r < rs: coord velocity: 0 [previous with r < rs]
dt
 free-falling objects rise upwards to rs?
or, time goes backwards?

• Types of singularity:
A metric relates the geometry of spacetime (ie physics) to the
coords used to pinpoint events (ie maths). Either can blow up:
physical singularity: spacetime becomes infinitely curved?
For example, the tip of a cone is an infinitely-curved physical /
geometric singularity.
coordinate singularity: the coord system fails but the physics is
well-behaved?
For example, a 2-D spherical surface of radius R has no physical
singularities - it's smoothly curved. A sensible way to map it is to
use the spherical polar angles q and f:

ds 2  R 2 dq 2  R 2 sin 2 q df 2 [from p. 46]

This is a well-behaved metric, with no singularities for any

values of q and f, which cover the whole surface.
An alternative (and less intelligent) way to measure from the
pole is to use the projected distance r = R sin q on a tangent
plane instead of the angle q. It certainly acts as a coord in the
northern hemisphere: pairs (r, f) uniquely specify points, which
is all a coordinate system has to do.
8. Black holes / Singularity at the Sch. radius
Lecture 13 84

r can be substituted for q to give the metric in (r, f) coords:

R2
ds  2
2
d r 2
 r 2
d f 2

R - r2
This metric has a singularity at r = R. The physical surface
remains well-behaved, so it is only a coordinate singularity.
Because we constructed that metric ourselves we can see why it
goes wrong: the coords just stop working at the equator. But if
we're given the metric without explanation, it's not so obvious.
• Which type is rs?
Re-visit the vertical drop, but with proper time instead:
1/2
dr r 
 - s  [(x) on p. 75]
dt r
This describes how r changes with time t experienced by the
falling object, rather than the time t of an observer at infinity.
There is no strange behaviour at or within rs. The object keeps
falling and reaches r = 0 in finite proper time (the equation is
easily integrated, see p. 92). It seems perfectly well-behaved,
and indicates that rs is merely a coordinate singularity, due to a
failure of the t coord. (We'll prove it in the next lecture by
finding alternative coords that eliminate the singularity at rs.)
8. Black holes / Singularity at the Sch. radius
Lecture 13 85
• Light cones
Divide the Sch. metric by dt2 and solve for dr/dt:
  ds 2 d 
2 2
 dr   rs   rs
2 2

   1 -   1 -  2 - r 
 dt   r  r   dt dt 2 
Look at the factor in {}. Remember that d2 is short for
d 2  dq 2  sin 2 q df 2 [from p. 51]
which is some squared things added together and so must be
either positive or zero. dt2 is another ordinary square. The
interval ds2 looks like a square but can be negative and, indeed,
for allowed (time-like or light-like) worldlines must satisfy
ds 2  0
Therefore the {} as a whole factor must be negative or zero.
Multiplying by 1 - rs/r gives a negative value if r > rs or a
positive value if r < rs (or zero in either case). Thus
r  rs r  rs
2 2 2 2
  dr   rs   dr   rs 
   1 -     1 - 
 dt   r   dt   r 
To plot the light cone at a given event, we need to know what
values of dt/dr (the slope on a t versus r spacetime diagram) are
allowed. This is the reciprocal of dr/dt, so
1 1
slope  slope 
r rs
1- s -1
r r
Outside rs, the light cone is steeper than a certain value. Inside
rs, the light cone is shallower than a certain value. At r = rs, the
sides of the light cone are vertical (infinite slope).
8. Black holes / Singularity at the Sch. radius
Lecture 13 86

For large r the light cone is upright with 45º sides, as in the
Minkowski spacetime of SR (p. 32). As r decreases the cone
narrows, to become infinitely thin vertically at rs. Then there's a
discontinuity (because of the singularity) to a wide-open light
cone on its side, facing towards r = 0. As r decreases further the
light cone closes to become infinitely thin horizontally at r = 0.
Look again at the Sch. metric:
-1
 r  2  rs 
ds 2  - 1 - s  dt  1 -  dr  r d 
2 2 2

 r   r

When r < rs, the coeffs of dt2 and dr2 swap signs. Recall that for
a diagonal metric (like this one) the negative coeff identifies the
time coordinate. So, for r < rs, r becomes the time coordinate
and t becomes a space coordinate! The light cones reflect this.
The Sch. radius rs is an event horizon passable only inwards.
Just as t inevitably increases outside the horizon, an object's r
inevitably decreases once inside. Then the object's unavoidable
future is r = 0 where (or rather, when!) its worldline terminates.
On the other hand, t can go in either direction inside. It follows
from the t equation of motion on p. 65: dt/dt = e/(1 - rs/r)
that both positive and negative energies e are possible inside the
horizon, since dt/dt can take either sign. Negative e is not
possible outside the horizon, where t is always increasing.
8. Black holes / PG coords
Lecture 14 87
Painlevé-Gullstrand (PG) coordinates
The Sch. time coord t (time at rest at large r) behaves badly at
r = rs but the proper time t of a free-falling object behaves well.
It is therefore just a coordinate singularity that can be eliminated
by replacing t with a different time coord. There are several
ways to do this. Here is P Painlevé and A Gullstrand's:
• The PG time t' (of an event): the time read from a free-falling
clock that was dropped from rest at large r and that happens to
fall past the event as it occurs. The PG coords are t' together
with the three spatial coords (r, q, f) of the Sch. metric.
Note that t' is well-defined whatever the observer's location and
motion, including inside the event horizon. Most importantly it's
measured locally to the event, unlike t. Now we'll relate t to t'.
If we start with the clock's reading t' at the event and subtract the
proper time tjourney of the clock's fall from its release at large r,
we get the clock's reading when it was dropped. This is also the
Sch. time when it was dropped, because until then it was at rest
at large r and so read Sch. time. If we then add the Sch. time
tjourney of the clock's fall, we get the Sch. time t of the event:
t  t  -t journey  t journey
and how it varies with the event's r coord:
dt dt  dt journey dt journey
 -  [differentiate]
dr dr dr dr
But we already know dr/dt and dr/dt for the journey of an object
falling from rest at large r, from (x) and (xi) on p. 75. Here they
are again, with the "journey" subscript added and the appropriate
choice of sign for an inward fall:
8. Black holes / PG coords
Lecture 14 88
1/2
dr r 
 - s 
dt journey r
1/2
dr r   r 
 -  s  1 - s 
dt journey r  r
Substituting these in, and multiplying by dr:
1/2
r  rs 
1/2

1/2    
r  rs 
dr  dt  -   dr
r
dt  dt     dr -
 rs   rs   rs 
 1 -  1 - 
 r   r
This can (with difficulty) be integrated to give a transformation
between t and t', but we won't bother because we can substitute
it straight into the Sch. metric to yield the PG metric:
1/2
 r  r 
ds  - 1 - s  dt 2  2  s  dt dr  dr 2  r 2 d 2
2

 r r
The physics is unchanged. It's still Schwarzschild spacetime, just
using different coords. Notice:
No more singularity at rs! (There is still one at r = 0.)
gtt is positive for r < rs but this does not mean t' is a space coord
inside the horizon. The "negative-g" rule applies to diagonal
metrics, and this one has a mixed term. In fact t' is manifestly a
time coord for all r, because it's measured by a (local) clock.
For r < rs every term on the RHS except the mixed term is
positive, but time-like or light-like (ie allowed) worldlines
require ds2  0. Therefore it is necessary that dt'dr < 0: future
(dt' > 0) motion inside the horizon must be inwards (dr < 0).
8. Black holes / PG coords
Lecture 14 89
• Light cones
Writing b for the mixed velocity* of the falling PG clock:
1/2
dr r 
b  s 
dt journey r
Divide the PG metric by dt'2 (like we did with the Sch. metric):
 ds 2 2 d 
2
 dr   dr 
2

    2 b     ( b - 1)   2 - r
2
2  (xii)
 
dt  
dt  dt dt  
The {} factor on the RHS is a t' version of the one for the Sch.
metric on p. 85 and, for the same reasons (ds2  0 and ordinary
squares  0), must be negative or zero for allowed worldlines:
2
 dr   dr 
   2 b    ( b 2
- 1)  0
 dt   dt 
As a function of dr/dt', the LHS is an upward parabola (+ve for
large |dr/dt'|) that is negative only between its two roots:

To find the roots (and the bounds on dr/dt'), factorise the LHS:
 dr   dr 
 dt   b - 1  dt   b  1  0
[roots dr/dt']

dr dr
  1- b and  -1 - b
dt  dt 

* It's "mixed" because it's the change of coord distance r w.r.t. proper time t.
8. Black holes / PG coords
Lecture 14 90
So the condition for allowed worldlines is

dr   rs  
1/2

-1 - b   1- b b    
dt    r  
As in the Sch. case (p. 85), the slopes of the light cones on a t' vs
r spacetime diagram are the reciprocals of dr/dt'. Since b < 1
outside and b > 1 inside the horizon, the light cones look like:

Now there is no discontinuity of behaviour at rs. As r decreases,

the initially-upright light cone tips over (and narrows) towards
r = 0. At rs the outward slope is vertical. Inside the horizon both
edges have -ve slope and all future worldlines are inward.

• Easier calculation of light cones

Actually we can start with (xii) - previous page - and simply say
the light cones are bounded by the worldlines of light (ds2 = 0)
travelling radially inwards and outwards (d2 = 0). This gives
the bounding dr/dt' more directly.
But, it relies on continuous behaviour (ie no singularity) at rs.
Try it with the Sch. metric (p. 85) and you miss how the light
cones go sideways inside the horizon!
8. Black holes / PG coords
Lecture 14 91
• The river model
There are other ways to get rid of the singularity at rs, but the PG
metric in particular has an intuitively-appealing property. The
metric can be written (prove by working backwards) as
ds 2  -dt 2  (dr  b dt )2  r 2 d 2

 dr 2
This looks like the (flat) Minkowski metric of SR in spherical
polars, but with an r' coord that "flows inwards" at the speed
1/2
r 
b  s 
r
of the falling clocks that defined the PG time t'. (A point of
constant r', ie dr' = 0, moves as dr/dt' = -b.) It's as if spacetime
itself flows inwards at this speed, and the PG clocks are carried
along inertially in the flow. The flow gets faster at smaller r, and
exceeds the speed of light* (b = 1) inside the horizon.
Particles and light move in this flowing spacetime like fish in a
river. For r > rs, fast-enough fish can overcome the flow and
swim upstream. For r < rs, even the fastest fish (light) cannot
make headway against the flow, and are carried downstream.
Is this real physics? It's no more fanciful than the cosmological
picture that "spacetime itself is expanding" after the Big Bang...

* Should we worry about the flow exceeding the speed of light? No. SR is always valid
locally (p. 48) so local measurements in the river will always be limited by the speed of
light. But attempts to measure non-locally (measuring "here" the speed of something
"there") can give strange answers, like b > 1. It's like at the edge of the observable
Universe, where galaxies recede at the speed of light: another non-local velocity.
8. Black holes / The central singularity
Lecture 15 92
The central singularity (r = 0)
We now know that rs is merely a coordinate singularity. There is
no extreme physics at rs, and the singularity can be removed by
changing the coords. What about the other singularity, at r = 0?
Tensor analysis: The Riemann curvature tensor is infinite at
r = 0, meaning that spacetime is infinitely curved. No change of
coords can eliminate the singularity. It is a real / physical /
geometric singularity, and is referred to as the singularity.
• A journey to the singularity
An observer falls from rest at large r (like a PG clock):
1/2
dr r 
 - s  [(x) from p. 75]
dt r
This equation is easily integrated to find the proper time to fall
from arbitrary r to the singularity:
3/2
2r  r 
t s  (xiii)
3  rs 
so the time spent inside the horizon is
2r
t s [r = rs]
3
For a solar-mass black hole (with a mass of the order of the
Sun's), t is a few microseconds. You'll need a much bigger black
hole to "enjoy" the experience of life inside.
8. Black holes / The central singularity
Lecture 15 93
• Tidal forces
An acceleration a can be defined for our observer:
d  dr  d  dr  dr
a (r )      [chain rule]
dt dt  dr  dt  dt
r *
 - s2 [dr/dt on previous page]
2r
Of course the observer does not feel the average acceleration,
being in free fall. But because a varies with r, different parts of
the observer's body try to fall with different accelerations, setting
up tidal forces. We can use a to work these out because r is a
proper distance for our observer (in the PG metric, keep t' and
the other space coords fixed and get ds2  ds2 = dr2).

If there's an acceleration difference Da along the distance Dr

between the observer's extremities, a tensile tidal force (see Q2)
of the order of
 da 
F ~ mDa  m   Dr [chain rule]
 dr 
will be set up. The force depends on their height Dr, and on the
tidal acceleration gradient da/dr. From a(r) above:
da rs
 3
dr r

* Interestingly, gives the Newtonian expression if you substitute r s.

8. Black holes / The central singularity
Lecture 15 94
The tidal gradient da/dr   as r  0: an example of an
extreme physical effect, and strong evidence for a physical
singularity at r = 0 even without knowledge of tensors.
The observer's body (and, soon afterwards, constituent atoms)
will be irresistably stretched and torn apart as the singularity is
approached, a phenomenon known as spaghettification.

• Will it hurt?
Assume we can endure a tidal acceleration gradient of
da
q
dr
The tidal gradient will reach this value at the "pain radius"
1/3
 rs 
rpain    [da/dr on prev. page]
 
q
The pain will last at most until r = 0, for a "pain time" given by
substituting rpain into (xiii) on p. 92:
3/2
2rpain 2
t pain  
3rs1/2 3q1/2
Remarkably, this is independent of the mass of the black hole. It
only depends on the observer's pain threshold. Reasonable
values of q [see Q29(d)] give tpain of the order of 0.1 s, similar to
a typical reaction time. Spaghettification probably doesn't hurt!
8. Black holes / The central singularity
Lecture 15 95
• Maximising survival time
The "free-fall from large r" trajectory doesn't maximise the time
experienced inside the black hole. From the Sch. metric:
dr 2 r 
dt  -ds 
2
-  s - 1 dt 2 - r 2 d  2
2

 rs   r 
 - 1
r 
 0 inside the horizon
To maximise t we need (a) d2 = 0 (ie radial motion) and (b)
dt = 0. From the t equation of motion for free fall:
dt e
 [see p. 65]
dt  2GM 
1 - 
 r 
ie, dt = 0  free fall with e = 0. So, to maximise experienced
time, get yourself into a radial e = 0 trajectory* asap then switch
your engines off. The maximum proper time is
dr
[subst r = rs sin2q]
0 0
t  - 1/2
 -2rs  sin 2 q dq
 /2
rs
 rs 
 - 1 
r 

 rs
2
which for all the fuss isn't much better than
2r
t s
3
for free fall from rest at large r (p. 92).

* An advanced optional thinking question: how do you get into an e = 0 trajectory?? A

closely-related question, given the t equation of motion (p. 65): how do you change your
motion in the t direction, which is a space coord inside the black hole?
8. Black holes / The central singularity
Lecture 15 96
• So, what happens at r = 0?
Nobody knows.
The singularity cannot be observed: it is "clothed" by the event
horizon.
All of the black hole's mass M (and anything that falls in) goes
there.
There are infinite tidal forces and an infinite spacetime
curvature.

It's worth remembering that, if a physical theory gives an infinite

answer, it's either a somewhat-abstract concept (eg infinite
temperature means that all of a system's microstates have equal
occupancy while it is in thermal equilibrium) or it means the
theory has broken down (eg the ultraviolet catastrophe of black-
body radiation showed that the classical Rayleigh-Jeans theory
had failed, leading to Planck's founding of quantum physics).
It's therefore likely that GR ceases to accurately describe physics
at the centre of a black hole. It is after all a classical theory: a
theory of quantum gravity (see L20) may smooth out the
all-at-once nature of the singularity and make it "fuzzy".
Nevertheless, this is speculation, and the only reliable answer so
far is the one at the top of this page.
8. Black holes / KS coords
Lecture 16 97
Kruskal-Szekeres (KS) coords
This is another way to visualise Sch. spacetime and avoid the
singularity at rs. Its advantages over PG are that the metric is
diagonal, its light cones are simple, and it has great conceptual
power. The disadvantages are that it is mathematically abstract
and not much use for calculating orbits.
You are advised to focus on the concepts rather than the maths!
We replace both t and r this time, with new coords u and v:
uv
t  rs ln
u -v
r  r / rs
 - 1 e  u - v
2 2

 rs 
The expression for t is straightforward but note that r is only
implicitly defined: it can't be written "r = ..."
Substitution into the Sch. metric and 2 pages of dedicated
algebra (Q30, only completists need attempt) yield the KS
metric:
4rs3 - r / rs
ds  -
2

r
e  dv 2 - du 2   r 2 d 2

There is no singularity at rs. (There is still one at r = 0.)

It still contains r - we can't get rid of r, so instead treat it as an
implicit function of u and v.
It's diagonal - no mixed terms. The coefficient of dv2 is always
negative, so v is always time-like and u is always space-like.
8. Black holes / KS coords
Lecture 16 98
• Light cones
Rewriting the KS metric and dividing by dv2:
r r / rs  ds 2 2 d 
2
 du 
2

   1 3 e  2 - r 
 dv  4rs  dv dv 2 

0
For the familiar reasons (ds2  0 and ordinary squares  0) the
second term on the RHS must be negative or zero for allowed
worldlines. This means
dv
1
du
The edges of light cones on a v versus u spacetime diagram are
at 45º everywhere - just like in flat Minkowski spacetime!
However, the other features of the spacetime diagram must be
severely deformed to accommodate this very-simple rule for
light cones.
To map out a KS spacetime diagram we accept (as for the Sch.
and PG metrics) that each point on a 2-D chart represents all
values of the angular coords: it will be a plot of v versus u alone.
To plot lines of constant r, investigate u and v   and 0 in the
r definition
r 
u 2 - v 2   - 1 e r / rs
 rs 
This describes hyperbolae with 45º asymptotes. For r > rs the
curves are more vertical than horizontal, and vice versa for
r < rs. For r = rs we get the asymptotes themselves, and the
singularity r = 0 becomes the curve v2 - u2 = 1.
8. Black holes / KS coords
Lecture 16 99

NB singularities
are usually drawn
as zig-zag lines

To plot lines of constant t, solve the t equation on p. 97 for v

assuming* u > v:
 et / rs - 1 
v   t / rs u
 e 1
These are straight lines through the origin, with slopes from -1
for t  - to +1 for t  .

* Repeat but assuming u < v instead to give lines of constant t inside the horizon - and
help advanced thinkers to answer the question in the footnote on p. 95.
8. Black holes / KS coords
Lecture 16 100
• Relation to the Rindler frame
Our plots of Sch. coords (t, r) on KS spacetime diagrams (v, u)
look a lot like plots of Rindler coords (T, X) on Minkowski
spacetime diagrams (t, x) from p. 34-37. The light cones and
event horizon (but not the singularity) are analogous. Both can
be used, in similar ways, to qualitatively study causal relations
between different observers, eg the probe P falling through the
horizon, or what the rest of the Universe looks like to P, or
whether P gets to see the singularity once inside the horizon.

(figs copied from pp. 36-37)

The Minkowski coords were "natural" for inertial frames in SR,

Rindler coords representing a more-complicated accelerating
frame. So it looks like KS coords are natural for spacetime
around a black hole, with Sch. coords representing a more-
complicated accelerating frame! Indeed, an observer at rest in
Sch. coords does feel an acceleration due to gravity.
8. Black holes / KS coords
Lecture 16 101
• The Kruskal extension (speculation alert!)
The Sch. coords sit in the top-right half ("east and north") of the
KS diagram. The bottom-left half ("west and south") is a
theoretical extension. The Rindler analogy suggests what the
extension represents: when prompted on p. 35, did you think
about the line t = -x in the Rindler frame (t = +x being the event
horizon)?

East quadrant: the rest of the Universe, ie normal space outside

the black hole, r > rs. Future worldlines (see the light cone...)
lead only elsewhere in normal space or into the black hole, but
past worldlines can lead from the south quadrant. There is no
causal connection at all (past or future) with the west quadrant.
North quadrant: the black hole, r < rs. Future worldlines lead
only to the singularity, but past worldlines lead from any of the
other three quadrants. Matter and light can enter through the
horizon but never leave.
8. Black holes / KS coords
Lecture 16 102
West quadrant: Exactly like the east quadrant (ie normal space)
but causally disconnected from it  it's another Universe!
Observers in the black hole can see both universes and meet
travellers from the other universe (shortly before being
spaghettified).
South quadrant: a time-reversed black hole, called a white hole.
Past worldlines lead only from the singularity, but future
worldlines lead to any of the other three quadrants. Matter and
light can leave through the horizon but never enter. An object
leaving the white hole follows the time-reverse of the trajectory
of an object falling into the black hole. If you look at the black
hole from outside, what you actually see is the white hole.
The Einstein-Rosen bridge: Consider the "slice" of spacetime
represented by the u axis. On the KS diagram, it's the horizontal
line through the diagram's origin where the horizons intersect.
Here's a diagram with contours of r marked:

As you move from right to left, see how r decreases from very
large values down to rs at the origin, then increases again in the
other universe. The embedding diagram along this path is just
two Flamm's paraboloids (p. 53) joined at their throats:
8. Black holes / KS coords
Lecture 16 103

It's called the Einstein-Rosen bridge, and it's the simplest

example of a wormhole in spacetime. Unfortunately, it can't be
used to travel between two universes. As we've already seen,
they are causally disconnected: all the worldlines through the
wormhole are space-like.
To add to the sci-fi fan's disappointment: the extra universe and
the white hole are valid solutions of Einstein's equation, but they
only appear for eternal black holes with no beginning in time:
the white-hole horizon is t  -. For astrophysical black holes
formed by the collapse and/or merger of stars sometime in the
finite past, the in-falling matter forms a boundary beyond which
the "vaccum solution" of Sch. spacetime is not valid.
9. Rotating black holes / Kerr metric
Lecture 17 104
9. Kerr (rotating) black holes
In L7 - L16 we studied the Schwarzschild solution for spacetime
outside a non-rotating, spherically-symmetric source of gravity.
But real astrophysical objects (including black holes) rotate. As
an example of how hard it is to find out how "matter tells
spacetime how to curve", 45 years passed between K
Schwarzschild's work and R Kerr's solution of Einstein's field
equation outside a rotating, axially-symmetric point-source of
gravity of mass M and angular momentum* (or spin) J.
The Kerr metric
In Boyer-Lindquist coords (t, r, q, f), the Kerr metric is
ds 2  gtt dt 2  g rr dr 2  gqq dq 2  gff df 2  2 gft df dt
where
 rr  2 a 2 rs r sin 2 q  2
gtt  - 1 - s 2  gff  r  a   sin q
2

 r  r 2
 
r2 ars r sin 2 q
g rr  gqq  r 2
gft  gtf  -
D 2
r
D  r 2  a 2 - rs r r 2  r 2  a 2 cos2 q
aJ /M the angular momentum* parameter

Positive f is defined to be in the direction of spin, so a  0.

r   (far away)  reduces to Minkowski (flat) metric
a  0 (no spin)  reduces to Sch. metric
rs  0 (no mass) but fixed a  reduces to flat spacetime
Mixed term dfdt  the sign of df/dt matters
* We'll use J or a for the spin angular momentum of the central mass, and L or l for the
orbital angular momentum of a particle moving around it.
9. Rotating black holes / Kerr metric
Lecture 17 105
• Singularities: infinite ds2 when r2 = 0 or D = 0
r2 = 0  r 2  a 2 cos 2 q  0  r = 0 and q = /2

This corresponds to the central singularity of Sch. spacetime

when a = 0, so it is a physical singularity. (Why does the value
of q matter if r = 0? Because in Boyer-Lindquist coords r = 0 is
not a point when a  0 - see Q34(c)).
D=0 
r 2 - rs r  a2  0
1/2
rs  rs  
2

Quadratic: r     - a 2

2  2  

The bigger solution r+ is rs when a = 0 (the Sch. limit), so it is a

coordinate singularity at the event horizon rH:
1/2
rH  GM   GM  - a 2  [event horizon, where D = 0]
2
 
Indeed, notice that grr changes sign at r < rH, suggesting that r
becomes time-like. (Although the rule about the sign of gmm
indicating the time coord only works for diagonal metrics, the
mixed term in the Kerr metric doesn't involve dr.)
For a rotating black hole (a  0), rH < rs.
The other solution r- is another coordinate singularity called the
Cauchy horizon. In the Sch. limit a = 0, r-  0 and vanishes.
The Cauchy horizon is not very relevant since it is inside the
event horizon.
9. Rotating black holes / Kerr metric
Lecture 17 106
• The Cosmic Censorship Principle
If a > GM there's no real value for rH and so no horizon. Without
an event horizon, there's no black hole. The r2 = 0 physical
singularity becomes a naked singularity, "un-clothed" by a
horizon. We could observe it, or visit it and return.
The conjecture that physics does not allow naked singularities is
called the cosmic censorship principle. For a Kerr black hole:
a  GM
This conjecture is not proven, but is very likely and widely
believed. For example, evidence suggests that an a > GM black
hole cannot form. The limiting case of a Kerr black hole with
a = GM is called an extremal black hole.

• The static limit

The time dilation of an object at rest relative to the observer at
infinity is given by dt 2 = -ds2 with dr = dq = df = 0:
1/ 2
 rr
dt  - g tt1/ 2 dt   1 - s 2  dt
 r 
so there's infinite time dilation (and redshift) when gtt = 0:
rs r
1- 0  r 2 - rs r  a2 cos2 q  0
r2
like the quadratic for the horizons but with cos2q, and solutions
1/2
rs  rs  
2

r     - a cos q 
2 2

2  2  
The r'- solution is inside the horizon, so r'+ = rE:
9. Rotating black holes / Kerr metric
Lecture 17 107
1/2
rE  GM   GM  - a 2 cos 2 q 
2
[static limit, where gtt = 0]
 

marks the static limit or infinite redshift surface. Unlike the

horizon rH it depends on q. At the poles (q = 0 and ) it
coincides with the horizon but otherwise is outside the horizon,
reaching rE = 2GM = rs at the equator. The space between the
horizon and the static limit is called the ergoregion.

• The ergoregion
In Sch. black holes the event horizon and static limit coincide. In
Kerr black holes there's the ergoregion in between, where escape
to infinity is still possible but there's some kind of problem with
time for static objects. What happens there?
Consider light (ds2 = 0) moving only in the f direction "along a
line of latitude" (dr = dq = 0). The Kerr metric becomes
0  gff df 2  2 gft df dt  g tt dt 2

which is a quadratic in df/dt, with solutions

* The figure has poetic licence: r is (of course) not a radius measured from the centre!
9. Rotating black holes / Kerr metric
Lecture 17 108
df 
gft   
1/2

gff 
   -  
1  1 - g tt  
dt gff   gf2t  
 
always +ve

Since nothing travels faster than light, particle worldlines are

bounded by
df
-   
dt
What signs can the metric coeffs have? See p. 104:
gff is always +ve, gtf is always -ve
gtt is -ve outside the static limit and +ve inside the static limit
So + is always positive. For large r > rE, - is negative and
df/dt (bounded by + and -) can have either sign. But, for
r < rE, both + and - are positive and df/dt must be positive.
 In the ergoregion, matter must orbit in the same f direction as
the spin of the black hole. (This resolves the problem of time for
static objects: objects cannot be static in the ergoregion, hence
the term "static limit"). However, it is outside the horizon and is
still free to move inward or outward in r (or escape to infinity).
This is an example of frame dragging - in the ergoregion, the
black hole's spin drags inertial frames around it so fast that not
even light can orbit the "wrong" way. (The fact that particles
moving with the spin can stay outside the horizon is due to
gravito-magnetism: the repulsive contribution to gravity between
co-moving masses that we briefly encountered in L1.)
9. Rotating black holes / Kerr orbits
Lecture 18 109
Orbits around Kerr black holes
For simplicity we'll only consider orbits in the equatorial plane
(q = /2, dq = 0) from now on. In Sch. spacetime we did this
without loss of generality but, because Kerr black holes have
axial not spherical symmetry, it's very much a special case here.
Here are the q = /2 versions of the results from the last lecture:
 rs  a 2 rs
g tt  -  1 -  gff  r  a 
2 2

 r r
[metric coeffs]
r2 rs a
g rr  gf t  -
D r
D  r 2  a 2 - rs r
1/2
rH  GM   GM  - a 2 
2
[event horizon, where D = 0]
 
rE  2GM [static limit, where gtt = 0]

It's also possible to prove this Very Useful Identity*:

gf2t - gff gtt  D sin 2 q [in general]
D [equatorial, q = /2]

• Equations of motion
The only coord that the metric coeffs depend on is r, but the
metric is not diagonal because gft  0. This means we can use
simplifications #1 and #3, but not #2, on p. 61-62:
3
dx b
 
b
gab
0 dt
is constant if xa = t or f

* It doesn't have a name, so I'll just call it the VUI.

9. Rotating black holes / Kerr orbits
Lecture 18 110
dt df
xa = t: gtt  gf t  -e
dt dt give the
dt df constants names
xa = f: gft  gff l
dt dt
dt
For large r, e   specific relativistic energy e
dt
df
l  r2  specific angular momentum* l
dt
Solve the simultaneous equations, with the aid of the VUI:
dt gff e  gft l

dt D [t and f equations of motion]

df - gft e - gtt l

dt D

For xa = r use simplification #3 from p. 62:

 df   df  dt 
2 2 2
 dt   dr 
-1  gtt    g rr    gff    2 gft   
 dt   dt   dt   dt  dt 
Subst for dt/dt and df/dt and manipulate using the VUI 
  dr 2 
 g rr    1 D  e gff  l gtt  2elgft
2 2
[r eqn of motion]
  dt  
We now have three equations of motion that can (in principle) be
integrated (numerically?) for given e and l to give the particle's
worldline r(t), f(t) and t(t). But they are somewhat unlovely.
We'll study two cases.

* Reminder: we use J or a for the spin angular momentum of the central mass, and L or l
for the orbital angular momentum of a particle moving around it.
9. Rotating black holes / Kerr orbits
Lecture 18 111
• Case 1: Free fall from rest at large r
From the large-r results on the previous page, and for the same
reasons as the Sch. case on p. 75, l = 0 and e =  = 1. From the
equations of motion:
df - gft rs a
 
dt D rD
dt gff r 2  a 2  a 2 rs / r
 
dt D D
1/2
dr r 
 ... algebra ...  -  s3 (a 2  r 2 ) 
dt r 
The particle has a non-zero angular velocity (in the direction of
the black hole's spin) despite having zero angular momentum* ...
At the horizon (r = rH), dt/dt and df/dt   while dr/dt is finite
 both coords t and f behave badly at the horizon; it's a
coordinate singularity. But
df df / dt rs a
 
dt dt / dt r  r 2  a 2  a 2 rs / r 
finite at the horizon

Shape of orbit:
df df / dt
1/2
a rr    at the horizon (D = 0)
 -  2 s 2
dr dr / dt Dr a 

 the observer at  sees the particle revolve around the black

hole at a finite rate as it approaches the horizon, but it wraps
around an infinite number of times.
f is basically a failed coordinate at the horizon!

* Indeed, a particle's angular velocity and angular momentum can be in opposite

directions!
9. Rotating black holes / Kerr orbits
Lecture 18 112
• Case 2: Negative-energy motion (e < 0)
Meaning: to bring a particle with e < 0 to r   (where e = 1 at
rest) requires more energy than the particle's own rest-mass. It is
energetically "cheaper" to abandon the particle and build a new
one at infinity! For Sch. black holes, we have already seen that
negative e is only possible inside the horizon (p. 86).
Rearrange the r equation of motion (p. 110) as a quadratic in e:

gff e2  2lgft e  l 2 gtt - Z D  0

2
 dr 
where Z  g rr    1
 dt 
is +ve outside the horizon, and solve for the allowed energies e
given l. After some algebra and the VUI:

D1/2 2
l  gff Z 
gft 1/2
e- l
gff gff

The minus sign gives e = -1 at r  , but e  1 at infinity so

this solution is absurd. Consider only the plus sign:

D1/2 2
 ff 
gft 1/2
e- l l  g Z
gff gff

gff > 0 and gft < 0 always, so both circled parts of the equation
must be positive. The only way e can be negative is if

l  gff Z 
1/2
gf t l  D1/2 2

which requires l < 0: the particle's angular momentum is directed

against the black hole's spin.
9. Rotating black holes / Kerr orbits
Lecture 18 113
In that case, e < 0 at the horizon (D = 0). For small -ve l, the
inequality holds only close to the horizon: D1/2 increases with r
and the RHS soon overtakes the LHS. As l becomes more -ve, r
must be bigger before the RHS beats the LHS, and the range
where e < 0 widens. The extreme case is when l is large and -ve,
in which case we can neglect the Z term on the RHS and
substitute for gft :
rs a
- l  D1/2 | l | [modulus to keep RHS +ve]
r
2
 rs a 
    D  r 2
 a 2
- rs r [square]
 r 
 a 2 rs a 2 
 (r - rs )  r   2   0 [factorise]
 r r 

 r  rs  rE [2nd factor has no +ve roots]

Summary: For Kerr black holes, negative energy e is

possible outside the horizon, but only in the ergoregion and
only for negative l.

• The Penrose process (using negative energy)

Send rocket R from a far-away base to the ergoregion, where R
dispatches payload P into an e < 0 trajectory. R returns to base,
while P falls through the horizon. R has lost a -ve energy
payload so it returns with more energy than it started with. The
payload's -ve energy and -ve ang. mom. are added to the black
hole's, reducing its mass and spin. This Penrose process
therefore mines the rotational energy of a "live" Kerr black hole.
9. Rotating black holes / Kerr orbits
Lecture 18 114

At the base, R's extra kinetic energy can be used to do work,

then a new P attached and the process repeated. This is a very
efficient energy generation scheme for a technologically-
advanced civilisation. (It's also good for waste disposal, if P is
filled with junk.) The ultimate limit is when all of the black
hole's rotational energy has been extracted, leaving a "dead"
Schwarzschild black hole.

• Electrically-charged black holes (for completeness)

The other characteristic a black hole can have besides mass and
spin is electric charge Q. However, charged black holes are of
theoretical interest only. Here are the names of their metrics:

Q=0 Q≠0

J=0 Schwarzschild Reissner-Nordström

J≠0 Kerr Kerr-Newman

* The arrows in the figure represent angular momentum, not angular velocity (which in
the ergoregion is, of course, always in the same direction as the black hole's spin).
10. GR and QM / Thermodynamics
Lecture 19 115
10. GR and quantum mechanics
The thermodynamics of black holes
The thermodynamic states of black holes are very simple.
Whereas the structure of an ordinary star encompasses huge
numbers of moving particles, the only characteristics of a black
hole are its mass M, spin J and electric charge Q. All other
information about what formed it, or fell in afterwards, is lost.
Wheeler (again!) expressed this as: “A black hole has no hair.”
If a black hole has no other degrees of freedom in its structure, it
has no statistical-mechanical microstates. So, what happens to
the entropy of matter that falls through the horizon? Do black
holes violate DS  0, the second law of thermodynamics?
• Irreversibility in black holes
Entropy is about irreversible processes, so what's irreversible
about a black hole? Although things can only pass inwards
through the event horizon, a black hole's mass doesn't always
increase: the Penrose process is a counter-example. However,
S Hawking showed that the area A of the event horizon (or the
combined area if several black holes interact) can never decrease
by classical physical processes. J Bekenstein then proposed that
black holes have an entropy proportional to the area: S  A. The
second law (for black holes + everything else) survives if
kB A
S [in general]
4G
4 k B GM 2 [Sch. black hole*, A = 4rs2]


* Because (p.50) the event horizon of a Sch. black hole has the geometry of a sphere of
radius rs and hence a surface area of 4rs2
10. GR and QM / Thermodynamics
Lecture 19 116
• Temperature of black holes
A thermodynamic system whose entropy S depends on its
internal energy U has a temperature T :
dU  TdS  T1 [1st law, đW = 0]
dS / dU
But in relativity mass is a form of internal energy: U = M. Since
S (previous page) depends on M, black holes have a temperature:
T [differentiate dS/dM]
8 k B GM
A black hole is a black body (it absorbs all incident radiation) so
it must emit black-body radiation according to Planck's law for
temp T! Hawking used this to oppose Bekenstein's entropy idea -
obviously nothing comes out of a horizon, so T = 0. But then he
discovered a quantum mechanism for black holes to radiate.
• Hawking radiation
According to quantum field theory, vacuum fluctuations
continually produce virtual particle-antiparticle pairs. One has
+ve energy +E and the other -ve energy -E. Since -ve energy is
forbidden outside the horizon, the particles exist only briefly
before recombining in a time given by the uncertainty principle:
DEDt ~  t ~ /E
But if they are so close to the horizon that the -E particle falls in
within this time, its energy is now allowed* and the particles
become real. The -E particle reduces the black hole's mass, and
the +E particle can escape to infinity as Hawking radiation.
Hawking calculated the temperature of a black hole from this
idea, and got the same answer as derived from Bekenstein's
entropy. Thus he changed his mind about Bekenstein's proposal.

* -ve e is allowed inside the horizon, see p. 86

10. GR and QM / Thermodynamics
Lecture 19 117

• Hawking's derivation
... is beyond us. But remarkably we can derive an approximate T
from the uncertainty principle. A virtual pair fluctuates into
existence just outside the horizon at x = x0, where r  rs + x.
Observe the pair in a reference frame free-falling from rest at
that point. How long does it take the virtual -E particle to reach
the horizon at x = 0 and become real? From (ix) on p. 75:
1/2
dr   r 
 -  e 2 - 1 - s   [vertical drop]
dt   r 
Close to the horizon x << rs so using the binomial approx:
 rs  1  x x
1 -
 r  1 -  1 -  1 - 
  1  x / rs   rs  rs
x0 x0
Start at rest at x = x0: 0  e2 -  e2 
rs rs
dr dx ( x0 - x)1/2
  -
dt dt rs1/2
Integrate from the starting point x = x0 to the horizon x = 0:
t 0 dx
0
dt  - rs1/2 
x0 ( x - x )1/2
0
0
1/2  ( x0 - x ) 
1/2
 time to horizon t  rs    2rs x0
1/2 1/2

 1/ 2 d
10. GR and QM / Thermodynamics
Lecture 19 118
The uncertainty principle DEDt ~ ħ allows the fluctuation to last
this long if
E ~  1/2 1/2
t 2rs x0
The +E particle with this energy undergoes gravitational redshift
as it travels to infinity, where it is observed to have energy E
1/2
 r 
E  1 - s  E [from p. 57, E = ħ]
 r
x01/2
 1/2 E [binomial approx on prev page]
rs
x01/2
 1/2  1/2 1/2  [independent of x0 !]
rs 2rs x0 2rs
Characteristic temp corresponding to this energy E = kBT:

T 
2k B rs 4k B GM
within 2 of Hawking's exact calculation!
• Black hole lifetime
If a black hole radiates, then (in a cold environment) it will lose
mass and eventually evaporate completely. In Q41 you'll use
Hawking's temp & Stefan's law ( radiated intensity) & the
area of the horizon ( radiated power) & U = M ( rate of
mass loss) to derive and solve a differential equation for dM/dt
for a Sch. black hole in a Universe at absolute zero.
3
 M 
 lifetime t0  2.11067   years
M 
ie a very very long time. mass of the Sun
10. GR and QM / Quantum gravity
Lecture 20 119
Quantum gravity

Attention: in this lecture (quantum gravity) we will

revert to ordinary c  1 units, with time in seconds.

GR is a classical theory that ignores quantum uncertainty.

Bekenstein and Hawking needed QM (quantum mechanics) -
their formulae include ħ - but they still used classical GR for the
gravity parts of their derivations. We need to go beyond GR to a
quantum theory of gravity to understand gravitational
phenomena where quantum "fuzziness" on small scales is
important - like the singularities of black holes.
The cube of physics illustrates how physical theories have been
developed to encompass phenomena represented by non-zero
values of G, ħ and c-1 (SR  Newton if c is infinite). Quantum
gravity is the missing corner where all three are included:
10. GR and QM / Quantum gravity
Lecture 20 120
• When do we need quantum gravity?
When QM and GR phenomena are both significant! Consider
the "effective size" of a point mass M under each theory.
QM: To localise a particle to within Dx, the uncertainty principle
says we give it a momentum uncertainty of Dp ~ ħ/Dx. But
according to E2 = p2c2 + M 2c4, a momentum of p = 3Mc ~ Mc
provides enough kinetic energy to create a new particle,
preventing us localising the original one. So (equating p and Dp)
the minimum measurable quantum size of "point" mass M is

DxQM ~ c  [a.k.a. the reduced Compton wavelength]

Mc
GR: Meanwhile the minimum gravitational size of mass M is
2GM
DxGR ~ rs  [the Sch. radius, in c  1 units]
c2
Note that quantum size  1/M whereas gravitational size  M:
10. GR and QM / Quantum gravity
Lecture 20 121
Need quantum gravity (QM and GR) for point masses where:
1/2
 c
c ~ rs  M ~    MP
G
This is called the Planck mass. The effective size (c or rs) of a
Planck-mass particle is of the order of the Planck length. Light
travels the Planck length in the Planck time.

• The Planck scale

The fundamental constants ħ, c and G combine (dimensional
analysis) to give the Planck units of length, time and mass*
1/2
 c
Planck mass M P     2.177 10-8 kg
G
1/2
 G 
Planck length LP   3   1.616 10-35 m
 c 
1/2
 G
Planck time TP   5   5.39110-44 s
c 
These are extreme values: LP and TP are ridiculously small for
any purpose, and MP is both ridiculously big (for a fundamental
particle) and ridiculously small (for a black hole). And we've just
found that this is the scale of quantum gravity! At this scale:
Gravitational and quantum phenomena are both important;
LP is the smallest meaningful length in physics;
A point particle is heavy enough to be sufficiently localised
(despite quantum uncertainty) to disappear inside its own event
horizon and form a "micro black hole";

* Other derived Planck units can be obtained from these three, eg Planck energy MPc2,
Planck area LP2, etc.
10. GR and QM / Quantum gravity
Lecture 20 122
Attempts to measure down to LP require so much energy that the
measurement process creates micro black holes;
A black hole is small enough for quantum effects to become
important - QM affects spacetime itself, not just particles;
Spacetime becomes like foam (Wheeler), or a bucket of dust
(Wheeler), or a bubbling sea of virtual black holes (Hawking), or
a weave of knots and tangles (Smolin), or whatever ...

• So, why explore quantum gravity?

Physics at the Planck scale needs a theory of quantum gravity.
Neither QM nor GR on its own is good enough, and they can't
both be right at this scale. But experimentally it's not a pressing
problem: no foreseeable experiment could probe such small
distances and times, or such heavy point particles. Current
theories of quantum gravity are therefore largely speculation,
untestable hypothesis, or even metaphysics.
Nevertheless, quantum gravity is needed to explain:
- what happens at the singularity of a black hole;
- what happens before time TP after the Big Bang;
- how black-hole evaporation ends, when M ~ MP;
- to reveal "unknown unknowns" in physics;
- to complete physics!
Unfortunately there is as yet no adequate theory of quantum
gravity, nor any prospect of experimental guidance: classical GR
remains our best theory of gravity. But we can look at some
features a theory of quantum gravity may have.
10. GR and QM / Quantum gravity
Lecture 20 123
• Origins of black-hole entropy
c3k B A
Bekenstein: S [from p. 115, in c  1 units]
4G
No-hair theorem: where are the microstates? Subst LP:
A
S  kB
(2 LP ) 2
One unit of entropy kB for every ~Planck area (2LP)2 of the event
horizon - something deep there! This suggests that a black hole
stores information uniformly on its horizon, in a form to be
determined by quantum gravity. This hypothesis is known as the
holographic principle, by analogy with the way a hologram
stores a 3-D image on a 2-D surface.
The principle can be generalised to say that the information in
the whole observable Universe is stored on its boundary: the
cosmological horizon where the Hubble velocity is c.
• The graviton
The force-carrying quantum of gravity, like the photon is for
electromagnetism. It has zero rest-mass like the photon, because
both forces are long range. Tidal displacement has two-fold
rotation symmetry, so the graviton is a tensor boson with spin 2.
(The electric field vector has one-fold rotation symmetry, so the
photon is a vector boson with spin 1.) Gravitons interact very
weakly, so there is no prospect of detecting them experimentally.
• The cosmological constant (dark energy) problem
Simple QM calculations predict a cosmological constant that is
~10120 bigger than observed. This has been called the worst
theoretical prediction in the history of physics! We'd like
quantum gravity to fix it...
10. GR and QM / Quantum gravity
Lecture 20 124
• Current theories of quantum gravity
The most well-known is string theory / superstrings / M-theory
(the same theory in different stages of development):
Particles are excitations of 1-D Planck-scale strings rather than
the traditional 0-D points.
The theory unifies all fundamental forces, not just gravity.
It attempts to eliminate free parameters (like particle masses,
charges, force strengths etc) from physics.
It introduces 6 or 7 extra Planck-sized dimensions to spacetime.
The maths has not been completed and is only approximate so
far. Consequently there are ~10500 possible topologies for the
extra compactified dimensions - a lot more free parameters!
Supersymmetry (hence superstrings) predicts new "partner"
particles, eg photinos, squarks, sleptons etc. The LHC hasn't
found any of them yet - how long do we wait?
It is controversial, highly speculative, and has metaphysical
baggage (extra dimensions, unfalsifiable multiverses, anthropic
reasoning - see "Occam's razor"). But, simplified cases correctly
yield the microstates needed for black-hole entropy, and it
reproduces the holographic principle.
It is not testable for the foreseeable future, so is it really physics?

Other theories of quantum gravity are available. The most well-

publicised is loop quantum gravity, in which spacetime itself is
quantised on the Planck scale.