Susskind Lectures on String Theory

March 15, 2024

version 2.3

These notes are from a series of lectures given by Leonard Susskind (one of the
founders of string theory). They follow the logic he used when he first started thinking
about strings. 1 I wrote these notes for graduate students interested in quantum
gravity. These notes are not verbatim; I watched the lectures and reproduced them
afterwards.
Part I of this document follows the lecture series by Susskind titled “String theory
and M-theory.” This is a basic introduction to bosonic string theory using the infi-
nite momentum frame description. In this “light cone” frame one is able to quantize
the non-relativistic degrees of freedom in the transverse plane using canonical quan-
tization. The massless excitations are examined and found to contain photon-like
and graviton-like particles, as well as tachyon states. We show that to obtain a zero
mass for the photon-like state requires that the critical dimension of spacetime be
D=26. The Veneziano amplitude is introduced in the context of meson scattering,
and is, remarkably, shown to result from the scattering amplitude of two open bosonic
strings. Part II follows the Susskind lectures titled “Topics in String Theory”. Part
II is incomplete.
If you find any errors or typos in this document please email them to me at
[email protected]. I hope you enjoy the notes.
1
Susskind gave this series of lectures for retired engineers and other scientists in his continuing
education course at Stanford. All of Susskind’s continuing education lectures are easily found on
YouTube. Even for advanced students these lectures are well worth watching.

1
CONTENTS CONTENTS

Contents

Contents 2

List of Figures 3

I Bosonic String Theory 5

1 Particle Energies 5
1.1 Non-relativistic particles . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.2 Composite particles . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.3 The infinite momentum frame . . . . . . . . . . . . . . . . . . . . . . 6

2 Infinite Momentum Frame 8

2.1 The continuum limit . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
2.1.1 Canonical quantization of the string . . . . . . . . . . . . . . 8
2.1.2 Helicity for massless particles . . . . . . . . . . . . . . . . . . 11
2.1.3 Low energy excitations of the string . . . . . . . . . . . . . . 12
2.1.4 Open vs. closed strings . . . . . . . . . . . . . . . . . . . . . . 13
2.2 Closed string spectrum . . . . . . . . . . . . . . . . . . . . . . . . . . 13
2.3 Critical dimension . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16

3 Open String Scattering 18

3.1 Mandelstam variables . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
3.2 The Veneziano amplitude . . . . . . . . . . . . . . . . . . . . . . . . 21
3.3 Bosonic open string scattering . . . . . . . . . . . . . . . . . . . . . . 21
3.3.1 Euler Beta function . . . . . . . . . . . . . . . . . . . . . . . 25

4 World sheet symmetries 26

4.1 Conformal mapping . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
4.2 Compactification . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
4.3 Winding number . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33

5 Kaluza-Klein 35
5.1 Kaluza-Klein metric . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
5.2 D-branes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37

2
LIST OF FIGURES LIST OF FIGURES

II Topics in String Theory 38

6 Fundamental Objects 38
6.1 The electron and the monopole . . . . . . . . . . . . . . . . . . . . . 39
6.2 D-branes and S-duality . . . . . . . . . . . . . . . . . . . . . . . . . . 39

7 The Schwarzchild metric 41

7.1 Black holes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
7.1.1 Information and entropy . . . . . . . . . . . . . . . . . . . . . 42
7.1.2 Black hole quantum mechanics . . . . . . . . . . . . . . . . . 43
7.1.3 Hidden degrees of freedom . . . . . . . . . . . . . . . . . . . . 45
7.2 Metric near the horizon . . . . . . . . . . . . . . . . . . . . . . . . . 46
7.2.1 Rindler space . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
7.3 Penrose diagrams . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
7.4 Black hole formation . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
7.5 Counting black hole microstates . . . . . . . . . . . . . . . . . . . . . 49

List of Figures
1 Four-particle interaction with incoming momenta k1 , k2 and outgoing
momenta q3 , q4 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
2 Feynman diagrams for S- (left) and t-channel (right) scattering with
time running from left to right. Making particle 4 incoming and 2
outgoing corresponds to u-channel scattering. (The gluon line here
represents many kinds of particles of different mass.) . . . . . . . . . 20
3 Bosonic open string scattering. Two open strings come together for a
time τ and then separate. Time runs from left to right in this figure. 21
4 Conformal mapping of the two-string interaction in Fig. 3 to the unit
disk. The interior of the disk is the interior of the shaded string region
in Fig. 3 and the injection points are shown as exaggerated cutouts in
the disk. These are mapped to infinitesimal points on the exterior of
the disk. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
5 s- and t-channel scattering maps. The τ parameter determines the
symmetric locations of the four vertex operators. For τ = 0 the points
would move closer and overlap on the left. For τ → ∞ the point on
the right map would move closer and overlap. . . . . . . . . . . . . . 27

3
LIST OF FIGURES LIST OF FIGURES

6 Linear fractional map w(z) = z+1z−1

.. . . . . . . . . . . . . . . . . . . . 29
7 Rindler spacetime diagram. The horizon of a black hole is the 45-
degree line defining the light cone from the origin. . . . . . . . . . . 48
8 Rindler metric, near and far from the horizon. . . . . . . . . . . . . . 48

4
1 PARTICLE ENERGIES

Part I

Bosonic String Theory

1 Energy of Non-relativistic and Relativistic Par-
ticles
1.1 Non-relativistic particles
The energy of a “particle” is given the by the kinetic energy of the particle plus the
rest mass of the particle plus whatever binding energy holds together the constituent
parts. Classically, we write this as

p2
E= + mc2 + B. (1)
2m

Here B represents the internal binding energy of the particle. If the particle remains
slow and is bound tightly, then this energy will never be observed and can be moved
to the LHS of (1) with a redefinition of the energy E ′ = E − mc2 − B giving

p2
E′ = . (2)
2m

For example, the energy of a cup of coffee would be the momentum of the center
of mass of the motion of the coffee plus the rest mass m of all the atoms in the coffee
plus the binding energy B of the water in the coffee. The rest mass m of the coffee
would include the internal motion of the atoms (heat). The binding energy of the
coffee would be the latent heat needed to evaporate all of the coffee and this would
go into B.

1.2 Composite particles

What separates “elementary” particles from composite particles? It is the relative
magnitude of the energy required to increase the internal angular momentum of the
internal constituents of the particle.2 If this number is a large fraction of the rest mass
2
In particle accelerator experiments a plot of mass squared m2 vs angular momentum J of pions,
for example, shows a linear relationship. All hadrons including for example the pions (two-quark
particles) and deltas (three-quark particles with spin 3/2) show the same behavior with the same
sloping trajectories.

5
1.3 The infinite momentum frame 1 PARTICLE ENERGIES

of the particle it is considered to be an elementary particle. If the required energy is a

very small fraction of the rest mass of the particle it is considered a composite particle.
Examples here are given by the electron, an elementary particle, and the cup of coffee,
which is clearly composite. An electron has spin ±h̄/2. Is it possible to “spin up”
an electron to a higher angular momentum state similar to the way hadrons can be
spun up, resulting in the Regge trajectories seen in particle accelerators? Classically
the energy of a rotating object is given by the kinetic energy E = Iω 2 /2 and the
angular momentum is given by L = Iω. Quantum mechanics dictates that angular
momentum is quantized in units of h̄. For a large, classical object like a basket ball
it is easy to increase the angular momentum by one unit of h̄ by slightly increasing
the spin rate ω. Clearly, this is a small fraction of the total energy of the basketball.
On the other hand an electron’s intrinsic angular momentum is h̄/2, and increasing
it requires at least one unit of h̄. In contrast to a macroscopic object, spinning up
the electron from h̄/2 to 3h̄/2 triples its angular momentum. The change in energy
would be given by E3/2 − E1/2 = 8Iω02 /2 = 8E1/2 or eight times the rest energy of
the particle, if an electron could be considered as a tiny spinning object (it cannot!).
Another way to look at this large change in energy is through the equation E = L/2I,
the energy is inversely proportional to the rotational inertia. The size of the electron
is known to be very small, on the order of 10−20 m, so that the energy required to
increase the angular momentum by one unit must be very large (on the order of the
Planck energy).

1.3 Relativistic particles and the infinite momentum frame

We have argued that composite particles can be “spun up” by increasing the angular
momentum of their constituent parts. We would like to consider now the motion
of a composite particle after it is given a large Lorentz boost. Starting from the
Klein-Gordon relation and setting c = 1 we have

E 2 = p2 + m2 (3)

we can approximate the energy in two ways. First we can assume that the particle
momentum is low and so the rest mass dominates the energy. Then we write

√ ( )1/2
p2 p2
2 2
E = p +m =m 1+ 2 ≈m+ (4)
m 2m

6
1 PARTICLE ENERGIES 1.3 The infinite momentum frame

giving the usual relationship upon restoring c: E = p2 /2m + mc2 . In the second
case we perform a Lorentz boost along the z-direction. Here the momentum along z
dominates the energy and we expand as
√
E = p2z + p2x + p2y + m2
( )1/2
p2x + p2y + m2
= pz 1 +
p2z
p2x + p2y m2
≈ pz + + . (5)
2pz 2pz

We can assume that as pz → ∞ the only dynamics that are left are those in the
perpendicular plane. In this case pz on the RHS is simply a constant and can be moved
to the LHS as an additive constant to the total energy. The Lorentz contraction along
the boost direction flattens the dynamics to the perpendicular plane with the total
momentum along the boost direction acting as an inertial mass

p2x + p2y p2x + p2y

←→
2pz 2M

with pz playing the role of M . A Lorentz boost of the system results in time dilation
and apparently slow dynamics from the point of view of the stationary observer due
to the large momentum. A rescaling of the time variable can effectively scale out
the pz or one can simply multiply (5) through by pz . Redefining the total energy as
E ′ = (E − pz )pz we have
′
p2x + p2y m2
E = + . (6)
2 2
Now the dynamics is given in terms of momentum in the plane perpendicular to
the boost and the internal additive energy of the particle is given by B = m2 /2.
At relativistic energies m2 is something that is measured, giving rise to the Regge
trajectories described earlier. The two important points to remember from this section
are (1) the internal energy goes as m2 in the infinite momentum frame, and (2) The
motion can be described classically in the plane perpendicular to the boost direction.

7
 2 INFINITE MOMENTUM FRAME

2 Relativistic String in the Infinite Momentum

Frame
2.1 Coupled spring system in the continuum limit
In this section we model the string as a collection of point masses connected by
springs. Consider an open spring system of finite length consisting of N point masses
connected by springs, with the equilibrium spacing a and spring constant κ. If the
system is boosted along the 3-direction the total energy in the 1,2-plane is given by

∑
N
mẋ2 ∑
N −1
κ(xn+1 − xn )2
n
E= +
n=1
2 n=1
2

where x ≡ (x, y) is the position of mass n. In the limit as N → ∞, a → 0, and the

spring constant κ → ∞ such that the product is finite aκ → constant, we can write
the energy as an integral over the length of the string. In particular

∑ ∫ ( )2
(xn+1 − xn )2 ∂x
κa(a∆n) → dσ (κa)
n
a2 ∂σ

where dσ = lima→0 a∆n, and ∆n = 1, and in the first term we write

∑ mẋ2 ∫
ẋ2
a∆n → dσ (m/a).
n
2a 2
√
This clearly requires that m → 0 such that m/a → constant. Redefining x = κax
will give the total energy
∫ [( )2 ( )2 ]
π
1 ∂x ∂x
E= dσ +
0 2 ∂t ∂σ

where the length was defined such that σ goes from 0 to π. This is convenient in
the Fourier expansion below, and is defined this way for the open string so that the
integral for closed strings will go in a loop from 0 to 2π.

2.1.1 Canonical quantization of the string

∫π [( ) ( ∂x )2 ]
2
Now we can quantize canonically using the Lagrangian L = 0 dσ 21 ∂x
∂t
− ∂σ
.
Susskind writes this with some constants in front that can be absorbed into the

8
2 INFINITE MOMENTUM FRAME 2.1 The continuum limit

definition of the field and also uses τ , the proper time.

∫ [( )2 ( )2 ]
π
1 ∂x ∂x
L= dσ − (7)
2π 0 ∂τ ∂σ

Note that this is the Lagrangian for a wave field and the Euler-Lagrange equation
is the wave equation. In order to Fourier analyze the string one must make a deci-
sion on the boundary conditions. At the end of the string the last mass point has a
force proportional to the stretch in the spring or mẍ ∝ ∂x/∂σ but the mass points
were tending to zero so the acceleration would be increasing without bound. There-
fore the boundary condition chosen is ∂x/∂σ = 0 or Neumann boundary conditions.
Expanding in a Fourier series we have

∑
∞
x(σ, τ ) = xn (τ ) cos(nσ) (8)
n=0
∑
∞
ẋ(σ, τ ) = ẋn (τ ) cos(nσ) (9)
n=0
∑∞
x,σ (σ, τ ) = − nxn (τ ) sin(nσ) (10)
n=0

Plugging these equations into the Lagrangian (7) and using the orthogonality of the
sines and cosines3 we will find an infinite collection of harmonic oscillators of frequency
ωn = n. Explicitly
∫ π [ ]
1 ∑ ∑
L = dσ ẋn cos(nσ)ẋn′ cos(n′ σ) − nxn sin(nσ)n′ xn′ sin(n′ σ)
2π 0 n,n′ n,n′
[ ]
1 2 1 ∑( 2
∞
)
= ẋ0 + ẋn − n2 x2n (11)
2 2 n=1

This is a set of harmonic oscillators of increasing energy. The oscillator labeled n = 1

has one unit of energy, n = 2 corresponds to two units. The zero point energy of a
harmonic oscillator is h̄ω/2 so for oscillator n the ZPE is n/2. The total energy of
the n-th oscillator in the m-th excited state is

1
Em = n(m + )
2
3
∫π ∫π
0
dx sin(nx) sin(mx) = δnm π2 and 0
dx cos(nx) cos(mx) = π
2 δnm + π2 δn0 δm0

9
2.1 The continuum limit 2 INFINITE MOMENTUM FRAME

To obtain the Hamiltonian from this Lagrangian one calculates the conjugate
momentum
pn = ∂L/∂ ẋn = ẋn /2

and rewrites the Lagrangian as

∞ ( )
1 2 ∑ 2 n2 x2n
L = ẋ0 + pn − . (12)
2 n=1
4

The Hamiltonian is
∑
H = pn ẋn − L
n=0
∑( n2 x2n
)
= p2n + (13)
n=1
4

where we have dropped the CM motion term. Let us write this in terms of the usual
creation and annihilation operators. Focusing on one of the terms we follow the usual
procedure for harmonic oscillator quantization
( ) ( nx ) ( nx )
n2 x2n n n
+ ẋ2 = + iẋ − iẋ
4 2 2

and identify each term with either the creation or annihilation operator. Using the
canonical quantization condition [x, p] = i and knowing we need [a− , a+ ] = 1 we
calculate the commutator
[( nx ) ( nx )] ni
+ ip , − ip = ([p, x] − [x, p])
2 2 2
= n (14)

so we define √
n i
a±
n = x̂n ∓ √ p̂n . (15)
2 n
These are the creation and annihilation operators. By adding and subtracting them
we can represent the position and momentum operators in terms of the creation and
annihilation operators

1 ( )
x̂n = √ a+ + a− (16)
n
√
n( + )
p̂n = i a − a− (17)
2

10
2 INFINITE MOMENTUM FRAME 2.1 The continuum limit

If we define the operators a± to correspond to excitations of the oscillators in x,

we can define b± operators for the excitations of the oscillators in y. Then we can
begin to discuss the excitations from the ground state of the string, denoted by
|0⟩. Note that |0⟩ is not the vacuum state, but the ground state of an existing
string. The first excited state is doubly degenerate and consists of the states a+ 1 |0⟩
and b+1 |0⟩. The second excited state is 5-fold degenerate and consists of the states
{a1 a1 |0⟩ , a+
+ +
1 b1 |0⟩ , b1 b1 |0⟩ , a2 |0⟩ , b2 |0⟩}. Remember here that an excites the n-th
+ + + + + +

mode of the oscillator in the x direction, so a+ 2 excites a mode with frequency ω = 2,

( + )2
and a1 excites the mode-1 oscillator twice, giving energy 1 + 1 = 2.
String theory is not consistent in 3+1 dimensions and requires higher dimensions.
These can be incorporated easily in our discussion by simply allowing the oscillator
to move in D+1 dimensions and adding the appropriate number of terms to the
Hamiltonian and defining their corresponding creation and annihilation operators.
Here we are identifying the excitations of the string in the transverse plane that
has only two dimensions. In real string theory there are more dimensions, but the
remaining dimensions are compactified.
The excitations of the string will be described by the creation and annihilation
operators acting on the ground state of the string |0⟩. Keep in mind that the energy
of the internal motion of the string is to be identified with the mass squared. If the
string is in its ground state we would write EGS → m20 .

2.1.2 Helicity for massless particles

Massless particles that have spin J can only have two spin states instead of the usual
2J + 1 states for a massive particle. This is because massless particles move at the
speed of light and cannot be brought to a rest frame in which a rotation operator can
act on the spin state to orient it in a direction perpendicular to its motion. The two
remaining spin states are ±J and are called helicity.
A photon is a spin-1 particle described by the vector potential Aµ . It has two spin
states of its angular momentum, |R⟩ and |L⟩, for left and right circularly polarized. In
classical electrodynamics we describe the electric field vector of Maxwell’s equations
as having two possible orientations along the direction of motion of the electromag-
netic wave. These are the linear polarizations along the x- or y-direction, which we
would label quantum mechanically as |x⟩ and |y⟩. To go from one description (lin-
ear polarization) to another (circular polarization) we use a linear combination with

11
2.1 The continuum limit 2 INFINITE MOMENTUM FRAME

complex coefficients.

|x⟩ = |R⟩ + i |L⟩ (18)

|y⟩ = |R⟩ − i |L⟩ (19)

Keep in mind that a linear combination of |x⟩ and |y⟩ with real coefficients would
simply be a linearly polarized light wave along some angle between the x- or y-
direction.

2.1.3 Low energy excitations of the string

The energy of the quiescent string in the light cone frame is EGS → m20 . It will
include zero-point oscillations, but those can be absorbed in m20 . The lowest energy
excitations of the string will involve using the creation operators to give one unit of
oscillation to the string,
1 |0⟩
a+ (20)

for example. It would increase the energy of the string by one in some units. We
would write
E = m20 + 1 (21)

The same would be true for exciting oscillations along the y-axis. We could combine
the creation operators to create a linearly polarized oscillation along an arbitrary
√
direction. For example, along a 45 degree angle would be (a+ +
1 + b1 )/ 2. Both a
+

and b+ transform as components of a vector. It is the transformation properties of

the operators that is important. Note that we could also create oscillations that
correspond to the photon with circular polarization by using
√
1 ± ib1 )/ 2
(a+ +
(22)

Importantly, we notice that there are only two states of this energy that are linear
combinations of a+ +
1 and b1 . In particular there is not a third state, and because
these states transform as vectors under rotation they represent polarization vectors
just as in the case of photons. One can have either linear polarization with linear
√ √
combinations (a+1 ± b1 )/ 2 or circular polarizations with (a1 ± ib1 )/ 2. So there are
+ + +

photon-like objects in the open string. But there is a problem, photons are massless.
They must be massless for Lorentz invariance to hold. This is a problem because
the energy of the string is to be associated with its mass squared, E ∼ m2 = 0 and

12
2 INFINITE MOMENTUM FRAME 2.2 Closed string spectrum

implies that we have

m20 = −1 (23)

a negative ground-state mass squared. This is called a tachyon and was originally
thought of as particles that had to move faster than the speed of light since E =
√
p2 + m2 implies the speed of the wave should be
√
∂ω/∂k = ∂E/∂p = p/ p2 + m2 > 1

This is unreasonable of course, and the correct interpretation is that the negative mass
squared means there is an unstable vacuum from the point of view of field theory. In
a field theory Lagrangian, L = (∂ϕ)2 − m2 ϕ2 with the potential V (ϕ) = +m2 ϕ2 , the
negative mass term corresponds to a potential that has no ground state.

2.1.4 Open vs. closed strings

If string ends are allowed to interact, then two open strings can approach each other
and join ends, making a longer string. Think of the quark model in hadronic physics
where quark-antiquark pairs are joined by stings and may interact in this way by
annihilating a quark-antiquark pair and making a longer string. If this interaction
is possible, then the ends of a single string may interact and form a closed string.
So open string interactions imply the existence of closed strings. This does not work
the other way around; closed strings do not have allow opening and so closed string
theory does not have to contain open strings. So string theory predicts the existence
of closed strings. We will find that open strings behave like photons, and closed
strings behave like gravitons.
If an open string can interact with a closed string it can open the closed string
to form an open string; it can absorb a graviton. Closed strings can do the same.
It appears that every kind of string-like object can absorb closed strings, and this is
why gravitons interact with everything.

2.2 Closed string spectrum

Susskind starts the lecture (number 4) with Noether’s theorem and states the fact
that the charge of a conserved quantity is the generator of the transformation that
gives the conserved quantity. This is standard basic QFT so I have not written it up
here.

13
2.2 Closed string spectrum 2 INFINITE MOMENTUM FRAME

The closed string spectrum is found by a similar analysis of the Lagrangian (7) but
instead of the Neumann boundary condition one uses periodic boundary conditions. If
the string is closed then the waves can be considered to run either left or right around
the string, i.e., in the positive or negative σ direction. Before we Fourier expand the
coordinates let us rewrite the Lagrangian to reflect the left- and right-moving waves.
∫ {[( ) ( )]2 [( ) ( )]2 }
2π
1 1 ∂x ∂x ∂x ∂x
L= dσ + + − (24)
2π 0 2 ∂τ ∂σ ∂τ ∂σ

In this equation the cross terms cancel and, except for the change in limits, we have
(7). In this In this case we let the length of the string go from 0 to 2π for convenience
and require that x(σ + 2π) = x(σ). The solutions with this boundary condition
include both sines and cosines, and we write
∑[ ]
x(σ, τ ) = x0 + xn (τ )einσ + x−n (τ )e−inσ
n>0

where right-moving waves are given by the exponential einσ and left-moving waves
are given by e−inσ . Following the same canonical quantization procedure above we
find a set of creation and annihilation operators

a+
n a+
−n b+
n b+
−n

where we now have operators that create waves moving to the left or right, indicated
by the subscripts. For the closed string we require that the right- and left-moving
waves have the same energy. This requirement can be stated in the form of the
integral
∫ {[( ) ( )]2 [( ) ( )]2 }
2π
1 ∂x ∂x ∂x ∂x
∆E = dσ + − − =0
0 2 ∂τ ∂σ ∂τ ∂σ

which clearly implies that

∫ 2π ( )( )
∂x ∂x
dσ = 0. (25)
0 ∂τ ∂σ

This result also follows from reparameterization invariance of the position along the
string. Mathematically the quantum mechanical wave function must be the same for
a reparameterization of σ. Here we consider an infinitesimal change σ → σ + ϵ and

14
2 INFINITE MOMENTUM FRAME 2.2 Closed string spectrum

write
ψ [x(σ + ϵ)] = ψ [x(σ)] . (26)

We should write this as a functional since σ varies over a continuous range and
any construction, of say the partition function, will be a functional integral over all
possible x(σ). In this case we expand (26), integrate over σ and find
∫ 2π
∂ψ ∂x
dσ ϵ = 0.
0 ∂x ∂σ
(∫ )
2π
This is an integral operator: dσ ∂x(σ)
0
∂
∂σ ∂x(σ)
|ψ⟩. But ∂
∂x(σ)
= −ip̂(σ) and the
momentum p(σ) = ẋ(σ) is proportional to ∂x
∂τ
so we again arrive at (25).
The lowest energy excitations of the string must have the same energy in left-
moving excitations as right-moving excitations. The lowest energy energy excitations
are then given by
1 a−1 |0⟩
a+ +

with a total of 2 units of energy, one left and one right. Similarly for
{b1 b−1 , a−1 b1 , a1 b−1 }. Now write them in terms of circular polarization states:
+ + + + + +

1 + ib1 )(a−1 + ib−1 ) |0⟩

(a+ + + +
(27)
1 − ib1 )(a−1 − ib−1 ) |0⟩
(a+ + + +
(28)
1 + ib1 )(a−1 − ib−1 ) |0⟩
(a+ + + +
(29)
1 − ib1 )(a−1 + ib−1 ) |0⟩
(a+ + + +
(30)

Equations (27) and (28) correspond to 2 units of angular momentum, and equations
(29) and (30) correspond to zero units of angular momentum. For a zero mass particle
of spin-2 we expect the two states m = ±2. These are the gravitons. We do not
expect zero angular momentum states. Equations (29) and (30) are particles called
the dilaton and axion, respectively. Neither of these particles has ever been observed.
These particles can be removed from the theory while keeping the gravitons.
In response to a question, Susskind discusses in some detail the question of what
is or is not a fundamental particle. Consider the example of the electron vs. the
magnetic monopole. From Dirac’s argument that the charges satisfy eg = 2π, the
magnetic charge must be large if e ≪ 1, and Feynman diagrams in QED converge
because the probability for every vertex contains a factor of e2 . In contrast, if g ≪ 1
then the Feynman diagrams would all diverge, and so one might expect that the elec-
tron is fundamental while the monopole is not. The argument can be turned around

15
2.3 Critical dimension 2 INFINITE MOMENTUM FRAME

if one considers that Maxwell’s equations have some interesting symmetry (duality)
and that if the ratio of e/g were varied and g ≪ 1 then one would consider the
monopole to be fundamental and the electron composite. Are strings fundamental?
Well, strings have a duality with D-branes, and they morph into each other as a
parameter of the theory is varied; from some viewpoints the strings are fundamental
and in others the D-branes are fundamental.

2.3 Critical dimension of the open bosonic string

The negative mass squared we found in (23) is problematic and needs an explanation.
We have previously ignored the zero point energy of the string, and this is where it
comes in. If we consider the photon-like excitation of the string that must be massless
we are left with explaining how the zero point motion could give rise to m20 = −1.
The zero point energy for a single oscillator is h̄ω/2, and if our string can vibrate in
D − 2 dimensions we should have a ZPE of (D − 2)h̄ω/2. The D − 2 comes from D
spacetime dimensions, minus one dimension for time, and minus one dimension for
the coordinate which contained the boost to very high momentum. The number of
dimensions in the transverse plane is D − 2. Using h̄ = 1 and ω = n we have

(D − 2) ∑
∞
n = −1. (31)
2 n=1

This looks bad because the sum is formally divergent. However, infinite sums can
be assigned numerical values. This is not trickery, there is physics involved. A
similar situation arises in a mathematical analysis of the Casimir force between two
conducting plates. In that case one needs to “regularize” the sum because arbitrarily
high frequencies of electromagnetic waves will not be contained by the plates and
the regularized sum is not divergent. It is, however confusing it may be, negative.
There are more and less sophisticated ways of arriving at the sum and we will do it
informally, using a mathematically inconsistent technique, but it will give the correct
answer. The sum is
S = 1 + 2 + 3 + 4 + 5 + 6 + ···

16
2 INFINITE MOMENTUM FRAME 2.3 Critical dimension

To make an alternating and therefore more convergent, but not convergent series, we
write

S = 1 + 2 + 3 + 4 + 5 + 6 + ···
T = 0 − 4 + 0 − 8 + 0 − 12 + · · · = −4S
S − 4S = 1 − 2 + 3 − 4 + 5 − 6 + · · ·

The last series is the Taylor expansion of 1/(1 + x)2 about x = 0 and setting x = 1.
We then have
1
S − 4S =
4
or S = −1/12 giving
∑∞
1
n=− . (32)
n=1
12

Adding infinite series as we have done with the additional zeros in T is inconsistent
and not mathematically rigorous.4 A rigorous derivation of this result requires use of
the Riemann zeta function. Plugging (32) into (31) we find

(D − 2)
=1
2 · 12

or D = 26. This means that if the string lives in 26 dimensions we can satisfy (23)
in a consistent manner. This is called the critical dimension of the string theory. In
supersymmetric string theories the number of critical dimensions is reduced. In some
sense it means that there are not arbitrarily high zero point oscillations.
Susskind uses a better argument to obtain this result and explains that the di-
vergent sum should be thought of as having a finite piece that can be combined with
the infinite P piece of the 2P (E − P ) that was obtained from the infinite momentum
frame. Then he calculates the sum using a convergence term.
∑ ∑
n = lim ne−nϵ
ϵ→0
n=1 n=1

Let the limit be understood, then we can write

∑ ∂ ∑ −nϵ
n=− e
n=1
∂ϵ n=1
4
Although it is interesting to note that Ramanujan had this derivation in one of his notebooks!

17
 3 OPEN STRING SCATTERING

where we note that the sum starts from n = 1. Performing the sum gives e−ϵ /(1−e−ϵ )
and expanding the exponentials to third order gives
[ ] [ ]
∂ (1 − ϵ + ϵ2 /2 − ϵ3 /6) ∂ 1 (1 − ϵ + ϵ2 /2 − ϵ3 /6)
− =− .
∂ϵ (ϵ − ϵ2 /2 + ϵ3 /6 − ϵ4 /24) ∂ϵ ϵ (1 − ϵ/2 + ϵ2 /6 − ϵ3 /24)

Keeping terms to order ϵ2 , and using 1/(1 − x) = 1 + x + x2 + · · · , gives

[ ]
∂ 1( )( )
− 1 − ϵ + ϵ /2 1 + ϵ/2 − ϵ /6 + ϵ /4
2 2 2
∂ϵ ϵ

where the ϵ2 /4 comes from (ϵ/2 − ϵ2 /6)2 , the x2 in the expansion. Then we have
[ ]
∂ 1( ) 1 1
− 1 − ϵ/2 + ϵ /12 −→ lim 2 −
2
∂ϵ ϵ ϵ→0 ϵ 12

which contains the formally infinite piece (infinity!) plus the finite piece −1/12.
The fermionic string reduces the number of required dimensions to D=10, and
removes the tachyon states.

3 Open String Scattering

3.1 Mandelstam variables
In particle scattering analysis it is useful to define the Mandelstam variables. Consider
a 4-particle scattering event with four identical particles with incoming and outgoing
four-momenta {k1 , k2 , q3 , q4 }. Momentum and energy conservation dictate that we
write k1 + k2 = q3 + q4 . If we relabel q3 = −k3 and q4 = −k4 we have

k1 + k2 + k3 + k4 = 0.

In this case, the figure would show all lines going in to the blob. The Klein-Gordon
relation can be written p2 − E 2 = kµ k µ = −m2 for each of the particle and provide
constraints. There are 16 independent variables in the four-momenta of each of the
particles. The constraint equations reduce the number of independent variables to
just two. In the center of mass frame there are ultimately only two variables that
describe the amplitude of scattering, the total energy ECM , and the angle through
which the scattered particles depart, θ. There is only one relevant angle if all four
particles are identical. To form invariants from the four-momenta there is only one

18
3 OPEN STRING SCATTERING 3.1 Mandelstam variables

k1
k2 q4

q3

Figure 1: Four-particle interaction with incoming momenta k1 , k2 and outgoing mo-

menta q3 , q4 .

thing to do, form scalars by squaring them. Consider that k1 and k2 are the incoming
particles. It is always possible to boost to the CM frame so that the particles are
coming at each other directly. Then ⃗k1 = −⃗k2 and

(k1 + k2 )2 = (⃗k1 + ⃗k2 )2 − (k01 + k02 )2

= 0 − (2ECM )2

This is defined as S = (2ECM )2 . Taking particles 3 and 4 would be the same as taking
1 and 2, gaining nothing new. Now form the invariant with particles 1 and 3. In this
case k01 + k03 = 0 because we have defined 3 as an outgoing particle. The momentum
transfer is then given by ⃗k1 − ⃗q3 = ⃗k1 + ⃗k3 . Working out this quantity in terms of the
angle of scattering gives
( )2 ( )
⃗k1 + ⃗k3 = 2 E 2 − m2 (1 − cos θ) .
CM

The combination of particles 2 and 4 gives the same relation as 1 and 3. The only
combination left is particles 1 and 4 giving
( )2 ( )
⃗k1 + ⃗k4 = 2 E 2 − m2 (1 + cos θ) .
CM

19
3.1 Mandelstam variables 3 OPEN STRING SCATTERING

k2 k2 k4
k4

k1 k3

k1 k3
Figure 2: Feynman diagrams for S- (left) and t-channel (right) scattering with time
running from left to right. Making particle 4 incoming and 2 outgoing corresponds
to u-channel scattering. (The gluon line here represents many kinds of particles of
different mass.)

Forming scalars from various combinations of the incoming and outgoing four-
momenta, one defines the Mandelstam variables

−s = (k1 + k2 )2
−t = (k1 + k3 )2 (33)
−u = (k1 + k4 )2

The final constraint is s + t + u = −4m2 , and the two independent quantities are
usually taken to be s, and t. The resulting amplitudes for the processes in Fig. 2 are
called s- and t- processes, and are given by
( )
2 1
As = g (34)
s2 − M 2
( )
2 1
At = g . (35)
t2 − M 2

Note that in s-channel scattering the incoming particles form a new bound state
that depends only on the CM energy. s-channel scattering is independent of outgoing
angle, they are all equally probable. This makes sense because they from a compound
state and then decay, and the compound state does not remember where it came from.
A very different physical process is the t-channel scattering where particles 1 and 2
exchange a particle and go out as particles 3 and 4. The t-process depends on the
angle of scattering, while s-processes do not. Note the intriguing symmetry between
s- and t-channel scattering.

20
3 OPEN STRING SCATTERING 3.2 The Veneziano amplitude

Figure 3: Bosonic open string scattering. Two open strings come together for a time
τ and then separate. Time runs from left to right in this figure.

3.2 The Veneziano amplitude

Historically in the 1960’s when meson scattering amplitudes were first being measured,
the two amplitudes for s- and t-channel scattering were thought to add in a calculation
of the amplitude. But the scattering amplitude from this calculation is twice what
is seen in experiments. Physicists at the time were trying various functions for the
amplitude that might result from an exchange of a large number of particles in the
channels. The result was that one was discovered that matched the data perfectly,
and has the same symmetry as the diagrams. It is called the Veneziano amplitude
and is given by
Γ(−s)Γ(−t)
AVeneziano = g 2 . (36)
Γ(−s − t)
How one arrives at the Veneziano amplitude from first principles is a triumph of string
theory. The calculation goes as follows.

3.3 Bosonic open string scattering

We now consider two strings as shown in Fig. 3. Two strings moving freely through
space come together and joint for a time τ . We need to calculate the scattering
amplitude for this process. We take the top string to have incoming momentum k1
and the bottom string incoming momentum to be k2 . We write the wave function of
the top string in the ground state of all the harmonic oscillators that make up the
string as
1 ∑N
ψt (k1 ; x1 , · · · , xN ) = eik1 N n=1 xn ψ0 (x1 , x2 , · · · )

21
3.3 Bosonic open string scattering 3 OPEN STRING SCATTERING

where xl is the position of the l-th mass point on the string. Do not confuse this with
the Fourier mode decomposition above where we quantized. The bottom string as

1 ∑2N
ψb (k2 ; xN +1 , · · · , x2N ) = eik2 N n=N +1 xn
ψ0 (xN +1 , xN +2 , · · · ).

The state ψ0 is the wave function of a collection of harmonic oscillators in their

ground states and is just a product of exponentials.Note that the CM is given by
∑
xCM = N1 N n=1 xn ; we describe the momentum state of the string in terms of the
motion of the center of mass. The crucial point in this analysis is the following. For
the time period in which the strings are attached, we set xN = xN +1 , providing a
constraint, and describe the wave function of the combined strings as

Ψ(k1 , k2 ; x1 , · · · , xN , xN , xN +1 , · · · , x2N ) =
1 ∑N −1 ∑2N
eik1 N (xN +1 + xn ) ik2 N
1
n=1 e n=N +1 xn
ψ0 (x1 , x2 , · · · , xN )ψ0 (xN , xN +2 , · · · ). (37)

At this point we must evolve the system in time using the Hamiltonian, Ψ(τ ) =
e−iHτ Ψ(0), which we have already expressed above. The amplitude for the event is
given by projecting the time-evolved state Ψ(τ ) onto the disconnected string states
with outgoing momenta k3 and k4 . According to the Feynman rules we must calculate
the amplitude5 by summing the projection for all possible times τ , giving
∫ ∞
A= dτ ψ(k3 , x1 , · · · , xN ) ψ(k4 , xN +1 , · · · , x2N ) e−iHτ Ψ(0). (38)
0

One can approach the calculation of eiHτ Ψ(0) by considering a few simple cases.
First, a one-dimensional quantum harmonic oscillator has ground state wave function
ϕ0 (x) = αe−mωx /2 with α = (mω/π)1/4 . This is a stationary state so Hϕ0 = ω2 ϕ0 ,
2

and e−iHτ ϕ0 = e−iωτ /2 ϕ0 . In the case of the string we have a collection of mass points
connected by springs and the Hamiltonian is given by

1 ∑ 2 1 ∑
N N
H= ẋl + mω 2 (xl+1 − xl )2 .
2m l 2 l

5
A path integral calculation of the scattering amplitude would involve an integral over all of the
trajectories X µ (τ ) for each of the mass points. Then the amplitude would be something like
∫ ∫
A = [DX µ ] e− dτ L

where exponential is the Wick-rotated action for the interacting strings.

22
3 OPEN STRING SCATTERING 3.3 Bosonic open string scattering

Focus on the mass point located at xl . This piece of the Hamiltonian will be

1 1 [ ]
hl = mẋ2l + mω 2 (xl − xl−1 )2 + (xl+1 − xl )2
2 2
1 2 1 ( )
=− ∂l + mω 2 ∆2l−1,l + ∆2l,l+1
2m 2

where we have defined ∆l,l−1 = xl − xl−1 . The wave function for the string is

Ψ = eik(···+xl−1 +xl +xl+1 +··· )/N e−

mω
2 [···+(xl+1 −xl )2 +(xl −xl−1 )2 +··· ] (39)

We need the action of the Hamiltonian on this piece of string moving with a momen-
tum k, i.e.
e−ihl τ Ψ(0)

so we need
hl eik(···+xl +··· )/N e−
mω
2 (···+(xl −xl−1 )2 +(xl+1 −xl )2 +··· )

to give us an eigenvalue so that the full exponential e−ih1 τ ψ = e−i(eval)τ ψ. Calculating,

we have

e(all that) ≡ eik(···+xl +··· )/N e− (···+∆2l−1,l +∆2l,l+1 +··· )

mω
2

∂l e(all that) = e(all that) [+mω∆l−1,l − mω∆l,l+1 + ik/N ]

∂l2 e(all that) = e(all that) ×

{ }
[+mω (∆l−1,l − ∆l,l+1 ) + ik/N ]2 − 2mω

so that
( )
(all that) 1 2 1 [ 2 ] (all that)
hl e = − 2
∂ + mω ∆l−1,l + ∆l,l+1 e2
=
2m l 2
k k2
+mω 2 ∆l−1,l ∆l,l+1 + i ω (∆l−1,l − ∆l,l+1 ) + +ω
N 2N 2 m

In the limit N → ∞ the spacing between points goes to zero, and ∆l,l±1 → 0, leaving
us finally with the result
k2
hl e(all that) → +ω
2N 2 m
so that we may write ( )
k2
−i
e−ih1 τ ψ → e
+ω τ
2N 2 m ψ.

23
3.3 Bosonic open string scattering 3 OPEN STRING SCATTERING

We have N → ∞ oscillators, giving

( )
k2 k2
−i
ψ = e−i 2N m τ e−iN ωτ ψ.
+ω Nτ
e 2N 2 m

In the limit, the mass points go to zero such that N m → 1 (a constant we define to
be unity), finally giving
k2
e−i 2 τ e−iN ωτ ψ

The wave function Ψ = ψt ψb will have evolution

2
Ψ(τ ) = e−iHτ Ψ(0) = e−iτ [ 2 (k1 +k2 )
1
+N ω ]
Ψ(0).

Recall that s = − (k1 + k2 )2 . The amplitude we seek (38) will have another term
with (k3 + k4 )2 and also has xN +1 = xN in the initial state before time evolution.
This particular term will give (k1 + k2 − k3 )2 . We still need to fill in the details here
(Wick rotation?). If we Wick rotate, then the infinite term N ω gives e−N ωτ → 0 and
we are left with terms like e−τ s . The calculation of the full amplitude (38) results in
∫ ∞
A= dτ eτ (s+1) (1 − eτ )−t−1 e−τ
0

which upon setting z = e−τ gives

∫ 0
A = − dz z −(s+1) (1 − z)−t−1
1
Γ(−s)Γ(−t)
= .
Γ(−s − t)

This is the Veneziano amplitude (36) for the scattering of two strings each in their
ground state which we recall is a tachyon state. This function has a name in math-
ematics, the Euler beta function β(a, b) ≡ Γ(a)Γ(b)/Γ(a + b). What we have just
demonstrated is remarkable; the scattering of two strings produces the amplitude of
s- or t-channel processes described by Veneziano for the scattering of mesons, and
does so analytically. This strongly suggests that there is validity in the idea that
strings are fundamental objects in nature.
Further confidence was gained by calculating the scattering amplitudes of the
photon-like and graviton-like (excited state) strings with themselves and with other
types of string states. In QFT the scattering amplitudes of photon-photon and
graviton-graviton scattering are distinctive and string theory reproduces those ampli-

24
3 OPEN STRING SCATTERING 3.3 Bosonic open string scattering

tudes.

3.3.1 Euler Beta function

The beta function is defined as ∫ 1
β(n, m) = dl ln−1 (1 − l)m−1 (40)
0
∫∞
We can construct a solution to this integral using the gamma function Γ(t) = o dt e−t tn−1 . First we write the
product of two gamma functions.
∫ ∞ ∫ ∞
Γ(n)Γ(m) = dt e−t tn−1 dq e−q q m−1
0 0

We want to manipulate this integral to contain the beta integral in (40). We need to cover the right upper quadrant
that is represented by the integral over t and q. We will perform a coordinate transformation (t, q) → (z, l). To obtain
the correct measure we use differential forms with the substitutions

t → zl
q → z(1 − l)

giving
∫ ∞ ∫ ∞ ∫ ∞ ∫ 1
dt dq −→ dz f (z, l) dl
0 0 0 0

where we need to find the function f (z, l). The Jacobian can be calculated in the usual way. Here we’ll use differential
forms and write

dt = l dz + z dl
dq = dz − z dl − l dz

dt ∧ dq = (l dz + z dl) ∧ (dz − z dl − l dz)

= l dz ∧ dz − l dz ∧ z dl − l dz ∧ l dz
+ z dl ∧ dz − z dl ∧ z dl − z dl ∧ l dz
= −l dz ∧ z dl + z dl ∧ dz − z dl ∧ l dz
= −lz dz ∧ dl + z dl ∧ dz + zl dz ∧ dl

⇒ dt dq = z dl dz

Our double integral is transformed to

∫ ∞ ∫ 1
Γ(n)Γ(m) = dz e−z z n+m−2 z dl ln−1 (1 − l)m−1
0 0

or ∫ 1
Γ(n)Γ(m) = Γ(n + m) dl ln−1 (1 − l)m−1 .
0

Finally we have the form of the beta function we desire

∫ 1 Γ(n)Γ(m)
dl ln−1 (1 − l)m−1 = .
0 Γ(n + m)

The LHS is the beta function β(n, m). Note that when doing this manipulation the substitutions must follow the

25
 4 WORLD SHEET SYMMETRIES

rules t + q = z and t = z f (l), and g = z g(l), where f (l) and g(l) are functions of l. This allows the gamma function
to come out.

4 World sheet symmetries

The trajectory for a point particle is a world line xµ (λ) where λ is a parameter along
the world line of the particle. A string sweeps out a world sheet. Each point along
the string is parameterized by xµ (τ, σ). The path integral is written as
∫
[Dx] eiS

with ∫ [( )2 ( )2 ]
1 ∂xµ ∂xµ
S= dτ dσ − (41)
2 ∂τ ∂σ

In order to make the path integral converge we Wick rotate τ → −iτ .6 An integral
over all surfaces is what defines the path integral for strings. For closed strings the
world sheet of a single string is a tube and interactions among closed strings look like
tubes that join into larger tubes, and separate again. This is the generalization of a
single vertex in point particle QFT. If one allows the strings to have any arbitrary
number of holes it becomes the generalization of the sum over all Feynman graphs
containing any number of loops and so on. So the path integral is a sum over all
possible surfaces that connect the starting positions and end positions of a set of
strings.
The equation of motion for (41) yields the wave equation. After Wick rotation
the signs of the derivative terms are the same and the equation of motion will instead
be the Laplace equation
∂x2µ ∂ 2 xµ
+ =0
∂τ 2 ∂σ 2
The symmetries of the Laplace equation include reparameterization of the coor-
dinates σ ′ = (σ, τ ) and τ ′ = (σ, τ ). In particular, the kinds of coordinate transforma-
tions that preserve the Laplace equation are conformally invariant transformations.
Conformal transformations preserve angles, and infinitesimal squares are mapped
∫ ∫ t2
The generic path integral in QFT is given by something like [Dx] ei t1 dt L(x,ẋ) where the
6
∫t
weight for each path is eiS(x) with S(x) = t12 dt L(x, ẋ) and therefore has the same magnitude
(they are points on the unit circle in the complex plane). Generally, this integral is hard to define
mathematically and is non-convergent. The standard prescription is analytically continuing the path
∫ ∫ −it
− −it 2 dτ L(x,ẋ)
integral by setting t = iτ so that the path integral becomes [Dx] e 1 = F (it1 , it2 ) and
then analytically continuing the result back.

26
4 WORLD SHEET SYMMETRIES 4.1 Conformal mapping

Figure 4: Conformal mapping of the two-string interaction in Fig. 3 to the unit disk.
The interior of the disk is the interior of the shaded string region in Fig. 3 and the
injection points are shown as exaggerated cutouts in the disk. These are mapped to
infinitesimal points on the exterior of the disk.

Figure 5: s- and t-channel scattering maps. The τ parameter determines the sym-
metric locations of the four vertex operators. For τ = 0 the points would move closer
and overlap on the left. For τ → ∞ the point on the right map would move closer
and overlap.

to infinitesimal squares. In particular the shaded region in Fig. 3 is topologically

equivalent to a unit disk with four special points (vertex operators, or insertions)
corresponding to the string pieces that come in from minus infinity and go out to
plus infinity. The special points are injection points for the momenta ki . The s- and
t-channel scattering

4.1 Conformal mapping

Conformal mapping is usually introduced in electrodynamics in physics to map 2D
electrostatics problems into one another as an aid in obtaining solutions.7 These
7
Incidentally, for a charge confined to 2D, where the flux is also confined to the 2D surface
containing the charge, the electrostatic force F ∼ 1/r, and the potential is then V ∼ ln r. This
is easy to see by contrasting with charges in 3D. The flux eminating from a 3D charge would be
measured by surrounding the charge with a sphere whose area is 4πr2 and so the flux escaping a

27
4.1 Conformal mapping 4 WORLD SHEET SYMMETRIES

techniques are standard methods used in elementary complex variables so we will

summarize the basics quickly. Consider a mapping w(z) that takes the complex z
plane to the complex w plane. Where z = x + iy and w(z) = u(x, y) + iv(x, y). We
assume that the mapping is one-to-one and onto (bijective). In defining the derivative
of the mapping, w′ (z) we examine the quantity

dw du + idv
w′ (z) = = .
dz dx + idy

For the derivative to be uniquely defined, it must be the same if approached along any
direction in the z plane. It is easiest to come in along the x and y axes independently,
so that dy = 0 and dx = 0, respectively. Setting these two derivatives equal yields
the Cauchy-Riemann equations

∂u ∂v ∂u ∂v
= =− . (42)
∂x ∂y ∂y ∂x

Mappings that satisfy the Cauchy-Riemann equations are called analytic or holomor-
phic. Setting the mixed partials of these equations equal shows that both of the
functions u(x, y) and v(x, y) satisfy the Laplace equation, ∇2 u = 0 and ∇2 v = 0.
We will now show that any analytic mapping is a conformal mapping, i.e. it
will preserve angles. For this we consider two small arrows in the z plane, δz =
′
ρeiθ and ∆z = ρ′ eiθ starting at the point z. The ratio of these displacements is
′
δz/∆z = (ρ/ρ′ ) ei(θ−θ ) , indicating that the angle between these displacements is
(θ − θ′ ). For an analytic mapping w(z), we stipulate that the derivative w′ (z) exists.
Then w(z + δz) = w(z) + δz w ′ (z) and w(z + ∆z) = w(z) + ∆z w′ (z). Therefore we
have the following ratios equal
δz δw
=
∆z ∆w
and the mapping is angle preserving. A few simple examples of analytic functions
are f (z) = z, and any power of z n , except 1/z, which has a pole at the origin, but
is analytic everywhere else. Typical functions such as ez and sin(z), are analytic.
An example of a non-analytic function is f (z) = z ∗ . It is easy to see that this
function does not satisfy the Caucy-Riemann equations. However, it is important in
string theory to consider functions that are anti-analytic. These are functions that
are analytic in the variable z ∗ , often denoted with a bar as z̄. The Cauchy-Riemann
equations for anti-analytic functions differ by a minus sign from the usual definition
small patch of area da on the sphere must be da/4πr2 . In 2D, the flux would pass throug a circle of
’area’ 2πr.

28
4 WORLD SHEET SYMMETRIES 4.1 Conformal mapping

z+1
Figure 6: Linear fractional map w(z) = z−1
.

(42). If f (z̄) = z ¯=x − iy, then u = x and v = −y, so that we must have ∂u ∂x
= − ∂y
∂v
.
There are two special mappings that we now consider. The first is the function
w(z) = log(z), where we write z = reiθ and consider the mapping of the right half-
plane. From the properties of the logarithm we have w(z) = log r + iθ. The vertical
line from the origin at z = 0 to +i∞ takes log r from −∞ −→ +∞ (this is the same
for any radial line from the origin out to infinity) at angle θ = π/2. In the w-plane
this is a horizontal line at height iπ/2. The vertical line from the origin to −i∞ is
mapped to a horizontal line at −iπ/2. Thus the right half z-plane is mapped into
a strip in the w-plane that extends horizontally from −∞ to +∞, going from −π/2
to +π/2 in the vertical direction. Half circles (fixed radius) in the right half-plane
are mapped to vertical lines in the strip. This mapping is useful for open strings
where the σ parameter ranges from 0 to π. If one maps the entire z-plane with the
logarithm one obtains a strip in the w-plane that is 2π wide, and where the top of
the strip is identified with the bottom of the strip. This describes a cylinder and is
useful for closed strings.
z+1
The second mapping we consider is the linear fractional mapping w(z) = z−1 .
This maps the point z = 0 to w = −1, and maps z → ∞ in any direction (the
“point at infinity”) to w = +1. The vertical y-axis is mapped to a half circle in the
w-plane going from -1 to +1. This can be seen by writing z = reiθ for θ = π/2, then
( )
w(z) = − 1+ir 1−ir
. As r → ∞ this goes to +1, and in between there is a small excursion
into the positive complex part of the w-plane. The negative y-axis is mapped to a half
circle in the negative imaginary direction. Along the positive x-axis the mapping hits
a pole at z = 1. The mapping takes (0,1) to (∞,1) and takes (1,∞) to (-1,−∞), never
covering u ∈ (−1, 1). Overall the mapping takes the left half-plane to the interior of
the disc, and the right half plane to the exterior of the disc.
The defining path integral for the open string scattering amplitude can then be

29
4.1 Conformal mapping 4 WORLD SHEET SYMMETRIES

written as
∫ (∏ )∫
∫ [
∂Xµ ∂X µ ∂X µ
] ∏
− dτ dσ + ∂σµ ∂X (j)
X µ (⃗
A= dzj DX e ∂τ ∂τ ∂σ
eikµ z)
(43)
j j

where the integral over the zj is the integral of the vertex operator positions on the
boundary of the unit disc that inject strings of momentum k (j) into the world sheet.
By conformal mapping three of these positions can be fixed, but the rest must be
integrated over. The number of vertex opertors corresponds to the number of strings
interacting. This is a genus 0 sheet, it captures all s- and t-processes and anything
that can be deformed topologically without tearing the sheet. So this takes care of
all tree-level processes.
The action in (43) is conformally invariant and the equation of motion for each
field X µ satisfies the Laplace equation. The path integrals are all Gaussian and can
be done independently for each of the 26 fields. The problem is therefore reduced
(j)
to solving the Laplace equation on the disc with the charges being the kµ . In the
(1) (4)
case where there are four strings interacting, we have kµ , · · · , kµ and 26 kinds of
“charge,” one for each dimension. What is being calculated is the analog of the
electrostatic energy for each field X µ . The final result is that the amplitude is given
by
∫ (∏ ) ∑
A= dzj e− j Ej (⃗z) (44)
j

where Ej (⃗z) is the energy of the j-th type charge which depends on the locations of
all the other injected charges of the same type. This is the simplest form of open
string theory and (43) can be taken as a definition of this theory.
This result (44) is calculated for strings in 26 dimensions and the mathematics
so far has been relatively simple. The real complexity in string theory comes in
with the compactification of these extra dimensions to leave only the 3+1 dimensions
that we see. The number of ways of compacting these dimensions is very subtle
and mathematically difficult. The compactification allows one to put gauge group
symmetries into string theory thus giving a connection the Standard Model. However,
the number of ways of compactifying results in a large number of possible “standard
models” of the order 10500 . The current problem with string theory is that there
seems to be no way of picking one and that they may occur randomly.

30
4 WORLD SHEET SYMMETRIES 4.2 Compactification

4.2 Compactification
Let us contrast the difference between a point particle and a string moving on a surface
embedded in a large flat space. A point particle that is constrained to remain in the
surface with no other forces on the particle will move along a geodesic. Classically,
we can divide up the trajectory of the particle into small pieces, and in the limit we
have a continuous line. Quantum mechanically the picture is different.
On a sphere of radius R a point particle with no other forces would travel on a
great circle. The energy of the center of mass of the particle would be mv 2 /2 = p2 /2m.
Because the angular momentum is quantized, and because L = mvR = pR, we have
that

E = L2 /2mR2
= L2 /I, (45)

where the moment of inertia is I = mR2 . For a string moving on the geometry of a
sphere things are much different. We have to consider the spatial extent of the string
due to its zero point motion. We could calculate the root mean square by calculating
√
Xrms = ⟨0|X 2 (σ)|0⟩. Using the creation and annihilation operators we have

∑ 1 ( )( + )
⟨0| √ √ ′ a+ −
n + an an ′ + a−
n′ |0⟩
nn′
n n
∑ 1
and the only contributing term will be a− + ′
n an′ with n = n . This will give n n which
∑nmax
is formally divergent. Again we regulate the sum using a cutoff n and look for
the dependency of the energy in (45) on the regulator. The integral approximation to
∑
the sum is nnmax ∼ ln(nmax ). On a sphere, when the higher ZP modes are included
in the string Xrms begins to cover the surface of the sphere and the CM of this “string
tangle” moves closer to the center of the sphere and in the limit of including all the
modes we find I → 0, and (45) becomes divergent. We interpret this to mean that
string motion in this geometry is not allowed; the sphere is a bad geometry for strings
to move in. In order to find a suitable geometry we have to consider the metric, gµν (x),
of the surface it moves in. Because the string itself has energy and can deform the
metric of the space, we need the change in the metric of the space to be well-behaved
with the regulator introduced above. This will require that the space be “Ricci flat.”
We can write this as
δgµν = Rµναα = Rµν . (46)

31
4.2 Compactification 4 WORLD SHEET SYMMETRIES

This is the variation of the metric with a change in the regulator. We want this to
be zero.
δgµν = Rµν = 0 (47)

Now consider the Einstein equations for the vacuum,8

1
Rµν − gµν R = 0.
2

The scalar curvature R = Rαα is the trace of the Ricci tensor. Raising one index in
the free field equation we have

1
Rµν g να − gµν g να R = 0
2
1
Rµ − gµα R = 0
α
2
1
Rµ − δµα R = 0.
α
2

Now form the trace by setting α = µ and summing, giving

4
R− R=0
2
R = 2R.

Clearly this can be satisfied only for R = 0, or the space is Ricci flat. This means that
the consistency condition (46) for strings moving in a spacetime background requires
that the background space be a solution of the free field Einstein equations. This is
remarkable.
Compactification of a surface can be done many ways, and we begin with an
example of a 2D spatial sheet of finite extent. If we take a rectangle and identify
the left side with the right side we obtain a cylinder. If we take the same rectangle
and identify the lhs with the rhs and the top with the bottom, we obtain a torus.
Topologically the two are identical if we remember the identifications required on the
rectangle. If we start with a 3D rectangular parallelopiped, infinite in one direction,
and identify the front with the back and the top with the bottom (the compact
8
The full Einstein field equations have the energy-momentum tensor Tµν on the rhs that contains
matter and everything other than the gravitational field. Solutions to the vacuum equations include,
in particular, gravitational waves, e.g. think about the vacuum Maxwell equations that admit wave
solutions. Other “solutions” include Schwarzchild, and FRW, but those have singularities. The de
Sitter spaces are solutions to the vacuum equations provided there is an additional cosmological
constant term.

32
4 WORLD SHEET SYMMETRIES 4.3 Winding number

directions), we have a torus at each point of the non-compact direction. This can be
imagined as a 1D line with a torus attached at each point of the line. Similarly we
could imagine another topology consisting of a line with a sphere attached at each
point of the line. However, we have seen above that a sphere is not a good geometry
for strings to move on. Toroids are the simplest and most direct way to compactify
dimensions, and this process is called “toroidal compactification.”
There are three parameters that describe a 2-torus. If we unwrap a 2-torus we
first cut it and unwrap it to form a cylinder, and then cut the cylinder lengthwise to
a rectangle of sides a and b. The three parameters are (1) the area ab, also called the
Kahler modulus, (2) the ratio of the sides a/b, is the complex structure modulus, and
(3) the identification “angle” or parameter that describes any twist applied to the
cylinder before reconnecting it to form the 2-torus is called the Dehn twist. These
are called the moduli, and make up the moduli space of the torus.9
Other Ricci flat manifolds called Calabi-Yao manifolds are good manifolds for
strings to move on. They appear to provide more of a connection to the particle
physics we know because they are less symmetric than tori, and the tori are too
symmetric.

4.3 Energy spectrum and winding number

Consider a point particle moving on a 2D surface which is compact in one dimension,
e.g. a long cylinder of radius r. It can have two components of momentum, one
along the non-compact direction and one along the compact direction. Momentum
along the compact direction is quantized. This can be seen using the usual periodic
boundary condition for a plane wave, eikx = eik(x+2πr) , and so 2πkr = 2πn for some
integer n. Therefore km = n/r. For a massless particle E = |p|, and so E ∼ n/r for
n ∈ Z. Because energy is mass, the mass spectrum of these particles will have spacing
1/r. As r → ∞, then the spacing between the states is very small. This particle can
also be a string of small spread.
Now consider a string wrapped around the compact dimension. Its length has an
energy density, so the number of times the string is wrapped around the compact
dimension gives its energy, and this goes as nr, where n is called the winding number.
Negative n correspond to the direction in which the string is wound. So we have
two types of mass spectra, the non-wound string that goes as n/r, and the wound
9
There are non-orientable surfaces such as the Klein bottle that can be made by identifying the
cylinder edges before reconnecting, but they are not good surfaces for strings to move on. The
surfaces must be Ricci flat.

33
4.3 Winding number 4 WORLD SHEET SYMMETRIES

string that goes as nr. Note that the spectrum of these two pictures exchange and
go into each other as r → 0 or r → ∞. These pictures are dual to each other
by exchanging the n of the quantized momentum with the winding number and
exchanging r ←→ 1/r. This is a symmetry of the theory. At the radius r = 1
the spectra are the same. As one reduces the compact dimension length scale below
r < 1, the non-wound theory just looks like the string theory with winding number
in a space with r > 1.10 This is called T-duality where the T is for torus.
The winding number is a conserved quantity. To show this we remember that
for closed oriented strings, there is a sense of direction along the string. Take two
strings of winding number w = 1 and w = −1 wrapped around the compact direction
in opposite directions. If the strings are allowed to interact and combine, we must
follow the rule that the sense of direction of a string cannot change along the string.
Then when the strings combine, the only way for them to do so is to form one string
that is no longer wound around the compact direction. The total winding number 0
is conserved. Drawing a picture is easy and it will be obvious. If instead we wrap
two strings both with w = 1, and let them combine to form one string it is easy to
see that the winding number of the single string is 2.
The mathematics of the T-duality can be described as follows. For an unwound
string with (quantized) momentum along the compact direction we write E = p =
n/r, but p ∼ dy/dτ where y(σ) is the coordinate in the compact direction at the
point σ along the string. So we can write
∫
dy(σ)
n=r dσ
dτ

where the integral over dσ accounts for different parts of the string moving at different
velocity. Similarly, we can consider the winding number of a string. The coordinate
along the string σ goes from 0 → 2π for a closed string. The circumference of the
compact space coordinate is 2πr, therefore we can write σ = y/r so that as y goes
from 0 → 2πr, σ goes from 0 → 2π. Here the energy E = w2πr is proportional to r
and we write ∫
1 dy(σ)
w= dσ .
r dσ
T-duality takes n → w, r → 1/r, τ → σ.
10
This implies that there is a smallest length scale.

34
5 KALUZA-KLEIN

5 The Kaluza-Klein Picture

Closed strings on a compact manifold exhibit repulsion and attraction; they behave
like charged particles. If two strings attract or repel each other in this way it is
natural to ask how. For simplicity we consider 5 dimensions, one time and four
space, one of which is compact. This is the Kaluza-Klein (KK) picture that is able to
combine electromagnetism with general relativity. The mathematics is also used in
string theory. Here we are considering closed strings. We will see that closed strings
with momentum number +n in the compact direction attract those with momentum
number −n, and likewise for strings with winding number +w and −w. Repulsion
occurs between objects whose momentum number or winding number is the same
sign.

5.1 Kaluza-Klein metric

The generalization to an extra compact dimension of space was first proposed by
Kaluza and later refined by Klein. The idea is simple. Instead of the usual GR
metric gµν where µ, ν ∈ {0, 1, 2, 3} we define a more general metric tensor
( )
gµν gµ5
gM N = . (48)
gµ5 g55

We want to find an analog electric or magnetic field to describe the attraction and
repulsion of these strings. Recall from electromagnetism that the fundamental object
⃗ = ∇ × A,
is the vector potential Aµ = (ϕ, A1 , A2 , A3 ) where B ⃗ and E ⃗ = −∇ · ϕ − ∂t A.
⃗
The vector potential Aµ is a four-vector. The photon is a vector particle and has
a vector index corresponding to the µ. The KK metric tensor contains two new
fields, gµ5 , a four-vector that can be identified with Aµ , and g55 , a scalar Φ called the
dilaton. The dilaton or g55 is the component of the metric around the 5-th direction
and tells you how big the 5-th direction is. If g55 is big or small, then so is the 5-th
direction. Both new fields are functions of the spacetime coordinates gµ5 = gµ5 (x)
and g55 = g55 (x). If g55 can vary, then the compact dimension can vary and looks like
a tube of varying radius. These represent waves in the compact direction or particles
associated with the dilaton. The quiescent vacuum would have no such waves and
the radius would be constant.
Recall the definition of the energy-momentum tensor Tµν , the time components
T0j correspond to the momentum flow in each spatial direction. In the KK metric the

35
5.1 Kaluza-Klein metric 5 KALUZA-KLEIN

terms gµ5 correspond to momentum flow in the 5-direction. The source of the analog
vector potential is the component of momentum going around the 5-direction. The
momentum quantum number ±n can be though of as electric charge corresponding to
the vector field gµ5 . This electric field is associated with the graviton. The momentum
number n, and the winding number w correspond to two kinds of electric charge,
giving rise to n-photons, and w-photons. Momentum photons are a piece of the
gravitational field gµ5 , gravitons which are polarized along the µ5-direction. The more
general picture including photons emitted and absorbed by the winding number are
as follows. Recall the spectrum of closed strings where we had to use level matching.
The creation and annihilation operators can now create oscillations in the usual three
( )±
directions of the non-compact space. These are ai±n where i ∈ {x1 , x2 , x3 } and
the left and right operators are given by ±n. The tachyon has -2 units of energy and
we ignore it. To get a non-tachyonic state we must create two units of energy and
−1 a+1 |0⟩, corresponding to gravitons polarized in two directions in
j+
form states like ai+
space. We can also form excitations like ai+ −1 a+1 |0⟩, which has one index and can
5+

be identified with the polarization of the photon since we have only one index in
this state. We can also form ai+ +1 a−1 |0⟩. The correct thing to do is to form linear
5+

superpositions and form two kinds of photon fields

( )
ai+ a5+
+1 −1 ± a −1 +1 |0⟩
i+ 5+
a

where + ↔ n-photons, and − ↔ w-photons. The n-photons are identified with gµ5 ,
gravitons, and the Ramond field bµ5 is associated with the w-photons or winding num-
ber. There is also a combination of creation operators that have no index structure
+1 a−1 |0⟩, the field quanta of the dilaton field Φ.
and correspond to a scalar particle, a5+ 5+

T-duality can be stated

n↔w
r ↔ 1/r
∂x/∂τ ↔ ∂x/∂σ
gµ5 ↔ bµ5 .

At the moment these are mathematical constructions and real particle candidates are
not known for all of these particles. However, there is plenty of room in physics for
these objects.

36
5 KALUZA-KLEIN 5.2 D-branes

5.2 D-branes
T-duality for open strings leads to the idea of Dirichlet branes or D-branes, discovered
by Polchinski. The D-branes can have different dimensions and a Dn-brane has
n dimensions. Consider the space R3 × S 1 where the compact space is S 1 with
coordinate y and take it to be very small. Suppose a string lives in this space. The
Neumann boundary condition on the string is ∂y/∂σ|0,π = 0, which means there is no
stretch of the string at the very endpoints. If we insist that T-duality holds for the
string, then in the dual space where S 1 is very large we have the dual condition that
∂y/∂τ |σ=0,π = 0, or that the string end points cannot move; the string is pinned in
the S 1 space. This object that has dimension 1 anchors the end of the string and can
be thought of as a 1-dimensional object that is heavy, because strings attached to it
cannot leave it. These objects have to be deformable to make sense from a general
relativistic point of view, and so they are objects that have their own dynamics.
A D0-brane is a point, a D1-brane is a string (though it is heavy and not like the
strings that are attached to it). A D2-brane is a membrane, and a D3-brane is a
3-dimensional space where strings attach. If we think about open string theory in our
3-dimensional world we would say the strings are attached to a space-filling D3-brane
where they move all about the space because the endpoints are attached to the whole
space. One can say that open string theory in which T-duality is imposed predicts
D-branes.
The strings attached to a D-brane have an orientation that gives rules for joining
and breaking of strings for scattering amplitudes. Each string attached to a brane
has an arrow along its length. If two strings attached to a D2-brane approach each
other the only ends that can join are the two of opposite orientation. In other words
two strings cannot join that have arrows coming out of the points. That would lead
to a string that had two arrows along its length of opposite direction. This leads to
strings conserving orientation number.
QCD has a beautiful connection to D-branes and string theory in the following
sense. Consider three separate D3-branes labeled R, G, B that are embedded in a
higher dimensional space. Then a string attached to the R3-brane has points labeled
r, r̄, where one has an arrow going into the brane and the other an arrow going out.
Strings can attach at one brane and go to another type of brane so there could be
rb̄ strings, and so forth. Incidentally there are 8 gluons and not 9 because one linear
combination can come together and disappear. For example the ends of an rr̄ gluon
can come together and lift off the sheet and thus is not stable. An isolated quark in
this picture would be a string that starts one of the RGB sheets and goes off to some

37
 6 FUNDAMENTAL OBJECTS

distant other brane of a different kind. These rules are the same as super Yang-Mills
QCD rules.
Supposing we have only one D3-brane the structure of the theory is like electro-
magnetism. In this picture a string with both ends attached to the brane is a photon
like object, and a string with one end in the brane and one end at infinity is a an
electron. A positron would have the opposite orientation. The theory predicts in this
picture that a D1-brane that ends on the D3-brane is a magnetic monopole. It is
heavier than the usual string that ends on the D3-brane that represents the electron
or positron.
Finally, if open strings are attached to D-branes, and if our physical world is
a D3-brane with open strings attached. Then open strings may describe all of the
Standard Model particles, and closed strings which do not attach to these branes may
describe the gravitons. These gravitons can wander freely over all of the spaces and
this may explain why gravitational forces are so much weaker than the others; gravity
can “leak out” into the other dimensions.

Part II

Topics in String Theory

This part of the document follows the first several lectures of Susskind’s “Topics in
String Theory” lectures.

6 Fundamental Objects
In quantum field theory there is a well-known picture associated with bosonization
that indicates there is difficulty in determining what objects are fundamental or not.
Consider two spin-1/2 fermions with coupling g. A field ψ describes the fermion field.
If g ≪ 1 then the fermions move freely and are considered the fundamental particles.
If the coupling is very strong, g ≫ 1, then the fermions look like a composite particle
that can have a ground state of spin 0 and look like a boson. In this picture the
boson would be considered the fundamental particle. The boson field ϕ is an effective
description of the fermions. The boson field creates pairs of fermions. The bosons
cannot be paired together to look like fermions because any pair of two bosons is also
a boson. However, there is a completely different way to think about the fermions in

38
6 FUNDAMENTAL OBJECTS 6.1 The electron and the monopole

terms of the bosons in terms of kinks in the boson field. The field ϕ smoothly jumps
over some small distance. A good example is a belt that is twisted 2π over some small
distance. Kinks are heavier than the bosons that make up excitations in the field ϕ,
and these kinks behave like fermions.
For some ranges of the parameters in the field theories ψ and ϕ it is easier to
think of ψ is the fundamental field and its building blocks are fundamental, and for
other parameters of the field theory the bosons are more useful to think of as the
fundamental objects. Here the bosons would behave more simply for example.

6.1 The electron and the monopole

A more sophisticated example is given by quantum electrodynamics. Here we think
of the electron as a fundamental particle. The fine structure constant α ∼ e2 , with
a numerical value of ≈ 1/137, is the coupling constant that describes the probability
of an electron emitting a photon if it is accelerated. The electron only emits photons
perhaps 1% of the time. It takes very large energies to probe the structure of the
electron and infer the cloud of virtual photons and e+ e− pairs that surround it. From
Dirac’s famous calculation of the charge of the magnetic monopole, g = 2π/e, we see
that the magnetic monopole will interact with photons much more strongly. In doing
so there will be a complicated cloud of virtual photons, and electron-positron pairs
that surround it. The monopole is therefore much heavier than the electron. If we
consider the electric charge e as a parameter, then as we vary the parameter from
very small to very large, the electron and the monopole trade places in the complexity
of the cloud that surrounds them. The monopole would become very light and the
electron very heavy. It is therefore not entirely clear then what object is considered
fundamental.

6.2 D-branes and S-duality

In string theory the presence of D-branes is part of the mathematical framework and
these objects must exist as discussed in the section above. We will now discuss two
versions of string theory. In one theory there are only even dimensional D-branes and
in the other only odd dimensional D-branes. If we consider first the odd dimensional
version, then there are no D0-branes, but there are D1-branes which we discussed
as being much heavier than fundamental strings. In string theory the parameter
that allows for duality is the string coupling constant, g. This constant controls the
probability of strings breaking and connecting.

39
6.2 D-branes and S-duality 6 FUNDAMENTAL OBJECTS

In a theory that contains D1 branes the same behavior is present, the D1 string
has lots of fundamental strings (F-strings) attached to it, and so it is a heavy object.
In this type of theory a fundamental string and a D1 string will morph into each other
based on the value of the coupling constant g of the theory. If g ≪ 1 the F-strings are
very light. As g becomes large the F-string gets heavier and so the D1 string cannot
“emit” as many F-strings and so it starts to look more simple. At the same time any
bits of string that break off from the F-string will form closed loops and be attracted
around the F-string. The closed loops form the particles like the graviton, etc. As
g ≫ 1 these become very massive and turn into black holes. The D1 strings then
get lighter and they make up the gravitons. This is called S-duality for “strength
duality”, the strength of the coupling constant. In string theory g can vary in space
and it satisfies field equations.
In a string theory that contains D0 branes they behave like heavy objects because
they have lots of fundamental strings attached to them. The look like little loops that
come out of and return to the D0 brane, like a fluffy ball. Imagine a 3D space where
one of the dimensions is compactified that looks like a thick sheet. As g starts to
increase, the compact dimension decompactifies. The D0 branes become lighter and
turn into the gravitons in the decompactified space. In 10D string theory this results
in 11 total dimensions. The F-strings in the original 2D space (with compactified 3rd
dimensioin) that form a loop for example will, upon g getting larger, form ribbons
that are stretched in the direction that is becoming decompactified. These are D2
branes. There are also D2 branes that float between the two noncompact spaces.
This is the nature of the dualities in string theory.
D3-branes can act as our usual 3D space. One can imagine that two strings
attached to the D3-brane can come together and join, move around, and then break
again. This looks like the particle scattering that we are familiar with from quantum
field theory.
There are other objects in string theory in addition to D-branes. There are fluxes,
and other objects. The spaces of real interest in string theory are Calabi-Yao man-
ifolds. They have handles or holes. A relatively simple Calabi-Yao manifold with
genus 500, then around each handle we can wrap fluxes, D-branes, and so forth. This
configuration space is what gives the roughly 10500 possible ground states for string
theory that result in the many different possible particle physics we see. It could
be that particle physics is so complicated because there are lots of moving parts in
the string theory, even though the underlying rules in the string theory are relatively
simple. If string theory is correct then particle physics may be so complicated that

40
7 THE SCHWARZCHILD METRIC

we will not be able to completely unravel. The figure 10500 is actually a conservative
number and of this large number there may be very many that look similar and so
we may never be able to tell the difference between individual choices, but we may
be able to only know the region of the configuration space we may be in.
Do we live in a compactified space? We don’t know. Old-fashioned cosmological
theories were that we live on the 3D surface of S 3 , so all dimensions were compact!
Observationally we can only say how flat the universe looks.

7 The Schwarzchild metric

Let us consider first the Euclidean and Minkowski metrics. The Euclidean metric is
ds2 = dx2 + dy 2 + dz 2 and is the usual generalization of the Pythagorean theorem.
There is no time in the Euclidean metric. The Minkowski metric is ds2 = c2 dt2 −
dx2 − dy 2 − dz 2 where time and space have different “signature.” In terms of proper
time we would write dτ 2 = dt2 − (dx2 + dy 2 + dz 2 )/c2 .11 It is seen here that the
speed of light is just a conversion factor for the dimensions. Light rays follow null
trajectories and in the Minkowski metric this means that dτ 2 = 0, and in one space
dimension we have 0 = dt2 − dx2 /c2 or dx/dt = ±c for left or right moving light
rays. The universal statement in physics is that the proper time along a light ray is
zero. We have just seen in Minkowski space, which defines special relativity, that this
implies the constancy of the speed of light.
From here on we set c = 1. Writing the Minkowski metric in spherical coordinates
we have dτ 2 = dt2 − (dr2 + r2 dθ2 + r2 sin2 θ dϕ2 ) which we write as dτ 2 = dt2 −
dr2 − r2 dΩ22 where dΩ22 is the metric for a unit 2-sphere dΩ22 = dθ2 + sin2 θ dϕ2 . The
Schwarzchild metric is a spherically symmetric line element describing a black hole
from the point of view of a stationary observer at spatial infinity. The time for the
observer at infinity is called coordinate time. The metric is given by
( )
2M G 1
2
dτ = 1− dt2 − ( ) dr2 − r2 dΩ22 (49)
r 1− 2M G
r

where rS = 2M G is the Schwarzchild radius and is the radial location of the horizon
of the black hole. We can examine how a radial light ray behaves by looking at the
11
There are two conventions when writing metrics. The invariant interval is what it is, invariant.
However if one is interested in time-like events one uses the proper time convention dτ 2 = dt2 −
dx2 − dy 2 − dz 2 , so that the proper time comes out positive. The proper-distance convention is for
investigating space-like phenomena and one puts in a sign change, writing ds2 = −dt2 + dx2 + dy 2 +
dz 2 , so that the proper distance is positive.

41
7.1 Black holes 7 THE SCHWARZCHILD METRIC

null trajectory and ignoring the angular part of the metric. We have then

dr ( rS )
=± 1−
dt r

and outside the horizon light rays, both ingoing and outgoing, have smaller and
smaller velocities in coordinate time. The constancy of the speed of light is broken in
curved space! From the perspective of the observer at spatial infinity, ingoing objects
never cross the horizon but rather get ever closer; the horizon has a “sediment” of
flattened objects just outside. Even a light ray shined into the black hole never
crosses the horizon from the perspective of the observer at spatial infinity. From the
perspective of the observer falling into a sufficiently large black hole the experience is
unremarkable and they simply cross the horizon. At r = rS there is a type of singular
behavior but the real coordinate singularity is at r = 0 where the tidal forces become
infinite. We will come back to the horizon and what happens there later.

7.1 Black holes

7.1.1 Information and entropy

The idea of entropy goes back to the nineteenth century and the work of Boltzmann
who put it on firm mathematical grounds. The more recent ideas in information
theory began with the work of Claude Shannon in the 1940’s. The information content
of a system can be determined by asking a series of “yes” or “no” questions about the
system and forming a string of 1’s and 0’s to represent the answers. For a macroscopic
object this list would be very large and could contain for example a list of whether
there is a molecule or not in each small subdivision of the object.
The essential concept behind entropy is that it is the amount of inaccessible in-
formation content in a physical system. Take the following example. Consider two
large tanks that are half full of water. In the first tank one drips in water in Morse
code that contains the Feynman lectures on physics, and in the other tank one drips
in the MTW treatise on gravitation. After some time the tanks become quiescent.
By all appearances the water in the tanks looks identical.12 In fact, the information
is still in the water, but it is contained in degrees of freedom so numerous and so
difficult to measure that it is for all practical purposes inaccessible. The entropy,
temperature, and energy of a system are related by dE = T dS. A modern definition
12
The rule in physics is that information never disappears it simply moves from one form to
another. For this rule one needs the concept of ergodicity in the phase space of the system where
phase space trajectories never cross; the trajectory is required to be microscopically time-reversible.

42
7 THE SCHWARZCHILD METRIC 7.1 Black holes

of the temperature of a system is the following. The temperature of a system is equal

to the change in energy of a system when adding one bit of information. If ∆S = 1,
then ∆E = T .
A black hole that absorbs energy (information, if thought of as the bits of infor-
mation that are arranged in the Planck squared array on the surface of the hole) will
eventually radiate that energy out. So the information that fell into the black hole is
evaporated back out through Hawking radiation. This fact has to be reconciled with
the notion that nothing can escape from behind the horizon. This is a major puzzle
of quantum theories of gravity.
The conservation of information is more fundamental than the conservation of en-
ergy or momentum. It is so fundamental that most people forget about it. Without
conservation of information there would be no conservation of energy or momentum.
Consider an object in empty space where there is no friction whatsoever. Suppose that
the natural law was that objects given a push will come to rest because they followed
equations of motion that said they did. It follows then that no matter what direction
the object was pushed it will come to rest without any interaction with anything. If
the object “forgets” where it came from there is no time reversal symmetry. From
time reversal symmetry follows energy conservation, so there is no energy conserva-
tion.13 Also, there is no momentum conservation if it comes to rest. Is translation
invariance broken? No, but then the statement that momentum conservation follows
from translation invariance is broken. For momentum conservation it is necessary to
have both information conservation and translation invariance.
In short, information conservation is probably the most fundamental idea in
physics. It is similar to, but distinct from, the idea of causality. These are both
meta-principles without which there is essentially no place to begin thinking about
physics. In classical physics the statement of information conservation is the same as
Liouville’s theorem that the phase space volume of a system does not change. In quan-
tum mechanics the conservation of information is called unitarity: orthogonal states
remain orthogonal with time, which means that distinctions are always preserved.
Unitarity is deeply connected with energy conservation in quantum mechanics.

7.1.2 Black hole quantum mechanics

Black holes are objects. Are they objects that violate quantum mechanics? Do things
fall into a black hole? If they go in then all the information they take into the black
13
Clearly if an object simply comes to rest and there is no friction to slow it down, then there is
no energy conservation.

43
7.1 Black holes 7 THE SCHWARZCHILD METRIC

hole is lost inside. If things don’t fall in it appears to be stuck at the horizon. However
black holes evaporate over time, so what happens to the information? Does it get
radiated out, is it stuck in the remnant of the evaporating black hole? There are many
such questions that were largely resolved in part by string theory. The Beckenstein
argument for an evaporating black hole is the following. First we must establish that
a black hole has a temperature and an entropy.
Using our fundamental definition above, that the temperature of a system is the
change in energy of a system due to adding one bit of information, we calculate the
change in energy of a black hole. If we can calculate this, then the change in the
radius of the horizon is simply related to the change in mass. To add one bit of
information to a black hole it is most simple to consider throwing in one photon. It
contains one bit of information in its polarization state. But if the black hole is large,
say a kilometer, we don’t want to throw in a photon of visible light that has a small
wavelength. Then one could ask whether the photon entered at an angle θ with the
black hole, etc. Instead we use light of wavelength near the size of the black hole so
that its location of entry is uncertain.14
Proceeding with the calculation we need the Schwarzchild radius rS = 2M G/c2 .
The energy of a photon of this radius is E = hν = hc/λ or the change in energy of
the black hole is ∆E = hc3 /2M G. Thus the temperature of a black hole of mass M
hc3
is TBH = 2M G
. A proper calculation gives

h̄c3
TBH = .
8πM G

Using ∆E = ∆M c2 we have the change in mass of the black hole upon introducing
one bit of information is ∆M = hc/2M G. The change in radius of the black hole is

∆rS = 2 (∆M ) G/c2

( )
hc
=2 G/c2
2M G
h
=
Mc
14
It is interesting to note that photons of wavelength larger than the size of the black hole will
not enter, but will be reflected out. See Susskind’s book on black holes for this calculation

44
7 THE SCHWARZCHILD METRIC 7.1 Black holes

and the change in area of the black hole is ∆A = 8πrS ∆rS or

2M G h
∆A = 8π
c2 M c
Gh
= 16π 3 .
c
√
The length associated with this change in area is L = Gh/c3 , the Planck length.
This means that the entropy of a black hole of finite size is given by the area of the
horizon divided by the Planck length squared. A proper calculation getting all of the
factors of 2 and π correct gives
Ac3
SBH = . (50)
4Gh̄
This is one of the most remarkable equations in physics. Unlike the entropy defined
in classical statistical mechanics that varies with the system volume, the quantum
mechanical entropy of a black hole is proportional to the area of the horizon, not
the volume of the black hole. Note that the Beckenstein-Hawking entropy has h̄ in
the denominator. In classical physics h̄ → 0, so classically a black hole can have
infinite entropy. In fact, in classical physics, any amount of entropy can be stored in
a vanishingly small energy because h̄ → 0. It is the presence of h̄ in the denominator
that prevents the black hole from having an infinite entropy. Finally, it is important
to point out how large is the black hole entropy. It is vastly larger than anything of
comparable size that is not a black hole.
How long does it take for a black hole to radiate away its mass? Using the Stefan-
Boltzmann law we can calculate the luminosity

dE
− = AT 4
dt

and using A ∝ rS2 ∼ m2 , T ∝ h/λ ∼ 1/m, and E = m, gives −m2 dm/dt = 1 or

t ∼ m3 . A solar mass black hole takes approximately 1043 seconds to evaporate,
much longer than the age of the universe.

7.1.3 Hidden degrees of freedom

What is the mechanism by which a black hole has entropy? Putting together a set of
principles including quantum mechanics and thermodynamics the result is that there
are many hidden degrees of freedom that give the black hole its entropy. These degrees
of freedom are not apparent in the classical description of black holes or gravity. The
understanding of classical general relativity is in a sense like fluid dynamics. Fluid

45
7.2 Metric near the horizon 7 THE SCHWARZCHILD METRIC

dynamics describes the smooth flow of a uniform fluid. It can also describe turbulent
flow and so forth, but it does not take into account the microscopic structure of
fluids. Only in special cases is this considered. Ordinary fluid flow is described by
the Navier-Stokes equations, a set of equations that are the effective theory. One
does put entropy in the flow, but one does not ask where it came from. But the
simple fact that there is entropy in a fluid tells us that there is a hidden microscopic
structure; it is a hint that there is something more. The temperature and entropy we
have discussed so far with black holes is also a hint that classical Einstein gravity is
missing hidden degrees of freedom to be found in a proper quantum theory of gravity.

7.2 Metric near the Schwarzchild radius

The space near the surface of the horizon of a black hole is unremarkable. We
want to show now that the metric there is that of flat space by several coordinate
transformations of the Schwarzchild metric (49).
First we draw some inspiration by working with some simple metrics. Consider
2D flat Euclidean space with metric ds2 = dy 2 + dx2 . In polar coordinates x = ρ sin θ
and y = ρ cos θ, the metric becomes ds2 = dρ2 + ρ2 dθ2 . In Minkowski space we have
ds2 = −dT 2 + dx2 and here we need hyperbolic functions. Defining x = ρ cosh ω and
T = ρ sinh ω, we again have15

ds2 = −ρ2 dω 2 + dρ2 . (51)

The coordinate ω is the “time”. This is called the Rindler metric for reasons that will
become clear below.
Now let us write the Schwarzchild metric in terms of proper distance so that we
can make our analogy with flat Minkowski space. We write
( )
2M G 1
ds = − 1 −
2
dt2 + ( ) dr2 + r2 dΩ22 (52)
r 1− 2M G
r
15
In polar coordinates in flat space the right triangle with hypotenuse ρ drawn from the origin
has x, y coordinates as given here. Similarly for the Minkowski space one can draw the same right
triangle, but the “radius” ω draws out a hyperbola. Dropping the perpendicular to the x-axis, one
again has the sides of the right triangle as x = ρ cosh ω and T = ρ sinh ω.

46
7 THE SCHWARZCHILD METRIC 7.2 Metric near the horizon

and consider the case where r ≈ 2M G. Writing r = 2M G + δ we have

( ) ( )
2M G r − 2M G
1− =
r r
≈ δ/2M G

and we could expand around this, but let us instead do a variable change instead
keeping in mind that δ is a positive number; we are expanding outside the horizon.
Dropping the angular part of (52) we have
( )
2M G dr2
ds = − 1 −
2
dt2 + ( ).
r 1 − 2Mr G
1 rdr2
= − (r − 2M G)dt2 +
r r − 2M G
1 dr2
≈− (r − 2M G)dt2 + 2M G (53)
2M G r − 2M G
√ √
If we use dρ = 2M G dr/ r − 2M G for a variable change then the second term goes
to dρ2 and we are halfway there. Integrating we find
∫ √ ∫
dr′
r
dρ = 2M G√
r′ − 2M G
√ √
= 2M G · 2 r − 2M G

or
ρ2 = 8M G(r − 2M G).

Then (53) goes to

1
ds2 = − ρ2 dt2 + dρ2
16M 2 G2
= −ρ2 dω 2 + dρ2
√
where we have defined ω = t/ 16M 2 G2 . Note that this metric is the flat metric
(51). Therefore spacetime is approximately flat near the event horizon in a small
local frame. This means that nothing crazy happens at the horizon for an observer
falling through.

47
7.3 Penrose diagrams 7 THE SCHWARZCHILD METRIC

t
II

x
III I

IV

Figure 7: Rindler spacetime diagram. The horizon of a black hole is the 45-degree
line defining the light cone from the origin.

Figure 8: Rindler metric, near and far from the horizon.

7.2.1 Rindler space

Rindler was the first to recognize that the metric in (51) could be interpreted in
terms of a horizon in a simple diagram called a Rindler space diagram. In hyperbolic
coordinates, the constant ρ lines (red) that asymptotically approach the light cone are
the accelerated observers. They come in from minus infinity and go out to positive
infinity at a fixed acceleration. They turn around when they reach the x-axis. The
45-degree line, which is just a light-cone in the usual description of this diagram, is
called the acceleration horizon. This is because the accelerated observers shown in
quadrant I can never receive a light signal from anything behind that light cone. In
this sense it represents a black hole horizon.

7.3 Penrose diagrams

description of Penrose diagrams

7.4 Black hole formation

collapsing light cone

48
7 THE SCHWARZCHILD METRIC 7.5 Counting black hole microstates

7.5 Counting black hole microstates

The following is a simplified argument for obtaining the entropy of a black hole from
string theory. In units where c = h̄ = 1 we have, dimensionally, length = time and
energy = 1/time, respectively. In particular this means that energy ~ 1/length, which
we know from simple formulas like E = hc/λ for the energy of a photon of wavelength
λ. Newton’s constant in natural units is then [G] = [F r2 /M 2 ] = El/E 2 = l2 or units
of length squared. This is the Planck length, so we write G ∼ lP2 . In black hole
physics or GR physics, the Beckenstein entropy formula is

SBH ∼ A/lP2 = A/G, (54)

with A the area of the horizon. We would like to obtain this formula from stringy
arguments. If the string coupling constant is g and this constant gives the amplitude
for a string to “emit” a closed loop (graviton) for absorption by another string, then
the probability of a graviton exchange will be g 4 . The amplitude for the exchange
process is g 2 because there are to events. If this is related to gravity we should be
able to connect these ideas by writing g 2 ∼ G but g is dimensionless. The only thing
it can be related to with dimension is the characteristic string length, lS , which is the
length of the wiggles on a ball of string, not the size of the ball of string. Then we
can write
g 2 lS2 = lP2 (55)

or just glS = lP . This should make some sense in that the coupling constant should
be less than one, and we have g = lP /lS and lS > lP .
We now need the entropy of a string, or approximate entropy of a string. The
easiest way to approximate it is to consider a simple cubic lattice of points that make
up space. The string can be any configuration from one point on the lattice to another
and can even go back on itself. Then in one space dimension the string can move one
of two ways from each point. In two space dimensions, the string can move in any one
of four directions. Clearly in d dimensions the string can go in any of 2d directions. If
the string length is L = lS n, n links long, then the number of configurations is (2d)n .
This gives an entropy for the string of SS = ln (2d)n = n ln 2d. This is essentially
SS ∼ n and from the string length equation we have

SS ∼ n = L/lS . (56)

Note the similarity with the Beckenstein formula, the characteristic parameter for a

49
7.5 Counting black hole microstates 7 THE SCHWARZCHILD METRIC

black hole is the area of the horizon, while for a string it is the string length. Simply
divide out the characteristic lengths to obtain a dimensionless quantity and we arrive
at the entropy.
The entropy of the black hole can be written SBH = M 2 G2 /G = M 2 G using the
Schwarzchild radius RS ∼ M G. Can we write the string entropy in this way? The
mass of the string is MS = n/lS where the mass of a segment of the string is 1/lS .
Then we can write MS = L/lS2 , and the entropy of the string as SS = L/lS = lS MS .
Using (55) and G = lP2 we can write SS = ms MS G.

50