The light-cone approach in String theory

Sarthak Duary
International Centre for Theoretical Sciences,Tata Institute of Fundamental Research, Shivakote,
Bengaluru 560089, India.

E-mail: [email protected]
Contents

1 The Classical Relativistic String 1

1.1 The Relativistic Point Particle 2
1.2 A classical theory of relativistic string- the Polyakov action 4
1.2.1 Symmetries of the Polyakov action 7
1.2.2 Gauge fixing 8
1.2.3 Field equations and boundary conditions from gauge fixed Polyakov
action 9
1.2.4 Solutions to the equations of motion 11
1.2.5 Noether currents 14
1.2.6 The Virasoro Constraints 17
1.2.7 Mass formula for the string 18

2 The Quantum String 20

2.1 Light-cone Quantization (LCQ) 25
2.1.1 Open string spectrum 28
2.1.2 Lorentz Symmetry 32
2.1.3 Closed string spectrum 33

A Conserved currents associated with the Poincaré transformation 34

B Lorentz algebra 35

1 The Classical Relativistic String

First, we study General relativity in M dimensions+ scalar fields+ cosmological constant.

Now, the action for this theory is given by

M √
Z  
1 mn i j
S[g, φ] = d x −g R − g ∂m φ ∂n φ Mij − Λ , (1.1)
2

where Λ is Cosmological constant; m, n takes values from 1 to M ; i, j’s are internal space
indices which takes values from 1 to D. Also, Mij 6= f (φ) and g = det(gmn ).
The properties of this action are the following.

• It’s invariant under diffeomorphism or reparameterization. Diffeomorphism is not a

symmetry, it’s an invariance i.e. it’s a redundancy of the description.

–1–
 • This action has a global symmetry.

φi → φi + ai (translations),
(1.2)
φi → Λi j φj such that ΛT M Λ = M .

If M = 
ID×D i = 1,· · · D, global symmetry group is SO(D).
−1 0 ... 0
0 1 . . . 0
 
If M = 
 .. .. . . ..  , global symmetry group is SO(1,D-1) which is also called Lorentz group.

 . . . .
0 0 ... 1
Translations together with Lorentz group form Poincaré group. Now, if we choose
M ≡ diag(−1, 1, · · · , 1) ∗ , then the internal space becomes Minkowski space.
| {z }
D−1
Equation of motion is given by
δS
=0
δgmn
(1.3)
1
=⇒ Rmn − gmn R + Λgmn = Tmn .
2
Here, the stress tensor(energy-momentum tensor) of the matter field is given by
 1 
Tmn = Mij ∂m φi ∂n φj − gmn g rs ∂r φi ∂s φj . (1.4)
2
1.1 The Relativistic Point Particle
Consider M=1 (Particles): x0 → τ, Mij → ηµν , φi (τ ) → X µ (τ ).
g = −e2 , R = 0. Now, we also denote ∂τ X µ ≡ Ẋ µ .
The action is
Z
1 1 
S[e, X] = dτ Ẋ µ Ẋ ν ηµν − 2eΛ . (1.5)
2 e
Equation of motion is
δS
=0
δe
=⇒ −e−2 Ẋ µ Ẋ ν ηµν − 2Λ = 0 (1.6)

=⇒ −e−2 Ẋ 2 − 2Λ = 0 . (where Ẋ µ Ẋ ν ηµν ≡ Ẋ 2 .)

Field e dosen’t have a kinetic term. We should integrate out e. we find equation of motion,
solve for e and we plug it back into the action eq.(1.5).

∗
In QFT, we choose global symmetry group to be SO(D) but not SO(1,D-1). Because, zero component of
the field will have wrong sign for the kinetic term. In QFT, we see that we can’t change the sign of the kinetic
term otherwise hamiltonian will be unbounded from below.

–2–
 √
2 2
Equation of motion gives e2 = − Ẋ =⇒ e = √−Ẋ .
2Λ 2Λ
Plug it back into the action eq.(1.5) we get effective action
Z p
Sef f [X] = − dτ −2ΛẊ 2 . (1.7)

Taking non-relativistic limit taking ηµν = diag(−1, 1, · · · , 1), we get

| {z }
D−1

~ 2
 dt 2  dX
Ẋ 2 = − + .
dτ dτ
Action will be
√
s !
√ Z ~ 2
 dX Z √ 2Λ  dX~ 2
S = − 2Λ dt 1− = dt − 2Λ + + ··· . (1.8)
dt 2 dt
√
Now, if 2Λ = m then we can interpret the second term of the action given by eq.(1.8) as
kinetic term of a point particle.
√
Substituting 2Λ = m into eq.(1.7) we get
Z p
Sef f [x] = −m dτ −Ẋ 2 . (1.9)

This describes the action of a point particle propagating in flat spacetime. So, now we have
the worldline and X ’s are mapping the worldline into a spacetime.
√
Now, substituting 2Λ = m into eq.(1.5) we get
Z
1 1 
S[e, X] = dτ Ẋ µ Ẋ ν ηµν − m2 e . (1.10)
2 e

From diffeomorphism invariance we know that the action under τ 0 = τ 0 (τ ). We can use this
to gauge fix e = 1.
Equation of motion becomes

− Ẋ 2 − m2 = 0 (constraint). (1.11)
∂L
Now, the momentum conjugate to X µ coordinate is given by Pµ = ∂ Ẋ µ
= Ẋµ .
This gives
P 2 + m2 = 0 . (1.12)
This is the dispersion relation for a free point particle with mass m.
Consider M=2 (Strings): Two coordinates (x0 , x1 ) → (τ, σ), scalar fields are φi → X µ (τ, σ),
Mij → ηµν and gmn → hmn . The action for this theory in M = 2 is given by
√
Z  
1 mn µ ν
S[h, X] = dσdτ −h R − h ∂m X ∂n X ηµν − Λ , (1.13)
2

–3–
where Λ is Cosmological constant; m, n takes values from 1, 2 ; µ, ν’s are internal space indices
which takes values from 1 to D. We take ηµν ≡ diag(−1, 1, · · · , 1).
| {z }
D−1
In two dimensions Riemann curvature tensor is given by
R
Rmnpq = (hmp hqn − hmq hpn ) . (1.14)
2
R
From this we get, Rnq = 2 gnq . The Einstein tensor vanishes i.e.

1
Rmn − gmn R = 0 . (1.15)
2
Now, substituting it in eq.(1.3) we get,

Λhmn = Tmn . (1.16)

The stress tensor(energy-momentum tensor) of the matter field is given by

 1 
Tmn = ηµν ∂m X µ ∂n X ν − hmn hpq ∂p X µ ∂q X ν . (1.17)
2
Now, contracting with hmn gives trace of Tmn which is

Tm m = hmn Tmn = 0.

So, this is a traceless energy-momentum tensor.

Now, we compute trace from eq.(1.16) we find Λ = 0. So, the only cosmological constant that
is consistent with equations of motion is Λ = 0.
Now, the first term from eq.(1.13) in two dimension gives
√
Z
1
dτ dσ −hR = χ = 2 − 2h .
4π Σ

This action only depend on the topological properties of the worldsheet Σ, with χ is known
as the Euler-characteristic of Σ (h is the number of genus). 2D gravity is dynamically trivial,
gives no dynamics. χ is just a number.
Therefore, we get
√
Z  
1 mn
S[hmn , X ] = d σ −h − h ∂m X ∂n X ηµν . ( where, τ = σ 0 , σ = σ 1 .) (1.18)
µ 2 µ ν
2

1.2 A classical theory of relativistic string- the Polyakov action

Now, we’ll introduce the Polyakov action which is also known as the string sigma-model
action. The action is expressed in terms of an auxiliary world- sheet metric hαβ (σ, τ ), which
plays a role analogous to the auxiliary field e introduced for the point particle. We will use

–4–
the notation hαβ for the world-sheet metric, whereas ηµν denotes a space-time metric. Also,
h = dethαβ and hαβ = (h−1 )αβ . In this notation the Polyakov action becomes
√
Z
T
SP [hαβ , X µ ] = − d2 σ −h(hαβ ∂α X µ ∂β X ν ηµν ) , (1.19)
2
where σ 0 = τ, σ 1 = σ. Here, µ, ν are internal space indices i.e. target space indices. Minkowski
metric ηµν ≡ diag(−1, 1, · · · , 1). Also, T = the tension = 2πα 1 0 2 µ
0 , [α ] =(length) . Here, X
| {z }
D−1
are scalar fields from the point of view of two-dimensional world-sheet R1,1 . The equation of
motion implies the vanishing of the world-sheet energy-momentum tensor Tαβ ,i.e.,
2 1 δSP
Tαβ = − √ =0 . (1.20)
T −h δhαβ
The Polyakov action is
√
Z
µ T
SP [hαβ , X ] = − d2 σ −h(hαβ ∂α X µ ∂β X ν ηµν ) . (1.21)
2
To evaluate the variation of the action, we use
√ 1√
δ −h = − −hhαβ δhαβ .
2
After taking the variation of the action, the result for the energy-momentum tensor takes the
form
1
Tαβ = ∂α X.∂β X − hαβ hργ ∂ρ X.∂γ X = 0 . (1.22)
2
Here, we introduce short-hand notation for the scalar products which are defined in the case
of a flat space-time by A.B = ηµν Aµ B ν . This is the equation of motion for hαβ . The equation
of motion for Xµ , obtained from the Euler-Lagrange equation is given by
∂LP T√
µ
=− −h2hαβ ∂β X ν ηµν , (1.23)
∂(∂α X ) 2
∂LP
∂α =0 , (1.24)
∂(∂α X µ )
√
∂α ( −hhαβ ∂β Xµ ) = 0 . (1.25)
Now the expression for the Laplacian for a general metric hαβ is
1 √
∇2 Xµ = √ ∂α ( −hhαβ ∂β Xµ ) . (1.26)
−h
Thus we see that our Euler-Lagrange equation is just telling us that ∇2 X µ = 0.
Now, the equation of motion for hαβ can be used to eliminate hαβ from the Polyakov action.
We take the § square root of minus the determinant of both sides of the eq.(1.44)
1
∂α X.∂β X = hαβ hργ ∂ρ X.∂γ X .
2
§
Because, we are in lorentzian signature, so we take minus sign.

–5–
This gives
q 1√
−det(∂α X.∂β X) = −hhργ ∂ρ X.∂γ X
2 (1.27)
√ 1
= −h hργ ∂ρ X.∂γ X .
p
=⇒ −det Gαβ
2
Here, we introduce induced metric on the world-sheet Gαβ = ∂α X µ ∂β X ν ηµν .
After eliminating hαβ we get the effective action which is called Nambu-Goto action given by
Z
SN G [X] = −T d2 σ −detGαβ .
p
(1.28)

Therefore the Polyakov action is the same as the Nambu-Goto action once we plug in the
equation of motion for the compensator (world-sheet metric)

SP |EOM = SN G .

Now, we restrict the σ domain to be from 0 to π. The image of the strip from τ − σ plane to
spacetime is no-longer be infinite. In fig.2 we see that the world-sheet for the free propagation
of an open string describes a rectangular surface and the free propagation of a closed string
sweeps out a cylinder.

Figure 1: The functions X µ (σ, τ ) describe the embedding of the string world-sheet in space-
time.

–6–
Figure 2: The world-sheet for the free propagation of an open string describes a rectangular
surface, while the free propagation of a closed string sweeps out a cylinder.

The integral appearing in the action given by eq.(1.28) describes the area of the world-
sheet. As a result, the classical string motion extremises the world-sheet area, just like classical
particle motion makes the length of the world line extremal by moving along a geodesic.

1.2.1 Symmetries of the Polyakov action

• Diffeomorphisms or reparameterization invariance: This is a gauge symmetry on the
worldsheet. We redefine the worldsheet coordinates as σ α → σ̃ α (σ, τ ). The fields X µ
transform as world-sheet scalars, while gαβ transforms in the manner appropriate for a
2d metric.
∂σ γ ∂σ δ
X µ (σ, τ ) → X̃ µ (σ̃) = X µ (σ), hαβ (σ) → h̃αβ (σ̃) = hγδ (σ) . (1.29)
∂ σ̃ α ∂ σ̃ β

• Poincaré invariance: This is a global symmetry on the worldsheet. Under this,

δX µ = aµν X ν + bµ . (1.30)

• Weyl Invariance: Under this symmetry, δX µ = 0, but the metric changes as

hαβ → eφ(σ,τ ) hαβ . (1.31)

Since,
√ √
−h → eφ −h , hαβ → e−φ hαβ (1.32)
cancels each other therefore the Polyakov action has weyl invariance. This local sym-
metry is the reason that the energy-momentum tensor is trace-less. This can be easily
seen from the defination of the energy-momentum tensor.
1
Tαβ = ∂α X.∂β X − hαβ hργ ∂ρ X.∂γ X (1.33)
2
δhαβ = φhαβ (1.34)

–7–
 Z
δS 2 1 δS
δS = δhαβ , using T αβ = − √
δhαβ T −h δhαβ
√
Z
T
=− d2 σ −hδhαβ T αβ
2 (1.35)
√
Z
T
=− d2 σ −h φ hαβ T αβ
2
=0
since there is no variation in S under a Weyl transformation. Therefore, the energy-
momentum tensor is trace-less

1.2.2 Gauge fixing

We would like to use our symmetries to even further simplify our Polyakov action before we
quantize, i.e. we would like to gauge fix. Poincaré transformations are global symmetries,
whereas diffeomorphisms and Weyl transformations are local symmetries. The local symme-
tries can be used to choose a gauge. We have to be a bit careful since we must remember to
impose the equations of motion (the EOM before gauge fixing) as constraints to our gauge-
fixed theory. Firstly, we have two reparameterizations to play with. We know that in two
dimensions the world-sheet metric hαβ which is a symmetric tensor has three independent
components. This means that we expect to be able to set any two of the metric components
to a value of our choosing. Diffeomorphisms give us the freedom to set two functions to
anything we like in the metric. We will choose to make the metric locally conformally flat,
meaning
hαβ = eφ ηαβ , (1.36)
where φ(σ, τ ) is some function on the worldsheet. Choosing a metric of the form (1.36) is
known as conformal gauge. We have only used diffeomorphism invariance to get to the metric
(1.36). We still have Weyl transformations. Clearly, we can use these to remove the last
independent component of the metric and set φ = 0 such that,

hαβ = ηαβ . (1.37)

We end up with the flat metric on the worldsheet in Minkowski coordinates. Therefore, using
diffeomorphisms and weyl we should be able to gauge fix the metric ¶ to be hαβ = ηαβ .
On a two dimensional worldsheet the local diffeomorphism and Weyl invariance can be used
to fix the metric hαβ by locally gauging away all its parameters. The possibility can be
anticipated already by counting degrees of freedom. For a D−dimensional worldsheet this
¶
Actually such a flat world-sheet metric is only possible if there is no topological obstruction. This is the
case when the world sheet has vanishing Euler characteristic. Examples include cylinder, torus etc. In two
dimensions, the Euler characteristic of a manifold is proportional to the integral of the Ricci curvature over
the manifold, we see that being able to extend the locally flat metric requires for the Euler characteristic of
the worldsheet to vanish.

–8–
counting goes as follows:
1
hαβ : D(D + 1) d.o.f.
2 (1.38)
diffeomorphism+ weyl : D + 1 d.o.f.

Therefore, 21 D(D + 1) − (D + 1) d.o.f. can’t be gauged away locally. Thus we see that if
D = 2 we can locally gauge away all parameters.
Now, using the choice of coordinates for the worldsheet as (τ, σ) we take
" #
−1 0
hαβ = . (1.39)
0 1

With this choice, h = dethαβ = −1. We have

hαβ ∂α X.∂β X = −∂τ X.∂τ X + ∂σ X.∂σ X

(1.40)
= −Ẋ 2 + X 02 ,
µ µ µ ν µ ∂X ν
where, ∂X µ µ ∂X µ 0µ 2 ∂X ∂X 02 ∂X
∂τ = ∂τ X = Ẋ , ∂σ = ∂σ X = X and Ẋ = ∂τ ∂τ ηµν , X = ∂σ ∂σ ηµν .
We write the Polyakov action in the remarkably compact form after gauge fixing
Z
T
SP GF [X] = d2 σ(Ẋ 2 − X 02 ) . (1.41)
2
1.2.3 Field equations and boundary conditions from gauge fixed Polyakov action
Z
T
SP GF [X] = d2 σ(Ẋ 2 − X 02 )
2
Z
T
=⇒ δSP GF [X] = d2 σ(2Ẋ.δ Ẋ − 2X 0 .δX 0 )
2 (1.42)
" #
..
Z
∂   ∂
= T d2 σ (X 00 − X).δX + X 0 .δX


Ẋ.δX − .
| {z } |∂τ {z ∂σ
1st term
}
Boundary terms

Since δX µmust be zero for any generic variation For every scalar field that we’ve on the
world-sheet
.. µ  ∂ ∂  µ
X 00µ − X = 0 =⇒ − X (σ, τ ) = 0 (1.43)
∂σ 2 ∂τ 2
must be true. Eq.(1.43) can also be obtained from eq.(1.25) after gauge fixing hαβ = ηαβ .
Since the metric on the world-sheet has been gauge fixed, the vanishing of the energy-
momentum tensor, that is, Tαβ = 0 originating from the equation of motion of the world-sheet
metric, must now be imposed as an additional constraint condition. Constraint condition gives
1
Tαβ = ∂α X.∂β X − hαβ hργ ∂ρ X.∂γ X = 0 . (1.44)
2
In the gauge hαβ = ηαβ the components of this tensor are
1
T01 = T10 = Ẋ · X 0 = 0, T00 = T11 = (Ẋ 2 + X 0 2 ) = 0 . (1.45)
2

–9–
These are called the Virasoro constraints. We see the vanishing of the trace of the energy-
momentum tensor
Tr T = η αβ Tαβ = T11 − T00 = 0 .
This is a consequence of Weyl invariance that we have seen earlier in eq.(1.35). We therefore
learn that the equations of motion of the string are the free wave equations eq.(1.43) subject
to the two constraints eq.(1.45) arising from the equation of motion T αβ = 0.
Now, we specify boundary conditions. A string can be either closed or open. We choose the
coordinate σ to have the range 0 ≤ σ ≤ π. The stationary points of the action are determined
by demanding invariance of the action under the shifts

X µ → X µ + δX µ . (1.46)

The boundary term

Z π Z ∞
∂  
T dσ dτ Ẋµ δX µ = 0. (by assumption.) (1.47)
0 −∞ ∂τ

We assume that this variations are actually dying off in the far past and far future. But in the
spatial boundary we’ve to be careful because our spatial direction is compact. So, we can’t
do the same thing that we usually do in QFT where we assume that the variations vanish at
infinity. Spatial boundary term
Z ∞ Z π Z ∞ h
∂ i
Xµ0 δX µ = −T dτ Xµ0 δX µ |σ=π − Xµ0 δX µ |σ=0


−T dτ dσ (1.48)
−∞ 0 ∂σ −∞

must vanish. There are several different ways in which this can be achieved. For an open
string these possibilities are illustrated in Fig.3.

Figure 3: Illustration of Dirichlet (left) and Neumann (right) boundary conditions. The
solid and dashed lines represent string positions at two different times.

• Closed string: In this case the embedding functions are periodic i.e.

X µ (τ, π) = X µ (τ, 0) . (1.49)

– 10 –
 • Open string with Neumann boundary conditions: We have

X 0µ (τ, π) = 0 , X 0µ (τ, 0) = 0 . (1.50)

• Open string with Dirichlet boundary conditions: In this case the positions of the two
string ends are fixed so that δX µ = 0, we have

X µ (τ, π) = Xπµ , X µ (τ, 0) = X0µ . (1.51)

1.2.4 Solutions to the equations of motion

To find the solution to the equations of motion and constraint equations we introduce world-
sheet light-cone coordinates, defined as

σ± = τ ± σ . (1.52)

In these coordinates the two-dimensional worldsheet metric take the form

ds2 = −dτ 2 + dσ 2 = −dσ + dσ − , (1.53)

where, " # " #

η++ η+− 1 01
=− . (1.54)
η−+ η−− 2 10
The derivatives take the form
1
∂± = (∂τ ± ∂σ )
2 (1.55)
∂τ = ∂+ + ∂− , ∂σ = ∂+ − ∂− ,

where we use
∂
= ∂± .
∂σ ±
In light-cone coordinates the field equations for X µ given by eq.(1.43) becomes

∂+ ∂− X µ (σ + , σ − ) = 0 . (1.56)

The general solution of the wave equation eq.(1.56) is given by

µ −
X µ (σ + , σ − ) = XR (σ ) + XLµ (σ + ) , (1.57)

which is a sum of right-movers (left-moving components) and left-movers (right-moving com-

ponents). Now, of course we can’t solve these equations unless we know the boundary condi-
tions. Now, we choose some boundary conditions. First, we want to solve eq.(1.56) assuming
closed string boundary conditions

X µ (τ, σ) = X µ (τ, σ + π) (periodic). (1.58)

– 11 –
First, we note that
µ
∂− X µ = ∂− XR
(1.59)
∂+ X µ = ∂+ XLµ .
µ
Now, the L.H.S of eq.(1.59) are periodic in σ, therefore ∂− XR and ∂+ XLµ are also periodic in
σ. So, we expand the derivatives in fourier modes as
∞
µ −
X
µ −2imσ
∂− XR = ls αm e
m=−∞
∞ (1.60)
µ −2imσ +
∂+ XLµ
X
= ls α̃m e .
m=−∞

µ
Here, XR and XLµ have dimensions of length in spacetime. It’s convenient to have fourier
µ µ
modes αm , α̃m to be dimensionless. Also, the world-sheet coordinates are taken to be dimen-
sionless . We can therefore associate a length scale, ls , which is related to the string tension
T and the open-string Regge slope parameter α0 by

ls2 1
α0 = , T =
2 2πα0
The string length scale ls is the natural length that appears in string theory. Integrating
eq.(1.60) we get
µ
µ − ils X αm −
XR (σ ) = xµR + ls α0µ σ − + e−2imσ ,
2 m
m6=0
µ (1.61)
ils X α̃m +
XLµ (σ + ) = xµL + ls α̃0µ σ + + e−2imσ .
2 m
m6=0

Here, xµR and xµL are constants. Now,

ils X 1 µ −2imσ−
X µ (σ + , σ − ) = (xR + xL )µ + ls (αµ0 σ − + α̃µ0 σ + ) + + α̃µm e−2imσ+ .


α me
| {z } 2 m
m6=0
not periodic in σ
(1.62)

We want X µ to be periodic in σ, but the second term gives

ls (αµ0 σ − + α̃µ0 σ + ) = τ (αµ0 + α̃µ0 ) + σ(−αµ0 + α̃µ0 ) , (1.63)

which is not periodic in σ. To make X µ periodic in σ we set coefficient of σ to be zero which

gives
αµ0 = α̃µ0 .

– 12 –
Now, we get
ils X 1 µ −2imσ− +

X µ (σ + , σ − ) = (xR + xL )µ + 2ls αµ0 τ + α me + α̃µm e−2imσ
2 m
m6=0
(1.64)
ils X 1 µ −2imσ− +

= xµ + 2ls αµ0 τ + α me + α̃µm e−2imσ .
2 m
m6=0

The requirement that X µ ’s are real functions implies that xµR , xµR and α0µ are real, while
positive and negative fourier modes are conjugate to each other

(αµm )? = αµ−m , (α̃µm )? = α̃µ−m , (1.65)

where ? represents complex conjugate.

Now, if we solve field equations for open string with Neumann boundary conditions we get,
X αµm
X µ (τ, σ) = xµ + ls α0µ τ + ils e−imτ cos mσ . (1.66)
m
m6=0

Now, we’ll prove this result given by eq.(1.66).

The open string with Neumann boundary conditions say that normal derivative of X µ must
vanish at the string boundary which is

X 0µ = 0 , for σ = 0 and σ = π . (1.67)

These are free boundary conditions which prevent from momentum from flowing off the ends
of the string. The general solution can be written as an expansion of Fourier modes. Here,
we denote these modes as αµm , α̃µm and write the right-moving and left-moving components
as
µ
µ − ls µ − ils X αm −
XR (σ ) = xµR + α0 σ + e−imσ ,
2 2 m
m6=0
µ (1.68)
ls ils X α̃m +
XLµ (σ + ) = xµL + α̃0µ σ + + e−imσ .
2 2 m
m6=0

This gives

0µ − ls ls X µ −im(τ −σ)
XR (σ ) = − α0µ − αm e ,
2 2
m6=0
(1.69)
ls ls X µ −im(τ +σ)
XL0µ (σ + ) = α̃0µ + α̃m e .
2 2
m6=0

Now, summing this two equations given by eq.(1.69) and setting

X 0µ = 0 , for σ = 0 , (1.70)

– 13 –
We get,
ls µ ls X µ
(α̃0 − α0µ ) + (α̃ m − αµm )e−2imτ = 0 ,
2 2 (1.71)
m6=0

α̃0µ = α0µ , µ
α̃m = µ
αm .

Let’s consider the other boundary condition, which is imposed at the other end of the string.
Summing this two equations given by eq.(1.69) and setting

X 0µ = 0 , for σ = π , (1.72)

We get,
X
ls αµm e−imτ sin mπ = 0 , (1.73)
m6=0

This can only be true if sin mπ = 0, which means that m must be an integer.
µ
Substituting, α̃0µ = α0µ , α̃m µ
= αm gives
X αµm
X µ (τ, σ) = xµ + ls α0µ τ + ils e−imτ cos mσ . (1.74)
m
m6=0

So, we obtain classical solution for both closed string in eq.(1.64) and open string for Neumann
boundary condition in eq.(1.74). Now, these classical solution is for any number of classical
fields X µ i.e. for any space-time dimensions. To begin the quantization of the string, we have
to study the conserved charges of the system and for that we have to use Noether’s theorem
which we discuss in next section sec.1.2.5.

1.2.5 Noether currents

For every global symmetry we can produce a conserved current jα , where α is a vector index.
Since we’re dealing with a field theory in two dimensions, the current has a vector index on
the world-sheet.

Global Symmetry → Conserved current jα i.e. ∂ α jα = 0 is satisfied (α = 0, 1) . (1.75)

Once we have a conserved current, we can produce a conserved charge which is defined by
integrating the zeroth component of the conserved current over all the space directions. In
this case we are going to get an integral over σ given by
Z
Q = dσj0 . (1.76)

We’ve to analyze carefully that Q is conserved or not because the space is compact.
For Polyakov action the global symmetries are Poincaré invariance. The Poincaré transfor-
mations
δX µ = aµν X ν + bµ . (1.77)

– 14 –
are global symmetries of the string world-sheet theory. Therefore, they give rise to conserved
Noether currents. The conserved currents associated with the Poincaré transformation of X µ
are
Pαµ = T ∂α X µ ,
(1.78)
J µν µ ν ν µ
α = T (X ∂α X − X ∂α X ) .

Detail workout of conserved currents are computed in appendix section A. Noether charges
corresponding to conserved current is given by
Z π Z π
Pµ = dσT X˙ µ = T dσ X˙ µ . (1.79)
0 0

Now, taking derivative with respect to τ

Z π .. µ
∂ µ
P =T dσ X . (1.80)
∂τ 0

For every scalar field that we’ve on the world-sheet from equation of motion we have
.. µ
X 00µ = X . (1.81)

Therefore we have,
Z π
∂ µ
P =T dσX 00µ
∂τ
Z0 π
∂X 0µ (1.82)
=T dσ
0 ∂σ
= T (X 0µ |σ=π − X 0µ |σ=0 ) .

This is zero for closed string because X µ is periodic in σ. For closed string the two contri-
butions cancel and we get zero. For open string X 0µ |σ=π − X 0µ |σ=0 vanishes only when we
have Neumann boundary condition on both sides. For open string with Dirichlet boundary
conditions P µ is not conserved. Because the string will attach to D-brane which will provide
momentum or take momentum from the string.
Now, we’ll compute total conserved momentum of a closed string by integrating the momen-
tum current over σ at τ = 0. Closed string mode expansion is given by
ils X 1 µ −2imσ− +

X µ (σ + , σ − ) = xµ + 2ls αµ0 τ + α me + α̃µm e−2imσ . (1.83)
2 m
m6=0

Therefore, the total momentum of a closed string is given by

2α0µ
Z π
dX µ (σ)
Pµ = T dσ = 2ls T α0µ π = , (1.84)
0 dτ ls

where, in the last step we use T = 1

2πα0 , ls2 = 2α0 .
2αµ
Now, we define ls
0
≡ pµ , where pµ is the momentum of the zero mode for closed string.

– 15 –
So, the total momentum of the closed string is the same as the momentum of the zero mode,
all the oscillators doesn’t contribute to this because they integrate to zero when we go from
0 to π. Centre of mass of the string xµ also don’t contribute as it doesn’t depend on τ . From
here we get a nice interpretation for zero mode α0µ which is
Pµ 1
α0µ = = ls pµ . (1.85)
2ls T π 2
Therefore, the closed string mode expansion in terms of pµ is given by
ils X 1 µ −2imσ− +

X µ (σ + , σ − ) = xµ + ls2 pµ τ + α me + α̃µm e−2imσ . (1.86)
2 m
m6=0

For open string with Neumann boundary conditions we have,

X αµm
µ
X (τ, σ) = x + µ
ls α0µ τ + ils e−imτ cos mσ . (1.87)
m
m6=0

Therefore, the total momentum of an open string is given by

αµ
Z π
dX µ (σ)
µ
P =T dσ = ls T α0µ π = 0 , (1.88)
0 dτ ls
where, in the last step we use T = 1
2πα0 , ls2 = 2α0 .
αµ
We define 0
ls≡ pµ , where pµ is the momentum of the zero mode for open string. Therefore,
the zero mode α0µ is
α0µ = ls pµ . (1.89)
Therefore, the open string mode expansion in terms of pµ is given by
X αµm
X µ (τ, σ) = xµ + ls2 pµ τ + ils e−imτ cos mσ . (1.90)
m
m6=0

Now, for Lorentz transformation in the target space

δX µ = aµν X ν . (1.91)

the conserved current is given by

J µν µ ν ν µ
α = T (X ∂α X − X ∂α X ) . (1.92)

Consrved charge i.e. the total angular momentum is given by

Z π
µν
J =T dσ(X µ X˙ ν − X ν X˙ µ ) . (1.93)
0

The mode expansion for angular momentum generators J µν of an open bosonic string is given
by
∞
X 1 µ ν
J µν = xµ pν − xν pµ − i (α α − α−nν
αnµ ) ≡ lµν + S µν . (1.94)
n −n n
n=1

– 16 –
Using the mode expansion for closed string X µ , the angular momentum can be written as
∞ ∞
X 1 ν µ µ
X 1 ν µ µ
J µν = (xµ pν − xν pµ ) − i αnν − i α̃nν



α−n αn − α−n α̃−n α̃n − α̃−n
n n (1.95)
n=1 n=1
µν µν µν
≡l +S + S̃ ,
The first term, lµν , is the orbital angular momentum of the string while the remaining pieces
S µν and S̃ µν is the angular momentum due to excited oscillator modes.

1.2.6 The Virasoro Constraints

We found all the solutions of the equation of motion consistent with the boundary conditions.
By gauge fixing the world-sheet metric we got the action given by
Z
T
SP GF [X] = d2 σ(Ẋ 2 − X 02 ) . (1.96)
2
In order to have the full physical picture we have to impose constraints coming from the
vanishing energy-momentum tensor. In the gauge hαβ = ηαβ the components of this tensor
are
1
T01 = T10 = Ẋ · X 0 = 0, T00 = T11 = (Ẋ 2 + X 0 2 ) = 0 . (1.97)
2
These are the Virasoro constraints.
These gives
Ẋ 2 ± 2Ẋ · X 0 + X 02 = 0
=⇒ (Ẋ ± X 0 )2 = 0
(1.98)
=⇒ (∂+ X)2 = 0 and (∂− X)2 = 0
=⇒ (∂+ XL )2 = 0 and (∂− XR )2 = 0 .
We have
T++ = ∂+ X µ ∂+ Xµ = 0
T−− = ∂− X µ ∂− Xµ = 0 (1.99)
T+− = T−+ = 0 .
We choose closed string boundary condition

X µ (τ, σ + π) = X µ (τ, σ) (periodic in σ.) (1.100)

Now, (∂+ XL )2 and (∂− XR )2 are also periodic in σ.

So, we decompose them in fourier modes
∞
X
(∂+ XL ) =2
2ls2 L̃m e−2imσ+ ,
m=−∞
∞ (1.101)
X
(∂− XR )2 = 2ls2 Lm e−2imσ− .
m=−∞

– 17 –
Here, we choose modes L̃m and Lm are dimensionless. As (∂+ XL )2 and (∂− XR )2 carries
dimension of (length)2 , so we put ls2 . Now, we want to compute L̃n as a function of α̃m .

1.We first set τ = 0

Z π
(1.102)
2. dσ(∂+ XL )2 |τ =0 e2inσ .
0
Now,
Z π Z π
2 2inσ
dσ(∂+ XL ) |τ =0 e = 2ls2 L̃n dσ
0 0 (1.103)
= 2ls2 L̃n π .

Therefore,
Z π
1
L̃n = 2 dσ(∂+ XL )2 |τ =0 e2inσ
2ls π 0
∞ ∞ (1.104)
1 X µ ν 1 X
=⇒ L̃n = α̃m α̃n−m ηµν = α̃ .α̃ .
2 m=−∞ 2 m=−∞ m n−m

where, we use mode expansion using eq.(1.60) in the last step.

Similarly we get
∞
1 X
Ln = αm .αn−m . (1.105)
2 m=−∞

1.2.7 Mass formula for the string

Constraints: L˜n = 0, Ln = 0, ∀n .
Now, let’s look at the n = 0 constraint. This contain α0 . From earlier computation we
learned that α0 is related to the momentum in space-time of the string. That should give us
some information about the mass of the string. The classical constraint L0 = L̃0 = 0 can be
used to derive an expression for the mass of a string.
Now,
1 1X
L̃0 = α̃02 + α̃m .α̃−m
2 2
m6=0
1 X
= α̃02 + α̃m .α̃−m (1.106)
2
m>0
1 X
= α02 + α̃m .α̃−m .
2
m>0

Also we have α0µ which is

Pµ
α0µ = . (1.107)
2ls T π

– 18 –
This gives
2
P2 −Mclosed
α02 = = (relativistic mass shell condition M 2 = −Pµ P µ .) (1.108)
(2ls T π)2 (2ls T π)2
Now, L̃0 = 0 gives
2
Mclosed X
= α̃m .α̃−m . (1.109)
8π 2 ls2 T 2
m>0
Now,
1 X
T = 0
, ls2 = 2α0 =⇒ α0 Mclosed
2
=4 α̃m .α̃−m . (1.110)
2πα
m>0
For L0 = 0 we get X
α0 Mclosed
2
=4 αm .α−m . (1.111)
m>0
In symmetric form we have
2 2 X
Mclosed = (αm .α−m + α̃m .α̃−m ) . (1.112)
α0
m>0

For open string

1 X
L0 = α02 + αm .α−m . (1.113)
2
m>0

α0µ is related to total momentum of the open string P µ as

α0µ = P µ ls . (1.114)

Therefore for open string we get

2 1 X
Mopen = αm .α−m . (1.115)
α0
m>0

Hamiltonian

World-sheet time evolution is generated by the Hamiltonian

Z π  T Z π
µ
H= Ẋµ P0 − L = (Ẋ 2 + X 02 )dσ , (1.116)
0 2 0

where
∂S
P0µ = = T Ẋ µ . (1.117)
∂ Ẋµ
This gives
∞
1 X
H= α−n .αn (open strings)
2 n=−∞
∞ (1.118)
1 X
H= (α−n .αn + α̃−n .α̃n ) (closed strings) .
2 n=−∞

– 19 –
2 The Quantum String

Now, we will move on to the quantization. For strings, we have a Quantum Field Theory
in two dimensions. We consider our approaches for quantizing this classical theory in the
canonical formalism.

1. Old covariant quantization or Gupta-Bleuler method of quantization :

In this method we first quantize the system and then impose the constraints that arise
from gauge fixing as operator equations on the physical states of the system. This is
based on a description in terms of the X µ coordinates only with restrictions on the phys-
ical Fock space corresponding to the Virasoro constraint conditions. These restrictions
are similar to the Gupta-Bleuer condition in QED, in which the classical coulomb gauge
constraint ∂µ Aµ is replaced by the requirement that the positive-frequency components
of the corresponding quantum operator should annihilate physical photon states. The
nice feature of this approach is that we never have to mention gauge choices when
quantizing and so it is manifestly covariant.

2. Light-cone quantization: Alternately, before we quantize we can use our symmetries

immediately to pick a particular gauge that solves the constraints. After solving the
constraints of the system we determine the space of physically distinct classical solutions.
We then quantize these physical solutions. This is manifestly unitary but also not
covariant.

3. BRST quantization or modern covariant quantization : It involves the intro-

duction of Faddeev-Popov ghosts, BRST symmetries etc.

Now, we review the old covariant quantization. We have D free scalar fields X µ whose
dynamics is governed by the action
Z
T
SP GF [X] = d2 σ(Ẋ 2 − X 02 ) . (2.1)
2
The momentum conjugate of the field X µ is given by
δL
P µ (τ, σ) = . (2.2)
δ Ẋµ

From lagrangian
T
L= (Ẋ 2 − X 02 ) , we get P µ = T X˙ µ . (2.3)
2

– 20 –
In classical field theory the natual object to study is Poisson brackets at equal-times but
different positions given by

{X µ (τ, σ), X ν (τ, σ 0 )}P.B. = 0 ,

δXµ (τ, σ)
{P µ (τ, σ), X ν (τ, σ 0 )}P.B. = η µν δ(σ − σ 0 ) , where = δ µν δ(σ − σ 0 ) ,
δXν (τ, σ 0 )
(2.4)
1
=⇒ {Ẋ (τ, σ), X (τ, σ )}P.B. = η µν δ(σ − σ 0 ) ,
µ ν 0
T
{P µ (τ, σ), P ν (τ, σ 0 )}P.B. = 0 .

For periodic variables X µ , P µ we use periodic version of the delta function given by
∞
0 1 X 2im(σ−σ0 )
0
δ(σ − σ ) → δP (σ − σ ) = e . (2.5)
π m=−∞

Plugging in the mode expansion we translate these into commutation relations for the Fourier
modes
µ
{αm , αnν }P.B. = imδm+n,0 η µν ,
µ
{α̃m , α̃nν }P.B. = imδm+n,0 η µν , (2.6)
µ
{αm , α̃nν }P.B. = 0 .

Using these relation we get

{Lm , Ln }P.B. = i(m − n)Lm+n . (2.7)

This is the algebra that Virasoro generators should satisfy in a classical conformal field theory.
This is called Witt algebra or classical Virasoro algebra.
Now, we’ll quantize the theory by replacing the Poisson brackets by commutator brackets via
the substitution
{·, ·}P.B. → −i[·, ·] . (2.8)
Now, we treat X µ as a quantum operator with eq.(2.4) replaced by canonical commutation
relations at equal τ .

[X µ (τ, σ), X ν (τ, σ 0 )] = 0

[P µ (τ, σ), X ν (τ, σ 0 )] = −iη µν δ(σ − σ 0 ) (2.9)
µ ν 0
[P (τ, σ), P (τ, σ )] = 0 .
µ µ
Now αm , α̃m and xµ , pν have commutation relations given by

[xµ , pν ] = iη µν
µ
[αm , αnν ] = mδm+n,0 η µν
µ
(2.10)
[αm , α̃nν ] = 0
µ
[α̃m , α̃nν ] = mδm+n,0 η µν .

– 21 –
 µ µ
Here, αm and α̃m look painfully close to harmonic oscillator creation and annihilation opera-
µ µ
tors, but not quite. The commutation relations of αm and α̃m are those of harmonic oscillator
creation and annihilation operators in disguise. The normalization is slightly different, here
extra m factor appears in the R.H.S.
We define
1 µ 1 µ
aµm = √ αm , ãµm = √ α̃m . (2.11)
m m
Reality condtion of X µ gives
µ
α−m µ†
= αm , aµ−m = aµ†
m (m > 0) . (2.12)

aµm and ãµm have commutation relations given by

[aµm , a†ν
n ] = δm,n η
µν
(2.13)
[ãµm , ã†ν
n ] = δm,n η
µν
.

So each scalar field X µ gives rise to two infinite towers of creation and annihilation operators
arising from right-moving modes αm and left-moving modes α̃m . With these commutation
relations we build the Hilbert space of our theory. We introduce a vacuum state of the string
|0i, defined by
µ µ
αm |0i = α̃m |0i = 0 for m > 0 (2.14)
If we work in momentum space, the vacuum carries another quantum number pµ , which is
the eigenvalue of the momentum operator. We write the vacuum as |0; pi, which still satisfies
eq.(2.14), along with
p̂µ |0; pi = pµ |0; pi . (2.15)
µ µ
Now, we build up the Hilbert space by acting with creation operators αm and α̃m with
m < 0. Any generic state arises from acting with any number of these creation operators on
the vacuum,
µ1 mµ1 µ2 mµ2 ν1 nν1 ν2 nν2
(α−1 ) (α−2 ) . . . (α̃−1 ) (α̃−2 ) . . . |0; pi . (2.16)
Each state in the Hilbert space is a different excited state of the string. Each has the inter-
pretation of a different particle in spacetime. As there’s an infinite number of ways to excite
a string there are an infinite number of different particles in this theory.

Ghosts
There’s a problem with the Hilbert space that we’ve constructed: it doesn’t have positive
norm. The presence of the spacetime Minkowski metric in the expression

[aµm , aνn † ] = η µν δm,n . (2.17)

The commutators of time components have a negative sign

[a0m , a0m† ] = η 00 = −1 , (2.18)

– 22 –
The state constructed with an odd number of time-like operators have negative norm.
For the state given by
|ϕi = a†µ1 †µ2 †µl
m1 am2 · · · aml |0i (2.19)
if an odd number of µ’s are equal to zero, then hϕ|ϕi < 0. A simple example of a negative
norm state is given by |ϕi = a†0
m |0i with norm

hϕ|ϕi = h0|a0m a†0

m |0i
= h0|[a0m , a†0
m ]|0i
(2.20)
= η 00
= −1 .

States with negative norm are called ghosts. There is a physical reason why there are these
ghosts states here and a physical resolution for how to get rid of them. Here, we are just
quantizing the free lagrangian. But we’ve to remember that we should not forget the con-
straints. So, we now have infinite number of negative norm states and we also have infinite
number of constraints. We can use these infinite number of constraints to project out the
infinite number negative norm states.
Now, the constraints L̃n and Ln are operators act on physical state in Hilbert space. These
vanish in a classical theory. Quantum analogue of this is L̃n and Ln annihilate all physical
states. Because L†n = L−n , it is therefore sufficient to require

Ln |ϕi = L̃n |ϕi = 0 for n > 0 . (2.21)

Now, when we go from classical constraints to quantum operators sometimes we are in trouble
because of ordering ambiguities. The problem is that, unlike for Lm with m 6= 0, the operator
L0 is not uniquely defined when we move to the quantum theory. There is an operator ordering
ambiguity arising from the commutation relations eq.(2.10). We don’t know what order to
put the αnµ operators in the quantum operator L0 or the α̃nµ operators in L̃0 .
1X
Lm = αn .αm−n αnµ , αn−m
ν
commute for m 6= 0 . (2.22)
2 n

But, for m = 0, the only possible ambiguity should be in the commutator. To make this
ambiguity manifest, we pick a choice of ordering. We define the quantum operators to be
normal ordered, with the annihilation operators αni , n > 0, moved to the right.
Therefore,
∞ ∞
1 X X 1
L0 = : αn · α−n : = α−n · αn + α02 . (2.23)
2 n=−∞ 2
n=1

and
∞ ∞
1 X X 1
L̃0 = : α̃n · α̃−n : = α̃−n · α̃n + α̃02 . (2.24)
2 n=−∞ 2
n=1

– 23 –
In particular,
L0 |ϕi = a |ϕi , L̃0 |ϕi = a |ϕi , (2.25)
for some constant a. Here, different orderings will only differ by the constant a. So, a can be
treated as normal ordering constant. We see from here L0 − L̃0 annihilate the physical state
(L0 − L̃0 ) |ϕi = 0 . (2.26)
This is called the level-matching condition of the closed string. It is the only constraint
that relates the left and right-moving modes which means we can’t just excite left-movers
without exciting right-movers.
Now, the Virasoro algebra in the quantum theory is
c
[Lm , Ln ] = (m − n)Lm+n + m(m2 − 1)δm+n,0 , (2.27)
12
where, c = D =number of free bosons i.e. the target space (space-time) dimensions. The
term proportional to c is a quantum effect. This means that it appears after quantization
and is absent in the classical theory. This term is called a central extension, and c is called a
central charge.
Now, let’s pretend that in the quantum theory any physical state can be annihilated by all
the constraints i.e. Lm |ϕi = 0 ∀m. This gives
hϕ|[Lm , Ln ]|ϕi = 0 . (2.28)
Choosing, n = −m
hϕ|[Lm , L−m ]|ϕi = 0
D
=⇒ 2m hϕ|L0 |ϕi + m(m2 − 1) hϕ|ϕi = 0 ∀m
12
D (2.29)
=⇒ 2ma hϕ|ϕi + m(m2 − 1) hϕ|ϕi = 0 ∀m
12
D
=⇒ 2ma + m(m2 − 1) = 0 ∀m .
12
This calculation is done under the assumption Lm |ϕi = 0 ∀m.
Now, let’s do it under the assumption Lm |ϕi = 0 for m > 0.
:0

hϕ|[Lm , L−m ]|ϕi = hϕ|Lm L−m |ϕi − hϕ|L  −m
 |ϕi
Lm (Lm |ϕi = 0 , m > 0) (2.30)

= ||L−m |ϕi ||2 (L−m = L†m ) .
Therefore,
D
||L−m |ϕi ||2 = 2ma + m(m2 − 1) , (2.31)
12
where |φi is normalized. The eq.(2.29) is wrong. Because, this gives infinite number of
equations for two variables a and D. The correct assumption is Lm |ϕi = 0 for m > 0. Now,
decoupling the ghosts give unique choice for the constant a = 1. It also restricts the number
of scalar fields that we have in the theory, D = 26.
In the next section 2.1, we’ll show how to quantize the string and derive these values for a
and D in lightcone gauge.

– 24 –
2.1 Light-cone Quantization (LCQ)
The constraints Lm and L̃m are quadratic in the oscillators. We want to find a way to
linearize the constraints and solve them. The construction is known as the Light-Cone
Quantization. We started out with the spacetime coordinates i.e. the scalar fields on the
world-sheet
X µ = (X 0 , X 1 , · · · , X D−2 , X D−1 ) . (2.32)
In terms of these coordinates we had our metric η µν = diag(−1, 1, · · · , 1). Now, we go to the
light-cone coordinates
X µ → (X + , X − , X 1 , X 2 , · · · , X D−2 ) , (2.33)
where, we define the two coordinates
1
X ± = √ (X 0 ± X D−1 ) . (2.34)
2
In lightcone coordinates, the spacetime Minkowski metric is
D−2
X
ds2 = −2dX + dX − + dX i dX i . (2.35)
i=1

The light-cone metric is  

0 −1 . . . 0
−1 0 . . . 0
 
ηµν =
 .. .. . . ..  , (2.36)
 . . . .
0 0 ... 1
where, µ, ν ∈ {+, −, i} and i ∈ {1, 2, · · · , D − 2}. Greek indices µ, ν runs over all target space
dimensions, while Latin indices i, j only runs over transversal directions.
In this notation, the inner product of two arbitrary vectors U = (U + , U − , U i ) and V =
(V + , V − , V i ) takes the form
D−2
X
U.V = −U + V − − U − V + + U iV i . (2.37)
i=1

Now, let’s go back to our open string mode expansion (Neumann boundary conditions at
both ends)
X αµn
X µ (τ, σ) = xµ + ls2 pµ τ + ils e−inτ cos nσ . (2.38)
n
n6=0

Negative norm states will be produced by +

αm −.
and αm
For open string X + is
 i X α+n −inτ 
X + (τ, σ) = x+ + ls2 p+ τ + e cos nσ .
ls p+ n
n6=0

– 25 –
In the light-cone coordinates on the worldsheet
Z
S[X] = T dσ + dσ − ∂+ X µ ∂− X ν ηµν . (2.39)

This action still has an infinite number of symmetries.

Under the conformal transformation

σ ± → σ̃ ± = f ± (σ ± ) (2.40)

the action is invariant.

Now,

σ̃ + = τ̃ + σ̃ = f + (σ + )
(2.41)
σ̃ − = τ̃ − σ̃ = f − (σ − ) ,

which gives
1h + + i
τ̃ = f (σ ) + f − (σ − )
2 (2.42)
1h i
σ̃ = f + (σ + ) − f − (σ − ) .
2
τ̃ is a function of σ + and σ − . This actually looks like an oscillator i.e. coordinate of a string.
τ̃ satisfies the free wave equation
∂+ ∂− τ̃ = 0 .
Therefore, τ̃ ∼ one of the X µ .
Now, the light-cone gauge corresponds to the choice
 i X α+n −inτ 
X + = x+ + ls2 p+ τ + e cos nσ = x+ + ls2 p+ τ̃ . (2.43)
ls p+ n
n6=0
| {z }
=τ̃

We can get rid of all the αn+ oscillators in terms of new variable τ̃ . This corresponds to setting
αn+ = 0 for n 6= 0. The light-cone gauge has eliminated the oscillator modes of X + . We are
left with αn− oscillators. In the following the tildes are omitted from the parameters τ̃ and σ̃,
replacing
(τ̃ , σ̃) → (τ, σ) =⇒ X + (τ, σ) = x+ + ls2 p+ τ .
Now, we look at the Virasoro constraints. Using Virasoro constraints
X
0 = (Ẋ ± X 0 )2 = −(Ẋ + ± X +0 )(Ẋ − ± X −0 ) − (Ẋ − ± X −0 )(Ẋ + ± X +0 ) + (Ẋ i ± X i0 )2
i
X
= −2ls2 p+ (Ẋ − −0
±X )+ (Ẋ i ± X i0 )2 ,
i
(2.44)

– 26 –
we get
1 X i
(Ẋ − ± X −0 ) = (Ẋ ± X i0 )2 . (2.45)
2ls2 p+
i
Now, we note that the formula for the open string
X αµn
X µ (τ, σ) = xµ + ls α0µ τ + ils e−inτ cos nσ
n
n6=0
∞
X ∞
X
=⇒ Ẋ µ = ls αµn cos nσe−inτ , X 0µ = −ils αµn sin nσe−inτ (2.46)
n=−∞ n=−∞
∞
1 µ  l X
s
=⇒ ∂± X µ = Ẋ ± X 0µ = αµ e−in(τ ±σ) .
2 2 n=−∞ n

We have
1 X i 1 X
(Ẋ − ± X −0 ) = (Ẋ ± X i0 )2 =⇒ ∂± X − = (∂± X i )2 . (2.47)
2ls2 p+ ls2 p+
i i

The mode expansion for X − for an open string is

X α−
X − (τ, σ) = x− + ls α0− τ + ils n −inτ
e cos nσ . (2.48)
n
n6=0

So, we can get the left hand side of eq.(2.47)

∞
− ls X − −in(τ +σ)
∂+ X = α e . (2.49)
2 n=−∞ n

The right hand side of the eq.(2.47) is

1 X i 2 1 X X X i i −i(k+m)(τ +σ) 1 XXX i i −in(τ +σ)
(∂ + X ) = α α
k m e = αn−m αm e
ls2 p+ 4p+ 4p +
i i k m i n m
1 XXX i i −in(τ +σ)
= + αn−m αm e .
4p n m i
(2.50)

Therefore, we obtain the explicit solution

D−2 ∞
1 X X i
αn− = i
αn−m αm
2p+ ls m=−∞
i=1
D−2 ∞
1 1 X X i i

= + : αn−m αm : −aδn,0 (2.51)
p ls 2 m=−∞ i=1
 1 D−2 ∞
1 X X
i i

= √ : αn−m αm : −aδn,0 .
p+ 2α0 2 i=1 m=−∞

– 27 –
We introduce a constant a here, due to ordering issue of operators for n = 0. Therefore, in
the light-cone gauge we get rid of all the troublesome oscillators. αn+ and αn− oscillators are
not there any more. Our physical hilbert space is only constructed from only the transverse
oscillators.
Let’s write our Physical Hilbert Space. So we have,

|ϕi = aim1 †1 aim2 †2 · · · aiml †l |0i , (2.52)

where, i ∈ {1, · · · , D − 2} . They all have positive norm and the theory is unitary.

2.1.1 Open string spectrum

The key to compute the spectrum is given by eq.(2.51). First, we identify α0− = ls p− , this is
the mass-shell condition. We get
D−2
X 2(N − a)
M 2 = −pµ pµ = 2p+ p− − p2i =
ls2
i=1
0 2
=⇒ α M = N − a , (2.53)
D−2
XX ∞
where ls2 = 2α0 , N = i
α−n αni , α0i = ls pi .
i=1 n=1

Now using
√ √
αn = nan , α−n = na†n (n > 0) ,
∞ D−2
X X (2.54)
we get N = nai† i
n an .
n=1 i=1

Now, we’ll compute the spectrum. The usual way to compute the spectrum is to go level by
level.
For N=0: We want to list all possible states with number operator eigen value equal to zero.
The vacuum is the state where we don’t act with any creation operators. To find the mass
of the state we perform
α0 M 2 |0i = −a |0i , α0 M 2 = −a . (2.55)
For N=1: The state is given by |ϕiiN =1 = ai†1 |0i . We have D − 2 states.
We have
α0 M 2 |ϕiiN =1 = (1 − a) |ϕiiN =1 , α0 M 2 = 1 − a . (2.56)
The first excited states looks like a (D − 2) component vector representation of the transverse
rotation group SO(D − 2). A transversely polarized vector, after Lorentz transformation
gets a longitudinal polarization unless it is massless. From the Wigner’s classification of
representations of the Poincaré group we see that the massive particles in R1,D−1 transform
under the little group SO(D − 1) of spatial rotations. This can be seen by going to the rest

– 28 –
frame of the particle by setting pµ = (p, 0 . . . , 0). Here, we can see how any internal indices
transform under the little group SO(D − 1) of spatial rotations. But the particles described
here have (D − 2) states. The massless representation of SO(1, D − 1) has D − 2 degrees of
freedom and looks like a vector. This means that massless particles get away with having
fewer internal states than massive particles. For example, in four dimensions the photon has
two polarization states, but a massive spin-1 particle must have three. Therefore, if we want
the quantum theory to preserve the SO(1, D − 1) Lorentz invariance, then these states will
have to be massless.

Lorentz Invariance =⇒ α0 M 2 = 1 − a = 0 =⇒ a = 1 . (2.57)

It forms a vector of SO(D − 2) which is a little group of massless particles for SO(1, D − 1).
Now, for N=0, we find a tachyon, whose mass is given by α0 M 2 = −1.
For N=2: N = 2 gives the first states with positive (mass)2 . The states are

ai† |0i, ai† aj†

1 |0i , a = 1, α0 M 2 = 1 . (2.58)
| 2{z } | 1 {z }
D−2 (D−2)(D−1)
2

(D − 1)(D − 2) (D + 1)(D − 2)
The total number of states is D − 2 + = .
2 2
Now, the symmetric traceless representation of SO(D − 1), tij , i, j = 1, · · · , D − 1 with tii = 0 has
D(D − 1) (D + 1)(D − 2)
−1= states.
2 2
(2.59)
So these states reassemble into a traceless representation of SO(D − 1).
The total number of open-string states with α0 M 2 = n − 1 , denoted cn is given by coefficient
of xn in tr xN , where N is the number operator
D−2
XX ∞
i
N= α−n αni . (2.60)
i=1 n=1

The generating function is given by

∞
X
G(x) = cn xn = tr xN . (2.61)
n=0

Now,
∞
Y
tr xN = tr xα−n .αn
n=1
∞ (2.62)
Y
n −(D−2) −(D−2)
= (1 − x ) = [f (x)] ,
n=1

– 29 –
where,
∞
Y
the classical partition function f (x) = (1 − xn ) .
n=1

The general state at level n is of the form

i †
aim1 †1 · · · amp p Si1 ···ip |0i (2.63)
P
where i mi = n. If cn is the number of states at level n, then cn is given by the formula
∞ ∞
X
n
Y 1
cn x = = 1 + (D − 2)x
(1 − xn )D−2
n=0 n=1
" # (2.64)
(D − 2)(D − 1)
+ + (D − 2) x2 + · · · .
2

A heuristic approach to compute the normal-ordering constant

Here, we give a heuristic approach is to compute the normal-ordering constant a appearing

in the definition of L0 directly.
D−2 ∞
1X X i i
L0 = α−n αn
2
i=1 n=−∞
D−2
" ∞ ∞
#
1 i i 1X X i i X
i i
= α0 α0 + : α−n αn : + αn α−n
2 2
i=1 n=1 n=1
(2.65)
D−2 ∞ D−2 ∞
1 XX 1 XX
= α0i α0i + i
: α−n αni :+ n
2 2
i=1 n=1 i=1 n=1
D−2 ∞ ∞
1 XX 1 X
= α0i α0i + i
: α−n αni : + (D − 2) n .
2 2
i=1 n=1 n=1

Therefore,
∞
1 X
− a = −1 = (D − 2) n , (2.66)
2
n=1
P∞
since a = 1 by requiring Lorentz invariance. Now, we calculate n=1 n by ζ-function regu-
larization, which is a formal way to regularize the sum ∞
P
n=1 n.

1. Zeta Function Regularization

The Riemann zeta function is defined, for Re(z) > 1, by the sum
∞
X 1
ζ(z) = . (2.67)
nz
n=1

– 30 –
But ζ(z) has a unique analytic continuation to all values of z. In particular,
1
ζ(−1) = − . (2.68)
12
Now, we will give a nice proof of this result given by eq.(2.68). From the reflection formula
of Riemann zeta function we have
 πz 
ζ(z) = 2z π z−1 sin Γ(1 − z)ζ(1 − z)
2
1 (2.69)
=⇒ ζ(−1) = 2 (−1)Γ(2)ζ(2) .
2π
Now,
∞
X 1
ζ(2) =
n2
n=1

which can be obtained from the fourier series of f (x) = x2 , −π < x < π .
For a fourier series on [−L, L] we have
∞
a0 X nπx nπx
f (x) = + an cos + bn sin . (2.70)
2 L L
n=1

Here, we get

1 π 2 π2
Z
a0 = x dx =
π −π 3
Z π
1 4(−1)n
an = x2 cos nxdx = (2.71)
π −π n2
Z π
1
bn = x2 sin nxdx = 0 .
π −π

So, we have
∞
2 π2 X cos nx
x = +4 (−1)n
3 n2
n=1
∞
π2 X (−1)n (−1)n
π2 = +4 (x = π) (2.72)
3 n2
n=1
∞
X 1 π2
=⇒ ζ(2) = = .
n2 6
n=1

π2
Therefore, substituting the value of ζ(2) = 6 in eq.(2.69) we get

1
ζ(−1) = − . (2.73)
12

– 31 –
2. Using cut-off
We introduce a cut-off   1.
∞ ∞ ∞
X X ∂ X −n
n −→ ne−n = − e
∂
n=1 n=1 n=1
∂ h − i
(2.74)
=− e (1 − e− )−1
∂
1 1
= 2− + O()
 12
The 1/ term diverges as  → 0. We renormalize this term to preserve the Weyl invariance
of the Polyakov action since it contributes to a cosmological constant on the worldsheet.
∞
X 1
n=− . (2.75)
12
n=1

From eq.(2.66) we get

D−2
a= . (2.76)
24
a = 1 gives D = 26.

2.1.2 Lorentz Symmetry

Previously we see that in LCQ Unitarity is manifest. But, Lorentz invariance could be broken.
The reason is that along the way starting from the Polyakov action to the LCQ and getting rid
of negative norm states we use a conformal transformation and we pretended that conformal
transformations were preserved by the quantization of the system. If that was true then
we get Lorentz invariance because we started with a Lorentz invariant theory. The theory
becomes Lorentz invariant if conformal invariance is not broken by Quantum corrections. In
LCQ the critical values for the normal ordering constant a and for the number D of spacetime
dimensions follow by the requirement of Lorentz invariance. The generators of the Lorentz
transformation are given by
Z π
J µν = T dσ(X µ X˙ ν − X ν X˙ µ ) . (2.77)
0

The mode expansion for angular momentum generators J µν of an open bosonic string is given
by
∞
X 1 µ
J µν = xµ pν − xν pµ − i (α αν − α−m ν µ
αm ) ≡ lµν + S µν . (2.78)
m −m m
m=1
The first term,lµν ,
is the orbital angular momentum of the string while the second term S µν is
the angular momentum due to excited oscillator modes representing internal spin. Classically,
these obey the Poisson brackets of the Lorentz algebra.
These should satisfy the Lorentz algebra

[J µν , J ρσ ] = iη µρ J νσ + iη νσ J µρ − iη µσ J νρ − iη νρ J µσ

– 32 –
in quantum theory.
Taking µ = i, ρ = j, ν = σ = − for calculating commutator.
Now, Lorentz algebra tells
[J i− , J j− ] = 0 . (2.79)
An explicit, long calculation gives
∞
i− j− 1 X
i j j i
[J ,J ]=− ∆m (α−m αm − α−m αm ) ,
(p+ )2
m=1 (2.80)
26 − D 1  D − 26 
where ∆m =m + + 2(1 − a) .
12 m 12
The calculation is worked out in appendix B. Lorentz invariance implies ∆m = 0 which gives
a = 1 and D = 26.

2.1.3 Closed string spectrum

From the previous section, we found that a = 1 and D = 26. Now, we analyze the spectrum.
For the case of the closed string, there are two sets of modes (left-movers and right-movers)
and the level-matching condition must be taken into account. The spectrum is easily found
from that of the open string, since closed- string states are tensor products of left-movers and
right-movers, each of which has the same structure as open-string states. The mass of states
in the closed-string spectrum is given by

α0 M 2 |ϕi = 4(N − 1) |ϕi = 4(Ñ − 1) |ϕi

α0 M 2 2 (2.81)
M |ϕi = (N − 1) |ϕi = (Ñ − 1) |ϕi ,
4
where,
X ∞
24 X ∞
24 X
X
i
N= α−n αni , Ñ = i
α̃−n α̃ni
i=1 n=1 i=1 n=1

and

on a physical state ϕ : N = Ñ =⇒ level-matching condition (L0 − L̃0 ) |ϕi = 0 .

The physical states of the closed string are as follows.

0 2
N = Ñ = 0 : It is the vacuum gives α M 4 = −1. Therefore the vacuum |0i is a tachyon.
N = Ñ = 1 : For the N = 1 level there is a set of 242 = 576 states of the form

a†i †j
1 ã1 |0i

corresponding to the tensor product of two massless vectors, one left-moving and one right-
moving. The states are massless
α0 M 2
= 0.
4

– 33 –
These states form a 24⊗24 representation of SO(24) . These decompose into three irreducible
representations

traceless symmetric ⊕ anti-symmetric ⊕ singlet (=trace) .

24
" #
1 †i †j †j †i 1  X †k †k  ij
(a ã + a1 ã1 ) − a1 ã1 δ |0i → traceless symmetric rank 2 tensor
2 1 1 24
k=1
1  †i †j †j †i

a ã − a1 ã1 |0i → anti-symmetric rank 2 tensor (2.82)
2 1 1
24
1 X †k †k
a1 ã1 |0i → trace .
24
k=1
Now, we denote rank 2 symmetric traceless tensor by Gij , rank 2 symmetric anti-symmetric
tensor by Bij and trace by Φ. Therefore, the number of components of Gij are 24×25 2 − 1,
the number of components of Bij are 24×23 2 and the number of component of Φ is one.
The symmetric traceless part in i and j transforms under SO(24) as a massless spin-two
particle, the graviton. The trace term is a massless scalar, which is called the dilaton. The
antisymmetric part transforms under SO(24) as an antisymmetric rank 2 tensor. This is
usually called the “Kalb-Ramond field” or the “2-form”.

A Conserved currents associated with the Poincaré transformation

Global symmetry transformation is

δX µ = aµν X ν + bµ . (A.1)

Conserved currents are given by

∂L
i jiα ≡ δX µ , (A.2)
∂(∂α X µ )
where,

α index to label world-sheet coordinates σ α ,

i index to label parameters in the symmetry transformation , (A.3)
µ index to label fields in the Lagrangian .

For translation invariance we denote the conserved current by Pµα . Here, δX µ = bµ .

From Polyakov action
√
Z
µ T
SP [hαβ , X ] = − d2 σ −h(hαβ ∂α X µ ∂β X ν ηµν ) , (A.4)
2
we get
∂LP T√
µ
=− −h2hαβ ∂β X ν ηµν . (A.5)
∂(∂α X ) 2

– 34 –
Choosing hαβ = ηαβ we get
bµ Pµα = bµ T ∂ α Xµ , (A.6)
where we drop the overall constant prefactor −1.
Therefore, conserved current associated with the translational invariance of X µ is

Pµα = T ∂ α Xµ . (A.7)
α . Infinitesimal trans-
Now, we denote the current associated with Lorentz invariance by Jµν
formation parameter is δX µ = aµν Xν . From eq.(A.2) we get
∂L
aµν Jµν
α
= δX µ
∂(∂α X µ )
= Pµα aµν Xν (A.8)
1
= − aµν (Xµ Pνα − Xν Pµα ) ,
2
where, we use that infinitesimal transformation aµν is antisymmetric i.e. aµν = −aνµ .
Therefore, we get
α
Jµν = (Xµ Pνα − Xν Pµα ) , (A.9)
where, we drop the overall constant prefactor.
The conserved currents associated with the Poincaré transformation of X µ are
Pαµ = T ∂α X µ ,
(A.10)
J µν µ ν ν µ
α = T (X ∂α X − X ∂α X ) .

B Lorentz algebra

The oscillators αn− is given by

D−2 ∞
!
1 1X X
αn− = √ i
: αn−m i
αm : −aδn,0
p+ 2α0 2 i=1 m=−∞
(B.1)
D−2 ∞
!
1 1X X  1
= + i i
: αn−m αm : −aδn,0 setting α0 =
p 2 m=−∞
2
i=1

Now, from the expression for αn− we see that [J i− , J j− ] involves terms quartic and quadratic
in the oscillators αni . The commutators involving the quartic terms vanish. So, we can make
the following ansatz for [J i− , J j− ] :
1
[J i− , J j− ] = − C ij (B.2)
(p+ )2
where
∞
j
X
C ij = i
∆m (α−m j
αm − α−m i
αm ). (B.3)
m=1

– 35 –
Here, ∆m are c-mumbers which we compute. Now,
k
h0|αm C ij α−m
l
|0i = m2 (δ ik δ jl − δ jk δ il )∆m . (B.4)

From eq.(B.4) we see that we can compute the coefficients ∆m by evaluating C ij from eq.(B.2).
We define E j = p+ E j− . We get
h 1 i
x− , + = i(p+ )−2
p
i (B.5)
[xi , αn− ] = + αni
p
[xi , E j ] = −iE ij .

With this we can compute the commutator [J i− , J j− ] explicitly. This gives

[J i− , J j− ] = −(p+ )−2 C ij ,
(B.6)
where C ij = 2ip+ α0− E ij − iE i pj + iE j pi − [E i , E j ] .

To evaluate matrix elements of the form of eq.(B.4) we need commutator relations for the
αn− . Comparing with the Virasoro generators Lm we evaluate that the p+ αn− satisfy the
commutation relations
hD − 2 i
− + − −
[p+ αm , p αn ] = (m − n)p+ αm+n + (m3 − m) + 2am δm+n
12 (B.7)
i + − i
[αm , p αn ] = mαm+n .

Now we use eq.(B.7) and the commutators

i
mαm+n
i
[αm , αnj ] = mδm+n δ ij , [αm
i
, αn− ] =
p+
to compute
k
h0|αm C ij α−m
l
|0i = h0| 2m(m + p+ α0− )δ ik δ jl + mpj pk δ il − mpj pl δ ik
m
 X 1 k −

+ m αm−n αni − δ ik p+ αm
n (B.8)
n=1
m

−
X 1 j l 
× δ jl p+ α−m −m α α |0i − (i ↔ j) .
p −p p−m
p=1

To evaluate this expression given by eq.(B.8) we proceed as follows.

Using
− −
 D−2 3
(p+ )2 h0|αm α−m |0i = 2m p+ α0− +

(m − m) + 2am (B.9)
12
as follows from
hD − 2 i
− + − −
[p+ αm , p αn ] = (m − n)p+ αm+n + (m3 − m) + 2am δm+n (B.10)
12

– 36 –
and the identities
m
+ −
X 1 j l m(m − 1)
p h0|αm α α |0i = pj pl + δ jl
n −n n−m 2
n=1
m m (B.11)
X 1 k i
X 1 j l
h0| α α α α |0i − (i ↔ j) = (m − 1)(δ il δ jk − δ jl δ ik )
n m−n n p −p p−m
n=1 p=1

we can compute eq.(B.8). Comparing the result with eq.(B.4) we get

26 − D 1  D − 26 
∆m = m + + 2(1 − a) . (B.12)
12 m 12

References

[1] P. Goddard, J. Goldstone, C. Rebbi and C.B. Thorn, Quantum Dynamics of a Massless Rela-
tivistic String, Nucl. Phys. B56 (1973) 109-135.

[2] P. Goddard, C. Rebbi and C.B. Thorn, Lorentz Covariance and the Physical States in Dual-
Resonance Models, Nuovo Cimento 12A (1972) 425-441.

[3] M.B. Green, J.H. Schwarz and E. Witten, Superstring Theory, vol. 1, Cambridge Univ. Press,
1987.

[4] K. Becker, M. Becker, J. H. Schwarz, String Theory and M-Theory:A Modern Introduction.

[5] P. West, Introduction to Strings and Branes.

[6] B. Zwiebach, A First Course in String Theory, Cambridge Univ. Press, 2009.

[7] J. Polchinski, An introduction to the Bosonic String, Vol 1.

[8] D.Lüst, S. Theisen and R. Blumenhagen, Basic Concepts of String Theory.

[9] T. Weigand, Introduction to String Theory Winter term 2011/12.

[10] Z.W. Liu Quantization of the Bosonic String Theory, Beijing String Discussion Class @ IHEP,
CAS.

[11] D.Tong, Lectures on String Theory.

[12] L. McAllister, Lectures on String Theory, Cornell University lecture note, 2010.

– 37 –