String Theory Exercises

Exercise 1.1
We consider a relativistic point particle moving in D-dimensional space-time with metric
Gµν and an action given by:
Z q
S = −m dτ −Gµν Ẋ µ Ẋ ν .

To find the equations of motion we vary the action:

Z
1  
δS = m dτ q δGµν Ẋ µ Ẋ ν + 2Gµν Ẋ µ δ Ẋ ν
2 −Gµν Ẋ µ Ẋ ν
Z
1  
= m dτ q ∂ρ Gµν Ẋ µ Ẋ ν δX ρ + 2Gµν Ẋ µ δ Ẋ ν .
2 −Gµν Ẋ µ Ẋ ν

By employing partial integration and using the fact that taking the variation and taking
the derivative with respect to τ commute and the fact that the variation δX µ vanishes
at the boundary, the second term can be rewritten so that the above equation becomes
the following:
  
Z µ ν µ
∂ρ Gµν Ẋ Ẋ ∂  G Ẋ
δS = m dτ  q − q µν  δX ρ .
∂τ
2 −Gµν Ẋ µ Ẋ ν −Gµρ Ẋ µ Ẋ ν

Since the variation in the action must be zero for any variation δX µ , we have, by the
fundamental lemma of of the calculus of variations, the following equation of motion:
 
µ ν µ
∂ρ Gµν Ẋ Ẋ ∂  G Ẋ
0=− q + q µρ .
µ ν ∂τ µ ν
2 −Gµν Ẋ Ẋ −Gµν Ẋ Ẋ

Now, the proper time that elapses between two events occurring at τ1 and τ2 respectively,
is given as follows:
Z τ2 q
dτ −Gµν Ẋ µ Ẋ ν .
τ1

The statement that τ is the proper time of the particle is equivalent to the statement
that the above integral is equal to τ2 − τ1 for any two values τ1 and τ2 in the range of the

1
parameter τ . But this statement is in turn equivalent to the following equation holding
true for all τ :
q
1 = −Gµν Ẋ µ Ẋ ν .
In the case where τ is the proper time, we can therefore rewrite the equation of motion
determined above as follows:
1 1
0 = − ∂ρ Gµν Ẋ µ Ẋ ν + ∂τ (Gµρ Ẋ µ ) = ∂ρ Gµν Ẋ µ Ẋ ν + ∂ν Gµρ Ẋ ν Ẋ µ + Gµρ Ẍ µ
2  2
1 1
= ∂ν Gµρ − ∂ρ Gµν Ẋ µ Ẋ ν + Gµρ Ẍ µ = (∂ν Gµρ + ∂µ Gρν − ∂ρ Gµν ) Ẋ µ Ẋ ν + Gµρ Ẍ µ
2 2
By multiplying through with the inverse metric Gσρ and performing the implicit summa-
tion we arrive at the geodesic equation:
1 ∂X µ ∂X µ ∂ 2 X σ
0 = Gσρ (∂ν Gµρ + ∂µ Gρν − ∂ρ Gµν ) + .
2 ∂τ ∂τ ∂τ

Exercise 1.2
The induced metric of a surface parametrized by n coordinates ξ α and embedded in a
space described by coordinates X µ is given by the following:
∂X µ ∂X ν
γαβ = ηµν .
∂ξ α ∂ξ β
The area of the surface is given as follows:
Z p
A = dn ξ | det(γ)|.

For a cylinder of length L and radius R embedded in three-dimensional Euclidean space

thus:
X 1 (ξ) = R cos(ξ 1 ), X 2 (ξ) =R sin(ξ 2 ), X 3 (ξ) = Lξ 2
0 ≤ 1 ≤ 2π, 0 ≤ ξ 2 ≤ 1,
the components of the induced metric are given as follows:
γ11 = R2 , γ22 = L2 , γ01 = γ10 = 0.
The area of the cylinder is given as follows:
Z 2π Z 1 √
1
A= dξ dξ 2 R2 L2 = 2πRL.
0 0
For a sphere of radius R embedded in three-dimensional Euclidean space thus:
X 1 (ξ) = R cos(ξ 1 ) sin(ξ 2 ), X 2 (ξ) =R sin(ξ 1 ) sin(ξ 2 ), X 3 (ξ) = R cos(χ2 )
0 ≤ 1 ≤ 2π, 0 ≤ ξ 2 ≤ π,
the components of the induced metric are given as follows:
γ11 = R2 sin2 (ξ 2 ), γ22 = R2 , γ01 = γ10 = 0.
The area of the sphere is given as follows:
Z 2π Z π q
1
A= dξ dξ R4 sin2 (ξ 2 ) = 4πR2 .
2
0 0

2
Exercise 1.3
The Lagrangian of the Nambu-Goto action is given as follows:
q
L = T (Ẋ · X 0 )2 − Ẋ 2 X 02 .

We obtain the conjugate momentum by differentiation:

∂L T2  
Πµ (τ, σ) = = Ẋ · X 0 gµν X 0ν − X 02 gµν Ẋ ν .
∂ Ẋ µ L
The conjugate momentum satisfies the following equations:

∂L T2  
Π · X 0 = Πµ (τ, σ)X 0µ = = Ẋ · X 0 X 02 − X 02 Ẋ · X 0 = 0,
∂ Ẋ µ L

2T4 
0 0ν 02

ν 0 0µ 02 µ

Π = 2 Ẋ · X gµν X − X gµν Ẋ Ẋ · X X − X Ẋ
L
T4  
= 2 (Ẋ · X 0 )2 X 02 − 2X 02 (Ẋ · X 0 )2 + X 04 Ẋ 2
L  
2 02 02 2 0 2
T4   T X X Ẋ − (Ẋ · X )
= 2 X 02 X 02 Ẋ 2 − (Ẋ · X 0 )2 =
L (Ẋ · X 0 )2 − Ẋ 2 X 02
= −T 2 X 02 .

The canonical Hamiltonian is given as follows:

T2  
H = Πµ Ẋ µ − L = (Ẋ · X 0 )2 − X 02 Ẋ 2 − L
L
(Ẋ · X ) − X 02 Ẋ 2
0 2
=Tq − L = L − L = 0.
(Ẋ · X 0 )2 − Ẋ 2 X 02

When working in a specific gauge, the above equations need not be automatically satisfied
and must be imposed as constraints.

Exercise 1.5
We wish to show that:
p 1p
δ − det(g) = − det(g) g µν δgµν .
2
We have that:
∂ p 1 ∂ det(g) C µν
− det(g) = − p =− p ,
∂gµν 2 − det(g) ∂gµν 2 − det(g)

where C µν denotes the cofactor of g and use has been made of Laplace’s formula:

det g = gµν C µν .

3
Next we make use of Cramer’s rule for inverse matrices:
C µν
g νµ = .
det(g)

Since g νµ is symmetric in its indices, we arrive at the following:

∂ p − det(g)g µν 1p
− det(g) = √ = − det(g)g µν .
∂gµν 2 − det g 2

From this we obtain the desired result:

p
p ∂ − det(g) 1p
δ − det(g) = δgµν = − det(g) g µν δgµν .
∂gµν 2

Exercise 1.8
We consider an infinitesimal coordinate transformation:

ξ α → ξ˜α (ξ) = ξ α − aα (ξ).

Under coordinate transformations the target space map X µ (ξ) transforms as a scalar:
˜ = X µ (ξ).
X̃ µ (ξ)

This fact can be used in making the following Taylor expansion, where only leading and
next-to-leading order terms are retained due to the infinitesimal nature of the transfor-
mation under consideration:

X̃ µ (ξ) = X̃ µ (ξ˜ + a) = X µ (ξ) + aα ∂α X µ .

It follows that the target space map satisfies the following:

δX µ (ξ) = X̃ µ (ξ) − X µ (ξ) = aα ∂α X µ .

Similarly we find that:

˜ + aγ ∂γ gαβ .
g̃αβ (ξ) = g̃αβ (ξ)

Under coordinate transformations the metric transforms as follows, where, as before,

higher-order terms are omitted:
γ δ γ δ
  
˜ = gγδ (ξ) ∂ξ ∂ξ γ ∂a ∂a
g̃αβ (ξ) = gγδ (ξ) δα + α δβ + β = gαβ (ξ) + gαδ (ξ)∂β aδ + gγβ (ξ)∂α aγ .
δ
˜ α
∂ξ ∂ξ ˜ β ∂ξ ∂ξ

We therefore have that the metric satisfies the following:

δgαβ (ξ) = g̃αβ (ξ) − gαβ (ξ) = aγ ∂γ gαβ + gαδ ∂β aδ + gγβ ∂α aγ .

4
Exercise 1.9
We consider the coordinate transformation (ξ 0 , ξ 0 ) → (ξ + , ξ − ) given as follows:

ξ± = ξ0 ± ξ1.

Expressed in terms of the new variables, the old variables are given as follows:

ξ+ + ξ− ξ+ − ξ−
ξ0 = , ξ1 = .
2 2
The new metric can be determined from the way metric tensors transform under coordi-
nate transformations:
∂η µ ∂η ν
ηµ̄ν̄ = ηµν
∂η µ̄ ∂η µ̄
We find that the components of the new metric are given as follows:
1
η+− = η−+ = − , η++ = η−− = 0.
2
By inverting the two-by-two matrix whose entries consist of the components of the metric,
we obtain the components of the inverse matrix:

η +− = η −+ = −2, η ++ = η −− = 0.

The derivatives with respect to the new variables can be determined from the way deriva-
tives transform under coordinate transformations:
∂ξ µ
∂µ̄ = ∂µ
∂ξ µ̄
We find that the derivatives of the new variables are given by:
1
∂± = (∂0 ± ∂1 ).
2

Exercise 1.10
By J α we denote a conserved current. It therefore satisfies the following:

∂α J α = 0.

We can integrate this equations as follows:

Z Z Z Z Z
 σ
0 = dσ∂α J = dσ(∂τ J + ∂σ J ) = dσ∂τ J + dσ∂σ J = ∂τ dσJ 0 + J 1 σ21 .
α 0 1 0 1

Assuming J 1 to have the same value at the two boundaries, the second term is zero and
we have that the charge
Z
Q = dσJ 0

5
is conserved.

We now consider the Polyakov Lagrangian density in flat gauge:

T
L = − η αβ ∂α X µ ∂β X ν ηµν .
2
The variation of this Lagrangian density is given as follows:

δL = −T η αβ ∂α (δX µ )∂β X ν ηµν .

The Polyakov action is translationally invariant. We therefore consider the following local
infinitesimal translation:

δX µ (ξ) = (ξ)bµ .

In this case the variation of the Lagrangian density is given as follows:

δL = −T η αβ ∂α ((ξ)bµ )∂β X ν ηµν = −∂α T η αβ bµ ∂β X ν ηµν

= −∂α T η αβ bµ ∂β X ν ηµν + ∂α T η αβ bµ ∂β X ν ηµν .

Since the first term is a total derivative and since the variation of the action with respect
to the transformation under consideration is zero, the second term must also be zero. We
conclude that translational invariance gives rise to the following conserved current:

J α = T η αβ bµ ∂β Xµ ηµν = bµ T ∂ α Xµ = −bµ Παµ .

The fact that this current is conserved can also immediately be seen from the equation
of motion:

∂α ∂ α X µ = 0.

We can integrate the zero-component of the conserved current to obtain the conserved
charge:
Z Z
0 µ
Q = dσJ = −b dσΠ0µ = −bµ pµ .

The see that the conservation of this charge is just momentum conservation.

The Polyakov action is also Lorentz invariant. Now we therefore turn our attention
to the following local infinitesimal translation:

δX µ (ξ) = (ξ)ω µν X ν ,

where ωνµ is the constant antisymmetric tensor of Lorentz transformations.

In this case the variation of the Lagrangian density is given as follows:

δL = −T η αβ ∂α ((ξ)ω µρ X ρ )∂β X ν ηµν = −T ∂α (ωνρ X ρ )∂ α X ν ηµν

= −∂α (T ωνρ X ρ ∂ α X ν ) + T ωνρ X ρ ∂α ∂ α X ν .

Because the Lorentz tensor is asymmetric we have that

ωνρ ∂α X ρ ∂ α X ν = 0

6
so that the second term in the variation of the Lagrangian density is equal to the following:

∂α (T ωνρ X ρ ∂ α X ν )

We conclude that the conserved current arising from the Lorentz invariance of the Polyakov
action is given as follows:
1 1
J α = ω µν T Xν ∂ α Xµ = ω µν T (Xν ∂ α Xµ − Xµ ∂ α Xν ) = − ω µν Jµν
α
.
2 2
The fact that this current is conserved can also be shown as follows by using the equations
of motion:

∂α J α = ω µν T ∂α (Xν ∂ α Xµ ) = ω µν T ∂α Xν ∂ α Xµ + ω µν T Xν ∂α ∂ α Xµ .

The first term is zero because the Lorentz tensor is asymmetric and the second term is
zero by the equations of motion so that the current is indeed conserved.
Integrating over the zero-component of the conserved current we obtain the conserved
current:
Z Z
1 µν   1
Q = dσJ = ω T dσ Xµ Ẋν − Xν Ẋµ = ω µν Jµν ,
0
2 2
where Jµν is the total angular momentum tensor, whose spatial components Jij are the
angular momenta of the string.

Exercise 1.11
A function f (x) with a periodicity of 2π can be expanded in a Fourier series expansion
as follows:
X
f (x) = fn einx .
n∈Z

The Fourier modes fn can be deduced by making use of the expansion:

Z 2π Z 2π Z 2π
1 −inσ 1 X
imσ −inσ 1 X
dσf (σ)e = dσ fm e e = fm dσei(m−n)σ
2π 0 2π 0 m∈Z
2π m∈Z 0

1 X
= fm 2πδm,n = fn
2π m∈Z

The mode expansions of the position field X µ (τ, σ) and momentum density Πµ (τ, σ) of
the bosonic closed string are given by the following:
ils X 1 µ −inξ− +
X µ (τ, σ) = xµ + ls2 pµ τ + √ (an e + ãµn e−inξ ),
2 n6=0 n

1 µ 1 X − +
Πµ (τ, σ) = p + √ (aµn e−inξ + ãµn e−inξ ),
2π 2π 2ls n6=0

7
where ξ ± = τ ± σ.
By integrating over the momentum density with respect to σ we obtain the momentum:
Z 2π
1 µ 2π X  Z 2π
Z Z 2π 
µ 1 µ −in(τ −σ) µ −in(τ +σ)
dσΠ (τ, σ) = p dσ + √ an dσe + ãn dσe = pµ .
0 2π 0 2π 2ls n6=0 0 0

The integral of σ over the position field is given by the following:

Z 2π Z 2π Z 2π  Z 2π Z 2π 
µ µ 2 µ ils X 1 µ −in(τ −σ) µ −in(τ −σ)
dσX (τ, σ) = x dσ + ls p τ dσ + √ an dσe + ãn dσe
0 0 0 2 n6=0 n 0 0

= xµ 2π + ls2 pµ τ 2π.

By taking the proper linear combination of the integrals over the position field and
momentum density we obtain the position:
Z 2π  
1 µ
dσ X (τ, σ) − ls τ Π (τ, σ) = xµ + ls2 pµ τ − ls2 pµ τ = xµ .
2 µ
0 2π
If the position field and momentum density are multiplied with a factor of e−inσ and then
integrated with respect to σ, the following results:
Z 2π Z 2π Z 2π
−inσ −inσ
µ
dσX (τ, σ)e =x µ
dσe 2 µ
+ ls p τ dσe−inσ +
0 0 0
 Z 2π Z 2π 
ils X 1 µ −imτ −i(n−m)σ µ −imτ −i(n+m)σ
√ am e dσe + ãm e dσe
2 m6=0 m 0 0
√
√ X 1 2ils π µ −inτ
aµm e−imτ δn,m + ãµm e−imτ δn,−m = − ãµ−n einτ ),


= 2ils π (an e
m6=0
m n

2π
1 µ 2π
Z Z
−inσ
µ
dσΠ (τ, σ)e = p dσe−inσ +
0 2π 0
X Z 2π Z 2π 
1 µ −imτ −i(n−m)σ µ −imτ −i(n+m)σ
√ am e dσe + ãm e dσe
2π 2ls m6=0 0 0

1 X µ −imτ 1
δn,m + ãµm e−imτ δn,−m = √ (aµn e−inτ + ãµ−n einτ ).


=√ am e
2ls m6=0 2ls

By taking the proper linear combinations of these integrals we obtain the modes of the
mode expansion:
 Z 2π Z 2π 
inτ ls µ −inσ in µ −inσ
e √ dσΠ (τ, σ)e − √ dσX (τ, σ)e =
2 0 2π 2ls 0
1 µ 1
(an + ãµ−n e2inτ ) + (aµn − ãµ−n e2inτ ) = aµn ,
2 2
 Z 2π Z 2π 
l in
e inτ
√s µ
dσΠ (τ, σ)e inσ
− √ µ
dσX (τ, σ)e inσ
=
2 0 2π 2ls 0
1 µ 2inτ 1
(a−n e + ãµn ) − (aµ− e2inτ − ãµn ) = ãµn .
2 2
8
Exercise 2.1
The canonical equal-time commutation relations are given as follows:

[X µ (τ, σ), Πν (τ, σ 0 )] = iδ(σ − σ 0 )δνµ ,

[X µ (τ, σ), X ν (τ, σ 0 )] = [Πµ (τ, σ), Πν (τ, σ 0 )] = 0.

From these commutation relations we can determine those of the modes, using the ex-
pressions for the modes in terms of the position field and momentum density found in
exercise 1.11:

Z 2π   Z 2π 
µ 1 µ 2 µ 0 0
[x , pν ] = dσ X (τ, σ) − ls τ Π (τ, σ) , dσ Πν (τ, σ )
0 2π 0
Z 2π Z 2π Z 2π Z 2π
0 1 0 1
= dσ dσ µ
[X (τ, σ), Πν (τ, σ )] = dσ dσ 0 iδ(σ − σ 0 )δνµ
2π 2π
Z0 2π 0 0 0
1
= dσ iδνµ = iδνµ ,
0 2π

  Z 2π Z 2π 
µ ls −imσ im −imσ
[αm , αnν ] = e imτ
√ µ
dσΠ (τ, σ)e − √ µ
dσX (τ, σ)e ,
2 0 2π 2ls 0
 Z 2π Z 2π 
inτ ls 0 ν 0 −inσ 0 in 0 ν 0 −inσ 0
e √ dσ Π (τ, σ )e − √ dσ X (τ, σ )e
2 0 2π 2ls 0
Z 2π Z 2π  
i 0 −i(mσ+nσ 0 ) µ ν 0 µ ν 0
=− dσ dσ e n[Π (τ, σ), X (τ, σ )] + m[X (τ, σ), Π (τ, σ )]
4π 0 0
Z 2π Z 2π  
i 0 −i(mσ+nσ 0 ) 0 µν 0 µν
=− dσ dσ e − niδ(σ − σ )η + miδ(σ − σ )η
4π 0 0
Z 2π
η µν η µν
= (−n + m) dσe−iσ(m+n) = (−n + m)2πδm,−n = mη µν δm,−n ,
4π 0 4π

  Z 2π Z 2π 
µ ls im
[α̃m , α̃nν ] = e imτ
√ µ
dσΠ (τ, σ)e imσ
− √ µ
dσX (τ, σ)e imσ
,
2 0 2π 2ls 0
 Z 2π Z 2π 
inτ ls 0 ν 0 inσ 0 in 0 ν 0 inσ 0
e √ dσ Π (τ, σ )e − √ dσ X (τ, σ )e
2 0 2π 2ls 0
Z 2π Z 2π  
i 0 i(mσ+nσ 0 ) µ ν 0 µ ν 0
=− dσ dσ e n[Π (τ, σ), X (τ, σ )] + m[X (τ, σ), Π (τ, σ )]
4π 0 0
Z 2π Z 2π  
i 0 i(mσ+nσ 0 ) 0 µν 0 µν
=− dσ dσ e − niδ(σ − σ )η + miδ(σ − σ )η
4π 0 0
Z 2π
η µν η µν
= (−n + m) dσeiσ(m+n) = (−n + m)2πδm,−n = mη µν δm,−n ,
4π 0 4π

9
 Z 2π   Z 2π  
µ ν 1 µ 2 µ 0 1 ν 0 2 ν 0
[x , x ] = dσ X (τ, σ) − ls τ Π (τ, σ) , dσ X (τ, σ ) − ls τ Π (τ, σ )
0 2π 0 2π
ls2 τ 2π
Z Z 2π  
0 µ ν 0 µ ν 0
=− dσ dσ [X (τ, σ), Π (τ, σ )] + [Π (τ, σ), X (τ, σ )]
2π 0 0
ls2 τ 2π
Z Z 2π  
0 0 µ 0 µ
=− dσ dσ iδ(σ − σ )δν − iδ(σ − σ)δν = 0,
2π 0 0

Z 2π Z 2π 
0 0
[pµ , pν ] = dσΠµ (τ, σ), dσ Πν (τ, σ ) = 0,
0 0

  Z 2π Z 2π 
µ ls −imσ im −imσ
[αm , α̃nν ] = e imτ
√ µ
dσΠ (τ, σ)e − √ µ
dσX (τ, σ)e ,
2 0 2π 2ls 0
 Z 2π Z 2π 
inτ ls 0 ν 0 inσ 0 in 0 ν 0 inσ 0
e √ dσ Π (τ, σ )e − √ dσ X (τ, σ )e
2 0 2π 2ls 0
Z 2π Z 2π  
i 0 i(nσ−mσ 0 ) µ ν 0 µ ν 0
=− dσ dσ e n[Π (τ, σ), X (τ, σ )] + m[X (τ, σ), Π (τ, σ )]
4π 0 0
Z 2π Z 2π  
i 0 i(nσ−mσ 0 ) 0 µν 0 µν
=− dσ dσ e − niδ(σ − σ )η + miδ(σ − σ )η
4π 0 0
Z 2π
η µν η µν
= (−n + m) dσeiσ(n−m) = (−n + m)2πδm,n = 0.
4π 0 4π
We have now shown which commutation relations for the modes follow from the canonical
equal-time commutation relations. But it is also possible to derive the canonical equal-
time commutation relations from the mode commutation relations just derived:

0 ils X 1 µ −in(τ −σ)
[X (τ, σ), Πν (τ, σ )] = xµ + ls2 pµ τ + √
µ
(an e + ãµn e−in(τ +σ) ),
2 n6=0 n

1 1 X
−im(τ −σ 0 ) −im(τ +σ 0 )
pν + √ (amν e + ãmν e )
2π 2π 2ls m6=0
1 µ i X X e−i(n+m)τ  i(nσ+mσ0 ) µ 0

= [x , pν ] + e [an , amν ] + e−i(nσ+mσ ) [ãµn , ãmν ]
2π 4π n6=0 m6=0 n
1 µ i X X e−i(n+m)τ  i(nσ+mσ0 ) µ 0

= iδν + e nδν δm,−n + e−i(nσ+mσ ) nδνµ δm,−n
2π 4π n6=0 m6=0 n
1 µ i µ X in(σ−σ0 ) 0 i µ i µ X in(σ−σ0 )
= iδν + δν (e + e−in(σ−σ ) ) = δν + δ e
2π 4π n6=0 2π 2π ν n6=0
1 X in(σ−σ0 )
= iδνµ e = iδ(σ − σ 0 )δνµ ,
2π n∈Z

10
 
0 ils X 1 µ −in(τ −σ)
[X (τ, σ), X (τ, σ )] = xµ + ls2 pµ τ + √
µ ν
(an e + ãµn e−in(τ +σ) ),
2 n6=0 n

ν 2 ν ils X 1 ν −im(τ −σ0 ) ν −im(τ +σ 0 )
x + ls p τ + √ (am e + ãm e )
2 m6=0 m
= ls2 τ [xµ , pν ] + ls2 τ [pµ , xν ]
ls2 X X e−i(n+m)τ  i(nσ+mσ0 ) µ ν −i(nσ+mσ 0 ) µ ν

− e [an , am ] + e [ãn , ãm ]
2 n6=0 m6=0 nm
ls2 X X e−i(n+m)τ  i(nσ+mσ0 ) µν −i(nσ+mσ 0 ) µν

=− e nη δm,−n + e nη δm,−n
2 n6=0 m6=0 nm
0 0
l2 X e−in(σ−σ ) + ein(σ−σ )
= − s η µν .
2 n6=0
n

Since the numerator is even in n and the denominator is odd and since the sum runs
over all positive and negative integers, all terms cancel pairwise so that the commutator
is equal to zero.

0 1 µ 1 X 0 0
µ ν
[Π (τ, σ), Π (τ, σ )] = p + √ (aµn e−in(τ −σ ) + ãµn e−in(τ +σ ) ),
2π 2π 2ls n6=0
1 ν 1 X 0 0
p + √ (aνm e−im(τ −σ ) + ãνm e−im(τ +σ ) )
2π 2π 2ls m6=0
1 X X −i(n+m)τ  i(nσ+mσ0 ) µ ν 0

= 22 e e [an , am ] + e−i(nσ+mσ ) [ãµn , ãνm ]
8π ls n6=0 m6=0
1 X X −i(n+m)τ  i(nσ+mσ0 ) µν 0

= 22 e e nη δn,−m + e−i(nσ+mσ ) nη µν δn,−m
8π ls n6=0 m6=0
η µν X X  in(σ−σ0 ) 0

= 22 n e + e−in(σ−σ ) .
8π ls n6=0 m6=0

Since the first factor in the sum is odd in n and the second factor is even and since the
sum runs over all positive and negative integers, all terms cancel pairwise so that the
commutator is equal to zero.

Now we turn our attention to the reality condition. The conjugate of the position field
and momentum density are given by the following:
−ils X 1 µ † in(τ −σ)
(X µ (τ, σ))† = (xµ )† + ls2 pµ τ + √ ((an ) e + (ãµn )† ein(τ +σ) )
2 n=±1,±2,... n
ils X 1 µ
= (xµ )† + ls2 pµ τ + √ ((a−n )† e−in(τ −σ) + (ãµ−n )† e−in(τ +σ) ),
2 n=∓1,∓2,... n
 
† 1 µ † 1 X 0 0
µ
(Π (τ, σ)) = (p ) + √ (aµn )† ein(τ −σ ) + (ãµn )† ein(τ +σ )
2π 2π 2ls n=±1,±2,...
 
1 µ † 1 X 0 0
= (p ) + √ (aµ−n )† e−in(τ −σ ) + (ãµ−n )† e−in(τ +σ ) .
2π 2π 2ls n=∓1,∓2,...

11
Since the modes are independent, we conclude that the reality conditions
(X µ (τ, σ))† = X µ (τ, σ), (Πµ (τ, σ))† = Πµ (τ, σ)
are equivalent to the reality conditions
(xµ )† = xµ , (pµ )† = pµ , (aµn )† = aµ−n , (ãµn )† = ãµ−n .

Exercise 2.2
The vacuum state |0; ki satisfies the following identities:
ls
aµ0 |0; ki = √ k µ |0, ki = ãµ0 |0; ki , aµn |0; ki = 0 = ãµn |0; ki , n > 0.
2
From these we can show that the vacuum state satisfies the requirements to being physical:
∞
!  
1 2 2 X 1 2 2
(L0 − a) |0; ki = l k + α−k · αk − a |0; ki = ls k − a |0; ki
4s k=1
4

The mass-shell condition tells us that k 2 = −M 2 and the no-ghost theorem that a = 1.
We therefore have that (L0 − a) |0; ki vanishes if the vacuum state has a mass squared
given by
4
M2 = − 2 .
ls
If this condition is satisfied, the vacuum state is physical. For, for n > 0, we have the
following:
!
1X 1 X X
Ln |0; ki = αn−k · αk |0, ki = αn−k · αk + αn−k · αk |0; ki
2 k∈Z 2 k=0,−1,−2,... k=1,2,3,...
!
1 X X
= αk · αn−k + αn−k · αk |0; ki ,
2 k=0,−1,−2,... k=1,2,3,...

where we have made use of the fact that αiµ and αjµ commute when i 6= −j. The first
sum is equal to zero because n − k is greater than zero, and the second because k is. So
Ln |0, ki vanishes. By adding tildes in the above equations we can similarly show that if
the mass squared is given as stated above, we have the following for n ≥ 0:
(L̃n − aδn,0 ) |0; ki = 0.
We now turn our attention to the level one states, which can generally be expressed in
the following form:
µ ν
|1; ki = ζµν α−1 α̃−1 |0; ki .
We therefore have the following:
∞
!
1 2 2 X µ ν
(L0 − a) |1; ki = l k + α−k · αk − 1 ζµν α−1 α̃−1 |0; ki
4s k=1
 
1 2 2 µ ν
= − ls M + α−1 · α1 − 1 ζµν α−1 α̃−1 |0; ki .
4

12
To evaluate the middle term, the following identity is useful:
µ ν ρ σ µ ν ρ µ
αm · α1 α−1 α̃−1 |0; ki = ηρσ αm α1 α−1 α̃−1 |0; ki = ηρσ αm α−1 α1σ α̃−1
ν
|0; ki + ηρσ η µσ αm
ρ ν
α̃−1 |0; ki
ρ µ ν σ µ ρ ν µ ν
= ηρσ αm α−1 α̃m α1 |0; ki + δρ α−1 α̃−1 |0; ki = αm α̃−1 |0; ki .

Setting m = 1, we see that level one states are eigenstates of α−1 · α1 with eigenvalue.
Using this in the equation above yields the following:
 
1 2 2 µ ν 1 µ
(L0 − a) |1; ki = − ls M + 1 − 1 ζµν α−1 α̃−1 |0; ki = − ls2 M 2 ζµν α−1 ν
α̃−1 |0; ki .
4 4
We see that in order to be physical, the level one states must be massless.
We now consider the case where n = 1:
1X µ ν
L1 |1; ki = α1−k · αk ζµν α−1 α̃−1 |0; ki
2 k∈Z
!
1 X X µ ν
= α1−k · αk + α1 · α0 + α0 · α1 + αn−k · αk ζµν α−1 α̃−1 |0; ki
2 k=−1,−2,... k=2,3,4,...
µ ν
= a0 · a1 ζµν α−1 α̃−1 |0; ki

Setting m = 0 in the identity shown above, we find that:

l
L1 |1; ki = ζµν α0µ α̃−1
ν ν
|0; ki = ζµν α̃−1 α0µ |0; ki = ζµν α̃−1
ν
√s k µ |0; ki .
2
We conclude that in order to be physical the level one states must satisfy the following:

ζµν k µ |0; ki .

We now consider the case where n > 1:

1X µ ν
Ln |1; ki = αn−k · αk ζµν α−1 α̃−1 |0; ki
2 k∈Z
!
1 X X µ ν
= αn−k · αk + αn−1 · α1 + αn−k · αk ζµν α−1 α̃−1 |0; ki
2 k=0,−1,−2,... k=2,3,4,...
1 µ ν
= αn−1 · α1 ζµν α−1 α̃−1 |0; ki
2
Using the identity from above with m = n − 1 we find the following:
1 µ ν 1 ν µ
Ln |1; ki = ζµν αn−1 α̃−1 |0; ki = ζµν α̃−1 αn−1 |0; ki = 0.
2 2
By adding tildes in the above equations we find that if the level one states are massless
and satisfy the condition that

ζµν k ν |0; ki = 0,

then they obey the following:

(L̃n − aδn,0 ) |0; 1i = 0.

13
The last task that remains is to find a basis for the level two states, physical and un-
physical. The general basis for the closed string Fock space is given by:
µ1 µ ν
α−m αµ2 ...α−m
1 −m2
p
α̃ν1 α̃ν2 ...α̃−n
p −n1 −n2
q
q
|0, ki .

Since we are considering two level states, we require that

p q
X X
mi = 2 = ni .
i=1 i=1

So we require that p = 1 and m1 = 2, or that p = 2 and m1 = m2 = 1. And the same

requirement applies to q and ni . Therefore the level two states have the following basis:
µ ν
Aµν α−2 α̃−2 |0; ki ,
µ ν ρ
Bµνρ α−1 α−1 α̃−2 |0; ki ,
µ ν ρ
Cµνρ α−2 α̃−1 α̃−1 |0; ki ,
µ ν ρ σ
Dµνρσ α−1 α−1 α̃−1 α̃−1 |0; ki .

Exercise 2.3
We consider a state of the following form:

|χi = L−1 |ηi

where |ηi obeys the following equations for n ≥ 0:

Ln |ηi = 0 = (L̃n − δn,0 ) |ηi .

For any physical state |φi we have the following:

χ̄ |φi = (L−1 |ηi)† |φi = hη| L1 |φi = 0.

By using the algebra for Ln and L̃n ,

D 3
[Lm , Ln ] = (m − n)Lm+n + (m − m)δm,−n = [L̃m , L̃m ],
12

[Lm , L̃n ] = 0,

we can show that |χi is a physical state. For all n ≥ 0 we have the following:

(L̃n − δ0,n ) |χi = (L̃n − δ0,n )L−1 |ηi = L−1 (L̃n − δ0,n ) |ηi = 0,

(Ln − aδ0,n ) |χi = Ln L−1 |ηi − δ0,n L−1 |ηi

D 3

= L−1 Ln |ηi + (n + 1)Ln−1 |ηi + n − n δn,1 |ηi − δ0,n L−1 |ηi
12
= (n + 1)Ln−1 |ηi − δ0,n L−1 |ηi .

For n > 0 both terms are zero, for n = 0 both terms are equal to |χi so that they cancel.

14
Exercise 2.4
We consider the following state:
√
2 µ
|ηi = mµ α̃−1 |0; ki
ls
where k 2 = 0 = k µ mµ .
This state satisfies the following for all n > 0:
√
1X 2 µ
Ln |ηi = αn−k · αk mµ α̃−1 |0; ki = 0
2 k∈Z ls
since either n − k or k will be greater than zero so that either αn−k or αk will kill the
vacuum state. When n = 0 the same argument holds true for all terms except the k = 0
term, so that we have the following:
√ √
1 2 µ 1 k 2 ls2 2 µ
L0 |ηi = α0 · α0 mµ α̃−1 |0; ki = mµ α̃−1 |0; ki = 0
2 ls 2 2 ls
since k 2 = 0.
The state |ηi satisfies the following for n > 2:
√
1X 2 µ
L̃n |ηi = α̃n−k · α̃k mµ α̃−1 |0; ki = 0
2 k∈Z ls
since either n−k or k will be greater than one and so that either α̃n−k or α̃k will commute
µ
with α̃−1 and kill the vacuum term. When n = 2 the same argument holds true for all
terms except the k = 1 term, so that we have the following:
√
1 2 µ mµ ηρσ ρ σ µ
L̃2 |ηi = α̃1 · α̃1 mµ α̃−1 |0; ki = √ α̃1 α̃1 α̃−1 |0; ki
2 ls 2ls
mµ ηρσ ρ µ σ mµ ηρσ η σµ ρ
= √ α̃1 α̃−1 α̃1 |0; ki + √ α̃1 |0; ki = 0.
2ls 2ls
When n = 1 the above argument tells us that all terms in the sum vanish except the
k = 0 and k = 1 terms, which for n = 1 are identical, so that we have the following:
√ √
2 µ 2
L̃1 |ηi = α̃0 · α̃1 mµ α̃−1 |0; ki = mµ ηρσ α̃0ρ α̃1σ α̃−1
µ
|0; ki
ls ls
√ √ √
2 ρ µ σ 2 σµ ρ 2mµ µ
= mµ ηρσ α̃0 α̃−1 α̃1 |0; ki + mµ ηρσ η α̃0 |0; ki = α̃0 |0; ki
ls ls ls
√
2 k µ ls
= mµ √ |0; ki = 0
ls 2
µ
since k mµ = 0.
For n = 0 the terms that do not vanish by the above argument are the k = 1 and k = −1
terms, which are identical, so that we have the following:
√ √
2 µ 2 ρ µ
L̃0 |ηi = α̃−1 · α̃1 mµ α̃−1 |0; ki = mµ ηρσ α̃−1 α̃1σ α̃−1 |0; ki
ls ls
√ √
2 ρ µ σ 2 ρ
= mµ ηρσ α̃−1 α̃−1 α̃1 |0; ki + mµ ηρσ η σµ α̃−1 |0; ki
ls ls
√
2 µ
= mµ α̃−1 |0; ki = |ηi .
ls

15
We conclude that for all n ≥ 0 we have the following:
 
L̃0 − δn,0 |ηi = 0.

From exercise 2.3 we now know that the following is a spurious state:
√
1X 2 µ
|χi ≡ L−1 |ηi = α−1−k · αk mµ α̃−1 |0; ki .
2 k∈Z ls

When k is greater than zero, αk will kill the vacuum state, when k is less than minus one
α−1−k will. This leaves us with the k = 0 and k = −1 terms, which are identical, so that
we have the following:
√ √
2 µ ls 2 µ ν µ
|χi = α−1 · α0 mµ α̃−1 |0; ki = α−1 · k √ mµ α̃−1 |0; ki = kν mµ α−1 α̃−1 |0; ki .
ls l
2 s

Exercise 2.5
The generators of Lorentz transformations for the covariantly quantized closed string are
given as follows:
Z 2π
µν
J = dσ : (X µ Πν − X ν Πµ ) :
0
Z 2π !
µ 2 µ ils X 1 µ −in(τ −σ) µ −in(τ +σ)
= dσ : x + ls p τ + √ (an e + ãn e ) ·
0 2 n6=0 n
!
1 ν 1 X
p + √ (aνm e−im(τ −σ) + ãνm e−im(τ +σ) ) : −(µ ↔ ν).
2π 2π 2ls m6=0

When we multiply the two parentheses, all terms containing a single exponential integrate
to zero so that we are left with terms containing either zero or two mode factors. And
of the terms with two modes, only those for which the σ part of the exponential factors
cancel integrate to something non-zero. We therefore have the following:
Z 2π
µν

1 ν
J = dσ xµ + ls2 pµ τ p +
0 2π
Z 2π X 1 
ils 1 µ ν µ ν −2inτ µ ν −2inτ µ ν
dσ : √ √ an a−n + an ãn e + ãn an e + ãn ã−n :
0 2 2π 2ls n6=0 n
− (µ ↔ ν)
∞  
µ ν
X 1 ν µ µ ν −2inτ µ ν −2inτ
=x p + ls2 pµ pν τ +i a a + an ãn e + ãn an e ν µ
+ ã−n ãn − (µ ↔ ν).
n=1
n −n n

We note that pµ commutes with pν , αnµ with α̃nν , and α̃nµ with αnν , so that the terms with
these pair of factors cancel when we subtract the same terms with µ and ν interchanged.
We therefore conclude that
∞  
µν µ ν
X 1 ν µ ν µ
J =x p +i a−n an + ã−n ãn − (µ ↔ ν).
n=1
n

16
We can use the expression just derived to determine the algebra of the Lorentz generator:
 ∞  
µν ρσ µ ν
X 1 ν µ ν µ
[J , J ] = x p + i a−n an + ã−n ãn − (µ ↔ ν),
n=1
n
∞   
ρ σ
X 1 σ ρ σ ρ
x p +i a a + ã−m ãm − (ρ ↔ σ)
m=1
m −m m
∞  
µ ν ρ σ
X 1 ν µ σ ρ ν µ σ ρ
= [x p , x p ] − 2
[a−n an , a−n an ] + [ã−n ãn , ã−n ãn ] − (µ ↔ ν) − (ρ ↔ σ).
n=1
n
Let us consider the terms separately:
[xµ pν , xρ pσ ] = xρ [xµ , pσ ]pν + xµ [pν , xρ ]pσ = iη µσ xρ pν − iη νρ xµ pσ ,

[aν−n aµn , aσ−n aρn ] = aσ−n [aν−n , aρn ]aµn + aν−n [aµn , aσ−n ]aρn = −nη νρ aσ−n aµn + nη µσ aν−n aρn ,

[ãν−n ãµn , ãσ−n ãρn ] = −nη νρ ãσ−n ãµn + nη µσ ãν−n ãρn .

Combining these results, we have the following:
∞  
µν ρσ µσ ρ ν νρ µ σ
X 1 νρ σ µ µσ ν ρ νρ σ µ µσ ν ρ
[J , J ] =iη x p − iη x p − − nη a−n an + nη a−n an − nη ã−n ãn + nη ã−n ãn
n=1
n2
− (µ ↔ ν) − (ρ ↔ σ)
 ∞   ∞ 
µσ ρ ν
X 1 ν ρ ν ρ

νρ µ σ
X 1 σ µ σ µ


=iη x p +i a−n an + ã−n ãn − iη x p +i a−n an + ã−n ãn
n=1
n n=1
n
− (µ ↔ ν) − (ρ ↔ σ)
 ∞   ∞ 
µσ ρ ν
X 1 ν ρ ν ρ

νρ µ σ
X 1 σ µ σ µ


=iη x p +i a−n an + ã−n ãn − iη x p +i a−n an + ã−n ãn
n=1
n n=1
n
 ∞   ∞ 
νσ ρ µ
X 1 µ ρ µ ρ

µρ ν σ
X 1 σ ν σ ν


− iη x p +i a a + ã−n ãn + iη x p +i a a + ã−n ãn
n=1
n −n n n=1
n −n n
 ∞   ∞ 
µρ σ ν
X 1 ν σ ν σ

νσ µ ρ
X 1 ρ µ ρ µ


− iη x p +i a a + ã−n ãn + iη x p +i a a + ã−n ãn
n=1
n −n n n=1
n −n n
 ∞   ∞ 
νρ σ µ
X 1 µ σ µ σ

µσ ν ρ
X 1 ρ ν ρ ν


+ iη x p +i a−n an + ã−n ãn − iη x p +i a−n an + ã−n ãn
n=1
n n=1
n
 ∞ 
X 1 ν ρ
=iη µσ xρ pν + i a−n an + ãν−n ãρn − (ρ ↔ ν)


n=1
n
 ∞ 
νρ µ σ
X 1 σ µ σ µ


− iη x p +i a−n an + ã−n ãn − (µ ↔ σ)
n=1
n
 ∞ 
νσ ρ µ
X 1 µ ρ µ ρ


− iη x p +i a−n an + ã−n ãn − (ρ ↔ µ)
n=1
n
 ∞ 
µρ ν σ
X 1 σ ν σ ν


+ iη x p +i a a + ã−n ãn − (ν ↔ σ)
n=1
n −n n
=iη µσ J ρν − iη νρ J µσ − iη νσ J ρµ + iη µρ J νσ .

17
We conclude that the Lorentz generators obey the Lorentz algebra:
[Jµν , Jρσ ] = −i(ηνρ Jµσ + ηµσ Jνρ − ηµρ Jνσ − ηνσ Jµρ ).

Exercise 2.6
Let |φi be a state in the closed string Fock space satisfying the following for all n ≥ 1:
Ln |φi = 0.
Ln satisfies the extended Virasoro algebra:
D 3
[Lm , Ln ] = (m − n)Lm+n + (m − m)δm,−n .
12
Letting n be any integer greater than zero, we therefore have the following:
D 3
0 =L−n Ln |φi = Ln L−n |φi + [L−n , Ln ] |φi = Ln L−n |φi − 2nL0 |φi − (n − n) |φi .
12
If L−n |φi = 0, we must have that
D 3
2nL0 |φi = − (n − n) |φi .
12
But this equation can at most be satisfied for one n ≥ 1. So L−n |φi can at most vanish
for one n ≥ 1. We conclude that the following cannot hold true for all n ≥ 1:
L−n |φi = 0.

Exercise 2.7
We consider a set of quantum operators Ln with the following algebra:
[Lm , Ln ] = (m − n)Lm+n + g(m)δm,−n .
The quantum operators also satisfy the following reality condition:
L†n = L−n .
By taking the conjugate transpose of the algebra equation we find the following:

†
†
(m − n)L−m−n + g(m)δm,−n = Lm Ln − Ln Lm = L−n L−m − L−m L−n
= [L−n , L−m ] = (−n + m)L−m−n + g(−m)δ−m,n .
From this equation we conclude that g(−m) = −g(m).
If by m, n, and k we denote integers such that m + n + k = 0, the Jacobi identity tells
us the following:
     
0 = Lm , [Ln , Lk ] + Lk , [Lm , Ln ] + Ln , [Lk , Lm ]
     
= Lm , (n − k)Ln+k + Lk , (m − n)Lm+n + Ln , (k − m)Lk+m
=(m − n − k)L0 + g(m) + (k − m − n)L0 + g(k) + (n − k − m)L0 + g(n)
= − (m + n + k)L0 + (n − k)g(m) + (m − n)g(k) + (k − m)g(n)
=(n − k)g(m) + (m − n)g(k) + (k − m)g(n).

18
By setting k = 1 and n = −m − 1 we arrive at the following:

0 =(−m − 1 − 1)g(m) + (m + m + 1)g(1) + (1 − m)g(−m − 1)

= − (m + 2)g(m) + (2m + 1)g(1) + (m − 1)g(m + 1).

This equation is equivalent to the following recurrence relation:

(m − 1)g(m + 1) = (m + 2)g(m) − (2m + 1)g(1).

We note that
c 3
g(m) = m + km
12
satisfies the recurrence relation:
c
g(1) = + k,
12

c 
(m − 1)g(m + 1) =(m − 1) (m + 1)3 + k(m + 1)
12
c
=(m − 1) (m3 + 3m2 + 3m + 1) + k(m2 − 1)
12
c c 
=(m + 2)( m3 + km) − (2m + 1) +k
12 12
=(m + 2)g(m) − (2m + 1)g(1).

That our solution is the most general can be seen from the fact that if some other odd
function

g(m) = C1 m + C3 m3 + C5 m5 + C7 m7 ...

is to be a solution, then, by the linearity of the recurrence relation, it most also hold true
that each of m5 , m7 ... individually fulfils the recurrence relation. But that this is not
possible can be seen by plugging g(m) = mn into the recurrence relation:

(m − 1)(m + 1)n = (m + 2)mn − (2m + 1).

By equating the coefficients of the mn terms we find that for n > 2 we must have that

n − 1 = 2.

We conclude that algebra of the operators Ln must be of the form:

c 
[Lm , Ln ] = (m − n)Lm+n + m3 + km δm,−n .
12

Exercise 2.8
The Virasoro generators are given as follows:
1X
Ln = : αn−k · αk : .
2 k∈Z

19
If we consider generators Lm and Ln with m 6= 0 6= n the normal ordering makes no
difference and can be left out, and the commutator of the generators will be given as
follows:
1 XX µ ν

[Lm , Ln ] = αm−j αj µ , αn−k αk ν
4 k∈Z j∈Z

1 XX  µ ν
 ν
 µ 
= αm−j , αn−k αk ν αj µ + αn−k αm−j , αk ν αj µ +
4 k∈Z j∈Z

µ ν µ ν
 
αm−j αj µ , αn−k αk ν + αm−j αn−k [αj µ , αk ν ]

1 XX
= (m − j)η µν δm−j+n−k=0 αk ν αj µ + αn−kν
(m − j)δνµ δm−j+k=0 αj µ +
4 k∈Z j∈Z

µ ν µ ν
αm−j jδµ δj+n−k=0 αk ν + αm−j αn−k jηµν δj+k=0
 
1X
= (k − n)αk · αm+n−k − kαn−k · αm+k + (k − n)αm+n−k · αk − kαm+k · αn−k
4 k∈Z

Since the sum runs over all integers, we are free to shift k by any integer in each of the
terms. By making the shift k → k − m in the second term and fourth term we get the
following:

1X
[Lm , Ln ] = (k − n)αk · αm+n−k + (m − k)αm+n−k · αk +
4 k∈Z

(k − n)αm+n−k · αk + (m − k)αk · αm+n−k .

If we in addition to the assumption that m 6= 0 6= n also impose the requirement that

m + n 6= 0, then all the modes will commute, in which case we have the following:
1 X
[Lm , Ln ] = (m − n) αm+n−k · αk = (m − n)Lm+n .
2 k∈Z

If n = −m, however, we pick up commutator terms when replacing αk · αm+n−k with

αm+n−k · αk . In this case, lumping all the commutator terms together in some unknown
function b(m), we have the following:

[Lm , L−m ] = 2mL0 + b(m).

Since [L0 , L0 ] = 0, only one case remains for us to consider, ie. that when m 6= 0 = n:
" ∞
#
1X 1 2 2 X
[Lm , L0 ] = αm−j · αj , ls p + α−k αk
2 j∈Z 4 k=1
∞
1 XX µ ν

= αm−j αj µ , α−k αk ν
2 j∈Z k=1
∞ 
1 XX
= (m − j)η µν δm−j−k=0 αk ν αj µ + α−k
ν
(m − j)δνµ δm−j+k=0 αj µ +
2 j∈Z k=1

20
 
µ µ
αm−j jδµν δj−k=0 αk ν + αm−j ν
α−k jηµν δj+k=0
∞  
1X
= k αk · αm−k − α−k · αm+k + αm−k · αk − αm+k · α−k
2 k=1
X∞   X
= k αk · αm−k − α−k · αm+k = kαk · αm−k .
k=1 k∈Z

By changing the sum index k → m − k we find the identity

X X
kαk · αm−k = (m − k)αk · αm−k .
k∈Z k∈Z

Hence it follows that

!
X 1 X X mX
kαk · αm−k = kαk · αm−k + (m − k)αk · αm−k = αk · αm−k .
k∈Z
2 k∈Z k∈Z
2 k∈Z

And so we conclude that for m > 0 we have the following:

mX
[Lm , L0 ] = αk · αm−k = mLm .
2 k∈Z

Having now considered all cases, we find that we can combine the results in a single
equation valid in all cases:

[Lm , Ln ] = (m − n)Lm+n + b(m)δm+n=0 .

By merely placing tildes on the modes and generators in the above derivation, we obtain
the commutator [L̃m , L̃n ]. Now, from the fact that αiµ and α̃jν commute for all i, j, µ,
and ν, it follows that Lm and L̃n commute for all m and n. By combining these results
with that of exercise 2.7, we conclude that the Virasoro generators satisfy the following
algebra:
c 
[Lm , Ln ] = (m − n)Lm+n + m3 + km δm,−n ,
12
c 
[L̃m , L̃n ] = (m − n)L̃m+n + m3 + km δm,−n ,
12

[Lm , L̃n ] = 0.

Exercise 2.9
We consider a normalized vacuum state |0; ki. The vacuum state satisfies the following:

αnµ |0; ki = α̃nµ |0; ki = 0 for n > 0,

pµ = kµ |0; ki ,

21
 h0; k|0; ki = 1.

We now consider the Virasoro generator Ln when n > 0. In this case the modes commute,
and either n − k or k will be greater than zero so that either αn−k or αk kills the vacuum
state:
1X
Ln |0; ki = αn−k · αk |0; ki = 0.
2 k∈Z

Now consider the case when L0 acts on the zero-momentum vacuum state. For k > 0
αk will kill the vacuum state, and for k = 0 α0 will pull out a factor of the momentum,
which in this case is zero, so that all terms vanish:
∞
X
L0 |0; 0i = α−k · αk |0; 0i = 0.
k=0

Now consider L−1 acting on the zero-momentum vacuum state. For k > 0 αk kills the
vacuum state, for k < −1 α−k−1 kills the vacuum state, and for k = 0 and k = −1 αk
and α−1−k , respectively, will pull out a factor of the momentum, which is zero, so that
all terms vanish:
1X
L−1 |0; 0i = α−1−k · αk |0; ki = 0.
2 k∈Z

We conclude that for all n ≥ −1

Ln |0; 0i = 0.

As a consequence of this equation holding true for n = 1 and n = −1 we have the

following:
 c  c 
0 = h0; 0| [L1 , L−1 ] |0; 0i = h0; 0| 2L0 + + k |0; 0i = h0; 0| + k |0; 0i .
12 12
It follows that c = −12k.
Now consider L−2 acting on the zero-momentum vacuum state. In this case only the
k = −1 term survives:
1X 1
L−2 |0; 0i = α−2−k · αk |0; 0i = α−1 · α−1 |0; 0i .
2 k∈Z 2

We therefore have the following:

1 X
h0; 0| [L2 , L−2 ] |0; 0i = h0; 0| L2 L−2 |0; 0i = h0; 0| α2−k · αk α−1 · α−1 |0; 0i .
4 k∈Z

Since the states {|i, 0i} are orthonormal, only the k = 1 term survives:
1 1
h0; 0| [L2 , L−2 ] |0; 0i = h0; 0| α1µ α1 µ α−1
ν
α−1 ν |0; 0i = h0; 0| [α1µ α1 µ , α−1
ν
α−1 ν ] |0; 0i .
4 4
Let us consider the commutator separately:

[α1µ α1 µ , α−1
ν
α−1 ν ] =[α1µ , α−1
ν ν
]α−1 ν α1 µ + α−1 [α1µ , α−1 ν ]α1 µ + α1µ [α1 µ , α−1
ν
]α−1 ν + α1µ α−1
ν
[α1 µ , α−1 ν ].

22
Of these terms, the first and second will kill the vacuum state, while the remaining two
are both equal to α1µ α−1 µ . We therefore have the following:

1 1 1 D
h0; 0| [L2 , L−2 ] |0; 0i = h0; 0| α1µ α−1 µ |0; 0i = h0; 0| [α1µ , α−1 µ ] |0; 0i = h0; 0| ηµµ |0; 0i = .
2 2 2 2
But we also have the following identity:
 
2 2
h0; 0| [L2 , L−2 ] |0; 0i = h0; 0| 4L0 + c + 2k |0; 0i = c + 2k.
3 3

Combining the two results, we find that 3D = 4c + 12k. When combining this result with
D
our previous result that c = −12k, we find that c = D and k = − 12 . We conclude that
the Virasoro generators satisfy the following algebra:
D
m3 − m δm,−n .


[Lm , Ln ] = (m − n)Lm+n +
12

Exercise 2.10
The momentum k and coefficient ζµν (k) of the N = Ñ = 1 state of the closed bosonic
string satisfy the following:

M 2 = −k 2 = 0, k µ ζµν (k) = 0, k ν ζµν (k) = 0.

The symmetric and traceless part Gµν (k) of ζµν (k) is given as follows:

1 1
ηµν − kµ k̄ν − k̄µ kν η ρσ ζρσ (k),


Gµν (k) = (ζµν (k) + ζνµ (k)) −
2 24
where k̄ is a vector satisfying the following:

k̄ 2 = 0, k · k̄ = 1.

From this we derive the following:

1 1 1
k µ Gµν (k) = (k µ ζµν (k) + k µ ζνµ (k)) − kν η ρσ ζρσ (k) + k 2 k̄ν + k · k̄kν η ρσ ζρσ (k)


2 24 24
1 1
= − kν η ρσ ζρσ (k) + kν η ρσ ζρσ (k) = 0.
24 24
And we derive the following:
1 1
η µν Gµν (k) = (η µν ζµν (k) + η νµ ζνµ (k)) − D − k · k̄ − k̄ · k η ρσ ζρσ (k)


2 24
1
=η µν ζµν (k) − (26 − 2) η ρσ ζρσ (k) = 0.
24
The N = Ñ = 1 state of the closed bosonic string is invariant under the following gauge
transformation of the coefficient:

ζνµ (k) → ζνµ (k) + kµ mν + m̃µ kν ,

23
where

k · m = k · m̃ = 0.

Under this gauge transformation the symmetric, traceless part of the coefficient trans-
forms as follows:
1 1
ηµν − kµ k̄ν − k̄µ kν η ρσ ζρσ (k) →


Gµν (k) = (ζµν (k) + ζνµ (k)) −
2 24
1 1
ηµν − kµ k̄ν − k̄µ kν η ρσ (ζρσ (k) + kρ mσ + m̃ρ kσ )


(ζνµ (k) + kµ mν + m̃µ kν + (µ ↔ ν)) −
2 24
1 1
ηµν − kµ k̄ν − k̄µ kν (η ρσ ζρσ (k) + k · m + m̃ · k)


= (ζνµ (k) + kµ mν + m̃µ kν + (µ ↔ ν)) −
2 24
1 1
ηµν − kµ k̄ν − k̄µ kν η ρσ ζρσ (k)


= (ζνµ (k) + kµ mν + m̃µ kν + (µ ↔ ν)) −
2 24
1
=Gµν (k) + (kµ mν + m̃µ kν + kν mµ + m̃ν kµ )
2
mν + m̃ν m̃µ + mµ
=Gµν (k) + kµ + kν .
2 2
We see that there is really just one free parameter in transforming Gµν (k), so that we
can set m → (m + m̃)/2 and consider the symmetric traceless part of the coefficient to
gauge transform as follows:

Gµν → Gµν (k) + kµ mν + kν mµ with k · m = 0.

The Fourier transform of Gµν (k) can be written as follows:

Z
1
Gµν (x) = D
dD kGµν (k)eik·x .
(2π)
It follows that
Z
ρ 1
∂ρ ∂ Gµν =− dD kk 2 Gµν (k)eik·x = 0,
(2π)D
Z
µ i
∂ Gµν = dD kk µ Gµν (k)eik·x = 0,
(2π)D
Z
µν 1
η Gµν = dD kη µν Gµν (k)eik·x = 0.
(2π)D
From the gauge transform of Gµν (k) it follows that Gµν (x) gauge transforms as follows:
Z
1
Gµν (x) = dD kGµν (k)eik·x →
(2π)D
Z
1
dD k (Gµν (k) + kµ mν + kν mµ ) eik·x
(2π)D
 Z 
1 D ik·x
=Gµν (x) + d kmν kµ e + (µ ↔ ν)
(2π)D
 Z 
1 D ik·x
=Gµν (x) + ∂µ d kmν e + (µ ↔ ν) .
(2π)D

24
Now we introduce the Fourier transform of m:
Z
1
ξµ = dD kmν .eik·x
(2π)D

The facts that k 2 = 0 and that m can be chosen feely so long as k · m = 0 is then
equivalent to our being free to choose ξ provided that ∂ µ ξµ = ∂ µ ∂µ ξν = 0, and so the
gauge transform of Gµν (x) can be written as follows:

Gµν (x) → Gµν (x) + ∂µ ξν + ∂ν ξµ with ∂ µ ξµ = ∂ µ ∂µ ξν = 0.

Exercise 2.11
The Virasoro generators for the ghost fields of the closed boson string are given as follows:
X X
L(gh)
n = (n − k) : bn+k c−k :, L̃(gh)
n = (n − k) : b̃n+k c̃−k : .
k∈Z k∈Z

The modes satisfy the following anti-commutation relations:

{bm , cn } = δm,−n , {b̃m , c̃n } = δm,−n .

And the remaining anti-commutators are zero.

By making use of the identities

[AB, C] = A{B, C} − {A, C}B, [A, BC] = {A, B}C − B{A, C}

we can determine the algebra of the ghost generators: We first consider the case when
n 6= 0 6= m so that the normal ordering can be left out:
X X
[L(gh) (gh)
m , Ln ] = (m − k)
k∈Z

25