MASSACHUSETTS INSTITUTE OF TECHNOLOGY

Department of Physics
String Theory (8.821) – Prof. J. McGreevy – Fall 2007

Solution Set 1
Classical Worldsheet Dynamics

Reading: GSW §2.1, Polchinski §1.2-1.4. Try §3.2-3.3.

Due: Friday, September 21, 2007 at 3 PM, in the 8.821 lockbox.

1. Weyl symmetry.1 A (global) Weyl transformation acts on a field theory

coupled to gravity by
F (x) 7→ λ−W (F )F (x)
where F is a field, and W (F ) is a number called the Weyl weight of F . W is
normalized so that W (gµν ) = 2.
(a) Find W (ϕ) so that

√ 1
Z
S= dD x g g µν ∂µ ϕ∂ν ϕ
2
is Weyl invariant.

ϕ(x) 7→ λ−W (ϕ) ϕ(x)

gµν 7→ λ−2 gµν
g µν 7→ λ+2 g µν
7 λ−2D det g
det g →
√
e ≡ g 7→ λ−D e
!
S 7→ Sλ−D+2−2W (ϕ) =S
2−D
W (ϕ) =
2

1
Problems 1-4 are due to Marty Halpern.

1
 (b) Consider coordinate transformations of the form x 7→ x′ = λ−1 x, which can
be called Einstein-scale transformations. How do ϕ and gµν transform?

∂x′ µ
= λ−1 δνµ
∂xν
ϕ(x) 7→ ϕ(x′ ) = ϕ(λ−1 x)
∂xρ ∂xσ
gµν 7→ gρσ (λ−1 x) = λ2 gρσ (λ−1 x).
∂x′µ ∂x′ν
(c) Write down the transformation of ϕ and gµν under simultaneous Weyl
and Einstein-scale transformations. How does the result for ϕ relate to the
engineering dimension you would assign ϕ in order to make S dimensionless?

ϕ(x) 7→ λ−W (ϕ) ϕ(λx), gµν 7→ gρσ (λx)

Altogether, this is just an ordinary scale transformation, where ϕ
has its engineering mass dimension.

2. Local Weyl Invariance. Consider a gravity plus matter system

S = Sgrav [gµν ] + Smatter [F, gµν ]

where Smatter is invariant under the local Weyl transformation

F (x) 7→ λ(x)−W [F ]F (x), gµν (x) 7→ λ−2 (x)gµν (x).

Using the equations of motion, show that the matter stress tensor
2 δSmatter
T µν = − √
g δgµν

is traceless:
T µν gµν = 0.

The infinitesimal form of the Weyl transformation on the metric,

with λ = eα is
δweyl gµν (x) = −2αgµν (x).
Now consider a general variation of Smatter :
 
δSmatter δSmatter
Z
D
δSmatter = d x δgµν (x) + δF (x)
δgµν (x) δF (x)

2
 around a classical solution for the matter fields, this is
  Z  
δSmatter det g µν
Z
D D
d x δgµν (x) = d x − T δgµν .
δgµν (x) 2

Now in the particular case of a weyl transformation, since were as-

suming Smatter is weyl invariant,
√
Z
0 = δweyl Smatter = dD xα(x) ( gT µν gµν ) .

Since α(x) is arbitrary, this implies

T µν gµν = 0.

Note that any (globally) Weyl-invariant matter action gives a trace-

less stress tensor in the flat space limit (gµν → ηµν ).

3. Symmetries of the Polyakov action. Recall the Polyakov action

1 √
Z
SP = − d2 σ −γγ ab ∂a X µ ∂b Xµ .
4πα ′

Think of this as the action for a system of matter (X) coupled to 2d gravity
(γab ).
(a) Show that the Polyakov action is locally Weyl-invariant.
In the first problem we learned that the weyl-weight for X in D = 2
is 2−D
2
= 0, so the local weyl transformation acts as

gab (σ) 7→ λ(σ)−2 gab (σ), X 7→ X

From the transformation of the metric, we have also

√ √
g 7→ λ−2 g

g ab 7→ λ2 g ab
√ ab √
gg 7→ gg ab
so SP is locally Weyl invariant. This is something special about
strings, compared to particles or membranes or...

3
(b) Obtain an expression for the stress tensor
2 δSP
Tab ≡ − √
γ δγ ab

and check that it is traceless. [You may want to use the relation

δ(det A) = det A tr (A−1 δA),

where A is any nonsingular matrix, and δA is any small deformation.]

For a scalar field φ coupled to a background metric by the lagrangian
density
L ≡ D a φDb φ − V (φ)
with action
√
Z
S= dD x gL

the stress tensor is

2 δS
T ab = − √ = D a φDbφ − g ab L.
g δgab

Making the replacement

1 X
φ = X µ , D = 2, V = 0, SP = − S[X µ ],
4πα′ µ

we get
 
ab 1 a µ 1 ab
b µ
T =− D X D Xµ − γ ∂c X ∂d Xµ , where D a X µ ≡ γ ab ∂b X µ .
4πα′ 2

Using γ ab γab = D = 2, the trace is

 
ab 1 ab µ 2 cd µ
Θ ≡ T γab = − γ ∂a X ∂b Xµ − γ ∂c X ∂d Xµ = 0.
4πα′ 2
This was guaranteed by local Weyl invariance via the previous prob-
lem.
(c) Use the equation of motion for γab to show that when evaluated on a solution

SP [γ|EOM , X] = SN ambu−Goto [X].

4
 The equations of motion for the worldsheet metric γab are

0 = T ab .

From part (a), (and non-degeneracy of γab ) this gives

1
0 = Tab ∝ ∂a X µ ∂b Xµ − γab γ cd ∂c X µ ∂d Xµ
2
which says
 
µ 1 cd
∂a X ∂b Xµ = γab γ ∂c X µ ∂d Xµ (W ilma)
2
The LHS of this last equation (which I will call (W ilma)) is the in-
duced metric
Γab ≡ ∂a X µ ∂b Xµ .
Taking the determinant of the both hand side of (W ilma) we learn
that  2
1 cd µ
Γ ≡ det Γab = det γab γ ∂c X ∂d Xµ .
ab 2
p p 1
=⇒ |Γ| = |γ| γ cd ∂c X µ ∂d Xµ
2
which is exactly the Polyakov lagrange density. So we find
1 1
Z p Z p
2 cd µ 2
SP [γ|EOM , X] = − d σ |γ|γ ∂c X ∂d X µ | EOM = − d σ |Γ| = SN G .
4πα′ 2πα′

4. Virasoro algebra (no central extension yet). Consider the classical me-
chanics of D two-dimensional free bosons X µ , which enjoy the canonical Poisson
bracket
{X µ (σ), P µ (σ ′ )} = η µν δ(σ − σ ′ )
with their canonical momenta P µ . Show that the objects
πα′ 2 ∂σ X µ
T±± ≡ ± W± , W±µ ≡ P µ ±
2 2πα′
satisfy two commuting copies of the Virasoro algebra

{T±± (σ), T±± (σ ′ } = (T±± (σ) + T±± (σ ′ ))∂σ δ(σ − σ ′ )

5
 {T±± (σ), T∓∓ (σ ′ )} = 0.
You may use the identity
1
f (σ)f (σ ′ )∂σ δ(σ − σ ′ ) = (f 2 (σ) + f 2 (σ ′ ))∂σ δ(σ − σ ′ ), ∀f (W ilbur).
2
[This problem may be an opportunity to apply the Tedious Problem Rule
discussed in class.]
First compute
 µ 2 µν
W± (σ), W±ν (σ ′ ) = ± η ∂σδ(σ − σ ′ )

2πα′
 µ
W± (σ), W∓ν (σ ′ ) = 0

This second equation tells us immediately that

{T±± (σ), T∓∓ (σ ′ )} = 0.

For the other one, we have to work:

2
2πα′  2

′
W± (σ), W∓2 (σ ′ )

{T±± (σ), T±± (σ } =
2
2
2πα′
  
2 ′
= W± (σ) · W± (σ ) ± ∂σ δ(σ − σ ′ )
2 2πα′
πα′  2
W± (σ) + W±2 (σ ′ ) ∂σ δ(σ − σ ′ )

=±
2
where the second step used the {AB, C} = A{B, C} + {A, C}B iden-
tity enjoyed by Poisson brackets, and the last step used the identity
(W ilbur) above. Comparing with the expression for T±± gives the
other commutator.

5. Spinning strings in AdS. [This problem might be hard.2 ] In this problem

we are going to think about the behavior of a string propagating in 5d anti
de Sitter space (AdS). Specifically, we are going to study and use some of its
2
It is taken from a recent paper. For privacy’s sake, let’s call the authors Steve G., Igor K, and
Alexandre P. No, that’s too obvious. Uhh... let’s say S. Gubser, I. Klebanov and A. Polyakov. OK,
it’s hep-th/0204041.

6
conserved charges. When these conserved quantities are large, they can be
compared to their counterparts in a dual 4d gauge theory.
(a) Write down the Nambu-Goto action for a bosonic string propagating in
AdS5 , whose metric (in so-called global coordinates) is

ds2 = R2 (− cosh2 ρdt2 + dρ2 + sinh2 ρdΩ23 ) ≡ Gij dX i dX j (1)

where dΩ23 denotes the line element on the unit three-sphere,

dΩ23 = dθ12 + sin2 θ1 (dθ22 + sin2 θ2 dφ2).

1
Z q
2
SN G = − d σ − det Γab
2πα′ ab

where the induced metric Γ is

 
Ẋ · Ẋ Ẋ · X ′
Γab = ,
X ′ · Ẋ X ′ · X ′ ab

where as usual Ẋ = ∂τ X, X ′ = ∂σ X, A · B = Gij Ai B j . So

   
Γτ τ = R2 1 + sinh2 ρ θ̇12 + sin2 θ1 θ̇22 + sin2 θ2 φ̇2

etc..
(b) The AdS background has many isometries. We will focus on two: shifts of t
(the energy) and shifts of φ (the spin). These isometries of the target space are
symmetries of the NLSM, and therefore lead to conserved charges. Using the
Noether method, write an expression for the conserved charge S which follows
from the symmetry φ → φ + ǫ, with ǫ a constant. Write an expression for
the conserved charge E which follows from the symmetry t → t + δ, with δ a
constant.
The Noether method says that:
Z
δS = d2 σ∂a ǫJ a .

With δϕ = ǫ(σ, τ ), the variation of the induced metric Γab is

δΓab = R2 sinh2 ρ sin2 θ1 sin2 θ2 ∂a ǫ∂b ϕ.

7
 √ √
Using δ Γ = 12 ΓΓab δΓab ,
1
Z √
δSN G =− d2 σ ΓΓab δΓab .
4πα′
More directly, this says that
1 1
J0spin = − ′
√ [Gφφ φ̇(X ′ )2 − Gφφ φ′ Ẋ · X ′ ]
2πα det Γ
So that the spin is
Z 2π
1 1
Z
spin
S≡ dσJ0 (σ, τ = const) = − ′
dσ √ [Gφφ φ̇(X ′ )2 −Gφφ φ′ Ẋ·X ′]
0 2πα det Γ

Similarly, if δt = ǫ(σ, τ ) the variation of the induced metric Γab is

δΓab = R2 cosh2 ρ∂a ∂b ǫ.

1 1
J0energy = − √ [Gtt ṫ(X ′ )2 − Gtt t′ Ẋ · X ′ ]
2πα det Γ
′

and the energy is

Z 2π
E≡ dσJ0energy (σ, τ = const) = ...
0

We’re going to consider a spinning folded string, which spins as a rigid rod
around its center, and lies on an equator of the S 3 in (1), θ1 = θ2 = π/2. The
center of the string is at the center of AdS, ρ = 0. Go to a static gauge t = τ ,
with ρ some function of σ. Consider an ansatz for the azimuthal coordinate

φ = ωt;

this describes a string spinning around the spatial sphere.

(c) Show that with these specifications the Nambu-Goto Lagrangian becomes
Z ρ0 q
R2
L = −4 dρ cosh2 ρ − (φ̇)2 sinh2 ρ,
2πα′ 0

where coth2 ρ0 = ω 2 , and the factor of 4 is because there are four segments of
the string stretching from 0 to ρ0 .

8
 Plugging in t = τ , ...

Ẋ 2 = R2 (− cosh2 ρ + sinh2 φ̇2 )

(X ′ )2 = R2 (ρ′ )2
Ẋ · X ′ = 0.
this leaves the action
R2
Z q
S=− d σ (cosh2 ρ − φ̇2 sinh2 ρ)(ρ′ (σ))2
2
2πα′

R2
 q
 ∂ρ 
Z
S=− dτ dσ   cosh2 ρ − φ̇2 sinh2 ρ.
2πα′ ∂σ
To be closed the string must fold back on itself and traverse ρ ∈
[0, ρ0 ) (ρ0 is the maximum value of ρ the string achieves) four times
(assuming the minimal number of folds). The maximum value of ρ
can be determined by demanding that the arg of the sqrt be positive:3

coth2 ρ0 = ω 2 .

Restricting the integral to run over four segments where ρ′ > 0, we

can get rid of the absolute value around ρ′ use the chain rule, and
multiply by 4, and we get the expression given.
(d) Show that the energy and spin of this configuration are
√ Z ρ
λ 0
cosh2 ρ
E=4 dρ p
2π 0 cosh2 ρ − ω 2 sinh2 ρ
√ Z ρ
λ 0
ω sinh2 ρ
S=4 dρ p
2π 0 cosh2 ρ − ω 2 sinh2 ρ
3
Why is this the right thing to do? If we included values of ρ where the sqrt took both signs we
would find complex energies which would be hard to interpret. A nice argument that some people
like to make is that the ‘fold’ (i.e. the place where ρ′ suddenly changes sign) is like a string endpoint
(which automatically has Neumann bc’s in ρ), and therefore it should move at the speed of light
(by the same argument as for Polchinski Problem 1.2 below). This gives the condition that the arg
of the sqrt vanishes at the fold. Alternatively again, there are more general solutions where the fold
is resolved by allowing the string to move in some other direction at that point, see e.g. the nice
paper by M. Kruczenski, hep-th/0410226.

9
where
R4
λ= .
α′2
Plugging into our expression for the spin, we get

R4 ω sinh2 ρρ′
Z 2π
1 Gφφ φ̇(X ′ )2 1
Z
S=− dσ p =− dσ p
2πα′ R2 |ρ′ | cosh2 ρ − ω 2 sinh2 ρ 2πα′ 0 R2 cosh2 ρ − ω 2 sinh2 ρ

ω sinh2 ρ
Z ρ0
R2
= −4 dρ p .
2πα′ 0 cosh2 ρ − ω 2 sinh2 ρ
For the energy we get

cosh2 ρρ′
Z 2π
1 Gtt ṫ(X ′ )2 R2
Z
E=− dσ = dσ
2πα′
p
2πα′ 0
p
R2 |ρ′ | cosh2 ρ − ω 2 sinh2 ρ cosh2 ρ − ω 2 sinh2 ρ

cosh2 ρ
Z 2π
R2
=4 dσ
2πα′ 0
p
cosh2 ρ − ω 2 sinh2 ρ
etc..
Note that I seem to have a sign difference in S relative to GKP. I’m
not going to chase it down because it can be undone by reversing
the sign of ω.
(e) Next we’re going to reproduce this solution from the Polyakov action, in
conformal gauge:
1
Z
SP = dτ dσGij ∂a X i ∂ a X j .
4πα′
In this description, a solution must also satisfy the equations of motion follow-
ing from varying the worldsheet metric, namely the Virasoro constraints:

0 = T++ = ∂+ X i ∂+ X j Gij

0 = T−− = ∂− X i ∂− X j Gij
Show that inserting
t = eτ, φ = eωτ, ρ = ρ(σ)
into the Virasoro constraints gives

(ρ′ )2 = e2 (cosh2 ρ − ω 2 sinh2 ρ).

10
(Here e is a bit of slop which can be adjusted to make sure period of σ is 2π.)
This leads to an expression for dσ
dρ
.
We’ve already seen that
X ′ · Ẋ = 0
automatically in this ansatz. The other Vir constraint is

0 = (X ′ )2 + (Ẋ)2 = R2 (ρ′ )2 + e2 (− cosh2 ρ + ω 2 sinh2 ρ) ,

where e = ∂τ t is a possible mismatch between the two labelling sys-

tems. This says
e2 (cosh2 ρ − ω 2 sinh2 ρ) = (ρ′ )2
which means that
dσ 1
= p (Arthur)
dρ e cosh2 ρ − ω 2 sinh2 ρ

Show that the (target-space) energy and spin are

Z 2π
R2
E= e dσ cosh2 ρ
2πα′ 0
Z 2π
R2
S= eω dσ sinh2 ρ.
2πα′ 0

Use your expression for dσ

dρ
to show that this reproduces the answers found
using the Nambu-Goto action.
From the Noether method,
2π
1 R2
Z
Jaspin =− G ∂
φφ a φ = − dσ eω sinh2 ρ
2πα′ 2πα′ 0

2π
1 R2
Z
Jaenergy =− G tt ∂a t = − dσ e cosh2 ρ
2πα′ 2πα′ 0
Using (Arthur) to change variables reproduces the expressions from
the NG action.
(f) [super-bonus challenge] Expand the integrals for E and S at large spin
(ω = 1 + 2η, η ≪ 1), and show that in that regime
√
E − S ∼ λ ln S.

11
 The fact that this expression is reminiscent of logarithmic scaling violations in
field theory is not a coincidence; the coefficient in front of the log is called the
’cusp anomalous dimension.’
This is a fun exercise in  Taylor
 expansion. A useful fact is that in
1 1
the limit given ρ0 ∼ 2 ln η . This leads to
√ Z √ Z
2 λ ρ0 cosh2 ρ 2 λ ρ0 2 2 2


E= dρ p = dρ cosh ρ 1 + 2η sinh ρ + O(η )
π 0 1 − 4η sinh2 ρ π 0

etc.. Eventually we get

√    
λ 1 1 0
E= + ln + O(η )
2π 2η η
√    
λ 1 1 0
|S| = − ln + O(η )
2π 2η η
Taking the difference gives
√  
λ S
E − |S| = ln √ + ...
π λ
Some other problems you might consider doing are
1. Polchinski Problem 1.2. Use the Virasoro constraints to show that the
endpoints of an open string (with Neumann boundary conditions) move at the
speed of light.
The Virasoro constraints are
X ′ · Ẋ = 0, (X ′ )2 + Ẋ 2 = 0,
which (classically) should hold at every point of the worldsheet. Neu-
mann boundary conditions means that at the spatial boundaries of
the worldsheet, the derivative in the normal direction to the bound-
ary vanishes:
∂σ X µ |bdy = 0.
Evaluating the second Vir constraint on the boundary, using (X ′ )2 =
0,
0 = Ẋ 2 |bdy
which says that Ẋ µ |bdy , the velocity of the endpoint, is a null vector,
i.e. the endpoint moves at the speed of light.

12
2. Polchinski Problem 1.1. (a) Pick static gauge τ = X 0 . Non-relativistic
means Ẋ i ≡ v i ≪ 1. The point-particle action is
Z √  
1 2
Z q Z
µ 2 4
Spp = −m dτ −Ẋ Ẋµ = −m dt 1 − v = dt −m + mv + O(v )
2

which clearly displays potential and kinetic terms.

(b) Again we use static gauge, and assume small velocity Ẋ i ≡ v i ≪ 1. It’s
convenient to call w i ≡ ∂σ X i (’w’ is for ’winding’). Then the induced metric is

−1 + v 2 w · v
 
Γab = .
w·v w2

The NG Lagrangian (not density) is then

1 √ 1
Z Z p
LN G = − ′
dσ − det Γ = − ′
dσ w 2 (1 − v 2 ) + (v · w)2
2πα 2πα
 
1 1 2 1
Z
2 4
=− dσ|w| 1 − v + (v · w) + O(v )
2πα′ 2 2w 2
1 (v ẇ)2
Z
2
= −T L + T dσ|w|(v − ) + ...
2 w2
R
where L ≡ dσ|w| is the length of the string. The first term here is obviously
just the energy from the tension. The second term looks a little funny until
we realize that the motion of the string along its extent is a gauge degree of
freedom, and introduce the transverse velocity
w·v
vT = v − w.
w2
Then we have
1
Z
LN G = −T L + T dσ|w|vT2 + ...
2
and the second term is clearly the kinetic energy of an object with mass T L.

13