A First Course in String Theory.

Second Edition.

Solutions for problems in Part I.

The following pages contain the solutions for problems to be found in Part I of the textbook
A First Course in String Theory, Second edition. The handwritten solutions are all due to
Jeffrey Goldstone. They are clear and elegant. If you distribute some solutions to students
please let them know that they are due to Goldstone. The rest of the solutions have been
typeset. They were mostly written by me, with some help by Stefanos Marnerides and Ian
Ellwood. Some of the solutions for the newer problems are due to Alan Guth. Other’s
contributions are indicated when applicable.
This file contains the solutions for all problems in Part I of the second edition, but
some edits remain to be done on the handwritten solutions concerning the numbering of the
problems and the equations they cite in the book (both of which changed from the first to
the second edition).

Barton Zwiebach
MIT
Cambridge, MA
26 May 2009
Quick Calculation 2.2.

a′0 = γ(a0 − βa1 )

a′1 = γ(−βa0 + a1 )
a′2 = a2
a′3 = a3

b′0 = γ(b0 − βb1 )

b′1 = γ(−βb0 + b1 )
b′2 = b2
b′3 = b3

b′0 = −γ(b0 − βb1 )

b′1 = γ(−βb0 + b1 )
b′2 = b2
b′3 = b3

Then
a′µ b′µ = −γ(a0 − βa1 )γ(b0 − βb1 ) + γ(−βa0 + a1 )γ(−βb0 + b1 ) + a2 b2 + a3 b3
h i
= γ 2 a0 b0 (−1 + β 2 ) + a1 b1 (1 − β 2 ) + a2 b2 + a3 b3
= −a0 b0 + a1 b1 + a2 b2 + a3 b3 = a0 b0 + a1 b1 + a2 b2 + a3 b3
= aµ bµ .

1
Quick Calculation 2.3.
Since we are leaving the coordinates x2 and x3 invariant, we only consider boosts along the
x1 axis, for which
x′0 = γ(x0 − βx1 ) .
To get
1
x′0 = x+ = √ (x0 + x1 ),
2
we would need
1
γ = √ , and β = −1 ,
2
p
which is not possible given that γ = 1/ 1 − β 2 .

More conceptually, under Lorentz transformations we must have

−(x′0 )2 + (x′1 )2 + (x′2 )2 + (x′3 )2 = −(x0 )2 + (x1 )2 + (x2 )2 + (x3 )2

If light-cone coordinates (x+ , x− , x2 , x3 ) defined a (primed) Lorenz frame we would have

−(x+ )2 + (x− )2 + (x2 )2 + (x3 )2 = −(x0 )2 + (x1 )2 + (x2 )2 + (x3 )2 (not true)

This requires
−(x+ )2 + (x− )2 = −(x0 )2 + (x1 )2 (not true)
Indeed, a short calculation shows that

−(x+ )2 + (x− )2 = −2 x0 x1 6= −(x0 )2 + (x1 )2 .

Light-cone frames are not Lorentz frames.

1
2.1. Exercises with units
(a) By definition, two charges of one esu each placed a distance 1cm from each other, repel
with a force of magnitude 1 dyne. Let q denote the charge in Coulombs that is equivalent
to one esu. We use the equations

1 |q1 q2 | 1 N m2
|F
| = with = 8.99 × 109 2
4π0 r2 4π0 C
and since 1 dyne is equivalent to 10−5 Newtons,

N · m2 q2
10−5 N = 8.99 × 109 → q = 3.334 × 10−10 C.
C2 (10−2 m)2

From this we conclude that 1 esu is equivalent to 3.34×10−10 Coulombs, or alternatively, 1

Coulomb is equivalent to 2.998×109 esu.

(b) The fundamental temperature τ is defined from τ1 = ∂S/∂E where S is entropy (unitless)
and E is energy. So, the fundamental temperature has units of energy. The Kelvin temper-
ature T arises from τ = kB T , so [kB T ] = [E]. The constant kB can be measured once we fix
the Kelvin scale. This is done by specifying arbitrarily the value for the temperature Ttp of
the triple point of water: Ttp ≡ 273.16K. Then kB can be measured, say from P V = N kB T ,
and from measurements on a dilute gas kB  1.38 × 10−16 erg/K. Just as Coulombs always
appear with 0 , so that Coulombs always cancel in formulae, Kelvin temperature T always
appears with kB , and K cancels.

(c) Recall that [e] = M 1/2 L3/2 T −1 , [~] = M L2 T −1 , and [c] = LT −1 . Therefore [~c] =
M L3 T −2 = [e2 ] and e2 /~c is dimensionless. With e = 4.802×10−10 esu, ~ = 1.504×10−27 erg·s,
and c = 2.997 × 1010 cm/s, we find

e2 1
= 7.297 × 10−3 ≈ .
~c 137
This is the fine structure constant.

1
2.2. Lorentz Transformation for light cone coordinates.
(a) The nontrivial Lorentz tranformations for this boost are:

x
0 = γ(x0 − βx1 )

x
1 = γ(x1 − βx0 ) .

We then find
x0
± x
1 γ
x
± = √ = √ (x0 − βx1 ± x1 ∓ βx0 )
2 2
γ(1 ∓ β) 0
= √ (x ± x1 ) = γ(1 ∓ β)x± ,
2
Therefore,



1−β + 1+β −
x
+ = x , x
− = x , x
2 = x2 , x
3 = x3 . (1)
1+β 1−β

The light-cone coordinates x+ and x− do not mix under boosts along x1 !

(b)

x2

x2 /

x1 /

θ
x1

Figure 1: A rotation of coordinates.

The new, rotated coordinates are given by

x
0 = x0
x
1 = cos θ x1 + sin θ x2
x
2 = − sin θ x1 + cos θ x2
x
3 = x3 .

1
After some algebra, we find
1 1 sin θ 2
x
+ = (1 + cos θ) x+ + (1 − cos θ)x− + √ x ,
2 2 2
1 1 sin θ 2
x
− = (1 − cos θ)x+ + (1 + cos θ) x− − √ x ,
2 2 2
sin θ sin θ
x
2 = − √ x+ + √ x− + cos θ x2 ,
2 2

3 3
x =x .

(c) For a boost with velocity parameter β along x3 , the Lorentz transformations are:

x
0 = γ(x0 − βx3 )

x
1 = x1
x
2 = x2
x
3 = γ(x3 − βx0 ) .

Therefore
x
0 + x
1 γ(x0 − βx3 ) + x1
x
+ = √ = √
2 2
γ + γβx3 1
= (x + x− ) − √ + (x+ − x− ) .
2 2 2
Working similarly for x
− , x
2 , and x
3 we obtain
1 1 γβ
x
+ = (γ + 1)x+ + (γ − 1) x− − √ x3 ,
2 2 2
1 1 γβ
x
− = (γ − 1)x+ + (γ + 1)x− − √ x3 ,
2 2 2

2 2
x =x ,
γβ
x
3 = γx3 − √ (x+ + x− ) .
2

2
2.4. Lorentz transformations as matrices.

• (a) If L1 and L2 are Lorentz transformations, then LT1 ηL1 = η, and LT2 ηL2 = η. Then

(L1 L2 )T η(L1 L2 ) = LT2 LT1 ηL1 L2 = LT2 ηL2 = η ,

so the result is shown.

• (b) Using η = LT ηL we find

 
−1 T −1
(L ) ηL = (L ) L ηL L−1 = ((L−1 )T LT ) η(LL−1 ) = η
−1 TT

which states that L−1 is a Lorentz transformation.

• (c) We are given that LT ηL = η, and we need to show that LηLT = η. Taking the
transpose of the starting equation gets us nowhere, since (LT ηL)T = LT ηL. But we
can get to where we want to go by multiplying both sides of the equation by Lη

(Lη)(LT ηL) = (Lη)(η)

(LηLT )ηL = L

where we used η 2 = I (I = identity matrix). Now if we multiply both sides on the

right by L−1 η have
(LηLT )ηL(L−1 η ) = L(L−1 η )
which gives the desired result,
LηLT = η .

(Solution by Alan Guth – small edits by B.Z.)

1
2.6. Constructing the T 2/Z3 orbifold.
(a) The last corner is at
√
3 3
z =1+e iπ/3
= +i .
2 2
This point is obtained by acting with T1 (z) = z + 1 on the corner at eiπ/3 or with T2 (z) =
z + eiπ/3 on the corner at 1. As illustrated in the figure, T2 identifies OA with BC and T1
identifies OB with AC. A fundamental domain for T1 and T2 , together with its boundary,
is provided by the parallelogram OBCA. Gluing the identified parallel edges we get the
oblique torus.

(b) The figure shows the action of R(z) = e2πi/3 z (a counterclockwise rotation by 120◦ ) on
the parallelogram OACB: we get the parallelogram OA′ C ′ B ′ . Drawing the short diagonal
on the parallelograms, we see that

△OAB → △OA′ B ′ , △ABC → △A′ B ′ C ′

Now note that the unit translation T1 takes △OA′B ′ on top of △OAB but with a coun-
terclockwise rotation by 120◦ degrees. Furthermore two successive applications of T1 (a
translation to the right by a distance equal to two) takes △A′ B ′ C ′ on top of △ABC with a
counterclockwise rotation by 120◦ degrees as well. The rotations are indicated by the arrows
around P and Q (those will turn out to be fixed points of the R action).

(c) One fixed point is clearly the origin z = 0 or point O (∼ A ∼ B ∼ C). It is simply
invariant under R. The fixed point P is at a coordinate z(P ) that comes back to itself after
action of R followed by a unit translation:

z(P ) = e2πi/3 z(P ) + 1 (1)

1
We readily solve for z(P )

1 = z(P )eiπ/3 (e−iπ/3 − eiπ/3 )

π
= z(P )eiπ/3 (−i · 2 sin )
iπ/3 −iπ/2
√ 3
= z(P )e e 3

We thus find
1 1 i
z(P ) = √ eiπ/6 = + √
3 2 2 3
The other fixed point Q is invariant under the action of R followed by a translation by two
units. Its coordinate z(Q) satisfies

z(Q) = e2πi/3 z(Q) + 2 .

Comparing with (1) we see that z(Q) = 2z(P ) solves this equation. Therefore
2 i
z(Q) = √ eiπ/6 = 1 + √ .
3 3
A fundamental domain for the Z3 action on the triangle △OAB is the shaded triangle
△P AB. A fundamental domain for the Z3 action on the triangle △ABC is the shaded
triangle △AQB. So a fundamental domain for the Z3 action on the torus is given by the
quadrilateral AQBP with the edges glued as indicated by the marks on the figure. One can
imagine folding this quadrilateral over the P Q segment, gluing P B to P A, and QB to QA.
The result is the equilateral pillowcase.

2
2.7. A more general construction for cones.
The problem suggests that we experiment with low values of M and N, so let’s follow
that suggestion. The lowest two integers consistent with N > M ≥ 2 are N = 3 and M = 2,
so we’ll try those. Then the identification becomes
“ ”
2πi 32 4π
z ∼ Rz , where R = e =e3i .

This rotation takes us more than halfway around the complex plane, so let’s see what happens
if we iterate it:
8π
z ∼ R z =⇒ z ∼ R̃ z , where R̃ = R2 = e 3 i = e2πi/3 .

Thus the complicated identification z ∼ R z implies the simpler identification, z ∼ R̃ z.

It is easy to see that the implication works the other way, too, since R̃2 = R. Thus the
identification z ∼ R z is equivalent to the identification z ∼ R̃ z, which is the case that we
have already studied. The fundamental domain is the cone with 0 ≤ arg(z) < 2π 3
. In this
case the value of M was irrelevant, since the identification was the same as it would have
been if M were equal to 1.

Will we find a similar simplification for the general case, with

“ ”
2πi M
R=e N
,

where M and N are relatively prime, with N > M ≥ 2 ? Using the hint, we assume that we
have found integers m and n such that

mM + nN = 1 .

To make use of an equation that involves the product mM, we are motivated to consider
Rm , which is then given by
“ ” “ ” “ ”
2πi mM 2πi 1−nN 2πi N1
−n
R̃ = R = e
m N
=e N
=e = e2πi/N .

The argument is completed as before: the identification z ∼ R z implies the identification

z ∼ R̃ z, and the identification z ∼ R̃ z implies z ∼ R z, since R̃M = R. So the problem
reduces to understanding the identification z ∼ R̃ z, which is the cone that we have already
understood.
To fully understand this problem, of course, we need to understand the hint. Good
discussions of Euclid’s algorithm can be found on the web at

http://mathworld.wolfram.com/EuclideanAlgorithm.html

1
and
http://en.wikipedia.org/wiki/Euclidean algorithm,
and perhaps in other places as well. The algorithm appeared in Euclid’s Elements in about
300 BC, making it one of the oldest known mathematical algorithms.
Suppose we have two positive integers a and b, with b < a, and we wish to find their
greatest common divisor GCD(a, b). Consider dividing a by b, obtaining an integer quotient
q1 and a nonnegative integer remainder r1 , with r1 < b. Then we can write
a = b q1 + r1 and r1 = a − b q1 .
From the right-hand equation one can see that any integer that divides both a and b will
divide r1 , and from the left-hand equation we can see that any integer that divides both b
and r1 will also divide a. It follows that the problem of finding GCD(a, b) is equivalent to
finding GCD(b, r1 ). The second problem is easier, since the numbers are smaller. The next
step is to iterate, dividing b by r1 . Then we can write
b = r1 q2 + r2 and r2 = b − r1 q2 .
The same arguments as before imply that GCD(r1 , r2 ) = GCD(b, r1 ), and now 0 ≤ r2 <
r1 < b < a. The process cannot continue forever, since a smaller integer is obtained at each
iteration. The only way that the process can terminate is that for some N, we will find
rN = 0:
rN −3 = rN −2 qN −1 + rN −1 and rN −1 = rN −3 −rN −2 qN −1 .
rN −2 = rN −1 qN and 0 = rN −2 − rN −1 qN .
But a remainder of zero implies that rN −1 divides rN −2 , and therefore
GCD(a, b) = GCD(rN −2 , rN −1 ) = rN −1 .
The last nonvanishing remainder is the GCD, completing the algorithm.
To obtain the theorem used above, we note that if the numbers a and b are relatively
prime, then the GCD is 1, so rN −1 = 1. The right-hand equation a few lines above then
reads
1 = rN −3 − rN −2 qN −1 ,
so 1 is expressed as a linear combination of r’s, with integer coefficients. Using the right-hand
equations from the list above, each r can be expressed in terms of earlier r’s, and ultimately
one has an expansion for 1 as a linear sum of a and b, as desired.

For the numerical example suggested, 187m + 35n = 1, the right-hand equations would
look like:
187 = 5 · 35 + 12
35 = 2 · 12 + 11
12 = 11 + 1

2
so we can re-assemble
1 = 12 − 11
= (187 − 5 · 35) − (35 − 2 · 12)
= (187 − 5 · 35) − (35 − 2 · (187 − 5 · 35))
= 3 · 187 − 16 · 35 .

Thus we have a solution, with m = 3 and n = −16. The answer is not unique, because we
could, for example, take m = 3 + 35 = 38 and n = −16 − 187 = −203.

Solution by A. Guth, 2/9/07

3
2.8. Spacetime diagrams and Lorentz transformations.
Consider the boost
x′0 = γ(x0 − βx1 )
x′1 = γ(x1 − βx0 ) .
The x′0 axis is the set of points with x′1 = 0, so it corresponds to the line x1 = βx0 . In other
words, it is a line that goes through the origin and the point x0 = 1, x1 = β. This point
gives x′0 = γ(1 − β 2 ) > 0.
The x′1 axis is the set of points with x′0 = 0, so it corresponds to the line x0 = βx1 . In other
words, it is a line that goes through the origin and the point x0 = β, x1 = 1. This point
gives x′1 = γ(1 − β 2 ) > 0.
In the figure we show the old and new axes, for β > 0 (left-side) and for β < 0 (right-
side). In both cases the angle φ between the axes (x0 and x′0 , or x1 and x′1 ) has magnitude
tan |φ| = |β|. Note that |φ| ≤ π/4.

Figure 1: Oblique axes for boosts.

1
2.9. Light-like compactification.
     
x x R
∼ + 2π (1)
ct ct −R
(a) We use the above to compute
ct + x 1
x+ = √ ∼ √ (ct − 2πR + x + 2πR) = x+
2 2
ct − x 1 √
x− = √ ∼ √ (ct − 2πR − x − 2πR) = x− − 2π( 2R)
2 2
We therefore have the light-cone identifications
√
x+ ∼ x+ , x− ∼ x− − 2π( 2R) .

The compactification does not involve light-cone time.

(b) With the Lorentz transformations x′ = γ(x − βct) and ct′ = γ(ct − βx), we find

x′ ∼ γ(x + 2πR − β[ct − 2πR]) = x′ + 2πRγ(1 + β)

ct′ ∼ γ(ct − 2πR − β[x + 2πR]) = ct′ − 2πRγ(1 + β)

thus
    s  
x′ x′ 1+β R
∼ + 2π . (2)
ct′ ct′ 1 − β −R
Comparing with (1) we learn that in the boosted frame the effective radius of compactifica-
tion is increased.

Now consider the compactification

    p 2 
x x R + Rs2
∼ + 2π
ct ct −R

(c) We search for a frame (x′ , ct′ )

x′ = γ(x − βct)
ct′ = γ(ct − βx)

in which this compactification is standard. For this we need ct′ ∼ ct′ . So,
p
ct′ ∼ γ(ct − 2πR − β [ x + 2π R2 + Rs2 ] )
p
= ct′ − γ(2π)(R + β R2 + Rs2 )

1
 
Thus, we need R + β R2 + Rs2 = 0, so
R
β=− . (3)
R2 + Rs2
Now examine the x
identification to find the radius:

x
∼ γ(x + 2π R2 + Rs2 − β(ct − 2πR))

x
∼ x
+ 2πγ( R2 + Rs2 + βR) .
Simplify,
 1  R2
γ( R2 + Rs2 + βR) =  ( R2 + Rs2 −  )
1 − β2 R2 + Rs2

R2 + Rs2 R2 + Rs2 − R2
= (  ) = Rs .
Rs R2 + Rs2



x ∼ x + 2πRs . The radius of compactification is Rs . It is interesting to note that
Thus,
γ = R + Rs2 /Rs . In the limit as Rs is small, this is a very large Lorentz factor γ  R/Rs .
2

(d) The required spacetime diagram is shown in Figure 1.

ct/
ct

2 2
2π R+R s
x

- 2πR /
x

Figure 1: The heavy dots are identified so that the x
axis must go through them. The world line
of the origin of the primed system is the ct
line. This origin is moving in the negative x-direction,
as the sign of β indicates.

(e) Light like compactification with radius R arises by boosting a standard compactification
with radius Rs with Lorentz factor γ  R/Rs , in the limit as Rs → 0.

2
2.10. A spacetime orbifold in two-dimensions.

(a) From Problem 2.2 we know that under a boost with velocity parameter β along the x1
direction, x+ and x− transform as
s s
1−β + 1+β −
x+ → x and x− → x .
1+β 1−β

Since these positive scale factors are inverses of each other we may define
s
1+β
eλ ≡ , −∞ ≤ λ ≤ ∞ .
1−β

λ is called the rapidity. Given two boosts with rapidities λ1 and λ2 , the rapidity of the
combined transformation is clearly λ1 + λ1 . The range −∞ ≤ λ ≤ ∞ arises because β ∈
(−1, 1), so eλ ∈ (0, ∞). We can write the identification as

(x+ , x− ) ∼ (e−λ x+ , eλ x− ) .

For arbitrary λ the orbifold fixed point is the origin x+ = x− = 0.

(b) Below we plot curves of constant a2 for two values of a2 . Note that there are upper and
lower branches.
larger a 2

x
− x0 x
+

line of constant a 2

1
x
larger a 2

Note that the quantity x+ x− is invariant under boosts since x+ and x− transform get mul-
tiplied by e−λ and eλ respectively under a Lorentz transformation. This implies that the
identification only equates points on the same x+ x− = a2 curve.

1
  2
(c) We have x+ = a2 /x− so that dx+ = − a
x−
dx+ . This yields
 2
2 + a
−ds = −2dx dx = 2 −
(dx− )2 > 0 .
x−

Thus ds2 < 0 and the curve is spacelike.

(d) From part (c) we have that −ds2 = 2a2 ( dx )2 so the appropriate root is
−
x−

√ dx−
2a −
x
To find the invariant length of the circle we must integrate the above differential from a
given x− λ −
∗ to its immediate image e x∗ . This gives an invariant length

Z eλ x− eλ x−
√ ∗
dx− √ ∗ √ √
2a = 2a log(x− ) = 2 a log(eλ ) = 2 aλ.
x−
∗
x− x−
∗

2
2.11. Extra dimensions and statistical mechanics.
(a) The states of the system are labeled by k and l and the partition function is given by
X   2  2  2 
~ kπ l
Z(a, R) = exp −β + .
k,l
2m a R

Here k = 1, . . . , ∞, while l = 0, 1, . . . , ∞. Due to the degeneracy of the energy levels we

must include an extra factor of 2 for each term in the l sum for l > 0. Defining
X ∞   2  2 
b ~ kπ
Z(r) ≡ exp −β ,
k=1
2m r

we have that  
b b
Z(a, R) = Z(a) 1 + 2Z(πR) . (1)
b
We need to evaluate Z(r) for small β. In this case, the terms in the sum vary very slowly
with k and many of them are nearly equal to one making the total sum very large. It is thus
possible to replace the sum with an integral that runs from zero to infinity:
Z ∞  2
 2  s
b β~ kπ mr 2
Z(r) ≃ dk exp − = ≫ 1, β → 0. (2)
0 2m r 2πβ~2

For sufficiently small β, (2) applies both for r = a and r = 2πR. So (1) gives:
s s s s
ma 2 m(πR) 2 ma 2 m(2πR)2
b
Z(a, R) ≃ Z(a) b
· 2Z(πR) ≃ × 2 = .
2πβ~2 2πβ~2 2πβ~2 2πβ~2
This partition function is just the product of two familiar factors, one using the length a and
the other using 2πR. For high temperature the partition function is approximately that of
a particle in a two-dimensional box with sides a and 2πR.

(b) The stated regime implies that the thermal energy kT satisfies the inequalities
~2 ~2
≫ kT ≫ .
mR2 ma2
b
We can use the high-temperature result (2) for Z(a), b
but not for Z(πR):
s   s    
2
ma2 b ma 2 β~
Z(a, R) = 1 + 2Z(πR) = 1 + 2 exp − + ...
2πβ~2 2πβ~2 2mR2

b
The leading correction arises from the first term in the sum that defines Z(πR). This is a
“low-temperature” small correction; the argument of the exponential is large and negative.