2 The e↵ective string and its quantization

2.1 Nambu-Goto theory

We begin by formulating a classical mechanical model of a relativistic string. Let ( 0 , 1 )

be an arbitrarily chosen non-degenerate coordinate system on the worldsheet of the string,
and X µ ( 0 , 1 ) be the embedding coordinates of the string in D-dimensional Minkowskian
spacetime. The action S[X] is a functional of X µ ( ), that is invariant with respect to
spacetime Poincaré symmetry which acts on X µ as a global symmetry, and also invariant
with respect to worldsheet reparameterization invariance which is viewed a gauge symmetry.
The simplest such action is the Nambu-Goto action
Z q
2
SNG [X] = T d det(@a X µ @b Xµ ) (2.1)

where the overall constant T is the string tension.

To illustrate the dynamics of the string, consider one that is extended along X 1 as well
as in time X 0 , that fluctuates slightly in X i directions, i = 2, · · · , D 1. We can work in
the static gauge by choosing 0 = X 0 , 1 = X 1 , and expand the Nambu-Goto action as
Z p
SNG [X] = T dX 0 dX 1 (1 @0 X i @0 X i )(1 + @1 X j @1 X j ) + (@0 X i @1 X i )(@0 X j @1 X j )
Z ⇢
T ⇥ ⇤
= dX dX 0 1
T+ (@0 X i )2 (@1 X i )2 + O((@X ? )4 ) .
2
(2.2)
i
We see that small fluctuations behave like D 2 massless scalar fields X propagating along
the string, such that long wave length modes are weakly interacting, with nonlinear e↵ects
suppressed by (@X ? )4 .
It is useful to generalize the Nambu-Goto action (2.1) to a spacetime general coordi-
nate invariant action, that describes a classical relativistic string propagating in a (possibly
curved) background metric Gµ⌫ (X),
Z q
2
S[X] = T d det(Gµ⌫ (X)@a X µ @b X ⌫ ). (2.3)

As an application, consider a rotating string in 2 + 1 dimensions. It is convenient to work

in polar coordinates, in which the spacetime Minkowskian metric takes the form

ds2 = dt2 + dr2 + r2 d✓2 . (2.4)

We will now identify the worldsheet coordinates 0 = t, 1 = r to parameterize a section of

the string. ✓ is now regarded as a function of t, r. The Nambu-Goto action now takes the

10
form Z p
S= T dtdr 1 r2 (@t ✓)2 + r2 (@r ✓)2 . (2.5)

A simple class of solution to the Euler-Lagrange equation takes the form ✓ = ✓0 +!t, describ-
ing a spinning string extended in the radial direction. Such a string cannot extend infinitely.
A possible configuration is a closed folded string that extends to r = R. Demanding that
the action is stationary with respect to variation of R determines R = 1/!. Note that the
coordinate r ranging from 0 to R describes a quarter of the folded string. Using the canonical
formalism, we can evaluate the energy and angular momentum of the solution,
Z R
dr 2⇡T
E = 4T p = ,
0 1 r !2 2 !
Z R (2.6)
r2 ! ⇡T
J = 4T dr p = 2.
0 1 r2 !2 !

Thus, we find the relation between energy and angular momentum of the folded spinning
spin,
E 2 = 4⇡T J. (2.7)
A more general rotating string solution can have ripples propagating along the radial direc-
tion that increase the total energy at a given angular momentum. A mass/energy-angular
momentum relation of the form (2.7) is known as the Regge trajectory. Its relevance in the
spectrum of hadrons was first suggested by Chew and Frautschi in 1961, hinting that certain
families of spinning hadrons may be viewed as states of spinning color flux strings.

2.2 The e↵ective string

The Nambu-Goto action is the universal action that describes motions of the classical rel-
ativistic string in a limit where the string is long, and fluctuations of the string have long
wave lengths as well. It is subject to corrections that can be described as additional terms
in the action that are symmetry invariants built out of the extrinsic curvature of the string.
We will write hab = @a X µ @b Xµ for the induced metric on the worldsheet. The extrinsic
curvature of the worldsheet may be characterized in terms of the second fundamental form,

⌦µab = ra @b X µ , (2.8)

where ra is the covariant derivative with respect to the metric hab . ⌦µab is symmetric in (ab)
and as a vector in the Minkowskian spacetime is orthogonal to the worldsheet. It obeys the
relation
⌦µac ⌦bdµ ⌦µad ⌦bcµ = Rabcd , (2.9)

11
where Rabcd is the Riemann tensor associated with hab . Note that the equation of motion
that would follow from the Nambu-Goto action can be written as

hab ⌦µab = 0. (2.10)

The action of a more general relativistic string theory may be organized according to a
derivative expansion, in which we assign weight 0 to @X and weight 1 to each additional
derivative. With this definition of weight, the Nambu-Goto action has weight zero. The first
Lorentz and reparameterization invariant terms of nonzero weight one can write down are
of the form Z
p ⇥ ⇤
d2 h ↵1 (hab ⌦µab )2 + ↵2 hac hbd ⌦µab ⌦cdµ . (2.11)

One linear combination of these two terms is proportional to the Euler characteristic,
Z
1 p
= d2 hR, (2.12)
4⇡
which is a topological term that does not a↵ect the equation of motion. Further more, any
term proportional to the Nambu-Goto equation of motion, i.e. of the form hab ⌦µab Lµ in the
Lagrangian density, can be removed by a field definition X µ / Lµ to first order. Thus, up
to a field definition, both term in (2.11) can be ignored.
In fact, the first nontrivial possible modifications to Nambu-Goto action in the derivative
R p R 2 p
expansion arise at weight 4, through terms such as d2 h (⌦µab ⌦ab 2
µ ) and d h (⌦µab ⌦ab⌫ )2 .
Note that our classical theory of the relativistic string does not allow for splitting or
joining of strings. We refer to strings that cannot split or join as free strings. In a free string
theory, the ripples on a string may well be interacting as massless fields in 1+1 dimensions,
but they cannot “leave” the string.
Before a direct attempt at quantization, there are good reasons to believe that consistent
relativistic quantum free string theories do exist, such as that of the color flux tube in SU (N )
Yang-Mills theory in the infinite N limit. In a quantum string theory, we expect to have
a Wilsonian e↵ective action on the worldsheet, that admits a derivative expansion of the
form described above. One may alternatively formulate the dynamics of the string in terms
of a 1PI e↵ective action that captures the scattering matrix of transverse fluctuations of an
(infinitely) long extended string.
If the strings can split or join, then the ripples on the string may well leave the string in
a scattering process, and the scattering matrix of fluctuations on a long string would not be
unitary by itself.

12
2.3 The Polyakov formalism

The simplest covariant string action, the Nambu-Goto action, appears to be that of inter-
acting massless scalars in the static gauge, with couplings to all derivative orders. It is not
immediately obvious whether there is a corresponding quantum theory that reduces to the
Nambu-Goto theory in the classical limit. More generally, one could also ask whether there
is a quantum theory whose Wilsonian e↵ective action is of the Nambu-Goto form plus higher
derivative corrections as described in the previous section.
A simpler route to quantization of the Nambu-Goto action exploits the worldsheet repa-
1
rameterization invariance. Defining the string tension as T = 2⇡↵ 0 , the Nambu-Goto action

is classically equivalent to the Polyakov action

Z
µ 1 p
SP [gab , X ] = 0
d2 g g ab @a X µ @b Xµ (2.13)
4⇡↵
where gab is a dynamical metric field on the worldsheet, g ab is the inverse metric and g ⌘
det(gab ). The Polyakov action admits worldsheet di↵eomorphism and Weyl rescaling gab !
e2!( ) gab as gauge symmetries. Upon imposing the equation of motion for gab , namely

4⇡ SP 1 1
Tab ⌘ p ab
= 0
@ a X µ @ b Xµ gab g cd @c X µ @d X µ (2.14)
g g ↵ 2
vanishes, which is equivalent to
gab / hab = @a X µ @b Xµ , (2.15)
SP reduces to the Nambu-Goto action SNG . In the Polyakov action, there are D free mass-
less scalar fields X µ propagating in a general worldsheet geometry. However, due to the
Weyl gauge symmetry, locally gab is gauge equivalent to a flat metric. Thus, locally on the
worldsheet we can work in the conformal gauge, by choosing a suitable coordinate system
a
in which gab = ⌘ab . At the level of classical equation of motion, this means that we can
set the action to S[⌘ab , X µ ], so that X µ are now free massless scalars on a 2D Minkowskian
worldsheet, subject to the constraint Tab = 0 which follows from the gab equation of motion.
The latter is known as the Virasoro constraint. In other words, the classical dynamics of the
Nambu-Goto string is equivalent to the conformal gauge equations
@ a @a X µ = 0, Tab = 0. (2.16)

2.4 A first attempt at quantization

Formally one may proceed to quantize the Polyakov action by writing down the path integral
representation of the partition function
Z
µ
Z = [Dgab DX µ ] e SP [gab ,X ] , (2.17)

13
where we now work in Euclidean signature on the worldsheet, with the Euclidean version of
the Polyakov action Z
1 p
SP = 0
d2 g g ab @a X µ @b Xµ . (2.18)
4⇡↵
(2.17) as is written is ill defined, due to the di↵eomorphism and Weyl invariance of SP .
To proceed, we shall fix the gauge invariance by setting gab = ĝab for an arbitrarily chosen
fiducial metric ĝab . This is possible locally on the worldsheet. We will return to global issues
later.
An infinitesimal di↵eomorphism corresponding to a vector field v a together with an
infinitesimal Weyl transformation acts on gab by

gab = ra vb rb va + 2 !gab . (2.19)

The (partially) gauge fixed form of the path integral is now written as
Z
µ
Zĝ = [DX µ ]ĝ FP [ĝab ] e SP [ĝab ,X ] , (2.20)

where FP [ĝab ] is the Faddeev-Popov determinant,

Z Z p
Sgh [ĝ,e 1
FP [ĝab ] = [Debab Dca D⌘]ĝ e b,c,⌘]
, Sgh [ĝ, eb, c, ⌘] = d2 ĝ ebab (r̂a cb + r̂b ca 2ĝ ab ⌘).
4⇡
(2.21)
e a a
Here bab , c and ⌘ are Grassmannian fields. c and ⌘ are ghosts that transform like the gauge
parameters v a , !, whereas ebab is the anti-ghost that transforms like the stress-energy tensor
Tab . r̂a is the covariant derivative defined with respect to the fiducial metric ĝ, that acts on
cb as a vector field. We can integrate out ⌘ and write
Z Z p
a Sgh [ĝ,b,c] 1
FP [ĝab ] = [Dbab Dc ]ĝ e , Sgh [ĝ, b, c] = d2 ĝ bab r̂a cb , (2.22)
2⇡
where bab is now a symmetric traceless tensor field, i.e. ĝ ab bab ⌘ 0.
A consistent gauge fixing would require Zĝ to be invariant under gauge transformation of
ĝ. While the actions SP [ĝ, X] and Sgh [ĝ, b, c] are indeed di↵eomorphism and Weyl invariant,
it is not obvious that the measure [DX µ Dbab Dca ]ĝ would be, due to the need for UV regular-
ization. In fact, while it is possible for the regulator to preserve di↵eomorphism invariance,
it is generally not Weyl invariant. Under ĝab ! e2! ĝab , Zĝ transforms by
⇢ Z
D 26 p ⇥ ab ⇤
Ze2! ĝ = Zĝ exp d2 ĝ ĝ @a !@b ! + !R(ĝ) , (2.23)
24⇡
where D is the target spacetime dimension, i.e. the number of X µ fields, and R(ĝ) is the
Ricci scalar of the fiducial metric. We will leave the derivation of this important result to
section 3.6.

14
 We see that if D = 26, Zĝ is Weyl invariant and the quantization is consistent in the
sense that it respects the gauge symmetries of the Polyakov action. The resulting worldsheet
theory is a free quantum field theory of the scalar fields X µ and the ghost fields bab , ca . This
results in the critical bosonic string theory, which we discuss in section 2.5.
If D 6= 26, the Weyl anomaly presents a problem: either we must accept that the Weyl
mode of the worldsheet metric is a dynamical field, which introduces extra degrees of freedom
in the worldsheet theory, or we must modify the Polyakov action to compensate for the Weyl
anomaly. The former option results in the so-called noncritical string theory, which we will
return to in chapter 6. The latter option would be the natural choice if we insist on looking for
a D-dimensional Poincaré invariant string theory. Indeed, we could add to SP the “composite
Liouville action”
Z
26 D p ⇥ ⇤
SCL [gab , '] = d2 g g ab @a '@b ' 'R(g) , (2.24)
24⇡
where ' is any scalar operator constructed out of X µ and gab that has Weyl transformation
of the form
' ! ' + !, (2.25)
SCL
so that under the Weyl transformation e transforms by a factor that cancels against
(2.23). The simplest choice of ' is
1
'= log(g ab @a X µ @b Xµ ). (2.26)
2
Formally, then, we have a gauge fixed path integral
Z
Z = [DX µ ]ĝ FP e SP SCL
(2.27)

Note that SSL is a highly nonlinear interacting action in X µ , and it is not immediately evident
whether such a UV complete quantum theory exists. Nonetheless, if such a worldsheet indeed
exists, it should admit a Wilsonian e↵ective action of the form SP +SCL plus higher derivative
corrections.
Let us examine the composite Liouville term (2.24) in some more detail. In the conformal
gauge gab = ⌘ab in Lorentzian signature, it can be written as
Z
26 D (@a @b X µ @ b Xµ )(@ a @c X ⌫ @ c X⌫ )
SCL = d2 . (2.28)
24⇡ (@d X ⇢ @ d X⇢ )2
A static long string extending in X 1 direction can be described by the classical field con-
figuration1 X 0 = ⌧ , X 1 = , with the other components X i set to zero, i = 2, · · · , D 1.
1
Here it is viewed as a solution to the equation of motion, not to be confused with the static gauge
condition.

15
Expanding around this configuration, we can write
Z
26 D
SL = d2 (@a @b X i @ b X i )(@ a @c X j @ c X j ) + · · · , (2.29)
96⇡
where · · · represents higher derivative terms that involve products of more than four fields.
The quartic term exhibited in (2.29) gives a nontrivial contribution to the 2 ! 2 S-matrix
element of massless fluctuations of the long string,
out
hk, q3 ; `, q4 |i, q1 ; j, q2 iin = ik j` (q1 q3 ) (q2 q4 ) + i` jk (q1 q4 ) (q2 q3 )
i (26 D)↵02 3 (2.30)
+ (2⇡)2 2 (q1 + q2 q3 q4 ) s ij k` + t3 ik j` + u3 i` jk + · · · ,
s 192⇡
where qi ⌘ (qi⌧ , qi ) are the momentum of the 4 massless excitations, with qi⌧ = |qi |, and
s = (q1 + q2 )2 , t = (q1 q3 )2 , u = (q1 q4 )2 are the Mandelstam variables in 1+1
dimensions. Note that for elastic 2 ! 2 scattering in 1 + 1 dimensions, the kinematic
constraints are such that one of s, t, u must be zero.
One must be cautious in translating between scattering amplitudes of massless quanta
on a long string in the conformal gauge versus static gauge, as the notion of space and time
variables are di↵erent in the two gauges. This leads to a nontrivial phase di↵erence, known as
“gravitational dressing”, between the long string S-matrix in the two gauges. For instance,
it explains why for D = 26 the worldsheet theory is free in the conformal gauge, but there is
nontrivial purely elastic scattering of the long string excitations in the static gauge, as one
can already seen from the tree level Nambu-Goto theory.
However, the gauge di↵erence does not account for the interaction due to SCL , for D 6= 26.
As we have seen in section 2.2, there are no weight 2 e↵ective couplings analogous to (2.29)
one could add to the Nambu-Goto action that preserve the target space Poincaré symmetry.
In fact, one can show that in the Nambu-Goto formalism, the nontrivial contribution to the
long string S-matrix element as in the second line of (2.30) arises as a 1-loop amplitude in
the ↵0 expansion.2

2.5 Critical string theory

Let us now focus on the D = 26 case, where our quantum string theory is defined by the
path integral (2.20) that is independent of the choice of fiducial metric ĝab , at least locally
2
Such a computation requires a UV regularization scheme that respects target space Poincaré symmetry,
and the inclusion of counter terms that respect the symmetry. One such scheme is dimensional regularization.
If we take the worldsheet dimension to be d = 2 ✏, the 1-loop 2 ! 2 amplitude in Nambu-Goto theory
would also receive a log divergence proportional to stu, which can be canceled by a d-dimensional Einstein-
Hilbert counter term that is no longer topological for away from d = 2. When restricted to d = 2, the term
proportional to stu is identically zero, and one is left with a finite term as in the second line of (2.30).

16
on the worldsheet. The space of quantum states of a closed string can be constructed out
of states of the 1+1 dimensional free quantum field theory consisting of X µ , bab , and ca on
a circle. However, the presence of ghost fields gives rise to negative norm states. A proper
account of the physical states requires consideration of BRST cohomology. We will defer the
discussion of BRST formalism to chapter 4.
For now, we would like to find an alternative gauge choice in which the FP ghosts are
non-dynamical. This is achieved through the light cone gauge
1
⌧ = X + ⌘ p (X 0 + X 1 ), g= 1, g = g ⌧ ⌧ = 1. (2.31)
2
Under di↵eomorphism and Weyl transformation, the gauge fixing conditions vary by

X+ = v a @a X + = v⌧ ,
1
g g = g ab gab = 2ra v a + 4 !, (2.32)
g = 2 !g 2r v .

Let ca and ⌘ be the FP ghosts associated with a and !. The ghost action is such that
upon integrating out the b-ghosts associated with X + and g, c⌧ is set to zero and ⌘ is
set to be equal to 12 r c . The remaining ghost coupling does not contain a kinetic term,
therefore the ghost fields are not dynamical and can be omitted in the light cone gauge.
Now the Polyakov action takes the form
Z
1
SP = d⌧ d 2g ⌧ a @a X + g ab @a X i @b X i , (2.33)
4⇡↵0

where X ⌘ p12 (X 0 X 1 ), and the index i is summed from 2 to D 1. On a closed string

worldsheet, is periodic. A priori, the periodicity of may be ⌧ -dependent, but as we will
see below the circumference of the string in our gauge is a conserved quantity. In any case,
we can decompose X (⌧, ) into its Fourier modes in . Integrating out the nonzero Fourier
modes of X sets
@ g ⌧ = 0. (2.34)
Note that (2.31) still leaves a residual di↵eomorphism gauge transformation of the form
( , ⌧ ) ! ( 0 , ⌧ 0 ), with
0
= + f (⌧ ), ⌧ 0 = ⌧. (2.35)
We can now use this transformation to set g ⌧ = 0, and so gab = ⌘ab .
The zero Fourier mode of X in is canonically conjugate to
I
+ 1
p = p = d . (2.36)
2⇡↵0

17
 1
Note that this agrees with jaµ = 2⇡↵ 0 @a X
µ
being the worldsheet Noether current associated
with target spacetime translation symmetry, and the corresponding conserved charge pµ =
H
d j⌧µ is the spacetime momentum. The Noether charge associated with translation in X + ,
in particular, is given by
I I
1 1 ⇥ i 2 i 2
⇤
p = d @ ⌧ X = d (@ ⌧ X ) + (@ X ) , (2.37)
2⇡↵0 4⇡↵0
where in the second step we used the equation of motion for g ⌧ ⌧ .3 This expression coincides
with the worldsheet Hamiltonian in the light cone gauge.
Now we can construct the Hilbert space of a single closed string state as that of the D 2
free bosons X i , subject to a residual gauge redundancy of translation symmetry in . The
latter amounts to a gauge constraint on the physical Hilbert space, namely that the states
must be translation invariant in .
The Fock space of the free bosons X i on the -circle is spanned by states of the form
|pi ; {nik , n
eik }k 1 i, (2.38)
where nik is the number of massless quanta of X i that carry k units of momenta each around
the circle, and neik is the number of quanta that carry k units of momenta. pi labels the
eigenvalue of the conserved charge associated with target space translation current jai , namely
I
i 1
p = d @⌧ X i . (2.39)
2⇡↵0
The worldsheet energy of the state (2.38) is
"D 1 #
(pi )2 1 XX
p = + 0 + (nik + n
eik ) + A . (2.40)
2p+ ↵p i=2 k 1

In deriving this we used (2.36) which says that p+ is the string tension times the circumference
H
` = d in the light cone gauge. A is a constant that appears in the oscillator ground
state Camisir energy E0 = 2⇡A `
. A standard calculation of the ground state energy, after
subtracting a UV divergence that can be canceled by a local counter term in the free boson
field theory, gives A = D12 2 .
Our gauge condition still leaves a residual gauge transformation given by (2.35) with
constant f . The corresponding gauge constraint amounts to the condition that the total
worldsheet momentum (translation charge in ) vanishes, namely
D
X1 X D
X1 X
N⌘ nik = eik .
n (2.41)
i=2 k 1 i=2 k 1

3
To be more precise, after gauge fixing g ⌧ ⌧ in the quantum theory, the equation of motion for g ⌧ ⌧ is not
zero but a BRST exact operator, whose expectation value on physical states vanishes.

18
We will refer to N as the (total) oscillator level. (2.40) can now be written in the manifestly
Lorentz invariant form ✓ ◆
µ 2 4 D 2
p pµ = m = 0 N . (2.42)
↵ 24
So far, in the light cone gauge quantization, we have not made explicit use of D = 26, which
as asserted was essential for the preservation of Weyl invariance in the quantum theory. The
Weyl anomaly for D 6= 26 manifests itself in the light cone gauge as a Lorentz anomaly.
In particular, the Lorentz generators J i built out of the corresponding Noether currents
do not quite obey the Lorentz algebra relation [J i , J j ] = 0. A simpler way to see the
Lorentz anomaly is to note that the level N = 1 states with pi = 0 transform in v ⌦ v of
the SO(D 2) that rotates the X i ’s, where v is the vector representation. If D 6= 26, the
N = 1 states have nonzero mass m, and must organize into a representation of the massive
little group SO(D 1) that contains SO(D 2) as a subgroup. But no representation of
SO(D 1) reduces to v ⌦ v with respect to the SO(D 2) subgroup. The only way for
this to be consistent with D-dimensional Lorentz symmetry is for the N = 1 states to be
massless, which requires D = 26.
Indeed, the D = 26 critical bosonic string theory is a consistent quantum theory at the
level of free strings. It will not be a consistent quantum theory of interacting strings due to
the presence of tachyons at oscillator level N = 0. Nonetheless, we will see later that the
interacting critical bosonic string theory makes sense as a classical string field theory in 26
dimensional spacetime.

19