1 Prelude

1.1 The case of quantum gravity

A classical theory of gravity is that of a dynamical metric field with di↵eomorphism invariance
or general coordinate invariance as a gauge symmetry. Such a theory may be formulated
based on an action functional, that of the Einstein-Hilbert form, possibly with a cosmological
constant as well as higher derivative couplings that involve Lorentz invariant contractions of
the Riemann tensor and its covariant derivatives.
It is widely believed that quantum mechanics is the correct framework for describing
all physical phenomena, including gravity. It is difficult to imagine how a classical theory
of gravity could be compatible with the quantum nature of all known elementary particles.
Indeed, one may consider a gedankenexperiment in which the superposition of quantum
states of a particle leads to a superposition of states of macroscopic objects that source
di↵erent gravitational fields, and conclude that gravitational fields ought to be quantized as
well.
It seems to be a good idea then to try to formulate a theory of gravity within the ordinary
quantum mechanical framework, based on a Hilbert space of physical states, Hamiltonian
time evolution, and Hermitian operators as observables. Naively, one may wish to associate
a Hilbert space with every spacelike slice of the spacetime manifold, but this is not a priori
meaningful, as the spacetime metric ought to be thought of as the expectation value of certain
field operators in a quantum state. In fact, I do not know any sensible way to formulate
a quantum theory of gravity in a given closed universe. For an open universe, however, it
may be possible to associate a Hilbert space to a certain asymptotic spacetime geometry, in
which normalizable quantum states describe local excitations.
A quantum theory of gravity should admit gravitons, namely, massless particle-like ex-
citations of spin 2, whose interactions at long distances are governed by general relativity.
One may begin with a vacuum state, which may correspond to Minkowskian spacetime in
the absence of a cosmological constant, or anti-de Sitter spacetime in the presence of a nega-
tive cosmological constant. The excited quantum states should involve the same asymptotic
geometry. In asymptotically Minkowskian spacetime, either the in-states or out-states of
gravitons as well as other possible particles should span the entire Hilbert space, and the
change of basis relating in- and out-states is the S-matrix. The S-matrix is expected to be a
well defined unitary operator, modulo the issue of infrared divergence which occurs in four
dimensional spacetime.
In asymptotically anti-de Sitter spacetime, there should be normalizable quantum states
that describe graviton and other particle excitations. The physical observables analogous

6
to the S-matrix in anti-de Sitter spacetime are transition amplitudes with sources on the
conformal boundary.

1.2 Weinberg-Witten theorem

A fundamental difficulty in formulating a theory of gravitons in the framework of a Poincaré

invariant, local quantum field theory is illustrated by Weinberg-Witten theorem, which as-
serts that the existence of local conserved stress-energy tensor operator Tµ⌫ (x) is incompatible
with the existence of massless particles with spin greater than 1.
For simplicity of the discussion let us work in 4-dimensional Minkowskian spacetime,
although the argument below can be straightforwardly generalized to higher dimensions.
We begin by considering the state |~p, hi of a single massless particle with null momentum
~ Here p0 = |~p|, p̂ is the unit vector along p~, and J~ is the
pµ = (p0 , p~) and helicity h = p̂ · J.
angular momentum vector. The normalization is such that
h~p0 , h|~p, hi = 3
(~p p~0 ). (1.1)
The matrix element of the stress-energy tensor
p0 )·x
h~p0 , h|Tµ⌫ (x)|~p, hi = ei(p h~p0 , h|Tµ⌫ (0)|~p, hi, (1.2)
by definition, obeys Z
0
h~p , h| d3~x T0µ (x)|~p, hi = pµ 3 (~p p~0 ). (1.3)

Note that the (unintegrated) matrix element (1.2) should have a continuous p~0 ! p~ limit, in
R
order for the energy-momentum density hf |T0µ (x)|f i of a wave packet |f i ⌘ d3 p~f (~p)|~p, hi
of finite width in position space to be localized in space.
Taking the p~0 ! p~ limit on (1.3) and dropping the spatial volume factor on both sides
then gives
pµ
lim h~p0 , h|Tµ 0 (0)|~p, hi = . (1.4)
0
~ !~
p p (2⇡)3
The analogous limiting matrix element of Tµ⌫ (0) can be recovered from Lorentz covariance,
0 pµ p⌫
lim h~
p , h|T µ⌫ (0)|~
p , hi = . (1.5)
~0 !~
p p (2⇡)3 p0
Now for any pair of di↵erent null momenta p, p0 , we can go to a Lorentz frame in which
p~0 = p~. The two particles of identical helicity h now have opposite angular momenta
J~0 = J~ = hp̂. Under rotation by angle ✓ around p~ axis, the states |~p, hi and |~p0 , hi
transform into
b p, hi = eih✓ |~p, hi,
R(✓)|~
(1.6)
b p0 , hi = e ih✓ |~p0 , hi.
R(✓)|~

7
The matrix element of the stress-energy tensor, on the other hand, transforms according to
b
Rµ ⇢ (✓)R⌫ (✓)h~p0 , h|T⇢ (0)|~p, hi = h~p0 , h|(R(✓)) † b p, hi = e2ih✓ h~p0 , h|Tµ⌫ (0)|~p, hi.
Tµ⌫ (0)R(✓)|~
(1.7)
⇢
Here Rµ (✓) is the corresponding rotation matrix in the vector representation of the Lorentz
group, and has eigenvalues 1, ei✓ , e i✓ . We see that the matrix element of Tµ⌫ (0) can
be nonzero only if h = 0, ± 12 , or ±1. For other values of helicity of the particle, we would
conclude that the matrix element of Tµ⌫ (0) is identically zero for all p~ 6= p~0 , which contradicts
(1.5).
This simple argument implies that a quantum theory of gravitons (helicity h = 2) can-
not admit a local stress-energy tensor operator, and thus cannot be described within the
framework of a local quantum field theory.

1.3 Yang-Mills theory at large N

In a confining nonabelian gauge theory, say the four dimensional pure Yang-Mills theory
with SU (N ) gauge group, flux tube/string states can be created by acting on the vacuum
with Wilson line operators, of the form
 Z
W (C) = TrR P exp i Aµ dxµ . (1.8)
C

Here R is a representation of the gauge group, which we shall take to be the fundamental
representation of SU (N ). P exp is the path ordered exponentiation, along a closed spacelike
path C. In this case, the flux tube is charged under a ZN -valued 1-form “center” symmetry. If
we compactify one spatial dimension, then 1-form center symmetry gives rise to an ordinary
(0-form) ZN global symmetry, and the Wilson line W (C) with C wrapping the compact
spatial direction creates a flux tube state that carries one unit of the ZN charge.
For any closed loop C, W (C) is a single trace operator. In the large N limit, with finite
’t Hooft coupling gY2 M N , or more precisely finite confining scale ⇤, standard diagrammatic
counting arguments indicate that correlation functions to leading order factorize into prod-
ucts of two-point functions of single-trace operators, with corrections suppressed by powers
of 1/N . This in particular indicates that to leading order in 1/N , the flux string cannot
split. The amplitude of a single flux string splitting into two, or emitting a glueball (thought
of as a short closed flux string), is suppressed by 1/N .
Thus, in the N ! 1 limit, we expect the flux string to behave like a free string the-
ory, in the sense that the flux strings cannot join or split. The modes of fluctuations of
the string, however, may well be interacting. On a long flux string, we expect at least
two massless Nambu-Goldstone boson fields that correspond to the spontaneously broken

8
translation symmetry in the two transverse directions. There is a low energy e↵ective the-
ory governing the interaction of these Nambu-Goldstone bosons, that further subject to the
full four-dimensional Poincaré symmetry that is partially spontaneously broken by the flux
string.
In the next chapter, we will analyze the e↵ective theory of fluctuations of the string
that respects spacetime Poincaré symmetry. To find a UV-complete quantum theory of the
flux string, even at infinite N as a free relativistic string theory, has proven to be difficult,
for reasons we will describe. However, we will see that a simple quantum theory of free
relativistic strings can be formulated in not 4, but 26 spacetime dimensions. This is known
as the critical bosonic string theory.