HIGHER-SPIN SYMMETRY

IN QUANTUM GRAVITY
AND CONFORMAL FIELD THEORY
Simone Giombi

DESY Theory Colloquium

Hamburg, May 10, 2017
Spin
• In Quantum Mechanics, particles can carry spin:
intrinsic angular momentum quantized to be integer
or half-integer multiple of Planck constant

• Particles of arbitrarily high spin (massive or

massless) theoretically possible
[Majorana ’32; Dirac ’36; Wigner ’39]
Familiar “Low-spin” particles
• Known elementary particles have small values of the spin
quantum number
Particles and fields
• In the formalism of relativistic quantum field theory,
these particles correspond to the quantum
fluctuations of fields
Higher-spin fields
• It is natural to generalize to higher spin fields
• For integer spin s, we need a symmetric traceless
rank-s tensor

• The problem of writing consistent relativistic equations

describing propagation of higher-spin fields, and their
interactions, is a non-trivial one and has a long history
[Majorana ’32; Dirac ’36; Wigner ’39; Fierz-Pauli ’39; Rarita-Schwinger ’41;
Bargmann-Wigner ’48; Fronsdal ‘78…]

Bohr and Pauli amused by a spinning top

Massive Higher-Spins from String Theory
• The spectrum of string theory
includes infinite towers of
massive higher-spin particles
• “Regge trajectories”: linear
relation M2 vs J
• In the ‘60s various mesons and
hadrons were discovered, and
observed to organize in similar
“trajectories”
• String theory started its life as a
model of strong interactions
Higher-Spins and String Theory
• Later, String Theory emerged as a leading candidate for
a consistent theory of quantum gravity, as well as all
other forces [Scherk-Schwarz; Yoneya; Green-Schwarz]
But: needs SUSY and 10 space-time dimensions
• Graviton: lowest mode of closed string. s=2, m=0
• Solves problem of UV (high energies) divergences in
quantizing gravity
• The UV finiteness of the theory is closely related to the
presence of infinite towers of higher-spin particles in the
spectrum!
Massless Higher-Spins
• For massless spinning particles of s>0, Poincare symmetry
implies they have only two polarizations (in d=4): helicity
h=+s, -s. [Wigner]
• Massless fields of spin s ≥ 1 are gauge fields. Need gauge
symmetry to decouple the unphysical polarizations
• Linearized gauge transformations
Massless Higher-Spins and String Theory
• In the small tension limit, or at very
high energies, the masses of
higher-spin states should become
negligible
• A very large higher-spin gauge
symmetry should emerge
• Current formulation of String Theory may describe a broken
phase of an underlying theory with higher-spin gauge
symmetry [Gross ’88]
• Higher-spin gauge theory should correspond to an unbroken,
maximally symmetric phase of quantum gravity/String Theory
Free massless higher-spins
• Fronsdal (1978) obtained the gauge invariant Lagrangian
and equations of motion that describe the consistent
propagation of free massless particles of any spin
• The equations are consistent in a flat, de Sitter (dS) or anti-
de Sitter (AdS) space-time. These are the maximally
symmetric solutions of Einstein equations with zero, positive
or negative cosmological constant
• Can we let these particles interact consistently?
Interactions?
• The problem of writing consistent interactions of massless
higher-spin fields is a difficult one
• Gauge symmetry severely constrains the possible
interactions
• In a flat space-time, No-Go theorems: powerful S-matrix
arguments [Weinberg; Coleman-Mandula] essentially forbid
interactions. HS gauge symmetry imply the existence of
conserved charges that are too constraining to allow a
non-trivial S-matrix
• A consistent theory of interacting massless higher-spins in
flat space probably does not exist, or has highly unusual
features (e.g. strong non-localities)
“Yes-Go”: Vasiliev theory
• However, assuming non-zero cosmological constant ,
Fradkin and Vasiliev (‘87) explicitly constructed
consistent cubic interaction vertices of higher-spin fields
• In a series of works (’90-’92), Vasiliev discovered a fully
non-linear, consistent theory of interacting massless
higher-spin fields. Non-zero cosmological constant is
essential. No smooth flat-space limit
• Originally developed in 4-dimensional space-time, but
generalizations to arbitrary dimensions were later
constructed [Vasiliev ‘03]
Vasiliev theory in 4d
• It involves a remarkable construction using ideas from
twistor theory and non-commutative field theory
Vasiliev equations

• They have a vacuum solution corresponding to anti-de

Sitter (or de Sitter) space-time
• Spectrum of the theory found by solving linearized
perturbations around vacuum. One finds

• Essentially, A encodes the metric and all other higher-spin

fields, and B contains the scalar field and the curvatures
(Weyl tensors) of the higher spin (HS) fields
Spectrum of Vasiliev theory

• Crucial feature: consistency with non-linear gauge

symmetries requires that the spectrum contains an infinite
tower of higher spin fields
• Spectrum is just like a single Regge trajectory. Much simpler
than full-fledged string theory
• A truncation to a minimal higher-spin theory with even spins
s=0,2,4,6,… is possible
• A massless spin 2 field is always part of the spectrum: the
graviton!
Vasiliev Higher-Spin Gravity
• So, Vasiliev theory may be viewed as a peculiar
theory of gravity which generalizes Einstein’s theory by
including an infinite set of massless fields of all spins
• Classical equations of motion are complicated (infinitely many
auxiliary fields), but fully known. Interactions in principle can
be read-off order by order

• Structure of the equations is almost completely fixed by the

higher-spin symmetry, up to free parameter q0
 q0 = 0: Type A Theory
 q0 = p/2: Type B Theory
 General 0<q0< p/2: Parity Breaking family of theories
Higher-Spin Quantum Gravity
• Quantization is not yet well-understood (no Lagrangian
known!), but due to the infinite dimensional gauge
symmetry, one may speculate the theory could define a
consistent, finite model of quantum gravity
• Recent evidence for one-loop finiteness of the theory:
vacuum energy is free of logarithmic divergences
[SG, Klebanov ‘13; SG, Klebanov, Safdi ‘14]

• Logarithmic divergences cancel when appropriately

regulating the sum over the higher-spin tower!
The AdS/CFT correspondence
• Higher-spin gravity in anti-de Sitter space plays an
important role in the context of the so-called
“AdS/CFT correspondence”
[Maldacena; Gubser-Klebanov-Polyakov; Witten]
• A remarkable conjecture relating theories of quantum
gravity in anti-de Sitter space to quantum field theories
in one (or more) lower dimension
• The QFT is a conformal field theory (CFT): it is invariant
under scale transformations. More generally, a CFT in
d-dimensional space-time is invariant under the
conformal group SO(d,2)
 AdS
• Anti-de Sitter space is a higher-dimensional
hyperboloid. It has constant negative curvature

 boundary

• It has the topology of a disk (or ball), and there is a notion

of a conformal boundary of the space
• The isometry group of AdSd+1 is isomorphic to the
conformal group SO(d,2) of the CFTd
AdS/CFT
• An exact equivalence, or duality, between quantum
gravity in AdS and a CFT that can be thought of as
living at the boundary of AdS

Holographic duality: quantum gravitational physics

“encoded” in the QFT at the boundary.
 AdS/CFT dictionary

• There is a general dictionary relating the physics in the

“bulk” (AdS) and “boundary”
• Operators in the CFT are dual to fields in the AdS theory
AdS/CFT dictionary
• Basic observables: correlation functions of local
operators in the CFT

“Witten diagrams”

E.g.: Cubic coupling of the graviton determines the 3-point

function of the stress energy tensor in the CFT.
• The coupling constant in the bulk is related to N,
roughly speaking the number of degrees of freedom in
the CFT. Large N corresponds to weak coupling on the
gravity side
Gauge-String duality
• The original and best understood example

• Mapping of parameters

• Large N expansion: loop expansion on string theory side

• Large l (strong coupling in gauge theory): large radius AdS,
supergravity approximation works well
• Large N, strongly coupled gauge theory  Classical Gravity
• At small l , the dual string theory becomes hard (small radius,
or small string tension)
Higher-Spin AdS/CFT
• As l  0 in the SYM theory, we expect massless
higher-spin gauge fields to emerge in the AdS dual
(tensionless strings). But spectrum is much more
complicated than that of Vasiliev higher-spin theory

• What is the CFT dual of Vasiliev higher-spin gravity in

AdS?

• Need Vector-like CFTs, rather than matrix-like!

Free O(N) vector model
• Consider a simple free theory of N massless scalar fields in 3
space-time dimensions

• This defines a free CFT

• is an N-component scalar field, transforming as a vector
under the O(N) symmetry of the action: vector model
• For applications to AdS/CFT, we focus on the O(N) singlet
sector: all operators invariant under O(N) symmetry
[Klebanov, Polyakov ’02]
• One could also consider complex fields and U(N) singlet sector
Higher-Spin Symmetry
• Easy to see that there are conserved currents of all even spins

• s=2 is the stress-energy tensor

• Exactly conserved currents  symmetries generated by
conserved charges . E.g.

• Infinite dimensional. A charge of s>2 requires whole tower

Single-trace operators
• At large N, there is a separation between single-trace and
multi-trace operators. In the vector model, the “single-trace”
operators are the O(N) invariant bilinears
• The conserved currents , together with the scalar bilinear

give the full “single-trace” spectrum

• In the well-understood AdS/CFT dictionary, single-trace
operators correspond to single-particle states in the bulk
• “Multi-trace” operators correspond to multi-particle states
Higher-spin AdS/CFT
• In AdS/CFT dictionary, conserved currents are dual to gauge
fields (bulk gauge symm.  boundary global symm.)

• In our case we get for the bulk scalar m2 = -2 (in units of

AdS radius). This is precisely the value predicted by the
Vasiliev equations
• The operator spectrum of the free vector model is
precisely in one-to-one correspondence with the spectrum
of (type A) Vasiliev theory!
Higher spins AdS/CFT
• The dual higher spin fields must be interacting in order
to reproduce the non-vanishing correlation functions of
HS currents in the CFT

• The Newton’s constant of the AdS theory scales as

GN ~ 1/N (in units of AdS radius)
• The large N limit corresponds as usual to weak
interactions in the bulk
Higher-spin AdS/CFT
• So we come at the natural conclusion that Vasiliev higher-
spin gravity in AdS should be exactly equivalent to a free
vector model CFT in one less dimension
• From AdS/CFT perspective, it is not surprising that higher-
spin gauge theory in AdSd+1 exist!
• Note that both sides of the duality are in principle accessible
perturbatively in 1/N: a “weak/weak” AdS/CFT duality
• May lead to a derivation of AdS/CFT in a simple setting
• Tests:
3-point functions [SG, Yin ‘09-’11]
Symmetry arguments [Maldacena, Zhiboedov ’11]
Matching one-loop partition functions
[SG, Klebanov ‘13; SG, Klebanov, Safdi ‘14; SG, Klebanov, Tseytlin ’14;
Gunaydin, Skvortsov, Tran ‘16; SG, Klebanov, Tan ‘16…]
The interacting O(N) vector model

• At low energies, the field theory flows to an interacting

CFT (provided mass term suitably tuned). This is the
Wilson-Fisher IR fixed point
• This fixed point is relevant for understanding the vicinity
of 2nd order phase transitions in various statistical
systems with O(N) symmetry
• N=1 corresponds to the 3d Ising CFT
• Strongly coupled in d=3. May be studied
by large N expansion, or Wilson-Fisher
epsilon-expansion
The interacting O(N) vector model
• At large N, the quartic interaction can be viewed as a “double-
trace” deformation of the free theory
• The scalar bilinear fifi has D=2+O(1/N) in the IR CFT
• Relevant entry of the AdS/CFT dictionary [Klebanov, Witten]
implies that the AdS dual of the interacting O(N) model is the
same Vasiliev theory dual to free vector model, with an
alternate choice of boundary condition for the bulk scalar
[Klebanov, Polyakov ’02]
 Type A HS gravity D=1 b.c.

 Type A HS gravity D=2 b.c.

Weakly broken higher-spin symmetry
• At large N, the higher-spin currents are almost conserved

• This implies the anomalous dimensions are

• Anomalous dimension of current correspond to masses
generated at loop level in the bulk

• The graviton (s=2) remains massless as it should (CFT stress

tensor remains conserved)
Higher-spin symmetry at finite N?
• HS symmetry is weakly broken at large N. What happens at finite
N?
• For finite N, we can use the e-expansion : formally study the theory
as a function of continuous dimension d

• Using weakly broken HS symmetry, can compute scaling

dimensions perturbatively in e near d=4 and d=2, and use
combined data to obtain estimates at d=3 (“two-sided” Pade
approximants)
HS scaling dimensions in d=3
• Combining d=4-e and d=2+e, as well as some input
from large N expansion, we obtain the following
estimates [SG, Kirilin ‘16]

• N=1 and N=2 are treated separately: NLsM cannot be

used in this case. Ordinary Pade’s give
Approximate HS symmetry?
• In all cases, anomalous dimensions appear to be very
small. Approximate HS symmetry at finite N ?
• Other quantities apparently very close to free field value.
E.g. for N=1 (3d Ising)

[SG, Klebanov ’14]

[El-Showk et al ’13] [Fei, SG, Klebanov,Tarnopolsky ’15]

• Are all these facts related to the approximate HS

symmetry? Can we use it to constrain further CFT data?
The critical Gross-Neveu model

• In 2<d<4, it defines an interacting CFT. 3d theory accessible

by large N expansion, or e-expansion near d=2 and d=4
• It also has a higher-spin gravity dual, following very similar
arguments to scalar model [Sezgin, Sundell; Leigh, Petkou]

 Type B HS gravity D=1 b.c.

(HS symm. weakly broken)

 Type B HS gravity D=2 b.c.

Chern-Simons vector models
• Interesting generalization: couple the vector models to U(N)
(or O(N)) Chern-Simons gauge theory

• Natural in AdS/CFT context: gives explicit way to implement

restriction to U(N) singlet sector
• Chern-Simons level k is quantized, does not run. We still get
CFTs even in the presence of the gauge interactions

[SG et al ‘11; Aharony et al ’11]

Chern-Simons vector models
• Look at the large N ‘t Hooft limit

• We can add quartic fermion or scalar self-interactions. This

yields generalizations of the Wilson-Fisher and Gross-
Neveu critical theories
Chern-Simons vector models
• The spectrum of “single-trace” operators is as simple as
in the ungauged versions (CS term does not add any
new local operators)

• In the critical models, the spectrum is the same, with

the dimension of the scalar operators changed as usual
due to the “double-trace” flow
Weakly broken HS symmetry
• A remarkable fact is that at large N, and for any finite ‘t Hooft
coupling, the higher-spin operators are still approximately
conserved

• So they acquire anomalous dimensions starting at 1/N order,

for any l (unlike familiar example of YM theory in d=4)

• Conjecture: the AdS duals must be the parity breaking

Vasiliev higher-spin theories with: [SG et al ‘11; Aharony et al ’11]
Weakly broken HS symmetry
• One can also show that the spin-0 bilinears have scaling
dimensions D=2+g0(l)/N+…in the CS+fermion and CS+critical
scalar, and D=1+g0(l)/N+… in the CS+scalar and CS+critical
fermion
• The weakly broken HS symmetry can be used to fix all planar
3-point functions [Maldacena, Zhiboedov; Aharony, Gur-Ari, Yacoby]

With:
Bose/fermi duality
• Note that as increase the coupling , fermion structures
are turned into (critical) scalar structures and vice-versa!
Expression mapped to each other by , or:

• Same structure of the 3-point functions is obtained from the

bulk Vasiliev theory [SG, Yin ’11; Vasiliev ‘16] . Duality manifest in
the bulk: fermionic and bosonic theory have the same bulk
dual up to a relabeling of the q0 parameter
“3d Bosonization duality”
3d Bosonization duality
• Passed various tests at leading order at large N:
 2- and 3- point functions [Aharony, Gur-Ari, Yacoby]
 Thermal free energy on R2 [Minwalla et al; Aharony, SG, Gur-Ari,
Maldacena, Yacoby]
 Relation to SUSY dualities [Gur-Ari, Yacoby]
 Explicit evaluation of HS anomalous dimensions to 1/N order
[SG, Guru-Charan, Kirilin, Prakash, Skvortsov]

• In terms of rank and level of the CS theory, this duality

amounts to a generalization of level-rank duality of the CS
theory
3d Bosonization duality

• It is plausible that this duality holds for finite N, k [Aharony]

• Abelian versions of these dualities were recently investigated
[Seiberg et al; Karch, Tong; Murugan, Nastase], and
shown to be part of
a “web” of dualities in 3d CFTs (including “particle/vortex”
duality)
• Interesting abelian examples
Conclusions
• Interacting theories of massless higher spin particles
can be constructed if the cosmological constant is non-
zero

• They involve infinite towers of fields of all spins,

including the s=2 graviton. They are, in particular,
models of (quantum?) gravity. Might be related to an
“unbroken” phase of string theory

• The Vasiliev theory in AdS is conjectured to be exactly

dual to simple vector model CFTs. Perhaps the simplest
version of AdS/CFT duality
Conclusions
• Via AdS/CFT, higher-spin theories provide exact
gravitational duals not only to free theories, but also to
interacting theories which are of interest in statistical
mechanics and condensed matter, such as the Wilson-
Fisher O(N) model, the Gross-Neveu model, and
generalizations involving Chern-Simons gauge fields

• New non-perturbative dualities, such as the “3d

bosonization”, were motivated and suggested by the
holographic duality to higher-spin gravity

• It is likely that higher-spin theories are relevant more broadly

for applications of AdS/CFT to condensed matter
THANK YOU!
 Anomalous dimensions of the HS currents
• The structure of the HS non-conservation equation can
be used, by same methods as before, to extract the
explicit expressions for the anomalous dimensions of
the weakly broken HS operators [SG, Guru-Charan, Kirilin,
Prakash, Skvortsov ’16]
• In both fermionic and scalar model, we find the result
 Anomalous dimensions of the HS currents
• The expression vanishes for s=1 and s=2 as expected
• It is precisely consistent with the bose/fermi duality
under

• Note at large spin, the anomalous dimensions now grow

logarithmically with spin, as expected in a gauge theory

• This logarithmic dependence disappears in the strong

coupling limit, where the scaling dimensions turn into
those of the critical O(k-N) (or GN) model, consistently
with the 3d bosonization duality