Prepared for submission to JHEP

A Conformal Dispersion Relation:

Correlations from Absorption
arXiv:1910.12123v2 [hep-th] 3 Jul 2020

Dean CarmiΨ Simon Caron-Huotσ

Ψ Institute
of Physics, École Polytechnique Fédérale de Lausanne (EPFL), Rte de la Sorge,
BSP 728, CH-1015 Lausanne, Switzerland
σ Department of Physics, McGill University, 3600 Rue University, Montréal, QC Canada
H3A 2T8
E-mail: [email protected], [email protected]

Abstract: We introduce the analog of Kramers-Kronig dispersion relations for

correlators of four scalar operators in an arbitrary conformal field theory. The cor-
relator is expressed as an integral over its “absorptive part”, defined as a double
discontinuity, times a theory-independent kernel which we compute explicitly. The
kernel is found by resumming the data obtained by the Lorentzian inversion formula.
For scalars of equal scaling dimensions, it is a remarkably simple function (elliptic
integral function) of two pairs of cross-ratios. We perform various checks of the dis-
persion relation (generalized free fields, holographic theories at tree-level, 3D Ising
model), and get perfect matching. Finally, we derive an integral relation that relates
the “inverted” conformal block with the ordinary conformal block.
Contents
1 Introduction 2

2 Preliminaries 5
2.1 Review of amplitude dispersion relation 5
2.1.1 Dispersion relation from the Froissart-Gribov formula 7
2.2 Review of CFT kinematics 8
2.2.1 CFT dispersion relation from Lorentzian inversion formula 11

3 Computing the CFT dispersion relation kernel 11

3.1 Performing the ∆ integration in d = 2 12
3.1.1 The Contact term KC 13
3.1.2 The Bulk term KB 14
3.2 Main result from Legendre PPPQ sum 15
3.3 Match with d = 4 and differential equation 17
3.3.1 Contact term 18
3.3.2 Bulk term 18
3.4 Differential equation for unequal scaling dimensions 19

4 Direct proof of dispersion relation 22

4.1 Contour deformation trick 23
4.1.1 Why two variables? 26
4.2 Convergence and subtractions 27
4.2.1 Subtracted dispersion relation 27
4.2.2 Keyhole contour near cross-channel singularity 28

5 Checks and discussion 29

5.1 Numerical check for generalized free fields 29
5.2 Holographic correlators 30
5.3 An integral relation between conformal blocks 32
5.4 3D Ising model and analytic functionals 32

6 Conclusion 35

A Identities for spin sums 36

1
B Inverted block from harmonic function when a = 0, b = 2
37

–1–
1 Introduction
The conformal bootstrap has enjoyed remarkable success in the last decade, employ-
ing both numerical [1–4] and analytic [5–8] methods to solve general consistency con-
ditions. Some of the primary methods of the analytic bootstrap include: light-cone
expansions of the crossing equations, large N expansions, AdS/CFT, and causality
constraints.
Implications of causality are often effectively captured by dispersion relations,
following the work of Kramers and Kronig in optics. These authors related (in
1926) the dispersive (real) and absorptive (imaginary) part of the index of refraction,
exploiting analyticity of the index of refraction in the upper-half complex frequency
plane. Dispersion relations were later used to try and constrain the relativistic S-
matrix [9–11]. This was an important tool for physicists in the 1950’s and 60’s who,
in the absence of a microscopic theory, attempted to solve or “bootstrap” the strong
interactions using consistency with the principles of causality, unitarity, and crossing;
a program which waned down at the time with the advent of QCD as a microscopic
description of the strong force.
Dispersion relations are typically most useful when one knows more about the
absorptive part than the real part. For the strong force at low energies, the imaginary
part is often saturated by narrow resonances, leading to phenomenologically interest-
ing sum rules [12]. It may also happen that the imaginary part (or the absolute value
of the amplitude) is the only quantity measured experimentally. Theoretically, the
imaginary part enjoys useful properties such as positivity (for example in the forward
limit), related to probabilities being nonnegative; applications include the first proof
of irreversibility of renormalization group flow in four spacetime dimensions [13]. In
perturbative scattering amplitudes, absorptive parts can be efficiently computed in
terms of lower-order amplitudes through the Cutkosky rules, a foundational insight
that is now built into successful methods such as generalized unitarity [14–17]. Given
that crossing symmetry and general principles appear to be particularly powerful in
conformal field theories, it is natural to expect a CFT dispersion relation to be a
useful tool in constraining CFT correlators.
In this paper we derive a dispersion relation for CFT 4-point correlators G(z, z̄):
Z 1
t
G (z, z̄) = dwdw̄K(z, z̄, w, w̄)dDisc[G(w, w̄)] (1.1)
0

where we separate the t and u channel contributions and a possible finite sum of
non-normalizable blocks (see section 4.2.1):

G(z, z̄) = G t (z, z̄) + G u (z, z̄) + (non-norm.) . (1.2)

The input dDisc[G(z, z̄)] represents the double-discontinuity of the correlator around
z̄ = 1, defined below, and is interpreted physically as its absorptive part. We notice

–2–
that the correlator is a function of two cross-ratios (z, z̄): the kernel K(z, z̄, w, w̄)
is thus a function of two pairs of cross-ratios, one pair being integrated over (with
w, w̄ real in the integration region). This is to be contrasted with more familiar
Kramers-Kronig type dispersion relations, in which a single variable is integrated
over. We will argue that such a complication is unavoidable if we insist that the
input be the “absorptive part” dDisc[G], as the analytic properties of the correlators
G(z, z̄) entangle its two arguments.
The existence of a formula such as (1.1), reconstructing correlators from (dou-
ble) discontinuities, is suggested by the Lorentzian inversion formula of [18–20]. That
formula reconstructs operator product expansion data from knowledge of the discon-
tinuities dDisc[G(z, z̄)] of the CFT 4-point correlator, and has been used notably
to streamline light-cone and large-N expansions. Examples suggest that a crude
approximation to the dDisc (ie. including the simplest few exchanged operators)
can lead to accurate results to the OPE data itself. These examples range from the
low-twist spectrum in 3D Ising and related models [21–25], mean field theory [26],
the calculation of Witten diagrams in strongly coupled (holographic) gauge theories
[27, 28], as well as defect CFTs and certain finite temperature effects [29, 30].
We find it is extremely encouraging that good first approximations to the dDisc
are easy to come by. This begs the question of systematic improvement. One lim-
itation of the Lorentzian inversion formula is that it is difficult to iterate it. For
example, its output cannot simply be fed back into it, in a way that would lead to
successively better approximations (while the formula produces a generating function
for the spectrum, computing the dDisc requires resolving the dimensions of individ-
ual operators, a step which requires a numerically difficult analytic continuation).
The dispersion relation (1.1) offers a step forward, since it enables crossing equations
to be formulated directly on the positive dDisc. As we will see, it will also circumvent
technical limitations regarding convergence at low spins.
In this paper we derive the dispersion relation (1.1), and in particular the kernel
K entering it, by resumming the OPE data extracted via the Lorentzian inversion
formula. The result can be split into a two-dimensional bulk integral KB and a
one-dimensional contact integral KC 1 :

dρw
K(z, z̄, w, w̄) = KB θ(ρz ρ̄z ρ̄w − ρw ) + KC δ(ρw − ρz ρ̄z ρ̄w ) (1.3)
dw

where θ(x) is the unit step function and δ(x) is the Dirac δ-function. In the case of

√
1 1−√1−z
The ρ-variables, defined in eq. (2.13), is: ρz ≡ 1+ 1−z
, and similarly for z̄, w, w̄.

–3–
operators of equal external scaling dimensions, our main result is the explicit form:
1  z z̄ 3/2 (w̄ − w)( w1 + w̄1 + z1 + z̄1 − 2) 3
1 3
KB = − 3 x 2 F1 ( 2 , 2 , 2, 1 − x),
2
64π ww̄ ((1 − z)(1 − z̄)(1 − w)(1 − w̄)) 4
1/2 (1.4)
1 − ρ2z ρ̄2z ρ̄2w 1 − ρz ρ̄z ρ̄2w

4 1
KC = .
π w̄2 (1 − ρ2z )(1 − ρ̄2z )(1 − ρ̄2w ) (1 − ρz ρ̄w )(1 − ρ̄z ρ̄w )
The first involves a rather special combination of cross ratios:
ρz ρ̄z ρw ρ̄w (1 − ρ2z )(1 − ρ̄2z )(1 − ρ2w )(1 − ρ̄2w )
x≡ . (1.5)
(ρ̄z ρ̄w − ρw ρz )(ρz ρ̄w − ρw ρ̄z )(ρz ρ̄z − ρw ρ̄w )(1 − ρw ρz ρ̄w ρ̄z )
The bulk integral contributes only for ρw < ρz ρ̄w ρ̄z (due to the step function), and
is proportional to a hypergeometric function, which can equivalently be written as
a combination of elliptic integral functions, see eq. (3.24). The contact integral,
proportional to a δ-function, is effectively integrated over a single variable w̄ ∈ [0, 1].
An alternative but equivalent form, which unites the bulk and contact terms, is given
in eq. (4.5).
We find it remarkable that a function of four complex variable can be written
in closed form as in eq. (1.4). As we will see in section 4, each factor plays a role,
and K above is arguably the simplest possible kernel able to fulfil the difficult task
assigned to it.
The outline of the paper is as follows. In section 2.1 we review the amplitude dis-
persion relation and the Froissart-Gribov inversion formula, and how one can derive
the former from the latter. This exercise will prepare us for the more difficult case of
the CFT dispersion relation. In section 3 we show the full details of derivation of the
CFT dispersion relation in d = 2 for scalars with equal external scaling dimensions.
We obtain an analytic result for the kernel, in terms of elliptic integral functions. The
same kernel is valid in any dimension, and we show that in section 3.3 that indeed
repeating the calculation in d = 4 yields the same kernel. In section 3.4 we derive
the dispersion relation for unequal external scaling dimension. The kernel satisfies a
differential equation, giving Taylor expansions for it. For a specific simple case, a = 0
and b = 12 , we also find an analytic form for the kernel. In section 4 we establish the
validity of the dispersion relation by a direct contour deformation argument. This
allows to overcome some of the original assumptions, and in particular we obtain
a subtracted dispersion relation that is valid in any unitary CFT. In section 5 we
explore possible applications of the dispersion relation: to strong coupling N = 4
SYM, to obtain novel identities relating inverted and conformal blocks, and to the
3D Ising model and new bootstrap functionals. We conclude by discussing future
directions in section 6.

Note added: While this paper was being completed, the work [31] appeared on arxiv
who introduced a single-variable dispersion relation that reconstructs correlators

–4–
from a single-discontinuity. This appears to be quite distinct from the formulas
considered here: the input in this case (to our knowledge) is neither sign-definite nor
admits a physical interpretation as an absorptive part.

2 Preliminaries
2.1 Review of amplitude dispersion relation
Dispersion relations enable to construct a function from a knowledge of it’s dis-
continuities. The most common type of a dispersion relation is the single variable
dispersion relation, where one variable is being integrated over. For definiteness, we
will discuss this here in the context of the relativistic 4-particle scattering amplitude,
although the reader may wish to keep in mind that the construction is more general.
We will review two derivations, the first involving a contour deformation argument
which is perhaps the most familiar.

Figure 1. Left: The 4-particle scattering amplitude M(s, t), with external momenta pi .
Right: t-channel tree level exchange diagram of particle with spin J .

Consider the 4-particle scattering amplitude M(s, t) for scalars with mass m
(fig. 1 left). M(s, t) is a function of the two Mandelstam variables s ≡ −(p1 + p2 )2
and t ≡ −(p1 − p3 )2 , with the energy conservation constraint s + t + u = 4m2 . For s
constant and in a suitable range, the complex t-plane has the structure depicted in
fig. 2, with two branch cuts along the real axis for t > t0 and t < 4m2 − s − t0 . These
are called the s- and t-channel cuts (the second condition corresponding to u > u0 ).
The single variable dispersion relation to be considered is:
Z ∞
1 dt0
M(s, t) = Disct0 [M(s, t0 )] + (t ↔ u) (2.1)
2π t0 t0 − t
The integral runs over the branch cuts of M(s, t0 ), and i Disct0 [M(s, t0 )] ≡ M(s, t0 +
i0) − M(s, t0 − i0) is the discontinuity across the cuts in the t0 -plane. Note that the
variable s just goes along for the ride2 .
2
One could alternatively write a dispersion relation in the s-plane, with fixed t.

–5–
 t

Figure 2. Left: The amplitude can be written as a contour integral by using Cauchy’s
theorem. Right: Upon deforming the contour, there will be contributions from the branch
cuts and from the arcs at infinity.

A common way to derive this is to start with a contour integral in the complex
0
t plane surrounding the point t (see Fig 2); by Cauchy’s residue theorem:
dt0
I
1
M(s, t) = M(s, t0 ). (2.2)
2πi t0 − t
Then one deforms the contour of integration as in Fig 2. If M(s, t0 ) decays fast
enough at |t0 | → ∞ that the arcs at infinity can be neglected, only the branch cuts
contribute, reproducing eq. (2.1) as desired.
What if M(s, t0 ) does not decay fast enough? If it is polynomially bounded, one
can still obtain a subtracted dispersion relation. The idea is to improve the behavior
on large arcs by subtracting the amplitude at some reference t = t∗ :
Z ∞  
1 0 1 1
M(s, t)−M(s, t∗ ) = dt 0 − Disct0 [M(s, t0 )]+u-channel, (2.3)
2π t0 t − t t0 − t∗

which has improved convergence since the bracket ∼ 1/|t0 |2 . One can generalize by
applying more subtractions as needed. A perhaps more illuminating way to write
t−t∗
this is to use elementary algebra to rewrite the bracket as (t0 −t)(t0 −t ) , and divide both
∗
sides by (t − t∗ ); the once-subtracted dispersion relation (2.3) becomes:
Z ∞
dt0 M(s, t0 )
 
M(s, t) 1
= Disct0 + u-channel. (2.4)
t − t∗ 2π t0 t0 − t t0 − t∗
This is nothing but the original dispersion relation, now applied to the rescaled
function3 M(s, t)/(t − t∗ ). This viewpoint will be useful below.
In the amplitude context, the dispersion integral generally runs over unphysical
regions of the (s, t0 )-plane, where Disct0 M is neither positive-definite nor physically
measurable. An exception is for the range 0 ≤ s < 4m2 in a theory with mass gap
3
In this form, we have assumed that t∗ is inside the integration range, so that the discontinuity
contains a term −2πδ(t0 − t∗ )M(s, t∗ ).

–6–
m: there the discontinuity is positive-definite, and is a smooth extrapolation (to
imaginary angles) of physically measurable t- and u-channel scattering amplitudes.
This is an important result of Martin, used in his celebrated proof of the Froissard
bound on the high-energy growth of total cross-sections [32]. The CFT dispersion
relation discussed in this paper will share the nice features of this special region.
For more on applications of scattering amplitude dispersion relations, the reader
may consult [9–11]. For a more recent application, see the following works on the
S-matrix bootstrap [33, 34].

2.1.1 Dispersion relation from the Froissart-Gribov formula

We turn to a perhaps less familiar derivation of the dispersion relation, starting from
the Froissart-Gribov formula expressing partial wave coefficients from the disconti-
nuity of the amplitude.
Consider the partial wave decomposition of the amplitude in the s-channel (Fig. 1
right), for definiteness working in d = 4 dimensions:
∞
1X 4m2 − s
M(s, t(z)) = (2J + 1)aJ (s)PJ (z), z ≡ cos θ, t(z) = (1 − z).
2 J=0 2
(2.5)
Physically, θ is the scattering angle and the coefficients aJ (s) encode the decompo-
sition of the amplitude into spherical harmonics at a given energy-squared s.
R1 2δJJ 0
Using the orthogonality of the Legendre polynomials, −1 dzPJ 0 (z)PJ (z) = 2J+1 ,
one may readily obtain a “Euclidean inversion formula” expressing the coefficients
as an integral over the amplitude. A less obvious formula, first derived by Froissart
and Gribov [35, 36], expresses the same data in terms of the discontinuity of the
amplitude:

1 ∞ 0
Z
t
aJ (s) = dz QJ (z 0 ) Disct M(s, t(z 0 )), aJ (s) = atJ (s) + (−1)J auJ (s), (2.6)
π 1

where QJ (zs0 ) is the Legendre function of the second kind. atJ (s) and auJ (s) are
the contributions from the t-channel and u-channel cuts respectively. The Frossart-
Gribov formula plays a foundational role in Regge theory, as it establishes analyticity
in spin of the partial waves (as well as providing quantitative large-spin estimates).
A proof of eq. (2.6) starts from the orthogonality relation, rewriting the integral
over z ∈ [−1, 1] as a contour integral using that PJ ∝ Disc QJ . One then deforms the
contour exactly as in fig. 2 above (see [18] for recent discussion with two derivations).
The Froissart-Gribov formula and dispersion relation are thus closely related, and it
should come as no surprise that one can derive either one from the other.
To go the other way, the trick is simply to plug the coefficient obtained from
eq. (2.6) into the partial wave sum in eq. (2.5), and interchange the summation and

–7–
integration:
∞ Z ∞
X 1
M(s, t(z)) = (2J + 1)PJ (z) dz 0 QJ (z 0 )Disct M(s, t(z 0 )) + (t ↔ u)
J=0
2π 1
Z ∞ ∞
1 0 0
X
= dz Disct M(s, t(z )) (2J + 1)PJ (z)QJ (z 0 ) + (t ↔ u). (2.7)
2π 1 J=0

The latter sum then turns into the following identity (for |z| < |z 0 |): 4

∞
X 1
(2J + 1)PJ (z)QJ (z 0 ) = (2.9)
J=0
z0 −z

which is recognized as the kernel of the dispersion relation (2.1). (One needs only
0 dt0
the change of variable zdz
0 −z 7→ t0 −t .)
0
We call the measure tdt 0 −t , which multiplies the discontinuity, the “kernel”. In-

terestingly, even though the form of the special functions PJ and QJ changes in
a complicated way as a function of spacetime dimension, and the left-hand-side of
eq. (2.9) acquires a measure factor [(1 − z 02 )/(1 − z)2 ](d−4)/2 , one can show that the
sum produces the same right-hand-side in any dimension. This was to be expected
physically, since the dimension simply did not enter the earlier derivation anywhere.
The reader may wonder why one would want to derive a dispersion relation
starting from the Froissart-Gribov formula (2.6), as opposed to simply writing down
0
the more elementary Cauchy kernel tdt 0 −t . The reason is that the substitution PJ 7→

Disc QJ underlying the former has a group-theoretical explanation (ie. both func-
tions satisfy the same Casimir differential equation), whereas writing down the
Cauchy kernel requires an educated guess. For conformal correlators, the group-
theoretic approach was successfully carried out in ref. [18], whereas the guessing
approach turns out to be much more challenging.

2.2 Review of CFT kinematics

In this paper we will focus on a correlator of four scalar primary operators in a CFT.
This can be written as a function of cross ratios z and z̄ multiplied by an overall
factor which is determined by the conformal symmetry:
 2 a  2 b
x14 x14
x224 x213
hO1 (x1 )O2 (x2 )O3 (x3 )O4 (x4 )i = 1 1 G(z, z̄) , (2.10)
(x212 ) 2 (∆1 +∆2 ) (x234 ) 2 (∆3 +∆4 )
4
This can be proved by combining the following two equations:
∞ 1
dz 0 PJ (z 0 )
Z
1X 1
(2J + 1)PJ (z)PJ (z 0 ) = δ(z − z 0 ), and QJ (z) = . (2.8)
2 2 −1 z0 − z
J=0

The latter shows that PJ (z) equals the discontinuity across the cut of QJ (z).

–8–
where we defined the differences of the external scaling dimensions:

a = 21 (∆2 − ∆1 ), b = 21 (∆3 − ∆4 ) (2.11)

and the cross ratios z, z̄ are defined through:

x212 x234 x223 x214
u = z z̄ = , v = (1 − z)(1 − z̄) = . (2.12)
x213 x224 x213 x224
We will often use the so-called radial or ρ-coordinates of ref. [37],
√ √
1− 1−z 1 − 1 − z̄ 4ρz 4ρ̄z
ρz ≡ √ , ρ̄z ≡ √ , z= 2
, z̄ = (2.13)
1+ 1−z 1 + 1 − z̄ (1 + ρz ) (1 + ρ̄z )2
which provide a double cover of the complex z-plane.
We will be focusing on the s-channel operator product expansion (OPE):
X
G(z, z̄) = f12OJ,∆ f43OJ,∆ GJ,∆ (z, z̄) (2.14)
J,∆

where fijO are the OPE coefficients and GJ,∆ (z, z̄) are s-channel conformal blocks
for exchange of a primary operator with spin J and scaling dimension ∆, and its
descendants. For our purposes the OPE may also be written as an integral over
principal series representations (harmonic functions), in which the scaling dimension
is continuous (see [38–40]):
∞ Z d
+i∞
X 2 d∆
G(z, z̄) = c(J, ∆)FJ,∆ (z, z̄) + (non-norm.). (2.15)
J=0
d
2
−i∞ 2πi

The “non-normalizable” part includes the s-channel identity operator as well as a

possible finite sum of F functions for scalar operators with dimension less than d/2.
The CFT data is then encoded in the poles of cJ,∆ , which occur on the real axis of
the complex ∆ plane at the position of the physical scaling dimensions, and whose
residue are the squared OPE coefficients:

f12OJ,∆ f43OJ,∆ = −Res∆0 =∆ c(J, ∆0 ) .

 
(2.16)

The F stand for harmonic functions, which combine a block and its shadow
1 
FJ,∆ (z, z̄) = GJ,∆ (z, z̄) + #GJ,d−∆ (z, z̄) , (2.17)
2
with a specific coefficient that will not be important below. (It ensures that F is
single-valued in Euclidean space where z̄ = z ∗ , a necessary condition for the F ’s to
form a complete orthogonal basis.)
Using the orthogonality for FJ,∆ (z, z̄), one may readily write an Euclidean inver-
sion formula expressing the OPE data cJ,∆ as an integral over correlators, in analogy

–9–
to that for Legendre polynomials discussed below eq. (2.5). Instead we will use the
Lorentzian inversion formula, which reconstructs the same data from an “absorptive
part” [18–20]
Z 1
t κJ+∆
c (J, ∆) = dwdw̄ µ(w, w̄) G∆+1−d,J+d−1 (w, w̄) dDisc[G(w, w̄)] (2.18)
4 0

where the integration region is the square 0 ≤ w, w̄ ≤ 1, the normalization and

measure are

Γ( β2 − a)Γ( β2 + a)Γ( β2 − b)Γ( β2 + b) − w)(1 − w̄))a+b

w − w̄ d−2 ((1
κβ = , µ(w, w̄) = ,
2π 2 Γ(β − 1)Γ(β) ww̄ (ww̄)2
(2.19)
and the OPE data itself is the sum of t- and u-channel contributions (as in eq. (2.6)):

c(J, ∆) = ct (J, ∆) + (−1)J cu (J, ∆). (2.20)

The u-channel contribution may be obtained by applying the integral (2.18) to the
correlator with operators 1 and 2 swapped.
Notice that the conformal block G∆+1−d,J+d−1 (w, w̄) appearing in the inversion
formula above is not the usual block, it has the roles of J and ∆ reversed; we may
call it the “inverted block”. (This reversal is a Weyl reflection of the so(d, 2) Lie
algebra.) This is analogous to the substitution PJ 7→ QJ in eq. (2.6). One can draw
a close analogy between between 4-point CFT correlators and 4-particle amplitude
scattering amplitudes, see Table 1.
The “dDisc” is primarily defined as a expectation value of the double-commutator
1
− 2 h0|[O2 , O3 ][O1 , O4 ]|0i, divided by the normalization factor in eq. (2.10). It can be
computed as a double discontinuity, or difference between three analytic continua-
tions, around the point z̄ = 1:

dDisc [G(ρ, ρ̄)] ≡ cos(π(a+b))G(ρ, ρ̄)

− 21 eiπ(a+b) G(ρ, ρ̄−1 −i0) − 21 e−iπ(a+b) G(ρ, ρ̄−1 +i0) (2.21)

where we assume 0 < ρ, ρ̄ < 1. This represents a discontinuity since ρ̄z and ρ̄−1z map
onto the same cross-ratio z̄, see eq. (2.13).
Physically, the dDisc is interpreted as an absorptive part because it represents
one minus the survival probability of a certain state. In particular it is positive-
definite by unitarity, see section 2.2 of [18]. In holographic theories, the double dis-
continuity effectively puts bulk propagators on-shell (as seen in specific tree and one-
loop examples, see ref. [27]), furthering the analogy with DiscM and the Cutkowski
rules. The idea that it is sometimes easier to approximate the dDisc than the cor-
relator itself, as reviewed in introduction, motivates us to try and reconstruct the
correlator itself from this data.

– 10 –
 R∞ 0
M(s, t) CJ (cos θ) QJ (cos η) Mt (s, t) = t0 tdt
0 −t Disc[M(s, t)]
t
R1
G(z, z̄) FJ,∆ (z, z̄) G∆+1−d,J+d−1 (z, z̄) G (z, z̄) = 0 dwdw̄K(z, z̄, w, w̄)dDisc[G(w, w̄)]

Table 1. Analogous quantities between the 4-particle scattering amplitude (top row) and
the CFT 4-point correlator (bottom row). The right most column shows the dispersion
relation.

2.2.1 CFT dispersion relation from Lorentzian inversion formula

Given the formula which extracts OPE data from the absorptive part (dDisc) in
eq. (2.18), it is only natural to insert it back into the OPE to obtain a dispersion
relation for the correlator itself. This is the procedure which led in subsection 2.1.1
to a dispersion relation for scattering amplitudes. We thus plug eqs. (2.20) and (2.18)
inside eq. (2.15):

G(z, z̄) = G t (z, z̄) + G u (z, z̄) + (non-norm.) (2.22)

where
∞ Z
t
X d∆ κJ+∆
G (z, z̄) = FJ,∆ (z, z̄)
J=0
2πi 4
Z 1
× dwdw̄ µ(w, w̄) G∆+1−d,J+d−1 (w, w̄) dDisc[G(w, w̄)] (2.23)
0

and similarly for G u (z, z̄). Exchanging the order of integrals and sum then gives a
dispersion relation in the form quoted in eq. (1.1), that is:
Z 1
t
G (z, z̄) = dwdw̄K(z, z̄, w, w̄)dDisc[G(w, w̄)], (2.24)
0

where the kernel is now given explicitly as:

∞ Z d
+i∞
µ(w, w̄) X 2
K(z, z̄, w, w̄) = d∆ κJ+∆ FJ,∆ (z, z̄)G∆+1−d,J+d−1 (w, w̄). (2.25)
8πi J=0 d
2
−i∞

This is a key formula, and the main goal of this paper will be to evaluate this kernel
K(z, z̄, w, w̄) explicitly.5 The integrand consists of the Euclidean harmonic function
F , times the inverted block G and times κ∆+J (the latter turns out to be crucial).

3 Computing the CFT dispersion relation kernel

In this section we analytically perform the sum-integral (2.25), thus obtaining the
kernel of the dispersion relation. A few observations will simplify this endeavour:
5
The kernel K reported in introduction equals K here upon symmetrization in w ↔ w̄.

– 11 –
 • We expect the kernel K to be independent of space-time dimension, because
eq. (2.24) is a mathematical identity which should hold for any two-variable
function G(z, z̄) satisfying certain analyticity properties (that are dimension-
independent). Indeed this is what happened in eq. (2.9) for the amplitude
dispersion relation. We will thus now set d = 2, where the blocks are simpler,
and verify in subsection 3.3 that the same result is obtained in d = 4.

• In a generic CFT, the integral (2.18) only converges to the OPE data for large
enough spin. Even for a unitary theory, it may fail for J = 0 and/or J = 1.
It is unclear how to improve the Lorentzian inversion formula to reach these.
Our strategy will be to first glibly ignore this issue and assume convergence.
After the kernel is obtained, in the next section (see 4.2.1) we will extend its
validity by means of a subtraction.

• We will first perform the sum assuming identical external operator dimensions;
this will require rather nontrivial identities. We will then realize that the
agreement between the d = 2 and d = 4 sums amount to interesting differential
equations, which will largely explain the form of the result and help attack the
general case.

3.1 Performing the ∆ integration in d = 2

Our first step to compute (2.25) is to perform the ∆ integral. The idea, as shown
in Fig. 3, is to close the contour and use the residue theorem to get a sum over the
residues of the poles. We will need the explicit form of conformal blocks in d = 2:

k∆−J (z)k∆+J (z̄) + k∆+J (z)k∆−J (z̄)

GJ,∆ (z, z̄) = ,
1 + δJ,0 (3.1)
β
β β
kβ (z) ≡ z 2
2 F1 ( 2 + a, 2
+ b, β, z).

Plugging into eq. (2.25), this yields two terms for the block G(w, w̄), and four terms
for F (z, z̄) the average of block and shadow. We can use the w↔w̄ symmetry of the
correlator to remove one of the former, and shadow symmetry of the other factors
to neglect the shadow symmetrization, reducing the number of terms to 2:6

g 1+i∞
Z
µ(w, w̄) X
K(z, z̄, w, w̄) = d∆ κJ+∆ k∆−J (z)k∆+J (z̄)kJ−∆+2 (w)k∆+J (w̄)+(z↔z̄).
4πi J 1−i∞
(3.2)
From now on until subsection 3.4 we consider the case of equal external scaling
dimensions: a = b = 0. We close the integration contour in the ∆ plane with a semi-
circle at |∆| → ∞, fig. 3. The integrand of eq. (3.2) has the following asymptotic
6
P P∞ AJ (z,z̄)
For conciseness we define the J sum with a tilde as fJ AJ (z, z̄) ≡ J=0 1+δJ,0 .

– 12 –
Figure 3. Left: The original integration contour of the principle series representation.
One can close the contour either to the left or to the right, depending on the behaviour of
the integrand at |∆| → ∞. The integrand has poles on the real ∆ axis.

behaviour as |∆| → ∞:
1+ J J
−J J ∆
16 ρw 2 ρ̄w2 ρz 2 ρ̄z2 (ρz ρ̄z ρ−1
w ρ̄w )
2
κJ+∆ k∆−J (z)k∆+J (z̄)kJ−∆+2 (w)k∆+J (w̄) → p .
π (1 − ρ2w )(1 − ρ̄2w )(1 − ρ2z )(1 − ρ̄2z )
(3.3)
−1
From this we see that when the cross-ratios are such that ρz ρ̄z ρw ρ̄w = 1, the ∆
integral is divergent and the kernel will have a contact term proportional to a delta
function. We compute this contact term in the next subsection. Otherwise, the
magnitude of ρz ρ̄z ρ−1
w ρ̄w determines whether we close the ∆-contour to the left or
to the right. Thus we expect our kernel to contain a step function as well, ie. both
“bulk term” and “contact terms” as in eq. (1.3), which we reproduce for convenience:
dρw
K(z, z̄, w, w̄) = KB θ(ρz ρ̄z ρ̄w −ρw )+KC δ(ρw −ρz ρ̄z ρ̄w )+K∅ θ(ρw −ρz ρ̄z ρ̄w ). (3.4)
dw
The notation K∅ anticipates that the third term vanishes.

3.1.1 The Contact term KC

Performing the ∆ integral using the asymptotics in eq. (3.3) gives a delta function:
Z 1+i∞
1 ∆
d∆(ρ−1
w ρz ρ̄z ρ̄w )
2 = δ(ρ
−1
w ρz ρ̄z ρ̄w − 1). (3.5)
4πi 1−i∞
The J sum from eqs. (3.2) and (3.3) is then simply a geometric sum:

g 1+ J2 J2 − J2 J2
X 1 − ρz ρ̄z ρ̄2w
ρw ρ̄w ρz ρ̄z + (z↔z̄) = ρz ρ̄z ρ̄w (3.6)
J
(1 − ρz ρ̄w )(1 − ρ̄z ρ̄w )

where we have used the constraint from the δ-function to eliminate ρw from the
result. Combining eqs. (3.2)-(3.6) gives the result for the contact term of the kernel:
16 1 ρ2w δ(ρw − ρz ρ̄z ρ̄w ) 1 − ρz ρ̄z ρ̄2w
K⊃ p . (3.7)
π (ww̄)2 (1 − ρ2w )(1 − ρ̄2w )(1 − ρ2z )(1 − ρ̄2z ) (1 − ρz ρ̄w )(1 − ρ̄z ρ̄w )

– 13 –
Dividing by the δ-function and Jacobian dρ w
dw
included in eq. (3.4), this gives the
formula recorded in the introduction, namely:
1/2
1 − ρ2z ρ̄2z ρ̄2w 1 − ρz ρ̄z ρ̄2w

4 1
KC (z, z̄, w̄) = . (3.8)
π w̄2 (1 − ρ2z )(1 − ρ̄2z )(1 − ρ̄2w ) (1 − ρz ρ̄w )(1 − ρ̄z ρ̄w )
The notation choice (3.4), with the Jacobian factored out, allows to directly integrate
out w, leaving a single integral over w̄:
Z 1
t
G (z, z̄) = dw̄ KC (z, z̄, w̄) dDisc[G(w, w̄)] . (3.9)
contact 0 ρw =ρz ρ̄z ρ̄w

3.1.2 The Bulk term KB

We now move on to compute the kernel when ρz ρ̄z ρ−1 w ρ̄w 6= 1. From eq. (3.3) we
see that when the cross-ratios are in the regime ρz ρ̄z ρ−1w ρ̄w > 1, we can close the
contour to the left (ie. Re(∆) < 1), and the contribution from the arc at infinity
will give zero. Likewise, when ρz ρ̄z ρ−1
w ρ̄w < 1 we close the contour to the right (ie.
Re(∆) > 1) in order to drop the contribution from the arc at infinity, fig. 3.
Now we use the residue theorem to compute the ∆ integral as a sum over residues
of all the poles of the integrand of eq. (3.2). Each one of the four hypergeometric
functions k’s has a tower of poles, and also κβ has a tower of poles. Performing the
residue analysis, we find a few remarkable cancelations which significantly simplify
the analysis. The first major simplification is that the poles of the conformal blocks
always cancel in pairs after summing over J, and thus they give a zero contribution.
This is the same mechanism as underlies the cancellation of spurious poles in the
harmonic decomposition (2.15), see [18, 19, 41]. Furthermore, κ∆+J does not have
any poles on the right (see eq. (2.19) with a = b = 0), thus all the poles cancel. The
kernel is identically zero in this region!

K(z, z̄, w, w̄) = 0 for ρw > ρz ρ̄z ρ̄w . (3.10)

In other words the kernel is proportional to a unit step function K(z, z̄, w, w̄) ∝
θ(ρz ρ̄z ρ−1
w ρ̄w − 1). This was expected physically, since the Lorentzian inversion for-
mula is known to commute with the lightcone expansion: the step function ensures
that the z → 0 limit of the correlator is determined by the w → 0 limit of the dDisc.
In the kinematics in which we close to the left, the kernel is non-zero. Again the
spurious poles of the conformal blocks cancel out, but now there is a tower of double
poles coming from κ∆+J . These can be exhibited from the definition:

1 Γ4 ( β2 ) cot2 ( πβ
2
) 1
κβ ≡ = − . (3.11)
2π 2 Γ(β)Γ(β − 1) π2 κ2−β
Since β = ∆ + J, we can label the poles by a positive integer m:

∆pole = −J − 2m, with m = 0, 1, 2, . . . ∞. (3.12)

– 14 –
Thus from eqs. (3.2), (3.11), and the residue theorem, we have:
∞  
2 1 X g X d k−2J−2m0 (z)k−2m0 (z̄)k2J+2m0 +2 (w)k−2m0 (w̄)
KB = 2 2 2 +(z↔z̄),
π w w̄ J m=0 dm0 2π 2 κ2m0 +2
m0 =m
(3.13)
0
where we took the derivative with respect to m (as required by the residue theorem
for the case of double poles), and then plugged the integer value m. Now we notice
that J appears in only two hypergeometric functions; in fact the J-sum is telescopic
and can be computed exactly, see eq. (A.4). Performing the J-sum first we thus
obtain
∞
d Γ2 (2m0 + 1)
 
4 1 X
KB = 2 2 2 D2 k−2m0 (z)k−2m0 (z̄)k−2m0 (w̄)k2m0 +2 (w) ,
π w w̄ m=0
dm0 2Γ4 (m0 + 1)
m0 =m
(3.14)
where D2 is a first-order differential operator acting on z, z̄ and w and defined in
eq. (A.3). To summarize, the dispersion kernel K defined by eq. (2.24) is now written
explicitly in the form (3.4) with the contact term (3.8) plus the bulk part (3.14), the
latter still to be simplified. It remains to perform the sum m over the tower of
d
poles. This sum seems formidable: the summand is a derivative dm 0 of a product of

4 hypergeometric functions. Amazingly, it can be performed exactly!

3.2 Main result from Legendre PPPQ sum

We will now perform the sum in eq. (3.14), namely:
∞
d Γ2 (2m0 + 1)
X  
S≡ k−2m0 (z)k−2m0 (z̄)k−2m0 (w̄)k2m0 +2 (w) . (3.15)
m=0
dm0 2Γ4 (m0 + 1)
m0 =m

To get some intuition, we first notice that, near w̄ → 1, each term has at most
a logarithmic singularity. This is because k−2m0 (w̄) is polynomial for integer m0 ; a
d
singularity can only appear when the dm 0 derivative acts on k−2m0 (w̄),

dk−2m0 (w̄) Γ2 (m + 1)
= Pm (w̄ˆ ) × 12 log(1 − w̄) + (non-singular) (3.16)
dm0 m0 =m Γ(2m + 1)

where for conciseness in this section we use a hat notation in which ŵ ≡ w2 − 1.

Let us first focus on the coefficient of the log term, that is the discontinuity around
w̄ = 1.
We also notice that plugging m0 = m = integer, the hypergeometric functions
reduce to Legendre functions:

Γ2 (m + 1) Γ(β)
k−2m (z) = Pm (ẑ), kβ (z) = 2 Q β (ẑ). (3.17)
Γ(2m + 1) Γ2 ( β2 ) 2 −1

– 15 –
where Pm (ẑ) and Qm (ẑ) are Legendre polynomials and Legendre functions of the
second kind, respectively7 . Thus the log part of the sum becomes:
∞
X
S 1
= (2m + 1)Pm (ẑ)Pm (z̄ˆ)Pm (w̄ˆ )Qm (ŵ). (3.18)
2
log(1−w̄)
m=0

So we must now compute this (2m + 1)Pm Pm Pm Qm sum. Luckily, the coefficient
(2m + 1) in the sum is the canonical coefficient which often appears with Legendre
functions! Encouragingly, we further notice that a similar sum appeared for the
scattering amplitude dispersion relation in eq. (2.9), involving (2m+1)Pm Qm —which
can be realized in the limit z̄, w̄ → 1 of the current one.
It turns out that such sums (with precisely the coefficient (2m + 1)) have been
evaluated in the mathematics literature, dating back to Watson who computed a PPP
sum [42]. Specifically, we use the result in eqs. (3.8)-(3.10) of [43], who computed
the P P P P sum. To uplift his result to our P P P Q sum in eq. (3.18), we need
simply replace the PJ (ŵ) with a QJ (ŵ), which can be done using the single-variable
dispersion relation in the footnote below eq. (2.7). In fact this step is completely
trivial: the PPPP sum given in [43] is defined in the interval w̄ˆ ∈ [−1, 1], and has
square-root branch points at the boundary. The function whose discontinuity is this,
has exactly the same functional form, but now viewed as a function of the complex
plane minus the interval. The result of the sum is thus (see also [44, 45]):

4ρw ρ̄w ρz ρ̄z K(x)

1
S = p ,
2
log(1−w̄) π (ρ̄w ρz − ρw ρ̄z )(ρ̄w ρ̄z − ρw ρz )(ρz ρ̄z − ρw ρ̄w )(1 − ρw ρ̄w ρz ρ̄z )
(3.19)
where K(x) is the elliptic integral of the first kind:
Z π
2 dθ π
K(x) ≡ p = F ( 1 , 1 , 1, x),
2 2 1 2 2
(3.20)
2
0 1 − x sin θ

and x is the following combination of ρ’s, recorded previously in eq. (1.5):

ρz ρ̄z ρw ρ̄w (1 − ρ2z )(1 − ρ̄2z )(1 − ρ2w )(1 − ρ̄2w )

x≡ . (3.21)
(ρ̄w ρz − ρw ρ̄z )(ρ̄w ρ̄z − ρw ρz )(ρz ρ̄z − ρw ρ̄w )(1 − ρw ρ̄w ρz ρ̄z )

We are not quite done yet—recalling eq. (3.15), we need to account for the derivative
d/dm0 , or, equivalently, we need to find the function whose log term is eq. (3.19).
This appears to be a difficult task, and so we try instead to make an educated
guess. As boundary data, one can directly show that eq. (3.15) should be regular
at ρw → ρz ρ̄z ρw̄ , corresponding to x → 1. The 12 log(1 − w̄) term we have found
corresponds to log x as x → 0. Our guess is to look for a second solution to the same
7
We pick the branch of Qm (x) which is analytic at large m, where it decays like x−m−1 ; we
warn the reader that this is not the default branch picked by e.g. Mathematica.

– 16 –
hypergeometric differential equation, but satisfying these other boundary conditions.
In fact there is a unique candidate, which turns out to be also an elliptic function:

−πK(1 − x) = K(x) log(x) + non-singular. (3.22)

This equation states that the coefficient of the log singular terms of an elliptic func-
tion is itself an elliptic function, with a changed argument. Our educated guess,
extending eq. (3.19), is thus:

−4ρw ρ̄w ρz ρ̄z K(1 − x)

S=p . (3.23)
(ρ̄w ρz − ρw ρ̄z )(ρ̄w ρ̄z − ρw ρz )(ρz ρ̄z − ρw ρ̄w )(1 − ρw ρ̄w ρz ρ̄z )

A numerical evaluation of eq. (3.15), or its series expansion at small w̄, both confirm
that this ansatz is correct!
We are now done; the bulk term in the kernel is obtained as KB = π42 w21w̄2 D2 S,
from eq. (3.14). Performing some simplifications, this gives us the form recorded in
eq. (1.4), namely8 :
3
1  z z̄  2 (w̄ − w)( z1 + z̄1 + w1 + w̄1 − 2) 3
1 3
KB (z, z̄, w, w̄) = − 3 x 2 F1 ( 2 , 2 , 2, 1−x).
2
64π ww̄ (1 − z)(1 − z̄)(1 − w)(1 − w̄)

4

(3.25)
An equivalent expression, suitable for integrating with respect to ρ-variables, is:
 3/2  
1 dρw dρ̄w ρz ρ̄z 1 1 1 1
KB (z, z̄, w, w̄)dwdw̄ = − + + + −2
π (ρw ρ̄w )3/2 (1 − ρ2z )(1 − ρ̄2z ) z z̄ w w̄
(ρw − ρ̄w )(1 − ρw ρ̄w ) 3
× p x 2 2 F1 ( 12 , 32 , 2, 1 − x).
2 2
(1 − ρw )(1 − ρ̄w )
(3.26)

3.3 Match with d = 4 and differential equation

We will now similarly derive the dispersion relation in d = 4 spacetime dimension
and show that it equals the one in d = 2, due to interesting identities. Since the
steps are very similar, we omit details and emphasize the few changes. The conformal
blocks in d = 4 are given by:

z z̄  
GJ,∆ (z, z̄) = k∆−J−2 (z)k∆+J (z̄) − k∆+J (z)k∆−J−2 (z̄) . (3.27)
z̄ − z
8
This hypergeometric function is a linear combination of elliptic integrals of the first and second
kind:
1 3 4  
2 F1 ( 2 , 2 , 2, 1 − x) = K(1 − x) − E(1 − x) (3.24)
π(1 − x)

– 17 –
The extra prefactor, different measure, and shift in the argument of k functions (to
∆ − J − 2) lead to mild changes in eq. (3.2):
1 z z̄ w̄ − w
K (d=4) =
4πi z − z̄ w3 w̄3
∞
X 2+i∞
Z (3.28)
× d∆ κJ+∆ k∆−J−2 (z)k∆+J (z̄)kJ−∆+4 (w)k∆+J (w̄) + (z ↔ z̄).
J=0 2−i∞

As in the d = 2 case, we close the contour in the ∆ plane and pick up the residues
of the poles, being careful with the behavior at infinity which gives rise to contact
terms.

3.3.1 Contact term

We first compare the contact terms, which originate from the large-∆ asymptotics
given in eq. (3.3). Following the steps leading to eq. (3.6) we find that the kernels
match due to the following identity:
∞  
w̄ − w z z̄ X −1 −1 J
−1 −1 J
ρw ρ (ρw ρ̄w ρz ρ̄z ) 2 − ρ̄z (ρw ρ̄w ρ̄z ρz ) 2
ww̄ z − z̄ J=0 z ρw =ρ̄w ρz ρ̄z

g∞
 
J J
X
= (ρw ρ̄w ρ−1 −1
z ρ̄z ) + (ρw ρ̄w ρ̄z ρz )
2 2 (3.29)
J=0 ρw =ρ̄w ρz ρ̄z

which is rather surprising but can be verified by explicit computation on both sides.
We thus find that the contact term in d = 4 matches that of d = 2:
(d=2) (d=4)
KC (ρ̄w , ρz , ρ̄z ) = KC (ρ̄w , ρz , ρ̄z ). (3.30)

3.3.2 Bulk term

The agreement for the bulk term will be rather more remarkable. Again we find that
spurious poles from the blocks cancel out pairwise, so we only need to keep the poles
from κ in eq. (3.28), which are in the left-hand ∆-plane. The summation over J can
be performed similarly to eq. (3.14), and leads to a different operator acting on the
same sum S defined in eq. (3.15):
(d=4) 4 1
KB = D4 S (3.31)
π 2 w2 w̄2
with
z z̄ w − w̄  zw(1 − w) z̄w(1 − w)  zw(1 − z) z̄w(1 − z̄) 
D4 ≡ − ∂w − ∂z + ∂z̄ ,
z − z̄ ww̄ z−w z̄ − w z−w z̄ − w
(3.32)
instead of D2 given in eq. (A.3). Remarkably, however, it is possible to verify using
the explicit form of S in eq. (3.23) that the two kernels agree:
 
D4 − D2 S = 0. (3.33)

– 18 –
As a result, the 4d bulk kernel is equal to the 2d one!
(d=4) (d=2)
KB (z, z̄, w, w̄) = KB (z, z̄, w, w̄). (3.34)

In summary, we showed that the dispersion relation is the same in d = 4 and d = 2.

This strengthens our intuition that the dispersion relation should not depend on the
space-time dimension d; it would be interesting to show this in other dimensions.
The agreement between the d = 2 and d = 4 kernels gives us an interesting first-
order differential equation satisfied by the P P P Q sum S. Turning the logic around,
we can now use this differential equation to help determine the kernel in the general
case of unequal scaling dimensions.

3.4 Differential equation for unequal scaling dimensions

We turn to the case of a generic 4-point correlator hO1 . . . O4 i of scalars with unequal
scaling dimension: a = 12 (∆2 − ∆1 ) 6= 0 and b = 21 (∆3 − ∆4 ) 6= 0. We will be brief
and emphasize the main points. There is formally no change to eq. (3.2), namely:

g 1+i∞
µ(w, w̄)(a,b) X
Z
(a,b) (a,b) (a,b) (a,b) (a,b) (a,b)
K = d∆ κJ+∆ k∆−J (z)k∆+J (z̄)kJ−∆+2 (w)k∆+J (w̄)+(z↔z̄),
4πi J 1−i∞
(3.35)
where we have simply made explicit the dependence on a and b of the various factors.
One may easily derive the contact term, by making a simple replacement in eq. (3.8):
  a+b
(a,b) (1 − w)(1 − w̄) 2
(0,0)
KC (z, z̄, w, w̄) = KC (z, z̄, w, w̄). (3.36)
(1 − z)(1 − z̄)

The bulk term comes from poles of κ∆+J , since spurious poles from the conformal
blocks cancel in pairs just as in the a = b = 0 case. As opposed to that case, however,
the poles of κ∆+J are now single poles instead of double poles. After performing the
J sum using the identity in eq. (A.2), we find the generalization of eq. (3.14):
 
(a,b) (a,b) (a,b) (a,b) (b,a) (b,a)
KB (z, z̄, w, w̄) = µ (w, w̄) D2 Sa + S−a + Sb + S−b (3.37)

where D2 , given in eq. (A.3), is the same differential operator as before, and we have
defined:
∞ (a,b) (a,b) (a,b) (a,b)
(a,b)
X sin(2πa0 ) Γ21+2a0 +2m k−2m−2a0 (z)k−2m−2a0 (z̄)k−2m−2a0 (w̄)k2m+2+2a0 (w)
Sa0 ≡
m=0
2π m! Γ1+2a0 +m Γ1+a0 −b+m Γ1+a0 +b+m sin(π(a − b)) sin(π(a + b))
(3.38)
using the notation Γx ≡ Γ(x). The sum in eq. (3.38) contains products of four
hypergeometric functions which cannot be reduced to Legendre functions. Thus it
may seem hopeless to try to compute it directly. However, we may say a lot about
the result using differential equations.

– 19 –
 A key observation is that dimension-independence still holds, that is:
(a,b)
(D2 − D4 )Sa0 = 0. (3.39)

We could prove this using hypergeometric identities to rewrite the derivatives as

shift on the index m of the k functions, and showing that m sum becomes telescopic;
it may also be readily verified order by order in w. Notice that both D2 and D4
are first-order differential operators (and independent of a and b). In fact, thanks
(a,b)
to the manifest permutation symmetry of Sa0 in (z, z̄, w̄), this identity and its
permutations give two linearly independent differential equations. The fact that a
function is annihilated by two first-order equations implies that it factors through
the two variables which represent its zero-modes, up to an overall factor:
√
(a,b) z z̄ww̄ (a,b)
Sa0 = 1/2+a+b × S̃a0 (x, y) (3.40)
y

where x is in eq. (1.5), reproduced here for convenience, and y is:

ρz ρ̄z ρw ρ̄w (1 − ρ2z )(1 − ρ̄2z )(1 − ρ2w )(1 − ρ̄2w )

x= , (3.41)
(ρ̄w ρ̄z − ρw ρz )(ρ̄w ρz − ρw ρ̄z )(ρ̄z ρz − ρ̄w ρw )(1 − ρw ρz ρ̄w ρ̄z )
(1 − ρz )(1 − ρz̄ )(1 − ρw )(1 − ρw̄ ) p
y= = (1 − z)(1 − z̄)(1 − w)(1 − w̄) . (3.42)
(1 + ρz )(1 + ρz̄ )(1 + ρw )(1 + ρw̄ )
(a,b)
The sums S̃a0 are further constrained by second-order differential equations, which
encode that the SL(2,R) Casimir eigenvalue with respect to each of the four variable
are the same as can be seen from eq. (3.38). From these we find two equations on S̃:
 (a,b)
0 = x2 (1 − x)∂x2 − x2 ∂x + y 2 ∂y2 + y∂y − 12 x2 (1 − y 2 )∂x ∂y + ( 14 − a2 − b2 ) S̃a0 ,

 (a,b)
0 = (x2 (1 − y)2 + 4xy)∂x ∂y − 2y∂y − 4ab S̃a0 .


(3.43)
(a,b) 1/2+a0
These two, together with the boundary condition that Sa0 (x, y) ∝ x (1+O(x))
as x → 0, with a constant easily determined from the m = 0 term in eq. (3.38),
completely determine the functions S.
Before discussing solutions, let us make an observation about the a = b = 0 case:
the second equation can be used to fix the y dependence of each term recursively in a
(a,b)
series in x; when ab = 0, it implies that the solution is independent of y: ∂y S̃a0 = 0.
√
The first equation then reduces to that satisfied by the elliptic function xK(1 − x).
With this method it is thus straightforward to derive the result (3.23) which we
previously only guessed. The key is the identity in eq. (3.39), which states that the
kernels in d = 2 and d = 4 are the same and which leads to eq. (3.40).
(a,b)
Instead of looking at the individual sums S̃a0 we now focus on the specific
combination in eq. (3.37) and the actual kernel. It is convenient to explicitly act

– 20 –
with the differential operator D2 on the prefactor in eq. (3.40). In a convenient
normalization the kernel is then
3
(a,b) 1  z z̄  2 (w̄ − w)( z1 + z̄1 + w1 + 1
w̄
− 2) (a,b)
KB =− K̃B (x, y) (3.44)
64π ww̄ y 3/2+a+b
where  
(a,b) (a,b) (b,a) (b,a)
K̃B (x, y) ≡ −8πx2 ∂x S̃a(a,b) + S̃−a + S̃b + S̃−b . (3.45)

From eqs. (3.43) we derive differential equations satisfied by K̃:

 (a,b)
0 = x2 (1 − x)∂x2 − 2x∂x + y 2 ∂y2 + y∂y − 21 x2 (1 − y 2 )∂x ∂y + ( 94 − a2 − b2 ) K̃B (x, y),

 (a,b)
0 = (x2 (1 − y)2 + 4xy)∂x ∂y − 6y∂y − 4ab K̃B (x, y).


(3.46)
These conditions ensure that the dispersion relation commutes with the s-channel
quadratic Casimir.9 While we have not been able to solve these in closed form, we
can state the following results:
(a,b)
• The kernel K̃B (x, y) is regular around x = 1. While this is not true for
the individual sums in eq. (3.38) (each has a logarithmic singularity), this is a
special property of the combination in eq. (3.37). The kernel is then the unique
regular solution to (3.44) with the boundary condition
(a,b) y
lim K̃B (x, y) = 1 − 4(a + b)2 + 16ab . (3.47)
x→1 y+1

(We could get the constant by combining the eqs. (3.46) into a single fourth-
order one with no y-derivatives, and solving it along the y = 1 line in terms of
4 F3 hypergeometric functions.)

• The limits as y → 0 and y → ∞ are regular, and equal to a simply generaliza-

tion of eq. (3.8):10
(a,b)
lim K̃B (x, y) = (1 − 4(a + b)2 )x3/2+a+b 2 F1 ( 21 + a + b, 32 + a + b, 2, 1 − x),
y→0
(a,b)
lim K̃B (x, y) = (1 − 4(a − b)2 )x3/2+a−b 2 F1 ( 12 + a − b, 32 + a − b, 2, 1 − x).
y→∞
(3.48)

• A Taylor series in (1 − x) can be obtained using just the second of eqs. (3.46),
y
together with the previous limits; each term is polynomial in 1+y .
9
We found that the resulting dispersion in fact commutes with the quadratic Casimir in any
dimension. The dimension-dependence of the Casimir is a first-order differential operator which it
might be interesting to relate to the constraint in eq. (3.39).
10
The limits x → 0 and y → 0 (or y → ∞) do not commute: to compare with the x → 0 limit
given below eqs. (3.43), one needs to carefully cross the region x ∼ y → 0.

– 21 –
In summary, for unequal scalar operators, the kernel takes the form in eq. (1.1), with
the contact term given explicitly in eq. (3.36), and bulk term in eq. (3.44) implicitly
described by the above.
Let us briefly comment on the special case: a = 0 and b = 21 , where the bulk
term KB identically vanishes. (This could be seen directly from the lack of poles of κ
in eq. (2.19).) This corresponds physically to a case where the double-discontinuity
(2.21) is effectively a single discontinuity!11 Only the contact term (3.36) remains.
We observe also that it is free of square roots (when written in terms of ρ’s). (More
generally, when a is integer and b is half-integer, the contact kernel K̃C (z, z̄, w̄) does
not contain square roots.) The dispersion relation then reduces to

1 1 dρ̄w 1 − ρz ρ̄z ρ̄2w (1 − ρw )(1 − ρ̄w )

Z
t,(0, 12 )
 
G (z, z̄) = dDisc G(w, w̄) ρw =ρz ρ̄z ρ̄w
.
π 0 ρ̄2w (1 − ρz ρ̄w )(1 − ρ̄z ρ̄w ) (1 − ρz )(1 − ρ̄z )
(3.49)
Upon further inspection, this could be recognized as a single-variable dispersion
relation of the form of section 2.1, taken with fixed value of the ratio ρ̄w /ρw and
acting on a certain rescaling of the correlator. This ratio more generally will play an
important role in the next section.

4 Direct proof of dispersion relation

Having now obtained its kernel, we will now prove directly that the dispersion integral
(1.1) indeed reconstructs correlators. This may be viewed as a theorem in complex
analysis, independent of the CFT origin of the formula. This will show directly
the validity of the formula in any dimension, and will enable us to go beyond the
situations where the Lorentzian inversion formula converges.
We begin by observing that the contact term and bulk term of the kernel (see
eqs. (3.4), (3.8) and (3.25)), are not independent, disparate entities. Rather, they
combine into the discontinuity of a single “pre-kernel”:
3
1 w̄ − w (z z̄ww̄) 2 ( z1 + z̄1 + w1 + w̄1 − 2) 3 1 3
Kpre (z, z̄, w, w̄) = − 3 x
2
2 F1 ( 2 , 2 , 1, x).
32 ww̄ ((1 − w)(1 − w̄)(1 − z)(1 − z̄)) 4
(4.1)
Near x = 1, the hypergeometric function above satisfies:

1 3 2 1 1 3
2 F1 ( 2 , 2 , 1, x) = − 2 F1 ( 2 , 2 , 2, 1 − x) log(1 − x) + (non-singular). (4.2)
π(1 − x) 2π
Using that
(ρz − ρ̄z ρw ρ̄w )(ρ̄z − ρz ρw ρ̄w )(ρ̄w − ρw ρz ρ̄z )(ρz ρ̄z ρ̄w − ρw )
1−x= , (4.3)
(ρ̄z ρ̄w − ρw ρz )(ρz ρ̄w − ρw ρ̄z )(ρz ρ̄z − ρw ρ̄w )(1 − ρw ρz ρ̄w ρ̄z )
11
We thank Dalimil Mazac for this observation.

– 22 –
we see that the pre-kernel has both a pole and branch cut at ρw → ρz ρ̄z ρ̄w , whose
residue and discontinuity precisely match, respectively, the contact term and bulk
terms:
1
KC = Resρw =ρz ρ̄z ρ̄w [Kpre ], KB = Discρw →ρz ρ̄z ρ̄w [Kpre ] (ρw < ρz ρ̄z ρ̄w ). (4.4)
2π
This enables to combine these terms into a single contour integral:
Z Z 1
t 1
G (z, z̄) = dw dw̄ Kpre (z, z̄, w, w̄) dDisc[G(w, w̄)] (4.5)
2πi Cw 0

where Cw is a “keyhole” contour going from the origin to the origin counter-clockwise
around its maximum wmax corresponding to ρw = ρz ρ̄z ρ̄w (similar to the contour Cσ
in fig. 4(a)). The existence of such a pre-kernel is very suggestive of a contour
deformation argument leading to the dispersion relation.

4.1 Contour deformation trick

We will now describe a contour in two complex variables, which, fortunately for us,
takes on a simple factorized form in suitable variables. The “good variables”, as
suggested by the degenerate case in eq. (3.49), are the geometric mean and ratio of
ρ-coordinates:
√ p √ p
σz = ρz ρ̄z , ηz = ρz /ρ̄z , σw = ρw ρ̄w , ηw = ρw /ρ̄w . (4.6)
Physically, in the Euclidean cylinder, σ is a radial coordinate and η = eiθ is an
angular variable. The singularities which will be relevant for our argument are shown
in fig. 4. The complete list of singularities of the kernel and pre-kernel come from
where x = 0, 1, ∞, namely:
x=0: ηw ∈ {±σw , ±σw−1 }, σw ∈ {0, ±ηw , ±ηw−1 , ∞},
x=1: ηw ∈ {±σz , ±σz−1 }, σw ∈ {±ηz , ±ηz−1 }, (4.7)
x = ∞ : ηw ∈ {±ηz , ±ηz−1 }, σw ∈ {±σz , ±σz−1 }.
Notice that each ηw -plane singularity is reflected four-fold: by η 7→ η −1 , which is
parity w↔w̄, and by η 7→ −η, which interchanges the t and u channels (ie. swaps
operators 1 and 2 in the four-point correlator).
We will now see that the dispersion relation can be derived starting from the
identity: I
0= Kpre (z, z̄, w, w̄)G(w, w̄) (4.8)
Cσ ×{|ηw |=1}

where the original contour, shown in fig. 4, is a product of a keyhole in σw (similar

to eq. (4.5)), times the unit circle ηw = 1, and then deforming the contour.12 This
will hinge on several properties that the pre-kernel (4.1) (remarkably!) combines:
12
The variable ηw was called w in ref. [18] and a similar contour deformation starting from the
unit circle |ηw | = 1 was used there to derive the Lorentzian inversion formula.

– 23 –
 1. It is odd under w↔w̄ due to the factor (w − w̄).

2. The branch cut at x = ∞ is only logarithmic.

1
3. It has a simultaneous double pole when (σw , ηw ) = (σz , ηz ), e.g. (ηw −ηz )(σw −σz )
.

4. An analogous pole at ηw = −ηz is canceled by the factor ( z1 + z̄1 + 1

w
+ 1
w̄
− 2).

5. The pre-kernel is symmetrical under ρ̄w 7→ ρ̄−1 x

w (x 7→ x−1 ) in the region x < 1.
This symmetry survives for the average of the two branch choices after going
around x = ∞.

Notice that each factor in the pre-kernel (4.1) has some role to play.
The vanishing of the integral along the unit circle |ηw | = 1 (4.8) is basically due
to symmetry property 1. This is valid for generic 0 < z < z̄ < 1. Note however that
property 2 is also implicitly used here, since at fixed σw the unit circle contour would
not be well-defined due to a branch cut at x = ∞. However, thanks to property 2,
the discontinuity across that cut cancels when integrated along the σw keyhole. Only
the two-dimensional contour is well-defined.

σw ηw
ηw
∞ ηz ∞ ∞ 1 0 0 1 ∞ ∞1
σz Cσ σw σz ηz

(a) (b)

Figure 4. The integration contour used in eq. (4.8) to prove the dispersion relation is a
product of a keyhole and a circle. (a) σw -plane: the keyhole Cσ starts and ends at σw = 0.
(b) ηw -plane: the integral over the circle |ηw | = 1 vanishes and is equal to the sum of the
pole and cuts in its interior. The pole gives minus the correlator G(z, z̄) and the cuts give
the dispersion integral. Values of x are shown in light gray above the axis.

The trick now is to deform the ηw contour inward from the unit circle. Property
3 ensures there is a pole at ηw = ηz , with residue −G(z, z̄). There is no branch cut
at this point (thanks to point 2, ie. there is an expansion similar to (4.2) around
x → ∞, and we are already taking a discontinuity in σ), so we can keep shrinking
the contour until it hits the cut at the smaller radius ηw = σz .
In doing so, one might worry about a reflected pole at ηw = −ηz , denoted by a
circle in fig. 4(b). Its residue would be the u-channel correlator G(z/(z−1), z̄/(z̄−1)).

– 24 –
Property 4 ensures that this undesired pole does not contribute, since the numerator
z z̄
of the pre-kernel (4.1) vanishes when (w, w̄) = ( z−1 , z̄−1 ):
   
1 1 1 1 z − 1 z̄ − 1 1 1
+ + + −2 = − − + + . (4.9)
z z̄ w w̄ z z̄ w w̄
This explains the role of this mysterious factor!
It remains to show that the cut organizes into dDisc’s. We organize the cut
into four segments. On the positive axis there is the Regge region (0, σw ) and the
Euclidean region (σw , σz ). They connect at ηw = σw or ρ̄w = 1, where a lightcone is
crossed (x214 x223 = 0). Each has a u-channel reflection on the negative axis. In the
Regge region we have 0 < ρw < 1 < ρ̄w with the constraint ρw ρ̄w < ρz ρ̄z . To map it
to our reference region (inside the unit square) we simply need to use the symmetry
under inversion ρ̄w 7→ ρ̄−1
w in property 5 (which interchanges ηw and σw ):
13
to write
it to a form similar to eq. (4.5)
I Z 1
eq. (4.8) ⊃ dw̄Kpre × G(ρw , ρ̄−1
w −i0). (4.10)
Cw 0

where Cw is the keyhole covering 0 < ρw < ρz ρ̄z ρ̄w . Notice that there is a Regge region
above the axis, and one below the axis. The kernel is identical in both (because its
branch point at infinity is only logarithmic), so the two sides of the t-channel region
simply replace:
G(ρw , ρ̄−1 −1 −1
w −i0) 7→ G(ρw , ρ̄w −i0) + G(ρw , ρ̄w +i0) ≡ G + G

(4.11)
where the notation emphasizes that we have gone a full circle around w̄ = 1 in the
original cross-ratios. The Euclidean region can similarly be combined, however in this
case we do not need to change variables but we need to use a symmetry of the kernel.
A subtlety is that the pre-kernel (4.1) (after x has been around ∞) has a log branch
point at the boundary x = 0 between the Regge and Euclidean regions; however,

we need the average between the two sides of the real axis (Kpre + Kpre = 2Kpre ,
property 5), leaving the desired double-discontinuity:
I Z 1
dw̄ Kpre (G + G  ) − Kpre + Kpre 



G
Cw 0
I Z 1 I Z 1 (4.12)

= dw̄Kpre × (G + G − 2G) ∝ dw̄Kpre × dDisc[G] .
Cw 0 Cw 0

The cut segments on the negative real axis similarly organize into a double discon-
tinuity around the u-channel limit w̄ → ∞.
Let us summarize. We have proved a general result on single-valued functions of
two complex variables G(ρz , ρ̄z ). We call a function “single-valued” if it satisfies the
following:
13 1 3
It may be verified through the hypergeometric identity, valid for x < 1: 2 F1 ( 2 , 2 , 1, x) =
(1 − x)3/2 2 F1 ( 12 , 32 , 1, x−1
x
).

– 25 –
 
• It is analytic in a cut plane C \ [1, ∞) ∪ (−∞, 0] for each variable ρz and ρ̄z

• It is devoid of branch cuts when restricted to the Euclidean region ρ̄z = (ρz )∗

• It satisfies G(ρz , ρ̄z ) = G(ρ̄z , ρz ) = G(ρ−1 −1

z , ρ̄z ) in the Euclidean region.

These properties are satisfied by any CFT correlator (as reviewed in [18]). (The third
condition is simply because of the way the ρ variables cover the u, v cross-ratios.)
The Euclidean OPE limit is (ρz , ρ̄z ) → (0, 0) and the Regge limit is (ρz , ρ̄z ) → (0, ∞)
(both of which map to (z, z̄) → (0, 0) but on different sheets). Then we showed:

Theorem Let G(ρz , ρ̄z ) be a single-valued function of two complex variables, which
vanishes sufficiently fast in the Euclidean and Regge limits. Then the function can
be recovered from its double-discontinuity
Z 1
t u t
G(z, z̄) = G (z, z̄) + G (z, z̄), G (z, z̄) ≡ dwdw̄K(z, z̄, w, w̄)dDisc[G(w, w̄)],
0
(4.13)
with the kernel as quoted in introduction (eq. (1.3)). The necessary rate of vanishing
can be estimated from convergence at σw = 0, and along the small arc at ηw = 0
which connect the t- and u-channel Regge limits in the preceding argument. By
expanding Kpre in these limits, we find that these arcs can be ignored provided that
G(z, z̄) vanishes faster than (z z̄)1/2 in both limits. (Convergence as w̄ = 1 also naively
requires a singularity no worse than (1 − w̄)−3/4 , however in reality this is naturally
resolved by retaining certain arcs in the contour there, shown below, and there is no
real constraint there.14 )

Viewing eq. (4.13) as a result in complex analysis, rather than a result in con-
formal theory, will be helpful for generalizations below.

4.1.1 Why two variables?

The kernel in (4.13) is quite nontrivial, and it is interesting to ask whether a disper-
sion relation with a simpler kernel than could have been possible.
It is of course possible to fix one variable, say z, and simply reconstruct the
correlator from its discontinuity in z̄ using the logic of Cauchy’s theorem, as usually
done for amplitudes (see section 2.1). However, such a formula will not feed on the
double discontinuity, which has a clear physical interpretation as an absorptive part.
Rather, it would feed on the correlator in regions such as (z, z̄) ∈ (0, 1)×(1, ∞),
whose physical interpretation remain unclear to us. We take the viewpoint that the
physical goal of a “dispersion relation” is to reconstruct data from some kind of
14
The cross-channel arcs near w̄ = 1, unlike arcs near w̄ = 0, do not invalidate the physical
interpretation of the formula as building on the dDisc.

– 26 –
“absorptive part”. One could try to repeat the process with respect to say z̄ to try
and get a second discontinuity, but the basic issue which we couldn’t solve is that
this wouldn’t avoid unphysical regions. Variables (ρz , ρ̄z ) suffer from the same issue.
One might hope to get more p appealing formulas by choosing better variables,
perhaps integrating over η = ρz /ρ̄z in fig. 4b with σ fixed. Indeed, taking η
negative does take us to the physical u-channel (which is why this variable was so
useful above). The issue however is that for the integrand to organize into a dDisc,
there would have to be a corresponding integral where η is fixed and σ is integrated
over (to provide the Euclidean correlator part of the dDisc). The formula obtained
in this paper achieves this by having a two-dimensional integral over both ηw and
σw , and a nontrivial symmetry when they are exchanged (see eq. (4.10)).

4.2 Convergence and subtractions

We are now positioned to overcome the limiting assumptions made in section 3, and
obtain a subtracted dispersion relation that is applicable in an arbitrary unitary
CFT.

4.2.1 Subtracted dispersion relation

Let us first see how the theorem (4.13) clarifies the non-renormalizable terms in the
formula (1.2) quoted in the introduction. One such mode that is generically present
is the s-channel identity exchange, which leads to limz,z̄→0 G(z, z̄) = 1,. This violates
the assumptions of the theorem.
A solution is simply to apply the theorem to the function [G − 1], which is
also single-valued, has exactly the same double-discontinuity, but vanishes faster in
the Euclidean OPE limit z, z̄ → 0. In general, harmonic functions (single-valued
combinations of blocks and their shadows) should be subtracted for each operator
of dimension less than 1 (the same as in the harmonic analysis formula (2.15) with
d = 2). This explains the non-normalizable terms in eq. (1.2).
 
There remains the question of whether the function G(z, z̄) − (non-norm.) van-
√
ishes faster than z z̄ in the Regge limit (the limit as z, z̄ → 0 with z̄ on a second
√ 1−J
sheet), corresponding to z z̄ with exchange of a spin J = 0 excitation [40]. Uni-
tarity implies only that the correlator stays bounded, so in general this will not be
the case. This reflects the fact that the Lorentzian inversion formula (which was the
starting point of the preceding section) may fail to converge to the OPE data for
spins J ≤ 1 [18].
The theorem (4.13) offers a simple way out: apply it to a rescaled correlator
z z̄
Gu/v = G (1−z)(1−z̄) . This is similar to the amplitude subtraction in eq. (2.4). Since
G is bounded in both the Euclidean and Regge limits (in any unitary CFT), this
rescaled correlator vanishes like z z̄ in both limits, and amply satisfies the assumptions

– 27 –
of the theorem. Explicitly showing the t- and u-channel contributions, this gives:
Z 1  
z z̄ ww̄
G(z, z̄) = dwdw̄ K(z, z̄, w, w̄) dDisc G(w, w̄)
(1 − z)(1 − z̄) 0 (1 − w)(1 − w̄)
Z 1 h i
z z̄ 0
+ dwdw̄ K( z−1 , z̄−1 , w, w̄) dDisc ww̄G (w, w̄) ,
0
(4.14)
0
where G denotes the correlator with operators 1 and 2 interchanged. The subtracted
dispersion relation (4.14) is a main result of this paper: it is guaranteed to converge
in any unitary CFT. Notice that the “non-normalizable” terms are gone: the extra
power of z z̄ has made their subtraction unwarranted (and incorrect).
The price for better convergence at w, w̄ → 0 is poorer convergence near the
cross-channel limit w̄ → 1. This is addressed shortly; the bottom line is that 1/(1−w̄)
means that the dDisc operation, which normally suppresses double-trace operators
in the t-channel, will leave unsuppressed the lowest double-twist family (ie. t-channel
operators of twist ∆0 − J 0 ≈ 2∆ext ). (Out of possible other choices, we chose uv so
that the dDisc still suppresses higher double-twists.)

4.2.2 Keyhole contour near cross-channel singularity

We finally address convergence near w̄ → 1. A basic fact is that the original in-
tegration contour (see fig. 4) does not touch that point, so there can’t be any real
divergence there. Rather, the contour encircles that point, and any apparent diver-
gence at w̄ = 1 is an artifact of incorrectly shrinking the circle to zero size.
The solution is to integrate ρ̄w over a “keyhole” type contour. We write the
result in full in the case of identical external operators. This is best done in the
following variables. First, we parametrize the integration in terms of t and ρ̄w by
setting ρw = ρz ρ̄z ρ̄w t, so the dispersion relation becomes:
Z 1 Z 1 Z
dwdw̄ K(z, z̄, w, w̄) dDisc[G̃(w, w̄)] 7→ dt dρ̄w JK(z, z̄, w, w̄)F
0 0 ρw =ρz ρ̄z ρ̄w t
A,B+ ,B−
(4.15)
dw dw̄
where J = is a Jacobian for the change of variable (w, w̄) 7→ (t, ρw ),
ρz ρ̄z ρ̄w dρ w dρ̄w

and G̃ stand for either of the combinations entering in eq. (4.14). Recall that the
kernel is a sum of a bulk and contact part, which are supported on 0 < t < 1 and
t = 1, respectively; they have simple expressions in ρ-coordinates, eq. (3.26).
If regularization were not needed, the ρ̄w contour would be simply the inter-
val [0, 1], and the correlator F = dDisc[G̃] evaluated for w, w̄ given in terms of t, ρ̄w .
Keeping the full key-hole contour, it is instead the sum of a regulated interval [0, ρmax ]
and two half-circles [ρmax , ρ−1
max ]. Explicit parametrizations and corresponding inte-
grands are shown in figure 5.

– 28 –
 B+ ρw̄
region ρ̄w correlator F
0 ρmax 1 ρ−1 A ρ̄w ∈ [0, ρmax ] dDisc[G](t, ρ̄w )
max −iθ
A B+ ρ̄w = (ρmax )e − 21 G (t + i0, ρ̄w )
iθ
B− ρ̄w = (ρmax )e − 21 G  (t − i0, ρ̄w )
B−
Figure 5. Keyhole contour in the ρ̄w variable to avoid cross-channel singularity (at w̄ = 1).
The angle runs over θ ∈ [0, π]. The radius of the circle shrinks as ρmax → 1.

The −1/2’s in the formula originate from the dDisc, which we recall (see eq. (2.21))
for identical operators (the case considered here) is dDisc[G] = G − 12 G − 21 G  . Va-
√
lidity of the formula require that the contour not enclose the poles at ρ̄w = 1/(ρz t)
√
and 1/(ρ̄z t); for real ρz , ρ̄z this is simply achieved by requiring ρmax ≥ max(ρz , ρ̄z ).
In practice, in numerical examples below we chose ρmax = 0.9 (adequate for z, z̄ <
0.997), and we verified that the integral is independent of ρmax . The t ± i0 notation
indicates that the t-contour must avoid a branch point on the real axis (at t = ρ2max ).15
(We note that the integrand is not analytic at the point ρ̄w = ρ−1 max where B± meet.
At this point, the integrand matches onto the Euclidean correlator part of dDisc[G]
at ρ̄w = ρmax .) We find that the keyhole integral is quite practical numerically.

5 Checks and discussion

In this section we illustrate various checks and possible applications of the formula.

5.1 Numerical check for generalized free fields

A first sanity check is to compare both sides of the dispersion relation in the simple
example of generalized free field. We consider:

G(w, w̄) = up1 v p2 = (ww̄)p1 ((1 − w)(1 − w̄))p2 (5.1)

Then taking the double discontinuity (for non-integer exponents), gives for the t and
u channel contributions respectively:

dDisct [G(u, v)] = 2 sin2 (πp2 )(ww̄)p1 ((1 − w)(1 − w̄))p2 ,

dDisct [G(u/v, 1/v)] = 2 sin2 (π(p1 + p2 ))(ww̄)p1 ((1 − w)(1 − w̄))−p1 −p2 . (5.2)

Now we can plug this on the right hand side of Eq. (4.13). There is a nonempty range
of p1 and p2 for which that formula converges without subtlety, namely: p1 > 21 and
− 34 < p2 < 34 − p1 . Computing numerically these integrals for various values of p1
15
In practice, we parametrize t ± i0 ≡ τ ± iτ (1 − τ ) where τ ∈ [0, 1] is a real integration variable.
The offset  doesn’t need to be small, and we used  = 1 in all numerical examples.

– 29 –
and p2 in this range (and various values of z, z̄) we found perfect match with the
LHS of eq. 4.13!
If we relax the condition on p1 , we need the subtracted dispersion relation (4.14).
And if p2 is such that convergence is not satisfied at 1 (or ∞), we need to use the
keyhole contour in figure 5. Again we find perfect agreement, for example when
p1 = p2 = 0, or p2 = 2.25 with either p1 = 1 or p1 = 2.25 (in the later case there
is only the t-channel cut). In particular, the first test confirms that the subtracted
dispersion relation (4.14) correctly reconstructs even the identity exchange, from the
dDisc of the correlators times u/v.
For unequal scaling dimensions, we did not attempt numerics because we do not
have a closed form for the kernel (3.44). However, we performed numerical tests in
the special case a = 0 and b = 12 using eq. (3.49), and also found perfect agreement.

5.2 Holographic correlators

Figure 6. The tree-level exchange diagram for the N = 4 stress tensor multiplet.

The double-discontinuity is particularly simple to compute in holographic theo-

ries, as it is saturated at tree-level by exchange of a finite number of light single-trace
operators. In some sense the dispersion relation give novel “closed form” expressions
for tree-level holographic correlators as an integral over conformal blocks.
For illustration, consider the correlator of stress-tensor multiplets in planar
N = 4 at large ’t Hooft coupling, dual to tree-level gravity in AdS5 × S5 . We
follow the notation of [27]. (In short: because of supersymmetry, the stress tensor
lies in a supermultiplet which includes scalars of dimensions ∆i = 2 in the [0, 2, 0]
representation of the SU(4)R global symmetry. We study the correlation functions of
these scalars, projected onto the s-channel [0, 4, 0] representation, which is known to
determine all other representations. This projection has good high-energy behavior,
allowing to use the unsubtracted dispersion relation.) The double-discontinuity in
this limit is saturated by exchange of a single t-channel (super)conformal block cor-
responding to the stress tensor multiplet, or graviton exchange in the bulk (fig. 6);

– 30 –
it admits a particularly concise form [27, 46]:
  2
w − 4w3 + 3w4 − 2w4 log w
 
(1) 1
dDisc[G(w, w̄)] = dDisc × ≡ X(w) .
1 − w̄ (1 − w)4
(5.3)
The correlator itself is the only single-value function with this double-discontinuity
(and correct Regge behavior) and is given as [47, 48] (see [49] for the D̄ functions):

u2
G(z, z̄)(1) = − u4 D̄2422 (u, v),
v
D̄2422 = ∂u ∂v (1 + u∂v + v∂v )Φ(1) (u, v), (5.4)
1−z
2Li2 (z) − 2Li2 (z̄) +log[ 1−z̄ ] log(z z̄)
Φ(1) (u, v) = .
(z − z̄)
We would like to see here how the dispersion relation reconstructs (5.4) starting from
the elementary dDisc given above it.
First we note that the dDisc is naively zero (no branch cut), so it is really a
sort of delta-function around w̄ = 1. This can be seen explicitly from the keyhole
contour in fig. 5: only the semi-circles survive. In fact it is possible to directly
integrate numerically over the semi-circles and compare (successfully) with (5.4).
Let us see how the integral could be done analytically. In fact, the two half-circles
would precisely cancel each other were it not for the fact that the kernel has a log.
So dDisc[1/(1 − w̄)] is effectively −2π 2 times a sort of δ-function which extracts the
coefficient of log(1 − w̄) in the kernel, and eq. (4.15) becomes
Z wmax  2

− 2u (1 − w)(u/w + 1 − v) + (u-channel)
G(z, z̄)(1) = dw
3/2 × X(w) (5.5)
0 u2 − 2wu(1 + v) + w2 (1 − v)2

where wmax = (1+u√v)2 is where ρw = ρz ρ̄z . Adding the u-channel term simply cancels
the (1 − v) term in the numerator, and doubles the remaining u/w term. A comment
is in order: the integration endpoint is a branch point of the denominator, so it
appears that by shrinking the circles to get a δ-function we have created a new
divergence as w = wmax . One can show that the proper treatment simply amounts
Rw R0
to integrating w itself on a keyhole, 0 max 7→ 12 0 , as in fig. 4a. The integral is then
unambiguous, and can also be checked numerically (giving a nontrivial confirmation
of the form of the dispersion relation).
The form of the integrand of eq. (5.5) makes manifest the fact that the integral
gives a combination of dilogarithms and simpler functions — the most complicated
R
part can be written in the form d log(· · · ) log(w), to which standard integration
algorithm can be applied, see for example [50]. (Since the square root has two branch
points with respect to w, one has to first go to a double-cover where the square root
is gone.) It could be interesting to use this method to help understand the functions
which can appear at higher loops.

– 31 –
5.3 An integral relation between conformal blocks
The validity of the dispersion relation predicts a new relation between harmonic
functions and the inverted block which enters the Lorentzian inversion formula (and
accessorily rederive the latter). Recall the dispersion relation:
Z
G(z, z̄) = dwdw̄ K(w, w̄, z, z̄) dDisc[G(w, w̄)] + (t ↔ u) (5.6)

The Euclidean inversion formula, in the conventions of [18], is:

Z
c(J, ∆) = N (J, ∆) dzdz̄ µ(z, z̄) FJ,∆ (z, z̄) G(z, z̄) (5.7)
Eucl

Γ(J+ d−2 )Γ(J+ d )K

J,∆
where the normalization is N (J, ∆) ≡ 2πΓ(J+1)Γ(J+d−2)K
2 2
J,d−∆
. Plugging eq. (5.6)
inside eq. (5.7) gives:
Z
c(J, ∆) = N (J, ∆) dwdw̄ µ(w, w̄) dDisc[G(w, w̄)]×
Z (5.8)
µ(z, z̄)
dzdz̄ FJ,∆ (z, z̄)K(w, w̄, z, z̄) + (t ↔ u)
µ(w, w̄)

where we exchanged integration orders, and multiplied and divided by µ(w, w̄). This
has precisely the form of the Lorentzian inversion formula (2.18): Euclidean inversion
plus dispersion relation gives Lorentzian inversion. The interesting thing is that
comparison reveals the following identity:
Z
4N (J, ∆) µ(z, z̄)
G∆+1−d,J+d−1 (w, w̄) = dzdz̄ FJ,∆ (z, z̄)K(w, w̄, z, z̄) (5.9)
κJ+∆ Eucl µ(w, w̄)

This is predicted to hold for any d, J, and ∆. The integration is over the complex z
plane, i.e z̄ = z ∗ . This equation is the analog of the relation between the Legendre
polynomials of the first and second kind (See table 1):
Z 1
1 dyPJ (y)
QJ (x) = . (5.10)
2 −1 y−x

While we haven’t checked this relation in the general case, in appendix B we verify
it in the special case that d = 2 and (a, b) = (0, 12 ). It would be interesting to
understand the relationship between this identity and the light transform of [20].

5.4 3D Ising model and analytic functionals

In the 3D Ising model, we now present numerical tests for the correlator of four
Z2 -odd operators (hσσσσi), and discuss a possible way to reorganize the crossing
equations.

– 32 –
 A straightforward exercise (if somewhat technical) is to numerically integrate
the subtracted dispersion relation in eq. (4.14), using the OPE data tabulated in [3]
to compute dDisc[G] as a sum over t-channel blocks. It is important conceptually
that the t-channel OPE commutes with the dispersion relation. The basic reason is
that the interior of the integration region, w, w̄ ∈ (0, 1), lies within the convergence
radius of the OPE. It is also important to be careful near endpoints which lie at
the boundary of convergence [51]. In our case these are the collinear limit w = 0
and Regge corner w, w̄ → 0. Due to absolute convergence of the (subtracted) kernel
against the full correlator, and thanks to positivity of the OPE, we expect that the
operations safely commute.
The most important numbers (with uncertainty in the last digit, see [3]) are:

∆σ = 0.518149, ∆ = 1.41263, fσσ = 1.051854. (5.11)

We used the 3D→2D dimensional reduction formulas of ref. [52] to efficiently compute
the 3D conformal blocks. Breaking the contributions into those of dominant opera-
tors and families, we find for example the correlator at a specific point (z, z̄) = ( 12 , 14 )

v ∆σ G( 21 , 14 ) = 0.368781 + 0.29685 + 0.16493T + 0.00568[σσ]0,≥4 + 0.000300 + 0.00018

≈ 0.83672 (5.12)

where the superscript label the cross-channel operator(s), and the last term collects
all the other operators ([σσ]1 and []0 families) recorded in [3]. Some comments are in
order about the lowest twist trajectory [σσ]0 . We separated the spin-2 contribution
(stress-tensor T ) from the others of spin J ≥ 4. Naively, one may have expect
the whole trajectory to be suppressed by a sin2 (π(τ /2 − ∆σ )) factor, however, as
explained below eq. (4.14), the subtracted dispersion relation involves dDisc[Gu/v]
which prevents that cancelation for the lowest twist (the usual suppressions still
operate for [σσ]1 ). As indicated in fig. 7a, by evaluating the contribution of these
operators up to spins O(20) and fitting to a power-law, we find that we can accurately
resum the trajectory (the fit in the figure used spins 20,22,24).
Performing the similar calculation at crossing-related points ( 21 , 34 ) we find:

v ∆σ G( 12 , 34 ) = 0.616351 + 0.19655 + 0.02951T − 0.00597[σσ]0,≥4 + 0.000110 + 0.00003

≈ 0.83659. (5.13)

The difference of these two numbers is a crossing equation, satisfied at the 10−4 level:

X( 12 , 14 ) = −0.00013, X(z, z̄) ≡ v ∆σ G(u, v) − u∆σ G(v, u) .

(5.14)

For comparison, calculating the same correlators vG at these particular points us-
ing the Euclidean OPE (and same OPE coefficients), we find a compatible value
vG = 0.83657 (with a change of ±2 in the last digit between the two crossing-related

– 33 –
positions). The agreement convincingly shows that the dispersion relation indeed re-
constructs the correlator. It is presently not clear whether the (small, but significant)
10−4 error in (5.12) is due to numerical integration or truncation of the spectrum.
Conceptually, one may be concerned that the sensitivity to the lowest twist
trajectory means that the formula requires more than the absorptive part. However,
in practice, the lowest-twist data is particularly well understood from the Lorentzian
inversion formula. A very crude approximation (simply feeding the identity and  into
the inversion formula, following [23]), for example reproduces the OPE coefficients
of [3] to per-mil accuracy; we used this approximation in the above, for spins 12
and higher. Conceptually, one may view the first four contributions in eq. (5.12) as
accurately parametrized (to per-mil level) simply by three parameters: (∆σ , ∆ , fσσ ).
It is amusing to try to constraint this crude model, for example for a given ∆σ
one can find (∆ , fσσ ) which minimizes the error in the crossing relation (and in
the twist of the stress tensor). Preliminary investigations yield a curve(s) passing
through the numerical bootstrap solution (5.11), with values of ∆ differing by less
than ±0.01 when considering different crossing equations (we did not observe any
kink). Possibly, to close the system and also fix ∆σ by such methods, one will need
to consider mixed correlators.

J X(1/2,1/4)
8 12 16 20 24
0.000 0.15
●
● ● ●
J=0 J=2
● ●
-0.001
●
21.21 0.10
-
vG(1/2,1/4)

-0.002 ●
3.474
J
-0.003 ● 0.05
-0.004
● Δ
-0.005 ● 2 4 6 8 10 12
-0.006 -0.05

(a) (b)

Figure 7. (a) High-spin tail of the contribution of the lowest-twist family [σσ]0,J to the
correlator in eq. (5.12). Operators of spin 28 and higher were resummed using the power-
law fit. (b) “Bootstrap functional” obtained by evaluating the contribution of spin-0 and
spin-2 exchanged operators of dimension ∆ to the dispersion integral X( 12 , 14 ) in eq. (5.14),
with external dimensions fixed to ∆σ . A rescaling envelope was applied. Note that the
curves have double zeros at non-leading double-twist dimensions.

It is interesting that the contributions of individual blocks are very different

between the Euclidean OPE and dispersion relation: the dispersion relation is not
a term-wise rewriting of the OPE. This becomes particularly sharp if we plot the
contribution to a crossing equation, say X( 12 , 14 ), from a given cross-channel operator.

– 34 –
This gives a “bootstrap functional”, shown in fig. 7b, which must be orthogonal to
the OPE data. That particular functional has double zeros at all (non-leading)
double-twist operators, and is mostly positive (with the exception of the identity
contribution, and some lowest-twist operators at high spin, not shown). In contrast,
Euclidean functionals display no such oscillatory behavior.
One can create a few more functionals of this type. For example, t↔u cross-
ing symmetry is not manifest because of the subtraction (4.14), giving a nontrivial
constraint:
z z̄


X̃(z, z̄) ≡ G(z, z̄) − G z−1 , z̄−1 = 0. (5.15)
A special case includes the Regge limit, for example the correlator should be real for
imaginary ρ’s:

Im G(ρz = ia, ρ̄z = −ib) = 0, (0 < a < 1 < b real). (5.16)

Such functionals will likely not form a complete basis (all have double zeros at the
double-twists, unlike some of those in [53]), but it would be interesting to compare
and perhaps combine them with other functionals like those found in [54, 55].

6 Conclusion
In this work we obtained a dispersion relation for four-point correlators of conformal
field theories, reconstructing them from an “absorptive part” (double discontinuity).
It’s kernel (given in eqs. (3.4), (3.8) and (3.25)) was found by explicitly resuming
the Lorentzian inversion formula of [18–20]. For non-equal external operators, a
differential equation was obtained, eq. (3.46). A subtracted dispersion relation, in
eq. (4.14), overcomes the limitations of the inversion formula fully reconstructs the
correlators in an arbitrary (unitary) conformal field theories. Various tests were
performed, including in holographic theories and the 3D Ising model.
The dispersion relation holds for d ≥ 2. For d = 1 there is only one cross ratio
z, and thus one could expect a simpler dispersion relation. Ref. [56] obtained a
crossing symmetric inversion formula in d = 1. The kernel in this inversion formula
is quite complicated for general scaling dimensions, precisely because it needs to
give rise to a crossing symmetric correlator. Combining a known d = 1 inversion
formula with the methods that we presented in this note, one can obtain a d = 1
dispersion relation [57] containing the double-discontinuity. It is also possible [57] to
obtain dispersion relations for boundary/defect CFTs by starting from the Lorentzian
inversion formulas of [58, 59]. Investigating the flat space limit of the dispersion
relation would also be interesting, as well as comparison with momentum space
approaches (for example [60]).
In our view, the most appealing feature is that the “absorptive part” (or dDisc)
on which the dispersion relation feeds can often be rather accurately approximated by

– 35 –
just the simplest exchanges, as discussed below eq. (5.14). This strongly suggests that
this is the right data around which to build a systematic expansion. Our hope is that
the dispersion relation presented here will help achieve that, as crossing symmetry
can now be directly formulated as a constraint on the dDisc. After subtracting the
simplest exchanges, the remainder of the dDisc should be a small, positive, and
regular function on a square (0<z, z̄<1). Finding how to “close” the equations and
bootstrap this function is in our view a key next question.

Acknowledgments

We thank Lorenzo Di Pietro, Shota Komatsu, Petr Kravchuk, Dalimil Mazac, Joao
Penedones, Balt van Rees and David Simmons-Duffin for discussions. DC thanks
McGill university and Caltech for hospitality where most of this work was done.
Part of this work was done during the Bootstrap 2019 conference at Perimeter in-
stitute and supported by the Simons foundation. Work of SCH is supported by the
National Science and Engineering Council of Canada, the Canada Research Chair
program, the Fonds de Recherche du Québec - Nature et Technologies, and the Si-
mons Collaboration on the Nonperturbative Bootstrap. DC is supported by the
European Research Council Starting Grant under grant no. 758903.

A Identities for spin sums

In this appendix we collect some of mathematical results that we use in the main
text. In Eq. 3.13 we used the following J sum which contains 2 hypergeometric
functions, which one can easily check by series expanding in w:
∞
X k−2J−2m0 (z)k2J+2m0 +2 (w)
J=0
1 + δJ,0
zw  m02 k−2m0 +2 (z)k2m0 +2 (w)  k−2m0 (z)k2m0 +2 (w)
= k−2m0 (z)k2m0 (w) − −
z−w 4(4m02 − 1) 2
 
1 zw   1
= 0
(1 − w)∂w − (1 − z)∂z − k−2m0 (z)k2m0 +2 (w).
2m + 1 z − w 2
(A.1)

The idea behind the first equality is that (1/z − 1/w) times the summand on the
first line can be rewritten, using hypergeometric identities, in terms of k-functions
with shifted arguments so as to turn the sum into a telescopic one. The second line
similarly re-interpret shifted k-functions in terms of derivatives on simple term. The
above identity is valid for arbitrary offset m0 , but a = b = 0. For non-zero a and
b, the second line form of the above identity becomes more complicated, but the

– 36 –
derivative form remains unchanged:
∞ (a,b) (a,b)
X k−2J−2m0 (z)k2J+2m0 +2 (w)
1 + δJ,0
J=0
 (A.2)
1 zw   1
(a,b) (a,b)
= 0
(1 − w)∂w − (1 − z)∂z − k−2m0 (z)k2m0 +2 (w).
2m + 1 z − w 2

Introducing the differential operator (the subscripts stands for d = 2)16 :

zw   z̄w  
D2 ≡ (1 − w)∂w − (1 − z)∂z + (1 − w)∂w − (1 − z̄)∂z̄ − 1, (A.3)
z−w z̄ − w
this allows us to rewrite the sum (3.13) in a concise form:
∞
X k−2J−2m0 −2a (z)k−2m0 −2a (z̄)k−2m0 −2a (w̄)(k2J+2m0 +2+2a (w)
+ (z↔z̄) (A.4)
J=0
1 + δJ,0
1 h i
= D2 k−2m0 (z)k−2m0 (z̄)k−2m0 (w̄)k2m0 +2 (w) . (A.5)
2m0 + 1

1
B Inverted block from harmonic function when a = 0, b = 2

In order to check the integral relation in Eq. 5.9, we fix d = 2 and a = 0, b = 12 . In

this case the dispersion relation contains only the contact term KC and not the bulk
term KB , as we saw in Eq. 3.49. Thus Eq. 5.9 becomes17 :
π
κ
Z
2 J+2−∆ µ(z, z̄) dρw
G∆−1,J+1 (w, w̄) = dzdz̄ FJ,∆ (z, z̄)δ(ρw − ρ̄w ρz ρ̄z )KC
1 + δJ,0 Eucl µ(w, w̄) dw
(B.1)
1
For a = 0, b = 2
the 2 F1 ’s of the conformal block basically simplifies to powers of ρ:

1
GJ,∆ (z, z̄) = (k∆−J (z)k∆+J (z̄) + k∆+J (z)k∆−J (z̄))
1 + δJ,0
∆−J ∆+J ∆+J ∆−J
22∆ (ρz ) 2 (ρ̄z ) 2 + (ρz ) 2 (ρ̄z ) 2 22∆ r∆ (eiθJ + e−iθJ )
= = (B.2)
1 + δJ,0 (1 − ρz )(1 − ρ̄z ) (1 + δJ,0 )|1 − reiθ |2

where in the last equality we changed to polar coordinates ρz = reiθ , and used the
fact that in Euclidean we have ρ̄z = ρ∗z . Similarly, we can write KC and the inverted
16
In terms of the ρ coordinates:
   
ρz ρw (1 − ρ2w )∂ρw − (1 − ρ2z )∂ρz ρz̄ ρw (1 − ρ2w )∂ρw − (1 − ρ2z̄ )∂ρ̄z
D2 ≡ + − 1.
(ρz − ρw )(1 − ρw ρz ) (ρz̄ − ρw )(1 − ρw ρz̄ )

17 4N (J,∆) 2(1+δJ,0 )
Where we used κJ+∆ = πκJ+2−∆ .

– 37 –
block in polar coordinates. Plugging these ingredients back in Eq. B.1, we can now
check that Eq. B.1 gives a correct result:
Z
1 µ(z, z̄) dρw
π dzdz̄ FJ,∆ (z, z̄)δ(ρw − ρ̄w ρz ρ̄z )KC
κ G
2 J+2−∆ ∆−1,J+1
(w, w̄) Eucl µ(w, w̄) dw
q
5
− 23 iθJ −iθJ ρw
ρw ρ̄w (1 − ρw ρ̄w )(ρw ρ̄w ) 2
2
−J Z 2π Z ∞ ∆
r +r 2−∆ (e + e )δ(r − ρ̄w
)
= 2−∆ ∆ ∆ 2−∆ dθ dr √
(ρw ) 2 (ρ̄w ) 2 + (ρw ) 2 (ρ̄w ) 2 0 0 4πr5 | ρw ρ̄w − e−iθ |2
(1 − ρw ρ̄w ) 2π (eiθJ + e−iθJ )dθ
Z
= J √
4π(ρw ρ̄w ) 2 0 | ρw ρ̄w − e−iθ |2
(1 − ρw ρ̄w ) (Z J−1 + Z −J−1 )dZ
Z
= J √ √ =1 (B.3)
4πi(ρw ρ̄w ) 2 |Z|=1 (Z − ρw ρ̄w )(Z −1 − ρw ρ̄w )

Where in the third line we performed the r integral over the delta function, in the
third line wrote the remaining θ integral as a contour integral in the complex Z ≡ eiθ
plane, and then used the residue theorem in the complex Z plane to get 1.

References
[1] R. Rattazzi, V. S. Rychkov, E. Tonni and A. Vichi, Bounding scalar operator
dimensions in 4D CFT, JHEP 12 (2008) 031, [0807.0004].
[2] S. El-Showk, M. F. Paulos, D. Poland, S. Rychkov, D. Simmons-Duffin and A. Vichi,
Solving the 3D Ising Model with the Conformal Bootstrap, Phys. Rev. D86 (2012)
025022, [1203.6064].
[3] D. Simmons-Duffin, The Lightcone Bootstrap and the Spectrum of the 3d Ising CFT,
JHEP 03 (2017) 086, [1612.08471].
[4] D. Poland, S. Rychkov and A. Vichi, The Conformal Bootstrap: Theory, Numerical
Techniques, and Applications, 1805.04405.
[5] Z. Komargodski and A. Zhiboedov, Convexity and Liberation at Large Spin, JHEP
11 (2013) 140, [1212.4103].
[6] A. L. Fitzpatrick, J. Kaplan, D. Poland and D. Simmons-Duffin, The Analytic
Bootstrap and AdS Superhorizon Locality, JHEP 12 (2013) 004, [1212.3616].
[7] L. F. Alday and A. Zhiboedov, An Algebraic Approach to the Analytic Bootstrap,
JHEP 04 (2017) 157, [1510.08091].
[8] L. F. Alday, Large Spin Perturbation Theory for Conformal Field Theories, Phys.
Rev. Lett. 119 (2017) 111601, [1611.01500].
[9] A. Martin, Scattering Theory: Unitarity, Analyticity and Crossing, Lect. Notes
Phys. 3 (1969) 1–117.
[10] R. J. Eden, P. V. Landshoff, D. I. Olive and J. C. Polkinghorne, eds., The analytic
S-matrix. Cambridge University Press, 1966.

– 38 –
[11] G. R. Screaton, ed., Dispersion Relations. Interscience publishers inc., 1961.
[12] M. A. Shifman, ed., Vacuum structure and QCD sum rules. Amsterdam,
Netherlands: North-Holland, 1992.
[13] Z. Komargodski and A. Schwimmer, On Renormalization Group Flows in Four
Dimensions, JHEP 12 (2011) 099, [1107.3987].
[14] Z. Bern, L. J. Dixon, D. C. Dunbar and D. A. Kosower, One loop n point gauge
theory amplitudes, unitarity and collinear limits, Nucl. Phys. B425 (1994) 217–260,
[hep-ph/9403226].
[15] R. Britto, F. Cachazo and B. Feng, Generalized unitarity and one-loop amplitudes in
N=4 super-Yang-Mills, Nucl. Phys. B725 (2005) 275–305, [hep-th/0412103].
[16] R. Britto, F. Cachazo, B. Feng and E. Witten, Direct proof of tree-level recursion
relation in Yang-Mills theory, Phys. Rev. Lett. 94 (2005) 181602, [hep-th/0501052].
[17] H. Elvang and Y.-t. Huang, Scattering Amplitudes, 1308.1697.
[18] S. Caron-Huot, Analyticity in Spin in Conformal Theories, JHEP 09 (2017) 078,
[1703.00278].
[19] D. Simmons-Duffin, D. Stanford and E. Witten, A spacetime derivation of the
Lorentzian OPE inversion formula, JHEP 07 (2018) 085, [1711.03816].
[20] P. Kravchuk and D. Simmons-Duffin, Light-ray operators in conformal field theory,
1805.00098.
[21] L. F. Alday, J. Henriksson and M. van Loon, Taming the epsilon-expansion with
large spin perturbation theory, JHEP 07 (2018) 131, [1712.02314].
[22] L. F. Alday, J. Henriksson and M. van Loon, An alternative to diagrams for the
critical O(N) model: dimensions and structure constants to order 1/N2, 1907.02445.
[23] S. Albayrak, D. Meltzer and D. Poland, More Analytic Bootstrap: Nonperturbative
Effects and Fermions, 1904.00032.
[24] C. Cardona, S. Guha, S. K. Kanumilli and K. Sen, Resummation at finite conformal
spin, JHEP 01 (2019) 077, [1811.00213].
[25] W. Li, Closed-form expression for cross-channel conformal blocks near the lightcone,
1906.00707.
[26] J. Liu, E. Perlmutter, V. Rosenhaus and D. Simmons-Duffin, d-dimensional SYK,
AdS Loops, and 6j Symbols, 1808.00612.
[27] L. F. Alday and S. Caron-Huot, Gravitational S-matrix from CFT dispersion
relations, 1711.02031.
[28] S. Caron-Huot and A.-K. Trinh, All tree-level correlators in = supergravity: hidden
ten-dimensional conformal symmetry, JHEP 01 (2019) 196, [1809.09173].
[29] M. Lemos, P. , M. Meineri and S. Sarkar, Universality at large transverse spin in
defect CFT, JHEP 09 (2018) 091, [1712.08185].

– 39 –
[30] L. Iliesiu, M. Kologlu, R. Mahajan, E. Perlmutter and D. Simmons-Duffin, The
Conformal Bootstrap at Finite Temperature, 1802.10266.
[31] A. Bissi, P. Dey and T. Hansen, Dispersion Relation for CFT Four-Point Functions,
1910.04661.
[32] A. Martin, Unitarity and high-energy behavior of scattering amplitudes, Phys. Rev.
129 (1963) 1432–1436.
[33] M. F. Paulos, J. Penedones, J. Toledo, B. C. van Rees and P. Vieira, The S-matrix
bootstrap II: two dimensional amplitudes, JHEP 11 (2017) 143, [1607.06110].
[34] M. F. Paulos, J. Penedones, J. Toledo, B. C. van Rees and P. Vieira, The S-matrix
Bootstrap III: Higher Dimensional Amplitudes, 1708.06765.
[35] M. Froissart, Phys. Rev. 123, 1053 (1961) .
[36] V. N. Gribov, Zh. Eksp. Teor. Fiz. Sov. Phys. JETP 14, 478,1395 (1961) .
[37] M. Hogervorst and S. Rychkov, Radial Coordinates for Conformal Blocks, Phys.
Rev. D87 (2013) 106004, [1303.1111].
[38] V. K. Dobrev, V. B. Petkova, S. G. Petrova and I. T. Todorov, Dynamical
Derivation of Vacuum Operator Product Expansion in Euclidean Conformal
Quantum Field Theory, Phys. Rev. D13 (1976) 887.
[39] V. K. Dobrev, G. Mack, V. B. Petkova, S. G. Petrova and I. T. Todorov, Harmonic
Analysis on the n-Dimensional Lorentz Group and Its Application to Conformal
Quantum Field Theory, Lect. Notes Phys. 63 (1977) 1–280.
[40] M. S. Costa, V. Goncalves and J. Penedones, Conformal Regge theory, JHEP 12
(2012) 091, [1209.4355].
[41] N. Gromov, V. Kazakov and G. Korchemsky, Exact Correlation Functions in
Conformal Fishnet Theory, JHEP 08 (2019) 123, [1808.02688].
[42] G. G.N. Watson, Notes on Generating Functions of Polynomials: (3) Polynomials of
Legendre and Gegenbauer, Journal of the London Mathematical Society s1-8 (1933) .
[43] A. Baranov, On Series Containing Products of Legendre Polynomials, Mathematical
Notes, vol. 80, no. 2, 2006. Translated from Matematicheskie Zametki, vol. 80, no.
2, 2006, .
[44] H. van Haeringen, A class of sums of Gegenbauer functions: Twenty-four sums in
closed form, Journal of Mathematical Physics 27, 938 (1986) .
[45] J. S. M. Rahman, M, Sums of products of ultraspherical functions, Journal of
Mathematical Physics 26, 627 (1985) .
[46] F. Aprile, J. M. Drummond, P. Heslop and H. Paul, Quantum Gravity from
Conformal Field Theory, JHEP 01 (2018) 035, [1706.02822].
[47] E. D’Hoker, D. Z. Freedman, S. D. Mathur, A. Matusis and L. Rastelli, Extremal
correlators in the AdS / CFT correspondence, hep-th/9908160.

– 40 –
[48] G. Arutyunov and S. Frolov, Four point functions of lowest weight CPOs in N=4
SYM(4) in supergravity approximation, Phys. Rev. D62 (2000) 064016,
[hep-th/0002170].
[49] F. A. Dolan and H. Osborn, Conformal four point functions and the operator
product expansion, Nucl. Phys. B599 (2001) 459–496, [hep-th/0011040].
[50] E. Panzer, Algorithms for the symbolic integration of hyperlogarithms with
applications to Feynman integrals, Comput. Phys. Commun. 188 (2015) 148–166,
[1403.3385].
[51] J. Qiao and S. Rychkov, Cut-touching linear functionals in the conformal bootstrap,
JHEP 06 (2017) 076, [1705.01357].
[52] M. Hogervorst, Dimensional Reduction for Conformal Blocks, JHEP 09 (2016) 017,
[1604.08913].
[53] D. Mazac, Analytic bounds and emergence of AdS2 physics from the conformal
bootstrap, JHEP 04 (2017) 146, [1611.10060].
[54] M. F. Paulos, Analytic Functional Bootstrap for CFTs in d > 1, 1910.08563.
[55] D. Mazˇ c, L. Rastelli and X. Zhou, A Basis of Analytic Functionals for CFTs in
General Dimension, 1910.12855.
[56] D. Mazac, A Crossing-Symmetric OPE Inversion Formula, 1812.02254.
[57] D. Carmi and A. Vichi, More on CFT dispersion relations, Work in progress .
[58] P. Liendo, Y. Linke and V. Schomerus, A Lorentzian inversion formula for defect
CFT, 1903.05222.
[59] D. Mazac, L. Rastelli and X. Zhou, An Analytic Approach to BCFT, 1812.09314.
[60] M. Gillioz, X. Lu and M. A. Luty, Scale Anomalies, States, and Rates in Conformal
Field Theory, JHEP 04 (2017) 171, [1612.07800].

– 41 –