Polyakov Bootstrap in Conformal Field Theory

• Polyakov bootstrap is a conformal field theory framework that uses local primaries, OPE, and crossing symmetry to enforce cancellation of spurious contributions.
• It evolved from Polyakov’s 1974 non-Hamiltonian method to modern approaches using Mellin amplitudes, Witten diagrams, and crossing-symmetric expansions.
• The method underpins advances in AdS/CFT, boundary CFT, and mixed-system analyses through self-consistent analytic sum rules and dispersion relations.

Polyakov bootstrap denotes a family of conformal-bootstrap formulations whose historical nucleus is A.M. Polyakov’s 1974 “non-Hamiltonian” approach to conformal quantum field theory, in which local primaries, the operator product expansion (OPE), conformal invariance, and four-point crossing symmetry are treated as the dynamical principle that determines scaling dimensions and OPE coefficients. In modern usage, the term most often refers to a crossing-symmetric expansion of correlators in Polyakov blocks, exchange Witten diagrams, or Mellin amplitudes, with OPE consistency imposed by cancellation of spurious contributions. In a distinct historical usage, it can also denote the Migdal–Polyakov or “old conformal” bootstrap based on self-consistent skeleton expansions of dressed two- and three-point functions (Rychkov, 2 Sep 2025, Gopakumar et al., 2016, Liendo et al., 2021).

1. Historical formation and terminological scope

Recent literature makes clear that “Polyakov bootstrap” has more than one historically valid meaning. This suggests that the term should be interpreted contextually rather than as the name of a single formalism.

Usage	Core object	Characteristic constraint
Migdal–Polyakov or “old conformal” bootstrap	Dressed 2- and 3-point functions	Skeleton self-consistency
Polyakov’s 1974 non-Hamiltonian bootstrap	Local primaries and OPE data	Four-point crossing symmetry
Modern Polyakov bootstrap	Crossing-symmetric exchange basis	Cancellation of spurious terms or poles

The historical setting emphasized by Rychkov begins in the late 1960s and early 1970s, when strong interactions, critical phenomena, scale invariance, and conformal invariance were being studied in a common conceptual arena. Polyakov’s 1970 letter “Conformal symmetry of critical fluctuations” is singled out as a turning point because it derived the forms of 3- and 4-point functions from conformal invariance, argued that critical-point correlators should be conformally invariant, and checked this against the 2D Ising model. Rychkov explicitly identifies Polyakov’s 1974 paper, “Nonhamiltonian approach to conformal quantum field theory,” as “the birth of modern conformal bootstrap” (Rychkov, 2 Sep 2025).

The same historical account distinguishes Polyakov’s contribution from those of the Rome and Sofia schools. The Rome group developed conformally covariant OPEs, shadow formalism, conformal partial waves, and early unitarity bounds, while Mack and the Sofia group developed harmonic-analysis and crossing-kernel structures that later look strikingly modern. Rychkov’s point is not that Polyakov alone had crossing or OPE, but that he most forcefully fused them into a self-contained dynamical program (Rychkov, 2 Sep 2025).

A separate line of development is the “old” conformal bootstrap of Migdal and Polyakov, revived with modern multi-loop technology. In that framework the unknowns are exact scaling dimensions and dressed couplings appearing in conformally covariant propagators and 1PI vertices, and the bootstrap equations are skeleton self-reproduction conditions rather than four-point crossing equations. The literature explicitly distinguishes this program from the modern crossing-based bootstrap, even though both inherit Polyakov’s broader bootstrap philosophy (Liendo et al., 2021).

2. Dynamical principle: OPE, crossing, and conformal invariance

The core 1974 idea, as reconstructed by Rychkov, is structurally close to the modern conformal bootstrap. One postulates a complete set of local primary operators, uses the OPE as the fundamental algebraic structure, exploits the fact that conformal invariance fixes the coordinate dependence of OPE coefficient functions up to constants, reconstructs higher-point correlators from OPE data, and then imposes crossing symmetry of the four-point function as the nontrivial consistency condition. Rychkov quotes Polyakov’s summary of the program: “Finally, our program consists in calculating all the functions $C$ to within a few constants, substituting the operator expansion into the four-point function and finding the unknown constants from the crossing-symmetry requirement” (Rychkov, 2 Sep 2025).

In modern notation, the reduced scalar four-point function is expanded as

$$g(u,v)=1+\sum_{\Delta,\ell} \lambda_{\Delta,\ell}^2\, g_{\Delta,\ell}(u,v),$$

and crossing takes the form

$$v^{\Delta_H}\left[1+\sum_{\Delta,\ell} \lambda_{\Delta,\ell}^2 g_{\Delta,\ell}(u,v) \right] = u^{\Delta_H}\left[1+\sum_{\Delta,\ell} \lambda_{\Delta,\ell}^2 g_{\Delta,\ell}(v,u)\right].$$

This is the channel-equality form of bootstrap that later became standard (Rychkov, 2 Sep 2025).

What distinguishes Polyakov’s own implementation from later Euclidean block bootstrap is not the credo but the representation. Rychkov stresses that Polyakov worked in Lorentzian signature, used dispersion relations, and organized the problem in terms of “algebraic amplitudes” and “unitary amplitudes.” These were meant to reconcile crossing, OPE, and acceptable singularity structure. Polyakov’s proposal was to sum $s$-, $t$-, and $u$-channel unitary amplitudes and require cancellation of the logarithmic terms that violate the OPE. This is the conceptual ancestor of the modern emphasis on manifestly crossing-symmetric building blocks and on removing unphysical contributions (Rychkov, 2 Sep 2025).

The same source also emphasizes that Polyakov already treated the OPE as more than a formal asymptotic device, making “a very important observation that OPE is not asymptotic but has a finite radius of convergence.” In historical terms, that observation anticipates a central assumption of later conformal-bootstrap practice (Rychkov, 2 Sep 2025).

3. Mellin space, Witten diagrams, and Polyakov blocks

The modern revival of the Polyakov bootstrap occurs most sharply in Mellin space. The 2016 Polyakov–Mellin construction replaces a one-channel conformal-block expansion by a crossing-symmetric sum of exchange Witten diagrams,

$${\cal A}(u,v)= \sum_{\Delta, \ell} c_{\Delta, \ell}\bigg( W_{\Delta,\ell}^{(s)}(u,v)+W_{\Delta,\ell}^{(t)}(u,v)+W_{\Delta,\ell}^{(u)}(u,v) \bigg),$$

and interprets OPE consistency as the cancellation of terms such as $u^{\Delta_\phi}$ and $u^{\Delta_\phi}\log u$, which are spurious unless produced by actual operators in the spectrum (Gopakumar et al., 2016).

In Mellin space this becomes especially transparent. A refined crossing-symmetric ansatz is

$$\mathcal A(u,v)=\int \frac{ds\,dt}{(2\pi i)^2}\,u^s v^t \Bigg[ \sum_{\Delta,\ell} c_{\Delta,\ell} \Big(M^{(s)}_{\Delta,\ell}(s,t)+M^{(t)}_{\Delta,\ell}(s,t)+M^{(u)}_{\Delta,\ell}(s,t)\Big) +M^{(c)}(s,t) \Bigg]\rho_{\Delta_\phi}(s,t),$$

with

$$g(u,v)=1+\sum_{\Delta,\ell} \lambda_{\Delta,\ell}^2\, g_{\Delta,\ell}(u,v),$$0

The Mellin measure has poles at

$$g(u,v)=1+\sum_{\Delta,\ell} \lambda_{\Delta,\ell}^2\, g_{\Delta,\ell}(u,v),$$1

so OPE consistency becomes the requirement that the coefficients of these spurious poles vanish after summing over exchanged operators and contact terms (Gopakumar et al., 2018).

A major refinement of this program is the realization that exchange Witten diagrams alone are generally incomplete. Contact Witten diagrams are required both in the large-$$g(u,v)=1+\sum_{\Delta,\ell} \lambda_{\Delta,\ell}^2\, g_{\Delta,\ell}(u,v),$$2 holographic bootstrap and in the $$g(u,v)=1+\sum_{\Delta,\ell} \lambda_{\Delta,\ell}^2\, g_{\Delta,\ell}(u,v),$$3-expansion beyond the lowest orders. In this literature, contact terms are not optional decorations but part of the correct crossing-symmetric basis. The same work also derives explicit Mellin-space formulas for crossed-channel Witten diagrams expanded in $$g(u,v)=1+\sum_{\Delta,\ell} \lambda_{\Delta,\ell}^2\, g_{\Delta,\ell}(u,v),$$4-channel partial waves, including a compact very-well-poised $$g(u,v)=1+\sum_{\Delta,\ell} \lambda_{\Delta,\ell}^2\, g_{\Delta,\ell}(u,v),$$5 expression that plays the role of a Witten-diagram crossing kernel (Gopakumar et al., 2018).

The subsequent problem was uniqueness. “The Unique Polyakov Blocks” resolves the contact-term ambiguity by introducing cyclic Polyakov blocks and fixing the polynomial ambiguity through Regge-type boundedness. In the cyclic basis the exchange ambiguity polynomial $$g(u,v)=1+\sum_{\Delta,\ell} \lambda_{\Delta,\ell}^2\, g_{\Delta,\ell}(u,v),$$6 is set to zero, producing canonical Mellin-space Polyakov blocks for arbitrary exchanged twist and spin. The same paper emphasizes the relation between cyclic amplitudes, Mellin-space dispersion relations, direct-channel double-twist data, and analytic bootstrap functionals (Sleight et al., 2019).

This Mellin-space framework is not merely structural. In holographic large-$$g(u,v)=1+\sum_{\Delta,\ell} \lambda_{\Delta,\ell}^2\, g_{\Delta,\ell}(u,v),$$7 CFTs, Polyakov–Mellin bootstrap has been pushed through $$g(u,v)=1+\sum_{\Delta,\ell} \lambda_{\Delta,\ell}^2\, g_{\Delta,\ell}(u,v),$$8, with contact terms corresponding to bulk $$g(u,v)=1+\sum_{\Delta,\ell} \lambda_{\Delta,\ell}^2\, g_{\Delta,\ell}(u,v),$$9 interactions, closed-form large-spin anomalous dimensions, and reconstruction of bubble and triangle AdS loop amplitudes from CFT data. In that sense the Polyakov bootstrap becomes a loop-level bootstrap for AdS/CFT (Ghosh, 2018).

4. Functional formulations and exact lower-dimensional realizations

In one dimension, the relationship between ordinary crossing, analytic functionals, and Polyakov blocks becomes fully explicit. For the $$v^{\Delta_H}\left[1+\sum_{\Delta,\ell} \lambda_{\Delta,\ell}^2 g_{\Delta,\ell}(u,v) \right] = u^{\Delta_H}\left[1+\sum_{\Delta,\ell} \lambda_{\Delta,\ell}^2 g_{\Delta,\ell}(v,u)\right].$$0 crossing equation,

$$v^{\Delta_H}\left[1+\sum_{\Delta,\ell} \lambda_{\Delta,\ell}^2 g_{\Delta,\ell}(u,v) \right] = u^{\Delta_H}\left[1+\sum_{\Delta,\ell} \lambda_{\Delta,\ell}^2 g_{\Delta,\ell}(v,u)\right].$$1

one can construct complete functional bases $$v^{\Delta_H}\left[1+\sum_{\Delta,\ell} \lambda_{\Delta,\ell}^2 g_{\Delta,\ell}(u,v) \right] = u^{\Delta_H}\left[1+\sum_{\Delta,\ell} \lambda_{\Delta,\ell}^2 g_{\Delta,\ell}(v,u)\right].$$2 adapted to generalized free boson or fermion spectra. These bases satisfy the decomposition

$$v^{\Delta_H}\left[1+\sum_{\Delta,\ell} \lambda_{\Delta,\ell}^2 g_{\Delta,\ell}(u,v) \right] = u^{\Delta_H}\left[1+\sum_{\Delta,\ell} \lambda_{\Delta,\ell}^2 g_{\Delta,\ell}(v,u)\right].$$3

and the associated Polyakov block is

$$v^{\Delta_H}\left[1+\sum_{\Delta,\ell} \lambda_{\Delta,\ell}^2 g_{\Delta,\ell}(u,v) \right] = u^{\Delta_H}\left[1+\sum_{\Delta,\ell} \lambda_{\Delta,\ell}^2 g_{\Delta,\ell}(v,u)\right].$$4

The resulting sum rules

$$v^{\Delta_H}\left[1+\sum_{\Delta,\ell} \lambda_{\Delta,\ell}^2 g_{\Delta,\ell}(u,v) \right] = u^{\Delta_H}\left[1+\sum_{\Delta,\ell} \lambda_{\Delta,\ell}^2 g_{\Delta,\ell}(v,u)\right].$$5

are exactly the Polyakov sum rules arising from cancellation of spurious double traces. In this setting the Polyakov bootstrap is not an ansatz but an exact reformulation of crossing, provided one imposes the Regge boundedness automatically satisfied by unitary correlators (Mazac et al., 2018).

The same paper identifies the Polyakov blocks with crossing-symmetrized $$v^{\Delta_H}\left[1+\sum_{\Delta,\ell} \lambda_{\Delta,\ell}^2 g_{\Delta,\ell}(u,v) \right] = u^{\Delta_H}\left[1+\sum_{\Delta,\ell} \lambda_{\Delta,\ell}^2 g_{\Delta,\ell}(v,u)\right].$$6 exchange Witten diagrams. In the bosonic case the contact-term ambiguity is fixed by requiring the corresponding analytic functionals to satisfy the correct Regge/swapping conditions. In the fermionic case there is no analogous ambiguity at the same Regge-bounded level, so the Polyakov blocks are canonical (Mazac et al., 2018).

An analogous bridge exists in boundary CFT. For scalar two-point functions in a half-space, one can construct a basis of linear functionals $$v^{\Delta_H}\left[1+\sum_{\Delta,\ell} \lambda_{\Delta,\ell}^2 g_{\Delta,\ell}(u,v) \right] = u^{\Delta_H}\left[1+\sum_{\Delta,\ell} \lambda_{\Delta,\ell}^2 g_{\Delta,\ell}(v,u)\right].$$7 adapted to the generalized free field solution with Dirichlet or Neumann boundary conditions. The resulting functional bootstrap equations are “essentially equivalent to a Polyakov-type approach to the bootstrap of BCFTs.” The BCFT Polyakov blocks are manifestly crossing-symmetric combinations built out of physical exchange plus towers of generalized-free-field states, and the functionals literally compute the coefficients of those towers. In the Dirichlet case the Polyakov blocks coincide with exchange Witten diagrams; in the Neumann case a contact-term ambiguity remains and is fixed by the functional analysis (Kaviraj et al., 2018).

5. Charged, mixed, and higher-point extensions

The Polyakov bootstrap has also been extended beyond the single identical-scalar four-point problem. For charged correlators on the line, the crossing equation becomes matrix-valued in representation space. The construction of a generalized-free-field-dual functional basis leads to a charged Polyakov bootstrap in which each irrep $$v^{\Delta_H}\left[1+\sum_{\Delta,\ell} \lambda_{\Delta,\ell}^2 g_{\Delta,\ell}(u,v) \right] = u^{\Delta_H}\left[1+\sum_{\Delta,\ell} \lambda_{\Delta,\ell}^2 g_{\Delta,\ell}(v,u)\right].$$8 carries its own generalized-free spectrum,

$$v^{\Delta_H}\left[1+\sum_{\Delta,\ell} \lambda_{\Delta,\ell}^2 g_{\Delta,\ell}(u,v) \right] = u^{\Delta_H}\left[1+\sum_{\Delta,\ell} \lambda_{\Delta,\ell}^2 g_{\Delta,\ell}(v,u)\right].$$9

and the correlator admits a Polyakov expansion

$s$0

Master functionals and a dispersion relation then encode the charged Polyakov sum rules. The paper’s explicit point is that this is a charged version of the Polyakov bootstrap, with representation-dependent spurious generalized-free towers and symmetry-dependent contact terms (Ghosh et al., 2021).

For arbitrary mixed systems of 1D correlators, Polyakov blocks can be defined for exchanged operators carrying both a scaling dimension and an OPE orientation vector. The resulting sum rules are diagonalized by correlators in tensor-product generalized free theories, and the basis can be changed again so that mixed systems involving elementary and composite operators in a single field theory are diagonalized instead. An explicit example is the $s$1 system in a single generalized free field, which provides the first non-trivial examples of optimal bounds saturated by a mixed-correlator solution of this type (Ghosh et al., 2023).

The same logic has now been carried to higher-point functions. A 2025 construction generalizes the functionals constituting the Polyakov bootstrap of four-point correlators to five-point functions in one-dimensional CFTs. Crossing-symmetric five-point Polyakov blocks are built from double-exchange, single-exchange, and contact Witten-diagram ingredients, and OPE consistency produces new sum rules controlled by two classes of functionals,

$s$2

which govern double-twist and triple-twist families. The conceptual claim is that higher-point correlators encode the data of infinitely many four-point correlators, so the five-point Polyakov bootstrap is already a bootstrap of infinitely many four-point data at once (Antunes et al., 7 Aug 2025).

6. Relation to other bootstrap programs and unresolved questions

A recurrent source of confusion is the relation between Polyakov bootstrap and the standard numerical conformal bootstrap. A 2023 thesis on bootstrap methods in theoretical physics is explicit on this point: it is only tangentially relevant to the Polyakov bootstrap proper. Its CFT discussion develops the standard conformal bootstrap in the numerical and functional sense—crossing symmetry, OPE associativity, conformal blocks, numerical functionals, analytic functionals, SDP/SDPB, JuliBoots, correlator bounds, gap maximization, and OPE coefficient bounds—but it does not explicitly develop Polyakov blocks, crossing-symmetric Polyakov expansions, Mellin bootstrap, exchange Witten-diagram decompositions, or Polyakov conditions for cancellation of spurious terms (Zheng, 2023).

A second confusion concerns the “old conformal bootstrap.” The Migdal–Polyakov framework revisited in recent work is a concrete calculational scheme based on conformally covariant propagators and exact 1PI vertices, with bootstrap equations of the form

$s$3

or their $s$4 analogues. This is a conformal Schwinger–Dyson or skeleton bootstrap for dressed two- and three-point functions, not a Polyakov-block or Mellin-space bootstrap for four-point crossing (Liendo et al., 2021).

The modern Polyakov bootstrap itself contains several open structural issues. One is contact-term ambiguity: the literature now treats it as central rather than peripheral, and several uniqueness prescriptions are formulated precisely to control it (Gopakumar et al., 2018, Sleight et al., 2019). Another is completeness: in some mixed-correlator and higher-point constructions the sum rules are tested extensively and used successfully, but full equivalence to crossing is not always proved in complete generality (Ghosh et al., 2023, Antunes et al., 7 Aug 2025). Rychkov also notes broader unresolved questions, including the precise relation between Polyakov-style constructions and other modern analytic approaches, even while arguing that Polyakov should be regarded as the key originator of conformal bootstrap as a dynamical principle (Rychkov, 2 Sep 2025).

In that restricted but precise sense, the Polyakov bootstrap is best understood not as a single algorithm but as a lineage of crossing-symmetric bootstrap ideas. Its enduring content is the attempt to write correlators in a basis where crossing is manifest, OPE consistency is imposed by removing unphysical contributions, and the data of a conformal field theory are determined without privileging a particular channel from the outset.