Abstract
We prove that in any unitary CFT, a twist gap in the spectrum of operator product expansion (OPE) of identical scalar quasiprimary operators (i.e. \(\phi \times \phi \)) implies the existence of a family of quasiprimary operators \({\mathcal {O}}_{\tau , \ell }\) with spins \(\ell \rightarrow \infty \) and twists \(\tau \rightarrow 2 \Delta _{\phi }\) in the same OPE spectrum. A similar twist-accumulation result is proven for any two-dimensional Virasoro-invariant, modular-invariant, unitary CFT with a normalizable vacuum and central charge \(c > 1\), where we show that a twist gap in the spectrum of Virasoro primaries implies the existence of a family of Virasoro primaries \({\mathcal {O}}_{h, {\bar{h}}}\) with \(h \rightarrow \infty \) and \({\bar{h}} \rightarrow \frac{c - 1}{24}\) (the same is true with h and \({\bar{h}}\) interchanged). We summarize the similarity of the two problems and propose a general formulation of the lightcone bootstrap.
Similar content being viewed by others
Data availability statement
Data sharing is not applicable to this article as no datasets were generated or analyzed.
Notes
In this paper we use the word “quasiprimary” to denote primary w.r.t. the global conformal group, even in \(d>2\) where the word primary is often used for this purpose.
Operators of twist exactly equal or close to \(\tau _{\textrm{gap}}\) need not exist. In other words, we do not assume that \(\tau _{\textrm{gap}}\) is the maximal twist gap.
For completeness, we mention two further ideas which were used in the past to argue for the twist accumulation. These ideas appear even more nontrivial to make rigorous and they will not play a role in our work. (1) Ref. [1] phrased its argument in terms of a spectral density in the twist space, which is a linear functional acting by integrating against continuous functions \(f (\tau )\), and defined as a \(z \rightarrow 0\) limit of a family of functionals. However [1] did not provide a proof that the limit exists, and this appears quite nontrivial. See the discussion in [12], footnote 26. (2) Ref. [2], Section 3.5, gave an argument following the ideas of Alday-Maldacena [13] which view the CFT 4pt function as a 4pt function in a massive 2D theory.
Since the convergent expansion (2.2) consists of positive terms, the series defining \(g_{\tau \geqslant \tau _{*}}\) is convergent and \(g_{\tau \geqslant \tau _{*}}\) is well defined.
For \(d = 2\) we have \(\nu = 0\) and the coefficient \(\frac{(\nu )_{\ell }}{(2 \nu )_{\ell }}\) is defined as \(\lim \nolimits _{\nu \rightarrow 0^+} \frac{(\nu )_{\ell }}{(2 \nu )_{\ell }} = \frac{1 + \delta _{\ell , 0}}{2}\)
Surprisingly, we do not know any other general rigorous proof of this result.
The proposition remains true in this formulation, as it is clear from its proof.
Collinear blocks are blocks associated with collinear primaries, which are annihilated by the generator \({\bar{L}}_1\) of the 2D global conformal group acting in the \(z, {\bar{z}}\) plane, but not necessarily by the generators \(L_1\) and \(K_i\). Every conformal quasiprimary multiplet splits into infinitely many collinear primary multiplets.
In this section “primary” stands for Virasoro primary.
The limit \(\beta \rightarrow 0\), \(\beta \rightarrow \infty \), \({\mathfrak {b}}= C\) with a nonzero C would also work for the purpose of showing \(T\leqslant A\).
We use here \(\log (x + C) - \log (x) = \int _x^{x + C} \frac{d y}{y} \leqslant C / x\).
In [5] the first equality in Eq. (2.11) is the statement of vacuum dominance, true for fixed \(\beta \) for \({\bar{\beta }} \rightarrow 0\) (in their convention of \(\beta \) and \({\bar{\beta }}\)). But then it is assumed to be true in \(\beta \rightarrow \infty \) in the final step of the argument.
In [5] the second equality in Eq. (2.11), assuming it remains true in the double lightcone limit, tells us how the vacuum dominance can be achieved from the crossed channel perspective.
To see why this treatment is subtle we consider the following toy function \(f (\beta ) = \sin (e^{k \beta })\) with k fixed but arbitrarily large. \(f (\beta )\) is an O(1) function because it is bounded by 1. However, its derivative \(f' (\beta ) = k e^{k \beta } \cos (e^{k \beta })\) is not bounded, indeed it grows exponentially fast in \(\beta \). This argument tells us that we may not differentiate after estimate. We believe the argument of [7] can be made rigorous after a more careful treatment basically because there are no factors like \(f(\beta )\) appearing in the error terms of their Eq. (2.13).
“Twist” here has nothing to do with \(\tau = \Delta - \ell \) in the rest of the paper.
To complete the list of analogies, we note that Theorem 3.1 may also be argued by expressing the vacuum character in the direct channel of the partition function as an integral of crossed-channel characters, the spectral density in this integral representation being a nonzero function at \(h, {\bar{h}} > (c - 1) / 24\), see [10] App. B.1.
To estimate Pochhammer ratios in this equation and the next one, the following lemma is useful: Let \(x, y > 0\) and let k be a nonnegative integer such that \(y + k \geqslant x\). Then \( (x)_n / (y)_n \leqslant (1 + n / y)^k\) for all \(n \geqslant 0\). For a proof, rewrite \((x)_n / (y)_n\) as \(\frac{(x)_n}{(y + k)_n} \times \frac{(y + k)_n}{(y)_n}\). The first factor is \(\leqslant 1\). For the second factor we use the identity \(\frac{(y + k)_n}{(y)_n} = \frac{\Gamma (y + k + n) \Gamma (y)}{\Gamma (y + k) \Gamma (y + n)} = \frac{(y + n)_k}{(y)_k}\). The last expression is \(\leqslant (1 + n / y)^k\).
References
Fitzpatrick, A.L., Kaplan, J., Poland, D., Simmons-Duffin, D.: The analytic bootstrap and AdS superhorizon locality. JHEP 12, 004 (2013). https://doi.org/10.1007/JHEP12(2013)004. arXiv:1212.3616 [hep-th]
Komargodski, Z., Zhiboedov, A.: Convexity and liberation at large spin. JHEP 11, 140 (2013). https://doi.org/10.1007/JHEP11(2013)140. arXiv:1212.4103 [hep-th]
Parisi, G.: How to measure the dimension of the parton field. Nucl. Phys. B 59, 641–646 (1973). https://doi.org/10.1016/0550-3213(73)90666-4
Callan, C.G., Jr., Gross, D.J.: Bjorken scaling in quantum field theory. Phys. Rev. D 8, 4383–4394 (1973). https://doi.org/10.1103/PhysRevD.8.4383
Collier, S., Lin, Y.-H., Yin, X.: Modular bootstrap revisited. JHEP 09, 061 (2018). https://doi.org/10.1007/JHEP09(2018)061. arXiv:1608.06241 [hep-th]
Afkhami-Jeddi, N., Colville, K., Hartman, T., Maloney, A., Perlmutter, E.: Constraints on higher spin CFT\(_{2}\). JHEP 05, 092 (2018). https://doi.org/10.1007/JHEP05(2018)092. arXiv:1707.07717 [hep-th]
Benjamin, N., Ooguri, H., Shao, S.-H., Wang, Y.: Light-cone modular bootstrap and pure gravity. Phys. Rev. D 100(6), 066029 (2019). https://doi.org/10.1103/PhysRevD.100.066029. arXiv:1906.04184 [hep-th]
Kravchuk, P., Qiao, J., Rychkov, S.: Distributions in CFT. Part II. Minkowski space. JHEP 08, 094 (2021). https://doi.org/10.1007/JHEP08(2021)094. arXiv:2104.02090 [hep-th]
Pappadopulo, D., Rychkov, S., Espin, J., Rattazzi, R.: OPE convergence in conformal field theory. Phys. Rev. D 86, 105043 (2012). https://doi.org/10.1103/PhysRevD.86.105043. arXiv:1208.6449 [hep-th]
Kusuki, Y.: Light cone bootstrap in general 2D CFTs and entanglement from light cone singularity. JHEP 01, 025 (2019). arXiv:1810.01335 [hep-th]
Collier, S., Gobeil, Y., Maxfield, H., Perlmutter, E.: Quantum Regge trajectories and the Virasoro analytic bootstrap. JHEP 05, 212 (2019). https://doi.org/10.1007/JHEP05(2019)212. arXiv:1811.05710 [hep-th]
Qiao, J., Rychkov, S.: A Tauberian theorem for the conformal bootstrap. JHEP 12, 119 (2017). https://doi.org/10.1007/JHEP12(2017)119. arXiv:1709.00008 [hep-th]
Alday, L.F., Maldacena, J.M.: Comments on operators with large spin. JHEP 11, 019 (2007). https://doi.org/10.1088/1126-6708/2007/11/019. arXiv:0708.0672 [hep-th]
Hogervorst, M.: Dimensional reduction for conformal blocks. JHEP 09, 017 (2016). https://doi.org/10.1007/JHEP09(2016)017. arXiv:1604.08913 [hep-th]
Karateev, D., Kravchuk, P., Serone, M., Vichi, A.: Fermion conformal bootstrap in 4d. JHEP 06, 088 (2019). https://doi.org/10.1007/JHEP06(2019)088. arXiv:1902.05969 [hep-th]
Fitzpatrick, A.L., Kaplan, J., Walters, M.T., Wang, J.: Eikonalization of conformal blocks. JHEP 09, 019 (2015). https://doi.org/10.1007/JHEP09(2015)019. arXiv:1504.01737 [hep-th]
Li, D., Meltzer, D., Poland, D.: Conformal collider physics from the lightcone bootstrap. JHEP 02, 143 (2016). https://doi.org/10.1007/JHEP02(2016)143. arXiv:1511.08025 [hep-th]
Seiberg, N.: Notes on quantum Liouville theory and quantum gravity. Prog. Theor. Phys. Suppl. 102, 319–349 (1990). https://doi.org/10.1143/PTPS.102.319
Maloney, A., Witten, E.: Quantum gravity partition functions in three dimensions. JHEP 02, 029 (2010). https://doi.org/10.1007/JHEP02(2010)029. arXiv:0712.0155 [hep-th]
Hartman, T., Mazáč, D., Rastelli, L.: Sphere packing and quantum gravity. JHEP 12, 048 (2019). https://doi.org/10.1007/JHEP12(2019)048. arXiv:1905.01319 [hep-th]
Dixon, L.J., Friedan, D., Martinec, E.J., Shenker, S.H.: The conformal field theory of orbifolds. Nucl. Phys. B 282, 13–73 (1987). https://doi.org/10.1016/0550-3213(87)90676-6
Ponsot, B., Teschner, J.: Liouville bootstrap via harmonic analysis on a noncompact quantum group, arXiv:hep-th/9911110
Ponsot, B., Teschner, J.: Clebsch–Gordan and Racah–Wigner coefficients for a continuous series of representations of U(q)(sl(2, R)). Commun. Math. Phys. 224, 613–655 (2001). https://doi.org/10.1007/PL00005590. arXiv:math/0007097
Yin, X.: Aspects of two-dimensional conformal field theories. PoS TASI2017, 003 (2017). https://doi.org/10.22323/1.305.0003
Dotsenko, V., Jacobsen, J.L., Lewis, M.-A., Picco, M.: Coupled Potts models: self-duality and fixed point structure. Nucl. Phys. B 546, 505–557 (1999). https://doi.org/10.1016/S0550-3213(99)00097-8. arXiv:cond-mat/9812227
Antunes, A., Behan, C.: Coupled minimal conformal field theory models revisited. Phys. Rev. Lett. 130(7), 071602 (2023). https://doi.org/10.1103/PhysRevLett.130.071602. arXiv:2211.16503 [hep-th]
Simmons-Duffin, D.: The lightcone bootstrap and the spectrum of the 3d Ising CFT. JHEP 03, 086 (2017). https://doi.org/10.1007/JHEP03(2017)086. arXiv:1612.08471 [hep-th]
Liu, J., Meltzer, D., Poland, D., Simmons-Duffin, D.: The Lorentzian inversion formula and the spectrum of the 3d O(2) CFT. JHEP 09, 115 (2020). https://doi.org/10.1007/JHEP09(2020)115. arXiv:2007.07914 [hep-th]. [Erratum: JHEP 01, 206 (2021)]
Caron-Huot, S., Gobeil, Y., Zahraee, Z.: The leading trajectory in the 2+1D Ising CFT, arXiv:2007.11647 [hep-th]
Caron-Huot, S.: Analyticity in spin in conformal theories. JHEP 09, 078 (2017). https://doi.org/10.1007/JHEP09(2017)078. arXiv:1703.00278 [hep-th]
Simmons-Duffin, D., Stanford, D., Witten, E.: A spacetime derivation of the Lorentzian OPE inversion formula. JHEP 07, 085 (2018). https://doi.org/10.1007/JHEP07(2018)085. arXiv:1711.03816 [hep-th]
Mukhametzhanov, B., Zhiboedov, A.: Analytic Euclidean bootstrap. JHEP 10, 270 (2019). https://doi.org/10.1007/JHEP10(2019)270. arXiv:1808.03212 [hep-th]
Mukhametzhanov, B., Zhiboedov, A.: Modular invariance, Tauberian theorems and microcanonical entropy. JHEP 10, 261 (2019). https://doi.org/10.1007/JHEP10(2019)261. arXiv:1904.06359 [hep-th]
Ganguly, S., Pal, S.: Bounds on the density of states and the spectral gap in CFT\(_{2}\). Phys. Rev. D 101(10), 106022 (2020). https://doi.org/10.1103/PhysRevD.101.106022. arXiv:1905.12636 [hep-th]
Pal, S., Sun, Z.: Tauberian-Cardy formula with spin. JHEP 01, 135 (2020). https://doi.org/10.1007/JHEP01(2020)135. arXiv:1910.07727 [hep-th]
Pal, S.: Bound on asymptotics of magnitude of three point coefficients in 2D CFT. JHEP 01, 023 (2020). https://doi.org/10.1007/JHEP01(2020)023. arXiv:1906.11223 [hep-th]
Mukhametzhanov, B., Pal, S.: Beurling-Selberg extremization and modular bootstrap at high energies. SciPost Phys. 8(6), 088 (2020). https://doi.org/10.21468/SciPostPhys.8.6.088. arXiv:2003.14316 [hep-th]
Pal, S., Sun, Z.: High energy modular bootstrap, global symmetries and defects. JHEP 08, 064 (2020). https://doi.org/10.1007/JHEP08(2020)064. arXiv:2004.12557 [hep-th]
Lamouret, Q.: Analytical bootstrap and double twist operators. Master thesis, Ecole Normale Supérieure (2022)
Li, W.: Lightcone expansions of conformal blocks in closed form. JHEP 06, 105 (2020). https://doi.org/10.1007/JHEP06(2020)105. arXiv:1912.01168 [hep-th]
Li, W.: Factorized lightcone expansion of conformal blocks. JHEP 05, 128 (2021). https://doi.org/10.1007/JHEP05(2021)128. arXiv:2012.09710 [hep-th]
Kos, F., Poland, D., Simmons-Duffin, D.: Bootstrapping the \(O(N)\) vector models. JHEP 06, 091 (2014). https://doi.org/10.1007/JHEP06(2014)091. arXiv:1307.6856 [hep-th]
Penedones, J., Trevisani, E., Yamazaki, M.: Recursion relations for conformal blocks. JHEP 09, 070 (2016). https://doi.org/10.1007/JHEP09(2016)070. arXiv:1509.00428 [hep-th]
Acknowledgements
Ideas of Quentin Lamouret [39] played a role at the genesis of this project. We thank Ying-Hsuan Lin, Dalimil Mazáč, Balt van Rees and Sylvain Ribault for useful discussions. When the project got initiated, SP was at the Institute for Advanced Study and he acknowledges a debt of gratitude for the funding provided by Tomislav and Vesna Kundic as well as the support from the grant DE-SC0009988 from the U.S. Department of Energy. SP also acknowledges the support by the U.S. Department of Energy, Office of Science, Office of High Energy Physics, under Award Number DE-SC0011632 and by the Walter Burke Institute for Theoretical Physics. JQ is grateful for the funding provided by École Normale Supérieure - PSL and the Simons Bootstrap Collaboration when this project was initiated in the workshop “Bootstrap 2022” at Porto University. JQ is supported by the Swiss National Science Foundation through the National Centre of Competence in Research SwissMAP and by the Simons Collaboration on Confinement and QCD Strings. SR is supported by the Simons Foundation grant 733758 (Simons Bootstrap Collaboration).
Author information
Authors and Affiliations
Corresponding author
Additional information
Communicated by Sakura Schafer-Nameki.
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Appendices
Approximate Factorization of Conformal Blocks
In this appendix we will prove Lemma 2.7. We will find it convenient to prove a slightly different version of the lemma:
Lemma A.1
(Approximate factorization, alternative version) Let \(a, b \in (0, 1)\), \(a < b\). Let \(d\geqslant 2\) and assume that \(\Delta , \ell \) satisfy the unitarity bounds. If \(\ell = 0\), assume in addition that \(\Delta > \nu = \frac{d - 2}{2}.\) Then the conformal block \(g_{\tau , \ell } (z, {\bar{z}})\) satisfies the following two-sided bound
where \(K_d (a, b)\) is a finite constant independent of \(\tau , \ell \) and of \(z, {\bar{z}}\) in the shown range.
Let us quickly see why this version implies Lemma 2.7. We have an identity:
We also have \(1 \leqslant G_{\beta } (\rho ) \leqslant \frac{1}{\sqrt{1 - \rho ^2}}\) (Lemma A.4 below). Hence the denominator of (A.1) and the denominator of (2.29) are related to each other by
where C(a) does not depend on \(\tau \) or \(\ell \).
We will first prove the 2D case (App. A.1), then give an inductive proof in \(d \geqslant 3\) dimensions (App. A.3), using Hogervorst’s dimensional reduction formula [14], which expands d-dimensional blocks in terms of \((d - 1)\)-dimensional ones (App. A.2).
Remark A.2
Following the arguments below carefully, it’s easy to see that \(K_d (a, b) = 1 + O (a)\) for \(a \rightarrow 0\) and b fixed.
Remark A.3
There is another possible way of proving Lemma A.1 but we have not managed to realize it. In any dimension, one can expand the conformal block into collinear blocks:
Here the coefficients \(c_{n, k}\) are positive for a group-theoretical reason: we decompose the \(\text {SO} (2, d)\) representation into irreducible representations of the subgroup \(\text {SL} (2; {\mathbb {R}})\) (which preserves the z-axis), and \(c_{n, k}\) is the sum of squared inner products:
where the sum is over an orthonormal basis of collinear primary states \(\psi \) with collinear twist \(h = \frac{\Delta + \ell }{2} + k\) and scaling dimension \(\Delta + n + k\).
Lemma A.1 would follow if we had an upper bound on \(c_{n, k}\) implying the inequality
Such a bound can be easily derived in 2D, but we do not know how to derive it in \(d \geqslant 3\). There have been papers deriving explicit expressions for the coefficients \(c_{n, k}\) (or coefficients of similar expansions where z is replaced by u and \({\bar{z}}\) by \(1 - v\)) [27, 29, 30, 40, 41], but we did not manage to apply them. We leave realization of this strategy as an open problem.
1.1 \(d = 2\)
Recall that the ratio \((\nu )_{\ell } / (2 \nu )_{\ell }\) is defined for \(\nu = 0\) as in footnote 7. In our normalization (2.26), the 2D conformal blocks are given by
For \(\ell = 0\), the upper and lower bounds are obvious with constants 1 and in the full range of \(z, {\bar{z}}\) (in \(d = 2\) there is no pole \(1 / (\Delta - \nu )\)).
Consider then the case \(\ell \geqslant 1\). Since \(k_{\beta } (z) > 0\) for \(z \in (0, 1)\), \(\beta \geqslant 0\), the lower bound \(g^{(2)}_{\tau , \ell } (z, {\bar{z}}) \geqslant {\frac{1}{2}} k_{\tau } ({\bar{z}}) k_{\tau + 2 \ell } (z)\) follows by dropping the second term in the block.
For the upper bound, let us write
We will use identity (A.2). We have the following lemma about \(G_{\beta } (\rho )\):
Lemma A.4
For any \(0 \leqslant \beta _1 \leqslant \beta _2 < \infty \) we have
Proof
\(G_{\beta _1} (\rho ) \leqslant G_{\beta _2} (\rho )\) follows from the monotonicity of coefficients in the series expansion of \(_2 F_1\). We also have \(G_0 (\rho ) = 1\) and \(G_{\infty } (\rho ) = \frac{1}{\sqrt{1 - \rho ^2}}\). \(\square \)
Using (A.2), the ratio X is rewritten as
using \({\bar{\rho }} \leqslant \rho \), and Lemma A.4 for the other two factors. Hence X is uniformly bounded by C(a). This implies the upper bound with \(C_2 (a, b) = 1 + C (a)\). The \(d = 2\) case is complete.
1.2 Dimensional reduction formula
We will need Hogervorst’s dimensional reduction formula. We will state and use this result using blocks \(g_{\Delta , \ell }\) labeled by the dimension and spin, not by the twist and spin as in the rest of the paper. Hopefully no confusion will arise.
Theorem A.5
(Hogervorst [14], Eq. (2.24)) Conformal blocks in d dimensions can be expanded in a series of conformal blocks in \(d - 1\) dimensions, as
where (\(j = \ell - 2 p\))
Some remarks are due here. First, the existence of an expansion like (A.12) can be inferred very generally from decomposing a representation of the d-dimensional conformal block under the \((d - 1)\)-dimensional conformal group. It also follows that the coefficients \({\mathcal {A}}^{(d)}_{n, j} (\Delta , \ell )\) have to be nonnegative if \((\Delta , \ell )\) satisfies the d-dimensional unitarity bounds. These condition are verified by the above expressions. Moreover, these coefficients are regular (no poles) for \(\ell = 0, \Delta > \nu \) and \(\ell \geqslant 1\), \(\Delta \geqslant \ell + 2 \nu \).
Hogervorst [14] derived recursion relations satisfied by the coefficients \({\mathcal {A}}^{(d)}_{n, j} (\Delta , \ell )\). He then guessed (A.13) as the (unique) solution of those recursion relations. He then could check up to very high recursion order that (A.13) indeed remains compatible with the recursion. The complete proof that (A.13) is a solution of Hogervorst’s recursion relations is so far missing, but there is hardly any doubt that they are correct.
It may perhaps be possible to prove (A.13) by comparing with the pole structure of Zamolodchikov-like recursion relations [42] which have been established rigorously [43].
1.3 \(d \geqslant 3\) lower bound
We will argue by induction using (A.12). The needed lower bound is obtained by keeping the term \(n = 0, j = \ell \) and throwing out the rest (recall that (A.12) is a positive sum). We have
From (A.13), it can be seen that:
Hence by induction we get the lower bound
1.4 \(d \geqslant 3\) upper bound
Here we will prove the upper bound in \(d \geqslant 3\)
Note that the pole \(\Delta = \nu \) can be approached only by the scalar blocks. Thus for \(\ell \geqslant 1\) the bound says that the blocks are uniformly bounded.
As discussed in Sect. A.3, the first term in the reduction formula, \({\mathcal {A}}_{0, \ell }^{(d)} (\Delta , \ell ) g^{(d - 1)}_{\Delta , \ell } (z, {\bar{z}})\), gives upon induction exactly the main term in the bound we want to obtain. Let Y be the sum of all the other terms, i.e.
To show the induction step, we need to show a bound
with C(a, b, d) independent of \(\ell , \Delta \).
Using (A.13) we have
In the third line we bound \(\left( \frac{\Delta + \ell }{2} \right) _n \Big / \left( \frac{\Delta + \ell + 1}{2} \right) _n \) by 1 and we also use
(actually \(= 1\) for \(\ell = 0\), while to see this for \(\ell \geqslant 1\) is easy using the unitarity bound \(\Delta - \ell - 2 \nu \geqslant 0\)).
We next study the second line in (A.21), denote it \(F_n\). We need to pay attention to the pole at \(\Delta = \nu \) and spurious poles at \(\Delta = 1\) and \(\Delta = \nu + 1 / 2\). For this reason we consider \(n = 0, 1\) and \(n \geqslant 2\) separately. We have
Considering separately the cases \(\ell = 0\), \(\ell = 1\) and \(\ell \geqslant 2\) and using the unitarity bounds in each case, it’s easy to see that
with C uniform in \(\Delta , \ell \), j.
Finally for \(n \geqslant 2\) we can write
We can estimate the first factor byFootnote 19
the second one by 1, and the last one by
We conclude that for all n, \(\Delta , \ell , j\) we have
Using the discussed estimates for the second and third line in (A.21), we get an upper bound for the ratios of the \({\mathcal {A}}\) coefficients in (A.20):
For the ratios of \((d - 1)\)-dimensional conformal blocks in (A.20) we use the bounds from the previous induction step (the upper bound in the numerator and the lower bound in the denominator). We thus get
where \(C' = K_{d - 1} (a, b) \left( 1 + \frac{\theta (d - 1 \geqslant 3) }{\Delta + 2 n - \nu _{d - 1}} \right) \) is unformly bounded in \(\Delta , \ell \) since \(\nu _{d - 1} = \nu - 1 / 2\).
For further simplification of (A.31) we use the identity (A.2). In addition to Lemma A.4, we will need another lemma about \(G_{\beta }\):
Lemma A.6
For \(\beta _1 \geqslant \beta _2 > 0\) we have
Proof
We use the integral representation of \({}_2 F_1\) valid for \(\text {Re} (c)> \text {Re} (b) > 0\), \(|\arg (1 - x) |< \pi \):
Taking \(a = \frac{1}{2}, b = \frac{\beta }{2}, c = \frac{\beta + 1}{2}\) and \(x = \rho ^2\) we get
We replace \(t^{\beta _1 / 2}\) under the integral by a larger quantity \(t^{\beta _2 / 2}\), and infer the stated bound. \(\square \)
By (A.2), we rewrite the ratio of k’s in (A.31) as
By Lemma A.4, we have
Plugging these estimates into (A.31), we estimate it as
Combining (A.30) and (A.38), and using the inequality
we get
The point of this bound is that it only depends on \(n, p, \rho , {\bar{\rho }}\) but is uniform in \(\Delta , \ell \). Also the dependence on n, p is factorized.
We now sum (A.40) over p, n to get a bound on the ratio (A.20). Although p in Hogervorst’s formula is bounded by \([\ell / 2]\) we will extend the sum over all p from 0 to \(\infty \). The sum over p is convergent when \({\bar{\rho }} / \rho < 1\) while the sum over n is convergent when \(\rho {\bar{\rho }} < 1\). Since we are assuming that \(0 \leqslant {\bar{z}} \leqslant a< b \leqslant z < 1\), we have
So the sums over n and over p are both bounded by a constant which only depends on a, b and d. We thus obtain the needed bound (A.20). This completes the induction step.
Proof of Lemma 2.14
Here we will prove Lemma 2.14, i.e. the improved logarithmic bound on the conformal blocks:
for all \((\tau , \ell )\) satisfying \(\tau \leqslant \tau _{\max }, \ell \leqslant \ell _{\max }\) and the unitarity bounds, i.e. \(\tau \geqslant \nu \) for \(\ell = 0\) and \(\tau \geqslant 2 \nu \) for \(\ell \geqslant 1\). We will show that the constant factor B in the upper bound is finite and depends only on a, b, \(\tau _{\max }\) and \(\ell _{\max }\).
We will divide the proof into two cases: (a) \(1 \leqslant \ell \leqslant \ell _{\max }\), (b) \(\ell = 0\). We will derive (B.1) with constants \(B_1\) for (a) and \(B_2 \) for (b). In the end we take \(B = \max \{ B_1, B_2 \}\).
1.1 \(1 \leqslant \ell \leqslant \ell _{\max }\)
In this case the range of \(\tau \) we are concerned about is given by
As explained in Remark 2.8(c), Lemma 2.7 implies a two-sided bound for \(2 \nu \leqslant \tau \leqslant \tau _{\max }\)
where \(F_{\tau , \ell }\) is given in (2.28), the constants \(C_1\), \(C_2\) are finite and they only depend on a, b and \(\tau _{\max }\). Using the upper bound for \(g_{\tau , \ell } (z, {\bar{z}})\) and the lower bound for \(g_{\tau , \ell } (b, a)\), we get
here we used the explicit form of \(F_{\tau , \ell }\) in the last step. Then using the lower bounds \(a^{\tau / 2} \geqslant a^{\tau _{\max } / 2}\), \(k_{\tau + 2 \ell } (b) \geqslant b^{\tau _{\max } / 2 + \ell _{\max }}\) for the denominator, and the upper bounds \(z^{\beta / 2} \leqslant 1\), \(k_{\beta } (z) \leqslant {}_2 F_1 (\beta / 2, \beta / 2; \beta ; z) \leqslant {}_2 F_1 (\beta _{\max } / 2, \beta _{\max } / 2; \beta _{\max }; z)\) for the numerator, we get
in the regime where \(0< {\bar{z}} \leqslant a< b \leqslant z < 1\), \(2 \nu \leqslant \tau \leqslant \tau _{\max }\) and \(1 \leqslant \ell \leqslant \ell _{\max }\). Here the coefficient \(C' {:}{=}C_2 / (C_1 a^{\tau _{\max } / 2} b^{\tau _{\max } / 2 + \ell _{\max }})\) is finite and depends only on a, b, \(\tau _{\max }\) and \(\ell _{\max }\). For the hypergeometric function \({}_2 F_1\) we use the fact (mentioned in Remark 2.8(e)) that
This asymptotics together with the monotonicity of \({}_2 F_1 \left( \frac{\beta }{2}, \frac{\beta }{2}; \beta ; z \right) \) imply that
where \(C''<\infty \) only depends on \(0<\beta <\infty \) and \(b>0\). Combining (B.5) and (B.7) we get
in the regime \(0< {\bar{z}} \leqslant a< b \leqslant z < 1\), \(2 \nu \leqslant \tau \leqslant \tau _{\max }\) and \(1 \leqslant \ell \leqslant \ell _{\max }\). Here \(B_1: = C' C''\) is finite and depends only on a, b, \(\tau _{\max }\) and \(\ell _{\max }\). This finishes the proof of \(\ell \geqslant 1\) case.
1.2 \(\ell = 0\)
The \(d = 2\), \(\ell = 0\) case can be shown exactly as \(\ell \geqslant 1\) in the previous section, since in this case the \(\Delta \rightarrow \nu \) pole is absent. In the rest of this section we consider \(d \geqslant 3\).
By upper bound in Lemma 2.7 and \({\bar{\rho }} \leqslant {\bar{z}}\) we get the upper bound of \(g_{\Delta , 0} (z, {\bar{z}})\) for \(\Delta > \nu \):
where \(K_d (a, b)\) is a finite constant independent of \(\Delta \). Then by the similar analysis as the \(\ell \geqslant 1\) case, we get an upper bound on \(g_{\Delta , 0} (z, {\bar{z}})\):
where \(D_1\) is a finite constant, depending only on a, b and \(\tau _{\max }\). The main subtlety here is the \((\Delta - \nu )^{- 1}\) term which blows up when \(\Delta \rightarrow \nu \). To show that the ratio \(g_{\Delta , 0} (z, {\bar{z}}) / g_{\Delta , 0} (b, a)\) has the \({\bar{z}}^{\tau / 2} \log \left( \frac{1}{1 - {\bar{z}}} \right) \) bound with the coefficient uniform in \(\Delta \), we need a lower bound of \(g_{\Delta , 0} (b, a)\) which also blows up as \((\Delta - \nu )^{- 1}\) when \(\Delta \rightarrow \nu \). The lower bound in Lemma 2.7 does not show this behavior. So we need an improved lower bound of \(g_{\Delta , 0} (b, a)\) which features the \((\Delta - \nu )^{- 1}\) singularity.
For this we use the dimensional reduction formula (A.12), restricting to \(\ell = 0\):
where the reduction coefficients are given by
When \(\Delta > \nu \), (B.11) is a positive sum, so \(g_{\Delta , 0}^{(d)}\) is bounded from below by the \(n = 1\) term:
This estimate gives us the needed lower bound of \(g_{\Delta , 0}^{(d)} (b, a)\) since it has the \((\Delta - \nu )^{- 1}\) factor. The whole prefactor in front of \(g^{(d - 1)}_{\Delta + 2, 0} (z, {\bar{z}})\) is bounded from below by
where \(D_2 {:}{=}\frac{\nu ^3}{32 \left( \tau _{\max } - \nu + \frac{1}{2} \right) (\tau _{\max } + 1)} > 0\). Now take \(z = b\) and \({\bar{z}} = a\), the estimate of \(g^{(d - 1)}_{\Delta + 2, 0} (b, a)\) is similar to the previous section:
Here in the first line we used the lower bound of Lemma 2.7 and the fact that \({\bar{\rho }}^{(\Delta + 2) / 2} \geqslant ({\bar{z}} / 4)^{(\Delta + 2) / 2} \geqslant 2^{- \tau _{\max } - 2} {\bar{z}}^{(\Delta + 2) / 2}\) for \(\Delta \leqslant \tau _{\max }\), in the second line we used the explicit form of \(F_{\Delta + 2, 0} (b)\), in the third line we bounded \(a^{(\Delta + 2) / 2}\), \(4^{(\Delta + 2) / 2}\), \(b^{\Delta + 2}\) and \({}_2 F_1\) by their minimal values in the range \(\nu \leqslant \Delta \leqslant \tau _{\max }\). \(D_3>0\) because \(a, b \in (0, 1)\) and \(\tau _{\max } < \infty \).
Putting these together, we have for \(0 \leqslant {\bar{z}} \leqslant a< b \leqslant z < 1\) and \(\nu < \Delta \leqslant \tau _{\max }\):
where \(B_2 {:}{=}\frac{D_1}{D_2 D_3} (1 + \tau _{\max } - \nu )\) is finite and depends only on a, b and \(\tau _{\max }\).
Rights and permissions
Springer Nature or its licensor (e.g. a society or other partner) holds exclusive rights to this article under a publishing agreement with the author(s) or other rightsholder(s); author self-archiving of the accepted manuscript version of this article is solely governed by the terms of such publishing agreement and applicable law.
About this article
Cite this article
Pal, S., Qiao, J. & Rychkov, S. Twist Accumulation in Conformal Field Theory: A Rigorous Approach to the Lightcone Bootstrap. Commun. Math. Phys. 402, 2169–2214 (2023). https://doi.org/10.1007/s00220-023-04767-w
Received:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s00220-023-04767-w