SciPost Submission Page
Positivity, low twist dominance and CSDR for CFTs
by Agnese Bissi, Aninda Sinha
This is not the latest submitted version.
This Submission thread is now published as
Submission summary
Authors (as registered SciPost users):  Aninda Sinha 
Submission information  

Preprint Link:  https://arxiv.org/abs/2209.03978v2 (pdf) 
Date submitted:  20221014 03:51 
Submitted by:  Sinha, Aninda 
Submitted to:  SciPost Physics 
Ontological classification  

Academic field:  Physics 
Specialties: 

Approach:  Theoretical 
Abstract
We consider a crossing symmetric dispersion relation (CSDR) for CFT four point correlation with identical scalar operators, which is manifestly symmetric under the crossratios $u,v$ interchange. This representation has several features in common with the CSDR for quantum field theories. It enables a study of the expansion of the correlation function around $u=v=1/4$, which is used in the numerical conformal bootstrap program. We elucidate several remarkable features of the dispersive representation using the four point correlation function of $\Phi_{1,2}$ operators in 2d minimal models as a testbed. When the dimension of the external scalar operator ($\Delta_\sigma$) is less than $\frac{1}{2}$, the CSDR gets contribution from only a single tower of global primary operators with the second tower being projected out. We find that there is a notion of low twist dominance (LTD) which, as a function of $\Delta_\sigma$, is maximized near the 2d Ising model as well as the nonunitary YangLee model. The CSDR and LTD further explain positivity of the Taylor expansion coefficients of the correlation function around the crossing symmetric point and lead to universal predictions for specific ratios of these coefficients. These results carry over to the epsilon expansion in $4\epsilon$ dimensions. We also conduct a preliminary investigation of geometric function theory ideas, namely the BieberbachRogosinski bounds.
Current status:
Reports on this Submission
Anonymous Report 2 on 20221123 (Invited Report)
 Cite as: Anonymous, Report on arXiv:2209.03978v2, delivered 20221123, doi: 10.21468/SciPost.Report.6188
Report
The paper focuses on the description of certain properties of fourpoint functions in 2D minimal models. In particular, it considers the Taylor coefficients around the crossingsymmetric point in crossratio space. Starting from a positionspace dispersion relation, the paper contains partial arguments that these coefficients should satisfy certain universal properties, such as signdefiniteness.
Perhaps the main interesting insight of the paper is that when the Taylor coefficients are computed using the schannel OPE data, convergence is better when one uses the OPE inside the dispersion relation rather than directly. In fact, the coefficients are wellapproximated by the contribution of a single twist tower.
However, little effort is spent on making this more precise. For example, suppose we neglect all operators of twist above some $\tau$. What is an upper bound on the error? It is stated that lowtwist dominance together with the dispersion relation explains the universal features. However, this logic is somewhat circular since there is no independent proof of lowspin dominance. It would be more correct to say that the universality (observed empirically in the paper) is evidence for low spindominance.
I am also not sure if the quantities that the paper analyses, namely the Taylor coefficients at $u=v=1/4$, are of much independent interest. As stated in the article, the numerical conformal bootstrap does utilize such expansion, but the coefficients appearing in particular correlators (such as in minimal models) are not needed in that context.
While the paper does contain some interesting insights and computations, it does not contain groundbreaking results and it is unclear if the method/results generalize beyond the considered examples (minimal models and $\epsilon$ expansion). For these reasons, I do not recommend it for publication in SciPost.
Anonymous Report 1 on 20221118 (Invited Report)
 Cite as: Anonymous, Report on arXiv:2209.03978v2, delivered 20221118, doi: 10.21468/SciPost.Report.6153
Report
In this paper the authors define in the framework of 2d diagonal unitary minimal models
a function $F(u,v)$ which is symmetric under the exchange of the two variables
as a consequence of the crossing symmetry of the fourpoint function characterized by
the crossratios u and v. They note that the coefficients of the Taylor expansion
\[F(u,v)=\sum_{p,q} c_{p,q} x^p y^q\] about the symmetric point $ u=v=1/4$ with
$ x=u+v1/2,
y=(u1/4)(v1/4)$, have some surprising and unexpected properties, in particular the sign of $c_{p,q}$ is uniquely fixed by the parity of $ p+q$. These properties do not depend on unitarity, since they are also true, as the authors point out, in the 2d Yanglee model. The authors show that these properties can be understood using a suitable dispersion relation of the
mentioned symmetric function. An approximate expression of these Taylor coefficients has also
been obtained.
The results are interesting and most probably could be extended to a much larger class
of conformal models, even with D>2, I have however some (minor) remarks:
1) F(u,v) is defined in (2.4) for the case of Isng model, but the same
notation is used for more general cases. For thesake of clarity it is necessary to
split equation (2.4) into two parts, the general definition and the specific form for
the Ising case.
2) The interested reader could wish to know how the fourpoint functions have been
obtained, so it is necessary to add a suitable reference for the Ising model, the
minimal models reported in in eq. (4.1) and in Appendix B (mentioning the paper of
Cardy where the YangLee fourpoint function was first obtained).
In conclusion, the submitted paper is interesting and also well written, so it
deserves publication once the above minimal corrections are made.
Requested changes
1) Split equation (2.4)
2) Add references as specified in the report
Author: Aninda Sinha on 20221124 [id 3063]
(in reply to Report 1 on 20221118)Thank you for the suggestions. We will incorporate them.
Author: Aninda Sinha on 20221124 [id 3064]
(in reply to Report 2 on 20221123)Anonymous on 20221125 [id 3078]
(in reply to Aninda Sinha on 20221124 [id 3064])Let me clarify my suggestion. It has been shown (in https://arxiv.org/pdf/1208.6449.pdf) that when a positionspace correlator is approximated by the OPE, keeping all operators up to dimension $\Delta$, the error is exponentially small in $\Delta$. My suggestion was to perform a similar analysis for the situation when the OPE is used inside the dispersion relation. I.e. what can we say about the error in general when only operators of twist up to $\tau$ are kept. In the present form, the paper observes empirically, in a particular example where the exact correlator is explicitly known, that when only the leading twist tower is kept, the error is small. However, no attempt is made at analyzing this phenomenon in general. For this reason, the claims of the paper come short of a proof. For example, in arguing for signdefiniteness of $\phi_{pq}$ after eq. (6.4), the authors say "Higher twist operators will be subleading in $\epsilon$ but to conclude that they are truly subleading we will have to assume that their OPE coefficients do not overwhelm the smallness of the factor." Have the authors considered using results on OPE convergence to justify this assumption?
I disagree that the paper contains an argument whose logical structure is "lowtwist dominance" $\Rightarrow$ "a nontrivial conclusion". In fact, it is not stated what the precise meaning of lowtwist dominance is. Indeed, the term is not standard in the literature and as far as I can see, it is used by the authors in the sense "we get the right answer if we only keep operators of low twist". However, this is precisely what they claim the conclusion is. Indeed, the logical content of the paper is that it gives empirical evidence that lowtwist operators dominate the contributions to $\phi_{pq}$. So lowtwist dominance is a conclusion, not an assumption.
Regarding application to other theories, I am not sure what the authors are expecting to learn. The analysis of the present paper was only possible because the fourpoint function of minimal models is known explicitly. What then, are we going to learn about more interesting theories where the exact fourpoint function is not known? This is related to my point 1 above. If it were proved on general grounds that only lowtwist operators contribute significantly, one could apply this to interesting theories where the exact correlator is not known. On the other hand, if we only focus on theories where the exact correlator is known, we can simply expand it around $u=v=1/4$ and forget about the dispersion relation. So it is unclear what the benefit of the presented method is in general.