SciPost Submission Page
A plane defect in the 3d O(N) model
by Abijith Krishnan, Maxim A. Metlitski
This Submission thread is now published as
Submission summary
Authors (as registered SciPost users):  Abijith Krishnan 
Submission information  

Preprint Link:  scipost_202304_00015v2 (pdf) 
Date accepted:  20230718 
Date submitted:  20230701 01:41 
Submitted by:  Krishnan, Abijith 
Submitted to:  SciPost Physics 
Ontological classification  

Academic field:  Physics 
Specialties: 

Approach:  Theoretical 
Abstract
{It was recently found that the classical 3d O$(N)$ model in the semiinfinite geometry can exhibit} an ``extraordinarylog" boundary universality class, where the spinspin correlation function on the boundary falls off as $\langle \vec{S}(x) \cdot \vec{S}(0)\rangle \sim \frac{1}{(\log x)^q}$. This universality class exists for a range $2 \leq N < N_c$ {and MonteCarlo simulations and conformal bootstrap} indicate $N_c > 3$. In this work, we extend this %analysis {result} to the 3d O$(N)$ model in an infinite geometry with a plane defect. We use renormalization group (RG) to show that in this case the extraordinarylog universality class is present for any finite $N \ge 2$. We additionally show, {in agreement with our RG analysis}, that the line of defect fixed points which is present at $N = \infty$ is lifted to the ordinary, special (no defect), and extraordinarylog universality classes by $1/N$ corrections. %in agreement with our RG analysis. {We study the ``central charge" $a$ for the $O(N)$ model in the boundary and interface geometries and provide a nontrivial detailed check of an $a$theorem by Jensen and O'Bannon.} {Finally, we} revisit the problem of the O$(N)$ model in the semiinfinite geometry. We find evidence that at $N = N_c$ the extraordinary and special fixed points annihilate and only the ordinary fixed point is left for $N > N_c$.
Published as SciPost Phys. 15, 090 (2023)
List of changes
We would like to thank the Referee for the careful reading of our manuscript and for their comments.
1  In page 4 below eq 1.3, it is said that "believed that the scaling dimension ∆ϵ < 2 for all finite N ≥ 1". It might be worth pointing out the existence of rigorous bootstrap bounds which show this for many values of N, including moderately large N.
Upon the Referee’s suggestion, below Eq. (1.3) we have added a citation to Ref [12] (a bootstrap study identifying the O(N) islands) and Ref [13] (a review of the largeN results).
2  In page 16 it is said that "the interface ordinary fixed point is equivalent to two decoupled boundary ordinary fixed points for each side of the interface". Reminding the reader of eq. 1.4 could be helpful.
We have added a reference to Eq. (1.4), as suggested by the Referee.
3  It is used in multiple occasions that the special interface transition is essentially equivalent to no defect for all N. On the other hand, the boundary special transition has nontrivial anomalous dimensions with the respect to the bulk operators. Is there a simple way to argue for the triviality of the interface with an action similar to 1.4 or 2.2 for the special boundary class?
We can start with two decoupled boundaries each at its special fixed point (assuming N < N_c ~ 5). However, the dimension of the boundary O(N) vector at the boundary special fixed point is quite small (please see table I in Ref. [1]). Even for N=1, this dimension is ~0.36 and the dimension decreases with N. Thus, the analog of the coupling u in Eq. (1.4) is highly relevant. We cannot track the RG flow induced by u, however, the natural guess is that it is towards the trivial interface, which, importantly, has only a single relevant O(N) singlet perturbation.
4  The folding construction is used a few times in the text. Could one have anticipated eq. 5.37 from 3.34 + the folding argument?
We were very intrigued by the apparent doubling of the shift of the g^3 coefficient in the betafunction from the single boundary to the interface. One way to uncover the origin of this doubling would be to pursue the strategy explained on page 24:
``From the form of the action (5.1) a direct computation of the coefficient b in beta(g) requires the knowledge of the fourpoint function of the tilt operator t_i at the normal fixed point. (This should be compared to the computation of the coefficient alpha_{bound}, which relies only on the twopoint function of t_i and the knowledge of the coefficient s. In addition, a number of higher order counterterms in the action, omitted in
Eq. (5.1), such as e.g. delta L_{bound} ~ pi^2 pi_i t_i, would have to be fixed by the requirement of O(N) invariance. We do not pursue this route to computing b here.”
Instead of pursuing this plan, we chose to extract the g^3 coefficient in the betafunction from the boundary order parameter dimension at the special transition in d=3+epsilon. However, it would be interesting to pursue the calculation sketched above in future work.
5  Reference https://inspirehep.net/literature/1837673 computes conformal data for the O(N) conformal interface in the epsilon expansion (for the ordinary and special transitions ). Maybe these results could be mentioned. Perhaps some of the them are useful to compare with?
We thank the Referee for pointing out this paper. Since we do not use the OPE coefficients at the ordinary/special transition anywhere in our paper, we don’t see a natural place to cite this reference (we also don’t cite older papers on the epsilon expansion for the ordinary/special transition, since we don’t use the corresponding results.)
Submission & Refereeing History
You are currently on this page