{"id":"daaf6bc2-0ffc-4c3d-a424-d871dbb23e65","arxiv_id":"2411.09792","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":3,"one_line_summary":"For 2D CFTs with boundary-only symmetry breaking, the entanglement asymmetry is log|G| with power-law corrections for finite groups, (dim G/2) log log(ℓ/ε) for compact Lie groups, and it drops from log|G| to 0 at t=ℓ/2 after a global quench.","lead":"This paper computes how much a global symmetry is broken inside a subsystem when the symmetry is broken only at a boundary, using 2D boundary conformal field theory. It predicts a new logarithmic-of-logarithmic growth for continuous symmetries and a sharp symmetry restoration after a global quench.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Compact-Lie claim overreaches: Eq (2.21) assumes the unbroken subgroup H is finite, but Eq (1.8) is stated for all compact Lie G; for continuous H the saddle manifold has dim(H)(n−1) zero modes and the leading coefficient should be (dimG−dimH)/2, not dimG/2.","rationale":"I read the paper in good faith. The finite-group equilibrium derivation is a clean extension of BCFT technology and the quench section reproduces the known entanglement-entropy benchmark before proposing the asymmetry result, which gives genuine support to the method. The central weakness is the compact-Lie claim. The reader correctly flagged that Eq (2.21) assumes a finite unbroken subgroup while Eq (1.8) is stated for all compact Lie groups. I agree, and I sharpen the concern: the finite-H assumption is not merely an unproven input but changes the leading coefficient when H is continuous. If a boundary preserves a continuous subgroup, the saddle manifold has positive dimension, the Hessian in Eq (2.22) is degenerate along those directions, and the Gaussian integral produces a prefactor with exponent (dimG − dimH)(n−1)/2 rather than dimG(n−1)/2. The free-boson example in Section 2.5 sidesteps this by taking G to be the already-broken U(1) copy rather than the full U(1)×U(1) symmetry, so it does not test the continuous-H case. The same limitation affects the O(1) term in Eq (2.26), where log|H| is undefined for a continuous H. I do not see a comparable flaw in the finite-group result: the power-law correction and analytic continuation are standard, and the S3→Z2 example correctly shows that the leading constant becomes log(|G|/|H|). The quench factorization is heuristic, as the authors state, but it is benchmarked by Eq (3.13) and the absence of vacuum overlap in Eq (3.21) is a standard BCFT fact. The paper should either restrict the compact-Lie claim to finite unbroken subgroups or generalize the derivation to continuous H, with the expected replacement dimG → dimG − dimH in the leading log-log coefficient. Conditional acceptance remains the right outcome, with this caveat made explicit.","tokens_in":25947,"tokens_out":18561,"duration_ms":192477,"concrete_test":"Use the vertex-operator method of Section 2.5 on a free compact boson with full symmetry G = U(1)_L × U(1)_R and a boundary preserving the diagonal U(1), so H = U(1) is continuous. Compute the entanglement asymmetry from Eq (2.5) exactly or via the saddle integral on the coset G/H ≅ U(1). If the leading term is 1/2 log log(ℓ/ϵ), equal to (dimG − dimH)/2, Eq (1.8) overcounts by a factor of 2 for this geometry; if it is log log(ℓ/ϵ), the finite-H restriction in Eq (2.21) is harmless. The analogous check for SU(2) WZW with a U(1)-preserving Cardy boundary condition distinguishes (dimG−dimH)/2 = 1 from dimG/2 = 3/2.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The most load-bearing problem is in the compact-Lie derivation of Section 2.3. Equation (2.20) is an ansatz for Zn({gi})/Zn, and Eq (2.21) restricts the unbroken subgroup H to be finite. The subsequent saddle-point treatment counts |H|^{n-1} isolated saddles and yields Eq (2.26), whose leading term is dim(G)/2 log log(ℓ/ϵ). But Eq (1.8) and the abstract state the result for every compact Lie group with no such restriction. Whenever a conformal boundary preserves a continuous subgroup H — for example SU(2)→U(1), or U(1)×U(1) broken to the diagonal U(1) — the set {gi∈H, ∏gi=e} is a continuous saddle manifold of dimension dim(H)(n−1), and the Hessian in Eq (2.22) has zero modes along it. The Gaussian integration must then be performed over the coset directions only. Repeating the steps of Eqs (2.23)–(2.26) gives a leading term (dimG − dimH)/2 log log(ℓ/ϵ), not dimG/2 log log(ℓ/ϵ). This changes the coefficient of the main sublogarithmic growth, not merely an O(1) constant. Equation (2.21) is therefore not a technical convenience: as stated, Eq (1.8) is false for boundary conditions with a continuous residual symmetry. The finite-group result and the quench benchmark (Eq (3.13)) are not affected by this concern.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The manuscript studies the entanglement asymmetry ΔS_A in 2D boundary conformal field theory, in the setting where a global symmetry G is preserved in the bulk but broken by a conformal boundary condition. For finite groups the authors derive ΔS_A = log|G| − (ε/ℓ)^{2Δ_*} W(Δ_*) + o(ℓ^{−2Δ_*}) by expressing the replicated partition function in terms of boundary-changing operator two-point functions. For compact Lie groups they derive ΔS_A^{(n)} = (dim G/2) log log(ℓ/ε) + O(1) using a saddle-point integration over the symmetry group, and for a global quantum quench they predict ΔS_A(t) ≈ log|G| for 0<t<ℓ/2 and 0 for t>ℓ/2, using a chiral/anti-chiral splitting of the replica symmetry defect. The paper includes worked examples in the three-state Potts model and the compact boson, and reproduces the known Calabrese–Cardy entanglement entropy result as a benchmark of the new defect method.","tokens_in":26309,"tokens_out":11219,"duration_ms":106986,"significance":"If the results are correct, the paper provides universal leading and subleading formulas for entanglement asymmetry under boundary-only symmetry breaking in 2D CFT, including a qualitatively new log log(ℓ) growth for compact Lie groups that contrasts with the log ℓ growth found for bulk symmetry breaking in Ref. [37]. The finite-group derivation is transparent, parameter-free in the relevant CFT data, and gives an explicit, falsifiable prediction involving the boundary-changing operator dimension Δ_*. The dynamical prediction of a sharp symmetry restoration at t=ℓ/2 is a crisp statement that can be tested in lattice models. The use of BCFT boundary-changing operators and the benchmark reproduction of Eq. (3.13) are genuine strengths. The main weakness is that the compact-Lie statement is broader than the derivation, which assumes a finite unbroken subgroup.","major_comments":[{"comment":"The derivation in §2.3 restricts the unbroken subgroup H to be finite (Eq. (2.21)), but Eq. (1.8) and the abstract state the result for every compact Lie group G. When H is continuous, the set {g_i ∈ H : ∏ g_i = e} is a continuous saddle manifold of dimension dim(H)(n−1), and the Hessian in Eq. (2.22) has zero modes along it. The Gaussian integration must be performed over the coset directions only, and the leading term becomes (dim G − dim H)/2 log log(ℓ/ε), not dim G/2 log log(ℓ/ε). Thus Eq. (1.8) is incorrect as stated for common boundary conditions such as SU(2)→U(1) or U(1)×U(1)→diagonal U(1). The theorem should either be restricted to finite residual symmetry, or the continuous-H case should be treated by collective-coordinate integration with the resulting coefficient reported. The U(1) example in §2.5 has H trivial and therefore does not test this issue.","section":"§2.3, Eqs. (2.21) and (1.8)"},{"comment":"The compact-Lie calculation relies on the power-law ansatz Z_n({g_i})/Z_n = (ℓ/ε)^{−β_n({g_i})} without derivation. Since β_n is assumed to be independent of ℓ and cutoff-independent, this is a substantive assumption about the form of the multi-defect correlation function, not a trivial consequence of scale invariance when several boundary-changing operators are involved. The paper should either prove or explicitly state this as a hypothesis, and clarify the regime of validity. This issue is separate from the finite-H restriction, but it is equally load-bearing for Eq. (1.8).","section":"§2.3, Eq. (2.20)"},{"comment":"The quench result (3.24) rests on two unproven approximations: the factorization of the full path integral into independent cylinder amplitudes in the β→0 limit (Eq. (3.4)), and the absence of vacuum overlap between open-string Hilbert spaces with different boundary conditions (Eq. (3.21)). These are called reasonable in the text, but they are not derived. The benchmark reproduction of the entanglement entropy (3.13) supports the method, but it does not by itself validate the replacement of the symmetry defect inside the Z_n amplitude that leads to (3.22)–(3.24). The authors should state these as explicit assumptions and discuss the finite-β corrections, or provide a derivation of the factorization and of Eq. (3.21) within BCFT.","section":"§3.1.1–3.1.2, Eqs. (3.4) and (3.21)"}],"minor_comments":[{"comment":"Eq. (2.12) as written is not an identity: the left-hand side contains the weights 2(n−k), while the right-hand side does not. The counting leading to the factor n in Eq. (2.13) should be clarified, since the natural counting of configurations with two boundary-changing operators for a given pair gives n(n−1).","section":"§2.2, Eq. (2.12)"},{"comment":"The analytic continuation in n is imported from Ref. [47] for Δ_*<1/4 and then extended to all larger Δ_* by analyticity of s′(1). This extension should be stated more carefully, since the original derivation in Ref. [47] does not cover all values used here.","section":"§2.2, text after Eq. (2.19)"},{"comment":"The ellipses in Eqs. (2.46)–(2.47) and (2.49) are used without specifying the order of neglected terms. The authors should state whether the omitted terms are O(1/log(ℓ)) or of some other definite order.","section":"§2.5, Eqs. (2.46)–(2.49)"},{"comment":"The chiral/anti-chiral split of the replica Z_n symmetry defect is an important new construct, but its justification is only sketched in the footnote. A short discussion of its physical interpretation and of the non-locality caveat mentioned there would be helpful for readers who want to apply the method.","section":"§3.1, footnote 2"},{"comment":"The symbol L appears as an IR cutoff parameter but is not defined in the main text. Please define it explicitly when it is first used.","section":"§3.1.1, Eq. (3.5)"}],"recommendation":"major_revision","confidential_remarks":"The manuscript overlaps with the concurrent work of Ref. [62], which the authors thank for sharing results; the editor may want to ensure that the novelty relative to that work is clearly documented. The main technical issue is the overreach of the compact-Lie claim: Eq. (1.8) is stated for all compact Lie groups, but the derivation assumes a finite unbroken subgroup, and the continuous-H case changes the leading coefficient. This is fixable by restricting the theorem or by treating the zero modes properly with a collective-coordinate calculation."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Punchline: the finite-group part is a careful, genuine advance, but the compact-Lie claim is stated more broadly than the derivation supports.\n\nWhat is actually new: boundary-only symmetry breaking gives ΔS_A = log|G| − (ε/ℓ)^{2Δ_*} W(Δ_*) for finite G, with the power law fixed by the lowest-dimension boundary-changing operator. The log log ℓ growth for compact Lie groups is also new, if you restrict to the case where the unbroken subgroup is finite. The quench result for finite groups, ΔS_A(t) ≃ log|G| for t < ℓ/2 and 0 for t > ℓ/2, is a clean statement. The method-of-images split of the symmetry defect into chiral/anti-chiral parts is a useful trick that will be reused.\n\nWhat the paper does well: the finite-group derivation is built on standard BCFT technology, the analytic continuation from [47] is appropriate, and the Potts model example is concrete. The quench section first reproduces the known entanglement entropy, Eq (3.13), before going to the asymmetry — that is a good benchmark. The authors also flag the approximations in the quench section, which is honest.\n\nWhere the soft spots are: Eq (1.8) is the main problem. The derivation in Section 2.3 assumes the unbroken subgroup H is finite (Eq (2.21)), but the abstract and Eq (1.8) make no such restriction. If H is continuous — say, SU(2) broken to U(1) — the saddles {h_i ∈ H, ∏ h_i = e} form a continuous manifold, the Hessian has zero modes along H, and the Gaussian integral gives (dim G − dim H)/2 log log(ℓ/ε), not dim G/2. This changes the leading coefficient, not just an O(1) term. The U(1) example in Section 2.5 only computes the broken copy of the symmetry (effectively H = {e}), so it does not test the continuous-H case. The quench calculation also rests on heuristic steps — the β→0 factorization into independent cylinders and the no-overlap assumption in Eq (3.21) — which are reasonable but not proven. That is a moderate weakness, not a fatal one.\n\nWho it is for: people working on entanglement asymmetry, BCFT, symmetry-resolved entanglement, and quantum quenches. It deserves a serious referee. My recommendation: send it to peer review, but require the authors to fix the compact-Lie statement — either restrict Eq (1.8) to finite unbroken H, or derive the general coefficient dim(G/H)/2.","headline":"Solid finite-group result and useful BCFT technology, but the compact-Lie formula overstates its generality: the derivation needs finite unbroken H, and the coefficient becomes (dim G − dim H)/2 for continuous H.","tokens_in":26871,"tokens_out":8095,"would_cite":true,"duration_ms":75755,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["81T40","81P40"],"pacs":["11.25.Hf","03.67.Mn"],"model":"deepseek-v4-flash","headline":"Boundary-only symmetry breaking in two-dimensional conformal field theory makes entanglement asymmetry approach log|G| for finite groups and grow as log log ℓ for compact Lie groups.","keywords":["entanglement asymmetry","boundary conformal field theory","symmetry defects","boundary condition changing operators","global quantum quench","Renyi entanglement asymmetry","quantum Mpemba effect","compact Lie group symmetry"],"falsifier":"Numerically compute the entanglement asymmetry in a lattice model with a $U(1)$ symmetry broken only at the boundary (for example a critical free-boson chain with Dirichlet boundary) and check whether $\\Delta S_A^{(n)} - \\frac{\\dim G}{2}\\log\\log(\\ell/\\epsilon)$ approaches the $O(1)$ constant in Eq. (2.26) as $\\ell$ grows; any residual $\\ell$ dependence at fixed $n$ would falsify the power-law ansatz (2.20). In the Potts model, the same check is the prediction that the correction to $\\log 3$ decays as $\\ell^{-2/15}$ for the $|B+C\\rangle$ boundary and as $\\ell^{-4/3}$ for $|A\\rangle$.","tokens_in":25741,"feed_emoji":"⚛️","tokens_out":11969,"duration_ms":97912,"temperature":0.7,"pith_summary":"The paper asks how a global symmetry $G$ that is preserved everywhere except at a physical boundary shows up in the entanglement asymmetry of an interval attached to that boundary. For a finite group $G$, the authors establish that the asymmetry approaches $\\log |G|$ from below, with a universal power-law correction $(\\epsilon/\\ell)^{2\\Delta_*} W(\\Delta_*)$ controlled by the lowest-dimension boundary-changing operator. For a compact Lie group $G$, they establish a leading log-log growth, $\\Delta S_A^{(n)} = \\frac{\\dim G}{2}\\log\\log(\\ell/\\epsilon)+O(1)$, with the $O(1)$ terms determined explicitly—a signature that the symmetry is broken only at the boundary rather than in the bulk. They also study a global quench from a symmetry-broken boundary state and find that the asymmetry stays at $\\log |G|$ until $t=\\ell/2$ and then vanishes, meaning the symmetry is restored exactly when the subsystem becomes thermal. These results matter because they turn boundary symmetry breaking into a measurable, universal entanglement signature and extend the method of images to non-local symmetry defects.","feed_headline":"Boundary-only symmetry breaking yields log-log entanglement asymmetry","feed_subtitle":"Finite groups saturate at log|G| with a power-law correction; Lie groups grow as log log ℓ; a global quench restores symmetry at t=ℓ/2.","key_machinery":"The central object is the topological symmetry defect line $U(g)$: topological in the bulk but anchored at the boundary in a way that is not topological when the boundary breaks $G$. Deforming these lines lets each replica contribution $Z_n(\\{g_i\\})$ reduce to a correlation function of boundary-condition-changing operators on the disk-with-hole obtained by the conformal map $w=((z+\\ell)/(\\ell-z))^{1/n}$. For finite groups, the asymmetry is dominated by configurations with two such operators, whose two-point function $(\\epsilon/n\\ell)^{2\\Delta_b}[\\sin(\\pi k/n)]^{-2\\Delta_b}$ supplies the power-law correction; for compact Lie groups the same defect picture feeds a saddle-point Gaussian integral over $G$ whose Hessian produces the log-log growth. For the quench, the new machinery is splitting the $\\mathbb{Z}_n$ symmetry defect into chiral and anti-chiral parts, continuing to Lorentzian time, and evaluating the resulting cylinder partition functions in the $\\beta\\to0$ limit; the same method reproduces the known entanglement entropy growth of a global quench as a check.","core_discovery":"The central claim is that in a two-dimensional boundary conformal field theory where a non-anomalous global symmetry $G$ is broken only by the boundary condition, the Rényi entanglement asymmetry of an interval of length $\\ell$ attached to the boundary obeys explicit universal formulas. For finite $G$, Eq. (1.7) gives $\\Delta S_A = \\log |G| - (\\epsilon/\\ell)^{2\\Delta_*} W(\\Delta_*) + o(\\ell^{-2\\Delta_*})$, where $\\Delta_*$ is the smallest scaling dimension among boundary-changing operators generated by the broken group elements and $W(\\Delta_*)$ is read from Eq. (2.19). For compact Lie $G$, Eq. (1.8) with the complete $O(1)$ terms in Eq. (2.26) gives leading behavior $\\frac{\\dim G}{2}\\log\\log(\\ell/\\epsilon)$, in contrast to the $\\log \\ell$ growth found when symmetry is broken in the bulk. After a global quench, Eq. (3.24) gives $\\Delta S_A(t) \\simeq \\log |G|$ for $0<t<\\ell/2$ and $\\simeq 0$ for $t>\\ell/2$ in the $\\beta\\to0$ limit. The paper argues that these behaviors are generic within BCFT and are decided only by the group, the boundary condition, and the lowest boundary operator that changes the boundary condition.","pith_inferences":["If the formulas are robust beyond the continuum limit, the exponent $2\\Delta_*$ and the coefficient $N_*/|G|$ in Eq. (2.19) could serve as a boundary order parameter: scanning boundary conditions for one CFT would map exactly which subgroup of $G$ is broken.","The sharp jump at $t=\\ell/2$ is a $\\beta\\to0$ artifact; a finite-$\\beta$ computation would smooth the step, and that crossover is where a quantum Mpemba effect specific to boundary symmetry breaking, if present, would show up.","The same chiral/anti-chiral defect-splitting technique could be applied to other replica-based observables such as reflected entropy or charged moments, and to symmetry-breaking interfaces instead of boundaries.","The continuous-group quench is not worked out; a natural check is whether the asymmetry falls continuously from its initial value or shows the same sharp transition at $t=\\ell/2$."],"forward_implications":["For finite $G$, the leading term $\\log|G|$ is universal and the first subleading term is a negative power law with exponent $2\\Delta_*$, so the approach to the symmetric value carries information about the boundary operator content.","For compact Lie groups, boundary-only breaking is distinguished from bulk breaking by the $\\log\\log(\\ell/\\epsilon)$ scaling; the universal $O(1)$ term depends on $\\mathrm{Vol}(G)$, the Haar measure near the identity, and the unbroken finite subgroup $H$.","After a global quench, the asymmetry stays at $\\log|G|$ for $t<\\ell/2$ and drops to zero for $t>\\ell/2$ in the $\\beta\\to0$ limit, so the symmetry is restored at the time when the entanglement entropy becomes extensive and thermal.","For a semi-infinite interval the same result gives $\\Delta S_A^{(n)}=\\log|G|$ for all times, so the symmetry is never restored in that geometry because the system does not locally thermalize.","The chiral/anti-chiral defect-splitting method reproduces the known quench entanglement entropy as a benchmark and is then used for the asymmetry; the paper states it should apply to settings beyond entanglement asymmetry."],"supporting_citations":[{"why":"Introduces the entanglement asymmetry as the measure of symmetry breaking that the paper computes.","marker":"[1]"},{"why":"Supplies the conformal-frame and twist-operator geometry for an interval attached to a boundary.","marker":"[2]"},{"why":"Computes the asymmetry for bulk symmetry breaking in CFT; its saddle-point Hessian method is adapted here and its log-linear result is the contrast for the log-log law.","marker":"[37]"},{"why":"Establishes the universal leading log|G| asymmetry for matrix product states, matching the finite-group leading term.","marker":"[36]"},{"why":"Provides the analytic continuation in the Rényi index and the value of s'(1) used to reach the replica limit in Eq. (2.19).","marker":"[47]"},{"why":"Gives the global-quench entanglement entropy in 1+1 CFT that the new image method reproduces as a benchmark.","marker":"[52]"},{"why":"Fixes the Potts boundary states and boundary-condition-changing operator dimensions used in the Z3 examples.","marker":"[50]"},{"why":"Identifies the Potts fixed boundary conditions with boundary states, setting the Δ* values in the examples.","marker":"[49]"},{"why":"Computes the non-equilibrium asymmetry for discrete groups in spin chains, providing the consistency check for the quench result and the bound ΔS_A ≤ log|G|.","marker":"[18]"}],"fun_headline_variants":["Boundary-only symmetry breaking yields log-log entanglement asymmetry","Boundary-induced breaking in CFT yields log-log asymmetry","Finite groups saturate, Lie groups log-log in boundary CFT","Symmetry breaking by boundary: log-log entanglement asymmetry","Quench restores symmetry at t=ell/2 in BCFT"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The compact-Lie and quench results rest on two assumptions the paper states but does not prove: that the relevant replica quantity is an exact power law in the subsystem size with no further size dependence, and that in the fast-quench limit the path integral separates into independent cylinders so that different boundary conditions have no vacuum overlap.","fun_headline_variants_meta":{"raw":{"variants":["Boundary-only symmetry breaking yields log-log entanglement asymmetry","Boundary-induced breaking in CFT yields log-log asymmetry","Finite groups saturate, Lie groups log-log in boundary CFT","Symmetry breaking by boundary: log-log entanglement asymmetry","Quench restores symmetry at t=ell/2 in BCFT"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000838,"raw_usage":{"total_tokens":3701,"prompt_tokens":1038,"completion_tokens":2663,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":654,"completion_tokens_details":{"reasoning_tokens":2577}},"tokens_in":654,"tokens_out":2663,"duration_ms":19864,"temperature":1.0,"reasoning_tokens":2577,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-12T20:20:13.480278+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Numerically compute the entanglement asymmetry in a lattice model with a $U(1)$ symmetry broken only at the boundary (for example a critical free-boson chain with Dirichlet boundary) and check whether $\\Delta S_A^{(n)} - \\frac{\\dim G}{2}\\log\\log(\\ell/\\epsilon)$ approaches the $O(1)$ constant in Eq. (2.26) as $\\ell$ grows; any residual $\\ell$ dependence at fixed $n$ would falsify the power-law ansatz (2.20). In the Potts model, the same check is the prediction that the correction to $\\log 3$ decays as $\\ell^{-2/15}$ for the $|B+C\\rangle$ boundary and as $\\ell^{-4/3}$ for $|A\\rangle$.","supporting_citations":[{"cited_title":"Cardy,Boundary conditions, fusion rules and the verlinde formula, Nuclear Physics B 324 (1989) 581","cited_arxiv_id":null,"evidence_quote":"Fixes the Potts boundary states and boundary-condition-changing operator dimensions used in the Z3 examples."}],"review_version":1}