{"id":"380e6234-96a4-4c1b-984c-476c3218e47d","arxiv_id":"2601.02459","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":6,"one_line_summary":"The 2D O(N>2) nonlinear sigma model has complex-conjugate nontrivial fixed points, realized in non-Hermitian Heisenberg spin chains, forming a complex CFT universality class.","lead":"The authors show that the 2D O(N>2) nonlinear sigma model, normally asymptotically free, has complex-coupling fixed points described by a complex conformal field theory, and they locate these in non-Hermitian spin-chain models. The result suggests engineered dissipation can relax a system into a long-range-entangled critical state.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Loop-crossing exponent at N=3 is only marginally irrelevant (Re∆=2.04) and the spin-1 fit used predicted dimensions as targets; need a direct finite-size test of the crossing RG eigenvalue to support the NLSM-to-CCFT port.","rationale":"The reader's CONDITIONAL verdict is appropriate. The most load-bearing assumption is the irrelevance of loop crossings in the NLSM, which the paper supports via the Coulomb-gas prediction Re∆_{ℓ=4}=2.04 at N=3. This value is close to 2, and the numerical optimization used the predicted scaling dimensions to locate the finite-size critical point, so the two headline agreements (ϵ and ℓ=1) are by construction not independent confirmations. However, the paper provides independent evidence from the central charge (extracted from entanglement entropy) and from the ladder model with exact non-invertible symmetry, where only a single complex parameter is tuned and the spectrum matches the CCFT predictions. These considerations neither prove nor disprove the loop-crossing relevance; they highlight the need for a direct test of the crossing operator's RG eigenvalue at the complex fixed point. A finite-size scaling study of the ℓ=4 singlet gap without the second tuning (or a two-loop beta-function computation for the complex NLSM) would settle whether the port is valid. Until then, CONDITIONAL—rather than full acceptance—remains the correct verdict.","tokens_in":22372,"tokens_out":20567,"duration_ms":210247,"concrete_test":"At the finite-size fixed point (J2,K)+ in Eq. (6), compute the low-energy spectrum for L=16, 20, 24, 32 (DMRG or exact diagonalization) and extract the ℓ=4 singlet gap. Fit its effective scaling dimension ∆_eff(L) to ∆_∞ + c L^{Re(y)} cos(Im(y) ln L + φ). Confirm that ∆_∞ ≈ 2.04 and Re(y) < 0 (e.g., Re(y) ≈ -0.04) with systematic convergence as L grows. Alternatively, compute the two-loop beta function for the complex-coupling O(3) NLSM and check for a pair of complex zeros matching the Coulomb-gas exponents; a negative or zero Re(y) would invalidate the port.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim that the O(N>2) NLSM flows to the loop-model CCFT rests on the loop-crossing operator (singlet component of ℓ=4 watermelon) being irrelevant at the complex fixed point. Eq. (4) gives Re∆_{ℓ=4} ≈ 2.04 at N=3, just above marginal, and this value is obtained by analytic continuation from N≤2 in a model where crossings are absent. The near-marginality is mirrored in the numerics: the spin-1 chain required a second tuned complex parameter (J2,K) to suppress loop crossings, and the optimization (SM Sec. II, Eq. S3) minimized residuals against the predicted ℓ=1 and ϵ dimensions, so the agreement of those operators is not independent evidence. If the true Re∆_{ℓ=4} were ≤2 (e.g., due to corrections at the complex fixed point or a misidentification of the loop-crossing operator), the NLSM would not flow to the CCFT and the 'single relevant singlet / single complex parameter' genericity claim would fail. The ladder model with exact non-invertible symmetry avoids crossings but does not test the NLSM port.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper claims that the two-dimensional O(N>2) nonlinear sigma model (NLSM), when the coupling g is allowed to be complex, possesses a pair of complex-conjugate fixed points described by a complex CFT (CCFT) with complex central charge c(N). This is obtained by analytically continuing the known O(N) loop-model fixed points from N≤2 to N>2 and by arguing that loop crossings—the singlet component of the ℓ=4 watermelon operator—are irrelevant for N>2. The authors then locate the N=3 CCFT numerically in non-Hermitian spin-1 Heisenberg chains, in a K1-K2 chain, and in a spin-1/2 ladder with an exact non-invertible symmetry, comparing the complex spectrum and the biorthogonal entanglement entropy with Coulomb-gas predictions. They further propose a Lindbladian whose no-click trajectories realize the non-Hermitian Hamiltonian, so that dissipative evolution can prepare the CCFT vacuum.","tokens_in":22720,"tokens_out":6246,"duration_ms":69216,"significance":"If the central claim holds, it is conceptually significant: it would show that asymptotic freedom is lost in the complex coupling plane, that a generic non-Hermitian O(N>2) model can flow to a CCFT with a single complex relevant coupling, and that dissipative no-click dynamics can prepare a long-range entangled critical state. The numerical work is extensive: several distinct microscopic models are studied, and the non-fitted quantities—central charge, current, stress tensor, and several watermelon multiplets—show reasonable agreement. The Lindblad construction is explicit and physically motivated. However, the NLSM-to-loop-model port rests on the marginal irrelevance of loop crossings, and the main numerical confirmation uses the predicted ϵ and ℓ=1 dimensions as fitting targets. These issues are load-bearing and require additional evidence before the strong claims can be accepted.","major_comments":[{"comment":"The crossover from the O(N) loop model to the NLSM rests entirely on the loop-crossing operator (singlet component of the ℓ=4 watermelon) being irrelevant at the CCFT. At N=3, Eq. (4) predicts Re∆_lc ≈ 2.04, only 0.04 above marginal. The numerical spectrum in Fig. 3(d) gives Re∆ for D0 as 1.94, actually below 2. Because the critical point was located by fitting the predicted ℓ=1 and ϵ dimensions, this near-marginal/relevant value is not an independent check. Please provide a direct finite-size determination of the loop-crossing RG eigenvalue, e.g. the L-dependence of the D0 gap at fixed (J2,K), and state whether Re∆_lc > 2 in the thermodynamic limit. If it is ≤2, the NLSM does not flow to the CCFT and the central claim fails.","section":"Predictions for the O(N) NLSM, Eq. (4), Fig. 2(b)"},{"comment":"The finite-size critical point (J2,K) is found by minimizing the cost function with R1 and R2 set to zero against the predicted ∆ϵ and ∆ℓ=1. Reporting these two operators in the main text as 'good agreement' is circular. The independent confirmation consists of the central charge, J, T, and the ℓ=2,3,4 watermelon operators (with the caveat in Comment 1). Please label ϵ and A as fitting targets, remove them from the confirmation table, and quantify the agreement using only the non-fitted data. This is essential for the non-perturbative claim.","section":"Supplemental Material Sec. II, Eq. (S3)"},{"comment":"The genericity claim—single relevant singlet and single complex tuning parameter—is in tension with the actual numerical procedure, which tunes two complex parameters (J2,K) or (K1,K2) and uses an optimization weight w in Eq. (S3). The explanation is that loop crossings are weakly irrelevant, but this means the identified point is not shown to lie on the one-complex-parameter critical manifold. Please demonstrate either that a one-complex-parameter family through (J2,K)+ remains critical in the thermodynamic limit, or provide a quantitative estimate of the residual loop-crossing coupling at L=14 and its RG flow. Without this, the 'generic' statement is not supported by the numerics.","section":"Microscopic model and numerical results, Eq. (6)"},{"comment":"The spin-1/2 ladder is an exact realization of the dilute Temperley-Lieb/loop model and therefore excludes loop crossings by construction. It strengthens the loop-model side but cannot test the NLSM port. Please state this limitation explicitly where the ladder is presented as confirmation of the NLSM CCFT, and avoid presenting it as independent evidence for the NLSM-to-CCFT claim.","section":"End Matter, Ladder model with exact non-invertible symmetry"}],"minor_comments":[{"comment":"The table of identified CCFT states appears to be duplicated in panels (a) and (b)/(c) of Fig. 3; please remove the duplicate and ensure the table is a single panel.","section":"Fig. 3(d)"},{"comment":"Define clearly that R2 is complex and that the cost function uses |R2| after the feature-scaling weight w; the notation f(J2,K)=|R1|+w|R2| is only presented in the text and could be made explicit.","section":"Eq. (S3)"},{"comment":"The phrase 'unnecessary transition' is used without a precise definition. Please define what is meant in terms of the RG flow or the phase diagram.","section":"Discussion"},{"comment":"The claim that the system 'naturally relaxes' to the CCFT should be qualified: no-click postselection has a success probability that decays exponentially in time and system size. This is not a fatal issue, but it should be stated to calibrate the proposed state-preparation protocol.","section":"Monitored dynamics, Eq. (8)-(9)"},{"comment":"For a numerics-heavy Letter, please include a data/code availability statement, especially since exact diagonalization and DMRG details are only partially described in the Supplemental Material.","section":"General"}],"recommendation":"major_revision","confidential_remarks":"The paper is potentially important, but the loop-crossing port is the crux and it is only marginally safe at N=3. The fact that the fitted operators are also the headline confirmed operators needs to be addressed transparently. I would ask for a direct finite-size test of the loop-crossing RG eigenvalue and a clear separation of fitted versus predicted quantities before publication."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Quick take: this is a real step forward. The CCFT itself was already in the loop-model literature, but the paper does three genuinely new things: it argues the O(N>2) NLSM flows to those fixed points in the complex plane, it realizes them in microscopic non-Hermitian Heisenberg chains, and it sketches a Lindblad no-click preparation protocol. The numerical work is extensive and mostly convincing: the spectrum of watermelon operators, current, stress tensor, and the complex central charge from entanglement entropy all agree with the analytic continuation, and the DMRG central charge holds up to L=40. The ladder model with exact non-invertible symmetry is a particularly clean test — no loop crossings, single complex parameter, and it lands on the same CCFT.\n\nThe soft spots are real but proportionate. First, the NLSM-to-loop-model port hangs on the singlet component of the l=4 watermelon operator being irrelevant at N=3, with ReDelta approx 2.04. That is one epsilon above marginal. If corrections at the complex fixed point push it below 2, the NLSM flows elsewhere and the 'asymptotic freedom lost' framing collapses. The paper acknowledges this but does not test the crossing RG eigenvalue directly. Second, the spin-1 chain fixed point is located by optimizing residuals against the predicted l=1 and epsilon scaling dimensions (SM Eq. S3), so the 1% agreement of those two operators is not independent evidence. The other operators and central charge are independent, so the paper is not circular overall, but the two headline numbers in the main-text table should be labeled as fit targets. The ladder model's single-parameter tuning is a genuine check of the genericity claim, which helps.\n\nThe absence of error bars and the L=14 ED are minor at this stage; the DMRG central charge helps. Citation pattern is fine — Ref. [22] is the same group's earlier loop-model result and is properly credited. The self-citation is legitimate.\n\nWho is this for? Condensed-matter theorists working on non-Hermitian criticality, monitored dynamics, and CFT. It deserves a serious referee. I would send it to review, with the expectation that the loop-crossing relevance and the fit-versus-prediction distinction get highlighted. The central existence claim looks credible; the genericity claim needs a direct finite-size test of the crossing RG eigenvalue.","headline":"Serious numerical evidence for a complex O(3) CFT in non-Hermitian spin chains, but the NLSM port rests on a marginally irrelevant operator and two headline exponents are fit targets, not predictions.","tokens_in":23203,"tokens_out":2196,"would_cite":true,"duration_ms":21695,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"The two-dimensional O(N) nonlinear sigma model acquires a pair of complex-conjugate fixed points when the coupling is allowed to be complex, described by a complex conformal field theory.","keywords":["complex conformal field theory","O(N) nonlinear sigma model","asymptotic freedom","non-Hermitian spin chains","O(N) loop model","complex central charge","Lindblad dynamics","entanglement entropy"],"falsifier":"Measure the finite-size scaling of the loop-crossing coupling in the non-Hermitian spin-1 chain: if Re Δ_{ℓ=4} ≤ 2, the near-marginal operator will grow with system size and the spectrum will drift away from the predicted CCFT; alternatively, a high-precision transfer-matrix calculation of the ℓ=4 watermelon scaling dimension at N=3 would settle whether the loop-model fixed point carries over to the NLSM.","tokens_in":22210,"feed_emoji":"🧲","tokens_out":5618,"duration_ms":54109,"temperature":0.7,"pith_summary":"This paper argues that asymptotic freedom in the 2D O(N) nonlinear sigma model for N>2 is not the whole story: if the coupling constant is allowed to be a complex number, the theory flows to a pair of complex-conjugate fixed points governed by a complex conformal field theory (CCFT) with a complex central charge. The fixed point is generic, having a single relevant singlet operator, so it should arise in any non-Hermitian model with O(N) symmetry after tuning one complex parameter. The authors confirm this at N=3 by exact diagonalization of two non-Hermitian spin-1 Heisenberg chains and a spin-1/2 ladder, matching the predicted complex central charge and scaling dimensions. They also construct a Lindbladian whose no-click dynamics realize the non-Hermitian Hamiltonian; since the CCFT vacuum is the slowest-decaying eigenstate, dissipative evolution naturally relaxes into the CFT state, offering a route to preparing long-range entangled states.","feed_headline":"Complex couplings unlock a critical fixed point in 2D magnets","feed_subtitle":"The textbook 2D O(N) model flows to a complex conformal field theory when the coupling is allowed to be complex.","key_machinery":"The carrying object is the O(N) loop model's Coulomb-gas data, analytically continued to N>2: the fixed-point branches g̃± = 1 ± e(N) with e(N) = (1/π)cos⁻¹(N/2) give closed-form central charge and scaling dimensions that are real for N≤2 and move into a complex-conjugate pair for N>2. The link to the NLSM is provided by the loop-crossing operator, the singlet component of the ℓ=4 watermelon operator, which the paper computes to be irrelevant for all N>2 (Re Δ_{ℓ=4} rises from about 2.04 at N=3 to 2.5 as N→∞). That irrelevance is what makes the loop-model CCFT the generic critical behavior of the non-Hermitian NLSM rather than an artifact of the loop-model truncation.","core_discovery":"The central claim is that the 2D O(N) nonlinear sigma model, for N>2, possesses a pair of complex-conjugate fixed points at complex values of the coupling g, obtained by analytic continuation of two branches of real fixed points that annihilate at N=2. At these points the theory becomes a CCFT with central charge c±(N) = 1 − 6(1−g̃±)²/g̃±, where g̃± = 1 ± (1/π)cos⁻¹(N/2); for N=3 this gives c ≈ 1.51 ± 0.158i. The fixed point is generic: it has a single relevant singlet operator, the energy operator, so it requires tuning only one complex parameter. The port from the loop model to the NLSM rests on the irrelevance of loop crossings, identified with the singlet component of the ℓ=4 watermelon","pith_inferences":["Editorial inference: If the complex fixed point exists beyond the one-loop approximation, the NLSM on the real axis may pass near the complex fixed point and exhibit slow 'walking' RG flow or weakly first-order behavior, analogous to other complex-fixed-point phenomena.","Editorial inference: The near-marginal loop-crossing exponent at N=3 (Re Δ=2.04) makes the universality claim delicate; a high-precision transfer-matrix or Monte Carlo calculation of the ℓ=4 watermelon dimension would directly sharpen or refute the port from loop models to the NLSM.","Editorial inference: The ladder model with exact non-invertible symmetry may be a cleaner experimental and numerical platform for observing the CCFT, since it eliminates the slow finite-size convergence caused by loop crossings.","Editorial inference: The dissipative preparation protocol relies on postselecting no-click trajectories; a natural extension, noted as open by the authors, is whether the CCFT can appear as the steady state of the full Lindbladian without postselection."],"forward_implications":["Asymptotic freedom is lost in the complex coupling plane: a wide region of initial couplings flows to the CCFT rather than to the trivial g=0 fixed point.","The CCFT is the generic critical point of any non-Hermitian O(N>2)-symmetric model, requiring only a single complex tuning parameter.","Numerical agreement at N=3 in spin-1 chains and a spin-1/2 ladder supports the analytic continuation of the N≤2 real fixed points to complex N>2.","Because the CCFT vacuum is the longest-lived state, no-click dissipative dynamics of the proposed Lindbladian relaxes into a CFT state, providing a route to preparing long-range entangled states through engineered dissipation.","Perturbative beta functions at any fixed loop order should generically have zeros away from the positive real axis; whether these align with the CCFT points is a testable consistency check on perturbation theory."],"fun_headline_variants":["Complex conformal fixed point found in 2D magnets","Non-Hermitian spin chains hit a complex critical point","Dissipative chains relax into a CFT state via complex couplings","Beyond asymptotic freedom: complex fixed point in O(N)"],"cache_read_input_tokens":2304,"weakest_assumption_plain":"The argument stands on the claim that loop crossings—the singlet component of the ℓ=4 watermelon operator—are irrelevant for N>2; at N=3 the computed exponent Re Δ_{ℓ=4} ≈ 2.04 is barely above the marginal value of 2, so if this near-marginal operator is actually relevant or exactly marginal, the NLSM would flow elsewhere and the CCFT would not be the generic fixed point.","fun_headline_variants_meta":{"raw":{"variants":["Complex conformal fixed point found in 2D magnets","Non-Hermitian spin chains hit a complex critical point","Dissipative chains relax into a CFT state via complex couplings","Beyond asymptotic freedom: complex fixed point in O(N)"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000891,"raw_usage":{"total_tokens":3733,"prompt_tokens":851,"completion_tokens":2882,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":595,"completion_tokens_details":{"reasoning_tokens":2814}},"tokens_in":595,"tokens_out":2882,"duration_ms":18527,"temperature":1.0,"reasoning_tokens":2814,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-03T12:30:03.181401+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Measure the finite-size scaling of the loop-crossing coupling in the non-Hermitian spin-1 chain: if Re Δ_{ℓ=4} ≤ 2, the near-marginal operator will grow with system size and the spectrum will drift away from the predicted CCFT; alternatively, a high-precision transfer-matrix calculation of the ℓ=4 watermelon scaling dimension at N=3 would settle whether the loop-model fixed point carries over to the NLSM.","supporting_citations":[],"review_version":1}