{"id":"8ee8506d-9e79-4d20-880f-59e70b5bb18c","arxiv_id":"2502.02030","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":4,"one_line_summary":"A HOTRG-based algorithm computes entanglement entropy for arbitrary subsystem sizes and recovers the central charge c=0.49997(8) for the critical Ising chain.","lead":"This paper presents a tensor network method to compute entanglement entropy for any subsystem size in a 1D quantum system, and tests it on the transverse-field Ising model at criticality. It reproduces the known central charge c=1/2 with high precision, suggesting the method is a robust alternative to Monte Carlo, especially for models with sign problems.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Central arbitrary-size claim rests on unpublished tensor definitions and h<=2 tests; missing algorithm specification is the key weak point.","rationale":"I read the paper as proposing a novel HOTRG-based construction for subsystem-size-dependent entanglement entropy and benchmarking it on the critical 1D Ising model. The reported central charge c = 0.49997(8) is an impressive numerical result, and the convergence with D_cut in Table 1 is in the right direction. My stress-test focus is different from the reader's weakest_assumption: rather than the fit-range choice, I find the missing algorithm definition to be the most load-bearing concern. The manuscript itself flags this by referring to [17] for tensor definitions, and [17] is unpublished. Without the full construction, the claim of 'arbitrary subsystem size' cannot be checked. The numerical tests only exercise h <= 2, so even the empirical support does not cover the general case. This is not an accusation of error; it is a statement that the evidence as presented is incomplete. The reader's verdict is CONDITIONAL, and I agree that the paper should be accepted only conditionally, with the condition being the release of the full algorithm details and a more robust error analysis. I therefore recommend no change to the verdict, but I want the primary condition to be explicit: the arbitrary-size algorithm must be specified and tested beyond h <= 2 before the central claim is taken as established.","tokens_in":7062,"tokens_out":14566,"duration_ms":161392,"concrete_test":"Independently re-derive the boundary-factor construction for a Hamming-weight-3 subsystem (e.g., L = 32, l = 7) using only Eqs. (7) and Figs. 4-7, contract it exactly with the core matrix for small N, and compare the resulting S_A with exact diagonalization of the L = 32 critical Ising chain. If the construction cannot be uniquely read off from the manuscript, or if the computed S_A disagrees with exact diagonalization, the arbitrary-size claim is not established.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim is a new TRG algorithm for entanglement entropy of subsystems of arbitrary size, but the algorithm is not fully specified in this paper. In Section 2.1, after Eq. (6), the authors write 'See [17] for details on definitions,' and reference [17] is listed as 'Work in progress, (forthcoming)'. The boundary-factor construction in Figs. 4-7 and the binary algorithm of Eq. (7) therefore cannot be independently checked from the manuscript. The numerical benchmark does not close this gap: Section 3 restricts all calculations to subsystem sizes with Hamming weight h <= 2, i.e. l = 2^m + q with q = 0, 2^0, ..., 2^{m-1}. The cases h > 2, which exercise the full boundary-factor logic and the claimed O(D^3 h) contraction, are never tested. If the deferred definitions contain a mistake, or if the contraction pattern for h > 2 differs from the illustrated h <= 2 cases, the 'arbitrary size' claim fails even though the reported c = 0.49997(8) agrees with theory. A second, independent weakness is the post hoc fit range 7 <= l <= 768 in Section 3: the quoted error in Eq. (10) is a max-deviation with k fixed, so it does not include the systematic uncertainty from the choice of lower cutoff. This fit-range issue affects the strength of the benchmark, but the missing algorithm specification is more load-bearing for the central methodological claim.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The manuscript proposes a tensor renormalization group (TRG) method, based on the higher-order TRG (HOTRG) algorithm, to compute the entanglement entropy of a subsystem of arbitrary size in a one-dimensional quantum system. The reduced density matrix of a Gibbs state is represented as a (1+1)-dimensional tensor network, and the subsystem entropy is obtained from a 'trimmed' network in which boundary isometries are dropped using the isometry property U^dagger U = I. A binary decomposition of the subsystem size l is used to build a core matrix C and boundary factor B. The method is applied to the critical quantum Ising chain with L = 1024 and time extent alpha = 16, for subsystem sizes with Hamming weight h <= 2. The extracted central charge is c = 0.49997(8) for D = 96, in agreement with the theoretical value c = 1/2.","tokens_in":7292,"tokens_out":7949,"duration_ms":70944,"significance":"If the proposed algorithm is fully specified and validated, it would fill a genuine gap in the TRG toolbox: it would be the first TRG method to compute entanglement entropy for subsystems of arbitrary size, not just half of the system, with a claimed cost of O(D^3 h). The numerical benchmark is clean in the tested regime: the alpha -> infinity limit is checked in Fig. 8, the bond-dimension dependence in Table 1 is benign, and the extracted central charge is very close to the CFT prediction. However, the manuscript is not self-contained: the key algorithm definitions are deferred to a 'forthcoming' paper, and all numerical tests are restricted to Hamming weight h <= 2. These two limitations are load-bearing for the central claim and must be addressed before the result can be accepted as a general method.","major_comments":[{"comment":"The generalized algorithm for arbitrary subsystem size l is not fully specified in this manuscript. The definitions of the core matrix C and boundary factor B are delegated to reference [17], which is listed as 'Work in progress, (forthcoming)'. The rules for filling the blank boxes in Fig. 5 and the contraction patterns of Fig. 7 are presented pictorially, but the text does not give enough detail for an independent implementation. Since the central claim is a general TRG method for arbitrary l, this missing specification is load-bearing. The numerical tests in Section 3 cover only Hamming weight h <= 2 (l = 2^m + q with q = 0, 2^0, ..., 2^{m-1}); cases with h > 2, which would exercise the full boundary-factor construction and the claimed O(D^3 h) cost, are not presented. The authors should include the complete definitions (e.g., in an appendix) or make the forthcoming reference available before publication.","section":"Section 2.2 (after Eq. (7), Figs. 4-7)"},{"comment":"The fit range 7 <= l <= 768 is selected after observing that the effective central charge c(l) in Fig. 10 visibly deviates from the expected constant behavior for small l. This post hoc choice, together with the error estimate based on the maximum deviation with k fixed, does not account for the systematic uncertainty associated with the lower cutoff. The quoted central charge c = 0.49997(8) therefore carries an error bar that is likely underestimated. The authors should document the sensitivity of c to the fit range (for example, by repeating the fit with several lower cutoffs) or provide an independent criterion for selecting the cutoff.","section":"Section 3, Eq. (10) and Fig. 10"}],"minor_comments":[{"comment":"The sentence 'To keep K and K' finite, we need to take the limit beta -> infinity' is imprecise; the standard Suzuki-Trotter limit is N -> infinity with tau = beta/N fixed, which also sends beta to infinity. Please clarify.","section":"Section 2.1, Eq. (4)"},{"comment":"The entropy invariance statement S_A = -tr(U rho' U^dagger log(U rho' U^dagger)) = -tr(rho' log rho') is correct, but the text would be clearer if it explicitly stated that the nonzero spectrum is preserved under the isometric embedding U and that zero eigenvalues do not contribute to the von Neumann entropy.","section":"Section 2.2, trimming argument"},{"comment":"The caption and text should clarify the normalization in Fig. 8: it says 'All results are normalized by dividing them by the data at alpha = 1024', but it is not explicitly stated that the plotted quantity is S_A(alpha)/S_A(alpha=1024).","section":"Section 3, Fig. 8"},{"comment":"The statement that 'The theoretical form accurately describes the data in the whole region' is in apparent tension with the later exclusion of small-l data; please reconcile these statements, for instance by specifying that the deviation is visible only in the derivative quantity c(l).","section":"Section 3, Fig. 9 vs. Fig. 10"},{"comment":"The claim that 'no general TRG algorithm for subsystems of arbitrary size is yet known' should be substantiated with a brief discussion of how the prior TRG entanglement studies [12-15] are limited (for example, to half-space subsystems).","section":"Section 1, Introduction"},{"comment":"Please specify the range of m and q for which c(l) is computed, and note explicitly that l' is the next data point in the sequence for fixed q.","section":"Section 3, Eq. (9)"}],"recommendation":"major_revision","confidential_remarks":"The manuscript is a proceedings-style report, and the central novelty (arbitrary-size subsystem entropy) is not self-contained: the key algorithm definitions are in a 'forthcoming' paper. I recommend that the editor require the full definitions be included or made available before publication, and that a fit-range sensitivity analysis be added for the numerical claim. The algorithmic idea is promising and the benchmark is clean in the tested regime, so the paper could become acceptable after these revisions."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The core idea is genuinely new: prior TRG entanglement entropy work handles half-space bipartitions, and the binary-expansion construction of the boundary factor B and core matrix C for arbitrary subsystem size looks like a real extension. The derivation of the reduced density matrix tensor network from Suzuki-Trotter is standard but competently done, and the trimming argument dropping boundary isometries via U^dagger U = I is mathematically sound. The numerical benchmark is also solid as far as it goes: the convergence in alpha is checked properly, the observed S_A(l) = S_A(L-l) symmetry is a good sanity check, and the extracted central charge c = 0.49997(8) at D=96 matches theory almost too well. The table showing D-dependence moving monotonically toward 0.5 is credible. The soft spot is the one the stress test flags, and it is real. Section 2.1 says 'See [17] for details on definitions' for the transfer matrix T, and reference [17] is listed as 'Work in progress.' That means the precise form of the rank-4 tensor, including how the physical indices and the Trotter decomposition map into it, cannot be checked. The boundary-factor construction in Figs. 4-7 is described, but the box structures depend on b_k and the general contraction pattern is only illustrated for specific cases. The numerical section further restricts to Hamming weight h <= 2, i.e. l = 2^m + q with q a single power of two. So the claimed O(D^3 h) contraction and the h > 2 cases that exercise the full binary logic are never tested. If the deferred definitions contain a subtle sign error or the contraction pattern differs for h > 2, the central methodological claim would fail even though the benchmark is clean. The benchmark alone cannot close that gap because it only exercises the special cases. The fit-range selection is a second, milder issue. The effective central charge c(l) visibly deviates for small l, and the fit range 7 <= l <= 768 is chosen after seeing that deviation. The quoted error is a max-deviation with k fixed, so it does not include the systematic uncertainty from the cutoff choice. In a one-benchmark proceedings paper this is acceptable, but it keeps the significance moderate. Overall, the paper is a promising proceedings contribution, not a complete method paper. The authors are clearly thinking seriously and the algorithm as sketched is coherent. The missing details and the untested h > 2 cases are not flaws in what is presented; they are gaps in what can be independently verified. A serious referee should engage with this, because the idea is worth pursuing and the benchmark is encouraging. For my own work I would not cite it yet, pending the full paper with the tensor definitions and h > 2 tests. Recommendation: send to peer review, but require the algorithm specification and ideally one h > 2 test before acceptance.","headline":"The binary-expansion construction is a genuine extension of TRG to arbitrary subsystem sizes, and c = 0.49997(8) is an impressive benchmark, but the paper defers the key tensor definitions to an unpublished work and only tests h <= 2, so the arbitrary-size claim is not yet fully demonstrated.","tokens_in":860,"tokens_out":1233,"would_cite":false,"duration_ms":30305,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"A tensor-network algorithm extracts central charge $c = 0.49997(8)$ from entanglement entropy.","keywords":["tensor renormalization group","entanglement entropy","quantum Ising model","central charge","HOTRG","reduced density matrix","lattice field theory","conformal field theory"],"falsifier":"Recompute the effective central charge $c(l)$ at larger bond dimensions $D > 96$ and larger system sizes; if the deviation at small $l$ from a constant effective central charge does not shrink toward $c = 1/2$ as $D$ grows, the choice of the fitting range $7\\le l\\le 768$ is responsible for the agreement, and the extracted central charge would shift outside the reported error when the window is varied.","tokens_in":6777,"feed_emoji":"🧮","tokens_out":14845,"duration_ms":120040,"temperature":0.7,"pith_summary":"This paper proposes a tensor renormalization group algorithm that computes the entanglement entropy of a single connected segment of arbitrary length in a one-dimensional quantum system. The reduced density matrix of a Gibbs state is written as a $(1+1)$-dimensional tensor network, and the higher-order tensor renormalization group (HOTRG) coarse-grains it while trimming isometries that do not affect the entropy. The authors test the method on the critical one-dimensional quantum Ising model and extract the central charge $c = 0.49997(8)$ at bond dimension $D = 96$, matching the theoretical value $c = 1/2$. If correct, this is the first general TRG construction for subsystem-size-dependent entanglement entropy, avoiding the replica trick and the $s\\to 1$ extrapolation used in Monte Carlo computations.","feed_headline":"Tensor-network entropy algorithm reproduces central charge 1/2","feed_subtitle":"A new tensor-network construction handles subsystems of any size and matches c = 1/2 in the critical Ising model.","key_machinery":"The central object is the tensor network representation of the reduced density matrix $\\rho_A$ of a Gibbs state, built from the transfer matrix of the $(1+1)$-dimensional classical Ising model obtained by splitting the imaginary-time evolution into small steps. Coarse-graining proceeds with the higher-order tensor renormalization group (HOTRG): pairs of tensors are contracted into a new tensor $T' = U^\\dagger M U$ using an isometry $U$ built from the $D$ largest eigenvectors of $M^\\dagger M$. Because $U$ satisfies $U^\\dagger U = I$ but not $U U^\\dagger = I$, isometries attached to the boundary of the subsystem can be removed without changing the entanglement entropy $S_A = -\\operatorname{Tr}(\\rho_A \\log \\rho_A)$. The generalized algorithm decomposes the trimmed network into a core matrix $C$ and a boundary factor $B$, whose shape is fixed by the binary digits of the subsystem size $l$ (and $l-1$); contracting $C$ and $B$ gives $\\rho_A$ at cost $O(D^3 h)$, where $h$ is the Hamming weight of $l$.","core_discovery":"The central claim is that entanglement entropy as a function of subsystem size can be obtained directly from tensor coarse-graining, without the replica trick. The paper constructs the reduced density matrix $\\rho_A$ of a Gibbs state as a tensor network, applies HOTRG to it, and shows that the isometries attached to the boundary of the subsystem can be removed using $U^\\dagger U = I$, leaving a trimmed network composed of a core matrix and a boundary factor. The core matrix is a coarse-grained reduced density matrix of the half-space subsystem, while the boundary factor is assembled from isometries according to the binary representation of the subsystem size $l$ and of $l-1$; contracting the two gives $\\rho_A$ with cost $O(D^3 h)$, where $h$ is the Hamming weight of $l$. For the critical transverse-field Ising model on a 1024-site lattice with a large time direction, the computed entanglement entropy follows the finite-size conformal scaling formula $S_A(l,L) = (c/3)\\log(L\\sin(\\pi l/L)) + k$, and fitting yields $c = 0.49997(8)$ for $D = 96$, in agreement with $c = 1/2$. The paper takes this benchmark as evidence that the method is valid for arbitrary subsystem sizes and suitable for zero-temperature quantum systems.","pith_inferences":["A natural next test is to apply the same trimmed-network construction to models where Monte Carlo has a sign problem; if the tensor network can be formed, the method would yield entanglement entropy where the replica-based Monte Carlo route fails.","The reported error bar is computed with the fit parameter $k$ held fixed and the fitting window chosen after seeing $c(l)$, so the $0.49997(8)$ figure should be interpreted as conditional on that window; a window-independent estimate would be a stronger check.","The boundary-factor decomposition is specific to a one-dimensional subsystem whose boundary is two points; extending the idea to two spatial dimensions would require a new rule for trimming the network along a one-dimensional boundary, and the binary-decomposition structure does not carry over directly.","If the method survives those extensions, the holographic relation between entanglement entropy and geometry noted in the outlook could become a concrete lattice-level test rather than a formal correspondence."],"forward_implications":["For the critical 1D quantum Ising model, the extracted central charge converges toward $c = 1/2$ as the bond dimension grows: $0.4998(2)$ at $D = 64$, $0.4999(1)$ at $D = 80$, and $0.49997(8)$ at $D = 96$.","The same algorithm applies to any subsystem size $l$, not only half the system, with the post-coarse-graining contraction cost scaling as $O(D^3 h)$ where $h$ is the Hamming weight of $l$.","Entanglement entropy can be obtained without the replica trick, so no extrapolation to Rényi index $s\\to 1$ is needed.","At a time direction 16 times the spatial length, the entropy is effectively the ground-state value, so zero-temperature entanglement entropy is accessible in this framework.","The authors expect the method to extend to higher-dimensional theories and to studies of phase transitions using entanglement entropy as a probe."],"supporting_citations":[{"why":"Supplies the higher-order tensor renormalization group procedure that the new algorithm uses for coarse-graining the reduced-density-matrix network.","marker":"[16]"},{"why":"Provides the tensor-network definitions of the transfer matrix and reduced density matrix on which the construction relies.","marker":"[17]"},{"why":"Gives the quantum-field-theory formula for entanglement entropy that motivates the scaling form used to fit the data.","marker":"[1]"},{"why":"Supplies the conformal-field-theory finite-size scaling formula used as the fitting function.","marker":"[2]"},{"why":"Establishes that entanglement entropy can be computed within a tensor renormalization group framework, the line of work this paper extends.","marker":"[12]"},{"why":"Computes entanglement entropy in the O(2) model with a TRG-based method, demonstrating prior subsystem-size handling.","marker":"[13]"},{"why":"Shows how to estimate the central charge from Rényi entanglement entropy, the benchmark against which the extracted $c$ is compared.","marker":"[14]"},{"why":"Evaluates Rényi entropies of the $(1+1)$-dimensional O(3) model with TRG, another prior step toward general TRG entanglement computations.","marker":"[15]"}],"fun_headline_variants":["Tensor renormalization computes entropy without replica trick","Central charge 1/2 from tensor-network entanglement entropy","HOTRG-based entropy matches Ising central charge","No replica trick: tensor RG yields Ising c=1/2","Entanglement entropy via tensor RG: central charge 1/2"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The reported central charge depends on excluding small subsystem sizes from the fit, on the assumption that the visible non-constant behavior of the effective central charge at small $l$ is a lattice artifact; if that non-constant behavior is a genuine physical effect, the fitted value $0.49997(8)$ would be biased and the quoted error would not capture the bias.","fun_headline_variants_meta":{"raw":{"variants":["Tensor renormalization computes entropy without replica trick","Central charge 1/2 from tensor-network entanglement entropy","HOTRG-based entropy matches Ising central charge","No replica trick: tensor RG yields Ising c=1/2","Entanglement entropy via tensor RG: central charge 1/2"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000953,"raw_usage":{"total_tokens":4066,"prompt_tokens":947,"completion_tokens":3119,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":563,"completion_tokens_details":{"reasoning_tokens":3046}},"tokens_in":563,"tokens_out":3119,"duration_ms":20332,"temperature":1.0,"reasoning_tokens":3046,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-09T13:37:31.075916+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Recompute the effective central charge $c(l)$ at larger bond dimensions $D > 96$ and larger system sizes; if the deviation at small $l$ from a constant effective central charge does not shrink toward $c = 1/2$ as $D$ grows, the choice of the fitting range $7\\le l\\le 768$ is responsible for the agreement, and the extracted central charge would shift outside the reported error when the window is varied.","supporting_citations":[{"cited_title":"Hayazaki, D","cited_arxiv_id":null,"evidence_quote":"Provides the tensor-network definitions of the transfer matrix and reduced density matrix on which the construction relies."}],"review_version":1}