{"id":"fbda193a-fd17-464c-b787-12be2f689374","arxiv_id":"1908.02761","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":8,"one_line_summary":"At the many-body localization transition in a 20-site spin chain, logarithmic negativity decays exponentially with the ratio of block separation to block size, while mutual information decays as a power law, independent of block size.","lead":"A numerical study of a disordered spin chain finds that at the many-body localization transition, quantum entanglement between separated blocks depends only on the ratio of block separation to block size. This provides direct evidence of scale invariance at the transition and constrains theories of the many-body localization transition.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Scale-invariance claim is anchored to h=3.25 read off the same nearest-neighbor collapse that Fig. 4 then uses, with no finite-size scaling or independent h_c(L=20) estimate to rule out an accidental crossing.","rationale":"The reader's CONDITIONAL verdict is the right calibration. The proposed scale-invariant behavior is a plausible and interesting numerical observation; the visual collapse in Fig. 4 is real evidence, and the log-l self-entanglement analysis in Fig. 5 provides some corroboration. However, the load-bearing assumption, that h=3.25 is the genuine L=20 transition point rather than a point where finite-size curves happen to cross, is under-supported. Because the same normalized quantities are used both to select h and to demonstrate the effect, a selection effect is present. My proposed check would settle this by using an independent critical-point estimate and a quantitative collapse metric across system sizes. If the check passes, the central claim is substantially strengthened; if not, the paper should be treated as evidence for a finite-size crossover only, not for universal scale invariance.","tokens_in":11835,"tokens_out":10676,"duration_ms":121376,"concrete_test":"Compute the normalized bond negativity and mutual information for L=14,16,18,20 over a disorder grid around h=2.5-4.0, and define a collapse metric chi(h,L)=max over d/l of the standard deviation across block sizes divided by the mean across block sizes, excluding l values with fewer than several valid block pairs. Then (i) determine h*(L) minimizing chi and compare it with an independent estimate of h_c(L) from the average level-spacing ratio or from the crossing of mid-spectrum entanglement entropy; (ii) check whether the minimum chi decreases with L. If h*(L) disagrees with the independent h_c(L), or the collapse does not sharpen with L, the h=3.25 collapse is better explained as a finite-size crossing than as evidence of global scale invariance.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The paper's core statement is that at the transition point the normalized bond negativity and mutual information depend only on d/l, so that curves for different block sizes collapse. The only value of h for which this is shown is h=3.25 at L=20, and that value is not established independently. In the Results section (after Eq. 4 and in Fig. 3), h=3.25 is inferred from the fact that the nearest-neighbor normalized bond quantities for different l cross/collapse; the same class of normalized data is then used in Fig. 4(b,e) to claim scale invariance. This makes h=3.25 a selected value rather than a tested prediction. Since the Model section states that h_c is suspected to lie between h=3.5 and 5, the choice of 3.25 needs finite-size justification. Without a collapse-quality metric, without bootstrap or error estimates, and without comparing L=14,16,18,20 for the Fig. 4 quantities, the observed bundle of curves at one disorder strength is not separated from a finite-size crossing. The later log-l analysis in Fig. 5 covers multiple L, but it does not validate the collapse itself. Moreover, because two disjoint blocks of size l in an open chain of length L satisfy d<=L-l, the large-d/l tail of the collapse in Fig. 4 is supported only by small l, so the quoted functional forms are not tested across all block sizes.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies the scale-invariance of entanglement and correlation measures at the many-body localization (MBL) transition in a disordered Heisenberg chain. For L=20, exact diagonalization is used to compute the logarithmic negativity and mutual information between two disjoint blocks of equal size l separated by distance d. The authors report that at disorder strength h≈3.25, the normalized block measures collapse onto a single curve as a function of d/l, with logarithmic negativity decaying exponentially and mutual information decaying as a power law. They also report a logarithmic scaling of the self-entanglement near the transition and argue that these findings reveal a scale-invariant, multipartite entanglement structure in critical eigenstates, with implications for tensor-network descriptions of the MBL transition.","tokens_in":12130,"tokens_out":10133,"duration_ms":103274,"significance":"If correct, the result provides a concrete and falsifiable signature of scale invariance at the MBL transition and a clear distinction between the decay of logarithmic negativity and mutual information, which would constrain renormalization-group and tensor-network theories of the transition. The use of two controllable length scales (block size and separation) is a natural and powerful probe, and the visual collapse in Fig. 4 is striking. The paper also releases code (quimb) that enables independent verification of the numerics. However, the central claim rests on a single disorder strength at a single system size, with the critical field selected from the same data that is later used to claim collapse, and no quantitative collapse metrics or error estimates are provided. These issues temper the significance of the result as it stands.","major_comments":[{"comment":"The identification of h=3.25 as the transition point for L=20 is made by the observed collapse of the nearest-neighbor normalized quantities in Fig. 3, and the same value is then used in Fig. 4(b,e) to demonstrate the scale-invariant decay. This is a circular selection: the collapse in Fig. 4 is not an independent test but a consequence of choosing h to make the nearest-neighbor data collapse. The text states 'We infer that h∼3.25 corresponds to the transition point for this total system size of L=20, which matches previous studies [73],' but reference [73] is a software paper and does not provide an independent estimate of h_c. Since the Model section itself quotes the suspected transition range h~3.5-5, the manuscript needs an independent determination of h_c(L=20)—for example from level statistics or entanglement entropy scaling—before the collapse can be attributed to criticality. As written, the scale invariance claim is not tested but imposed.","section":"Results, Fig. 3 and text after Eq. (4)"},{"comment":"The exponential and power-law fits to the collapsed data in Fig. 4 are reported without any uncertainties, goodness-of-fit measures, or comparison to alternative forms. The statement in the Fig. 4 caption that a stretched exponential 'was not found to be as natural as a power law' is not supported by any quantitative criterion. Furthermore, no error bars or collapse-quality metric (e.g., variance of the curves, a Q-test, or bootstrap analysis) are provided, so the reader cannot judge whether the residual l-dependence in Fig. 4(b,e) is statistically significant. A quantitative collapse measure is essential for the central claim of scale invariance.","section":"Results, Fig. 4 and Eqs. (5)-(6)"},{"comment":"The scale-invariance claim is demonstrated only at L=20. Although Fig. 5 examines the self entanglement for L=14,16,18,20 and shows a peak in the log-law coefficient, the bond quantities E~d(l) and I~d(l) are never analyzed as functions of L. Consequently, it is unknown whether the data collapse improves with system size or whether the fitted parameters (λE≈0.45, αI≈0.5) drift. Without a finite-size scaling analysis, the observed bundle of curves at one disorder strength and one system size cannot be distinguished from a finite-size crossing.","section":"Results, Fig. 4"},{"comment":"The accessible range of d/l depends strongly on the block size l because two disjoint blocks of size l in an open chain of length L satisfy d ≤ L-l. For L=20, l=10 admits only d/l=1, while l=1 admits d/l up to 19. Thus the large-d/l tail of the collapse in Fig. 4 is supported almost entirely by the smallest block sizes, and the exponential (Eq. 5) and power-law (Eq. 6) forms are not tested uniformly across all block sizes. The authors should either restrict the quoted functional forms to the common range of d/l for all l, or present the data separated by l to show the collapse holds in overlapping ranges.","section":"Results, Fig. 4"},{"comment":"The text says that the coefficients CE, CI, λE, and αI 'might all be functions of L, l and h,' but then quotes single values (CE~2, λE~0.45, CI~0.3, αI~0.5). If these coefficients carry any l-dependence, the apparent collapse in Fig. 4 after normalizing by l would be a trivial scaling property rather than a universal signature. The authors must state explicitly that the fits in Eqs. (5)-(6) assume l-independent constants, and should report the fitting residuals as a function of l to demonstrate that the collapse is not an artifact of the normalization.","section":"Eqs. (5)-(6) and the paragraph after them"}],"minor_comments":[{"comment":"The partial transpose in the definition of logarithmic negativity is written as ρ_AB^{T_X} with 'subsystem X' undefined; it should specify TX = TA or TB.","section":"Eq. (3)"},{"comment":"The word 'Blocked' in the caption 'Average Nearest Neighbour Blocked Logarithmic Negativity' appears to be a typo; likely 'block' or 'block-entanglement' was intended.","section":"Fig. 3 caption"},{"comment":"The caption reports mean fitting uncertainties for the three coefficients but the main text does not refer to them; providing analogous uncertainties for the fits in Eqs. (5)-(6) would improve the manuscript.","section":"Fig. 5 caption"},{"comment":"The sentence 'The transition point, h_c, between these two phases is suspected to lie between h∼3.5−5 [42,69]' is in tension with the later use of h=3.25; the discrepancy should be acknowledged and discussed in the text.","section":"Model section, paragraph 2"}],"recommendation":"major_revision","confidential_remarks":"The paper's central claim rests on a single disorder strength and a single system size, with the critical field selected from the same data that is later used to demonstrate collapse. The authors should be encouraged to add an independent h_c estimate, error bars, collapse-quality metrics, and a finite-size scaling analysis for the bond quantities. The citation of the author's own software paper [73] for the transition point is misleading and should be replaced with a proper estimation study. These points are fixable within the scope of the manuscript, so major revision rather than rejection is recommended."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Dear colleague,\n\nThe short version: this is a genuinely new way to look at the MBL transition—using both block size and separation as independent length scales—and the reported collapse at h=3.25 is visually clean. But the central scale-invariance claim is not as solid as the abstract suggests, because the critical field is read off the same nearest-neighbor collapse that later testifies to scale invariance, and the whole demonstration is for L=20 only.\n\nWhat's new: the two-length-scale data collapse of normalized logarithmic negativity and mutual information as a function of d/l is a clever probe. The exponential-vs-power-law distinction between quantum and total correlations at the presumed critical point is also new and, if correct, would be a useful constraint on strong-disorder RG and tensor-network descriptions. The log scaling of the self entanglement in Fig. 5 is a nice supporting observation.\n\nWhere it gets soft: h_c=3.25 is inferred from the Fig. 3 collapse of the same normalized bond quantities that are then used in Fig. 4 to claim scale invariance. The contrast with h=1 and h=8 helps, but there is no independent estimate of h_c(L=20), no finite-size scaling of the Fig. 4 quantities, and no collapse-quality metric. The fits of Eq. (5) and (6) are reported without uncertainties or goodness-of-fit statistics. And since d ≤ L−l, the large-d/l tail of the collapse is carried by small blocks only, so the functional forms are not tested across all block sizes. The TNSLQ approximation for blocks with 2l>12 is an added caveat. No code or data is provided, which lowers reproducibility.\n\nEven with those caveats, the idea is solid and the data are suggestive. I'd like to see a revision that establishes h_c independently, shows collapse quality as a function of h, adds error bars and L=14,16,18 data for the Fig. 4 curves, and addresses the l-dependence of the tail. As is, it deserves a serious referee, but the strongest version of the claim is not yet quantitatively established.\n\nI'd bring it to our next reading group—the debate about what constitutes scale invariance at the MBL transition is worth having.","headline":"A clever two-length-scale probe of the MBL transition, but the scale-invariance claim rests on a self-selected critical point at a single system size.","tokens_in":12728,"tokens_out":4421,"would_cite":false,"duration_ms":45976,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["81P40","82B27","82B44"],"pacs":[],"model":"deepseek-v4-flash","headline":"At the many-body localization transition, the logarithmic negativity between two blocks decays exponentially with the ratio of separation to block size, while mutual information decays as a power law, showing that critical eigenstates are…","keywords":["many-body localization","scale invariance","logarithmic negativity","mutual information","entanglement transition","disordered spin chain","critical eigenstates","tensor networks"],"falsifier":"Compute the same quantities for a larger chain, say $L=24$ or $L=28$, at several disorder strengths around $h=3.25$, and check whether the data collapse persists at a disorder strength that is independently identified as critical (for example, from level statistics or other standard probes); if no choice of $h$ yields collapse for all block sizes at larger $L$, the scale-invariant exponential and power-law ansatzes are falsified.","tokens_in":11541,"feed_emoji":"⚛️","tokens_out":5738,"duration_ms":54452,"temperature":0.7,"pith_summary":"This paper claims that at the many-body localization (MBL) transition, the entanglement structure of critical eigenstates is scale invariant: the logarithmic negativity between two equal-sized blocks decays exponentially with the ratio of separation to block size, while the mutual information decays as a power law. The authors probe this by computing these quantities in a disordered spin chain for a range of block sizes and separations, and find that at disorder strength $h=3.25$ (the inferred transition point for $L=20$) the data for different block sizes collapses onto a single curve. The claim matters because scale invariance is a hallmark of a continuous phase transition, and identifying it in the entanglement structure constrains theories of the MBL transition and opens the door to tensor-network simulations of critical eigenstates.","feed_headline":"Scale-invariant entanglement found at MBL transition","feed_subtitle":"At critical disorder, quantum correlations depend only on block separation relative to block size.","key_machinery":"The central object is the pair of length scales, block size $l$ and separation $d$, combined into the normalized separation $\\tilde{d}=d/l$. By averaging the normalized logarithmic negativity $\\tilde{E}=E/l$ and mutual information $\\tilde{I}=I/l$ over pairs of blocks with the same $\\tilde{d}$, the authors test whether all dependence on $l$ disappears at the transition. The mechanism is the data collapse: at $h=3.25$ the curves for different $l$ converge onto a universal exponential for logarithmic negativity and a power law for mutual information, which is the signature of scale invariance.","core_discovery":"The central discovery is that the bond logarithmic negativity and the mutual information between disjoint blocks of equal size $l$, expressed as functions of the normalized separation $\\tilde{d}=d/l$, become independent of $l$ at the MBL transition point. Specifically, the logarithmic negativity decays exponentially as $\\tilde{E}_{\\tilde{d}}(l) = C_E e^{-\\tilde{d}/\\lambda_E}$ with fitted $C_E\\approx 2$ and $\\lambda_E\\approx 0.45$, while the mutual information decays as $\\tilde{I}_{\\tilde{d}}(l) = C_I \\tilde{d}^{-1/\\alpha_I}$ with fitted $C_I\\approx 0.3$ and $\\alpha_I\\approx 0.5$. This scale invariance is accompanied by a logarithmic growth of the average self-entanglement with block size at the transition, in contrast to linear growth in the ergodic phase and area-law (constant) behavior deep in the localized phase. The authors interpret this as evidence for a scale-invariant, multipartite entanglement structure in critical eigenstates.","pith_inferences":["A natural next step would be to check whether the scale-invariant collapse persists for larger system sizes than $L=20$; if it moves with $L$, the observed $h=3.25$ collapse might be a finite-size artifact rather than a true critical point.","The power-law decay of mutual information with exponent near $1/2$ could be related to a logarithmic growth of entanglement entropy when integrated over separations; testing the probability distribution of negativity instead of just the mean would further probe the multipartite structure.","The exponential decay length $\\lambda_E\\approx 0.45$ in units of $d/l$ might be interpretable as a critical correlation length in the ratio variable, which real-space renormalization group schemes should predict if they respect scale invariance."],"forward_implications":["If correct, critical eigenstates near the MBL transition have a scale-invariant entanglement structure, meaning any theory of the transition must reproduce the exponential decay of negativity and the polynomial decay of mutual information with normalized separation.","The observed collapse provides a parameter-free way to identify the transition point in finite systems, independent of the usual finite-size scaling with total system size.","The logarithmic growth of self-entanglement at the transition suggests a connection to conformal field theory or infinite-randomness fixed points, and may support a multiscale entanglement renormalization ansatz (MERA) description of critical eigenstates, enabling tensor-network simulations beyond exact diagonalization.","The distinction between the decay of quantum correlations (exponential) and total correlations (power law) at the same point implies that the critical state is not captured by simple quantum-classical equivalences, constraining strong-disorder renormalization group approaches."],"supporting_citations":[{"why":"Defines the logarithmic negativity used as the central quantum correlation measure.","marker":"[55]"},{"why":"Provides the TNSLQ method for computing logarithmic negativity beyond $2l>12$ and the previous identification of $h=3.25$ as the transition point.","marker":"[73]"},{"why":"Supplies the earlier estimate of the transition region and the scaling framework that this work builds on.","marker":"[42]"},{"why":"Together with [42], situates the suspected transition location between $h\\sim 3.5$ and $5$.","marker":"[69]"},{"why":"Establishes the monogamy bound $E_A \\ge E_{AB}$ used to compare self and bond entanglement.","marker":"[74]"},{"why":"Provides the contrasting result that at an infinite-randomness fixed point logarithmic negativity and mutual information scale identically, against which the exponential-versus-power-law distinction is highlighted.","marker":"[68]"},{"why":"Represents the strong-disorder renormalization group theory of the MBL transition that the observed scale invariance constrains.","marker":"[49]"},{"why":"MERA states are proposed as the natural description of scale-invariant critical eigenstates, extending efficient simulation into the critical region.","marker":"[77]"}],"fun_headline_variants":["Scale-invariant entanglement at MBL transition","Entanglement negativity reveals scale invariance at MBL","MBL critical point shows scale-free entanglement","Scale-invariant correlations at many-body localization transition","Critical eigenstates entangle scale-invariantly"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The transition point for $L=20$ is identified as $h=3.25$ from the same nearest-neighbor data collapse (Fig. 3) that is then used to demonstrate scale invariance in Fig. 4, so if the true critical point for that system size were different or shifts significantly with system size, the collapse could be coincidental.","fun_headline_variants_meta":{"raw":{"variants":["Scale-invariant entanglement at MBL transition","Entanglement negativity reveals scale invariance at MBL","MBL critical point shows scale-free entanglement","Scale-invariant correlations at many-body localization transition","Critical eigenstates entangle scale-invariantly"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000165,"raw_usage":{"total_tokens":1226,"prompt_tokens":895,"completion_tokens":331,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":511,"completion_tokens_details":{"reasoning_tokens":262}},"tokens_in":511,"tokens_out":331,"duration_ms":4242,"temperature":1.0,"reasoning_tokens":262,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T14:35:09.264549+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Compute the same quantities for a larger chain, say $L=24$ or $L=28$, at several disorder strengths around $h=3.25$, and check whether the data collapse persists at a disorder strength that is independently identified as critical (for example, from level statistics or other standard probes); if no choice of $h$ yields collapse for all block sizes at larger $L$, the scale-invariant exponential and power-law ansatzes are falsified.","supporting_citations":[{"cited_title":"Computable measure of entanglement,","cited_arxiv_id":null,"evidence_quote":"Defines the logarithmic negativity used as the central quantum correlation measure."},{"cited_title":"Fast Computation of Many-Body Entanglement","cited_arxiv_id":"1809.01685","evidence_quote":"Provides the TNSLQ method for computing logarithmic negativity beyond $2l>12$ and the previous identification of $h=3.25$ as the transition point."},{"cited_title":"Many-body localiza- tion transition: Schmidt gap, entanglement length, and scaling,","cited_arxiv_id":null,"evidence_quote":"Supplies the earlier estimate of the transition region and the scaling framework that this work builds on."},{"cited_title":"Many-body localization edge in the random-field Heisenberg chain,","cited_arxiv_id":null,"evidence_quote":"Together with [42], situates the suspected transition location between $h\\sim 3.5$ and $5$."},{"cited_title":"Dis- tributed entanglement,","cited_arxiv_id":null,"evidence_quote":"Establishes the monogamy bound $E_A \\ge E_{AB}$ used to compare self and bond entanglement."},{"cited_title":"Entanglement negativity in random spin chains,","cited_arxiv_id":null,"evidence_quote":"Provides the contrasting result that at an infinite-randomness fixed point logarithmic negativity and mutual information scale identically, against which the exponential-versus-power-law distinction is highlighted."},{"cited_title":"Analytically solvable renormalization group for the many-body local- ization transition,","cited_arxiv_id":null,"evidence_quote":"Represents the strong-disorder renormalization group theory of the MBL transition that the observed scale invariance constrains."},{"cited_title":"Class of quantum many-body states that can be efficiently simulated,","cited_arxiv_id":null,"evidence_quote":"MERA states are proposed as the natural description of scale-invariant critical eigenstates, extending efficient simulation into the critical region."}],"review_version":1}