{"id":"19a65f65-99e8-4863-8b40-c2b02307e836","arxiv_id":"2501.03076","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":4,"one_line_summary":"For two free-fermion lattices connected by a few quantum point contacts, the entanglement entropy of typical excited eigenstates grows only linearly with subsystem size, not extensively.","lead":"This paper computes how much quantum entanglement a small piece of a free-fermion system can hold when it talks to a reservoir through a few narrow contacts. It finds the entanglement grows only with the system's edge length, not its area, which means thermalization in such geometries looks very different from the usual expectation.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The central Eq. (6) scaling rests on Metropolis sampling at T=E; if the sampled ensemble is not microcanonical at the reported energy per particle E, the sub-extensive entropy result is an artifact.","rationale":"The reader's weakest assumption--that the Metropolis samples are typical eigenstates at energy E--is the right target. Eq. (6) is the central claim; it is not derived analytically, and all supporting figures (Figs. 5-8) are averages over 30 MMC-generated eigenstates. If the MMC ensemble is not representative of the microcanonical ensemble at the reported E, then the linear-in-L_A E scaling is not established. The text itself is ambiguous about whether the plotted E is the Metropolis temperature or the measured average energy, and no energy variances are given. For free fermions, canonical energy fluctuations are controlled by the specific heat; at the lowest reported E=0.05 and N~900, relative fluctuations are expected to be O(1), so the low-energy points may mix states with substantially different entanglement. This is a concrete correctness risk, not a stylistic objection.\n\nThe alpha inconsistency is real but secondary: Eq. (8) is labeled a conjecture, and even if it fails to interpolate to the ground state, Eq. (6) could still describe the finite-energy data. However, it is worth noting as corroborating evidence that the numerical characterization is not quantitatively consistent across limits.\n\nThe proposed test--direct microcanonical sampling in a narrow energy window--would settle the concern. If the microcanonical averages reproduce the MMC averages, then Eq. (6) survives this concern and the remaining issues are about precision and extrapolation. If not, the central claim should be downgraded. Since the reader already issued a CONDITIONAL verdict, no change to the verdict is needed; the concern is exactly the one on which that condition should hinge.","tokens_in":12283,"tokens_out":10294,"duration_ms":107020,"concrete_test":"Use the same single-particle Hamiltonian for LA=9,15,25 and LB=41, and sample Slater determinants directly within a narrow microcanonical energy window (e.g., relative width +/-5%) centered on each reported MMC energy E. Compute S_A from correlation matrices for 30 such states and compare the average to the MMC average from the paper. Also report the standard deviation of the MMC energy distribution; if sigma_E/E > 0.1 or the microcanonical average differs from the MMC average by more than the standard error, the Eq. (6) slope is not established.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central numerical claim, Eq. (6), is established only through the Metropolis Monte-Carlo procedure of Sec. III. The paper asserts that the 30 sampled eigenstates are 'typical' excited states at energy per particle E, but the sampler is canonical at temperature T=E, not microcanonical at fixed E. For noninteracting fermions, the many-body density of states is strongly energy dependent, and the energy fluctuations of a canonical sample at T=0.05 for N~900 are not negligible compared to E; states with significantly higher energy can dominate the averaged entanglement entropy. Because S_A increases with E, any such contamination inflates the slope alpha in Eq. (6) and could produce the apparent L_A E scaling without reflecting a typical eigenstate at energy E. The absence of reported energy variances, autocorrelation times, or comparisons to microcanonical sampling makes this the load-bearing assumption. A secondary internal inconsistency--the per-QPC ground-state coefficient b~0.238 in Eq. (4) disagrees with the finite-energy alpha~0.6 in Eq. (8), so the proposed interpolation cannot accommodate both limits--reinforces that the scaling law is not yet quantitatively secure.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The manuscript studies the bipartite entanglement entropy of noninteracting fermions on two 2D square lattices coupled by a small number of quantum point contacts (QPCs). In the ground state, each QPC contributes a logarithmic term b log L_A (Eq. (4)). For finite energy per particle E, the paper claims that a typical excited eigenstate has entanglement entropy S_A = a L_A + alpha m L_A E for small m and low E (Eq. (6)), so the entropy is sub-extensive in the linear size L_A and each additional QPC contributes an entropy proportional to L_A E. The authors also conjecture an interpolating form, Eq. (8), Delta S_A = alpha log[(1/E) sinh(L_A E)], which would connect the ground-state logarithmic behavior to the finite-energy linear behavior. The numerical evidence is based on Metropolis Monte-Carlo generation of 30 eigenstates per parameter set, followed by averaging of the entanglement entropy.","tokens_in":1749,"tokens_out":1908,"duration_ms":56047,"significance":"If the central claim holds, the result is genuinely interesting: it would demonstrate that the geometry of the coupling between a subsystem and a reservoir can change the scaling of entanglement entropy in an excited eigenstate, in contrast to the usual extensive volume-law expectations for ETH-satisfying systems. The paper leverages established free-fermion techniques and is clearly presented. The strongest part is the numerical demonstration that for small numbers of QPCs the entropy grows with L_A rather than L_A^2, which is a concrete and falsifiable prediction. However, the numerical evidence is weakened by the sampling methodology and by an internal inconsistency between the ground-state and finite-energy coefficients; these issues must be addressed before the claims can be accepted.","major_comments":[{"comment":"The central claim, Eq. (6), is about typical excited eigenstates at energy per particle E, but the numerical states are generated by a Metropolis Monte-Carlo procedure sampling at temperature T = E. This is a canonical ensemble, not a microcanonical one. For noninteracting fermions with a strongly energy-dependent many-body density of states, the energy fluctuations of a canonical sample can be significant, and because the entanglement entropy increases with energy, the averaged entropy may be dominated by higher-energy states. The paper does not report the variance of the sampled energy, the acceptance statistics, or a comparison of the canonical average with the microcanonical average at the same mean energy. Without such a validation, the numerical data do not directly test the stated eigenstate typicality assumption. I request either microcanonical (fixed-energy) sampling, exact diagonalization on small lattices, or at least a quantitative demonstration that the canonical distribution at T = E yields the same mean entropy as the typical eigenstate at energy E.","section":"Sec. III, Numerical Methods"},{"comment":"There is a load-bearing internal inconsistency between the ground-state coefficient b in Eq. (4) and the finite-energy coefficient alpha in Eq. (8). In the limit L_A E -> 0, Eq. (8) reduces to alpha log(L_A), so Eq. (8) can only accommodate the ground-state term m b log L_A if alpha = b. The paper reports b approximately 0.238 (Fig. 4) and alpha approximately 0.6 (Fig. 8), a factor of about 2.5 difference. The claim that Eq. (8) accommodates the second term of both equations (4) and (6) is therefore not correct as stated. The authors should either reconcile the two coefficients, show that the universal curve has a crossover with an effective alpha that changes with L_A E, or restrict the interpolation claim to a regime where the discrepancy is explained.","section":"Sec. II, Eq. (8) and associated text"},{"comment":"The slopes Delta S_A that form the basis for Eq. (5) and Fig. 8 are computed from only the first four data points (first three for E = 1.0), with no justification for this truncation beyond the onset of saturation. Because the saturation effect is present even at low E, the fitted slope alpha is sensitive to the number of points included and to the chosen energy and lattice sizes. No error bars are given for any averaged entropy or for the fitted slopes, despite the data coming from only 30 Monte-Carlo-generated eigenstates per point. The universal scaling claim would be much stronger if the fit were done with a full functional form including saturation, or at least with a systematic check that the slopes are stable as the fit range is varied, together with some estimate of statistical uncertainty.","section":"Sec. II, Figs. 5-8 and surrounding text"},{"comment":"The abstract states as fact that it is shown that the entropy scales as S_A ~ L_A E, but the body repeatedly qualifies the results as approximate and calls Eq. (8) a conjecture. More importantly, the claim in the abstract that the entropy is sub-extensive is only demonstrated numerically for small m and low E; the crossover to the ground-state logarithmic behavior is not observed because of computational difficulty, as the authors acknowledge in Sec. IV. The manuscript should clearly separate the numerically demonstrated scaling of Eq. (6) from the conjectured interpolation of Eq. (8), and the abstract should be phrased accordingly. This is a presentation issue, but it affects the reader's ability to judge what is established.","section":"Abstract and Sec. IV, Discussion and Conclusion"}],"minor_comments":[{"comment":"There is a typographical error in the definition of the energy: 'E = (E_ex - E0)/N where is the many-fermion ground state energy' is missing the symbol E0 before 'is'. This should read 'where E0 is the many-fermion ground state energy'.","section":"Sec. III, Numerical Methods"},{"comment":"The phrase 'we refer to the average energy E following this procedure as the MMC energy' is confusing because E was already introduced as the excited-state energy per particle. It would be clearer to define a separate symbol for the ensemble-averaged energy, for example E_bar, and to state explicitly whether the reported values of E in Figs. 5-8 are the desired target energies or the measured ensemble averages.","section":"Sec. III, Numerical Methods"},{"comment":"The caption states that entropy for finite energy states shows proportionality to L_A and that for comparison the ground state entropy is proportional to log L_A, without mentioning that the finite-energy data are averaged over 30 Monte-Carlo states. The reader should be reminded of this in the caption.","section":"Sec. II, Fig. 2 caption"},{"comment":"The notation in Eq. (8) uses log without specifying the base; since the rest of the paper uses natural log (as in log L_A for the ground state), it would be helpful to state that all logarithms are natural.","section":"Sec. II, Eq. (8)"}],"recommendation":"major_revision","confidential_remarks":"The paper addresses a timely and interesting question, and the numerical observation of sub-extensive entropy for small QPC number is worth pursuing. However, the sampling procedure is the main technical weakness: the paper implicitly equates a canonical Monte-Carlo average at T = E with a microcanonical typicality statement, and this is not trivial for noninteracting fermions with a rapidly growing density of states. The internal inconsistency between b and alpha also needs to be resolved. With a revised numerical analysis including error bars and a microcanonical check, plus a clearer separation of demonstrated results from conjectures, the paper could be suitable for publication. I recommend major revision rather than rejection."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"New here: the paper shows numerically that for two 2D free-fermion lattices coupled by m QPCs, the entanglement entropy of a small subsystem in a typical excited eigenstate grows as S_A = a L_A + alpha m L_A E, sub-extensive for fixed m. The collapse of the per-QPC entropy onto a 1D finite-temperature scaling form (Fig. 8) is a fresh observation and ties the problem to known CFT results for gapless 1D systems. The ground-state log-L_A per QPC behavior is also reproduced. To their credit, the authors explicitly label Eq. (8) as a conjecture and admit they cannot reach the LAE << 1 crossover numerically.\n\nThe soft spots are real but not fatal. The Metropolis sampling at T = E is not the same as fixing E; however, for these parameters (N ~ 10^3, T = 0.05) the canonical energy fluctuations should be a small fraction of the total energy, so I suspect the bias is mild. The paper should still report the energy variance and ideally do a microcanonical check, because the claim is specifically about typical eigenstates at fixed energy. More serious is the coefficient mismatch: the fitted alpha ~ 0.6 does not reduce to the measured ground-state b ~ 0.238 when Eq. (8) is evaluated at small LAE. That means the proposed interpolation is quantitatively inconsistent with the E = 0 limit the authors want it to connect to. Also, all slopes come from the first three or four m values with no error bars, and the abstract says 'shown' where the body later says 'conjecture.'\n\nIf the scaling holds, it is an interesting counterexample to the usual ETH intuition that geometry only affects relaxation times, not equilibrium entanglement. The audience is researchers working on ETH in free-fermion systems and on geometry dependence of entanglement. I would send it to peer review; a serious referee should push for the microcanonical comparison, error bars, and a resolution of the alpha/b discrepancy. The main observation is plausible enough to deserve that scrutiny.","headline":"Sub-extensive L_A E scaling in QPC-coupled fermions is new and plausible, but the evidence needs error bars, a microcanonical check, and the alpha/b mismatch resolved before I fully trust it.","tokens_in":13073,"tokens_out":4199,"would_cite":false,"duration_ms":41531,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":["03.65.Ud","05.30.-d","05.30.Fk"],"model":"deepseek-v4-flash","headline":"The paper claims that a small number of quantum point contacts makes the entanglement entropy of typical excited free-fermion eigenstates sub-extensive, $S_A = a L_A + \\alpha m L_A E$, not extensive.","keywords":["eigenstate thermalization hypothesis","free fermions","quantum point contact","entanglement entropy","sub-extensive scaling","conformal field theory scaling","Metropolis Monte Carlo eigenstate sampling","anomalous area law"],"falsifier":"A concrete check is to diagonalize the Hamiltonian of Eq. (1) exactly for $L_A=9$, $L_B=41$, and $m=1,2,3$, select eigenstates with energy per particle $E$ near 0.05, compute $S_A$ from the correlation matrix, and verify that the slope per contact is $\\alpha L_A E$; a mismatch would show that the Metropolis Monte-Carlo ensemble is not representative of typical eigenstates.","tokens_in":12091,"feed_emoji":"⚛️","tokens_out":12010,"duration_ms":106330,"temperature":0.7,"pith_summary":"The paper asks whether a free-fermion system that satisfies ETH through entanglement alone still has thermal, extensive entropy when its reservoir is connected through only a small number of quantum point contacts. It argues that it does not: in a typical excited eigenstate of two 2D free-fermion lattices joined by $m$ contacts, the entanglement entropy of the smaller subsystem is $S_A = a L_A + \\alpha m L_A E$, linear in the subsystem's linear size rather than its area. This makes the finite-energy behavior the counterpart of the ground-state result in which each contact contributes $\\sim \\log L_A$; the paper conjectures that one formula, $\\Delta S_A = \\alpha \\log[(1/E)\\sinh(L_A E)]$, interpolates between the two. If correct, the result shows that restricting the boundary between a subsystem and its reservoir can suppress the entropy of an eigenstate that would otherwise look thermal.","feed_headline":"Few quantum contacts keep free-fermion entropy sub-extensive","feed_subtitle":"Each added contact contributes entropy proportional to subsystem size times energy, a one-dimensional signature.","key_machinery":"The key object is the s-wave scattering channel formed by each quantum point contact: the in-going and out-going fermion modes on the two lattices are combined into a single spinor with periodic boundary conditions, and bosonized into one right-moving boson on an interval of length $L_A$. Each well-separated contact therefore acts as an independent 1D gapless entanglement channel, giving a per-contact entropy $\\sim \\log L_A$ in the ground state and $\\sim L_A E$ at finite energy. The dimensionless product $L_A E$ plays the role of the aspect ratio of space to imaginary time in a 1D conformal field theory, and the conjectured Eq. (8) is the interpolation formula that collapses the per-contact increments onto one universal curve.","core_discovery":"The central claim is Eq. (6): for a typical excited eigenstate of two 2D free-fermion lattices connected by $m$ quantum point contacts, the entanglement entropy of the smaller subsystem is $S_A = a L_A + \\alpha m L_A E$ in the regime of small $m$ and low $E$. The first term allows for ground-state degeneracy entropy; the second says each contact contributes the finite-energy entanglement entropy of a gapless 1D system, $\\sim \\alpha L_A E$. The paper also conjectures the full interpolation $\\Delta S_A = \\alpha \\log[(1/E)\\sinh(L_A E)]$ per contact, which reduces to $\\log L_A$ as $E\\to 0$ and to $L_A E$ at larger $L_A E$. The authors contrast this sub-extensive entropy with classical ergodic and quantum chaotic expectations, where boundary geometry affects relaxation times but not extensive equilibrium entropy.","pith_inferences":["Beyond the paper, a quench experiment that starts with uncoupled lattices and then switches on the QPCs would test whether the growth of $S_A(t)$ is controlled by contact geometry rather than by fast internal thermalization.","Beyond the paper, if Eq. (8) is the correct interpolation, the per-contact entropy increment should be independent of the detailed shape of each contact as long as contacts are separated by more than the inverse Fermi momentum $k_F^{-1}$; varying contact spacing in the numerics would test this independence.","Beyond the paper, an exact microcanonical computation at fixed energy per particle on small lattices, rather than Metropolis Monte-Carlo sampling at $T=E$, would remove the main source of sampling bias and place the formula on firmer footing.","Beyond the paper, in any realistic solid phonon and radiative coupling add many parallel entanglement channels, so the clean sub-extensive signature would likely be masked; the paper notes this, and the natural experimental target would be a cold-atom or mesoscopic device where QPCs are the only coupling between the two systems."],"forward_implications":["For small $m$ and small $E$, the entanglement entropy of the smaller lattice scales as $L_A$, not $L_A^2$, so the reduced state is less mixed than the volume-law and thermal expectations.","Each additional well-separated contact contributes an independent entropy increment: $\\sim b \\log L_A$ at zero energy and $\\sim \\alpha L_A E$ at finite energy, with the crossover controlled by $L_A E$.","Saturation occurs once $\\alpha m L_A E$ reaches the extensive value $L_A^2 E_{\\rm sat}$, giving a contact-number threshold $m_{\\rm sat}\\sim 0.5 L_A/E$; beyond it, the QPC restriction no longer limits entropy.","In the special case $L_A=L_B$, a single QPC can already produce extensive entropy because of the near-degenerate spectrum, so the sub-extensive rule applies specifically to a smaller subsystem coupled to a larger bath.","The sub-extensive entropy result implies that boundary geometry can alter an equilibrium property of an eigenstate, not just relaxation times, for free-fermion systems satisfying ETH."],"supporting_citations":[{"why":"Establishes that a small subsystem of a noninteracting-fermion eigenstate has an approximately thermal reduced density matrix, the ETH premise the paper tests in a QPC geometry.","marker":"[1]"},{"why":"Introduced the zero-dimensional area law for a single point contact in a gapless fermionic system, supplying the $\\log L_A$ single-contact ground-state entropy.","marker":"[8]"},{"why":"Confirmed the ground-state scaling $S_A = a L_A + b m \\log L_A$ for two 2D lattices coupled by multiple QPCs, the baseline this paper extends to finite energy.","marker":"[9]"},{"why":"Provides the weak-coupling bosonized model of a boundary impurity whose entanglement grows as $\\log L$, the basis for treating each QPC as a 1D channel.","marker":"[10]"},{"why":"Shows numerically that a defect in a 1D gapless fermion system produces $\\log L$ entanglement with a coupling-dependent prefactor, supporting the per-contact increment.","marker":"[12]"},{"why":"Supplies the crossover scaling in the dimensionless ratio $L_A E$ used to collapse the finite-energy data and to motivate the interpolation formula Eq. (8).","marker":"[15]"},{"why":"Gives the correlation-function method used to compute entanglement entropies from free-fermion eigenstates.","marker":"[26]"},{"why":"Derives the anomalous area law for fermion entanglement, the sub-extensive ground-state behavior that multiple QPC channels build up.","marker":"[13]"}],"fun_headline_variants":["Free-fermion entropy stays sub-extensive via few QPCs","Each QPC adds 1D entropy scaling: S_A ~ L_A E","Quantum contacts keep free-fermion entanglement sub-extensive","Few QPCs give free-fermion entropy a 1D thermodynamic shape","Sub-extensive entropy: quantum contacts limit free-fermion entanglement"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is that eigenstates generated by Metropolis Monte-Carlo sampling at temperature $T=E$ are representative of typical excited eigenstates with energy per particle $E$; if the sampling is biased, the reported sub-extensive scaling would be an artifact.","fun_headline_variants_meta":{"raw":{"variants":["Free-fermion entropy stays sub-extensive via few QPCs","Each QPC adds 1D entropy scaling: S_A ~ L_A E","Quantum contacts keep free-fermion entanglement sub-extensive","Few QPCs give free-fermion entropy a 1D thermodynamic shape","Sub-extensive entropy: quantum contacts limit free-fermion entanglement"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000251,"raw_usage":{"total_tokens":1607,"prompt_tokens":1044,"completion_tokens":563,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":660,"completion_tokens_details":{"reasoning_tokens":467}},"tokens_in":660,"tokens_out":563,"duration_ms":5645,"temperature":1.0,"reasoning_tokens":467,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-10T21:53:29.433345+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"A concrete check is to diagonalize the Hamiltonian of Eq. (1) exactly for $L_A=9$, $L_B=41$, and $m=1,2,3$, select eigenstates with energy per particle $E$ near 0.05, compute $S_A$ from the correlation matrix, and verify that the slope per contact is $\\alpha L_A E$; a mismatch would show that the Metropolis Monte-Carlo ensemble is not representative of typical eigenstates.","supporting_citations":[{"cited_title":"Entanglement Entropy Scaling Laws and Eigenstate Typicality in Free Fermion Systems","cited_arxiv_id":"1409.1224","evidence_quote":"Establishes that a small subsystem of a noninteracting-fermion eigenstate has an approximately thermal reduced density matrix, the ETH premise the paper tests in a QPC geometry."},{"cited_title":"Scaling of Entanglement Entropy in Point Contact Free Fermion Systems","cited_arxiv_id":"1402.5437","evidence_quote":"Confirmed the ground-state scaling $S_A = a L_A + b m \\log L_A$ for two 2D lattices coupled by multiple QPCs, the baseline this paper extends to finite energy."},{"cited_title":"Zero dimensional area law in a gapless fermion system","cited_arxiv_id":"0711.0957","evidence_quote":"Provides the weak-coupling bosonized model of a boundary impurity whose entanglement grows as $\\log L$, the basis for treating each QPC as a 1D channel."},{"cited_title":"Entanglement entropy in a boundary impurity model","cited_arxiv_id":"cond-mat/0408366","evidence_quote":"Shows numerically that a defect in a 1D gapless fermion system produces $\\log L$ entanglement with a coupling-dependent prefactor, supporting the per-contact increment."},{"cited_title":"Eigenstate entanglement: Crossover from the ground state to volume laws","cited_arxiv_id":"1905.07760","evidence_quote":"Supplies the crossover scaling in the dimensionless ratio $L_A E$ used to collapse the finite-energy data and to motivate the interpolation formula Eq. (8)."},{"cited_title":"”Entanglement Entropy of Fermions in Any Dimension and the Widom Conjec- ture,” Phys","cited_arxiv_id":null,"evidence_quote":"Derives the anomalous area law for fermion entanglement, the sub-extensive ground-state behavior that multiple QPC channels build up."}],"review_version":1}