{"id":"76ebe305-1070-4b66-a0ca-ccb68c5d33fe","arxiv_id":"2509.03514","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":4,"one_line_summary":"On a Rydberg-atom Sierpiński gasket, the ground state gains spin flips one at a time and the extra spin is delocalized over equivalent sublattice sites, matching exact-diagonalization predictions.","lead":"Researchers trapped 88Sr atoms in a Sierpiński gasket fractal and studied a quantum Ising model with long-range interactions. They found phases where single spins flip one by one, with the flipped spin spread across equivalent sites, which may help control quantum states.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The delocalized-superposition claim is not certified by the data: the symmetry-breaking disorder/defect model used to fit the experiment is strong enough to localize the extra spin, and only populations are measured.","rationale":"The reader's weakest assumption identifies the threefold symmetry of the ideal gasket as the load-bearing element; our concern is the same issue sharpened: the perturbations required to fit the experimental data are so large that they are likely to destroy the superposition, and the experiment measures only populations, leaving the coherence entirely unverified. Thus the central claim—delocalized single spin-flip superposition—is not supported by the presented evidence. This does not invalidate the paper's other contributions (exact diagonalization of the ideal model, the observed cascade of number states, the extensive experimental characterization), but it means the headline superposition claim is conditional on additional coherence evidence or a suitably weakened interpretation. The reader's CONDITIONAL verdict remains appropriate; therefore no verdict adjustment is needed.","tokens_in":20541,"tokens_out":9982,"duration_ms":109920,"concrete_test":"Exactly diagonalize the 9-site Hamiltonian with the calibrated disorder+defect model (1% Gaussian positional disorder, site 4 displaced outward by 10% of the nearest-neighbor distance) at the ⟨n↑⟩=4 operating point (V = 110 Ω, Δ/Ω = 4.1). Compute the single-particle density matrix restricted to the three edge sites and evaluate (i) the fidelity to the ideal W-state (|e1⟩+|e2⟩+|e3⟩)/√3 and (ii) the sum of off-diagonal coherences. If the fidelity is close to 1/3 (the value for a localized state) or the coherences are below ~0.05, the disorder model used to reproduce the data does not support a delocalized superposition; conversely, if the fidelity remains high, the concern is dispelled.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The headline claim is that the fourth and fifth spin flips form delocalized superpositions over equivalent edge sites. This is proven by ED only for the ideal threefold-symmetric 9-site Hamiltonian. The experimental confirmation, however, consists exclusively of site-resolved excitation probabilities, which cannot distinguish a coherent superposition from a classical mixture or a localized state. The paper itself reports that the raw data break the theoretical symmetry and require a model with 1% Gaussian position disorder plus a 10% outward displacement of site 4 to reproduce the measurements (Methods, 'Disorder and defects'; Extended Data Fig. 4). These are not small perturbations: for van der Waals interactions V ∝ r^{-6}, a 10% displacement changes the nearest-neighbor interaction by ~60%, i.e., by ~66 Ω at V = 110 Ω. This far exceeds the transverse field Ω that creates the superposition, so the ground state in the four-spin-up manifold should localize on the lowest-energy edge site; the off-diagonal coherences of the single-particle density matrix are suppressed. The statement that 'the additional spin up remains in a superposition and the quantum state is not destroyed' (main text) is not supported by any reported coherence measurement, and the strong symmetry-breaking terms needed to fit the data make it doubtful. If the realized state is localized or a mixture, the central claim of a delocalized single spin-flip superposition over all equivalent edge sites is not established.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies the transverse-field Ising model with long-range van der Waals interactions on a first-generation Sierpiński gasket (9 sites) and higher generations, combining exact diagonalization, variational mean-field, quantum Monte Carlo, and a symmetry-based SIM-GRAPH method. The central claim is that, in the quantum regime, the ground state exhibits phases in which spin flips occur one by one and the added spin is not localized on a single site but delocalized in a superposition over equivalent sublattice sites (e.g., four spin-ups with the fourth spin coherently spread over three edge sites). The authors support this with exact-diagonalization phase diagrams and with experimental measurements on 88Sr atoms in optical tweezers, reporting good agreement after including positional disorder and a site-4 defect in the theoretical model.","tokens_in":20907,"tokens_out":3213,"duration_ms":35723,"significance":"If established, the result would be significant: it would demonstrate that fractal geometry combined with long-range interactions can produce stable, sublattice-selective many-body states with one-by-one spin-flip cascades, and that such states are accessible in a Rydberg-atom simulator. The paper has notable strengths: exact diagonalization on the 9-site system is exact; the second-generation results are cross-checked with VMF and QMC; the experimental methods are described in exceptional detail, including laser noise characterization, pulse-shape calibration, and SPAM corrections; and the code and data availability statements are explicit. However, the central 'delocalized superposition' claim is not directly certified by the experimental data, and the paper's own disorder/defect model introduces symmetry-breaking terms that are large enough to threaten the superposition interpretation.","major_comments":[{"comment":"The statement that 'the additional spin up remains in a superposition and the quantum state is not destroyed' is not supported by the reported measurements. The experimental confirmation consists exclusively of site-resolved excitation probabilities P_i[↑]. Such populations cannot distinguish a coherent superposition from a classical mixture or from a localized state. The theoretical superposition is a property of the ED ground state of the ideal 9-site Hamiltonian, but no coherence witness (e.g., parity oscillations, off-diagonal density-matrix elements, or entanglement entropy extracted from the many-body state) is measured. The manuscript should either explicitly restrict the superposition claim to the ideal theoretical model and describe the experimental confirmation as population-level agreement, or provide a coherence-sensitive measurement or a quantitative estimate of the coherenc","section":"Main text, 'Comparison between theory and experiments'"},{"comment":"The model used to reproduce the experimental asymmetry includes 1% Gaussian positional disorder and a 10% outward displacement of site 4. For V = 110 Ω and V ∝ r^{-6}, a 10% displacement changes the nearest-neighbor interaction by roughly 60%, i.e., about 66 Ω, far exceeding the transverse field Ω that generates the superposition. This is not a perturbative symmetry breaking. The paper asserts that the superposition 'remains robust, with an asymmetric amplitude,' but no calculation of the one-body density matrix coherences (or any other off-diagonal observable) under this disorder/defect model is provided. Without such a calculation, the claim that the experimentally realized state retains coherent delocalization rather than localizing or becoming a mixture is unsubstantiated. The authors should compute and report the off-diagonal elements of the reduced single-particle density matrix in","section":"Methods, 'Disorder and defects'; Extended Data Fig. 4"},{"comment":"The paper's own benchmark shows that SIM-GRAPH fails to identify the 4- and 5-spin-up superposition phases for the first-generation gasket, compressing them into a narrow transition region (Extended Data Fig. 3b). For the second-generation system, the phase diagram in Fig. 3a is computed mainly with SIM-GRAPH, and the text claims that the ⟨n↑⟩=7 state contains 'four spins in a superposition on twelve sites along the inner edge.' Since SIM-GRAPH's projection assumes symmetric sites are occupied equally and is acknowledged to miss superposition states at phase boundaries, the higher-generation delocalization claim rests on a method whose central limitation is the very phenomenon being claimed. The authors should provide an ED or QMC calculation of the relevant off-diagonal correlations at the specific parameters for the second-generation ⟨n↑⟩=7 state, or clearly label this prediction as a","section":"Methods, 'SIM-GRAPH'; Extended Data Fig. 3"},{"comment":"The text states that the ⟨n↑⟩=12 plateau predicted by SIM-GRAPH is 'metastable' and that the other methods show the true ground state is missed by SIM-GRAPH. This is an honest and important caveat, but it also means that SIM-GRAPH can produce a stable phase that is not the ground state. This strengthens the need for the previous comment: for every higher-generation superposition claim not independently checked with ED/QMC, the possibility of a similar artifact must be explicitly addressed, not only for the ⟨n↑⟩=12 plateau but for the ⟨n↑⟩=7 state as well.","section":"Main text, 'Higher-generation fractals'"}],"minor_comments":[{"comment":"The magnetization is defined as ⟨m⟩ = (⟨n↑⟩ − ⟨n↓⟩)/N = (2⟨n↑⟩ − N)/N, but in several places the text refers to '⟨n↑⟩ = 4,5' as if it were the number of spin-ups. This is clear from context but should be made consistent, e.g., by defining n↑ explicitly as the total number of Rydberg excitations.","section":"General notation"},{"comment":"The caption lists parameter values 'Δ/Ω = -4.5, 4.1, 10, 15, 21.8 with V = 110 Ω' but the first value is negative and corresponds to the zero-spin-up phase. Since the text says the first cross is outside the shown phase diagram, a sentence clarifying the negative-detuning point would help the reader.","section":"Fig. 2 caption"},{"comment":"The notation P⃗ = (P1, P0) and M = ((1−εp, εn),(εp, 1−εn)) is slightly confusing because the order of the vector entries is not defined in the text. Please specify that P1 corresponds to the detected 'Rydberg-excited' (absent) outcome and P0 to the 'ground' (present) outcome, or vice versa.","section":"Methods, 'SPAM Correction Procedure'"},{"comment":"The phrase 'unprecedented control of a cascade of phase transitions' is strong. Given the limitations on the superposition claim discussed above, the authors may wish to temper 'control' to 'access' or 'observe' in the abstract, or to clearly distinguish the theoretical cascade from the experimentally demonstrated population changes.","section":"Abstract and main text"},{"comment":"Reference [22] is a footnote-like comment on the Cayley tree; it should be moved to a proper footnote or integrated into the main text, as numbered footnotes in the reference list are unconventional.","section":"References"}],"recommendation":"major_revision","confidential_remarks":"The paper is potentially interesting and the experimental effort is impressive, but the central novelty—the delocalized quantum superposition of a single spin-flip—is not experimentally demonstrated, and the paper's own disorder/defect model introduces perturbations that, on a simple estimate, are large enough to localize the state. The authors should be asked to either provide direct evidence of coherence (e.g., an off-diagonal observable or an entanglement witness) or substantially restate the claims as 'delocalized population in a symmetric model' rather than 'superposition in the realized experiment.' The SIM-GRAPH limitations also need to be more carefully separated from the higher-generation claims."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"This paper is a real step forward for Rydberg quantum simulation on non-integer-dimensional lattices, and the sublattice-resolved one-by-one spin-flip cascade is a genuinely new observation. But the headline claim that the fourth and fifth excitations form delocalized superpositions is not established by the experimental data; the measurements are populations only, and the symmetry-breaking disorder/defect model needed to reproduce them is strong enough to localize the extra spin.\n\nThe ED on 9 sites is exact and the 24-site comparisons against VMF and QMC are honest. The experimental methods section is exemplary: laser noise, pulse shapes, SPAM corrections, and an independent C6 calibration from two-atom sweeps. The difference between classical and quantum phase diagrams is clean, and the cascade from 0 to 9 excitations is convincing.\n\nMain soft spot: coherence. Site-resolved probabilities cannot distinguish a coherent superposition from a classical mixture or a localized state. The paper reports that the data break the threefold symmetry and require 1% Gaussian position disorder plus a 10% outward displacement of site 4 (Methods). For a van der Waals interaction scaling as r^-6, a 10% displacement changes the nearest-neighbor interaction by roughly 60%, larger than the transverse field Omega. That is a large perturbation; it plausibly localizes the extra spin on the lowest-energy edge site. The sentence 'the additional spin up remains in a superposition and the quantum state is not destroyed' is therefore not backed by a coherence measurement or an entanglement witness. Second, SIM-GRAPH misses exactly the 4- and 5-spin-flip superposition states on the first-generation gasket (Extended Data Fig. 3), which is worrying because it is then used to predict those same phases in higher generations. Third, the 'excellent agreement' is weakened by post hoc disorder/defect parameters; the main phase diagrams show no statistical uncertainties. These are addressable, not fatal.\n\nThis is a worthwhile paper for the AMO and quantum-simulation community. The experiment is careful and the numerics are mostly solid. The central physical phenomenon—sublattice-resolved spin-flip cascade—is likely real; the superposition part needs more evidence or a softer claim. I would send it to peer review and ask the authors to either measure coherence (e.g., via a parity or entanglement witness) or explicitly state that the delocalization is a theoretical prediction for the ideal Hamiltonian, not a demonstrated experimental result.","headline":"Sublattice-resolved spin-flip cascade in a Rydberg fractal is real and nicely characterized, but the delocalized-superposition headline outruns the data: only populations are measured, and the symmetry-breaking disorder model strong enough to fit the experiment would localize the extra spin.","tokens_in":21421,"tokens_out":2572,"would_cite":true,"duration_ms":28204,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":["67.85.-d"],"model":"deepseek-v4-flash","headline":"On a Sierpiński-gasket array of Rydberg atoms, excitations appear one spin flip at a time, each added spin delocalized over all equivalent sites of a sublattice — a control regular lattices cannot offer.","keywords":["Sierpiński gasket","Rydberg atoms","transverse-field Ising model","long-range van der Waals interactions","single spin-flip control","optical tweezers","quantum simulation","entanglement entropy"],"falsifier":"Image single experimental shots in the 4-spin-up plateau and tally which edge site carries the fourth excitation: the superposition claim requires the fourth spin to appear on all three edge sites with symmetric probabilities. The same apparatus can test the symmetry dependence directly — displacing one corner outward by more than about 10% of the lattice spacing should localize the fourth spin and narrow or eliminate the plateau, exactly as the paper's disorder model predicts.","tokens_in":20461,"feed_emoji":"⚛️","tokens_out":11048,"duration_ms":95785,"temperature":0.7,"pith_summary":"This paper sets out to show that a fractal lattice changes how a many-body spin system passes through its phases. On a Sierpiński gasket — a triangle of triangles with Hausdorff dimension 1.58 — Rydberg atoms interacting through long-range van der Waals forces flip their spins one by one as the detuning is raised, rather than jumping between ordered states. The new feature is where each added spin goes: it delocalizes into a superposition over all equivalent sites of one sublattice (corners first, then edge, then bulk sites), so a single collective excitation can be addressed as a sublattice property. The authors support this with exact diagonalization, variational mean field, quantum Monte Carlo, a new symmetry-reduction solver, and direct imaging of 88Sr atoms arranged in the gasket, and argue that the effect is absent in regular lattices, where all sites are equivalent. If the claim holds, the fractal's geometry offers stepwise experimental control of a cascade of single-spin-flip phase transitions.","feed_headline":"A Rydberg fractal flips its spins one by one","feed_subtitle":"Each added spin spreads across an entire sublattice, giving stepwise control of a many-body cascade.","key_machinery":"The load-bearing structure is the Sierpiński gasket's sublattice hierarchy, defined by connectivity: corner sites have one neighbor, edge sites two, bulk sites three, and the sublattices hold unequal numbers of sites. The long-range van der Waals interaction V(r)=C6/r6 orders the filling by sublattice — corners first to minimize repulsion, then edge, then bulk — while the transverse field Ω turns the energetically degenerate classical configurations into symmetric superpositions over the equivalent sites of the active sublattice. The Rydberg blockade radius Rb=a(V/Ω)^{1/6} sets the competition between interaction and spin-flip drive. The paper's new solver, SIM-GRAPH, computes the automorphi","core_discovery":"In the quantum regime, the ground state of a long-range-interacting Ising model on a first-generation Sierpiński gasket has stable phases with exactly four and five spin-up excitations, the added spins being delocalized over all equivalent edge sites. Classically the fourth spin sits on one random edge site; with a transverse field it spreads into a symmetric superposition over the three edge sites, and the five-spin-up phase carries two extra spins in superposition. Magnetization, spin susceptibility, entanglement entropy, and real- and reciprocal-space correlations all show the same one-by-one cascade, and experiments with single 88Sr atoms in optical tweezers reproduce the predicted phase","pith_inferences":["The superposition's fragility is itself a tool: because 1% position disorder measurably localizes the fourth spin, single-shot statistics over edge sites could serve as an in-situ calibration of array symmetry — a metrology use the paper does not state.","If the sublattice-filling mechanism is generic, other fractals with imbalanced sublattices, such as the Sierpiński carpet or the dual triangular gasket, should show analogous one-by-one cascades; a direct numerical run would settle whether the corner-edge-bulk ordering is universal.","Should the transverse-field criticality on the gasket admit a free-fermion description, the one-by-one spin cascade would be the occupation-space staircase of a single-particle spectrum — a concrete route toward the Majorana connection the authors list only as an outlook."],"forward_implications":["The 4- and 5-spin-up phases are stable, addressable superposition states, so the first-generation gasket's full 0→9 excitation cascade can be stepped through one spin flip at a time by tuning Δ/Ω and V.","Which sublattice is activated — corner, edge, or bulk — is set by the interaction range and the applied fields, so the same array can be programmed to probe a chosen sublattice.","The cascade is not a nine-site accident: SIM-GRAPH finds the same one-by-one plateaus (6 ≤ ⟨n↑⟩ ≤ 15) in the 24-site second generation and repeats the pattern in the third, so larger gaskets should show it too.","The structure factor carries a distinct reciprocal-space fingerprint for each sublattice's length scale, giving an experimental signature that works even though fractals lack translational symmetry.","The authors identify quantum sensing and quantum information processing as the practical beneficiaries of controllable single spin-flips in a many-body system."],"supporting_citations":[{"why":"Establishes Rydberg atoms as a quantum simulator for interacting spin models, the platform the paper builds on.","marker":"[7]"},{"why":"Supplies the atom-by-atom assembly of defect-free arbitrary arrays used to create the nine-site gasket.","marker":"[9]"},{"why":"Provides the many-body framework for individually controlled Rydberg atoms and long-range interactions.","marker":"[12]"},{"why":"Demonstrates tunable Rydberg arrays realizing quantum Ising models, the Hamiltonian class the paper extends to a fractal geometry.","marker":"[17]"},{"why":"Sets the transverse-field Ising model with Rydberg blockade as the experimental paradigm the gasket experiments follow.","marker":"[18]"},{"why":"Shows that sublattice imbalance in the fractal-lattice Hubbard model yields exotic order, motivating the sublattice analysis.","marker":"[36]"},{"why":"Supplies the stochastic series expansion QMC method used as one of the four numerical benchmarks.","marker":"[45]"},{"why":"Adapts SSE-QMC to Rydberg arrays; the concrete QMC implementation used for the second-generation checks.","marker":"[46]"}],"fun_headline_variants":["Fractal spins flip one by one in a Rydberg lattice","Single spin-flips in a quantum fractal","Stepwise spin control in a fractal Rydberg system","Delocalized spin flips in a Sierpinski gasket","Rydberg fractal shows cascade of single spin flips"],"cache_read_input_tokens":2688,"weakest_assumption_plain":"The delocalization claim rests on the prepared state being the true ground state after a few-microsecond sweep and on the three edge (and later bulk) sites remaining equivalent; the paper itself reports that trap deformations and hologram rounding break this threefold symmetry, and that the measured asymmetry is reproduced only after adding 1% position disorder and a 10% outward displacement of one site.","fun_headline_variants_meta":{"raw":{"variants":["Fractal spins flip one by one in a Rydberg lattice","Single spin-flips in a quantum fractal","Stepwise spin control in a fractal Rydberg system","Delocalized spin flips in a Sierpinski gasket","Rydberg fractal shows cascade of single spin flips"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000192,"raw_usage":{"total_tokens":1186,"prompt_tokens":749,"completion_tokens":437,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":493,"completion_tokens_details":{"reasoning_tokens":364}},"tokens_in":493,"tokens_out":437,"duration_ms":4530,"temperature":1.0,"reasoning_tokens":364,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-05T10:51:11.613596+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Image single experimental shots in the 4-spin-up plateau and tally which edge site carries the fourth excitation: the superposition claim requires the fourth spin to appear on all three edge sites with symmetric probabilities. The same apparatus can test the symmetry dependence directly — displacing one corner outward by more than about 10% of the lattice spacing should localize the fourth spin and narrow or eliminate the plateau, exactly as the paper's disorder model predicts.","supporting_citations":[{"cited_title":"& Buchler, H.P","cited_arxiv_id":null,"evidence_quote":"Establishes Rydberg atoms as a quantum simulator for interacting spin models, the platform the paper builds on."},{"cited_title":"& Browaeys, A","cited_arxiv_id":null,"evidence_quote":"Supplies the atom-by-atom assembly of defect-free arbitrary arrays used to create the nine-site gasket."},{"cited_title":"& Lahaye, T","cited_arxiv_id":null,"evidence_quote":"Provides the many-body framework for individually controlled Rydberg atoms and long-range interactions."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Demonstrates tunable Rydberg arrays realizing quantum Ising models, the Hamiltonian class the paper extends to a fractal geometry."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Sets the transverse-field Ising model with Rydberg blockade as the experimental paradigm the gasket experiments follow."},{"cited_title":"The fractal-lattice Hubbard model","cited_arxiv_id":null,"evidence_quote":"Shows that sublattice imbalance in the fractal-lattice Hubbard model yields exotic order, motivating the sublattice analysis."},{"cited_title":"Graph theory","cited_arxiv_id":null,"evidence_quote":"Supplies the stochastic series expansion QMC method used as one of the four numerical benchmarks."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Adapts SSE-QMC to Rydberg arrays; the concrete QMC implementation used for the second-generation checks."}],"review_version":1}