{"id":"ffa1cab9-a258-48c8-82f7-2d7539584d90","arxiv_id":"2507.12131","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":1,"one_line_summary":"Extended doublons with drive-induced non-local interactions are experimentally observed, and their topologically protected transport through the bulk is demonstrated in a photonic Floquet lattice.","lead":"Using laser-written photonic waveguides, the authors realize extended two-particle bound states, called doublons, whose non-local interaction is generated by periodic driving, and observe them travelling through the bulk of a lattice in a topologically protected channel. The experiment demonstrates a route to studying few-particle topological phenomena without relying on mean-field descriptions.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Central mechanism is deferred to a missing SI; the four-step drive's claimed revival of non-local interactions is not derivable from the main text and cannot be verified from the preprint.","rationale":"After reading the paper in good faith, I find the central claim—that a four-step Floquet drive induces non-local interactions that form extended doublons and enable topologically protected transport through the bulk—is plausible but rests on a theoretical construction that is not present in the submitted text. The reader's weakest assumption correctly identifies the decomposition into four virtual Hamiltonians as the load-bearing step. I agree: every subsequent statement about doublon formation, the two-site extension, and the four-site-per-two-period motion is interpreted through that decomposition, and the main text provides no independent derivation. I considered whether the missing perturbation test for 'protected' is more load-bearing, but topological protection in Floquet systems is usually inferred from band-structure topology once the model is established; the primary risk is whether the model itself (with non-local interactions) is correct. Hence I focus on the mechanism. The concrete test I propose—an independent Floquet-Hamiltonian extraction and projection into the two-particle basis—would settle whether the virtual-Hamiltonian picture is a real effect or a mislabeling of standard single-particle Floquet physics. Because the concern is about missing justification rather than a demonstrated error, the appropriate verdict remains CONDITIONAL, matching the reader's assessment.","tokens_in":8763,"tokens_out":10238,"duration_ms":118457,"concrete_test":"Compute the exact one-period evolution operator U(T) for the driven 2D lattice in Fig. 2b with the stated parameters and diagonal detuning U=2√3; extract the effective Floquet Hamiltonian H_F = (i/T) log U(T), and project it onto the two-particle 1D SSH basis. Check whether H_F contains explicit finite-range density-density or correlated-hopping terms beyond the local Hubbard U. Then test two predictions: (i) the two-particle spectrum has a two-site-extended bound state at U=2√3 whose stroboscopic motion advances four sites in two periods as in Fig. 4; (ii) replacing U(T) by the time-averaged Hamiltonian exp(-i T Σ H_i/4) destroys that bound transport. If no non-local term appears in H_F, the 'drive-induced non-local interactions' interpretation is not supported.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The main text asserts that H_0 can be decomposed as H_0 = H_I + H_II + H_III + H_IV, with each H_i containing only non-local interaction terms that cancel in the sum, and that the four-step helical drive in Fig. 2b sequentially instantiates these H_i, thereby exposing the non-local interactions. This is the entire basis for the claim of drive-induced non-local doublon formation, yet the derivation is relegated to 'the Supplementary Information,' which is not present in the preprint. In the standard dimensional mapping used in the same paper (and Refs. 37,46), H_0 corresponds to a single particle on a 2D square lattice with horizontal and vertical nearest-neighbor hopping. Under the four-step protocol, the horizontal step is the operator sum over m,n of J a†_{m+1,n} a_{m,n}, which in the two-particle tensor-product basis equals (J a†_{m+1} a_m) ⊗ I_B—a single-particle kinetic term, not a non-local interaction. Thus, as stated, the decomposition appears inconsistent with the naive mapping unless the SI introduces a nontrivial regrouping (e.g., density-assisted hopping per site of the other particle). Because this construction is the load-bearing premise for identifying the observed transport as a doublon, the central claim cannot be assessed without that derivation. The experimental support is likewise deferred to missing Extended Data Figures XD1 and XD2, and no perturbation test is shown for the word 'protected.'","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper reports an experimental study of extended doublons—two-particle bound states with non-local interactions—in a photonic artificial solid. The authors start from an SSH-type 1D chain of two interacting particles and use the standard dimensional mapping to a 2D square lattice, where the Hubbard interaction becomes a detuning of the diagonal sites. Their central theoretical move is to decompose the non-interacting Hamiltonian H0 into four virtual Hamiltonians HI–HIV that individually contain non-local interaction terms which cancel when summed. They assert that a four-step helical Floquet drive sequentially instantiates these virtual Hamiltonians, thereby evading the cancellation and producing drive-induced non-local interactions. At a specific resonant interaction strength, they observe a wave packet that propagates four sites in two Floquet periods along the detuned diagonal, which they interpret as a topologically protected extended doublon moving through the bulk of the 1D system. Experimental comparisons across interaction strengths and band-resolved measurements are deferred to Extended Data Figures XD1 and XD2, and the key derivation is deferred to the Supplementary Information.","tokens_in":9008,"tokens_out":4773,"duration_ms":60200,"significance":"If the central derivation is valid and the deferred experimental data are present, the work would constitute a notable experimental step toward few-particle topological physics in photonic lattices. The notion of drive-induced non-local interactions by sequential exposure of mutually cancelling virtual Hamiltonians is conceptually interesting and goes beyond earlier single-particle Floquet topological experiments. The paper builds appropriately on prior theoretical work (Refs. 36 and 49) and uses a well-established experimental platform. However, as submitted, the manuscript does not allow the reader to verify the load-bearing mechanism: the decomposition of H0 into HI–HIV and the claim that each Floquet step instantiates one of them are deferred entirely to a missing Supplementary Information. The experimental evidence for the main claims is also deferred to Extended Data figures that appear as placeholders. No machine-checked proofs or reproducible code are provided, and the data-availability statement still contains a placeholder reference.","major_comments":[{"comment":"The decomposition H0 = HI + HII + HIII + HIV, with each H_i containing only non-local interaction terms that cancel when summed, is the entire mechanism for drive-induced non-local interactions, but it is not derived in the main text; the proof is deferred to \"the Supplementary Information\" (pp. 2 and 3). In the naive dimensional mapping used in the same paper, each directional hopping term in the four-step protocol is a single-particle kinetic operator (e.g., sum_{m,n} J a†_{m+1,n}a_{m,n} = (J a†_{m+1}a_m) ⊗ I_B), so a non-trivial regrouping is required for the decomposition to produce non-local interactions. Without that derivation, the central claim that the observed four-site motion is an extended doublon and not a single-particle edge state in the mapped lattice cannot be assessed.","section":"Full Text p.3, Eq. (1) and Fig. 2b"},{"comment":"The resonance condition is central: the text states p(2√3)=0 and uses this first zero to set the interaction strength for the main experiment (Fig. 4). However, the hopping probability p(U) onto the diagonal is not defined or derived in the main text, and the zeros at U=2√3 are not obtained from any equation shown to the reader. The band-structure panels in Fig. 3b–3e are qualitative and do not by themselves establish the quantitative resonance positions used in the experiment. This definition and derivation must be moved into the main text or an accessible appendix, rather than being left to the missing Supplementary Information.","section":"Full Text p.4, Fig. 3a"},{"comment":"The experimental evidence for the central quantitative claims is not present in the manuscript: the comparison across interaction strengths and the band-resolved measurements are deferred to \"Extended Data Figure XD1\" and \"Extended Data Figure XD2\", which appear as placeholders in the posted text. Figure 4 shows only a single resonance, and no perturbation, disorder, or back-propagation test is shown, so the word \"protected\" in the abstract and conclusion is not demonstrated by the data. If the protection claim is meant to follow from a topological invariant, the invariant calculation for the specific doublon channel along the detuned diagonal should be given in the main text, not only invoked qualitatively.","section":"Full Text p.5, Fig. 4 and Extended Data Figures XD1–XD2"}],"minor_comments":[{"comment":"The word \"detunded\" appears where \"detuned\" is intended; please correct the typo.","section":"Methods, Experimental configuration"},{"comment":"The data-availability statement ends with \"[reference follows]\"; a complete accession identifier should be provided before review.","section":"Data availability"},{"comment":"The text refers to \"the band structure of our system depicted in Figs. 3b-d,\" while the caption labels panels (b)–(e); please ensure the in-text panel references match the caption and that the U→∞ panel (e) is explicitly discussed.","section":"Full Text p.4 and Fig. 3"},{"comment":"The phrase \"as these entities are defined as bound bipartite quasi-particles with long-range interaction\" would benefit from a precise operational definition of \"long-range\" in the context of the implemented lattice, since the interaction is induced rather than an explicit long-range Hamiltonian term.","section":"Full Text p.2, Definition of Doublons"}],"recommendation":"major_revision","confidential_remarks":"The manuscript cannot be properly evaluated in its current form because the core derivation and the primary experimental comparisons are relegated to files that are not present in the posted preprint. If the full submission to the journal includes a complete Supplementary Information and the actual Extended Data figures, the editor should ensure that these materials are distributed to reviewers; otherwise the central claims are unverifiable."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague,\n\nThe headline: this is the first experimental claim of extended (two-site) doublons with drive-induced non-local interactions, and the central observation—a two-particle bound state moving through the bulk by four sites in two Floquet periods—is genuinely new. If it holds up, it is a nice step for few-particle topological physics in photonic lattices.\n\nWhat the paper does well: the dimensional mapping from two particles in 1D to one particle in 2D is standard (refs 37, 46), and the band structure analysis in Fig. 3 is present in the main text and gives a concrete prediction for the first resonance at U = 2√3. The experiment in Fig. 4 shows the expected propagation along the diagonal, and a comparison to numerical simulations is included for the target occupation (though only in Extended Data XD1). The writing is clear, and the authors are honest that the construction of the four virtual Hamiltonians is detailed in the Supplementary Information.\n\nThe soft spot is exactly there. The claim that H0 decomposes into four virtual Hamiltonians, each with only non-local interaction terms that cancel when summed, is the load-bearing premise. But the decomposition is not derived in the main text, and the SI is not included in the preprint. Worse, the stress-test note identifies a real tension: in the standard mapping, the horizontal step of the four-step drive is just kinetic hopping for one particle, (J a†_{m+1} a_m) ⊗ I_B, which is a single-particle term, not a non-local interaction. As written, the main text appears inconsistent with the naive mapping unless the SI introduces a nontrivial regrouping (e.g., density-assisted hopping). This is not a minor gap; it is the mechanism that produces the extended doublon. Until the SI is available, the central claim cannot be independently assessed.\n\nThe experimental evidence is also thinner than the text suggests. The word 'protected' in 'topologically protected motion' is asserted, but no perturbation test is shown; the only robustness evidence is the band structure. Data availability is a placeholder, and the Extended Data figures are referenced but not reproduced. These are fixable, but they are conditions on the claim, not quibbles.\n\nIf the SI delivers the decomposition and the data artifacts appear, this paper is likely a solid contribution. As submitted, it deserves peer review—the novelty and experimental effort are real—but the referee will need to see the derivation before the physics can be trusted.\n\nMy recommendation: send it to review, but with the explicit request that the SI be made available and the 'protection' claim be backed by either a perturbation experiment or a clear argument.","headline":"First experimental claim of extended doublons with drive-induced non-local interactions, but the load-bearing theory sits in a missing SI and the data artifacts are incomplete.","tokens_in":9562,"tokens_out":2691,"would_cite":false,"duration_ms":30583,"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 four-step periodic drive converts local interactions into non-local ones, creating extended, topologically protected two-particle bound states that move unidirectionally through the bulk of a photonic lattice.","keywords":["doublons","two-particle bound states","non-local interactions","periodic driving","topological transport","photonic lattices","dimensional mapping"],"falsifier":"Scan the interaction strength $U$ through the first resonance ($U = 2\\sqrt{3}$ in units of the hopping $J$): the theory requires pronounced doublon formation and unidirectional transport only in the immediate vicinity of the zeros of the hopping probability $p(U)$, with decay into the bulk away from them, and the doublon intensity profile to remain extended over exactly two sites. A measurement showing robust transport for all detunings, or a bound state contracted to a single site, would falsify the resonance and extended-doublon picture.","tokens_in":8523,"feed_emoji":"🔁","tokens_out":10077,"duration_ms":106312,"temperature":0.7,"pith_summary":"This paper reports the experimental observation of extended two-particle bound states, called doublons, whose binding arises from non-local interactions induced by periodic driving rather than from a static on-site potential. The authors show that two interacting particles on a one-dimensional alternating-bond chain can be mapped onto one particle moving on a two-dimensional square lattice, where the local interaction becomes a detuning along the diagonal. When the hopping is switched sequentially in four drive steps, the four virtual Hamiltonians that cancel in the static system are instantiated one by one, so the non-local interaction terms survive. At certain resonant values of the interaction strength the two triangular halves of the lattice decouple, creating a topological channel along the diagonal that transports the doublon unidirectionally through the bulk while keeping it extended over two sites. The measurements confirm this resonant formation and topologically protected motion, opening a route to studying few-particle topological states on established experimental platforms.","feed_headline":"Topological channel carries extended doublons through the bulk","feed_subtitle":"Two particles bind into an extended state and move four sites per two drive periods, protected against disorder","key_machinery":"The central object is the four-step helical drive that sequentially enables the four virtual Hamiltonians $H_I$ to $H_{IV}$. In the static picture these Hamiltonians sum to the ordinary hopping Hamiltonian $H_0$, and their non-local interaction terms cancel exactly; by switching on each for one quarter of the drive period $T$, the drive evades the cancellation and lets the non-local terms act on the two-particle wavefunction. Working through the dimensional mapping, a local on-site interaction in the one-dimensional chain becomes a detuning $U$ of the diagonal sites in the two-dimensional lattice, and the topological channel appears where the hopping probability $p(U)$ onto that diagonal vanishes. The vanishing points are the resonances at which the two triangular sub-lattices decouple, turning the diagonal into a boundary along which the edge state propagates.","core_discovery":"The paper claims that a non-adiabatic periodic drive, combined with the mapping of two-particle dynamics onto a two-dimensional lattice, endows two otherwise locally interacting particles with a non-local, long-range interaction, leading to the formation of extended doublons at specific resonant interaction strengths. These extended doublons are topological bulk states: in the mapped picture, the bound pair travels clockwise along the detuned diagonal in a square lattice with a winding-number-one band structure, moving by four lattice sites (two unit cells) after two drive periods while remaining extended over two sites. The authors argue that the motion is robust because it populates an anomalous topological edge state of the decoupled triangular lattice domains, and that the same channel exists for both signs of the interaction, i.e., for both repulsive and attractive forces.","pith_inferences":["Going beyond the paper, the same drive-induced non-local interaction mechanism could be applied to more than two particles: mapping an $N$-particle problem onto an $N$-dimensional lattice should produce extended bound states (trions, quartets) at the corresponding detuning resonances.","The decoupling condition $p(U) = 0$ resembles a destructive-interference resonance, suggesting that the interaction strength itself could act as a switch that gates doublon transport—a functionality the paper does not explicitly exploit but its data imply.","A direct test of the resonance ladder: tuning to the second, third, and later zeros of $p(U)$ should reveal additional decoupling channels at higher interaction strengths; the paper demonstrates only the first resonance, so this is a predicted but unmeasured consequence."],"forward_implications":["At the resonant interaction strengths, the two-particle bound state propagates by four lattice sites (two unit cells) in two drive periods, stays extended over two sites, and moves unidirectionally along the diagonal—the hallmark of topologically protected transport in the bulk.","The same topological channel exists for both repulsive and attractive interactions: changing the sign of the detuning does not destroy the doublon, only the direction of the mapping into the one-dimensional picture changes.","Off-resonant interaction strengths produce unstable doublons that dissipate into the bulk, while very strong interactions converge back to the resonant behaviour, establishing the resonance structure of the bound state.","Because the protocol only requires a periodic drive and a lattice with a local interaction term, the mechanism can in principle be transferred to cold atoms, electric circuits, and other wave-physics platforms that already support anomalous topological insulators."],"supporting_citations":[{"why":"Defines the alternating-bond one-dimensional chain that hosts the two-particle dynamics.","marker":"[44]"},{"why":"Supplies the dimensional mapping from two particles on a one-dimensional lattice to one particle on a two-dimensional square lattice, with the interaction as a diagonal detuning.","marker":"[37]"},{"why":"Extends the same dimensional mapping to correlated particles, justifying the interpretation of the two-dimensional dynamics as two-particle evolution.","marker":"[46]"},{"why":"Establishes anomalous edge states and the bulk-edge correspondence for periodically driven two-dimensional systems, underpinning the topological channel.","marker":"[47]"},{"why":"Provides the experimental realisation of the anomalous driving protocol in photonic lattices, the driving scheme used here.","marker":"[14]"},{"why":"Analyzes topological two-particle dynamics in a periodically driven lattice with on-site interactions, the theoretical basis for finite-interaction doublons.","marker":"[36]"},{"why":"Predicts the resonant interaction strengths by computing the hopping probability onto the diagonal in the driven insulator.","marker":"[49]"},{"why":"Describes the femtosecond laser-writing technique used to fabricate the photonic waveguides.","marker":"[50]"}],"fun_headline_variants":["Topological bulk transport for extended doublons","Extended doublons take a protected topological path","Drive-induced interactions enable topological doublon travel","Nonlocal driving moves doublons along topological channel"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The result depends on the decomposition of the hopping Hamiltonian into four virtual Hamiltonians whose non-local interaction terms cancel in the static system and are only revealed by sequential driving—a construction derived in the Supplementary Information rather than the main text.","fun_headline_variants_meta":{"raw":{"variants":["Topological bulk transport for extended doublons","Extended doublons take a protected topological path","Drive-induced interactions enable topological doublon travel","Nonlocal driving moves doublons along topological channel"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000748,"raw_usage":{"total_tokens":3312,"prompt_tokens":902,"completion_tokens":2410,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":518,"completion_tokens_details":{"reasoning_tokens":2350}},"tokens_in":518,"tokens_out":2410,"duration_ms":21351,"temperature":1.0,"reasoning_tokens":2350,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-06T16:53:11.981334+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Scan the interaction strength $U$ through the first resonance ($U = 2\\sqrt{3}$ in units of the hopping $J$): the theory requires pronounced doublon formation and unidirectional transport only in the immediate vicinity of the zeros of the hopping probability $p(U)$, with decay into the bulk away from them, and the doublon intensity profile to remain extended over exactly two sites. A measurement showing robust transport for all detunings, or a bound state contracted to a single site, would falsify the resonance and extended-doublon picture.","supporting_citations":[{"cited_title":"P., Schrieffer, J","cited_arxiv_id":null,"evidence_quote":"Defines the alternating-bond one-dimensional chain that hosts the two-particle dynamics."},{"cited_title":"& Osellame, R","cited_arxiv_id":null,"evidence_quote":"Supplies the dimensional mapping from two particles on a one-dimensional lattice to one particle on a two-dimensional square lattice, with the interaction as a diagonal detuning."},{"cited_title":"Photonic Bloch oscillations of correlated particles","cited_arxiv_id":null,"evidence_quote":"Extends the same dimensional mapping to correlated particles, justifying the interpretation of the two-dimensional dynamics as two-particle evolution."},{"cited_title":"S., Lindner, N","cited_arxiv_id":null,"evidence_quote":"Establishes anomalous edge states and the bulk-edge correspondence for periodically driven two-dimensional systems, underpinning the topological channel."},{"cited_title":"J., Zeuner, J","cited_arxiv_id":null,"evidence_quote":"Provides the experimental realisation of the anomalous driving protocol in photonic lattices, the driving scheme used here."},{"cited_title":"& Carusotto, I","cited_arxiv_id":null,"evidence_quote":"Analyzes topological two-particle dynamics in a periodically driven lattice with on-site interactions, the theoretical basis for finite-interaction doublons."},{"cited_title":"& Bauer, D","cited_arxiv_id":null,"evidence_quote":"Predicts the resonant interaction strengths by computing the hopping probability onto the diagonal in the driven insulator."},{"cited_title":"& Nolte, S","cited_arxiv_id":null,"evidence_quote":"Describes the femtosecond laser-writing technique used to fabricate the photonic waveguides."}],"review_version":1}