{"id":"9344981a-fe9b-43a8-9a9f-d1f74713e7a4","arxiv_id":"2608.05241","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":4,"one_line_summary":"Quench spectroscopy and correlation imaging show that sound, group, and phase velocities separate into a hierarchy across the superfluid-to-Mott-insulator transition in a 1D Bose-Hubbard gas.","lead":"Physicists in Innsbruck and Paris measured three different propagation speeds for ultracold atoms in a lattice and found that they separate as interactions strengthen. The result gives a new experimental way to read how strongly correlated quantum systems carry sound, correlations, and phase information.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The sound velocity is derived via 1/ξ ∼ Δ/(ℏv_s) with an unspecified O(1) prefactor; if the true prefactor is not 1, the Fig. 4 velocities and the claimed v_s^2=(π/2)v_pv_g invariant shift, so the central hierarchy claim is not yet fully pinned down.","rationale":"After comparing the paper's stated claims with the extraction procedures, I do not find a fatal flaw, but the most load-bearing weak point is the derivation of v_s. The central figure (Fig. 4) combines three velocities measured in different ways: v_p from fringe extrema, v_g from the thresholded envelope, and v_s from ξ and Δ through 1/ξ ∼ Δ/(ℏv_s). Of these, v_s is the only one not directly read off from the quench dynamics. Its value enters the headline invariant relation with the specific constant π/2, so any multiplicative error in v_s translates directly into the invariant ratio. The paper uses '∼', explicitly leaving the prefactor open. In the 1D Bose-Hubbard Mott phase the exponential decay of G(1)(x) is controlled by the lowest particle-hole excitation, and the relation between the decay length, the gap, and the velocity parameter is not guaranteed to have prefactor 1 for the lattice Green's function; non-universal form-factor factors or higher-band effects can appear. The DMRG simulations are in principle able to settle this: for the same U,J and γ they can compute ξ, Δ, and v_s independently and compare with the assumed relation. I therefore recommend the test above. If the test verifies C=1, the central claim is supported by an additional independent check; if not, the paper should either quote the corrected velocities or soften the invariant-relation claim. The reader's verdict of CONDITIONAL is appropriate; my concern does not change it, so the verdict stays UNCHANGED. I credit the paper for making data available, for DMRG reproduction of both spectra and correlation cones, and for the cross-check of v_p and v_g to within 5%; those independent supports are why the concern is conditional rather than rejection.","tokens_in":12583,"tokens_out":15523,"duration_ms":149533,"concrete_test":"Compute, from the same DMRG/TEBD simulations and for each γ reported in Fig. 4, three quantities: (i) ξ from a fit to the equilibrium G(1)(x) using exactly the paper's exponential-decay procedure; (ii) Δ from the spectral function or the particle-hole gap; (iii) v_s from the low-energy slope of the computed dispersion. Then evaluate C(γ)=ℏ v_s/(ξ Δ). If C(γ) deviates from 1 by more than the reported experimental uncertainty or varies significantly across γ, recompute the v_s data points in Fig. 4 with v_s^true = v_s^paper × C(γ), and re-plot both the hierarchy and v_s^2/(v_pv_g). If the corrected inset no longer approaches π/2 in the deep Mott regime, the invariance claim fails; if C(γ)≈1, the concern is retired.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The weakest load-bearing step is the extraction of v_s from equilibrium data. In the main text and Supplemental C, v_s is obtained from 1/ξ ∼ Δ/(ℏv_s), where ξ is the decay length of G(1)(x) and Δ is the excitation gap. The symbol '∼' leaves the prefactor unspecified. The low-energy sine-Gordon/Luttinger-liquid relation can carry a non-universal (and possibly γ-dependent) prefactor C when applied to the one-particle Green's function of the lattice model: 1/ξ = C(γ) Δ/(ℏv_s). Since v_s is the only one of the three velocities entering the central invariant v_s^2=(π/2)v_pv_g, a prefactor C≠1 shifts every plotted v_s by 1/C and changes the inset ratio to C^2 times its true value. A constant shift would only rescale the vertical axis, but a γ-dependent C would alter the ordering v_p>v_s>v_g and the apparent approach to π/2. The paper quotes no estimate for C and does not test the relation independently. This is especially relevant because in the deep Mott regime the fitted U and J are extracted from the same measured dispersion, so a theory input entering v_s is not cross-checked by another measurement. The DMRG calculations reproduce the spectra and correlation cones, but they are not used to verify the ξ–Δ–v_s relation; hence the central quantitative claim currently rests on an unquantified theoretical prefactor.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper reports an experimental study of 1D bosonic atoms in an optical lattice across the superfluid-to-Mott-insulator transition. Using quench spectroscopy, the authors reconstruct the quasiparticle dispersion relation and, independently, follow the space-time evolution of the one-body correlation function extracted from time-of-flight images. From the correlation cone they extract a phase velocity and a group velocity, while the sound velocity is obtained from the measured Mott gap and equilibrium correlation length using 1/ξ ∼ Δ/(ℏv_s). The three velocities are compared with analytic expressions derived in the deep Mott regime and are reported to satisfy a relativistic-like invariance relation v_s^2 = (π/2) v_p v_g. The central claim is that the three velocities, nearly degenerate near the critical point, separate into a hierarchy v_p > v_s > v_g as the gap opens, and that this hierarchy is quantitatively described by a massive relativistic quasiparticle dispersion.","tokens_in":12870,"tokens_out":7207,"duration_ms":74930,"significance":"If the central claim holds, this is a significant experimental advance: it demonstrates that a strongly correlated quantum gas supports multiple, independently measurable propagation velocities that are quantitatively linked through a common low-energy description. The paper's strengths include the use of direct space-time measurements of correlation functions, the independent cross-checks of the gap against lattice modulation spectroscopy and quantum Monte Carlo, the agreement of DMRG simulations with the measured spectra and correlation spreading to within 5% for v_p and v_g, and the availability of data and simulation codes. The result would establish propagation-velocity hierarchies as a genuine observable feature of strongly correlated quantum dynamics and would extend the classical concept of dispersive phase/group velocity separation to interacting quantum matter.","major_comments":[{"comment":"The sound velocity is extracted from the relation 1/ξ ∼ Δ/(ℏv_s), where the symbol ∼ leaves the numerical prefactor unspecified. For the one-particle Green's function of the lattice Bose-Hubbard model, the low-energy relation can carry a non-universal, and possibly γ-dependent, prefactor C(γ), i.e., 1/ξ = C(γ) Δ/(ℏv_s). Since v_s is the only one of the three velocities that enters the central invariant v_s^2 = (π/2) v_p v_g, a value C ≠ 1 rescales every plotted v_s by 1/C and modifies the inset ratio in Fig. 4 by 1/C^2. If C varies across the interaction range, the ordering v_p > v_s > v_g and the apparent approach to π/2 could be distorted. The DMRG simulations reproduce the spectra and correlation cones but are not used to verify the ξ–Δ–v_s relation. I request that the authors determine C(γ) from their DMRG calculations (e.g., by fitting the numerical G(1)(x) and the gap), and propagate the resulting systematic uncertainty into v_s and the invariant ratio, or otherwise provide an independent measurement of v_s, before the central quantitative claim can be regarded as fully supported.","section":"Main text, 'Extraction of the sound velocity'; Supplement C"},{"comment":"The manuscript does not consistently specify whether the velocities v_p, v_s, and v_g are defined from the single-particle dispersion E(k) or from the pair dispersion 2E_ph(k) used in Supplement E and G. The measured correlation cone is governed by pair excitations, as shown in Supplement E, where the phase-front velocity is 2E_{k*}/ℏk* and the envelope velocity is 2∂E/∂ℏk at the saddle point. The analytic formulas v_p = 2U/(πℏ), v_g = 6J/ℏ, and v_s = sqrt(6JU)/ℏ that are compared with the data in Fig. 4 correspond to these pair quantities. In the main text, however, v_s is introduced through E(k) = sqrt(Δ^2 + (ℏv_sk)^2) as if it were the single-particle dispersion. This ambiguity should be resolved explicitly, because the interpretation of the hierarchy and the comparison with the Lieb-Robinson velocity scale depend on whether single-particle or pair velocities are meant.","section":"Fig. 1a, Supplement E and G"},{"comment":"The statement that the three velocities are determined 'without adjustable parameters' is not strictly accurate. The group velocity extraction relies on the Gaussian smoothing widths σ_x and σ_t and on the threshold ε defined in Supplement F, and the text itself reports a ~15% systematic uncertainty in v_g from the choice of ε. These are not fixed by the theory but are analysis choices. The claim should be reformulated to say that no parameter is adjusted to force agreement with the analytic curves, while explicitly listing ε, σ_x, σ_t as analysis parameters whose systematic effects are included in the uncertainty budget.","section":"Main text, 'without adjustable parameters'; Supplement F"}],"minor_comments":[{"comment":"There is a typo in the sentence preceding Eq. (17): 'The corresponding phase velocity v_g = 2E_ph(k)/ℏk* is then v_p = 2U/(πℏ)' should read 'phase velocity v_p' rather than 'v_g'.","section":"Supplement G"},{"comment":"The main text defines G(1)(x,t) = ⟨ψ†(x,t)ψ(0,t)⟩, but the measured quantity is actually the integrated correlation G(1)(x,t) ∝ ∫ ⟨Ψ†(x+x')Ψ(x')⟩ dx', as stated in Supplement E. This distinction should be made in the main text to avoid confusion, particularly because the integrated quantity may contain additional averaging effects.","section":"Main text, definition of G(1)(x,t)"},{"comment":"The analytic lines in Fig. 4 are drawn using the measured U and J, but the values of J used at each point are not stated in the main text. Please provide the J(U) relation or indicate that J is obtained from the same dispersion fits that yield U.","section":"Fig. 4"},{"comment":"The description of the correlation-length extraction states that ln G(1)(x) is fitted with its intercept fixed to zero. Since the measured G(1)(x) may have an overall amplitude and possibly a subleading background, the fit procedure should be described in more detail, including the fit range and the sensitivity of ξ to those choices.","section":"Supplement C"},{"comment":"The schematic in Fig. 1a labels the vertical axis with both the Mott gap Δ and interaction energy U, but the relation between them (e.g., Δ ≈ U−6J in the deep Mott regime) is not given in the caption. Adding this relation would help the reader connect the schematic to the analytic formulas.","section":"Fig. 1a"}],"recommendation":"major_revision","confidential_remarks":"The central experimental platform and the dataset are impressive, and the DMRG agreement is encouraging. My main concern is the unquantified prefactor in the v_s extraction, which is load-bearing for the invariant relation. I do not see this as a fundamental flaw; it can be addressed with additional DMRG analysis. The factor-of-two definition issue is also fixable with clearer text. I recommend major revision rather than rejection."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Bottom line: this is a genuinely new experiment—nobody has independently tracked v_p, v_s, and v_g across the SF-MI transition in a 1D Bose-Hubbard gas before, and the DMRG agreement within 5% puts the correlation-cone measurements on solid ground. The paper deserves a serious referee. The weak joint is the extraction of v_s. It comes from 1/ξ ∼ Δ/(ℏv_s), and the ∼ hides an O(1) prefactor that is neither computed nor bounded. Since v_s is the only velocity that enters the invariant v_s^2 = (π/2)v_pv_g, a prefactor C≠1 would rescale every v_s and change the inset ratio by C^2. If C varies with γ, the ordering v_p > v_s > v_g itself could shift. The authors should either use the exact prefactor from Luttinger-liquid/sine-Gordon theory, extract v_s directly from the dispersion fit (they have U and J from the same spectra), or check the ξ–Δ–v_s relation in DMRG. They did not. The fact that the v_s data points line up with sqrt(6JU) suggests C is close to one in practice, but that is asserted, not shown. The v_g extraction carries a 15% systematic from the threshold ε—honestly reported, but worth keeping in mind. The LR sentence is a little loose: they call v_g the LR velocity scale, which is an upper bound, so saying the phase front is faster than it while 'consistent with locality' is confusing; a one-sentence clarification would fix it. What's good: v_p and v_g come from independent analyses of the space-time correlation maps; the gap cross-checks against LMS and QMC; U and J are extracted from the measured dispersion and then used as inputs to DMRG without extra adjustment; data are on Zenodo. The theory side is standard (parity with Refs. 7-10, 17), and the analytic large-U expressions are clean. This is a paper for the cold-atom and quantum-dynamics crowd, and it will be cited. Give it a serious referee; ask for the v_s prefactor to be pinned down before accepting.","headline":"A well-executed experiment with DMRG backup, but the sound velocity rests on an unquantified prefactor, so the headline invariant needs one more check.","tokens_in":13511,"tokens_out":4433,"would_cite":true,"duration_ms":44719,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":["03.75.Kk","67.85.Hj"],"model":"deepseek-v4-flash","headline":"In a 1D lattice gas, the phase, sound, and group velocities separate into a hierarchy as the Mott gap opens, with the sound velocity obeying v_s^2 = (pi/2) v_p v_g.","keywords":["velocity hierarchy","Bose-Hubbard model","quench spectroscopy","Mott insulator","superfluid-to-Mott transition","ultracold atoms","phase velocity","sound velocity"],"falsifier":"A direct measurement of v_s by Bragg spectroscopy, or by resolving the low-energy slope of the dispersion with higher spectral resolution, that disagrees with the values obtained from 1/xi ~ $\\Delta$/(hbar v_s) by more than the combined uncertainties would falsify the proposed relation; equivalently, observing multi-exponential decay of G^(1)(x) at equilibrium would indicate that the single correlation-length extraction is insufficient.","tokens_in":12344,"feed_emoji":"🌊","tokens_out":7411,"duration_ms":63665,"temperature":0.7,"pith_summary":"This paper reports the first experimental observation of a velocity hierarchy in matter waves: in a one-dimensional Bose-Hubbard gas, the phase velocity, sound velocity, and group velocity are nearly identical near the superfluid-to-Mott-insulator transition but become progressively separated as the interaction strength increases and the Mott gap opens, eventually satisfying v_p > v_s > v_g in the deep Mott regime. The authors independently measure all three velocities across the transition using quench spectroscopy and time-resolved correlation measurements, and find they are connected through the same quasiparticle dispersion. The measured velocities obey a relativistic-like invariance relation, $v_s^{2}$ = (pi/2) v_p v_g, analogous to $c^{2}$ = v_p v_g for a massive relativistic particle. If correct, this establishes propagation-velocity hierarchies as emergent signatures of strongly correlated quantum dynamics and provides a parameter-free test of the low-energy description of the Mott insulator.","feed_headline":"Three velocities emerge as a lattice gas turns insulating","feed_subtitle":"Quench spectroscopy shows phase, sound, and group velocities obey v_s^2 = (pi/2) v_p v_g.","key_machinery":"The central object is the low-energy quasiparticle dispersion of the 1D Bose-Hubbard model, E(k)=$\\sqrt$($\\Delta$^2+(hbar v_s k)^2) in the Mott regime, which changes from gapless linear to gapped relativistic-like as interactions increase. This dispersion carries the argument: it determines the sound velocity v_s as the slope parameter of the low-energy cone, the group velocity v_g = dE/(hbar dk) as the maximum of the derivative (sitting at k=pi/2 in the deep Mott limit), and the phase velocity v_p = E/(hbar k) at that same point. The experiment reconstructs this dispersion via quench spectroscopy (S(k,nu) from Fourier transforming the time-dependent momentum distribution), and reads v_g and v_p off the space-time evolution of the one-body correlation function G^(1)(x,t), while v_s is obtained from the equilibrium relation 1/xi ~ $\\Delta$/(hbar v_s). The analytical limits v_p=2U/(pi hbar), v_s=$\\sqrt$(6JU)/hbar, v_g=6J/hbar then yield the invariant $v_s^{2}$=(pi/2) v_p v_g, which the measured velocities confirm.","core_discovery":"The central discovery is that the low-energy quasiparticle dispersion of a strongly interacting 1D Bose-Hubbard gas, which evolves from a nearly linear phonon-like form near the superfluid transition to a massive relativistic-like form E(k)=$\\sqrt$($\\Delta$^2+(hbar v_s k)^2) deep in the Mott phase, gives rise to three distinct and independently measurable propagation velocities. Near the transition these velocities are nearly degenerate; as the gap $\\Delta$ opens, the dispersion bends and the velocities separate into a hierarchy v_p>v_s>v_g. The authors extract v_p from the phase interference fringes inside the correlation cone of G^(1)(x,t), v_g from the envelope of the same correlation spreading, and v_s from the independently measured gap and equilibrium correlation length via 1/xi ~ $\\Delta$/(hbar v_s). They show that in the deep Mott regime the measured velocities agree with the analytical expressions v_p=2U/(pi hbar), v_s=$\\sqrt$(6JU)/hbar, and v_g=6J/hbar, and consequently satisfy the invariant relation $v_s^{2}$=(pi/2) v_p v_g, which they verify without adjustable parameters. Phase-coherence fronts propagate faster than the Lieb-Robinson velocity scale yet remain consistent with locality, since phase fronts do not transmit information.","pith_inferences":["A testable extension is to measure v_s directly via two-photon Bragg spectroscopy and compare with the values extracted from the gap and correlation length; a mismatch beyond the quoted ~15% uncertainty would indicate that the single-mode relation 1/xi ~ Delta/(hbar v_s) needs correction.","The invariant v_s^2 = (pi/2) v_p v_g likely reflects a geometric factor relating the maximum group-velocity point (k=pi/2) to the low-energy cone in a 1D lattice; analogous factors should appear in other lattice geometries where the dispersion has a different shape, providing a route to test universality.","The observation that phase fronts can outrun the Lieb-Robinson cone while information remains causal suggests that correlation spreading in lattice systems should be characterized by at least two distinct velocities; future entanglement-based measurements could separate the information-carrying front from the phase-coherence front."],"forward_implications":["The velocity hierarchy is a continuous, quantitative signature of the Mott transition: it can be used as a probe of the gap opening and quasiparticle renormalization in lattice gases.","Because the three velocities are linked by the same dispersion, measuring any two together with the gap constrains the third; the relation v_s^2 = (pi/2) v_p v_g provides a parameter-free consistency check of the low-energy theory in the deep Mott regime.","Phase-coherence fronts exceeding the Lieb-Robinson velocity do not imply superluminal signaling; the observed pattern shows that such fronts can coexist with causal, locally consistent dynamics, so velocity hierarchies must be interpreted via their information-transport capacity.","The experimental methodology (quench spectroscopy plus time-resolved correlations) transfers directly to other lattice models, including higher-dimensional and disordered systems, where multiple excitation branches may produce richer velocity structures."],"supporting_citations":[{"why":"Supplies the quench-spectroscopy method used to reconstruct the quasiparticle dispersion from the time-dependent momentum distribution.","marker":"[18]"},{"why":"Provides the two-particle dispersion and predicts correlation propagation fronts in interacting Bose gases, the basis for identifying v_p and v_g.","marker":"[7]"},{"why":"Earlier experiment on light-cone-like correlation spreading that this work extends by connecting fronts to independently measured energy scales.","marker":"[12]"},{"why":"Gives the low-energy relation 1/xi ~ Delta/(hbar v_s) used to extract the sound velocity from the gap and correlation length.","marker":"[29]"},{"why":"Provides lattice modulation spectroscopy measurements of the gap and the 1D interaction parameter gamma used to tune the system.","marker":"[25]"},{"why":"Theoretical treatment of phase and group velocities in the Mott phase of the Bose-Hubbard model, the prior claim this experiment tests.","marker":"[17]"},{"why":"Describes how to obtain the equilibrium one-body correlation function and correlation length from static momentum distributions.","marker":"[30]"},{"why":"Defines the Lieb-Robinson velocity scale against which the measured phase-front speed is compared.","marker":"[1]"}],"fun_headline_variants":["Three velocities emerge in a quantum lattice gas","Phase velocity outruns sound and group in Mott insulator","Velocity hierarchy observed in strongly interacting matter waves","Quantum gas reveals three distinct propagation speeds","Matter waves display velocity splitting across transition"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The sound velocity is not measured directly; it is derived from the relation 1/xi ~ $\\Delta$/(hbar v_s) using an order-one prefactor from low-energy Luttinger-liquid and sine-Gordon theory, so if the correlation decay receives contributions beyond the single low-energy mode or the prefactor differs from unity, the derived v_s (and hence the hierarchy and the invariant relation) would shift.","fun_headline_variants_meta":{"raw":{"variants":["Three velocities emerge in a quantum lattice gas","Phase velocity outruns sound and group in Mott insulator","Velocity hierarchy observed in strongly interacting matter waves","Quantum gas reveals three distinct propagation speeds","Matter waves display velocity splitting across transition"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000734,"raw_usage":{"total_tokens":3287,"prompt_tokens":958,"completion_tokens":2329,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":574,"completion_tokens_details":{"reasoning_tokens":2261}},"tokens_in":574,"tokens_out":2329,"duration_ms":17252,"temperature":1.0,"reasoning_tokens":2261,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-08T17:24:22.206800+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"A direct measurement of v_s by Bragg spectroscopy, or by resolving the low-energy slope of the dispersion with higher spectral resolution, that disagrees with the values obtained from 1/xi ~ $\\Delta$/(hbar v_s) by more than the combined uncertainties would falsify the proposed relation; equivalently, observing multi-exponential decay of G^(1)(x) at equilibrium would indicate that the single correlation-length extraction is insufficient.","supporting_citations":[{"cited_title":"Barmettler, D","cited_arxiv_id":null,"evidence_quote":"Provides the two-particle dispersion and predicts correlation propagation fronts in interacting Bose gases, the basis for identifying v_p and v_g."},{"cited_title":"Haller, R","cited_arxiv_id":null,"evidence_quote":"Provides lattice modulation spectroscopy measurements of the gap and the 1D interaction parameter gamma used to tune the system."},{"cited_title":"Mokhtari-Jazi, M","cited_arxiv_id":null,"evidence_quote":"Theoretical treatment of phase and group velocities in the Mott phase of the Bose-Hubbard model, the prior claim this experiment tests."}],"review_version":1}