{"id":"64543b4f-1ef8-4283-933b-4a6128345624","arxiv_id":"2507.12153","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":6,"one_line_summary":"Interacting bosons in twisted-bilayer optical lattices form cluster-induced Mott-like and Bose-glass-like phases, including isolated non-percolating mobility islands and, for incommensurate twists, mobility from interlayer interactions alone.","lead":"This paper calculates the ground-state phases of ultracold bosonic atoms in two twisted square-lattice layers, finding that the interlayer coupling splits the lattice into two-site clusters and produces unusual Mott-like and Bose-glass-like phases. It matters because twisted-bilayer optical lattices have just been built in the lab, so these predictions can be tested experimentally.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Commensurate Bose-glass phase may be an artifact of the 10^-3 t⊥ connectivity cutoff, since weak inter-island couplings are nonzero and yield a global superfluid at T=0.","rationale":"The reader's weakest_assumption identifies the percolation criterion as the central weak point, which matches my concern, but the reader emphasizes the |b|>0.05 threshold and the absence of exact benchmarking, whereas I focus on the connectivity cutoff 10^-3 t⊥ and its direct conflict with the physics of a clean periodic lattice. In the commensurate case, the model has no disorder and no exactly zero couplings, so the only way to obtain a non-percolating island phase is to neglect weak links by fiat. This is a real internal tension: a periodic Josephson array with nonzero couplings, no matter how small, has a Bloch band and a finite superfluid stiffness at zero temperature. The paper's own criterion 'a slight change does not affect results' does not address an order-of-magnitude reduction of the cutoff, and the absence of a stiffness calculation leaves the claim unverified. For incommensurate angles the Bose-glass claim is more plausible because quasiperiodic potentials can induce localization, so the central claim does not collapse entirely even if the commensurate BG proves to be an artifact. The proposed test—computing the superfluid stiffness and reducing the cutoff—would settle the question directly. The paper is clear, the model is physically motivated, and the cluster Gutzwiller method is a reasonable variational approach; the production issues (watermark, numerical typo) are minor. Therefore the CONDITIONAL verdict from the reader remains appropriate; my concern does not change the verdict but sharpens the condition that should be verified.","tokens_in":10599,"tokens_out":7928,"duration_ms":102743,"concrete_test":"Using the same cluster Gutzwiller ansatz, compute the superfluid stiffness D = (1/N) d^2E/dφ^2 from a twisted-phase boundary condition at the commensurate BG point marked by the cross in Fig. 2(a), for N = 10x10, 20x20, 40x40 unit cells. If D extrapolates to a positive value as N increases, the BG phase is a cutoff artifact; if D -> 0 while compressibility kappa = dn/dμ > 0, the BG label is supported. As a second check, reduce the connectivity threshold from 10^-3 to 10^-6 t⊥ and recompute P at the same point: P>0 would confirm threshold dependence.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The most load-bearing weakness is the use of a nonzero connectivity cutoff to define the Bose-glass phase, especially at commensurate angles. In Sec. II.C, two S=1 sites are deemed connected only if their coupling exceeds 10^-3 t⊥; all weaker links are ignored. But in the actual Hamiltonian inter-layer couplings are Gaussian and never exactly zero, and for a commensurate twist the lattice is periodic. A periodic array of superfluid islands coupled by arbitrarily small but nonzero Josephson tunneling is a superfluid at T=0, with a finite (if tiny) superfluid stiffness. Thus the BG-like phase shown in Fig. 2(a) may be an artifact of this threshold: the islands are not truly disconnected, and the system is a very weak global superfluid. The paper does not compute the superfluid stiffness or any phase-twist energy, so this cannot be checked from the reported data. The incommensurate cases are on firmer ground because quasiperiodicity can localize, but the same cutoff-based percolation criterion is used there, and the interaction-only claim in Sec. IV also uses cluster-family averaging that coarse-grains the aperiodic potential.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies the ground-state phases of interacting bosons in twisted-bilayer square optical lattices, with both inter-layer hopping and inter-layer interactions. The authors introduce a cluster Gutzwiller ansatz in which pairs of sites with strong inter-layer coupling are treated as two-site clusters, solve the resulting mean-field equations self-consistently, and classify phases by the distribution and percolation of sites with non-vanishing local mean field b. For the commensurate angle θ(3,2) they report a sequence of Mott-like density-wave phases and a Bose-glass-like phase of isolated superfluid islands; for the incommensurate angle θ(2,1)+3° they report Bose-glass regimes both with inter-layer hopping and, in the interaction-only case, without inter-layer hopping.","tokens_in":10897,"tokens_out":4919,"duration_ms":58011,"significance":"If the central claims are correct, the paper would extend Bose-glass physics to periodic commensurate moiré lattices and predict interaction-induced mobility islands in aperiodic bilayers, both of which are timely for recent twisted-bilayer cold-atom experiments. The paper has clear strengths: an explicit variational ansatz adapted to the cluster geometry, concrete percolation observables F and P, and a transparent presentation of the model. However, the main qualitative conclusions rest on threshold-based mean-field criteria rather than on genuine order parameters such as superfluid stiffness or compressibility, and the commensurate Bose-glass claim in particular is vulnerable to the finite connectivity threshold used in the percolation analysis.","major_comments":[{"comment":"The identification of a Bose-glass phase at a commensurate twist angle rests on the connectivity threshold 10^-3 t⊥. Because the inter-layer couplings t⊥(j,j') are Gaussian and never exactly zero, the supposedly isolated S=1 islands in Fig. 2(b) are coupled by small but nonzero Josephson tunneling; in a periodic lattice at T=0 this generically gives a global superfluid with a small but finite superfluid stiffness, not an insulating Bose glass. The paper neither computes the superfluid stiffness nor the phase-twist energy, so the central abstract claim that a Bose-glass-like phase occurs even for commensurate angles is not established. I ask for a calculation of the twist-energy or superfluid stiffness as a function of system size and a systematic scan of the connectivity threshold in the range 10^-4 to 10^-2 t⊥.","section":"Sec. II.C and Fig. 2(a)"},{"comment":"The incommensurate phase diagrams are obtained on a single 40×40 system with a percolation criterion, but no finite-size scaling is reported. Percolation is intrinsically a finite-size notion, and without varying the system size (and ideally averaging over twist angles) the BG regimes in Figs. 4 and 5 could reflect the finite sample rather than a thermodynamic phase. In addition, the replacement of the continuous distribution of cluster couplings by four discrete family averages (0.875, 0.625, 0.35, and 0.013 t⊥) coarse-grains the aperiodic potential and can artificially disconnect sites that would be connected in the actual Hamiltonian. The paper should quantify the sensitivity of F and P to the binning and to the cluster-definition cutoff t⊥,cr.","section":"Sec. IV"},{"comment":"No benchmark validates the cluster Gutzwiller ansatz or the mean-field percolation criterion against exact small-system results, quantum Monte Carlo, or an independent measure of superfluidity. The thresholds |b|<0.05 and t⊥>t⊥,cr are fixed by hand, and the assertions that 'we have checked' and that 'a slight change does not affect any of our results' are not accompanied by data. Because every phase label in Figs. 2, 4, and 5 is defined by these thresholds, the paper should provide at least one small-system exact-diagonalization comparison and a threshold-sensitivity table for the reported phase boundaries.","section":"Secs. II.B and II.C"},{"comment":"For the t⊥=0 case, the connectivity rule of Sec. II.C is ambiguous: the rule was stated in terms of couplings larger than 10^-3 t⊥, which is not a meaningful scale when t⊥=0. If the percolation graph in Figs. 5(b)-(d) includes inter-layer links, the two layers are connected even though there is no inter-layer hopping; if it uses only intra-layer links, this should be stated explicitly. As written, the BG/SF boundaries in the interaction-only scenario are not reproducible from the stated algorithm.","section":"Sec. IV (interaction-only case)"}],"minor_comments":[{"comment":"The notation α̸=ᾱ in Eq. (1) is not defined before use; please state explicitly that ᾱ denotes the opposite layer.","section":"Sec. II.A"},{"comment":"In the list of cluster families for the incommensurate case, family 4 is given as t⊥,c/t⊥ ∈ [0.06, 0.2) but the assigned average is 0.013, which lies outside that interval; this appears to be a typo.","section":"Sec. IV"},{"comment":"References [5] and [7] appear to be the same article (Yankowitz et al., Science 363, 1059 (2019)) and should be merged or replaced with a distinct citation.","section":"References"},{"comment":"The definition P ≡ Nspan/N is ambiguous when more than one spanning cluster exists; please specify how Nspan is computed in such cases.","section":"Sec. II.C"},{"comment":"The phrase 'the dc sites have unit filling' near Fig. 3 appears to be a typo for 'cluster 1 sites' or 'the c=1 sites'.","section":"Sec. III"},{"comment":"The energy expression for uncoupled clusters in Eq. (3) omits the intra-layer hopping contribution that is later included as T_intra; this is appropriate for the 'uncoupled cluster' definition but should be stated more explicitly to avoid confusion.","section":"Sec. II.B"}],"recommendation":"major_revision","confidential_remarks":"The paper's central claims are interesting and potentially important, but the load-bearing phase identification needs substantially more evidence: at minimum a superfluid-stiffness or twist-energy calculation for the commensurate case, finite-size scaling for the incommensurate case, and sensitivity analyses for the thresholds. I would be happy to reconsider after these additions."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Here is my read on 2507.12153. The genuinely new piece is the cluster Gutzwiller approach and the phase diagrams it produces for interacting bosons in twisted-bilayer lattices, particularly the claim that interlayer interactions alone (no interlayer hopping) can create mobility islands for incommensurate twists. The formalism is clear enough to re-implement, the cluster construction is sensible, and the authors cite the relevant prior work including Refs. [25], [26], and [29], with appropriate distinction from [29].\n\nThe soft spot is the percolation cutoff. Two S=1 sites are considered disconnected when their coupling is below 10^-3 t⊥. In the commensurate case that is not reliable: the system is periodic, and any nonzero inter-island Josephson coupling should produce a global superfluid at T=0, even if weak. The Bose-glass-like phase in Fig. 2(a) is probably an artifact of discarding those weak links. The stress-test note lands correctly on this. The incommensurate case is on firmer physical ground because quasiperiodicity can localize, but the same cutoff-based percolation criterion is used there, and the cluster-family averaging coarse-grains the aperiodic potential. The mean-field cutoff |b|<0.05 is also arbitrary, and the paper never computes superfluid stiffness or compressibility, which are the proper order parameters for the SF/BG distinction. No finite-size scaling, no benchmark against exact small-system results. The watermark in the figures and a numerical typo also hint at rushed production.\n\nWhat the paper does well: the cluster decomposition is an appropriate way to handle the strongly coupled two-site pairs, and the phase diagrams, despite the cutoffs, identify parameter regimes experiments could probe. The t⊥=0 interaction-only case for incommensurate angles is the most defensible new result.\n\nWho should read it: cold-atom theorists and experimentalists working on twisted optical lattices. It is more a survey of possible phases than a quantitative prediction.\n\nRecommendation: send it to a serious referee, but expect a major-revision verdict. The authors need to benchmark against exact small-system results, perform finite-size scaling, test sensitivity to the cutoffs, and reassess the commensurate Bose glass. If the commensurate phase disappears but the incommensurate interaction-only islands survive, the paper still has value.","headline":"The cluster Gutzwiller phase diagrams are a useful map, but the commensurate Bose-glass phase probably dissolves once the weak inter-island links are included, and the numerics need benchmarking.","tokens_in":11396,"tokens_out":3225,"would_cite":false,"duration_ms":39161,"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":"Interacting bosons in twisted-bilayer optical lattices develop cluster-induced Mott-like and Bose-glass-like ground states, including non-percolating mobility islands that can form even at commensurate angles and even without inter-layer…","keywords":["twisted-bilayer optical lattices","Bose-Hubbard model","cluster Gutzwiller ansatz","Mott insulator","Bose glass","percolation","moire lattices","incommensurate twist"],"falsifier":"Compute, on a small twisted bilayer at $\\theta=\\theta(3,2)$, the superfluid stiffness and compressibility by an exact numerical method across the boundary where the percolation criterion switches from $P=0$ to $P>0$; if the region the paper labels Bose-glass shows nonzero stiffness, or the labelled Mott region shows finite compressibility, the central phase identification fails.","tokens_in":10384,"feed_emoji":"🌀","tokens_out":9075,"duration_ms":93830,"temperature":0.7,"pith_summary":"The paper studies interacting bosons in two stacked square optical lattices twisted by an angle $\\theta$, a geometry now reachable in cold-atom experiments. It argues that inter-layer tunneling splits the lattice into two-site clusters with different coupling strengths, and that this clusterization, captured by a tailored cluster Gutzwiller ansatz, generates a ground-state phase diagram far richer than the usual superfluid/Mott picture. The central findings are Mott-like insulating phases in which different clusters carry different integer fillings, and Bose-glass-like phases in which particles move freely inside disconnected pockets of sites but never percolate across the lattice. These glassy phases occur even at commensurate twist angles, and for incommensurate angles they can arise from inter-layer interactions alone, with no inter-layer hopping at all. The results matter because they identify clusterization, rather than disorder, as the organising principle behind the phases, giving experiments specific density-wave and island structures to look for.","feed_headline":"Twist alone makes a Bose glass in an optical lattice","feed_subtitle":"Inter-layer coupling splits sites into clusters, leaving particles mobile only inside pockets that never connect into a superfluid.","key_machinery":"The load-bearing object is the cluster Gutzwiller ansatz. Sites in opposite layers coupled by a hopping larger than a threshold $t_{\\perp,\\mathrm{cr}}$ ($0.03\\,t_\\perp$ for commensurate and $0.06\\,t_\\perp$ for incommensurate angles) are grouped into two-site clusters, each described by a variational state $|\\rho_c\\rangle=\\sum_{n_1,n_2} g_c(n_1,n_2)|n_1\\rangle|n_2\\rangle$ that can hold arbitrary correlated occupations of the two sites; the remaining non-clustered sites are treated by site-local Gutzwiller states. Minimising the resulting mean-field energy produces per-site order parameters $b_{\\alpha,s}=\\langle a_{\\alpha,s}\\rangle$, and a site is counted as superfluid ($S=1$) only if $|b|>0.05$. Two $S=1$ sites are considered connected if their coupling exceeds $10^{-3}t_\\perp$, so a percolation analysis using a standard cluster-labeling algorithm decides whether the $S=1$ sites span the lattice. Mott-like phases have zero superfluid fraction and zero percolation probability ($F=P=0$), superfluid phases have $F>0$ and $P>0$, and the intermediate non-percolating case $F>0$, $P=0$ is identified as the Bose-glass-like phase.","core_discovery":"The central claim is that site clusterization induced by inter-layer couplings is what governs the ground state of interacting bosons in a twisted bilayer. At commensurate angles the lattice contains a finite set of cluster types, each formed by two sites, one per layer, with a characteristic inter-layer hopping; the paper shows that these clusters fill sequentially and non-uniformly with chemical potential, producing Mott-like phases labelled by cluster populations such as $(1,1,1,2)$, and that between such lobes isolated superfluid 'quantum wheels' form and remain disconnected. At incommensurate angles the cluster families effectively acquire spread-out coupling strengths; the paper finds broad Bose-glass-like regions in which islands of nonzero superfluid mean field fail to percolate, and it shows that when inter-layer hopping is absent but inter-layer repulsion $V_\\perp$ is present, the aperiodic geometry alone produces the same glassy and superfluid phases. The authors' conclusion is that these Bose-glass-like phases are a generic consequence of clusterization in twisted geometries, with percolation of the superfluid islands marking the transition to superfluidity.","pith_inferences":["Replacing the $|b|>0.05$ cutoff with the superfluid stiffness would provide a cleaner test of the glassy classification; if the labelled Bose-glass islands carry finite stiffness, the phase boundaries would shift.","The same cluster decomposition could be applied to fermionic twisted bilayers by exchanging the two-site number basis for a spin or charge basis, potentially exposing analogous cluster Mott and island phases in Fermi gases.","Because the superfluid fraction $F$ and percolation probability $P$ increase monotonically with $t/U$, the paper's mechanism suggests a percolation-driven superfluid transition in the incommensurate case; measuring the critical $t/U$ as a function of twist angle would map the island-to-percolation crossover.","The $V_\\perp$-only Bose glass at incommensurate angles implies that even density-correlation measurements should show strong bunching in the strongly coupled clusters, a signature accessible to quantum-gas microscopy."],"forward_implications":["Cluster filling patterns such as $(1,1,1,2)$ imply that density-wave order is a generic signature of twisted-bilayer bosons; in-situ imaging should reveal clusters with distinct integer occupations in the same insulating lobe.","Bose-glass-like regions can appear in perfectly periodic commensurate lattices, so observing a compressible insulating island phase would not by itself signal disorder.","For incommensurate angles the phase diagram resembles that of a strongly disordered Bose–Hubbard model, so twisted bilayers can serve as a disorder-free quantum simulator of the Bose glass.","Because Bose-glass-like phases exist at $t_\\perp=0$ with only $V_\\perp\\neq0$, inter-layer interactions alone can localise bosons in an aperiodic lattice, a mechanism independent of single-particle potential disorder."],"supporting_citations":[{"why":"This reference supplies the twisted-bilayer optical potential model and the Gaussian forms of inter-layer hopping and interactions that the paper adopts.","marker":"[13]"},{"why":"This reference reports the experimental realisation of an atomic Bose–Einstein condensate in twisted-bilayer optical lattices, the platform the paper builds on.","marker":"[20]"},{"why":"This reference defines the Bose glass phase and its compressible-insulator character, which the paper uses to name its non-percolating island phase.","marker":"[22]"},{"why":"This reference introduces the mean-field percolation parameters $F$ and $P$ used to classify superfluid, Bose-glass, and Mott-like phases.","marker":"[25]"},{"why":"This reference documents the 'quantum wheels' in quasicrystalline potentials that the paper invokes as the commensurate-angle analogue of its isolated superfluid rings.","marker":"[26]"},{"why":"This reference shows that inter-layer interactions can produce a Bose glass without inter-layer hopping, the background result the paper extends to incommensurate angles.","marker":"[29]"},{"why":"This reference develops the superfluid-cluster percolation picture for disordered Bose–Hubbard models that motivates the island-percolation analysis.","marker":"[30]"},{"why":"This reference provides the cluster-labeling algorithm used to decide whether $S=1$ sites percolate across the lattice.","marker":"[32]"}],"fun_headline_variants":["Twist-induced clusters create Bose glass in bosonic lattice","Disconnected superfluid pockets emerge in twisted bilayer lattices","Clusterization turns twisted boson lattice into Bose glass","Even without hopping, twist's geometry makes a Bose glass","Inter-layer coupling splits sites, yields non-percolating superfluid"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The whole phase classification rests on treating a site as superfluid only when its local order parameter exceeds $|b|>0.05$ and two islands as connected only when their coupling exceeds $10^{-3}t_\\perp$, thresholds the paper does not verify against exact calculations, superfluid stiffness, or compressibility.","fun_headline_variants_meta":{"raw":{"variants":["Twist-induced clusters create Bose glass in bosonic lattice","Disconnected superfluid pockets emerge in twisted bilayer lattices","Clusterization turns twisted boson lattice into Bose glass","Even without hopping, twist's geometry makes a Bose glass","Inter-layer coupling splits sites, yields non-percolating superfluid"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.00052,"raw_usage":{"total_tokens":2509,"prompt_tokens":924,"completion_tokens":1585,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":540,"completion_tokens_details":{"reasoning_tokens":1503}},"tokens_in":540,"tokens_out":1585,"duration_ms":12317,"temperature":1.0,"reasoning_tokens":1503,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-06T16:52:30.548942+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Compute, on a small twisted bilayer at $\\theta=\\theta(3,2)$, the superfluid stiffness and compressibility by an exact numerical method across the boundary where the percolation criterion switches from $P=0$ to $P>0$; if the region the paper labels Bose-glass shows nonzero stiffness, or the labelled Mott region shows finite compressibility, the central phase identification fails.","supporting_citations":[{"cited_title":"Fractal spectrum in twisted bilayer optical lattice","cited_arxiv_id":"2404.08211","evidence_quote":"This reference reports the experimental realisation of an atomic Bose–Einstein condensate in twisted-bilayer optical lattices, the platform the paper builds on."},{"cited_title":"Fallani, J","cited_arxiv_id":null,"evidence_quote":"This reference introduces the mean-field percolation parameters $F$ and $P$ used to classify superfluid, Bose-glass, and Mott-like phases."},{"cited_title":"Johnstone, P","cited_arxiv_id":null,"evidence_quote":"This reference documents the 'quantum wheels' in quasicrystalline potentials that the paper invokes as the commensurate-angle analogue of its isolated superfluid rings."},{"cited_title":"Weak Superfluidity in Twisted Optical Potentials","cited_arxiv_id":"2406.12963","evidence_quote":"This reference shows that inter-layer interactions can produce a Bose glass without inter-layer hopping, the background result the paper extends to incommensurate angles."},{"cited_title":"Interaction induced moir\\'e systems in twisted bilayer optical lattices","cited_arxiv_id":"2405.20732","evidence_quote":"This reference develops the superfluid-cluster percolation picture for disordered Bose–Hubbard models that motivates the island-percolation analysis."},{"cited_title":"Hettiarachchilage, C","cited_arxiv_id":null,"evidence_quote":"This reference provides the cluster-labeling algorithm used to decide whether $S=1$ sites percolate across the lattice."}],"review_version":1}