{"id":"9cb07271-700b-4c4c-8d0d-d72e6cf6f082","arxiv_id":"2501.02248","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"A 120-atom defect-free mixed-species (85Rb/87Rb) optical tweezer array is assembled with 98.6% filling fraction and 14% full-array defect-free probability using an enhanced rearrangement algorithm with diagonal moves.","lead":"Researchers built defect-free arrays of 120 mixed-isotope rubidium atoms in optical tweezers, improving the filling fraction to 98.6% and the defect-free probability to 14% with upgraded hardware and diagonal moves in the rearrangement algorithm. The demonstration is a step toward two-species arrays large enough for quantum error correction and simulation.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Mixed-species defect-free claim rests on uncalibrated species assignment; a per-site misidentification rate near 0.3% could materially overstate both the 98.6% filling and 14% defect-free probability.","rationale":"I read the paper as a claim of a record-scale mixed-species defect-free array enabled by an upgraded setup and diagonal-move rearrangement. The experimental evidence is a 500-shot histogram and a cycle-dependence curve; no formal verification or raw data are included. The reader's weakest-assumption identification—uncalibrated species identification—is the same one I would choose, and I agree with the CONDITIONAL verdict. I considered the two other weaknesses noted by the reader: the 'five orders of magnitude' sentence is internally inconsistent (14% vs. 5% is a factor of ~3, not 10^5), and the averaging over cycles 4–10 is post hoc. Both are real but secondary; neither changes the central measurement as directly as an uncharacterized detection error would. I also checked the reported species-specific sub-array success rates (26% and 37%) against the combined 14%; the combined value is below the minimum, so there is no logical inconsistency there, and positive correlation is plausible. The single most load-bearing uncertainty is therefore the imaging/assignment fidelity, and a confusion-matrix measurement would settle it. The verdict should remain CONDITIONAL: accept the assembly procedure as demonstrated, but do not take the headline defect-free numbers at face value until the detection calibration is provided.","tokens_in":11923,"tokens_out":5271,"duration_ms":55302,"concrete_test":"Obtain the raw single-shot images behind Fig. 4(c) and compute a confusion matrix for species assignment. Concretely: (1) load single-species arrays of pure 87Rb and pure 85Rb and run the exact two-color imaging sequence used in Section III; the fraction of atoms detected in the wrong color channel gives the misassignment floor. (2) On an assembled mixed array, take two consecutive images without rearrangement and measure per-site flip rates in species labels and atom/no-atom classifications; this bounds classification noise. (3) Recompute the 120-site filling fraction and defect-free probability after excluding runs where the assignment is ambiguous by the measured threshold. If the per-site misassignment or flip probability exceeds 0.1%, the reported 14(2)% needs correction; if it is below 0.01%, the central claim is supported.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim—120-site mixed-species arrays with 98.6(1)% filling and 14(2)% defect-free probability—is established only through fluorescence images in which each site is classified as 87Rb, 85Rb, or empty. Section III and Fig. 4 give the resulting histograms but no measurement of the species-assignment error. The two isotope images are taken in successive exposures, so the relevant quantity is the per-site, per-shot probability that an occupied site is assigned the wrong species or that noise is counted as an atom. This is not supplied, and the paper's discussion of 'cross-talk heating' is about loss, not about classification. The headline numbers are exponentially sensitive to such errors: for 120 sites, an extra 0.3% apparent success per site multiplies the measured defect-free probability by about (0.989/0.986)^120 ≈ 1.44, so a 14% measurement could correspond to a true value near 10% if misassignment contributes at the 0.3% level. A misidentification rate at the same order as the 1.4% vacancy rate would therefore change the qualitative conclusion. The adjacent 'five orders of magnitude' comparison is arithmetically wrong (14% vs. ~5% is a factor of ~3), which further motivates an independent check of the raw data. Conditional acceptance is appropriate until a species-detection confusion matrix is reported.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper reports substantial experimental upgrades to a mixed-isotope rubidium tweezer array: a 20x20 static array produced by an 830 nm laser and SLM, improved movable tweezers with lower phase-noise driving, and an enhanced heuristic heteronuclear algorithm (HHA) that adds diagonal moves and supports multiple rearrangement cycles. The central result is the creation of defect-free two-dimensional mixed-species arrays of 120 single atoms (60 87Rb and 60 85Rb) in checkerboard and zebra patterns, with a claimed filling fraction of 98.6(1)% and defect-free probability of 14(2)% obtained by averaging over rearrangement cycles 4 to 10. The authors attribute the remaining limitation to cross-talk heating during two-species fluorescence imaging.","tokens_in":12182,"tokens_out":3956,"duration_ms":36025,"significance":"If the headline numbers are correct, this is the largest defect-free mixed-species atom array demonstrated to date and a clear step toward two-species platforms for quantum error correction and Rydberg-mediated quantum simulation. The work has real strengths: the headline statistics come from 500 experimental repetitions and multiple rearrangement cycles; the result is a direct measurement rather than a derived quantity with fitted parameters; and the paper includes a comparison of the enhanced HHA with the earlier HCOA algorithm, supporting the claim that rearrangement losses are not the dominant bottleneck. The enhanced algorithm is described in enough detail to be a useful contribution on its own. However, the quantitative centrality of the species-assignment fidelity means the missing calibration is a load-bearing issue rather than a presentation detail.","major_comments":[{"comment":"The central claims of 98.6(1)% filling fraction and 14(2)% defect-free probability depend on classifying each of 120 sites as 87Rb, 85Rb, or empty from successive fluorescence images. The paper reports no calibration of the species-assignment fidelity, no false-positive/false-negative analysis, and no error model for the classification. Because the defect-free probability is exponentially sensitive to per-site errors, a 0.3% per-site apparent success inflation would change the measured 14% by a factor (0.989/0.986)^120 ≈ 1.44, and a misidentification rate comparable to the 1.4% vacancy rate would qualitatively change the conclusion. The authors should report a confusion matrix for the species assignment, for example from sites with known occupation or from independently prepared single-species arrays, and propagate the resulting uncertainties into the headline numbers.","section":"Section III, Fig. 4"},{"comment":"The statement that the measured defect-free probability is 'approximately five orders of magnitude higher' than in the previous work is arithmetically inconsistent. The previous single-cycle probability was about 5%, and the present 14(2)% corresponds to a factor of roughly 3; even comparing the single-cycle 1.3% with the multi-cycle 14% gives a factor of about 10, not 10^5. This comparison should be corrected or removed.","section":"Section III, 'five orders of magnitude' sentence"},{"comment":"The text states that the defect-free probability increases rapidly, reaches a saturation point of 13.7%, and then exhibits fluctuations, yet the headline value 14(2)% is the average over cycles 4 to 10. The choice of this averaging window is not justified. The authors should either report the cycle-by-cycle values, state a prespecified selection criterion, or present the saturation value with its uncertainty instead of an average over an unexplained subset of the data.","section":"Section III, Fig. 4(c) and averaging window"}],"minor_comments":[{"comment":"There are several typos and infelicities: 'revivor' should likely be 'reservoir', 'braodband' should be 'broadband', 'arbitratry' should be 'arbitrary', and 'Gerchber Saxton' should be 'Gerchberg-Saxton'.","section":"Section II"},{"comment":"The heading reads 'DEFECT-FREE HOMONUCLEAR ARRA Y ASSEMBL Y', but the section reports heteronuclear array assembly; this appears to be a typo and should be corrected.","section":"Section III heading"},{"comment":"The text says the enhanced HHA reduces the number of moves by 50% 'for making a defect-free 120-atom arrays', while Fig. 3(d) plots moves as a function of 'Size of Array' for N x N arrays. Please clarify whether the horizontal axis is the number of filled sites or the total array size, and state the simulation parameters used for the 120-atom comparison.","section":"Section III, Fig. 3(d)"},{"comment":"The sentence 'the success rate for defect-free 87Rb(85Rb) atom sub-array is 26% (37%)' introduces a quantity that is not defined in the text; please define how a sub-array success rate is computed and how it relates to the global defect-free probability.","section":"Section III, sub-array success rates"}],"recommendation":"major_revision","confidential_remarks":"The paper is a solid experimental advance and the central assembly claim is likely correct, but the missing species-assignment calibration is a load-bearing omission for the quantitative headline results. The 'five orders of magnitude' arithmetic error further undermines confidence in the quantitative framing. Both issues are fixable within the scope of the manuscript, so I recommend major revision rather than rejection."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Short version: this is a genuine experimental step forward—120-site defect-free mixed-isotope array, 60+60, with a diagonal-move extension of the HHA and multi-cycle correction. The 98.6(1)% filling and 14(2)% defect-free numbers come from 500 repetitions and are plausible. The paper is worth taking seriously.\n\nWhat's new: the algorithm change is simple but useful; simulation shows roughly 50% fewer moves for 120-atom arrays, and the multi-cycle correction is what lets them push past the single-cycle 1.3% defect-free probability. They also improved vacuum lifetime to 29 s and transport waveforms. Credit where due: they compare against their own HHA and HCOA, and the equal filling fraction for both algorithms supports the claim that rearrangement loss is not the bottleneck.\n\nSoft spots, in order. First, the 'five orders of magnitude' sentence is arithmetically wrong: 14% versus roughly 5% is a factor of about 3, not five orders. That kind of error makes me check everything else. Second, the species-assignment fidelity is not reported. Fluorescence images are classified into 87Rb, 85Rb, and empty, with successive exposures, but there is no confusion matrix or false-positive estimate. The stress-test arithmetic is fair: a 0.3% per-site misassignment would inflate the defect-free probability by about 1.4 times. I don't think that kills the result—standard isotope discrimination in tweezer experiments is usually well above 99%—but the paper should say so. Third, the averaging window over cycles 4-10 is unexplained. Why not 3-10 or 5-10? The plot shows saturation then fluctuations, so the choice matters. Fourth, the claim that residual imaging heating limits the defect-free probability is inferred from observing cross-talk heating, not measured directly. Plausible, but it is an inference.\n\nThe missing raw data and code are a minor concern for a hardware paper; the 500-repetition statistics are the main support. This paper is for experimental atomic physics and quantum computing groups working on atom arrays; theory readers will mostly care about the algorithm.\n\nVerdict: send it to peer review. The central demonstration is a real milestone and the algorithm contribution is transferable. A referee should ask for the species-detection calibration, a fixed averaging window, and a corrected comparison sentence. Those are revisions, not red flags.","headline":"A real 120-atom mixed-species assembly milestone, with a useful diagonal-move algorithm upgrade; send to review, but ask for species-assignment calibration and fix an arithmetic error.","tokens_in":12777,"tokens_out":3032,"would_cite":true,"duration_ms":30653,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"The paper reports defect-free mixed-species atom arrays of 120 atoms, with 98.6% filling and 14% defect-free probability.","keywords":["atomic arrays","optical tweezers","mixed-species arrays","defect-free assembly","atom rearrangement","heuristic heteronuclear algorithm","85Rb","87Rb"],"falsifier":"Repeatedly image a stationary 120-atom mixed-species array under the same heteronuclear fluorescence sequence and count how often a site's assigned species changes between exposures; if the per-site misidentification rate is comparable to the reported 1.4% vacancy fraction, the filling fraction and defect-free probability are overestimated.","tokens_in":11732,"feed_emoji":"⚛️","tokens_out":6053,"duration_ms":55546,"temperature":0.7,"pith_summary":"This paper reports a route to building two-dimensional arrays of two different atomic species with no empty or misplaced sites. Starting from a 20-by-20 optical-tweezer reservoir, the authors use a movable tweezer to sort 60 atoms of 85Rb and 60 atoms of 87Rb into checkerboard and zebra patterns, achieving a filling fraction of 98.6(1)% and a defect-free probability of 14(2)% after repeated rearrangement cycles. The advance that makes this possible is an upgraded sorting algorithm that lets the moving tweezer travel diagonally as well as along rows and columns, preventing atoms from blocking one another and allowing multiple repair cycles. If the result stands, it gives mixed-species atom arrays a path beyond the tens-of-atoms scale, where independent control and readout of two species could support quantum error correction and Rydberg-mediated many-body studies.","feed_headline":"Two-species atom array hits 120 defect-free atoms","feed_subtitle":"New diagonal-move sorting lifts filling fraction to 98.6% and defect-free probability to 14%.","key_machinery":"The load-bearing object is the enhanced heuristic heteronuclear algorithm (HHA), a sorting routine that plans the near-fewest moves of a single movable optical tweezer to fill misplaced and empty sites in a dual-species array. Its key addition is diagonal movement: the movable tweezer can travel along eight azimuthal directions, instead of only along the rows and columns of the static array, by jointly modulating the rf signals driving two perpendicular acousto-optic deflectors. This lets atoms move past one another and eliminates logjams, and it makes repeated rearrangement cycles practical; the authors report that the required number of moves for a 120-atom target is reduced by about 50% relative to the original HHA in simulation.","core_discovery":"The paper claims that defect-free two-dimensional mixed-species atom arrays can be scaled from the previous 64-tweezer proof of principle to 120 atoms by combining three upgrades: a 400-site static tweezer array generated by a spatial light modulator, a low-phase-noise arbitrary waveform generator that improves atom transport fidelity, and an enhanced heuristic heteronuclear algorithm (HHA) that adds diagonal tweezer trajectories to the original row-and-column moves. With these changes the authors report assembling arrays of 60 85Rb and 60 87Rb atoms in checkerboard and zebra geometries, with filling fraction 98.6(1)% and defect-free probability 14(2)% averaged over rearrangement cycles four through ten, up from a single-cycle probability of 1.3% for the checkerboard pattern. They also report that the remaining limit is not atom loss during rearrangement but residual heating from the sequential fluorescence imaging needed to identify the two isotopes.","pith_inferences":["A natural testable extension is to interleave mid-circuit readout or continuous reloading with the multi-cycle algorithm; if per-cycle atom loss is as low as the paper implies, the 14% defect-free probability could be compounded toward near-certain assembly.","The imaging-heating limit suggests that switching to a single-shot, species-resolving readout, such as isotope-selective shelving, could raise both the filling fraction and the defect-free probability beyond the reported values.","Diagonal moves may also reduce transport distances in single-species assembly, where the same logjam argument applies, potentially lowering the heating budget for very large arrays."],"forward_implications":["Defect-free mixed-species arrays of 120 atoms, 60 per species, can now be produced with a single apparatus, making dual-isotope platforms competitive with single-species arrays at the hundred-atom scale.","Because rearrangement loss is reported to be negligible, further progress hinges on reducing fluorescence-imaging heating, for example through gentler or species-selective readout.","The enhanced HHA's diagonal moves should transfer directly to other two-species combinations, since the algorithm is not specific to rubidium isotopes.","Multiple rearrangement cycles are now viable for mixed-species assembly, with defect-free probability saturating near 14% after four cycles rather than improving indefinitely."],"supporting_citations":[{"why":"Original proof-of-principle HHA and 64-tweezer dual-isotope array; the baseline this work scales up.","marker":"[39]"},{"why":"Prior dual-isotope Yb defect-free array with 1% success; the comparison showing the need for better transport and moves.","marker":"[61]"},{"why":"HCOA algorithm used as the numerical and experimental comparison for filling fraction and move distance.","marker":"[62]"},{"why":"Cited for the limitation of restricting moves to rows and columns, motivating diagonal trajectories.","marker":"[63]"},{"why":"Heuristic cluster algorithm for near-fewest single-species moves; used to validate the upgraded movable tweezer.","marker":"[64]"},{"why":"Single-species assembly of more than 100 atoms; the benchmark for the homonuclear test with one rearrangement cycle.","marker":"[51]"}],"fun_headline_variants":["Defect-free mixed array grows to 120 atoms","120 mixed-species atoms assembled defect-free","Diagonal moves lift atom array to 98.6% fill","Two-species array scales to 120 defect-free atoms","Mixed-species array expands to 120 atoms"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The 98.6% filling fraction and 14% defect-free probability depend on the fluorescence imaging correctly identifying which isotope occupies every site, and the paper reports no calibration of that species-identification error rate.","fun_headline_variants_meta":{"raw":{"variants":["Defect-free mixed array grows to 120 atoms","120 mixed-species atoms assembled defect-free","Diagonal moves lift atom array to 98.6% fill","Two-species array scales to 120 defect-free atoms","Mixed-species array expands to 120 atoms"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000947,"raw_usage":{"total_tokens":4070,"prompt_tokens":996,"completion_tokens":3074,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":612,"completion_tokens_details":{"reasoning_tokens":2998}},"tokens_in":612,"tokens_out":3074,"duration_ms":21692,"temperature":1.0,"reasoning_tokens":2998,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-10T22:14:08.883756+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Repeatedly image a stationary 120-atom mixed-species array under the same heteronuclear fluorescence sequence and count how often a site's assigned species changes between exposures; if the per-site misidentification rate is comparable to the reported 1.4% vacancy fraction, the filling fraction and defect-free probability are overestimated.","supporting_citations":[{"cited_title":"Sheng, J.-Y","cited_arxiv_id":null,"evidence_quote":"Original proof-of-principle HHA and 64-tweezer dual-isotope array; the baseline this work scales up."},{"cited_title":"Nakamura, T","cited_arxiv_id":null,"evidence_quote":"Prior dual-isotope Yb defect-free array with 1% success; the comparison showing the need for better transport and moves."},{"cited_title":"Tao, L.-G","cited_arxiv_id":null,"evidence_quote":"HCOA algorithm used as the numerical and experimental comparison for filling fraction and move distance."},{"cited_title":"Schymik, V","cited_arxiv_id":null,"evidence_quote":"Cited for the limitation of restricting moves to rows and columns, motivating diagonal trajectories."},{"cited_title":"Sheng, J.-Y","cited_arxiv_id":null,"evidence_quote":"Heuristic cluster algorithm for near-fewest single-species moves; used to validate the upgraded movable tweezer."},{"cited_title":"Ohl de Mello, D","cited_arxiv_id":null,"evidence_quote":"Single-species assembly of more than 100 atoms; the benchmark for the homonuclear test with one rearrangement cycle."}],"review_version":1}