{"id":"d49b1cd7-bed9-4913-871a-6cd236c88a3a","arxiv_id":"2501.09554","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":6,"one_line_summary":"Crosstalk between parallel entangling gates in a 2D trapped-ion chip requires a distance-5 surface code, and with optimized parallelism logical error rates below 10^-10 are reachable with hundreds to thousands of ions.","lead":"This paper analyzes how crosstalk between parallel two-qubit gates in 2D trapped-ion arrays affects quantum error correction. It finds that a distance-5 surface code is the minimum that can handle this noise, and that optimized parallelism can reach logical error rates below 10^-10 with realistic hardware.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Eq. (12) uses a threshold-slope exponent (d+1)/2 fit to a depolarizing crosstalk model; it is not an upper bound for the paper's own d=5 XX-crosstalk simulation, so the 10^-10 scaling claims are likely optimistic.","rationale":"Good-faith summary: the paper's central claim that d≥5 is needed for the rotated surface code under two-qubit crosstalk is directly supported by Fig. 2, and the randomized-compiling step used to justify the Pauli-XX model is standard (Refs. [56,57]); I do not regard that as the main risk. The load-bearing weakness is the quantitative scaling law Eq. (12). The appendix explicitly fits it to a depolarizing crosstalk model and near-threshold slopes, and the paper itself flags the derivation as 'hand-waving' and notes the asymptotic exponent is ⌈d/4⌉. The formula then fails as an upper bound against the paper's own d=5 XX simulation. Because the abstract and Sec. V use Eq. (12) to claim 1e-10 logical error rates with hundreds-to-thousands of ions, this is a direct threat to the quantitative conclusions, although not to the distance-5 necessity result. The reader's conditional verdict remains appropriate; the scaling analysis needs to be redone with the correct error model and a valid bound. Hence verdict unchanged, with partial agreement on the weakest assumption.","tokens_in":23471,"tokens_out":21524,"duration_ms":206867,"concrete_test":"Run Stim with the XX crosstalk model (as in Fig. 2) for d=5,9,13,17, sweeping p~c down to pL ~ 1e-8, and extract the local exponent δ = d(log pL)/d(log p~c). If δ approaches ⌈d/4⌉ rather than (d+1)/2 at low pL, Eq. (12) is invalid in the target regime; also verify whether the RHS of Eq. (12) is ≥ simulated pL at every sampled point. A minimal check: evaluate Eq. (12) at the Fig. 2(a) pseudothreshold parameters and confirm it exceeds the simulated pL.","verdict_should_be":"UNCHANGED","load_bearing_attack":"In Fig. 2(a), the XX-crosstalk simulation shows pL ∝ p^2 for d=5 with a pseudothreshold at p ≈ 1.3e-5. Substituting the same parameters into Eq. (12) with p~c = 2(d-1)p = 38p and the idling term lower-bounded as 3t/8T ≥ p/2 gives RHS ≈ 0.015[(50.9p)/0.013]^3 ≈ 2e-6 at p=1.3e-5, an order of magnitude below the simulated pL = 1.3e-5. Hence Eq. (12) is not an upper bound, even at d=5. The origin is that Appendix A fits Eq. (12) to a two-qubit depolarizing crosstalk model (Fig. 6) using slopes extracted near the error threshold (Fig. 7), while the physical crosstalk is the XX Pauli channel used in the main text, and the target pL < 1e-10 lies far below threshold where the paper itself notes the asymptotic exponent is ⌈d/4⌉, not (d+1)/2. Since the bracket in Eq. (12) is below 1, the inflated exponent makes the RHS too small and the code distances d=17 and d=41 in Fig. 5(d) optimistic.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper analyzes crosstalk errors between parallel entangling gates in a 2D ion crystal and their impact on rotated surface code quantum error correction. It argues that the phonon-mediated two-qubit crosstalk considered here cannot be corrected by a distance-3 code, so a distance-5 code is necessary, and it supports this claim with Stim simulations of d=3 and d=5 rotated surface codes, showing pL ~ p for d=3 and pL ~ p^2 with a pseudothreshold for d=5. The paper also optimizes the parallelism level by balancing idling errors against crosstalk, studies slow- and fast-gate spatial regimes with concrete EASE pulse sequences, and proposes a unified scaling formula, Eq. (12), to predict that logical error rates below 10^-10 can be reached with distance d=17 or d=41 in a 2D crystal of hundreds to thousands of ions.","tokens_in":23852,"tokens_out":11232,"duration_ms":112126,"significance":"The main qualitative conclusion, that phonon-mediated two-qubit crosstalk forces a distance-5 rotated surface code rather than distance-3, is well supported by direct simulation under the stated stochastic Pauli noise model, and it is practically relevant for 2D ion-crystal QEC. The paper has concrete strengths: it uses a physically motivated crosstalk model, performs stabilizer-circuit simulations with Stim, designs actual pulse sequences with the EASE protocol for both slow- and fast-gate regimes, and releases the pulse-sequence data on figshare. The quantitative extrapolation to large code distances, however, rests on Eq. (12), a fitted formula that is presented as an upper bound but is not conservative even at d=5; the large-distance resource claims are therefore not yet established.","major_comments":[{"comment":"Eq. (12) is stated as an upper bound, but it is not an upper bound for the paper's own d=5 simulation. At the pseudothreshold of Fig. 2(a), taking p ≈ 1.3e-5 with pg=pi=pc=p and full parallelism k=d(d-1)=20 (so t ≈ 9 and p~c = 2(k-1)p = 38p), Eq. (12) gives pL ≈ 0.015[(p + 4.5p + 1.3*38p)/0.013]^3 ≈ 2.5e-6, whereas Fig. 2(a) reports pL ≈ 1.3e-5 at the same point. The formula under-predicts the simulated logical error rate by about a factor of five. This originates from the fact that Appendix A fits Eq. (12) to the depolarizing-crosstalk model of Fig. 6, while the main-text simulations use the XX Pauli crosstalk channel of Sec. II B. Consequently the extrapolations to d=17 and d=41 in Fig. 5(d), and the claim that pL < 1e-10 is reachable, are not currently supported.","section":"Section V, Eq. (12), Appendix A"},{"comment":"The exponent (d+1)/2 in Eq. (12) is extracted from slopes near threshold, but the paper itself notes in Appendix A that the asymptotic scaling of the crosstalk contribution is pL ~ pc^{⌈d/4⌉}. The target pL < 1e-10 is far below threshold, where the near-threshold exponent is not the relevant one. Since the bracket in Eq. (12) is below one in the extrapolated regime, replacing the crosstalk exponent with the asymptotic ⌈d/4⌉ value changes the required code distances; at minimum, the authors should justify the larger exponent for the deep-sub-threshold regime or present a conservative bound using the asymptotic exponent for the crosstalk term.","section":"Appendix A, Fig. 7"},{"comment":"The conversion of coherent crosstalk e^{i ε_ij X_i X_j} into a stochastic XX error with probability pc = ε_ij^2 relies on randomized compiling, but no explicit twirling circuit is given and no estimate of the residual coherent component is provided. This assumption is load-bearing: the d=5 necessity, the pL ~ p^2 scaling, and Eq. (12) all use a stochastic Pauli channel. A direct simulation with a coherent crosstalk rotation, or an explicit randomized-compiling sequence with finite sampling, is needed to confirm that the conclusions survive partial or imperfect twirling.","section":"Sec. II B"}],"minor_comments":[{"comment":"The notation 'J25, 1, 5K' in the text appears to be a rendering error for [[5,1,5]]; please correct the typesetting.","section":"Sec. III"},{"comment":"Fig. 5(b) reports an average crosstalk pc ≈ 1.1e-5, but the scaling curves in Fig. 5(d) are computed with pc = 1e-5; the rounding should be stated explicitly in the caption or text.","section":"Fig. 5 and Sec. V"},{"comment":"The statement that a distance-5 code is 'necessary' should be qualified as necessary for the rotated surface code under MWPM decoding with this crosstalk model; the evidence does not establish impossibility for all distance-3 codes or all decoders.","section":"Abstract and Sec. III"},{"comment":"The caption should state explicitly that both curves are predictions from Eq. (12) for d > 5 and are not direct simulations at those distances.","section":"Fig. 5(d)"}],"recommendation":"major_revision","confidential_remarks":"The paper fits the journal's scope and the distance-5 result is likely worth publishing after revision. The main concern is the scaling-law overclaim: Eq. (12) is presented as a bound but is neither a rigorous bound nor consistent with the d=5 XX-crosstalk simulation. The authors can address this by reframing Eq. (12) as a heuristic fit, adding a conservative bound or higher-distance simulations, and validating the randomized-compiling assumption. I see no citation or novelty issues."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Two things to know. First, the core claim that phonon-mediated two-qubit crosstalk forces a distance-5 surface code is real, and the direct simulations for d=3 vs d=5 back it up. Second, the paper's long-term scaling story -- reaching 1e-10 with d=17 or d=41 -- is likely optimistic, because Eq. (12) is fit to a two-qubit depolarizing crosstalk model, not the XX Pauli channel used in the main text, and it under-predicts the paper's own d=5 simulation by about a factor of six at the pseudothreshold. What's new: previous crosstalk work mostly considered laser-beam spillover or distance-3 codes. This paper treats the intrinsic phonon-mediated crosstalk between parallel entangling gates on a 2D crystal, shows d=3 cannot handle it (pL proportional to p, always above the physical error rate), and demonstrates a genuine pseudothreshold for d=5. The parallelism optimization in Sec. IV is a useful design rule: for short coherence times full parallelism wins, for long T serial wins, and there is an intermediate optimum. The slow vs fast gate regime split, with the EASE pulse sequence examples and a figshare deposit of the pulse data, is a real contribution. The small-distance simulations are internally consistent. Soft spots, in order. (1) Appendix A explicitly switches to a depolarizing crosstalk channel to fit Eq. (12), while the main text's physics is the XX channel. That is not a cosmetic choice: at p=1.3e-5, Eq. (12) gives pL <= ~2e-6 for d=5, whereas Fig. 2(a) shows pL = 1.3e-5. An inequality that fails at d=5 cannot support quantitative claims at d=41. (2) The exponent (d+1)/2 is extracted near threshold, but the target pL < 1e-10 is far below threshold; the authors themselves note the asymptotic exponent should be ceil(d/4). With the bracket below 1, using the larger exponent shrinks the RHS dramatically. The d=17 and d=41 curves are therefore best read as upper-bound placeholders, not predictions. (3) The randomized-compiling argument converts coherent crosstalk to incoherent XX errors; residual coherence would make the errors worse than the depolarizing model. This is a modeling assumption worth flagging, not a fatal flaw. The main message -- distance-5 is the practical minimum for this crosstalk -- survives. The scaling analysis needs rework before publication. I'd send it to peer review with a major-revision request focused on Eq. (12) and the large-distance extrapolations. Trapped-ion experimentalists and QEC theorists planning parallel gates will want this paper; I'd cite it for the distance-5 result, not for the 1e-10 numbers.","headline":"The distance-5 necessity claim is solid and useful, but the scaling extrapolation to 1e-10 logical error rates is built on a fitted formula that already fails against the paper's own d=5 data.","tokens_in":24356,"tokens_out":4479,"would_cite":true,"duration_ms":37980,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["81P68","81P73"],"pacs":[],"model":"deepseek-v4-flash","headline":"Crosstalk between parallel entangling gates in trapped ions is a two-qubit error that requires a distance-5 surface code.","keywords":["trapped-ion quantum computing","crosstalk errors","parallel entangling gates","surface code","quantum error correction","randomized compiling","logical error rate","two-dimensional ion crystals"],"falsifier":"Run the same EASE parallel-gate pulses on a small two-dimensional ion crystal, with and without randomized compiling, and perform quantum process tomography on two simultaneous gate pairs; if the crosstalk channel retains coherent off-diagonal terms, or if the measured $d=3$ logical error rate does not scale linearly with $p_c$ while the $d=5$ rate does not scale quadratically, the central claim fails.","tokens_in":23300,"feed_emoji":"⚛️","tokens_out":8912,"duration_ms":81215,"temperature":0.7,"pith_summary":"The paper establishes that the crosstalk between entangling gates run in parallel on a two-dimensional trapped-ion crystal is a correlated two-qubit error, and that a rotated surface code of distance 3 cannot correct it even in principle. Using a distance-5 code, one crosstalk event is correctable, so the logical error rate acquires a quadratic suppression and a pseudothreshold (where the logical qubit beats the physical qubits) can exist. The paper simulates logical error rates and coherence times under crosstalk, gate infidelity, and idling errors, and shows that the optimal number of parallel gate pairs balances crosstalk against idling. It also derives a unified scaling bound for the logical error rate and argues that both slow-gate and fast-gate regimes can reach logical error rates below $10^{-10}$ with hundreds to thousands of ions. The practical consequence is that near-term trapped-ion error correction should plan for at least distance 5 once parallel entangling gates are used.","feed_headline":"Parallel ion-gate crosstalk forces a distance-5 code","feed_subtitle":"Phonon-mediated two-qubit crosstalk defeats distance-3 codes; d=5 can reach break-even and 10^-10 logical error rates.","key_machinery":"The load-bearing objects are the rotated surface code with its four-layer CNOT schedule and the phonon-mediated Mølmer–Sørensen gate model. The CNOT schedule is ordered so that single-qubit errors propagate perpendicular to the logical operators; a two-qubit crosstalk event between ancilla gates becomes a weight-4 data-qubit error, which the schedule keeps correctable only when the code distance is at least 5. The gate model writes the evolution as spin-dependent displacements plus pair phases $\\Theta_{ij}$, so an undesired phase is the crosstalk and residual spin-phonon entanglement becomes dephasing; randomized single-qubit rotations convert the coherent phase into a stochastic Pauli $XX$ error. The EASE algorithm supplies the amplitude-modulated pulses used to compute realistic crosstalk, and the unified bound in Eq. (12) combines gate infidelity, idling error, and crosstalk per gate $\\tilde p_c=2(k-1)p_c$ into a single distance-dependent formula.","core_discovery":"The central claim is that phonon-mediated crosstalk between parallel entangling gates is a two-qubit $XX$-type error that a distance-3 surface code cannot correct, making distance 5 the minimum code for a pseudothreshold. The argument is made concrete by simulation of a $d=5$ rotated surface code: for equal physical error rates $p$, the logical error rate scales as $p_L \\propto p$ for $d=3$ and $p_L \\propto p^2$ for $d=5$, with a pseudothreshold near $p\\approx 1.3\\times10^{-5}$. Once the coherent crosstalk phase is converted to a stochastic Pauli error by randomized compiling, the crosstalk enters as a two-qubit error with probability $p_c=\\epsilon_{ij}^2$. Under EASE-designed parallel gates with realistic pulse noise, the average crosstalk is $p_c\\approx1.1\\times10^{-5}$, and the paper finds a break-even point where logical coherence time equals physical coherence time over a wide parameter range. The unified scaling bound $p_L \\le 0.015[(p_g + 3t/8T + 1.3\\tilde p_c)/0.013]^{(d+1)/2}$ then implies that logical error rates below $10^{-10}$ are reachable with a distance-41 code in the slow-gate regime or a distance-17 code in the fast-gate regime.","pith_inferences":["If randomized compiling is imperfect, residual coherence in the crosstalk would make the two-qubit rotation more damaging than the depolarizing model; an experiment that compares $d=3$ and $d=5$ logical error rates under full process tomography of the crosstalk channel would reveal this directly.","The fitted $r^{-5.87}$ decay of crosstalk in the fast-gate regime is a concrete prediction: a two-dimensional crystal experiment measuring undesired two-qubit phase as a function of ion-pair distance could confirm the boundary between the slow- and fast-gate regimes.","The scaling law suggests that codes with nonlocal connectivity, such as quantum LDPC codes laid out on the same crystal, could inherit the same crosstalk protection with fewer physical qubits; the paper notes long-range gates as an outlook without quantifying this.","Reporting crosstalk per parallel layer rather than per gate pair would make fault-tolerance comparisons clearer, since the effective crosstalk per gate grows as $2(k-1)p_c$ with the number of parallel pairs."],"forward_implications":["A distance-3 rotated surface code has no pseudothreshold under two-qubit crosstalk: once $p_c$ is fixed, $p_L$ is lower-bounded by a value proportional to $p_c$ no matter how low gate infidelity and idling errors go.","The optimal parallelism level is not full parallelism: with spatially uniform crosstalk, choosing $k=O(d)$ parallel gate pairs balances the $O(d^2/k)$ idling error against the $O(k)$ crosstalk contribution.","In the fast-gate regime, crosstalk decays polynomially with ion-pair distance, and sublattice scheduling with $l=4$ gives $\\tilde p_c=10^{-6}$, making a distance-17 code sufficient for $p_L<10^{-10}$.","Under realistic parameters with $p_g=10^{-3}$ and physical coherence times of $10^3$ to $10^5$ gate times, the optimized $d=5$ code exceeds the break-even point $T_L=T$ for a wide range of crosstalk rates.","The slow-gate regime requires a larger code, $d=41$ with thousands of ions, but still reaches the $10^{-10}$ logical error goal."],"supporting_citations":[{"why":"Supplies the EASE pulse-design algorithm used to construct the parallel entangling gates and to estimate crosstalk under amplitude and frequency noise.","marker":"[31]"},{"why":"Justifies converting coherent crosstalk rotations into stochastic Pauli errors by randomized single-qubit gates; the whole noise model depends on this conversion.","marker":"[56, 57]"},{"why":"Provides the spin-dependent-force Hamiltonian and the unitary form whose unwanted two-qubit phases define crosstalk and whose residual displacements define dephasing.","marker":"[52]"},{"why":"Supplies the rotated-surface-code CNOT scheduling and the standard asymptotic logical-error scaling that Eq. (12) extends to include crosstalk.","marker":"[43]"},{"why":"Provides the stabilizer-circuit simulator used for the numerical logical-error-rate results in the main figures and Appendix A.","marker":"[47]"},{"why":"Underpins the minimum-weight perfect matching decoder used in all syndrome-decoding simulations.","marker":"[45, 46]"}],"fun_headline_variants":["Crosstalk in parallel ion gates demands d=5 code","Trapped-ion crosstalk: distance-3 fails, d=5 works","Phonon-mediated crosstalk pushes surface code to d=5","Parallel ion-gate crosstalk forces larger code distance"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"Everything downstream assumes that randomized compiling fully converts the coherent crosstalk phase into a stochastic Pauli $XX$ error; if residual coherent two-qubit rotation remains, the crosstalk is stronger than the model and distance 5 may not be sufficient.","fun_headline_variants_meta":{"raw":{"variants":["Crosstalk in parallel ion gates demands d=5 code","Trapped-ion crosstalk: distance-3 fails, d=5 works","Phonon-mediated crosstalk pushes surface code to d=5","Parallel ion-gate crosstalk forces larger code distance"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000289,"raw_usage":{"total_tokens":1762,"prompt_tokens":1085,"completion_tokens":677,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":701,"completion_tokens_details":{"reasoning_tokens":601}},"tokens_in":701,"tokens_out":677,"duration_ms":6275,"temperature":1.0,"reasoning_tokens":601,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-10T19:53:59.349428+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Run the same EASE parallel-gate pulses on a small two-dimensional ion crystal, with and without randomized compiling, and perform quantum process tomography on two simultaneous gate pairs; if the crosstalk channel retains coherent off-diagonal terms, or if the measured $d=3$ logical error rate does not scale linearly with $p_c$ while the $d=5$ rate does not scale quadratically, the central claim fails.","supporting_citations":[{"cited_title":"Grzesiak, R","cited_arxiv_id":null,"evidence_quote":"Supplies the EASE pulse-design algorithm used to construct the parallel entangling gates and to estimate crosstalk under amplitude and frequency noise."},{"cited_title":"Leibfried, B","cited_arxiv_id":null,"evidence_quote":"Provides the spin-dependent-force Hamiltonian and the unitary form whose unwanted two-qubit phases define crosstalk and whose residual displacements define dephasing."},{"cited_title":"Bombin and M","cited_arxiv_id":null,"evidence_quote":"Supplies the rotated-surface-code CNOT scheduling and the standard asymptotic logical-error scaling that Eq. (12) extends to include crosstalk."},{"cited_title":"Edmonds, Maximum matching and a polyhedron with 0, 1-vertices, Journal of research of the National Bureau of Standards 69B, 125 (1965)","cited_arxiv_id":null,"evidence_quote":"Provides the stabilizer-circuit simulator used for the numerical logical-error-rate results in the main figures and Appendix A."}],"review_version":1}