{"id":"f955da9e-bbe8-420a-aea1-0879ac2c82d2","arxiv_id":"2412.04414","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":7,"one_line_summary":"Coherent errors plus syndrome measurement on a surface code create emergent unitary designs on the logical qubit above a finite threshold that also marks the optimal error-correction and entanglement transitions.","lead":"This paper shows that applying random single-qubit rotations to surface-code qubits, then measuring syndrome outcomes and correcting, produces random logical gates on the encoded qubit. Above a critical rotation density the gates form a 'unitary design', a type of randomness useful for benchmarking and tomography, and the transition matches the code's coherent-error correction threshold.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The thermodynamic-limit unitary design phase rests entirely on finite-size scaling of d=7–15 without error bars or bond-dimension convergence; larger-d checks are needed before the central claim is secure.","rationale":"Theorem 2 is rigorous and the analytical GHZ mapping in Sec. V D provides a genuine structural connection between the logical unitary ensemble and the purification transition. The numerical evidence for the design phase is credible but not conclusive: the entire phase diagram for Haar-random single-qubit unitaries is inferred from d≤15 scaling collapses with no error bars, no published code or data, and no stated MPS bond dimension or convergence checks. This is precisely the weakest assumption the reader identified, and it is load-bearing because the headline claim is the existence of a thermodynamic-limit unitary design phase. A larger-distance check, with controlled statistics and explicit bond-dimension convergence, would settle whether the observed crossings persist. The reader's CONDITIONAL verdict is therefore appropriate; my stress-test does not move the verdict.","tokens_in":29670,"tokens_out":17703,"duration_ms":180183,"concrete_test":"Recompute ∆(2) and ∆(4) for d=17 and d=19 (and, if feasible, d=21) with at least 1280 realizations, bootstrap error bars, and a stated MPS bond dimension converged until truncation error drops below 10^-8, then repeat the three-parameter collapse of Eq. (24). If the crossing point shifts by more than the current pc=0.87±0.02 or the collapse quality degrades, the thermodynamic-limit design phase is not established; if the collapse is stable, the finite-size concern is resolved.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim—that the projected ensemble becomes a unitary k-design for d→∞ when p>pc≈0.87—is supported only by the scaling collapse of ∆(k)(p,d) in Fig. 3 using d=7,9,11,13,15 and the ansatz Eq. (24). This is the load-bearing step: if that apparent crossing is a finite-size artifact, the protocol still realizes a well-defined projected ensemble (by Theorem 2), but there is no emergent Haar-random logical gate. The collapse silently assumes a single critical point common to all k, one correlation-length exponent ν, and that d≤15 already lies in the asymptotic scaling regime; none of these is proven. At p=1 the values of ∆(k) at d=15 are still O(0.1), so asymptotic convergence is not yet visible. The MPS purification-time data (Sec. V C, Fig. 7) used to anchor the coincident entanglement transition have the same limited distances and omit both the bond dimension and any convergence threshold. Since no code or data are supplied, these numerical gaps cannot be checked post hoc.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper proposes a protocol to generate random logical unitaries on encoded qubits by applying local coherent errors to a surface code, measuring syndromes, and applying a recovery operation. Theorem 2 gives sufficient conditions (even-weight stabilizers, CSS code with odd X and Z distances, and a physical unitary U commuting with time reversal T) under which the corrected post-measurement map acts on the logical subspace as a unitary operator whose distribution is independent of the input logical state. Numerically, using a tensor-network/staircase mapping, the authors find a finite-size crossing in the unitary k-design distance Δ^(k) (k = 2, 3, 4) at p_c ≈ 0.87 for random single-qubit unitaries applied with density p, coincident within error bars with the optimal-decoder error-correction threshold p_c,EC ≈ 0.88. The same p_c appears in an MPS purification-time entanglement transition and in Clifford simulations at larger distances. The paper concludes that for p > p_c the projected ensemble becomes a unitary k-design in the thermodynamic limit, with applications to shadow tomography, randomized benchmarking, and quantum cryptography.","tokens_in":29961,"tokens_out":9990,"duration_ms":101576,"significance":"If the central finite-size claim survives a more thorough numerical analysis, the result is significant: it provides a resource-lean route to non-Clifford random gates on logical qubits, and it identifies a new 'unitary design phase transition' that appears to coincide with the optimal coherent-error threshold and with a complexity transition in a classical decoding algorithm. The proof of Theorem 2 is a genuine generalization of earlier work on coherent Z errors and appears sound. The Clifford-simulation appendix (Appendix C) is a valuable cross-check at much larger distances, and the scaling collapses across k = 2, 3, 4 and across the design, error-correction, and purification diagnostics are consistent. However, the thermodynamic-limit claim currently rests on small-distance (d ≤ 15) finite-size scaling without error bars or MPS convergence data, so the numerical support is not yet commensurate with the strength of the central claim.","major_comments":[{"comment":"The claim of a unitary design phase for p > p_c in the thermodynamic limit is inferred entirely from the data collapse of Δ^(k) for d = 7, 9, 11, 13, 15 shown in Fig. 3. The plotted values at the largest available d and at p = 1 are still far from zero, so the collapse demonstrates a consistent crossing but not the asymptotic convergence Δ^(k) → 0 required for a design. The scaling ansatz (24), with p_c and ν fitted and the scaling function F_k determined by the collapse, is a reasonable but not conclusive inference, especially because no error bars are shown for Δ^(k). No larger-distance data for Haar-distributed single-qubit unitaries are presented: the Clifford data in Appendix C reach d = 161, but, as the paper notes, Clifford coherent errors cannot generate k ≥ 2 designs, so they cannot validate the k ≥ 2 design transition for the actual Haar-rotation ensemble. I ask for either a direct finite-size extrapolation of Δ^(k) to d = ∞ at several p > p_c showing that it vanishes, or larger-distance data for the non-Clifford ensemble, or a clear statement that the design phase is a conjecture supported only by this finite-size collapse.","section":"IV A, Eq. (24), Fig. 3"},{"comment":"The scaling analysis assumes a single critical point and a single correlation-length exponent for all moments k; this is stated as a conjecture immediately after Eq. (24). The central claim of convergence to the Haar distribution for all k depends on this unproven assumption. The data for k = 2, 3, 4 are consistent with a common p_c, but they do not test k > 4, where the ensemble cardinality bound (2^{n−1}) still permits designs up to k ≤ 2^{n/2}. To make the central claim robust, the authors should either prove that the threshold is k-independent, or present evidence for at least one higher moment (for example k = 5), or explicitly discuss how the scaling ansatz would detect a k-dependent threshold if one existed.","section":"IV A, Eq. (24)"},{"comment":"The purification-time data used to identify the coincident entanglement transition are obtained from MPS simulations, but the manuscript does not state the bond dimension used, the truncation error, or any convergence criterion. Without this information the critical point p_c = 0.88(1) and ν = 1.3(2) extracted from Fig. 7 cannot be assessed, and the associated claim that the MPS algorithm becomes inefficient above threshold is not fully supported. Please report bond dimensions and show convergence of τ/d with bond dimension at several values of p across the transition. The exact Clifford checks in Appendix C reach larger sizes, but they concern a different (Clifford) ensemble, so they do not resolve this gap for the Haar-rotation protocol studied in the main text.","section":"V C, Fig. 7"},{"comment":"The numerical claim that the unitary design threshold p_c exactly equals the optimal error-correction threshold p_c,EC is based on comparing two separately fitted critical points, p_c = 0.87(2) and p_c,EC = 0.88(1), whose uncertainties overlap. This is reasonable evidence, but the two quantities are extracted from the same finite-size data with the same scaling ansatz, so the overlap does not by itself establish equality in the thermodynamic limit. A joint scaling analysis or a simultaneous collapse of Δ^(k) and Δ_EC with a common p_c would be stronger; otherwise, the wording should be softened to 'consistent with coincidence'.","section":"IV B, Sec. IV A"}],"minor_comments":[{"comment":"Equation (A7) writes p(s) = (1/2)Tr(Π_0 U Π_s U^†), but the derivation gives p(s) = (1/2)Tr(Π_s U Π_0 U^†) = (1/2)Tr(U Π_0 U^† Π_s), as stated in Corollary 2.1. The displayed equation in Appendix A appears to be a typo and should be corrected.","section":"Appendix A, Eq. (A7)"},{"comment":"The captions state that data are averaged over 384–1280 realizations, but no error bars or standard errors are shown. Since the central inference is a scaling collapse, the absence of statistical uncertainty information should be addressed, at least by adding error bars to the main data points or reporting the sample-to-sample variance.","section":"Fig. 3 and Fig. 7 captions"},{"comment":"The caption and text use the phrase 'unitary design phase transition' in the Clifford setting, although the same appendix proves that Clifford coherent errors can only form a unitary 1-design. Please rephrase to 'unitary 1-design transition' or 'design transition for k = 1' to avoid implying that the Clifford data validate higher-moment designs.","section":"Appendix C, Fig. 11"},{"comment":"The notation U_{dec,s} is used both for the unitary operator and for the superoperator U_{dec,s}(·) = U_{dec,s}[·]U_{dec,s}^†. Please distinguish these explicitly, for example by writing U_{dec,s} for the channel and U_{L,s} or U_s for the unitary.","section":"Eqs. (22)-(23)"},{"comment":"The random decoder introduces a logical Pauli σ_rand,s chosen randomly for each syndrome. It would be helpful to state explicitly that these random choices are known to the experimentalist and can be classically tracked, since otherwise the projected ensemble is not operationally accessible.","section":"Sec. IV A, random decoder"}],"recommendation":"major_revision","confidential_remarks":"The manuscript does not provide code or data, yet the central numerical claim is the basis for the paper's main conclusion; I would encourage the editor to request a data/code availability statement as part of the revision. The overlap with Refs. [86] and [87] is acknowledged in the note added, and the present work's distinct contribution is the unitary-design framing and Theorem 2; if the finite-size scaling can be made more convincing, the paper would be a strong addition to the field."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Three things to know. The theorem is real and the generalization matters: prior work only handled Z rotations on surface codes, while Theorem 2 covers arbitrary single-qubit unitaries and general CSS codes with odd distances and even-weight stabilizers. The proof is clean, and the condition [U,T]=0 identifies a genuinely large class of physical operations. The numerical discovery—a unitary design transition at p_c ≈ 0.87 that coincides with the optimal EC threshold—is new, and the collapse across k=2,3,4 is consistent. The staircase mapping to a 1+1D monitored circuit, plus the analytical GHZ argument in Sec. V D, gives the phase coincidence a mechanism rather than leaving it as a numerical accident.\n\nSoft spots, in proportion. The thermodynamic-limit design phase rests entirely on finite-size scaling at d=7–15 with the ansatz Eq. (24). The paper states the single-critical-point assumption as a conjecture, which is honest, but the data in Fig. 3 have no error bars, and at p=1 the design distance at d=15 is still O(0.1), so the crossover to asymptotic convergence is not visible at the largest accessible size. I don't read this as a fatal flaw: Theorem 2 guarantees a well-defined projected ensemble regardless, and the Clifford large-distance results (d up to 161) do anchor the entanglement, purification, and EC transitions. But Clifford ensembles can only form 1-designs, so those larger-d data cannot directly support the unitary design phase itself. The MPS purification data in Fig. 7 omit the bond dimension and any convergence threshold, which weakens the third coincidence. No code or data are supplied; that is a reproducibility gap, not a correctness issue.\n\nCitation pattern looks fine, and the note added acknowledges concurrent related work honestly. The applications section is appropriately careful about fault tolerance and about the need for inverse-free randomized benchmarking.\n\nWho benefits: people working on logical randomized benchmarking, encoded shadow tomography, or decoder complexity transitions. The core theorem plus the protocol is enough to justify referee time.\n\nRecommendation: accept for peer review, with the referee explicitly asked to demand larger-distance design data (or a sharper analytic argument), stated MPS bond dimensions, and error bars on the scaling collapse.","headline":"Solid theorem and a genuinely new numerical transition, but the thermodynamic-limit design phase rests on d=7–15 scaling collapse, so the referee should push on larger distances and MPS convergence.","tokens_in":30462,"tokens_out":2562,"would_cite":true,"duration_ms":27211,"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":"Applying strong enough local rotations to a surface code makes syndrome measurements generate random logical gates that converge to a unitary design.","keywords":["unitary designs","surface codes","coherent errors","projected ensembles","measurement-induced phase transitions","quantum error correction","monitored circuits","randomness"],"falsifier":"Compute the k = 2 design distance Δ^(2)(p, d) for a fixed p above 0.9 at larger code distances d = 19, 23, 27 using the same matrix-product-state method; if the values stop decreasing with d, or the finite-size crossing point drifts outside the stated p_c ≈ 0.87 interval, the claimed design phase would be falsified.","tokens_in":29468,"feed_emoji":"🎲","tokens_out":5308,"duration_ms":51067,"temperature":0.7,"pith_summary":"The paper proposes that by deliberately applying local unitary rotations (coherent errors) to the physical qubits of a surface code and then measuring the error syndrome, one can generate random unitary gates on the encoded logical qubit with no auxiliary qubits and no magic-state distillation. Its central claim is that, for odd-distance rotated surface codes, the distribution of logical unitaries produced this way converges to a unitary k-design as the code distance grows, provided the density of coherent errors exceeds a finite threshold p_c ≈ 0.87. The same threshold is also the code's optimal error-correction threshold under coherent errors, and it appears as an entanglement phase transition in a related (1+1)-dimensional monitored circuit obtained by a staircase encoder-decoder mapping. The work proves a theorem (Theorem 2) giving conditions on the code and the physical unitary that ensure the syndrome-conditioned operations are genuinely unitary, and it supports the design phase with finite-size scaling of the design distance for k = 2, 3, 4.","feed_headline":"Coherent errors become random logical gates above p≈0.87","feed_subtitle":"For odd-distance surface codes, syndrome randomness yields Haar-like logical unitaries in the thermodynamic limit.","key_machinery":"The argument rests on three interlocking devices. First, Theorem 2's time-reversal symmetry condition: all logical information lives in time-reversal-odd operators, while syndrome measurements probe only even-weight stabilizer operators, so the measurement reveals nothing about the logical input state and the induced transformations are unitary by a standard fact about Kraus operators. Second, the choice of decoder: a random decoder that multiplies each recovery by a uniform logical Pauli turns the ensemble into an exact 1-design and lets the intrinsic k ≥ 2 randomness be studied, while the optimal decoder isolates the error-correction threshold. Third, the classical simulation: writing the surface code encoder and decoder as staircase circuits of CNOT gates converts the two-dimensional problem into a (1+1)-dimensional monitored circuit amenable to matrix-product-state simulation, in which the purification time of a reference qubit exhibits an entanglement phase transition at the same p_c.","core_discovery":"When a CSS code with one logical qubit, even-weight stabilizer generators, and odd X and Z distances is acted on by a physical unitary that commutes with time reversal T = K ∏ iY_j, the syndrome-dependent correction satisfies C_s Π_s U Π_0 = √p(s) U_{L,s} Π_0 with U_{L,s} unitary (Theorem 2). Thus the Born-rule randomness of syndrome outcomes gives a well-defined probability distribution over logical unitaries. The paper's numerical finding is that for odd-distance rotated surface codes with independent random single-qubit unitaries applied at density p, this projected ensemble approaches the Haar distribution on U(2) as d → ∞ for p > p_c ≈ 0.87, while below threshold it flows to a random Pauli ensemble (or identity under optimal decoding), with the same critical point governing the optimal error-correction threshold and the entanglement transition of the staircase-mapped monitored circuit.","pith_inferences":["The result suggests a general engineering principle: measurement-induced randomness in stabilizer codes can be harvested as a resource (random logical gates) rather than treated only as noise, and codes with even-weight stabilizers and odd-weight logical operators may exhibit similar design phases.","A testable extension is the many-body unitary case: Theorem 2 covers unitaries commuting with time reversal, such as three-body ZZZ gates, and whether the design transition persists for interacting coherent errors at finite density is open.","The threshold p_c ≈ 0.87 is high; one could ask whether biased rotations (for example, only Z-type) or noise-assisted strategies lower the threshold, or whether repeated rounds of the protocol near threshold yield effective fault tolerance, as the paper speculates.","The entanglement transition in the staircase circuit invites a spacetime-dual statistical-mechanics interpretation of the logical channel, which could give an analytical handle on the phase boundary beyond the numerical scaling collapse."],"forward_implications":["Randomized protocols requiring unitary 2- or 3-designs—classical shadow tomography, randomized benchmarking, quantum cryptography—could run directly on encoded logical qubits of surface codes without transversal Clifford gates or magic-state distillation.","The projected ensemble gives approximate designs with approximation error suppressed exponentially in code distance, so scaling up the distance reduces the error of any protocol built on these gates.","Below p_c the same circuit works as error correction: all logical unitaries converge to the identity under the optimal decoder, so the coherent errors are successfully corrected.","The classical matrix-product-state decoder is efficient below p_c and inefficient above it, meaning the unitary-design phase coincides with a computational-complexity transition for the simulation algorithm.","The random decoder's uniform logical Pauli scrambling guarantees an exact 1-design while leaving the higher-moment convergence to the intrinsic randomness of syndrome outcomes, separating decoder-dependent and intrinsic effects."],"supporting_citations":[{"why":"Established that single-qubit Z rotations on an odd-distance surface code yield logical Z rotations after syndrome correction, providing the original example of a projected ensemble of logical unitaries.","marker":"[31]"},{"why":"Introduced projected ensembles and deep thermalization, the conceptual template for emergent universal distributions from measurement outcomes.","marker":"[7]"},{"why":"Defined measurement-induced phase transitions in monitored random circuits, the standard framework used to interpret the entanglement transition in the staircase circuit.","marker":"[34]"},{"why":"Supplied the space-evolving block decimation method that motivates the staircase encoder-decoder mapping to a (1+1)-dimensional monitored circuit.","marker":"[48]"},{"why":"Computed coherent-error thresholds for planar-graph surface codes, giving the optimal error-correction threshold with which p_c is compared.","marker":"[49]"},{"why":"Generalized the unitarity result for surface-code corrections under arbitrary single-qubit rotations, extending the earlier Z-only construction.","marker":"[50]"},{"why":"Provided the cardinality bound on the size of the projected ensemble, used to argue that the ensemble can support designs up to moment k ≤ 2^{n/2}.","marker":"[55]"}],"fun_headline_variants":["Syndrome randomness turns coherent errors into Haar-random gates above p≈0.87","Logical qubits get Haar-random gates from syndrome randomness above p≈0.87","Coherent errors + syndrome readout = random logical gates above p≈0.87","Above p≈0.87, syndrome outcomes randomize logical gates from coherent errors","Haar-random logical gates emerge from coherent errors above p≈0.87"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The infinite-distance conclusions are inferred from scaling collapse of design distances computed at code distances 7 through 15, and the paper conjectures without proof that a single critical point controls all moments and survives in the thermodynamic limit.","fun_headline_variants_meta":{"raw":{"variants":["Syndrome randomness turns coherent errors into Haar-random gates above p≈0.87","Logical qubits get Haar-random gates from syndrome randomness above p≈0.87","Coherent errors + syndrome readout = random logical gates above p≈0.87","Above p≈0.87, syndrome outcomes randomize logical gates from coherent errors","Haar-random logical gates emerge from coherent errors above p≈0.87"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.001228,"raw_usage":{"total_tokens":5084,"prompt_tokens":1019,"completion_tokens":4065,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":635,"completion_tokens_details":{"reasoning_tokens":3959}},"tokens_in":635,"tokens_out":4065,"duration_ms":28243,"temperature":1.0,"reasoning_tokens":3959,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-11T21:23:45.816543+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Compute the k = 2 design distance Δ^(2)(p, d) for a fixed p above 0.9 at larger code distances d = 19, 23, 27 using the same matrix-product-state method; if the values stop decreasing with d, or the finite-size crossing point drifts outside the stated p_c ≈ 0.87 interval, the claimed design phase would be falsified.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Defined measurement-induced phase transitions in monitored random circuits, the standard framework used to interpret the entanglement transition in the staircase circuit."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Computed coherent-error thresholds for planar-graph surface codes, giving the optimal error-correction threshold with which p_c is compared."},{"cited_title":"Vovk and H","cited_arxiv_id":null,"evidence_quote":"Generalized the unitarity result for surface-code corrections under arbitrary single-qubit rotations, extending the earlier Z-only construction."}],"review_version":1}