{"id":"223c4499-5f3d-4061-bda6-04cc4ea19298","arxiv_id":"2507.09684","paper_version":1,"verdict":"CONDITIONAL","confidence":"HIGH","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":7,"one_line_summary":"For square GKP codes, a Kerr evolution exp(i*pi*n-hat^2/8) is an exact, envelope-preserving logical sqrt(Hadamard) gate, enabling ancilla-free magic state preparation.","lead":"Researchers show that a standard Kerr interaction, a quartic nonlinearity, can directly produce the \"magic\" non-Clifford states needed for fault-tolerant quantum computing with GKP bosonic codes, without using an auxiliary qubit. The gate is exact for finite-energy GKP states and the paper proposes a superconducting circuit to implement it.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Robustness claim rests on an unmodeled false-positive readout probability: the quoted 1e-3 is the wrong conditional (P(e|g), not P(g|e)), and SM Sec. III.B appears to swap the two; without a P(g|e) budget, the claimed quadratic loss suppression is not established.","rationale":"After checking the core identity, Eq. (6) is correct: U_K=exp(i*pi*n^2/8) has eigenvalues +1 and +i respectively on the n congruent to 0 and 2 (mod 4) Fock classes that support the square-GKP Hadamard eigenstates, and [n^2, E_Delta]=0 preserves the finite-energy envelope. The reader's verification of this is sound, and no issue was found with the exactness claim in the absence of errors. The load-bearing weakness is in the secondary but central applied claim of robustness against photon loss. The post-selection protocol's fidelity is controlled by the false-accept probability P(g|e), not by P(e|g). The manuscript quotes and simulates P(e|g)=10^-3, and SM Sec. III.B contains a sentence that, read literally, would minimize the wrong probability. This is not a stylistic issue: it determines whether the accepted states after N rounds are dominated by two-loss events (quadratic, as claimed) or by single-loss events that were falsely accepted (linear in gamma). Since the paper does not supply a P(g|e) budget, an experimental readout curve, or a simulation scan over P(g|e), the quantitative robustness claim is not yet established. This matches the reader's conditional verdict; no adjustment is needed, but the condition should specifically require the false-positive readout budget, not merely P(e|g).","tokens_in":17053,"tokens_out":38871,"duration_ms":446210,"concrete_test":"Extend the numerical simulation behind Fig. S3 to use an explicit two-outcome confusion matrix with independent parameters P(g|e) (false accept) and P(e|g) (false reject), and plot 1-F_sbs versus gamma for P(g|e)=10^-2, 10^-3, and 10^-4 while keeping P(e|g) fixed at 10^-3. If 1-F_sbs develops a floor scaling as gamma times P(g|e) rather than gamma^2, then the post-selection robustness claim requires a demonstrated operating point with P(g|e) much less than gamma; provide a measured or simulated dispersive-readout discrimination curve for that operating point. Additionally, re-report the fidelity value F_sbs about 0.996 with the correct conditional probability labels to remove the ambiguity.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The abstract's headline advantage is robustness against a single photon-loss event via SBS post-selection. The mechanism is: reject any round in which the ancilla reads e. This only protects the logical fidelity if the false-accept probability P(g|e) (read g when the true error flag is e) is small compared with the loss probability gamma. The main text quotes P(e|g)=10^-3 as the realistic readout-error rate, but that is the opposite conditional: it is the probability of discarding a good state, which costs success probability but not post-selected fidelity. SM Sec. III.B then says one can minimize the probability P(e|g) at the cost of increasing P(g|e), and asserts quality increases; that sentence is internally inconsistent unless the two conditionals are swapped. As written, the realistic-fidelity estimate F_sbs about 0.996 with P(e|g)=10^-3 does not specify P(g|e), so it does not demonstrate that single-loss events are rejected with high enough probability to sustain the O(gamma^2) scaling shown in Fig. 2b. If P(g|e) is comparable to or larger than gamma, accepted states contain a fraction about gamma times P(g|e) of single-loss contaminated events, producing a linear infidelity floor that invalidates the practical robustness claim. This is not an attack on Eq. (6); the gate identity is sound, but the paper's central applied claim is conditional on a readout error budget that is neither modeled nor demonstrated.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"This Letter proposes using the Kerr unitary U_K = exp(iπ n^2/8), generated by the quartic Hamiltonian H = K n^2/2, to implement the logical sqrt(H) gate on square GKP codes. Because U_K is diagonal in the Fock basis and the finite-energy GKP envelope E_Δ = exp(−Δ² n) is also diagonal, the envelope is preserved and the gate is claimed to be exact in the absence of errors, in contrast to cubic-phase gates. The authors then combine U_K with small-Big-small (SBS) error correction and post-selection to prepare GKP magic states robustly against photon loss, present simulations of infidelity versus loss parameter γ, discuss ancilla readout errors, and propose a SNAIL-based circuit QED implementation with K/2π ≈ −20 kHz and a gate time of 6.3 μs.","tokens_in":17286,"tokens_out":18879,"duration_ms":213259,"significance":"The central identity Eq. (6) is exact, clean, and parameter-free: U_K acts as sqrt(H)_L on the square-GKP code states because of their Fock-support structure, and it commutes with the finite-energy envelope. This gives a genuine qubit-independent route to a non-Clifford logical operation on GKP states, avoiding the envelope distortion that makes cubic-phase gates unsuitable. The combination with SBS post-selection is a natural way to turn single-photon-loss detection into quadratic infidelity suppression. These are real strengths. However, the practical robustness claims are presently weakened by an unresolved conflation of readout-error conditionals, by numerical simulations that lack convergence checks and error bars, and by an implementation section that does not clarify the relation between the n^2 Hamiltonian and the standard a†²a² Kerr term. These issues are local and fixable, so the underlying idea remains valuable.","major_comments":[{"comment":"The robustness claim under measurement errors is not established because the two conditional probabilities of the ancilla readout are conflated. In a post-selection protocol that accepts only g outcomes, the fidelity-relevant quantity is p(g|e), the probability of reading g when the ancilla is in e. The main text correctly says that only false-positive results with probability p(g|e) affect the fidelity, but the realistic estimate F_sbs ≈ 0.996 with P(e|g) = 10^{-3} uses the opposite conditional, which only reduces the success probability and does not directly enter the post-selected infidelity. SM Sec. III.B then states that one can minimize P(e|g) at the cost of increasing P(g|e) and that the quality of the post-selection increases; this is backwards, because increasing P(g|e) admits more error-contaminated states and lowers quality. Furthermore, Figure S3's symbols are labeled P(e|g), yet they appear to change the infidelity; if both error directions are included in the model, the labeling is wrong, and if not, the plotted infidelity cannot depend on P(e|g) alone. To support the claimed quadratic suppression and the quoted realistic fidelity, the authors need to specify a budget or model for p(g|e) and show that it is small compared with γ.","section":"Main text (measurement-error paragraph after Fig. 2) and SM Sec. III.B"},{"comment":"The numerical evidence for the central robustness claim lacks convergence checks and error bars. In Fig. 2b the Δ = 0.15 curve is stated to saturate at low γ, and the authors attribute this to numerical errors rather than physics. No truncation analysis (Fock-space cutoff, SBS-basis cutoff) or sampling statistics is reported, so the reader cannot tell whether the saturation is an artifact or a real effect. Since the paper's headline is the quadratic suppression of infidelity with γ, the quantitative support for that scaling should be backed by convergence data, at least for the curves used to claim the quadratic gain.","section":"Simulations and Fig. 2b"},{"comment":"The implementation section leaves unspecified the relation between the n^2 Hamiltonian in Eq. (6) and the effective self-Kerr Hamiltonian H = (K/2)a†²a² = (K/2)(n² − n) in SM Eq. (S14). The difference is a linear term −(K/2)n, which at the quoted gate time t = π/(4K) contributes exp(∓iπ n/8), a nontrivial phase-space rotation on the square-GKP code. The paper does not state whether this term is cancelled by a rotating-frame choice or by an explicit frame-correction step, and the factor-of-two convention for K affects the gate time. Please clarify the rotating frame and the definition of K used in the SNAIL extraction.","section":"Circuit QED Implementation and SM Eq. (S14)"}],"minor_comments":[{"comment":"There is a typo: \"Altough\" should be \"Although\".","section":"Introduction"},{"comment":"Reference [25] is malformed: \"B. B. E. A. e. a. Ding, A.Z., Quantum control of an oscillator with a kerr-cat qubit\" should be corrected to a proper author list and journal/year.","section":"References"},{"comment":"The caption says the x-axis values of γ differ between curves; please explain in the caption why the curves are horizontally offset or plot them on a common axis for direct comparison.","section":"Fig. 2b caption"},{"comment":"The phrase \"higher GKP size should outperform those with smaller size\" is ambiguous; \"larger GKP states (smaller Δ)\" would be clearer.","section":"Simulations, paragraph after Fig. 2"},{"comment":"Define P(e|g) and P(g|e) explicitly as conditional readout probabilities and state which one is the false-positive and which is the false-negative relative to the post-selection outcome, to avoid the present confusion.","section":"SM Sec. III.B"}],"recommendation":"major_revision","confidential_remarks":"The core gate identity is sound and the proposed direction is useful. The main revision should focus on the readout-error budget and the numerical convergence of Fig. 2b, plus a clarifying statement about the rotating-frame treatment of the linear Kerr term. I do not see grounds for rejection, but the practical claims need to be made rigorous before acceptance."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The gate identity is real: exp(i pi n^2/8) squares to H on the Fock classes supporting the Hadamard eigenstates, and because it commutes with the envelope, the finite-energy GKP state sees an exact logical sqrt(H). I verified the central algebra and the photon-loss commutation relation; both are clean. The novelty is genuine -- the cited literature had the Fock-support structure and the cubic-gate obstruction, but not this construction. Pairing it with SBS post-selection is a natural way to turn a detectable single-loss event into a rejected round.\n\nThe soft spots are all downstream of the gate. The biggest is readout-error accounting. The main text correctly says that only the false-accept probability p(g|e) affects post-selected fidelity, but then the realistic fidelity estimate quotes P(e|g)=10^-3 -- the opposite conditional. SM Sec. III.B appears to swap them too: minimizing P(e|g) at the cost of increasing P(g|e) would degrade post-selection, not improve it. If Fig. S3 is actually varying P(g|e), the labels are wrong. As written, the claimed F_sbs ~ 0.996 at gamma=10^-2 lacks a demonstrated P(g|e) budget, and if P(g|e) is comparable to gamma, the quadratic suppression becomes a linear infidelity floor. This is fixable -- the authors clearly understand the right conditional in the main text -- but it needs a careful revision and a model tracking both conditionals separately.\n\nThe numerics also need work: no error bars, no convergence checks, and the Delta=0.15 saturation is attributed to numerical error without showing it vanishes. The SNAIL operating point sits at K/K3=900, below the authors' own 10^3 criterion; they acknowledge this and suggest optimization, which is fine for a proposal but should be framed as such. No code or data is shipped, though the SM gives enough detail for a specialist re-implementation.\n\nOverall this paper deserves a serious referee. The central result is exact and the protocol is plausible; the robustness claims are conditional on standard noise parameters, not fitted constants. The likely outcome is a major revision centered on readout-error analysis and simulation quality, but the core idea is worth having in the literature.","headline":"A genuinely new exact sqrt(H) gate for GKP via Kerr, but the robustness claims need a readout-error budget before they fully land.","tokens_in":697,"tokens_out":833,"would_cite":true,"duration_ms":62186,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":["03.67.Pp","85.25.Cp"],"model":"deepseek-v4-flash","headline":"The paper establishes that a Kerr interaction of duration π/(4K) realizes the logical square-root-Hadamard gate for finite-energy square GKP grid states, exactly when no errors occur, and that combining it with small-Big-small error…","keywords":["GKP code","magic state preparation","Kerr interaction","square-root Hadamard gate","small-Big-small error correction","post-selection","photon loss","circuit quantum electrodynamics"],"falsifier":"A direct experiment would prepare a finite-energy square GKP state $\\lvert +Y_\\Delta\\rangle$, apply the Kerr drive for exactly $t_K=\\pi/(4K)$, and perform Wigner tomography: if the output deviates from $\\sqrt{H}_L\\lvert +Y_\\Delta\\rangle$ beyond the expected loss and noise budget, or if the Gaussian envelope's width or position changes, the claimed exactness is falsified. A second, protocol-level test is to measure the post-selected infidelity as a function of $\\gamma$: if it does not scale roughly as $\\gamma^2$ when the false-positive readout probability $p(g|e)$ is varied, the robustness claim is falsified.","tokens_in":16691,"feed_emoji":"⚛️","tokens_out":9580,"duration_ms":110938,"temperature":0.7,"pith_summary":"Magic-state preparation for GKP qubits usually leans on an auxiliary qubit, long sequences, or cubic gates that distort the finite-energy envelope. This paper proposes a single oscillator-only step: a quartic Kerr Hamiltonian $H=K\\hat{n}^2/2$ acting for time $\\pi/(4K)$, whose unitary $\\hat{U}_K=e^{i\\pi\\hat{n}^2/8}$ is exactly the logical square-root-Hadamard gate $\\sqrt{H}_L$ for square GKP codes. Because the envelope operator $\\hat{E}_\\Delta=e^{-\\Delta^2\\hat n}$ commutes with $\\hat{n}^2$, the finite-energy code states keep their Gaussian envelope, so the gate is exact in the absence of errors rather than approximate. Under photon loss, a single-loss event leaves the state outside the code space; the paper shows that small-Big-small error correction and post-selection restore an approximately quadratic error suppression, and it gives a concrete cavity-SNAIL circuit-QED implementation with a ~20 kHz self-Kerr and ~6.3 μs gate time.","feed_headline":"Kerr drive prepares GKP magic states with no auxiliary qubit","feed_subtitle":"A single quartic drive applies the logical square-root-Hadamard gate to finite-energy grid states; post-selection tames photon loss.","key_machinery":"The load-bearing object is the diagonal number-phase unitary $\\hat{U}_K=e^{i\\pi\\hat n^2/8}$, the square of the generator of the GKP Hadamard gate $\\hat F=e^{i\\pi\\hat n/2}$. It is generated by a quartic Kerr Hamiltonian $H=K\\hat n^2/2$, so that after time $t_K=\\pi/(4K)$ the oscillator has undergone a logical square-root Fourier transform on square GKP codes, whose Hadamard eigenstates live only on Fock numbers $4n$ and $4n+2$. What makes the construction cohere is the commutation $[\\hat n^2,\\hat E_\\Delta]=0$: the Gaussian envelope of a finite-energy GKP state is untouched, so the gate is exact for the physically relevant states, not only for ideal infinite-energy combs. The same object also dictates the error behaviour through $\\hat U_K\\hat a=e^{i\\pi/8}e^{-i\\hat n\\pi/4}\\hat a\\,\\hat U_K$, meaning a photon lost before the gate is compounded into a large rotation, whereas a photon lost during the gate leaves the state outside the code space and therefore detectable by the SBS protocol.","core_discovery":"The central claim is Eq. (6): for square GKP codes the unitary $\\hat{U}_K=\\exp(i\\pi\\hat{n}^2/8)$, generated by the Kerr Hamiltonian $H=K\\hat{n}^2/2$ with $\\hbar=1$, coincides with the logical gate $\\sqrt{H}_L=\\sqrt{\\hat F}$ on the code space. The argument exploits the fact that the Hadamard eigenstates of square GKP codes have support only on Fock numbers $4n$ and $4n+2$; on those supports $\\hat{U}_K$ produces exactly the phase pattern of a square-root Fourier transform. The finite-energy envelope $\\hat E_\\Delta=e^{-\\Delta^2\\hat n}$ commutes with $\\hat n^2$, so the envelope is preserved and the gate is exact in the absence of errors—unlike the cubic-phase gate $T_L$, which distorts the envelope. The paper further shows that photon loss during the Kerr evolution is detectable but not correctable, so in an offline magic-state-preparation setting it applies SBS post-selection to reject single-loss events, recovering an error infidelity scaling close to $\\sim\\gamma^2$, and it estimates a realistic total fidelity $F_{\\rm sbs}\\approx0.996$ with about 81% success probability including measurement errors.","pith_inferences":["Extension: because any function of $\\hat n$ commutes with the Gaussian envelope, the same Fock-space phase-engineering trick could yield other diagonal logical gates on square GKP codes, not only $\\sqrt{H}_L$, by choosing integer-valued phase polynomials modulo the code's Fock support.","Extension: the offline post-selection framing suggests a natural heralded-retry architecture: reject a failed preparation and rerun the Kerr gate, which could lower the magic-state cost per accepted state compared with distillation if the false-positive readout probability can be pushed well below $\\gamma$.","Extension: the paper's success metric could be refined by including finite-duration gate errors and Kerr nonlinearity calibration; a parameter sweep of SNAIL detuning versus gate time would identify an optimum for fixed cavity lifetime, something the paper leaves to future work.","Extension: since the scheme works for any oscillator with a tunable self-Kerr, it transfers to trapped-ion motional modes and acoustic resonators, although the readout and post-selection steps would need platform-specific error models."],"forward_implications":["Magic states of H-type can be prepared from a GKP $\\lvert +Y_\\Delta\\rangle$ state by a single oscillator-only drive, removing the auxiliary-qubit lifetime and leakage bottleneck that limits qubit-mediated gates.","The protocol is compatible with finite-energy GKP states as prepared in the lab: no re-encoding or envelope-restoration step is needed after the gate.","With SBS error correction and post-selection, the logical infidelity is suppressed approximately quadratically in the single-photon-loss parameter $\\gamma$; two post-selection rounds already give $F_{\\rm sbs}\\approx 2.4\\times10^{-4}$ at $\\gamma=10^{-3}$.","Including photon loss during initialization, gate, and error correction, plus measurement errors, the paper's parameter set ($\\Delta=0.36$, $\\gamma=10^{-2}$, $P(e|g)=10^{-3}$) yields $F_{\\rm sbs}\\approx0.996$ with success probability $\\approx81\\%$.","A cavity-SNAIL implementation with $K/2\\pi\\approx-20\\,$kHz gives a $6.3\\,\\mu$s gate time, and the associated cubic-Kerr term is mostly detectable by post-selection for $K/K_3\\gtrsim10^3$."],"supporting_citations":[{"why":"Defines square GKP codes, the logical operators, and the Fock-space support of Hadamard eigenstates used to identify $\\hat U_K$ with $\\sqrt{H}_L$.","marker":"[13]"},{"why":"Shows the cubic-phase gate distorts the finite-energy GKP envelope, establishing the contrast that motivates the Kerr gate.","marker":"[23]"},{"why":"Introduces the small-Big-small error-correction protocol for finite-energy GKP states, used here for error correction and post-selection.","marker":"[26]"},{"why":"Provides the photon-loss rate scaling and bosonic-code performance analysis that sets the fidelity and success-probability trade-offs.","marker":"[14]"},{"why":"Provides the dispersive-readout model and asymmetric-thresholding discussion used to account for measurement errors in post-selection.","marker":"[30]"},{"why":"Demonstrates a tunable self-Kerr with a SNAIL device for cat-state preparation, the implementation route adapted for the proposed gate.","marker":"[41]"}],"fun_headline_variants":["Kerr gate does sqrt-Hadamard for GKP magic states, no ancilla","No ancilla: Kerr interaction prepares GKP magic states","Kerr nonlinearity gives sqrt-Hadamard for GKP magic states","GKP magic states via Kerr: no aux qubit needed","Kerr-mediated sqrt-Hadamard: GKP magic states, no ancilla"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is that the auxiliary-qubit readout used for post-selection can be made to almost never report 'success' when a photon-loss error has occurred (a false-positive probability $p(g|e)$ much smaller than the loss rate $\\gamma$); the paper assumes a realistic $P(e|g)=10^{-3}$ and argues asymmetric thresholding can achieve this, but does not demonstrate it for the full protocol.","fun_headline_variants_meta":{"raw":{"variants":["Kerr gate does sqrt-Hadamard for GKP magic states, no ancilla","No ancilla: Kerr interaction prepares GKP magic states","Kerr nonlinearity gives sqrt-Hadamard for GKP magic states","GKP magic states via Kerr: no aux qubit needed","Kerr-mediated sqrt-Hadamard: GKP magic states, no ancilla"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000442,"raw_usage":{"total_tokens":2240,"prompt_tokens":944,"completion_tokens":1296,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":560,"completion_tokens_details":{"reasoning_tokens":1200}},"tokens_in":560,"tokens_out":1296,"duration_ms":13495,"temperature":1.0,"reasoning_tokens":1200,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-06T17:52:39.743900+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"A direct experiment would prepare a finite-energy square GKP state $\\lvert +Y_\\Delta\\rangle$, apply the Kerr drive for exactly $t_K=\\pi/(4K)$, and perform Wigner tomography: if the output deviates from $\\sqrt{H}_L\\lvert +Y_\\Delta\\rangle$ beyond the expected loss and noise budget, or if the Gaussian envelope's width or position changes, the claimed exactness is falsified. A second, protocol-level test is to measure the post-selected infidelity as a function of $\\gamma$: if it does not scale roughly as $\\gamma^2$ when the false-positive readout probability $p(g|e)$ is varied, the robustness claim is falsified.","supporting_citations":[{"cited_title":"Hastrup, M","cited_arxiv_id":null,"evidence_quote":"Shows the cubic-phase gate distorts the finite-energy GKP envelope, establishing the contrast that motivates the Kerr gate."},{"cited_title":"Royer, S","cited_arxiv_id":null,"evidence_quote":"Introduces the small-Big-small error-correction protocol for finite-energy GKP states, used here for error correction and post-selection."},{"cited_title":"Using a Kerr interaction for GKP magic state preparation","cited_arxiv_id":null,"evidence_quote":"Demonstrates a tunable self-Kerr with a SNAIL device for cat-state preparation, the implementation route adapted for the proposed gate."}],"review_version":1}