{"id":"e3adad1d-dc19-4753-a849-34661c9c5df6","arxiv_id":"2412.05171","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":6,"one_line_summary":"Spin hopping between dots with differently tilted quantization axes performs high-fidelity microwave-free single-qubit control in a silicon 2x2 array.","lead":"A 2x2 silicon quantum dot array is operated both as a four-qubit microwave-controlled processor and as a two-qubit device using microwave-free baseband spin hopping, with a measured hopping gate fidelity lower bound of 99.50(6)%. The result shows that hopping control, previously shown in germanium hole arrays, can be engineered in silicon and may reduce the heating and crosstalk that limit microwave-based scaling.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The hopping X90 fidelity lower bound is inferred from one RB branch whose even-parity twin contains an unresolved artifact; without an interleaved RB or leakage check, the comparison to EDSR rests on an unverified premise.","rationale":"The reader's weakest_assumption identifies the same unresolved even-parity RB artifact, and I agree that this is the main source of risk. I partially disagree with the framing that the odd branch is already fully 'conservative': it is conservative relative to the even branch, but if the artifact originates from leakage or non-Markovian errors that also affect odd parity, a single-exponential extraction of F_Clif,odd could still be inflated, so the 'lower bound' label depends on an unverified assumption. The paper is commendably transparent: it explicitly states that interleaved randomized benchmarking was not taken, names readout crosstalk and state leakage as possible origins of the artifact, and provides data and scripts. I found no internal inconsistency in the fidelity attribution formula itself, and no other load-bearing objection that would move the verdict. The appropriate disposition remains conditional acceptance pending the interleaved RB or leakage check, which is exactly the reader's verdict, so no adjustment is needed.","tokens_in":30742,"tokens_out":7027,"duration_ms":78671,"concrete_test":"Reanalyze the deposited Zenodo RB data: fit the odd-parity decay with the same oscillatory/leakage model used for the even branch and test whether F_Clif,odd shifts by more than 0.1%; if it does, the attributive formula F_X90 = 1 - (1-F_Clif,odd)/2 is not a reliable lower bound. In a new experiment, run interleaved randomized benchmarking of the X90 hopping gate on Q1 at Bext = -15.8 mT and compare the IRB fidelity with 99.50(6)%; also track charge-state populations during RB to bound leakage. If IRB agrees within error, the concern is settled.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim F_hop_X90 = 99.50(6)% (Methods, Randomized Benchmarking) is derived as 1 - (1 - F_Clif,odd)/2, attributing all error in the X/Z Clifford RB to the two X90 gates per Clifford. This is a valid lower bound only if Z rotations are effectively error-free and if the odd-parity RB decay faithfully represents the average error per Clifford. The text itself reports that the even-parity branch 'contains a small oscillatory artifact that persists even with substantial averaging' and that 'we are unsure of the true origin of this artifact'; leakage and readout crosstalk are named as candidate mechanisms. If leakage into S(2,0)/T(0,2) charge states or non-Markovian phase errors also affects the odd branch, a single-exponential fit of F_Clif,odd can be biased upward, and the quoted 99.50(6)% would not be a guaranteed lower bound. Choosing the lower-fidelity branch is conservative relative to the even branch, but it does not by itself rule out upward bias from the same unresolved mechanism. Since the headline comparison to EDSR (99.54(4)%) rests on this number, the missing interleaved RB or leakage quantification is the load-bearing gap.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"Unseld et al. report operation of a 2x2 28Si/SiGe quantum dot array in two control regimes: conventional micromagnet-based EDSR of four qubits at high field, and baseband hopping-spin control of single-electron spin qubits at reduced field. The central result is a quoted lower bound on the fidelity of an X90 hopping gate, F_hop_X90 = 99.50(6)%, extracted from odd-parity randomized benchmarking of Q1, which the authors compare with the device's average EDSR single-qubit fidelity of 99.54(4)%. The paper also reports improved Hahn-echo coherence at low field, absence of the transient pulse-induced resonance shift for hopping control, and a proposed periodic nanomagnet design for scalable two-dimensional hopping control.","tokens_in":30977,"tokens_out":6643,"duration_ms":76115,"significance":"If the hopping fidelity claim holds, this is a valuable experimental demonstration: it shows that a single-electron silicon spin qubit can be controlled with baseband-only pulses at fidelity competitive with resonant EDSR, avoiding microwave heating and transient crosstalk and opening a concrete route toward two-dimensional silicon spin arrays. Strengths of the paper include direct measurement of hopping-induced coherent oscillations and quantization-axis tip angles, an honest discussion of device limitations, and public availability of data and scripts. The nanomagnet architecture proposal is clearly labeled as illustrative but rests on plausible OOMMF simulations. The main caveat is that the headline fidelity depends on a single randomized-benchmarking branch with an unresolved oscillatory artifact.","major_comments":[{"comment":"The headline result F_hop_X90 = 99.50(6)% is derived from the odd-parity Clifford decay F_hop_Clif,odd = 99.01(11)%, while the even-parity branch yields F_hop_Clif,even = 99.49(7)% and, as the authors state, 'contains a small oscillatory artifact that persists even with substantial averaging' whose origin they are unsure of. The two branches differ by 0.48% in Clifford fidelity, far larger than the 0.06% error bar quoted on the X90 bound. Choosing the lower-fidelity branch is conservative only with respect to the even/odd branch difference; it does not protect against a mechanism such as state leakage, readout crosstalk, or non-Markovian phase error that could bias both branches. Because the comparison to EDSR rests on this number, the authors should provide an interleaved randomized benchmarking measurement of the X90 gate, a leakage or parity-loss check during the RB protocol, or an explicit model showing that the artifact cannot affect the odd branch. Without one of these, the claimed lower bound is not established at the stated precision.","section":"Methods, Randomized Benchmarking; Fig. 4(f)"},{"comment":"The abstract compares the one-qubit hopping lower bound on Q1 with the EDSR average fidelity averaged over all four qubits, 99.54(4)%. Only the Q1 hopping gate was benchmarked; the Q4 hopping gate could not be faithfully benchmarked because pulsing gate P3 degrades readout. A one-qubit result is therefore not equivalent to a device-average comparison. The authors should either benchmark hopping gates on additional qubits or restrict the claim to 'comparable to the EDSR fidelity of this device' with the one-qubit sample size explicitly stated; as written, the comparison in the abstract is stronger than the data support.","section":"Abstract and Section III"},{"comment":"The quoted uncertainty on F_hop_X90 is the fitting standard deviation of the odd-branch Clifford decay only. The unresolved even-branch artifact and the 0.48% discrepancy between the parity branches constitute a systematic uncertainty that is not reflected in the quoted 99.50(6)% value. At minimum, the systematic spread should be reported or the authors should explicitly state that the quoted uncertainty captures only statistical fit error and that the systematic contribution is unresolved.","section":"Methods, Randomized Benchmarking"}],"minor_comments":[{"comment":"The labels '(lower)' and '(upper)' in the caption are ambiguous; please rephrase to identify clearly which panel shows T2* and which shows T2^H.","section":"Fig. 3(b) caption"},{"comment":"The ordering of t2 and tadd in the first Rz factor of Eq. (1) should be double-checked against the pulse diagram in Fig. 4(a); as written it is not obvious why tadd appears only in the first factor.","section":"Methods, Eq. (1)"},{"comment":"Use a space between numerical values and units, e.g., 'fRabi ≈ 2 MHz' instead of 'fRabi ≈2 MHz'.","section":"Page 5 and throughout"},{"comment":"The sentence 'The data in a-b) was collected' should read 'The data in a-b) were collected'.","section":"Supplementary Information, Section V"},{"comment":"The shaded area is described as a 'generous estimate of uncertainty' but no quantitative definition is given; please specify what the shaded band represents and how the ±15 nm out-of-plane displacement was chosen.","section":"Fig. 3(a) caption"}],"recommendation":"major_revision","confidential_remarks":"The paper reports a genuinely interesting demonstration, and the central idea is sound. My main reservation is the load-bearing fidelity lower bound: it is extracted from one randomized-benchmarking branch with an unexplained artifact, and the error bar quoted in the abstract does not reflect the even/odd branch discrepancy. This should be fixable by additional analysis or by downgrading the claim to a conditional bound. I would not reject the manuscript, but the abstract currently overstates the robustness of the headline number."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague,\n\nRead the Unseld et al. paper on baseband control of silicon spin qubits in a 2x2 array. The real news: this is the first demonstration of hopping-spin single-qubit control in single-electron silicon, using micromagnet stray-field quantization-axis tips at low field. The germanium hole work (Wang et al.) is properly cited; this paper adds the silicon electron version plus a tileable nanomagnet architecture. The engineering content is solid: they characterize EDSR, exchange, crosstalk, then go to baseband hopping and show a hopping X90 lower bound of 99.50(6)%, comparable to their EDSR average 99.54(4)% on the same device. The coherence doubling at low field and the absence of transient PIRS are nice, believable observations. The nanomagnet proposal is speculative but clearly simulated and honestly flagged as an architectural sketch.\n\nThe weakest spot is exactly where the stress-test note points: the headline fidelity bound comes from the odd-parity randomized benchmarking branch, while the even-parity branch shows an oscillatory artifact the authors say they cannot explain. They choose the lower-fidelity branch, which is conservative in one sense—it avoids overstating—but it does not rule out a common-mode bias from leakage or non-Markovian errors. If that artifact affects both branches, the 99.50(6)% could be an overestimate. That is a real gap, but it is a measurement-methodology gap, not a fatal flaw. The paper is unusually transparent: they state the artifact, list candidate causes, and do not hide the poor exchange quality factors or readout degradation on certain gates. No circular reasoning; the micromagnet polarization is fit to qubit frequencies, but the hopping fidelity is a direct measurement.\n\nThe central claim (hopping gates work at competitive fidelity in silicon) holds up. The precise number has a soft spot. The paper is worth a serious referee: the experimental work is careful, the comparison to EDSR is fair, and the architecture proposal is useful for the community. A referee should ask for interleaved RB or a leakage check on the hopping gate, and for more detail on the even-parity artifact. I would send it to review with a request for that additional data, not desk-reject.\n\nFor your reading group: worth a look if you follow spin qubit control or nanomagnet architectures. I would cite it if I wrote about scaling silicon qubit control.\n\nRecommendation: accept for peer review, conditional on the RB artifact being addressed or explicitly bounded.\n\n—","headline":"First silicon single-electron hopping-spin qubit demonstration with a credible 99.5% X90 lower bound; the main residual risk is the odd-parity RB branch artifact, but the paper is honest and worth refereeing.","tokens_in":31648,"tokens_out":654,"would_cite":true,"duration_ms":9903,"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":"Baseband hopping gates control single-electron silicon spin qubits at 99.50% fidelity.","keywords":["spin qubits","silicon","baseband control","hopping gate","quantum dot array","EDSR","crosstalk","nanomagnet"],"falsifier":"Run interleaved randomized benchmarking on Q1 with the hopping X90 as the interleaved gate at the same low-field setting: if the interleaved fidelity is more than one standard error below the attributed $99.50(6)\\%$ bound, the assumption that all randomized-benchmarking error comes from X90 fails and the quoted number is not a true gate fidelity.","tokens_in":30489,"feed_emoji":"🧲","tokens_out":9747,"duration_ms":90800,"temperature":0.7,"pith_summary":"This paper tries to establish that single-electron spin qubits in silicon can be controlled entirely with baseband voltage pulses by hopping the electron between quantum dots whose quantization axes point in different directions, and that this control is already good enough for quantum processors. The load-bearing result is a lower bound of $F^{\\mathrm{hop}}_{X90}=99.50(6)\\%$ on the fidelity of the hopping X90 gate, measured on one qubit in a $2\\times2$ $^{28}$Si/SiGe quantum dot array. That value is statistically the same as the $99.54(4)\\%$ average fidelity of the same device's microwave-driven EDSR gates. The authors argue this matters because baseband hopping avoids the microwave heating and transient resonance shifts that make EDSR crosstalk hard to calibrate in two-dimensional arrays, and because the quantization-axis tips needed for hopping can be engineered with periodic nanomagnets. If true, this gives a concrete path for scaling silicon spin qubits into two dimensions without an external solenoid.","feed_headline":"Baseband hopping gates control silicon spin qubits at 99.50% fidelity","feed_subtitle":"Microwave-free control matches EDSR fidelity on a 2D array and avoids transient crosstalk.","key_machinery":"The hopping gate is a sequence of two transfers between dots with non-collinear quantization axes, executed by fast detuning ramps: the charge follows the ground state adiabatically while the spin direction is transferred diabatically, keeping its orientation rather than tracking the changing quantization axis. The rotation accumulates from Larmor precession during intervals $t_1$ and $t_2$ in the two dots, plus an idling time $t_{\\mathrm{add}}$ that keeps the rotation axis consistent across repeated cycles. The unitary $U=\\big(R_z(\\omega_{\\mathrm{init}}(t_2'+t_{\\mathrm{add}}))\\,R_\\theta(\\omega_{\\mathrm{tip}}t_1')\\,R_z(\\omega_{\\mathrm{init}}t_2')\\,R_\\theta(\\omega_{\\mathrm{tip}}t_1')\\big)^r$ is fitted to measured parity oscillations to extract the tip angle $\\theta$, and the X90 gate is calibrated by adjusting timing parameters to give a periodicity of four under repeated application. The fidelity claim comes from randomized benchmarking with a Clifford set built from X90 hopping gates and physical Z-rotations, using the odd-parity branch of the readout.","core_discovery":"The central claim is that a silicon single-electron spin can be coherently rotated by shuttling it back and forth between two dots in an inhomogeneous micromagnet stray field and letting it precess for controlled times in the two non-collinear quantization axes. At the low external field used for hopping, the micromagnet produces measured quantization-axis tips of $37.3(2)^\\circ$ and $47.5(2)^\\circ$ for the two measured dot pairs. By attributing all randomized-benchmarking error to the constituent X90 hopping gates, the odd-parity data give $F^{\\mathrm{hop}}_{X90}=99.50(6)\\%$, comparable to the EDSR average fidelity of $99.54(4)\\%$, while Hahn-echo coherence times nearly double and the hopping gate adds no transient pulse-induced resonance shift. The paper also proposes a periodic iron nanomagnet pattern that would create roughly $90^\\circ$ quantization-axis tips on a 100 nm dot pitch, so that hopping control could be engineered over an arbitrarily large two-dimensional array without a solenoid.","pith_inferences":["The paper's bound assumes the Z-rotation shoulders of each Clifford are nearly error-free compared with the X90; an interleaved randomized-benchmarking measurement of the hopping X90 alone would test that assumption directly.","If non-adiabatic charge transfer really is the dominant error, a straightforward prediction is that the same hopping gate with tunnel coupling increased to 50–100 µeV should beat 99.5% fidelity; the paper leaves this as an outlook rather than a tested claim.","The unexplained even-parity randomized-benchmarking artifact suggests parity-based benchmarking may be contaminated by state leakage or readout crosstalk; if so, odd-parity lower bounds should be treated as conservative bounds rather than point estimates in fault-tolerance analysis.","The periodic nanomagnet landscape of roughly 90° tips and decoherence sweet spots could double as a built-in parking and shuttling map for idling spins and long-range qubit transport, which the paper does not develop further."],"forward_implications":["Hopping gates are baseband operations that do not produce the transient, schedule-dependent phase pickup of EDSR; the hopping decoupling pulse contributes only a constant phase shift.","Operating at the low fields needed for hopping nearly doubles Hahn-echo coherence, from roughly 30 µs to 49 µs for Q1 and from 21 µs to 44 µs for Q4.","Because hopping addressability comes from local quantization-axis tips rather than frequency separation, the proposed nanomagnet array can work at zero external field and remove the solenoid from the cryostat.","Tip angles between 45° and 135° allow a single shuttling cycle to implement a Hadamard-class gate, making the design tolerant to magnet and dot placement variation.","The dominant estimated error, a 0.1–1% Landau–Zener probability of non-adiabatic charge transfer at the anticrossing, is attributed to limited tunnel-coupling tunability, so improving tunnel coupling should raise the fidelity beyond the quoted bound."],"supporting_citations":[{"why":"Supplies the hopping-spin gate protocol and baseband calibration method, originally demonstrated on germanium hole spins, that this paper transfers to single electrons in silicon.","marker":"[25]"},{"why":"Supplies the coherent shuttling and quantization-axis tip characterization methods used to fit the hopping oscillations.","marker":"[24]"},{"why":"Supplies the Pauli-spin-blockade initialization/readout, EDSR control, and coherent benchmarking procedures on a silicon multi-qubit array.","marker":"[9]"},{"why":"Supplies the modified Hahn-echo pulse sequence used to quantify pulse-induced resonance shift crosstalk for both EDSR and hopping gates.","marker":"[21]"},{"why":"Provides the previous temperature-dependent PIRS study whose trends this device's EDSR crosstalk measurements are compared against.","marker":"[22]"},{"why":"Supplies the X/Y and X/Z Clifford compilation tables used to convert measured Clifford fidelities into primitive gate fidelities for both control modes.","marker":"[43]"},{"why":"Provides the planar 2x2 28Si/SiGe quantum dot platform and gate layout on which the measurements are performed.","marker":"[26]"},{"why":"Supplies the modular nanomagnet design concept and material parameters adopted for the proposed periodic hopping-control architecture.","marker":"[31]"},{"why":"Supplies the micromagnetic simulation package used to compute the relaxed magnetization and stray-field landscape of the proposed nanomagnet array.","marker":"[35]"}],"fun_headline_variants":["Hopping control hits 99.5% fidelity in silicon qubits","2D silicon qubits tamed by microwave-free hopping","Baseband hopping doubles coherence in silicon spin qubits","Silicon qubits hop to 99.5% fidelity without microwaves","Microwave-free hopping matches EDSR at 99.5% fidelity"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The fidelity bound rests on the randomized-benchmarking analysis of the odd-parity branch being faithful, even though the even-parity branch shows an unexplained oscillatory artifact that could indicate state leakage or non-Markovian errors affecting both branches.","fun_headline_variants_meta":{"raw":{"variants":["Hopping control hits 99.5% fidelity in silicon qubits","2D silicon qubits tamed by microwave-free hopping","Baseband hopping doubles coherence in silicon spin qubits","Silicon qubits hop to 99.5% fidelity without microwaves","Microwave-free hopping matches EDSR at 99.5% fidelity"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000658,"raw_usage":{"total_tokens":3058,"prompt_tokens":1040,"completion_tokens":2018,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":656,"completion_tokens_details":{"reasoning_tokens":1927}},"tokens_in":656,"tokens_out":2018,"duration_ms":14485,"temperature":1.0,"reasoning_tokens":1927,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-11T20:50:39.275071+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Run interleaved randomized benchmarking on Q1 with the hopping X90 as the interleaved gate at the same low-field setting: if the interleaved fidelity is more than one standard error below the attributed $99.50(6)\\%$ bound, the assumption that all randomized-benchmarking error comes from X90 fails and the quoted number is not a true gate fidelity.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the hopping-spin gate protocol and baseband calibration method, originally demonstrated on germanium hole spins, that this paper transfers to single electrons in silicon."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the Pauli-spin-blockade initialization/readout, EDSR control, and coherent benchmarking procedures on a silicon multi-qubit array."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the X/Y and X/Z Clifford compilation tables used to convert measured Clifford fidelities into primitive gate fidelities for both control modes."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provides the planar 2x2 28Si/SiGe quantum dot platform and gate layout on which the measurements are performed."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the micromagnetic simulation package used to compute the relaxed magnetization and stray-field landscape of the proposed nanomagnet array."}],"review_version":1}