{"id":"43d8d8fd-8f6b-46ec-acf4-70981c694bdb","arxiv_id":"2412.15903","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":3,"one_line_summary":"Argon milling and oxygen descumming introduce magnetic Yu-Shiba-Rusinov bound states inside the superconducting gap of tantalum thin films.","lead":"This study used a scanning tunneling microscope at millikelvin temperatures to show that standard argon milling and oxygen cleaning steps used in making tantalum superconducting circuits can create magnetic impurities. The impurities appear as Yu-Shiba-Rusinov bound states inside the superconducting gap, which can shorten the lifetime of qubits built from tantalum films.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Near-gap YSR peaks (597.5/598.4 µeV) are only 2-3 µV from the coherence peak while the lock-in modulation is 20 µV, so they may not be resolved; the inner pair's 18 mT suppression also lacks a quantitative explanation.","rationale":"The reader's weakest-assumption analysis correctly identifies the magnetic-vs-nonmagnetic interpretation as the soft spot. I agree only partially: the more fundamental problem is that two of the three claimed YSR binding energies (597.5 and 598.4 µeV) are separated from the SIS coherence peak by less than the lock-in modulation amplitude, making their existence as distinct states unresolved. This is an internal-consistency issue that can be checked directly. The inner pair is a real in-gap feature, but the evidence that it is magnetic is the field dependence, and the proposed gap-edge-smearing mechanism cannot explain the disappearance of a state at 410 µeV while Δ(B=18 mT) ≈ 584 µeV. This is not a disagreement with current consensus; it concerns the quantitative support in the paper itself. The authors should be credited for a well-executed STM experiment and for discussing the implications for qubit decoherence, but those implications are speculative until the states are resolved and their magnetic origin is quantitatively demonstrated. I keep the verdict CONDITIONAL, with the added condition that higher-resolution spectroscopy and a quantitative field-dependence analysis be provided.","tokens_in":8235,"tokens_out":10672,"duration_ms":105112,"concrete_test":"Acquire dI/dV spectra on the same milled Ta samples with Vmod ≤ 2 µV (and ideally with a normal-metal tip, or at fields above the tip critical field, to remove the SIS convolution) and check whether the near-gap features remain as distinct peaks; if they do not, the 597.5/598.4 µeV binding energies are artifacts of the gap edge. For the inner pair, measure peak position and linewidth versus out-of-plane field up to Bc and compare with a quantitative YSR model including Zeeman and orbital pair-breaking effects; if the peak disappears at 18 mT even though ε is far below Δ(B), the magnetic-YSR assignment needs to be revisited.","verdict_should_be":"CONDITIONAL","load_bearing_attack":"Lock-in differential conductance was measured with Vmod = 20 µV. In the SIS geometry, a sample YSR state at energy ε appears at bias Δ_tip + ε, while the coherence peak sits at Δ_tip + Δ_sample. With Δ_tip = 208.1 µeV and Δ_sample = 600.3 µeV, the reported single-pair state at ε = 597.5 µeV lies only 2.8 µV from the coherence peak, and the outer pair at ε = 598.4 µeV lies 1.9 µV from it. These separations are an order of magnitude smaller than the modulation amplitude, so the lock-in spectra cannot resolve distinct peaks; the quoted near-gap binding energies are not supported by the stated energy resolution. The inner pair (410.2 µeV) is well resolved, but its suppression at ~18 mT is not quantitatively explained: with the Ginzburg-Landau gap at 18 mT being ~584 µeV, the state remains deep inside the gap, and a Zeeman shift of ~1 µeV is negligible. The paper's appeal to magnetic anisotropy (Ref. 28) is qualitative, so the magnetic-YSR interpretation of the only well-resolved in-gap signal is not established.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The manuscript reports low-temperature scanning tunneling microscopy on alpha(111) tantalum thin films that were exposed to air and then Ar+ milled, a standard in situ cleaning step in qubit fabrication. After 15 minutes of milling the authors observe a 'single-pair' of in-gap peaks with binding energy 597.5 ueV, located very close to the superconducting coherence peaks, and after an additional 30 minutes they observe a double-pair structure with binding energies 598.4 ueV and 410.2 ueV. These peaks are suppressed by out-of-plane magnetic fields of about 17-18 mT. The authors interpret the peaks as Yu-Shiba-Rusinov bound states originating from local magnetic moments associated with oxygen defects introduced by the cleaning procedure, and they discuss implications for dephasing and relaxation in tantalum-based superconducting qubits.","tokens_in":8511,"tokens_out":6591,"duration_ms":64953,"significance":"If the identification is correct, the result is practically significant: standard Ar-milling of tantalum surfaces may inadvertently create magnetic impurities that contribute to qubit decoherence, and the paper provides a plausible microscopic route connecting a specific cleaning step to in-gap bound states. The manuscript has quantitative strengths: the superconducting gaps and critical fields are extracted with Dynes and Ginzburg-Landau fits, the field-dependent spectra are shown, and the central observation of milling-induced in-gap states is clearly presented. However, the magnetic YSR interpretation is largely inferential: no atomic-resolution identification of the defects, no elemental analysis for oxygen, and no quantitative model for the field-induced suppression are provided. The practical relevance for qubit processing is nonetheless a valid and testable hypothesis that warrants further work.","major_comments":[{"comment":"The two near-gap binding energies are claimed to be within a few microvolts of the coherence peak, but the lock-in modulation amplitude is Vmod = 20 uV. With Delta_tip = 208.1 ueV and Delta_sample = 600.3 ueV, the single YSR state at epsilon = 597.5 ueV corresponds to a bias of 805.6 ueV, only 2.8 uV from the coherence peak at 808.4 ueV; the outer double-pair peak at epsilon = 598.4 ueV is 1.9 uV away. Because the lock-in signal is a convolution over a window of about 20 uV, these peaks cannot be resolved as distinct features with the stated resolution, so the quoted near-gap binding energies are not supported by the experimental parameters. The inner pair at 410.2 ueV is well separated and not affected by this criticism, but the near-gap values should be re-measured with smaller modulation or supported by a deconvolution/fitting analysis.","section":"Figure 2d and methods (lock-in settings)"},{"comment":"The suppression of the well-resolved inner YSR pair at about 18 mT is not explained quantitatively. Using the reported Ginzburg-Landau parameters, at B = 18 mT the sample gap is Delta(B) = 600.3 x (1 - (18/103.1)^2) ≈ 582 ueV, so the 410 ueV state remains about 170 ueV inside the gap, and a Zeeman shift of roughly 2 ueV is negligible. The qualitative argument that gap-edge smearing overdamps states near the coherence peak cannot account for the disappearance of a state this deep in the gap, and the appeal to magnetic anisotropy (Ref. 28) is not developed into a testable prediction. A quantitative model of the field suppression, or additional measurements such as in-plane field dependence, temperature dependence, or spin-polarized spectroscopy, is needed before the inner pair can be confidently identified as a magnetic YSR state.","section":"Figure 3 and text following it"},{"comment":"The abstract and introduction attribute the bound states to 'oxygen descumming and argon milling,' but the experimental procedure only Ar+ mills an air-exposed film; no oxygen-plasma descumming step is performed, and no un-milled control sample is measured. In addition, no elemental analysis (e.g., XPS or EDS) is provided to show that the defects contain oxygen. The attribution to oxygen impurities is therefore a hypothesis rather than a demonstrated result. The authors' own acknowledgment that atomic-resolution mapping is unattainable on these granular films is reasonable, but it does not replace the need for elemental or chemical characterization, or at least an explicit reframing of the oxygen mechanism as speculative.","section":"Abstract and experimental methods"}],"minor_comments":[{"comment":"The critical fields are reported inconsistently: the main text gives B_sample,c = 103.1 mT and B_tip,c = 1.59 T, while the supplement gives 101.7 mT and 1.51 T for the same quantities; these should be reconciled.","section":"Main text vs. Supplemental Figure S1"},{"comment":"The suppression field for the single-pair state is given as 17 mT in the text and in Figure 2e, while the double-pair state is said to be suppressed at 'approximately 18 mT'; the values should be made consistent or the difference should be explained.","section":"Figure 2 caption and text"},{"comment":"The spectrum in Figure 2c is labeled the 'pristine superconducting gap,' but it was acquired after Ar milling on a region without YSR features; the wording should be clarified to avoid implying a measurement on an unmilled surface.","section":"Figure 2c caption and text"},{"comment":"Binding energies such as 597.5 ueV and 598.4 ueV are quoted to 0.1 ueV precision, which is much finer than the lock-in modulation of 20 uV; please provide the fitting procedure or round the values to a precision consistent with the energy resolution.","section":"Binding-energy precision"}],"recommendation":"major_revision","confidential_remarks":"The paper reports an observation that is potentially important for the superconducting-qubit community, and the experimental data appear to be collected carefully. The main concern is overclaiming: the magnetic YSR and oxygen-defect interpretation is plausible but is not yet established by the present data, and one of the two key spectral features is not resolved at the stated lock-in settings. I would suggest requiring the authors to address the energy-resolution issue and the quantitative field-suppression issue before publication, and to soften the oxygen attribution unless supporting chemical analysis is added."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"You should know this paper reports something potentially useful but overinterprets its own data. The authors show that standard Ar+ milling of oxidized tantalum films creates sub-gap features in STM spectra, and they argue these are Yu-Shiba-Rusinov bound states from oxygen-induced local moments that could harm qubit coherence. The observation itself is a legitimate data point for the qubit materials community.\n\nWhat is genuinely good: the SIS gap calibration is careful, with Dynes fits and Ginzburg-Landau extrapolation giving Delta_sample = 600.3 ueV and Delta_tip = 208.1 ueV. The topography and XRD establish the alpha(111) phase. The inner pair at 410.2 ueV is well resolved and its disappearance with field is worth noting. The paper is also honest about its limits—no atomic-resolution defect mapping, no elemental analysis—and the oxygen attribution is explicitly hedged as “likely.”\n\nNow the soft spots, and they are real. The stress-test note is correct: with lock-in modulation Vmod = 20 uV, the near-gap peaks at 597.5 and 598.4 ueV sit only ~2-3 uV from the coherence peak. That separation is an order of magnitude below the modulation, so those two states cannot be resolved as distinct peaks. Quoting binding energies with sub-uV precision overstates the energy resolution. The only well-resolved in-gap feature is the inner pair at 410 ueV, but its collapse at ~18 mT is not quantitatively explained. At that field the GL gap is still ~580 ueV, the state is deep inside the gap, and the Zeeman shift is ~1 uV—negligible. The appeal to magnetic anisotropy and gap-edge smearing does not address why a state at 410 ueV would vanish while the gap is still wide open. So the magnetic-YSR attribution rests on the one resolved feature, and its field dependence is unexplained.\n\nThe paper would benefit from a serious referee precisely because the claim is plausible and the experimental platform is relevant, but the analysis needs to be tightened. A referee should ask for a proper deconvolution or lower modulation to substantiate the near-gap peaks, and a quantitative model for the field suppression of the inner pair. The connection to qubit decoherence remains speculative but is reasonable as a motivation. I would send this to peer review, but with major revision expected.","headline":"Potentially relevant observation of in-gap states in argon-milled tantalum films, but the near-gap peaks are likely unresolved and the magnetic-YSR interpretation is not established.","tokens_in":9050,"tokens_out":3711,"would_cite":true,"duration_ms":37616,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"Standard argon milling of tantalum thin films leaves oxygen-related magnetic defects that create Yu-Shiba-Rusinov bound states inside the superconducting gap.","keywords":["tantalum thin films","Yu–Shiba–Rusinov bound states","argon ion milling","oxygen defects","magnetic impurities","scanning tunneling microscopy","superconducting qubits","two-level systems"],"falsifier":"Expose identically prepared tantalum films to different oxygen levels before the same argon-milling step (for example, varied air-exposure times or an in situ capping layer), then count YSR peaks in dI/dV maps; if peak density does not increase with oxygen exposure, the oxygen-defect mechanism is wrong. Elemental mapping of individual defect sites would provide a complementary check.","tokens_in":8044,"feed_emoji":"🔬","tokens_out":8444,"duration_ms":70035,"temperature":0.7,"pith_summary":"The paper reports that a standard in situ cleaning step used to prepare tantalum superconducting devices—oxygen descumming followed by argon ion milling—creates oxygen-related defects that carry local magnetic moments. At 45 mK, scanning tunneling spectroscopy resolves these moments as Yu–Shiba–Rusinov (YSR) bound states inside the superconducting gap: single-pair states with a binding energy of 597.5 µeV after 15 minutes of milling, and double-pair states at 598.4 µeV and 410.2 µeV after 30 more minutes. Because YSR states are low-energy excitations, the paper argues they can act as parasitic two-level systems that contribute to dephasing and energy relaxation in tantalum-based qubits. The finding matters because the cleaning protocol is routine in qubit fabrication, so these magnetic defects may be present in devices that were assumed to have clean interfaces.","feed_headline":"Argon cleaning seeds magnetic bound states in tantalum","feed_subtitle":"Yu-Shiba-Rusinov peaks appear after the standard surface prep used in tantalum qubit fabrication, adding a decoherence source.","key_machinery":"The central object is the Yu–Shiba–Rusinov (YSR) bound state, a sub-gap excitation that forms when a local magnetic moment exchange-couples to a superconductor. The paper uses the standard formula $\\varepsilon = \\pm \\Delta (1-\\alpha^2)/(1+\\alpha^2)$, with $\\alpha = J S_{\\mathrm{imp}} \\pi \\rho_n$, to relate the peak position to the exchange coupling $J$ and the impurity spin; weak coupling puts the peaks near the coherence peaks, and higher spin produces more pairs. The measurement machinery is a superconducting tantalum tip forming a superconductor–insulator–superconductor tunnel junction with the film, with sample and tip gaps $\\Delta_{\\mathrm{sample}}=600.3\\,\\mu\\mathrm{eV}$ and $\\Delta_{\\mathrm{tip}}=208.1\\,\\mu\\mathrm{eV}$ extracted from field-dependent Dynes fits. The argument also relies on a gap-edge-blurring mechanism: a moderate out-of-plane field smears the density-of-states peak at $\\Delta$, letting near-gap YSR states decay into the continuum.","core_discovery":"On its own terms, the paper establishes that argon milling of oxidized tantalum films does not simply remove the native oxide; it also disrupts tantalum–oxygen bonds in a way that leaves unsatisfied charge and spin centers embedded in or under the surface. These centers exchange-couple weakly to the superconducting condensate and produce YSR peaks that sit asymptotically close to the quasiparticle coherence peaks. Extending the milling time brings the defects closer to the superconducting interface, increases their effective spin, and shifts the peaks toward the Fermi level, giving the double-pair spectrum. The peaks are suppressed by an out-of-plane field of roughly 17–18 mT, which the paper explains through the broadening of the gap edge: once the coherence peaks blur, near-gap bound states overlap continuum states and become overdamped. The authors conclude that common cleaning protocols introduce magnetic bound states into tantalum qubit devices and should be revisited.","pith_inferences":["A testable extension follows: the density of YSR states should track the oxygen content of the film; varying air exposure or oxygenation before milling and counting in-gap peaks would confirm the oxygen-moment attribution.","The same cleaning chemistry is used for other superconducting materials, so analogous magnetic bound states may be present in non-tantalum qubit films and deserve a similar search.","If near-gap YSR states are suppressed by small fields, then low-frequency magnetic noise in the device environment could effectively switch these states in and out of the gap, possibly explaining part of the noise seen in transmon coherence measurements."],"forward_implications":["If the cleaning-induced YSR states are present in real devices, fabricated tantalum qubits inherit low-energy magnetic excitations at the interface, adding a concrete decoherence channel.","Longer or more aggressive milling does not remove the problem; it makes it worse, generating higher-spin defects whose YSR peaks sit closer to the Fermi level.","The near-coherence-peak position means even weak perturbations—fields around 15–20 mT—can destabilize these states, so qubit operating environments need to be considered.","Noble-metal passivation layers that prevent the native oxide from forming would avoid the defect source altogether, pointing to a design change in metallization.","Because the defects are spatially random and often subsurface, device-level mitigation will need statistical or spatially averaged detection rather than atomic-precision defect control."],"supporting_citations":[{"why":"Original theory papers predicting sub-gap bound states from magnetic impurities in a superconductor.","marker":"14–16"},{"why":"A review that supplies the YSR energy formula and the tunneling-spectroscopy framework used to assign the peaks.","marker":"19"},{"why":"Introduces the tantalum transmon qubit whose fabrication motivates the cleaning protocol studied here.","marker":"5"},{"why":"Documents the Ta2O5 surface oxide and suboxide species that underlie the oxygen-defect picture.","marker":"6"},{"why":"Reports oxygen-vacancy-induced magnetism at superconducting interfaces, the prior evidence for local moments.","marker":"11"},{"why":"Show that higher-spin impurities produce multiple YSR pairs, supporting the double-pair assignment.","marker":"20,21"},{"why":"Describes the superconducting tantalum STM tip used to resolve the sub-gap spectra.","marker":"22"},{"why":"Analyzes the field response and magnetic anisotropy of YSR states, used to explain the 18 mT suppression.","marker":"28"},{"why":"Proposes noble-metal passivation to avoid native tantalum oxide, the alternative fabrication route the paper suggests.","marker":"32"}],"fun_headline_variants":["Argon milling spawns spin states inside tantalum films","Tantalum qubit cleaning seeds magnetic bound states","YSR peaks from argon descum threaten tantalum qubits","Cleaning tantalum for qubits creates spin impurities","Millikelvin STM reveals cleaning-induced magnetic states"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is that the in-gap conductance peaks are magnetic Yu–Shiba–Rusinov states caused by oxygen-generated defects; if they are instead non-magnetic impurity states or artifacts of the granular film, the paper's conclusion about cleaning-induced magnetism collapses.","fun_headline_variants_meta":{"raw":{"variants":["Argon milling spawns spin states inside tantalum films","Tantalum qubit cleaning seeds magnetic bound states","YSR peaks from argon descum threaten tantalum qubits","Cleaning tantalum for qubits creates spin impurities","Millikelvin STM reveals cleaning-induced magnetic states"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000302,"raw_usage":{"total_tokens":1676,"prompt_tokens":820,"completion_tokens":856,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":436,"completion_tokens_details":{"reasoning_tokens":777}},"tokens_in":436,"tokens_out":856,"duration_ms":8475,"temperature":1.0,"reasoning_tokens":777,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-11T10:57:51.713572+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Expose identically prepared tantalum films to different oxygen levels before the same argon-milling step (for example, varied air-exposure times or an in situ capping layer), then count YSR peaks in dI/dV maps; if peak density does not increase with oxygen exposure, the oxygen-defect mechanism is wrong. Elemental mapping of individual defect sites would provide a complementary check.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Analyzes the field response and magnetic anisotropy of YSR states, used to explain the 18 mT suppression."},{"cited_title":"\\ Wang , author X.-Q","cited_arxiv_id":null,"evidence_quote":"Proposes noble-metal passivation to avoid native tantalum oxide, the alternative fabrication route the paper suggests."}],"review_version":1}