{"id":"51f219e1-b286-49b3-9694-8cc08bcb565a","arxiv_id":"2412.16730","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":8.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":1,"one_line_summary":"Epitaxial alpha-tantalum grown directly on c-plane sapphire shows a large, power-independent microwave loss that disappears when a niobium interlayer or a damaged sapphire surface prevents direct epitaxial Ta/sapphire contact.","lead":"Superconducting tantalum films that grow epitaxially on sapphire were found to absorb about 60 times more microwave energy than similar niobium films, even though their crystal quality and electrical properties looked excellent. Inserting a thin niobium layer or deliberately damaging the sapphire surface before growth restored performance, showing the loss originates at the tantalum/sapphire interface.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Interface attribution rests on two controls that each change more than the interface; a geometry-sweep test is needed to confirm the loss scales with interface participation.","rationale":"The paper presents a careful, well-controlled experimental study with internal consistency: epitaxial Ta(111) on sapphire gives low, power-independent Qi, while the Nb interlayer and plasma-damaged sapphire both restore high Qi. The reproduction in Appendix A across additional samples and substrates strengthens the empirical correlation. I agree with the reader's conditional assessment that the empirical finding is solid but the causal attribution to the interface is not fully closed. My analysis identifies the same weakest assumption as the reader—that the controls differ only at the interface—and extends it by noting that Sample C changes both the interface and the Ta film orientation, so the two controls together still leave open the possibility that a property of epitaxial Ta(111) (rather than the Ta/sapphire interface per se) is the loss source. The proposed geometry sweep would provide a direct, quantitative test of the interface-loss mechanism by checking whether the excess loss tracks the interface participation ratio. If it does, the interface attribution is strongly supported; if not, the conclusion would need revision. This does not change the reader's verdict of CONDITIONAL, so the recommendation is UNCHANGED.","tokens_in":20474,"tokens_out":5241,"duration_ms":51955,"concrete_test":"Fabricate CPW resonators with several gap widths (e.g., 2, 3, 6, and 10 um) on the same epitaxial Ta-on-sapphire film (Sample A) and on the Nb-interlayer film (Sample B), keeping all other processing identical. Compute the electromagnetic field participation ratio at the metal/substrate interface for each geometry using a finite-element solver. If the excess loss (1/Qi_A - 1/Qi_B) scales linearly with the interface participation ratio, the attribution to the Ta/sapphire interface is confirmed. If 1/Qi_A is independent of geometry or does not track the calculated participation, the dominant loss likely resides in the Ta film bulk or top surface rather than the bottom interface.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim that the epitaxial Ta/sapphire interface is the source of loss is supported by Samples B and C, but neither control isolates the interface while holding the Ta film fixed. Sample B replaces the Ta/sapphire interface with Ta/Nb and Nb/sapphire interfaces; Sample C simultaneously amorphizes the sapphire and changes the Ta film from epitaxial (111) to textured (110). The paper argues from XRD, AFM, RRR, Tc, and TEM/EDS that Samples A and B are otherwise very similar, but microwave loss at an interface can be controlled by atomic-scale chemistry, strain, or electronic structure that need not affect Tc or RRR. In particular, the 5 nm Nb interlayer could alter the local field distribution, proximity coupling, or interfacial strain of the Ta film in ways that improve Qi without removing an intrinsic Ta/sapphire loss. Sample C's high Qi shows that roughness and disorder do not cause the loss, but because the film orientation also changes, it cannot separate interface effects from properties of epitaxial Ta(111) itself. The data establish a strong correlation between low Qi and the epitaxial Ta/sapphire interface, yet the strong claim 'only present in samples having an epitaxial Ta/sapphire interface' is not uniquely proven by the two controls as presented.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"This paper reports a systematic study of Nb and Ta films sputtered on c-plane sapphire at different growth temperatures, correlating structural, transport, and microwave-loss properties. The central finding is that epitaxial alpha-Ta(111) films grown directly on c-plane sapphire show markedly poor low-power internal quality factors (Qi ~ 8k) despite excellent structural and DC transport properties, whereas Nb films and lower-temperature Ta films are high-performing. To locate the loss, the authors compare three 630 C samples: Sample A (epitaxial Ta on sapphire), Sample B (epitaxial Ta on a 5 nm epitaxial Nb interlayer), and Sample C (Ta on an Ar-plasma-damaged sapphire surface, yielding textured film). Both modifications raise Qi to ~400-500k, and the paper attributes the loss to the epitaxial Ta/sapphire interface. The authors support this with XRD, AFM, RHEED, TEM/EDS, and transport data, and they reproduce the effect on additional samples in Appendix A. They further note that the loss is essentially power-independent, suggesting a non-TLS mechanism, and they discuss possible microscopic origins while leaving the mechanism open.","tokens_in":20596,"tokens_out":3394,"duration_ms":33492,"significance":"If the central claim holds, this is an important result for superconducting quantum devices: it identifies a specific, previously unrecognized loss channel at epitaxial Ta/sapphire interfaces and offers simple, practical mitigation routes (Nb interlayer or substrate-surface damage). The study is thorough in its characterization, provides detailed process parameters for reproducibility, and the appendices strengthen the conclusion by showing the effect across different substrate vendors, polish conditions, and surface preparations. The two-control experimental design is clever and the paper is candid about the open microscopic mechanism. However, the causal attribution to the interface itself is the load-bearing point, and it rests on the assumption that the controls differ only in the intended interfacial property, which is not uniquely established.","major_comments":[{"comment":"The two controls used to identify the interface as the loss source each change more than the interface alone. Sample B replaces the Ta/sapphire interface with Ta/Nb and Nb/sapphire interfaces, introducing a new material and possible proximity, strain, or field-distribution effects; Sample C simultaneously amorphizes the sapphire surface and changes the Ta film from epitaxial (111) to textured (110). The fact that two very different modifications both raise Qi by ~60x is consistent with an interfacial loss mechanism, but it does not uniquely prove that the loss resides specifically at the epitaxial Ta/sapphire interface, because the shared variable (absence of that interface) is not varied independently of other film and interface properties. A geometry-sweep test that varies the interface participation ratio (e.g., CPW gap width) on identically grown Samples A and B would provide a direct, quantitative test of the claim that the loss scales with interface participation. Without such a test, the abstract's statement that the authors 'determine that the source of loss is only present in samples having an epitaxial Ta/sapphire interface' is stronger than the evidence supports; the data establish a strong correlation, not a unique causal proof.","section":"Section III C, Fig. 3, Table II"},{"comment":"The inference that Samples A and B are 'very similar' and that therefore the loss must come from the interface relies on macroscopic quantities (XRD lattice constant, RRR, Tc, surface roughness). Microwave loss at an interface can be controlled by atomic-scale chemistry, interfacial strain, or electronic structure that need not affect Tc or RRR, as the paper itself acknowledges when discussing possible mechanisms. The claim that the loss is not due to Ta film structure or surface morphology is supported by the comparison between A and B, but the comparison cannot exclude the possibility that the 5 nm Nb interlayer alters the interfacial electronic structure or strain in the Ta film in a way that removes a loss channel located in the Ta near the interface rather than at the sapphire interface itself. The paper should either soften the causal language (e.g., 'suggests' instead of 'determines' in the abstract) or add a more direct test, such as increasing the number of geometries to vary interface participation or performing a thickness series with fixed interface, to separate interface from near-interface film loss.","section":"Section IV, first paragraph"}],"minor_comments":[{"comment":"The text contains a typo: 'Lorenzian' should be 'Lorentzian' in the description of the resonance line shape.","section":"Appendix E"},{"comment":"The word 'vorticies' is misspelled; it should be 'vortices' in the discussion and in the title of reference 51.","section":"Section IV and reference 51"},{"comment":"The chemical name in the text is misspelled: 'heximethyldisilazane' should be 'hexamethyldisilazane'.","section":"Section II A"},{"comment":"The photon-number estimation in Eq. (2) depends on the assumed attenuation and the extracted Q and Qc; the paper should clarify the systematic uncertainty in the absolute photon number, since the power-independence argument for the lossy samples relies on the relative scaling across the measured range.","section":"Section II C and Appendix E"}],"recommendation":"major_revision","confidential_remarks":"The paper is a solid experimental contribution with careful materials characterization and a reproducible dataset. The central result is likely of broad interest, but the headline claim in the abstract overstates the causal evidence. With a geometry-sweep test or a more cautious framing of the conclusion, the paper would be acceptable; as written, the load-bearing attribution requires additional support. The paper's fit to the journal scope is good for an applied physics or quantum hardware venue."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Two things to know. First, the empirical result is likely real and important: epitaxial α-Ta(111) grown directly on c-plane sapphire shows low-power Qi around 8k, roughly 60 times worse than the same Ta grown on a 5 nm epitaxial Nb interlayer or on plasma-damaged sapphire, and the loss is power-independent. Second, the paper is careful about structural and electrical characterization, and it reproduces the finding on additional substrates and preparations in Appendix A, which is what makes the observation credible.\n\nWhat's new: prior work used Nb buffer layers and studied TLS in nucleated versus non-nucleated alpha-Ta films, but I don't know of a controlled demonstration that the epitaxial Ta/sapphire interface itself is the dominant loss source. The paper also gives useful process details for reproducible low-loss Ta growth.\n\nWhat it does well: the growth temperature series shows the loss appears exactly when Ta becomes epitaxial (111), while Nb films prepared the same way stay high-Q. The two interface-modification experiments are a good design, and the structural (XRD, RHEED, AFM, TEM/EDS) and transport (RRR, Tc) data support the claim that Samples A and B are nearly identical Ta films differing mainly in the nucleation interface.\n\nThe soft spots are in the strength of the causal language, not in the data. As the stress-test note says, Sample B changes more than the interface: it adds Ta/Nb and Nb/sapphire interfaces, and a 5 nm Nb layer could alter the local field distribution or proximity coupling. Sample C changes the Ta film orientation from epitaxial (111) to textured (110), so it cannot separate interface effects from properties of the epitaxial film itself. The paper argues convincingly that roughness and disorder are not the cause, but the abstract's claim that the source of loss is \"only present in samples having an epitaxial Ta/sapphire interface\" is a bit ahead of what these two controls strictly prove. The microscopic mechanism remains open; the discussion handles this honestly. Minor weaknesses: few resonators per chip, no raw data or code released, and interface participation is estimated rather than directly varied.\n\nBottom line: this is a solid, useful paper for anyone working on tantalum-based qubits and microwave materials. It deserves a serious referee and likely publication after the authors either soften the causal claim or add one more test, such as a geometry sweep that changes interface participation. I'd cite it.","headline":"Genuinely useful experimental paper showing the epitaxial Ta/sapphire interface is a major microwave loss source, with causal language slightly ahead of the evidence.","tokens_in":21237,"tokens_out":2640,"would_cite":true,"duration_ms":23308,"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":"Microwave loss in epitaxial tantalum films on c-plane sapphire originates at the tantalum/sapphire interface, and can be eliminated by inserting a niobium interlayer or damaging the sapphire surface.","keywords":["tantalum thin films","microwave loss","c-plane sapphire","epitaxial interface","superconducting resonators","two-level systems","niobium interlayer","qubit materials"],"falsifier":"A direct probe of the interface would settle the claim: grow an epitaxial Ta film on sapphire while altering only the interfacial chemistry, for example by oxygen dosing or changing the surface termination, and check whether the low-Qi loss disappears while the Ta remains epitaxial (111). Alternatively, spectroscopic detection of in-gap quasiparticle states at the epitaxial Ta/sapphire interface (e.g. via a tunnel junction or X-ray standing wave) that does not correlate with the loss would contradict the interfacial-loss mechanism.","tokens_in":20208,"feed_emoji":"🔬","tokens_out":4961,"duration_ms":39655,"temperature":0.7,"pith_summary":"This paper claims that the poor microwave performance of high-quality epitaxial α-phase tantalum films on c-plane sapphire is caused by the epitaxial Ta/sapphire interface itself, not by the film's bulk quality, surface roughness, or ordinary two-level-system losses. The evidence comes from a systematic growth-temperature study: epitaxial Ta(111) films have excellent structural and DC electrical properties, yet their resonator quality factors are roughly 60 times lower than comparable Nb films. By controlling the nucleation interface, the paper shows that inserting a thin epitaxial Nb interlayer or deliberately damaging the sapphire surface with argon plasma restores high quality factors. This identifies a specific, controllable material interface as the loss source and offers practical steps for fabricating low-loss tantalum superconducting circuits.","feed_headline":"A single interface explains tantalum's microwave loss","feed_subtitle":"Adding a niobium interlayer or damaging the sapphire recovers resonator quality factors by about 60x.","key_machinery":"The experimental control is a three-way sample comparison: epitaxial α-Ta(111) grown directly on c-plane sapphire (Sample A), on a 5-nm epitaxial Nb(111) interlayer (Sample B), and on sapphire deliberately disordered by an in-situ argon plasma (Sample C). Because Nb films of nearly identical structure are low-loss and the Ta films in Samples A and B have nearly identical XRD, roughness, resistivity, Tc, and RRR, the only systematic difference is the Ta/sapphire interface; the ~60x Qi difference therefore localizes the loss to that interface. The measurement chain is narrow-gap (6/3 µm) CPW resonators at low photon occupation, with the diameter-correction method used to extract internal quality factor.","core_discovery":"The central discovery is that the epitaxial α-Ta(111)/Al2O3(0001) interface is a significant source of microwave loss in superconducting resonators, while structurally similar epitaxial Ta films nucleated on a 5-nm Nb(111) interlayer or on plasma-damaged sapphire show internal quality factors more than an order of magnitude higher. The loss is essentially power-independent over the studied range, indicating it is not a saturable two-level-system loss. The paper infers the interfacial origin by comparing three samples grown at the same temperature that differ only in the nucleation layer: Sample A (epitaxial Ta directly on sapphire) has low Qi, while Sample B (Nb interlayer) and Sample C (plasma-treated sapphire) recover high Qi despite similar Ta bulk properties. The microscopic mechanism is left open, with candidates including interface quasiparticle states, interfacial piezoelectricity, or vortex dynamics.","pith_inferences":["The results suggest that other epitaxial metal/sapphire interfaces used in superconducting circuits deserve the same kind of differential testing, since a material could be falsely blamed for loss that actually originates at its epitaxial interface with the substrate.","If the loss mechanism is an electronic interface state, adjusting the Fermi-level alignment or the oxygen stoichiometry at the interface (e.g. by oxygen dosing during growth) could suppress the loss while preserving epitaxy, which is a testable alternative to the interlayer and damage approaches.","The reproducibility data in Appendix A imply the effect is robust across substrate vendors and polish types, so the mitigation strategy should transfer across fabrication facilities, provided the growth-temperature calibration is done carefully."],"forward_implications":["Growing epitaxial α-Ta on a thin epitaxial Nb interlayer restores resonator internal quality factors to above 500k, providing a practical route to low-loss Ta capacitors without changing the Ta surface.","The loss is power-independent, so it will not be cured by operating at higher power or by TLS-mitigation strategies such as surface cleaning alone.","Substrate surface preparation matters not only for chemical cleanliness but for the crystallographic structure of the interface: plasma damage removes the lossy epitaxial interface at the cost of film texture and roughness.","High DC transport quality (RRR above 40, Tc near bulk) does not guarantee low microwave loss, so structural and DC characterization alone is insufficient to qualify a superconducting film for quantum circuits."],"supporting_citations":[{"why":"Establishes tantalum as a low-loss capacitor metal for transmons, the baseline that this paper's lossy epitaxial Ta films contradict.","marker":"[8]"},{"why":"Prior attribution of Ta performance to surface oxide properties; this paper's interfacial-loss result refines that picture.","marker":"[12]"},{"why":"Shows that a thin niobium underlayer nucleates body-centered-cubic α-Ta, providing the basis for Sample B.","marker":"[17]"},{"why":"Previous microwave characterization of Ta resonators grown on a niobium buffer layer, a comparison point for the interlayer approach.","marker":"[18]"},{"why":"Earlier study of two-level systems in nucleated and non-nucleated epitaxial α-Ta films; the non-saturable loss seen here contrasts with that work.","marker":"[19]"},{"why":"Provides the CPW resonator measurement methodology and the diameter-correction fit used to extract internal quality factors.","marker":"[28]"},{"why":"The specific resonator mask design targeting the metal-substrate interface, used for all loss measurements.","marker":"[34]"},{"why":"Supports the estimate of electric-field participation at the metal/substrate interface for the narrow-gap geometry.","marker":"[36]"},{"why":"The diameter-correction model that converts measured S21 spectra into accurate internal and external quality factors.","marker":"[37]"}],"fun_headline_variants":["Epitaxial Ta-sapphire interface is the microwave loss culprit","Ta's microwave loss traced to epitaxial sapphire interface","Niobium interlayer fixes tantalum's interface microwave loss","Epitaxial interface drives tantalum resonator loss","Plasma damage or Nb interlayer mitigate Ta microwave loss"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The attribution of the loss to the epitaxial Ta/sapphire interface assumes that the three samples differ only in the nucleation interface and not in hidden variables such as impurity incorporation, strain, or electronic structure changes induced by the Nb interlayer or plasma damage.","fun_headline_variants_meta":{"raw":{"variants":["Epitaxial Ta-sapphire interface is the microwave loss culprit","Ta's microwave loss traced to epitaxial sapphire interface","Niobium interlayer fixes tantalum's interface microwave loss","Epitaxial interface drives tantalum resonator loss","Plasma damage or Nb interlayer mitigate Ta microwave loss"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000183,"raw_usage":{"total_tokens":1338,"prompt_tokens":994,"completion_tokens":344,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":610,"completion_tokens_details":{"reasoning_tokens":263}},"tokens_in":610,"tokens_out":344,"duration_ms":3295,"temperature":1.0,"reasoning_tokens":263,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-11T10:15:45.646315+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"A direct probe of the interface would settle the claim: grow an epitaxial Ta film on sapphire while altering only the interfacial chemistry, for example by oxygen dosing or changing the surface termination, and check whether the low-Qi loss disappears while the Ta remains epitaxial (111). Alternatively, spectroscopic detection of in-gap quasiparticle states at the epitaxial Ta/sapphire interface (e.g. via a tunnel junction or X-ray standing wave) that does not correlate with the loss would contradict the interfacial-loss mechanism.","supporting_citations":[{"cited_title":"Place , author L","cited_arxiv_id":null,"evidence_quote":"Establishes tantalum as a low-loss capacitor metal for transmons, the baseline that this paper's lossy epitaxial Ta films contradict."},{"cited_title":"Crowley , author R","cited_arxiv_id":null,"evidence_quote":"Prior attribution of Ta performance to surface oxide properties; this paper's interfacial-loss result refines that picture."},{"cited_title":"Face \\ and\\ author D","cited_arxiv_id":null,"evidence_quote":"Shows that a thin niobium underlayer nucleates body-centered-cubic α-Ta, providing the basis for Sample B."},{"cited_title":"Urade , author K","cited_arxiv_id":null,"evidence_quote":"Previous microwave characterization of Ta resonators grown on a niobium buffer layer, a comparison point for the interlayer approach."},{"cited_title":"Alegria , author D","cited_arxiv_id":null,"evidence_quote":"Earlier study of two-level systems in nucleated and non-nucleated epitaxial α-Ta films; the non-saturable loss seen here contrasts with that work."},{"cited_title":"McRae , author H","cited_arxiv_id":null,"evidence_quote":"Provides the CPW resonator measurement methodology and the diameter-correction fit used to extract internal quality factors."},{"cited_title":"Calusine , author A","cited_arxiv_id":null,"evidence_quote":"Supports the estimate of electric-field participation at the metal/substrate interface for the narrow-gap geometry."},{"cited_title":"Khalil , author M","cited_arxiv_id":null,"evidence_quote":"The diameter-correction model that converts measured S21 spectra into accurate internal and external quality factors."}],"review_version":1}