{"id":"5aa73c2d-a221-435c-af81-d5ebaeee93e3","arxiv_id":"2507.17150","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":3,"one_line_summary":"Monolithic barium titanate microresonators with intrinsic quality factor above one million and waveguide loss of 0.32 dB/cm are demonstrated using a redeposition-free dry etch.","lead":"Researchers developed a dry-etching process for barium titanate photonic chips that removes etch residue, and used it to build microresonators with record quality factors above one million. The advance lowers optical loss in BTO waveguides to 0.32 dB/cm, a step toward practical high-speed electro-optic modulators and frequency comb sources on this material.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Intrinsic Q and 0.32 dB/cm loss claims depend on an unreported coupling model and group index; both inputs need explicit reporting or independent extraction to support the record.","rationale":"The central claim is that a redeposition-free Ar+Cl2 etch produces the first BTO resonators with intrinsic Q > 1 million and the lowest propagation loss (0.32 dB/cm). There is strong supporting evidence: systematic variation of straight fraction in racetracks, SEM/AFM showing smooth sidewalls, and a comparison table. The statistical error bars in Fig. 2c,d are provided, which is better than typical. However, the two headline numbers rest on a measurement analysis that is incompletely specified. The paper never states the group index or the coupling coefficient used to convert linewidths to intrinsic Q. This is precisely the reader's weakest assumption. I agree with the reader: the concern is real and load-bearing, but it is correctable by reporting or re-fitting raw data. The 'near-critical coupling' label and the given Qi/QL ratio imply a particular coupling state, but without the extinction ratio or coupling Q, an independent reviewer cannot verify that Qi is not, say, 0.9×10^6 under a different valid fit. Similarly, the loss quote is explicitly proportional to ng; since ng for a shallow-etched BTO rib can differ from the material index by a few to several percent, the 0.32 dB/cm value carries a systematic uncertainty not captured by the per-resonance scatter. I also note that the 0.32 dB/cm is a racetrack average, and the extrapolated straight loss is 0.3 dB/cm, so the abstract's phrasing is slightly imprecise but if anything conservative. The sidewall-angle language ('high verticality' vs ~60°) is a separate reporting inconsistency, but it does not directly threaten the measured Q or loss. Verdict remains CONDITIONAL: the physics is credible, but the reporting gaps should be closed before the record claims can be independently confirmed.","tokens_in":9070,"tokens_out":10291,"duration_ms":110419,"concrete_test":"Re-fit the raw transmission spectra behind Fig. 2a with a standard single-bus ring lineshape to extract the loaded QL and the coupling parameter x (from the resonance extinction ratio), then compute Qi = QL(1+x) for each resonance. Independently obtain ng from the measured free spectral range of the same devices via ng = λ²/(FSR·L). If the refitted Qi remains above 1×10^6 and the loss recomputed with the measured ng stays within ±10% of 0.32 dB/cm, the record claims are robust; if either fails, the reported values need revision.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The paper's headline records—Qi = 1.35×10^6 and α = 0.32 dB/cm—are extracted from transmission resonance linewidths, but the extraction relies on two unstated inputs. First, separating intrinsic from loaded Q requires the coupling state (e.g., the resonance extinction ratio or coupling Qc). Figure 2a only labels the device 'near-critical coupling' and reports Qi and QL; the reader cannot verify the coupling model. If the coupling state were mis-assigned, Qi could shift by tens of percent, directly threatening the '>1 million' claim. Second, converting Qi to loss via α = 2πng/(Qiλ)·10log10 e (Fig. 2d) requires the group index ng, which is never stated. The EO analysis uses n ≈ 2.26, but the group index of the 180-nm-etch rib waveguide is not provided; since α scales linearly with ng, a 10% value error changes 0.32 dB/cm to ~0.35 dB/cm. The same formula also treats the racetrack loss as uniform, so the quoted 0.32 dB/cm is a length-weighted average over straight and bent sections, not the straight-waveguide loss (extrapolated separately as ≈0.3 dB/cm). These are reporting gaps rather than demonstrated errors, but they are load-bearing for the central claim.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The manuscript reports a redeposition-free Ar+Cl2 dry etching process for monolithic barium titanate (BTO)-on-insulator waveguides, and uses it to fabricate microracetrack and microring resonators. The central claims are an intrinsic quality factor Qi = 1.35×10^6 (the first >1 million in BTO), a propagation loss as low as 0.32 dB/cm, and Mach–Zehnder modulators with VπL = 0.54 V·cm and an effective Pockels coefficient r_eff = 162 pm/V. The paper motivates the etch chemistry by showing that conventional wet cleans (RCA-1, piranha, KOH, BOE) are ineffective or damaging for BTO, whereas the Ar+Cl2 process yields smooth, vertical sidewalls with ~0.29 nm rms roughness. Loss values are extracted from resonance linewidths across many devices and resonances, and the straight-waveguide loss is extrapolated to ≈0.3 dB/cm from a series of racetracks with varying straight fraction.","tokens_in":9328,"tokens_out":3831,"duration_ms":37813,"significance":"If the reported results hold, this is a significant advance for BTO integrated photonics: it would remove a key fabrication bottleneck (sidewall roughness and residue) and demonstrate that long photon lifetimes are achievable in a strong electro-optic material. The statistical analysis over many resonances and the benchmarking against prior BTO platforms are strengths, as is the practical etch improvement that may transfer to other ferroelectric oxides. However, the headline claims depend on two unstated inputs—the coupling model used to separate Qi from QL and the group index ng used to convert Qi to loss—and on an unreported overlap integral for the r_eff extraction. These gaps must be closed for the record claims to be fully verifiable.","major_comments":[{"comment":"The reported Qi = 1.35×10^6 and QL = 0.8×10^6 are extracted from a single transmission resonance, but the manuscript does not state the coupling model used to separate intrinsic from loaded quality factor. It only says the device is 'near-critical coupling.' Please provide the resonance fitting function, the coupling Qc (or the resonance extinction ratio), and the uncertainty on Qi. This is load-bearing because a mis-assigned coupling state could shift Qi by tens of percent and would directly affect the '>1 million' headline claim.","section":"Optical characterization, Fig. 2a"},{"comment":"The conversion from Qi to loss uses α = 2πng/(Qiλ)·10log10 e, where the group index ng is never stated or measured. Since the loss scales linearly with ng, a 10% error in ng changes 0.32 dB/cm to about 0.35 dB/cm, so the quoted loss carries a proportional systematic uncertainty that is not quantified. Please report the value of ng used (with its source, e.g., a mode simulation or an independent measurement) and its uncertainty. In addition, the text and Fig. 2d treat the racetrack loss as uniform; the quoted 0.32 dB/cm is therefore a length-weighted average over straight and bent sections, not the straight-waveguide loss, which is separately extrapolated to ≈0.3 dB/cm. This distinction should be stated explicitly wherever the 0.32 dB/cm figure appears.","section":"Optical characterization, Fig. 2d"},{"comment":"The effective Pockels coefficient r_eff = 162 pm/V is obtained from r_eff = λg/(n^3 Γ VπL), but the overlap integral Γ is not reported, and the assumed refractive index n = 2.26 is given without uncertainty. Please provide the simulated Γ value (with simulation details) and the uncertainty in the Vπ measurement. Without Γ, the reader cannot independently verify the r_eff claim, which is one of the paper's three headline results.","section":"Electro-optic modulation analysis, r_eff extraction"}],"minor_comments":[{"comment":"The caption says 'Electro-beam lithography'; this should be 'Electron-beam lithography.'","section":"Fig. 1 caption"},{"comment":"The phrase 'whose the etch recipe is similar' contains a grammatical error; it should read 'whose etch recipe is similar.'","section":"Supplementary S1"},{"comment":"The error bars are described as statistical variation over resonances, but the number of devices measured per data point is not stated; please specify how many devices and how many resonances contribute to each point.","section":"Fig. 2c,d"},{"comment":"The benchmarking spans a wide range of wavelengths (632–1577 nm); since propagation loss is wavelength-dependent, a brief note on this caveat would make the comparison fairer and more informative.","section":"Fig. 2e and Table S1"},{"comment":"The phrase 'waveguide propagation loss as small as 0.32 dB/cm' could be misread by nonspecialists as the straight-waveguide loss. Consider rewording to 'racetrack effective propagation loss' or 'best measured propagation loss (including bend sections)' to avoid ambiguity.","section":"Abstract and main text"}],"recommendation":"major_revision","confidential_remarks":"The paper is a strong candidate for publication if the reporting gaps are addressed. The central claims are plausible, but the coupling model and group index are essential for verifying the Qi and loss records. I would encourage the editor to request the actual transmission data and fitting code as part of the revision, since the statistics and fits are the backbone of the paper. Also, the r_eff value depends on a simulation that should be reported. The manuscript's comparison to prior work is generally careful, but the wavelength spread in Table S1 should be acknowledged. No issues of novelty or authorship attribution came to my attention."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Short version: this is a genuine process-and-device advance. The Ar+Cl2 etch adapted from lithium niobate work gives residue-free, smooth-sidewall BTO waveguides, and the device numbers back that up: Qi = 1.35e6 and 0.32 dB/cm loss on a racetrack, with a statistical series across straight fractions and two bend radii. That is the first Qi>1M in monolithic BTO and, by a clear margin, the lowest loss reported on that platform. The modulator result (VpiL = 0.54 V·cm, reff ≈ 162 pm/V) is also competitive, and the benchmark table against prior BTO work is useful.\n\nI don't think the central claim is shaky. The main soft spot is that two inputs into the headline numbers are never stated explicitly. α is converted from Qi using α = 2πng/(Qiλ)·10log10e, but no group index is reported; n ≈ 2.26 appears in the EO section, but ng for this 180-nm rib could be a few percent off, and α scales linearly with it. Similarly, the paper reports Qi and QL for individual resonances but not the coupling Qc or the extinction-ratio/coupling model used to separate them. The label 'near-critical coupling' is not enough to verify. If the coupling state were mis-assigned, Qi could shift tens of percent. These are reporting gaps, not demonstrated errors, but for a record claim they matter.\n\nMinor things: the etch recipe details (Cl2 fraction, pressure, power, Cr mask removal conditions) are omitted, which limits reproducibility even for experts. The AFM sidewall angle of ~60° sits awkwardly with 'high verticality' — 60° is not vertical, and that phrase should be softened. The straight-waveguide loss of 0.3 dB/cm is honestly labeled as an extrapolation, but the 0.32 dB/cm 'best' figure is itself a length-weighted average over straight and bent sections, not a pure straight-waveguide number, so readers should not over-interpret it. There are also a few typos and a comma splice or two.\n\nWho it's for: anyone working on BTO or on monolithic ferroelectric photonics more broadly. It deserves a serious referee. I would accept it for review and ask for the missing reporting as major-but-remedial revisions.","headline":"Record BTO resonator results that are probably right but need fuller reporting of the Q-extraction inputs before the headline numbers are fully load-bearing.","tokens_in":9865,"tokens_out":1620,"would_cite":true,"duration_ms":16952,"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":"A redeposition-free dry etch for barium titanate-on-insulator photonics is claimed to produce microresonators with intrinsic quality factor above one million and propagation loss as low as 0.32 dB/cm, the lowest reported in any BTO-based…","keywords":["barium titanate","integrated photonics","microresonator","quality factor","electro-optic modulator","Pockels effect","dry etching","propagation loss"],"falsifier":"Remeasure the linewidth of the same 200 micrometre racetrack and convert to propagation loss using a group index independently determined from the resonator free spectral range; if the resulting $\\alpha$ is not within the reported statistical spread around 0.32 dB/cm, the headline loss is an artifact of the assumed $n_g$. Alternatively, fabricate identical waveguides with the Cl2 flow turned off: if the intrinsic Q does not drop when sidewall redeposition returns, the etch is not the cause of the low loss.","tokens_in":8857,"feed_emoji":"⚡","tokens_out":8357,"duration_ms":82761,"temperature":0.7,"pith_summary":"The paper claims that a dry-etch recipe for barium titanate-on-insulator waveguides removes the redeposition and roughness problem that has limited monolithic BTO photonics, and that this unlocks long photon lifetimes in a material with an extremely large Pockels coefficient. The central evidence is microracetrack resonators with intrinsic quality factor $Q_i = 1.35 \\times 10^6$ and a best waveguide propagation loss of $0.32$ dB/cm, both claimed as records for any BTO-based platform. A Mach-Zehnder modulator built on the same waveguides shows $V_\\pi L = 0.54$ V$\\cdot$cm and an effective electro-optic coefficient $r_{\\mathrm{eff}} = 162$ pm/V. If correct, the result matters because it puts a strong electro-optic material into the same loss regime as mature integrated photonics, making devices like low-loss modulators and frequency combs practical.","feed_headline":"Barium titanate resonators pass the million-Q mark","feed_subtitle":"A redeposition-free argon-chlorine etch gives 0.32 dB/cm loss and a 0.54 V·cm modulator in monolithic BTO.","key_machinery":"The central object is the etch process itself: Ar+ ion milling with a small Cl2 fraction, run in an inductively coupled plasma reactive ion etching tool with a chromium hard mask on a 340 nm a-oriented BTO film on SiO2. The Ar+ ions provide physical etching, and the Cl2 reacts with sputtered barium and titanium to form volatile byproducts that are pumped away instead of redepositing on the sidewalls. That mechanism directly carries the argument: it converts the usual source of scattering loss, namely sidewall redeposition and roughness, into a residue-free surface, which is what the high $Q_i$ and low propagation loss are attributed to. The quality-factor analysis across racetrack resonators with varying straight fraction, using $\\alpha = 2\\pi n_g/(Q_i \\lambda) \\cdot 10\\log_{10} e$, then separates straight-waveguide loss from bending loss.","core_discovery":"The paper's central claim is that adding a small fraction of Cl2 gas to Ar+ ion milling makes monolithic BTO etching redeposition-free, producing sidewalls that are smooth (0.29 nm rms) and sufficiently vertical (about 60 degrees), and that this fabrication advance is what allows BTO microresonators to reach intrinsic quality factors above 1 million for the first time. The authors report $Q_i = 1.35 \\times 10^6$ for a 200 micrometre-turn-radius microracetrack, $Q_i = 0.84 \\times 10^6$ for a 50 micrometre-radius ring, and a straight-waveguide propagation loss extrapolated to about 0.3 dB/cm, with 0.32 dB/cm as the best individual measurement. They also report that common wet-cleaning chemistries (RCA-1, hot piranha, KOH, buffered oxide etch) either damage BTO or fail to remove redeposited material, which is why the redeposition-free etch matters. The electro-optic demonstration supports the platform's purpose: a single-arm Mach-Zehnder interferometer with $V_\\pi = 1.44$ V and $V_\\pi L = 0.54$ V$\\cdot$cm, corresponding to $r_{\\mathrm{eff}} \\approx 162$ pm/V.","pith_inferences":["A direct dispersion measurement, for example from the resonator free spectral range, would supply $n_g$ explicitly; without it, the 0.32 dB/cm figure inherits any error in the assumed group index, since the quoted loss scales linearly with $n_g$.","If the same Ar+Cl2 chemistry works for other perovskite oxides whose etch products are involatile, it could generalize the redeposition-free sidewall result beyond BTO.","A test that separates sidewall scattering from material absorption, for example by varying waveguide width or measuring quality factor versus temperature, would show whether further loss reduction should come from etch refinement or from film quality.","Switching from planar poling at 120 V to sidewall electrodes could raise the effective electro-optic coefficient closer to bulk BTO values, making the modulator figure of merit even stronger."],"forward_implications":["BTO microresonators can hold photons for roughly a nanosecond at telecom wavelengths, which makes resonant nonlinear optics and electro-optic modulation practical in a monolithic platform.","The same etch produces a single-arm Mach-Zehnder modulator with $V_\\pi L = 0.54$ V$\\cdot$cm, so low switching voltage and low propagation loss no longer require a hybrid silicon or silicon-nitride host.","The loss values place BTO within reach of applications the paper enumerates: electro-optic and Kerr frequency combs, higher harmonic generation, narrow-linewidth lasers, and squeezed light sources.","Because the process avoids damaging wet cleans, it sidesteps the Curie-temperature constraint that made high-temperature residue removal impractical for BTO."],"supporting_citations":[{"why":"Supplies the Ar+ plus Cl2 etching approach for lithium niobate that the paper adapts to barium titanate.","marker":"[25]"},{"why":"Documents the non-volatile barium byproducts and high-temperature removal difficulty that motivate a redeposition-free etch.","marker":"[22]"},{"why":"Shows argon plasma etching for smooth LiNbO3 sidewalls and the wet-cleaning route that does not transfer to BTO.","marker":"[23]"},{"why":"Provides the lithium tantalate wet-cleaning comparison that proves ineffective for BTO sidewalls.","marker":"[24]"},{"why":"Reports previous monolithic BTO ring resonators with $Q_i$ around $5 \\times 10^5$, the baseline this work exceeds.","marker":"[16]"},{"why":"Reports absorption loss and Kerr nonlinearity in BTO waveguides, a recent loss benchmark this work improves on.","marker":"[18]"},{"why":"Provides earlier monolithic BTO modulator results and the formula used to extract $r_{\\mathrm{eff}}$ from $V_\\pi L$.","marker":"[15]"},{"why":"Gives frequency-dependent Pockels coefficients and the $r_{\\mathrm{eff}}$ expression used to interpret the modulator measurement.","marker":"[17]"},{"why":"Early BTO thin-film waveguide modulator, used as a historical benchmark in the loss comparison.","marker":"[14]"}],"fun_headline_variants":["BTO photonics hits Q>1M with redeposition-free etch","Monolithic BTO reaches Q>1M, 0.32 dB/cm loss","First BTO microresonators with Q above 1 million","Low-loss BTO integrated photonics passes Q>1M"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The entire loss and quality-factor report rests on the conversion from measured resonance linewidths to intrinsic Q and propagation loss using a coupling model and a group index $n_g$ that the paper does not state, so any error in those inputs scales the headline numbers proportionally.","fun_headline_variants_meta":{"raw":{"variants":["BTO photonics hits Q>1M with redeposition-free etch","Monolithic BTO reaches Q>1M, 0.32 dB/cm loss","First BTO microresonators with Q above 1 million","Low-loss BTO integrated photonics passes Q>1M"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000195,"raw_usage":{"total_tokens":1390,"prompt_tokens":1010,"completion_tokens":380,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":626,"completion_tokens_details":{"reasoning_tokens":300}},"tokens_in":626,"tokens_out":380,"duration_ms":4591,"temperature":1.0,"reasoning_tokens":300,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-06T14:55:14.289440+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Remeasure the linewidth of the same 200 micrometre racetrack and convert to propagation loss using a group index independently determined from the resonator free spectral range; if the resulting $\\alpha$ is not within the reported statistical spread around 0.32 dB/cm, the headline loss is an artifact of the assumed $n_g$. Alternatively, fabricate identical waveguides with the Cl2 flow turned off: if the intrinsic Q does not drop when sidewall redeposition returns, the etch is not the cause of the low loss.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the Ar+ plus Cl2 etching approach for lithium niobate that the paper adapts to barium titanate."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Documents the non-volatile barium byproducts and high-temperature removal difficulty that motivate a redeposition-free etch."},{"cited_title":"& Bernal, M.-P","cited_arxiv_id":null,"evidence_quote":"Shows argon plasma etching for smooth LiNbO3 sidewalls and the wet-cleaning route that does not transfer to BTO."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Reports previous monolithic BTO ring resonators with $Q_i$ around $5 \\times 10^5$, the baseline this work exceeds."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Reports absorption loss and Kerr nonlinearity in BTO waveguides, a recent loss benchmark this work improves on."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provides earlier monolithic BTO modulator results and the formula used to extract $r_{\\mathrm{eff}}$ from $V_\\pi L$."},{"cited_title":"et al.Barium titanate and lithium niobate permittivity and Pockels coefficients from megahertz to sub-terahertz frequencies","cited_arxiv_id":null,"evidence_quote":"Gives frequency-dependent Pockels coefficients and the $r_{\\mathrm{eff}}$ expression used to interpret the modulator measurement."},{"cited_title":"& Buchal, C","cited_arxiv_id":null,"evidence_quote":"Early BTO thin-film waveguide modulator, used as a historical benchmark in the loss comparison."}],"review_version":1}