{"id":"0ab946a2-a97d-4d56-80d4-82dd597d7713","arxiv_id":"2411.17360","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"Oxygen doping of sputtered ferroelectric Al0.73Sc0.27N reduces leakage current several-fold and reversibly changes the as-grown film polarity from nitrogen- to metal-polar.","lead":"This paper shows that adding oxygen during sputtering of ferroelectric AlScN films cuts electrical leakage by several times and can flip the film's as-grown polarization direction. This matters because high leakage near the switching field is a known obstacle for nitride ferroelectric memory and sensor devices.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The claimed bulk-O mechanism is not uniquely supported: surface O-rich protrusions and interface oxide appear at the same doping levels as the electrical changes.","rationale":"Both the reader and this pass converge on the same load-bearing assumption: the O measured by EDS/ToF-SIMS is electrically active bulk substitutional O. The paper deserves credit for multiple independent electrical signatures (P-E sharpening, J-E tail reduction, steady-state leakage) and PFM confirmation of polarity in some samples; those are real observations. The vulnerability is not the measurements but their attribution: the doping series is a single gas-flow knob, so bulk O content, surface protrusion density, interface chemistry, and film stress all change together. The surface protrusions are explicitly O-rich and appear exactly in the doping window where the leakage/polarity transitions occur, so surface/interface O is a concrete confounder. Because O quantification is self-described as qualitative and no atomic-column or site-occupancy evidence is provided, the central mechanism is underdetermined. The proposed check—removing the top surface layer before electrode deposition—separates the surface/interface contribution from the bulk contribution. If the effect survives, the paper's causal claim is much stronger; if not, the headline claim would need re-scoping to 'O-rich growth conditions alter surface/interface properties,' which is a different and less general result. The reader's CONDITIONAL verdict is appropriate; no change is needed.","tokens_in":12994,"tokens_out":4438,"duration_ms":46813,"concrete_test":"Fabricate two additional Nmix = 12 and Nmix = 15 films. Before depositing the Pt top electrode, remove the top ~20 nm (the surface protrusion/oxide layer) from one set by low-angle Ar-ion milling or reactive ion etching, leaving the other set as-grown; then repeat the leakage measurements of Figure 5c-d and the unipolar polarity test of Figure 6g-h on both sets. If the ~fourfold/order-of-magnitude leakage reduction and the M-polar as-deposited state persist after surface-layer removal, the bulk-O mechanism is supported; if they disappear or weaken substantially, the reported effects are dominated by surface/interface oxygen rather than bulk substitutional O.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The abstract and conclusion attribute the leakage reduction and polarity flip to O substituting on N sites in the bulk wurtzite lattice, but the experiments do not separate this bulk mechanism from surface/interface oxygen. Figure 3 shows that from Nmix = 8 sccm onward the film surface develops protruding particles that at Nmix = 12-15 sccm cover the surface almost homogeneously, and STEM-EDS (Figure 4) identifies these as O-rich amorphous oxynitrides; the same threshold (8-10 sccm) is where the J-V polarity test shows mixed and then M-polar as-deposited states (Figure 6). The 5 nm native oxide on the top surface (and any interface layer under the Pt electrode) can alter leakage through contact resistance and barrier heights, and the roughness increase at Nmix = 15 is itself a known source of PFM phase/amplitude contrast. O quantification is explicitly qualitative: SEM-EDS carries a ~1.2 at% O background from buried SiO2 and includes surface O, ToF-SIMS is only relative, and the authors note FIB and surface-oxidation can inflate the bulk O signal. No atomic-scale evidence (e.g., EELS or APT) shows that the measured O occupies N sites in grain interiors rather than grain boundaries, surface protrusions, or interfaces. If the electrical effects track the surface/interface O-rich layer instead of bulk substitutional O, the central technical claim—that gas-phase O doping of the bulk Al0.73Sc0.27N reduces leakage and sets polarity—does not follow.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The manuscript reports a systematic study of oxygen incorporation during reactive sputtering of ferroelectric Al0.73Sc0.27N thin films, achieved by mixing a N2/O2 source into the nitrogen gas flow. The claims are that O-doped films show an almost fourfold leakage-current reduction near the coercive field, roughly an order-of-magnitude reduction of steady-state leakage at sub-coercive fields, a polarity transition from nitrogen-polar to metal-polar as-deposited films with increasing O content, and only minor structural degradation (a <20% increase in XRD rocking-curve width). These claims are supported by P-E and J-E hysteresis measurements, quasi-static leakage measurements, unipolar J-V polarity tests, PFM phase/amplitude mapping, HRXRD, SEM/STEM imaging, and STEM-EDS elemental mapping. The authors attribute the leakage improvement to O filling N vacancies and forming deep-level DX centers, and attribute the polarity flip to O-driven destabilization of the N-polar growth mode.","tokens_in":13241,"tokens_out":3605,"duration_ms":37515,"significance":"If the claims hold, the paper provides a practically relevant and seemingly simple gas-phase doping route to reduce leakage and engineer the as-deposited polarity of ferroelectric AlScN, which is a key obstacle for memory and MEMS applications. The strengths of the paper include direct electrical measurements of leakage and switching behavior, independent PFM confirmation of the polarity transition, multi-technique structural characterization, and an unusually candid discussion of the quantitative limitations of the O-content measurements. The central weakness is that the bulk-substitutional mechanism is not uniquely established: the electrical changes occur at the same doping range in which pronounced surface protrusions and a native oxide appear, and the O quantification is explicitly qualitative. The significance is therefore conditional on separating bulk from surface/interface contributions.","major_comments":[{"comment":"The manuscript attributes the leakage reduction and polarity flip to oxygen incorporated on nitrogen sites in the bulk wurtzite lattice, but the data do not exclude an alternative explanation in which the electrical changes are caused or strongly modified by the O-rich surface protrusions, the ~5 nm native oxide, or interfacial layers. The surface protrusions first appear at Nmix = 8 sccm and cover the surface by Nmix = 12 sccm (Figure 3e-h), which is exactly the doping range where the mixed-polarity state and the second-stage leakage drop are observed (Figures 5 and 6). The authors themselves note that SEM-EDS carries a ~1.2 at% O background from buried SiO2, that ToF-SIMS is only relative, and that FIB and surface oxidation can inflate the bulk O signal. Without atomic-scale evidence such as EELS or atom-probe tomography showing O on nitrogen sites in grain interiors, the central claim that gas-phase O doping of the bulk reduces leakage and sets polarity is not uniquely demonstrated. I recommend adding a control experiment (e.g., a film with the surface protrusions removed or an interface-engineered stack) or direct site-specific chemical analysis to separate bulk from surface/interface effects.","section":"Results and discussion: Determination of O-content; Effects of O-doping on the surface and bulk microstructure"},{"comment":"The headline claim of 'almost a fourfold reduction of the leakage current near the coercive field' is not supported by a single quantitative measurement: the text states a 'four-fold drop (2x at the positive side and 2x at the negative side)', which is a product of two separate branchwise factors rather than a directly observed fourfold ratio in the same device or polarity state. In addition, Figure 5c-d reports leakage current densities without error bars, device-to-device statistics, or a statement of how many capacitors were measured per composition; the experimental section mentions 100 averages but not the number of devices. The authors should report the extracted leakage values for each polarity branch, with uncertainties and sample counts, and revise the abstract/conclusion wording if the fourfold factor is a combined metric.","section":"Figure 5 and accompanying text"},{"comment":"The polarity assignment in Figure 6 is based on unipolar J-V responses from two pristine pads per composition, and PFM phase maps are shown for only three of the eight compositions (Nmix = 0, 8, 15 sccm). Given the substantial surface roughness and protrusions at Nmix ≥ 12 sccm, which the authors acknowledge can affect PFM contrast and electrode contact, the conclusion that the films are 'entirely M-polar' at Nmix = 12 and 15 sccm would be strengthened by additional pads per sample, a statistical measure of the switching-peak asymmetry, and PFM data for the 12 sccm sample. As written, the evidence supports a qualitative polarity trend but not the quantitative strength of the claim.","section":"Results and discussion: As-deposited polarization reversal via O-doping"}],"minor_comments":[{"comment":"The phrase 'changing from nitrogen- to metal-polar orientation' is clear, but the abstract does not mention the mixed-polarity intermediate region reported in the main text; consider adding a qualifier such as 'via intermediate mixed-polar states' for consistency.","section":"Abstract and Introduction"},{"comment":"The error bars in Figure 2b for the in-plane a lattice parameter are large and asymmetric, but this is not explained in the caption; the text mentions the larger error only in passing. Please add the uncertainty source to the caption.","section":"Figure 2"},{"comment":"The statement 'It can be assumed that O atoms are likely to replace an N atom (ON) in a Sc-coordinated tetrahedra first' would be better phrased as a hypothesis, because the authors later note that calculations for Sc-tetrahedra distortions are missing; the current wording overstates the structural evidence.","section":"Results and discussion: Effects of O-doping on the surface and bulk microstructure"},{"comment":"The sentence 'The use of this separate sample set on sapphire was necessary to conduct as the sample has to be transparent for transmission measurements' contains a grammatical error ('to conduct as'); it should read 'was necessary because the sample has to be transparent'.","section":"Experimental section"},{"comment":"The PFM sub-figures are labeled (i), (ii), (iii) within each panel but the caption refers to 'sub-figures i-l', which is inconsistent with the figure layout showing (a-h) electrical data and (i-k) PFM data; please align the labels.","section":"Figure 6"}],"recommendation":"major_revision","confidential_remarks":"The paper is likely to be of interest to the applied-physics community and the claims are potentially important. The main risk is that the observed electrical changes correlate with surface/interface oxygen as much as with bulk substitutional oxygen, and the current manuscript does not provide the experimental separation needed to support the bulk mechanism. This is fixable with additional experiments or more carefully qualified claims, so I recommend major revision rather than rejection. The reviewer should ask the authors to provide statistics for the leakage measurements and to clarify how the 'fourfold' claim is defined."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Bottom line: this is a useful, incremental experimental study, not a breakthrough. The systematic O2-in-N2 flow series is new for ferroelectric AlScN, and the leakage reduction and polarity flip are supported by several independent measurements: J-E tails, quasi-static leakage, P-E loop sharpening, and PFM phase contrast. I would send it to a serious referee, but the revision should tighten the statistics and the mechanistic claims.\n\nWhat the paper does well: the doping series is clean, the structural characterization is fairly complete, and the authors are honest about the limits of their O quantification. They flag the SEM-EDS background, the 5 nm native oxide, and FIB-related O influx. The finding that the c-axis texture degrades less than 20% at the highest doping is credible and worth knowing. The polarity flip from N- to M-polar via mixed states is the most novel piece, and the PFM data make it believable.\n\nWhere I would push back: the stress-test note is partly right but overstates the danger. The paper already acknowledges the surface oxide, the protrusions, and the rough surface at high O flow. It does not, however, prove that the electrical changes come from O substituting on N sites in grain interiors rather than from surfaces, interfaces, or grain boundaries. The protrusions appear at Nmix = 8 sccm, the same window where the polarity starts to flip, so the correlation is suggestive but not causal as presented. STEM-EDS shows \"homogeneous\" O in the bulk, but EDS cannot distinguish substitutional O from O in grain boundaries or from FIB damage without atomic-scale evidence such as EELS or APT. The authors' own statements about surface oxidation and FIB contamination weaken the bulk-site claim further. That is a real soft spot, but not fatal: the paper mostly couches the mechanism as speculation, and the qualitative result stands even if the site occupancy is unresolved.\n\nThe bigger soft spot is statistical reporting. There are no error bars on the leakage or P-E data, and the \"fourfold\" near-coercive-field reduction appears to be assembled from a 2x improvement on each branch of one representative measurement. The order-of-magnitude sub-coercive reduction in the highest-doped films is more robust because it appears in separate quasi-static data, but I would still want repeated devices and confidence intervals before quoting the numbers in a citation.\n\nBottom line for you: this is a good experimental contribution for the AlScN device community, worth refereeing, and worth citing with care. The mechanism should be softened in revision, and the headline quantitative claims need statistics. If those are fixed, I would be comfortable with it in a solid applied-physics journal.","headline":"Useful incremental AlScN study: systematic O-doping lowers leakage and flips as-grown polarity, but the bulk-substitution mechanism and the headline numbers need tougher evidence before I'd trust them quantitatively.","tokens_in":13866,"tokens_out":2065,"would_cite":true,"duration_ms":24528,"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":"Oxygen doping of ferroelectric Al0.73Sc0.27N cuts leakage by up to an order of magnitude and flips as-grown polarity.","keywords":["ferroelectric AlScN","oxygen doping","leakage current","polarity control","reactive sputtering","wurtzite nitride","DX centers","thin films"],"falsifier":"A decisive check would be atom-column-resolved STEM-EELS across grain boundaries in the Nmix = 15 film: if most of the oxygen signal is confined to grain boundaries, surface protrusions, or the interface rather than the bulk lattice, the proposed O_N substitution and DX-center mechanism would not be the operative cause of the leakage drop.","tokens_in":12783,"feed_emoji":"⚡","tokens_out":6569,"duration_ms":55933,"temperature":0.7,"pith_summary":"This paper claims that deliberately adding oxygen during sputter deposition of ferroelectric Al0.73Sc0.27N thin films reduces leakage currents by almost a factor of four near the coercive field and by roughly an order of magnitude at sub-coercive fields, while leaving the crystal structure largely intact. The oxygen is introduced as a small fraction of the nitrogen gas flow and ends up distributed through the bulk of the film, where it is argued to occupy nitrogen sites and act as deep-level traps that suppress leakage. The doping also flips the as-deposited polarization direction from nitrogen-polar to metal-polar above a threshold oxygen concentration, giving process control over the initial domain state. If these effects hold up, oxygen flow becomes a simple knob for making ferroelectric nitride devices more practical.","feed_headline":"Oxygen doping cuts AlScN leakage currents fourfold","feed_subtitle":"Adding oxygen during sputtering also flips the as-grown polarity from nitrogen- to metal-polar.","key_machinery":"The load-bearing mechanism is substitutional oxygen on nitrogen sites (O_N) in the wurtzite lattice. Oxygen replaces nitrogen first in scandium-coordinated tetrahedra (favored by bond-energy arguments) and, at higher concentrations, in aluminum tetrahedra, forming deep-level DX-center-like traps that capture electrons and reduce leakage; the accompanying local lattice distortion and change in film stress are proposed to drive the N-polar to M-polar inversion above Nmix ≈ 8–10 sccm.","core_discovery":"The central discovery is that intentional oxygen incorporation, achieved by mixing a N2 with 2% O2 gas source into the reactive sputter gas, is a viable doping route for w-Al0.73Sc0.27N: it cuts switching-related leakage near the coercive field by almost fourfold and steady-state leakage at low fields by about one order of magnitude, and it reverses the as-grown film polarity from N-polar to M-polar as the Nmix flow rises from 8 to 15 sccm. XRD and STEM show the 0002 rocking-curve width increases by only about 20% at the highest doping, with no chemical segregation or Al2O3 interface layer, so the leakage and polarity improvements are not bought at the price of destroying the wurtzite texture. The authors attribute the initial leakage drop to filling of nitrogen vacancies by oxygen and the formation of deep DX-center-like acceptor states, and the later polarity reversal to lattice strain and changes in local bonding that destabilize the N-polar growth mode.","pith_inferences":["Editorial inference: if the leakage reduction really comes from O_N deep traps, then combining oxygen doping with other intentional dopants could allow independent tuning of leakage and polarity, a combination not explored in this paper.","Editorial inference: because the O quantification is qualitative, a quantitative calibration (for instance Rutherford backscattering) would be needed before the Nmix values could be transferred to another deposition tool.","Editorial inference: the polarity reversal near Nmix = 8–10 sccm is reported for as-deposited films; whether the mixed-polar domain state at the boundary is stable under cycling is not tested, so a natural follow-up is an endurance measurement on Nmix = 8 and 10 films."],"forward_implications":["O-doped Al0.73Sc0.27N capacitors can be switched with a smaller leakage tail, which directly improves the margin for reading and writing ferroelectric memory cells.","Setting the oxygen flow during sputtering selects the as-deposited polarity, so devices could be built with the preferred initial polarization orientation without a post-growth poling step.","Because bulk crystallinity and c-axis texture are largely preserved, the doping route is compatible with existing sputter-based fabrication flows.","The same gas-phase doping approach may reduce leakage in other wurtzite ferroelectrics such as Al1-xBxN and Al1-xYxN, which the paper argues share the relevant structural features."],"supporting_citations":[{"why":"Establishes wurtzite AlScN as a ferroelectric and defines the coercive-field range that motivates the leakage problem.","marker":"[1]"},{"why":"Provides the comparative leakage behavior of AlScN capacitors on Si substrates that this doping study aims to improve.","marker":"[16]"},{"why":"Documents oxygen as a deep-level impurity in III-nitrides, the defect picture the paper adopts for the leakage reduction.","marker":"[27]"},{"why":"Prior demonstration of oxygen-atom-incorporated ferroelectric AlScN capacitors for multi-level operation that this work extends with systematic doping and polarity control.","marker":"[28]"},{"why":"Theoretical prediction of defects and oxygen impurities in wurtzite Al1-xScxN that the bandgap and DX-center measurements are compared against.","marker":"[31]"},{"why":"Hybrid-functional calculations of DX centers in AlN and GaN that support the O_N acceptor-state interpretation.","marker":"[32]"},{"why":"Shows polarity inversion of AlN by dopants, the precedent for using oxygen to control as-deposited polarity.","marker":"[39]"},{"why":"Shows that oxygen in sputtering gas degrades AlN crystallinity, the contrast case this paper claims to avoid for AlScN.","marker":"[44]"}],"fun_headline_variants":["Oxygen cuts AlScN leakage 4x, flips polarity","O-doping in AlScN: 4x less leakage, polarity flip","Oxygen in AlScN: fourfold leakage cut, polarity reversal","Oxygen flips AlScN polarity, cuts leakage by 4x"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The paper's claims assume that the oxygen counted by SEM-EDS and ToF-SIMS is electrically active oxygen sitting on nitrogen sites inside the wurtzite lattice, and that the Nmix gas-flow setting reliably tracks that bulk doping level.","fun_headline_variants_meta":{"raw":{"variants":["Oxygen cuts AlScN leakage 4x, flips polarity","O-doping in AlScN: 4x less leakage, polarity flip","Oxygen in AlScN: fourfold leakage cut, polarity reversal","Oxygen flips AlScN polarity, cuts leakage by 4x"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000829,"raw_usage":{"total_tokens":3669,"prompt_tokens":1041,"completion_tokens":2628,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":657,"completion_tokens_details":{"reasoning_tokens":2546}},"tokens_in":657,"tokens_out":2628,"duration_ms":30575,"temperature":1.0,"reasoning_tokens":2546,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-12T12:12:04.985943+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"A decisive check would be atom-column-resolved STEM-EELS across grain boundaries in the Nmix = 15 film: if most of the oxygen signal is confined to grain boundaries, surface protrusions, or the interface rather than the bulk lattice, the proposed O_N substitution and DX-center mechanism would not be the operative cause of the leakage drop.","supporting_citations":[],"review_version":1}