{"id":"208a45c2-b3e8-43b9-a983-d48915567acc","arxiv_id":"2607.05617","paper_version":2,"verdict":"CONDITIONAL","confidence":"HIGH","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":4,"one_line_summary":"Pulse-driven accumulative switching in HZO obeys three domain-radius scaling regimes (α_local >1, ≈1, <1) set by the competition between field-on excitation and field-off relaxation.","lead":"Phase-field simulations show that pulse-driven polarization buildup in HZO follows three domain-growth regimes—superlinear, steady, then decelerating—tracked by a local kinetic exponent of the switched-domain radius. The map of how pulse amplitude, on/off times, and starting domain geometry move those regimes offers design knobs for low-power ferroelectric memory and neuromorphic synapses.","discovery_kind":"extension","skeptic_critique":{"model":"grok-4.5","headline":"The three-regime α_local claim is largely a geometric finite-size effect of the 80 nm box and R definition, not a robust nonequilibrium kinetic law.","rationale":"The Reader correctly flags the 2D continuum idealizations and calibrated Landau coefficients as the weakest modeling assumptions and assigns CONDITIONAL. That is right on external validity. The more load-bearing internal issue for the strongest claim itself is that the three α_local regimes are not cleanly separated from the finite computational domain and the particular definition of R. The paper already shows geometry controls the exponent (QCoD vs CeD) and attributes late deceleration to boundary interactions; a simple L-scaling test would settle whether any intermediate self-similar window survives in the continuum limit. Because the work remains a useful, coherent organizing result inside its model, the verdict stays CONDITIONAL rather than REJECT; the concrete size-scaling check is the minimal addition that would make the scaling laws more than box-size phenomenology.","tokens_in":20690,"tokens_out":567,"duration_ms":6509,"concrete_test":"Re-run the CeD series of Fig. 6/12 on L=160 nm and L=320 nm (same mesh density, same Eapp,n,max, Ton, Toff, R0). Extract the n-window where α_local stays within 0.9–1.1 and the n at which α_local first drops below 0.8. If that window lengthens proportionally to L (or the drop vanishes until walls hit the boundary), the decelerating regime is a finite-size artifact and the claimed universal three-regime law weakens.","verdict_should_be":"CONDITIONAL","load_bearing_attack":"The central claim is that R follows universal regimes α_local>1 → ≈1 → <1 set by excitation–relaxation competition (Eq. 7; Figs. 3d, 6, 11–12). Inside the model this is true, but the late-time α_local<1 branch is forced by the finite 80 nm domain with Neumann boundaries once the switched front approaches the edge (Results: “depletion of switchable polarization and increased boundary interactions”; CeD saturates after few pulses). Early α_local>1 is also geometry-tied: free circular walls (CeD/ED) accelerate while QCoD stays ≈1 under identical pulses (Fig. 3d). Thus the “unified scaling framework” largely re-describes free radial expansion until confinement, rather than an intrinsic continuum kinetic law independent of box size and R0. Without size-independent intermediate scaling or continuum-limit checks, the regimes cannot be treated as device-ready design laws for real polycrystalline HZO.","agreement_with_reader":"partial"},"referee_report":{"model":"grok-4.5","summary":"This computational paper uses a 2D time-dependent Landau–Ginzburg phase-field model of HZO to study accumulative polarization switching under trains of sub-coercive pulses. By varying initial domain geometry (single and multi-domain placements), pulse amplitude Eapp_n,max, Ton, and Toff, the authors track average polarization, accumulated polarization, an effective switched-domain radius R (and multi-domain variants Rc, Req), and a local kinetic exponent α_local = d ln(R−R0)/d ln n. They report three regimes—α_local>1 (superlinear acceleration), ≈1 (steady self-similar growth), and <1 (deceleration from confinement, depletion, and off-time back-switching)—and attribute the transitions to competition between field-driven Type-III domain-wall advance during Ton and spontaneous relaxation during Toff. The abstract and conclusion present this as a unified scaling framework with design guidelines for HZO memory and neuromorphic devices.","tokens_in":21013,"tokens_out":1478,"duration_ms":23075,"significance":"If the three-regime picture is robust beyond the simulated box and idealized continuum model, the work would give a practical kinetic language for pulse-protocol design in accumulative ferroelectric switching, which is relevant to multi-level FeFET/FTJ and neuromorphic weight update. Strengths include a systematic parameter sweep across domain geometries and pulse parameters, mutual consistency among morphologies, Pacc(n), R(n), and α_local, and a clear microscopic taxonomy of excitation/relaxation pathways. The local-exponent diagnostic is a useful, falsifiable way to report nonequilibrium domain growth under pulsed drive. The main significance is therefore methodological and interpretive within continuum phase-field kinetics, not yet a validated device law for polycrystalline HZO.","major_comments":[{"comment":"Model §II and Results (Figs. 3d, 6, 11–12; L = 80 nm × 80 nm, Neumann boundaries): the late-time α_local < 1 branch is tightly tied to geometric confinement and depletion once the switched front approaches the box edge (explicitly stated for CeD/ED). Early α_local > 1 is likewise geometry-selective (free circular walls accelerate; QCoD stays near unity under identical pulses). The claim of a “unified scaling framework” and device design guidelines therefore requires either (i) a documented size-independence / continuum-limit check showing an intermediate α_local ≈ 1 window that survives larger L and different R0, or (ii) an explicit reframing that the regimes are finite-domain kinetic signatures of free radial expansion until confinement. The brief remark that other lateral sizes give “qualitatively similar” accumulation is not sufficient without R(n) and α_local(n) for at least two size","section":"Model; Results Figs. 3d, 6, 11–12"},{"comment":"Abstract, §I motivation, and §IV conclusion assert “design guidelines for HZO-based memory and neuromorphic devices.” The model omits thermal noise, oxygen vacancies, grain boundaries, electrode interfaces, and 3D polycrystalline texture (Model, Eqs. 1–6; fixed Landau set α̂, β̂, γ̂ from prior calibration). Without quantitative comparison to measured accumulative switching (e.g., pulse-number dependence of remanent P or FeFET threshold shift under comparable Ton/Toff), or a clear statement of which predictions survive those omissions, the device-guideline claim overreaches the evidence. Either add a validation/benchmark section against published HZO pulse-train data or temper the claim to continuum-model design rules.","section":"Abstract; §I; §IV"},{"comment":"Eq. (7) and multi-domain Results (Figs. 8, 10): R, Rc, and Req are central observables, but the manuscript does not specify the operational definition used to extract radius from the discrete polarization field (threshold on Pn, area-equivalent radius πR² = A_switched, perimeter-based, etc.), nor how α_local is differentiated numerically from discrete n (smoothing, window, handling of staircase Ton/Toff structure). For multi-domain coalescence, Req’s construction is especially load-bearing for the claimed cooperative acceleration. Please define R extraction and α_local computation precisely and show that the >1 / ≈1 / <1 classification is stable under reasonable threshold and smoothing choices.","section":"Eq. (7); Results Figs. 8, 10"}],"minor_comments":[{"comment":"Notation for the local exponent alternates between α_local, α_loc, and α loc across text and figure captions (e.g., Fig. 3d vs. Eq. 7). Standardize.","section":"Eq. (7); Figs. 3, 6, 11"},{"comment":"QCoD is defined as “quarter-corner domain” in the Results text but as “quasi-centre domain” in the Fig. 2 caption. Correct the inconsistency.","section":"Results; Fig. 2"},{"comment":"Several figure panels are dense (e.g., multi-panel morphology sequences in Figs. 2, 7, 9). Adding scale bars, pulse-number labels on morphologies, and a short legend for color scale of Pn would improve readability.","section":"Figs. 2, 7, 9"},{"comment":"Minor grammar/typos: “inducsed” (Introduction), “RESUL TS” / “SUMMAR Y” spacing artifacts, “Toff = 0.4µ.” incomplete unit in Fig. 2 caption, and mixed µs/ns units for the same quantities without a conversion note.","section":"Introduction; section headers; Fig. 2"},{"comment":"The Type-I/II/III excitation–relaxation taxonomy is useful but should be cross-referenced more clearly to the prior Saha et al. usage versus what is newly quantified here, so readers can separate nomenclature from new kinetic results.","section":"Model; Fig. 1"}],"recommendation":"major_revision","confidential_remarks":"The study is a solid, incremental extension of the Saha et al. (2019) HZO phase-field setup rather than a first-principles or experimentally closed treatment. That is acceptable for a computational materials journal if claims are scoped to the model; the main editorial risk is overselling “device design guidelines” and “universal” scaling without size checks or data comparison. I would not reject on novelty alone, but I would not accept until the finite-size and claim-scope issues are addressed."},"author_rebuttal":null,"desk_editor":{"model":"grok-4.5","letter":"Punchline: this is a careful phase-field parameter study, not a new microscopic theory. They take the established TDLG/HZO setup (Saha et al. and the Type-I/II/III language) and add a local kinetic exponent α_local = d ln(R−R0)/d ln n on the switched radius. The new piece is the systematic three-regime map—superlinear, near-linear, then decelerating—and how seed geometry plus {E, Ton, Toff} move the crossovers.\n\nWhat they do well is the sweep. Single-domain seeds (center, edge, corner, quarter-corner), multi-domain DCC/CCC cases, and dense 2D maps of α_local versus amplitude, Ton, and Toff are mutually consistent: morphologies, Rc(n), Pacc, and α_local tell the same story. Early α_local > 1 is the interesting bit—repeated sub-coercive pulses progressively help irreversible wall motion—and longer Ton or higher E clearly extends that window while longer Toff kills accumulation. For people who already simulate multi-level FeFET or synapse pulse trains, those maps are usable design intuition inside the model.\n\nSoft spots, in proportion. The stress-test is half right: late α_local < 1 is heavily geometric once the front hits the 80 nm Neumann box, and free circular walls accelerate more than confined seeds under identical drive. The paper names confinement and depletion, then still sells a “unified scaling framework” and device guidelines a bit hard. No size-independence check, no thermal noise/defects/grains/interfaces, fixed Landau set from prior calibration, no experiment, no code. That does not break the internal claim; it caps how far you should export the numbers to real polycrystalline HZO.\n\nWho it is for: continuum FE modelers and HZO neuromorphic/device people who need pulse-protocol intuition. Not a general ferroelectrics audience. Math and numerics look coherent; citations track the right accumulation literature without weird gaps.\n\nI would send it to peer review. Expect revision that softens universality language and flags finite-size more clearly. Engage if you work in this niche; skim the maps if you only need the regime language.","headline":"Useful α_local regime maps on top of the Saha TDLG/HZO model; late-time deceleration is partly finite-size, and nothing is experiment-checked, but the pulse-parameter organization is real and worth refereeing.","tokens_in":21654,"tokens_out":586,"would_cite":true,"duration_ms":19148,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"grok-4.5","headline":"Sub-coercive pulse trains reverse HZO polarization through three domain-growth regimes set by a local kinetic exponent.","keywords":["HZO","accumulative polarization switching","domain-growth kinetics","local kinetic exponent","phase-field model","sub-coercive pulses","ferroelectric neuromorphic devices","domain-wall dynamics"],"falsifier":"Measure switched-domain radius versus pulse number on real HZO capacitors or FeFETs under matched sub-coercive pulse trains; if the extracted local exponent never exceeds one early on or never drops below one later, or if geometry and pulse-off time do not shift the transitions as predicted, the scaling framework fails.","tokens_in":21563,"feed_emoji":"⚡","tokens_out":746,"duration_ms":6749,"temperature":0.7,"pith_summary":"This paper argues that accumulative polarization switching in ferroelectric HZO under repeated sub-coercive pulses is not a single process but proceeds through three kinetic regimes of domain growth. Using a continuum phase-field model, the authors track the effective switched-domain radius and define a local kinetic exponent that measures how that radius grows with pulse number. Early on the exponent exceeds one, meaning each pulse expands domains faster than the last; later it settles near one for steady growth and then falls below one as walls hit boundaries, switchable material runs out, and relaxation during pulse-off intervals pulls walls backward. The transitions are controlled by the tug-of-war between field-driven excitation while the pulse is on and spontaneous back-switching while it is off, and they depend strongly on the starting domain geometry. Raising pulse amplitude or on-time lengthens the superlinear stage; lengthening the off-time favors relaxation and slows accumulation. The result is a scaling framework that links microscopic wall motion to macroscopic polarization build-up, offering concrete design rules for low-power multi-level memories and neuromorphic synapses based on HZO.","feed_headline":"Three scaling regimes govern pulse-driven HZO switching","feed_subtitle":"A local kinetic exponent maps superlinear, steady, then decelerating domain growth under sub-coercive trains","key_machinery":"The local kinetic exponent α_local = d ln(R − R0)/d ln n, extracted from the effective switched-domain radius in a time-dependent Landau–Ginzburg phase-field model; it classifies superlinear, self-similar, and decelerating growth and maps how pulse parameters and starting domain layout move the system between those regimes.","core_discovery":"The effective switched-domain radius R obeys distinct scaling regimes characterized by the local kinetic exponent α_local = d ln(R − R0)/d ln n: initially α_local > 1 (superlinear irreversible domain-wall growth under successive pulses), then α_local ≈ 1 (steady self-similar growth), then α_local < 1 (deceleration from geometric confinement, depletion of switchable polarization, and relaxation-induced back-switching). Transitions are set by competition between pulse-on excitation and pulse-off relaxation and by initial domain geometry.","pith_inferences":[],"forward_implications":[],"fun_headline_variants":["Local kinetic exponent maps three domain-growth regimes in pulsed HZO","Superlinear then steady then decelerating scaling in HZO pulse switching","Pulse-on excitation vs off relaxation sets HZO accumulative regimes","Initial domains and pulse timing tune HZO switched-radius scaling laws","Effective domain radius obeys distinct kinetic exponents under HZO pulses"],"cache_read_input_tokens":16512,"weakest_assumption_plain":"The claim rests on a two-dimensional continuum model with fixed material coefficients, polarization locked normal to the film, no thermal noise, and no defects, grains, or electrode interfaces, assumed sufficient to capture real nanoscale HZO kinetics.","fun_headline_variants_meta":{"raw":{"variants":["Local kinetic exponent maps three domain-growth regimes in pulsed HZO","Superlinear then steady then decelerating scaling in HZO pulse switching","Pulse-on excitation vs off relaxation sets HZO accumulative regimes","Initial domains and pulse timing tune HZO switched-radius scaling laws","Effective domain radius obeys distinct kinetic exponents under HZO pulses"]},"model":"grok-4.5","effort":"low","cost_usd":0.003828,"raw_usage":{"total_tokens":1294,"prompt_tokens":888,"num_sources_used":0,"completion_tokens":73,"cost_in_usd_ticks":38280000,"prompt_tokens_details":{"text_tokens":888,"audio_tokens":0,"image_tokens":0,"cached_tokens":256},"completion_tokens_details":{"audio_tokens":0,"reasoning_tokens":333,"accepted_prediction_tokens":0,"rejected_prediction_tokens":0}},"tokens_in":888,"tokens_out":73,"duration_ms":3519,"temperature":1.0,"reasoning_tokens":333,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-07-13T06:54:02.193889+00:00","model_set":{"reader":"grok-4.5"},"falsifier":"Measure switched-domain radius versus pulse number on real HZO capacitors or FeFETs under matched sub-coercive pulse trains; if the extracted local exponent never exceeds one early on or never drops below one later, or if geometry and pulse-off time do not shift the transitions as predicted, the scaling framework fails.","supporting_citations":[],"review_version":2}