{"id":"3df561d9-a72e-441a-b835-a181efb01030","arxiv_id":"2507.12353","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":6,"one_line_summary":"Nanoscale ferroelectric capacitors switch in the material-limited regime, with LBFO switching in about 150 ps, polycrystalline HZO in about 210 ps, and AlBN in about 20 ns.","lead":"By shrinking ferroelectric capacitors to nanoscale and measuring voltage and current directly, the authors report material-limited switching speeds as fast as 150 picoseconds in LBFO, about 210 picoseconds in HZO, and about 20 nanoseconds in AlBN. Why read: these are among the fastest polarization-switching times reported and point to multi-gigahertz ferroelectric memory and logic.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Material-limited claim rests on a single fixture; no control varying series resistance or pulse rise time is shown, so circuit independence of the ~150/210 ps times is not yet demonstrated.","rationale":"The reader's weakest-assumption analysis focuses on KAI model mis-specification, which is a real vulnerability in the criterion and in the t0 extraction. I agree with that concern, but it is not the most load-bearing risk for the headline claim: even a perfectly KAI-compatible dataset would not prove material-limited switching if the small-capacitor transients were still shaped by the measurement circuit. The paper's strongest evidence is the simultaneous VFE measurement: small capacitors show a nearly square VFE, and the HZO saturation is slower than the measured RCnon-switch and VFE rise time. This is credible support. However, the fastest LBFO point is close to the 100 ps instrument rise time, and no circuit-parameter sweep is presented. A skeptic can explain the 150 ps point as a convolution of a faster material response with the pulse edge, especially given the large 147 ± 49 ps uncertainty. The numerical criterion also contains a factor-of-ten discrepancy in its worked example, so it is not reliable as independent calibration. The proposed series-resistance and rise-time invariance test is a direct, decisive check: if the switching time is invariant under circuit changes at constant VFE, the central claim is solid; if it moves, the regime assignment is wrong. The reader's CONDITIONAL verdict already accommodates this missing validation, so no verdict change is needed.","tokens_in":13767,"tokens_out":12632,"duration_ms":157468,"concrete_test":"Using the same 200 nm LBFO and 400 nm HZO devices, measure the 10-90% switching time with external series resistors of 10, 50, and 200 Ω, adjusting Vin so the amplitude of VFE at the capacitor remains at 2.3 Vc, and repeat with pulse generator rise times of 50, 100, and 200 ps while recording VFE and IFE. If the extracted switching time remains 147 ± 49 ps and 211 ± 11 ps within uncertainty, the material-limited interpretation is confirmed; if it scales with Rs or rise time, the central claim fails.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central assertion is that the measured 10-90% polarization reversal times in the small-capacitor regime are intrinsic material properties, not artifacts of the drive/sense electronics. The paper supports this with two arguments: (i) VFE shows no droop for small capacitors (Fig. 2), and (ii) the 400 nm HZO saturation time (211 ± 11 ps) is about twice the measured RCnon-switch and VFE rise time (Fig. 4d). These are necessary but not sufficient. No experiment varies the circuit parameters -- series resistance, source impedance, or pulse rise time -- while holding the same device and VFE amplitude. For the headline LBFO value, the reported 150 ps (147 ± 49 ps) is only about 1.5x the 100 ps pulse rise time; without deconvolution or a rise-time sweep, a pulse-shape or convolution contribution to the extracted transient is not excluded. The KAI-based criterion (Supplementary Section III) is calibrated on the same data and a post-hoc 0.1 threshold, and its numerical example appears off by a factor of about 10 (main text gives a critical diameter of around 3 µm for Rs = 100 Ω, while the stated formula and parameters give about 0.33 µm). Thus the criterion cannot substitute for a direct circuit-independence check. If the small-capacitor switching time changes when Rs or rise time is changed, the 'material-limited' label is not established.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper reports measurements of the 10-90% polarization reversal time for nanoscale ferroelectric island capacitors in three material families: perovskite La0.15Bi0.85FeO3 (LBFO), polycrystalline Hf0.5Zr0.5O2 (HZO), and wurtzite Al0.92B0.08N (AlBN). Using a semi-coplanar waveguide fixture with simultaneous voltage and current readout, the authors observe a crossover from a circuit-limited regime, in which the switching time scales linearly with capacitor area, to a material-limited regime at sub-micron dimensions in which the switching time is nearly area-independent. They report ~150 ps switching in LBFO, a ~210 ps apparent fundamental limit in HZO, and ~20 ns switching in AlBN. They model the voltage transients with an RC circuit whose ferroelectric branch obeys the Kolmogorov-Avrami-Ishibashi (KAI) kinetics, extract Merz activation fields and transient pseudo-resistivities, and propose a criterion for reaching the material-limited regime based on the voltage drop caused by the peak ferroelectric switching current. The central claim is that the small-capacitor switching times are intrinsic material properties rather than artifacts of the measurement circuit.","tokens_in":14115,"tokens_out":6451,"duration_ms":76625,"significance":"If established, the paper would provide intrinsic 10-90% switching limits for three technologically relevant ferroelectric families, mark LBFO as the fastest electrically switched ferroelectric reported at ~150 ps, and give circuit designers a quantitative rule for entering the material-limited regime. The experimental platform is a genuine strength: simultaneous measurement of V_FE and I_FE, the RC model that reproduces the V_FE plateaus for large capacitors, the power-supply checks in Fig. S5, and the signal-integrity checks in Fig. S7 all go beyond what is typical in this literature. The reported scaling of energy-delay and power-density metrics is also a useful contribution for device benchmarking. However, the central claim that the observed sub-200 ps times are material-limited is not yet fully supported because no experiment varies the circuit parameters, and the criterion used to define the regime contains a numerical inconsistency and depends on KAI parameters fit from the same data. These issues are fixable but need to be addressed before the material-limited interpretation can be accepted.","major_comments":[{"comment":"The numerical example for the material-limited criterion is off by a factor of about ten. From the stated formula and parameters (n=3, Pr=40 microC/cm^2, Vin=1 V, t0=100 ps, ARs <= 8.44e-8 ohm cm^2), a circular capacitor with Rs=100 ohm has diameter ~0.33 micrometers, not \"around 3 micrometers\" as written. This is not a cosmetic error: the same formula predicts that a 1-micrometer-diameter capacitor with Rs=100 ohm violates the 10% voltage-drop condition, yet the experimental crossover in Fig. 1 occurs near 1 micrometer for HZO and LBFO. Please correct the numerical example and reconcile the criterion with the observed crossover, or state explicitly what series resistance corresponds to the experimental crossover.","section":"Main text, criterion paragraph after Fig. 3a"},{"comment":"The material-limited claim for the smallest capacitors is not yet supported by a direct circuit-independence control. The paper compares the 400 nm HZO switching time with the extracted RC_non-switch and the V_FE rise time, but it does not vary the series resistance, source impedance, or pulse rise time while holding the device and V_FE amplitude fixed. For the headline LBFO value of 147 +/- 49 ps, the pulse rise time is 100 ps, so the measured transient is only about 1.5 times the rise time; without deconvolution or a rise-time sweep, a pulse-shape contribution to the extracted switching time cannot be excluded. A control experiment with, for example, two different values of series resistance or pulse rise time, or an explicit deconvolution of the oscilloscope/pulse response, is needed to establish that the small-capacitor times are independent of the circuit.","section":"Fig. 4d; Methods; Ultra-fast Large Signal Ferroelectric Switching Measurements"},{"comment":"The KAI parameters t0 and n are fitted from the same polarization transients that are then used to define the material-limited regime, to interpret the IFE,max versus area slope, and to derive the material-limited criterion. This makes the regime assignment and the criterion partly self-referential. In particular, the 210 ps HZO saturation is interpreted as a material limit under the assumption that KAI kinetics hold down to the smallest devices, but polycrystalline HZO is known in the literature to show nucleation-limited switching behavior that deviates from the simple KAI form. Please add a sensitivity analysis with an alternative kinetics model (for example, nucleation-limited switching) or otherwise quantify how much the extracted t0 and the material-limited criterion change if the KAI assumption is relaxed.","section":"Fig. 2b,c; Supplementary Sections I and III"}],"minor_comments":[{"comment":"The phrase \"reduces 12(4)14 and the resulting current\" appears garbled; it looks like a corrupted citation or an incomplete reference to t0. Please rephrase.","section":"Main text, AlBN paragraph"},{"comment":"The caption states the switching time is plotted along with the RC_non-switch time constant and the V_FE rise time, but the numerical values of these two timescales are not given. Providing them in the caption or text would make the claimed factor-of-two separation easier to verify.","section":"Fig. 4d"},{"comment":"The extrapolated energy-delay values at 40 nm and 10 nm capacitor diameter are outside the measured range and should be explicitly marked as projections in the figure or text.","section":"Fig. 3b,c"},{"comment":"The caption says the figure shows \"the change in polarization (Delta P) as a function of capacitor diameter,\" but the figure appears to show polarization transients for different diameters; please clarify the caption.","section":"Fig. S4 caption"}],"recommendation":"major_revision","confidential_remarks":"The manuscript reports a timely and potentially important experimental result, and the core data are more substantial than many papers in this area. The major comments above are fixable: the numerical criterion error can be corrected, a circuit-variation control or deconvolution can be added, and the KAI sensitivity can be assessed. I do not see grounds for rejection. I would encourage the editor to request a revision that addresses these points."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Hi,\n\nRead the Chiang et al. submission on material-limited switching in nanoscale ferroelectrics. Bottom line: this is a serious experimental paper that reports genuinely fast switching (LBFO ~150 ps, HZO saturation ~210 ps) and an explicit criterion for reaching the material-limited regime. The simultaneous measurement of VFE and IFE on a semi-coplanar waveguide is a real advance, and the authors do a credible job of showing that the RC model reproduces the observed voltage droop in large capacitors. The extension to fluorite and wurtzite materials is new relative to Parsonnet et al.\n\nThe main soft spot is that the central claim—that the sub-200 ps times are intrinsic—rests on two arguments, neither of which is a direct circuit-independence check. They show VFE has no droop for small caps, and for the 400 nm HZO they show t_switch is about 2x the RC time constant and the VFE rise time. That's reasonable but not conclusive. For LBFO, 147 ± 49 ps is only about 1.5x the 100 ps pulse rise time, and no deconvolution or rise-time sweep is reported. The paper would be much stronger if the authors varied series resistance or pulse rise time on the same device and showed the extracted switching time doesn't move.\n\nThere's also an internal inconsistency in the criterion: the stated formula with Rs = 100 ohm gives a critical diameter around 0.33 µm, not ~3 µm as the text claims. That's a factor of ten and a referee will catch it. More broadly, the criterion is built on KAI parameters t0 and n fitted from the same transients and a post-hoc 10% threshold, so it's a useful design rule but not an independent prediction.\n\nThe data are not public, only available on request. That's common but frustrating for a claim of this importance; release of raw transients would help.\n\nIf I'm being fair: the scaling transition is supported by multiple independent observables (voltage droop, current saturation, activation field flattening), and the HZO saturation at 211 ± 11 ps is the most convincing piece. The paper deserves serious peer review, but the material-limited label should be provisional until the circuit sweep is done. I'd send it to a good journal, with a referee who checks the criterion arithmetic and asks for the rise-time control.\n\nBest,\n[You]","headline":"Impressive experimental paper with a plausible material-limited regime, but the central claim needs a direct circuit-parameter sweep to fully land.","tokens_in":14689,"tokens_out":2514,"would_cite":true,"duration_ms":27849,"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":"Nanoscale ferroelectric capacitors switch in 150 picoseconds.","keywords":["ferroelectric switching","material-limited regime","nanoscale capacitors","hafnium zirconium oxide","bismuth ferrite","aluminum boron nitride","polarization reversal","ultrafast measurement"],"falsifier":"Measure the 10-90% switching time of a fixed 400 nm HZO capacitor at high overvoltage while changing the series resistance from 50 Ω to 500 Ω; if the ~211 ps floor is genuinely material-limited, the switching time should be unchanged, whereas a circuit-limited reading predicts a proportional increase.","tokens_in":13602,"feed_emoji":"⚡","tokens_out":10224,"duration_ms":94464,"temperature":0.7,"pith_summary":"The paper's central claim is that ferroelectric capacitors small enough to outrun their measurement circuit switch at speeds set by the material itself, not by the wiring. By shrinking island capacitors from micron to sub-micron dimensions, the authors observe a clean crossover from circuit-limited to material-limited polarization reversal. In this regime, La-doped BiFeO3 switches in about 150 ps (the fastest reported for any ferroelectric), polycrystalline Hf0.5Zr0.5O2 reaches a floor of 211 ± 11 ps, and Al0.92B0.08N switches near 20 ns, capped by coercive and breakdown fields. The significance is that these numbers are intrinsic material properties, and they come with a quantitative criterion for when the measurement circuit can be ignored.","feed_headline":"Nanoscale ferroelectric capacitors switch in 150 picoseconds","feed_subtitle":"Shrinking capacitors below a micron removes circuit delays and exposes each material's intrinsic speed limit.","key_machinery":"The measurement structure is a ferroelectric-capacitor semi-coplanar waveguide (MFM-sCPW), an embedded vertical ferroelectric island capacitor that lets the authors record the voltage across the ferroelectric and the switching current simultaneously with picosecond resolution. The kinetic analysis rests on the Kolmogorov-Avrami-Ishibashi (KAI) model, $\\Delta P(t) = 2P_r[1-\\exp(-(t/t_0)^n)]$, with the characteristic time $t_0$ and exponent $n$ fitted to the measured polarization transients. The maximum switching current derived from that model defines the material-limited regime through the criterion $I_{\\mathrm{FE,max}} R_s \\le 0.1 V_{\\mathrm{in}}$, meaning the voltage dropped across the series resistance during switching stays below 10% of the drive voltage. Merz's law, $t_{\\mathrm{switch}} = t_0 \\exp(E_a/E)$, is used to extract the activation field, and the transient pseudo-resistivity $V_{\\mathrm{FE}}/I_{\\mathrm{FE}}$ is tracked during reversal.","core_discovery":"On its own terms, the paper establishes that the 10-90% polarization reversal time of a nanoscale ferroelectric capacitor can be made independent of the measurement circuit. The evidence is a size-dependent transition: for large capacitors the voltage across the ferroelectric droops while switching current is drawn, so the switching time scales linearly with area; below roughly one micron diameter the voltage step stays square, the peak switching current scales linearly with area as the KAI model predicts, and the extracted activation field plateaus. In that material-limited regime, La0.15Bi0.85FeO3 switches in about 150 ps, polycrystalline Hf0.5Zr0.5O2 saturates at 211 ± 11 ps, and Al0.92B0.08N requires about 20 ns because its higher coercive field cannot be overdriven without breakdown. The paper also reports that the instantaneous pseudo-resistivity drops below 10 Ω cm during switching and that energy-delay products scale favorably with shrinking area.","pith_inferences":["If the ~211 ps HZO floor is set by grain-boundary pinning rather than by intrinsic domain-wall velocity, then single-crystalline or texture-engineered HZO should switch faster; the paper does not test this directly.","The criterion implies that at fixed series resistance there is a maximum capacitor diameter for material-limited operation, so shrinking below that diameter should leave the switching time unchanged; a direct resistance-variation experiment could confirm this.","The extrapolation to sub-femtojoule dissipation at 40 nm diameter assumes the material-limited scaling continues unchanged; at such sizes the capacitor may enter a single-domain or depletion-limited regime that could break the trend.","AlBN's ~20 ns switching is set by the achievable overvoltage before breakdown, so its true material limit may be far faster; pulsed breakdown studies at higher fields could separate coercive-field limits from intrinsic kinetics."],"forward_implications":["A 400 nm polycrystalline HZO capacitor switches in 211 ± 11 ps independent of further voltage increase, implying a fundamental many-grain limit that would support operation near 5 GHz.","The criterion $I_{\\mathrm{FE,max}} R_s \\le 0.1 V_{\\mathrm{in}}$ gives device designers a quantitative bound on the area-series-resistance product, roughly $A R_s \\le 8.44 \\times 10^{-8}\\,\\Omega\\,\\mathrm{cm}^2$ for typical parameters.","In the material-limited regime the activation field extracted from Merz's law is constant with capacitor size, so intrinsic material parameters can be compared across devices without circuit de-embedding.","Energy-delay products scale more favorably with area in the material-limited regime, with femtojoule operation at 200 nm diameter and extrapolated sub-femtojoule dissipation at 40 nm diameter for LBFO.","The transient pseudo-resistivity of LBFO and HZO drops by eight orders of magnitude during switching, suggesting picosecond-scale nonlinear circuit elements."],"supporting_citations":[{"why":"Supplies the literature switching-time-versus-area data and the prior approach to intrinsic switching in BiFeO3.","marker":"[11]"},{"why":"Provides the previous fastest reported 10-90% switching time of 220 ps, which the LBFO result surpasses.","marker":"[19]"},{"why":"Provides the prior record hafnia switching time of 605 ps, which the HZO result surpasses.","marker":"[20]"},{"why":"Supplies the Avrami nucleation-and-growth kinetics underlying the KAI model used to fit polarization transients.","marker":"[21]"},{"why":"Extends the KAI model to ferroelectric domain switching and underpins the switching-current expression.","marker":"[22]"},{"why":"Formulates the polarization reversal kinetics used to derive peak switching current and the material-limited criterion.","marker":"[23]"},{"why":"Provides Merz's law, the exponential activation-field relation used to extract Ea and distinguish the regimes.","marker":"[27]"},{"why":"Ultrafast switching measurements on HZO used as the comparison point for speed and methodology.","marker":"[8]"},{"why":"Prior finite-element analysis of the same pillar structure, used to rule out edge field amplification in the scaling data.","marker":"[29]"},{"why":"Supplies the energy-delay benchmarking metric used to evaluate the technology implications.","marker":"[26]"}],"fun_headline_variants":["Ferroelectric switching hits material limit at 150 ps","Nanoscale capacitors expose intrinsic ferroelectric speed","Fastest ferroelectric switch: 150 ps in scaled capacitors","Size shrinks reveal true ferroelectric switching limit","Circuit-free switching: ferroelectrics reach 150 ps"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The interpretation rests on the Kolmogorov-Avrami-Ishibashi model with a single characteristic time and growth exponent fitted from the same transients used to extract switching parameters; if the true kinetics are nucleation-dominated or vary with capacitor area, the extracted material-limited times and the crossover criterion would not be reliable.","fun_headline_variants_meta":{"raw":{"variants":["Ferroelectric switching hits material limit at 150 ps","Nanoscale capacitors expose intrinsic ferroelectric speed","Fastest ferroelectric switch: 150 ps in scaled capacitors","Size shrinks reveal true ferroelectric switching limit","Circuit-free switching: ferroelectrics reach 150 ps"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000214,"raw_usage":{"total_tokens":1443,"prompt_tokens":980,"completion_tokens":463,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":596,"completion_tokens_details":{"reasoning_tokens":383}},"tokens_in":596,"tokens_out":463,"duration_ms":5306,"temperature":1.0,"reasoning_tokens":383,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-06T16:48:09.898892+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Measure the 10-90% switching time of a fixed 400 nm HZO capacitor at high overvoltage while changing the series resistance from 50 Ω to 500 Ω; if the ~211 ps floor is genuinely material-limited, the switching time should be unchanged, whereas a circuit-limited reading predicts a proportional increase.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provides Merz's law, the exponential activation-field relation used to extract Ea and distinguish the regimes."}],"review_version":1}