{"id":"c7a749e5-35ad-4cff-b9dd-67d03984c0f0","arxiv_id":"2411.16652","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"Ni/Al0.68Sc0.32N/4H-SiC ferroelectric capacitors switch up to 1000 °C and show stable retention and endurance at 600-800 °C, a promising step for high-temperature non-volatile memory.","lead":"Capacitors made of nickel, aluminum scandium nitride, and silicon carbide switched their electric polarization at temperatures up to 1000 °C and held data for many hours at 600 °C. The result targets a long-unmet need for non-volatile memory that survives extreme heat in aerospace, automotive, and energy exploration systems.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The 1000 °C switching claim rests almost entirely on negative-direction J-E and ND PUND data; positive-direction switching is not demonstrated above 800 °C, leaving thermally activated charge injection and interface trapping unexcluded as sources of the observed peaks.","rationale":"The reader's weakest-assumption analysis identifies the same load-bearing risk: the 1000 °C claim is supported only by negative-direction J-E peaks and ND PUND data, while positive-direction data are absent or compliance-limited above 800 °C. The manuscript itself acknowledges in Figure S6 that positive EC is not marked at 900 °C and 1000 °C, and in Figure 2(d) that PU-sequence 2Pr is unreliable above 600 °C. Given this, the central claim of 'robust ferroelectric switching up to 1000 °C' should not be accepted at face value without further evidence ruling out non-ferroelectric charge dynamics. However, the claim is not internally contradicted: the negative-direction data are consistent with ferroelectric switching, prior work has shown AlScN thermal stability after 1000 °C annealing, and room-temperature PUND saturation is demonstrated across ten devices. The right outcome is a conditional acceptance requiring the proposed pulse-width scaling test and an explicit attempt to recover positive-direction switching at 900–1000 °C. Therefore the reader's CONDITIONAL verdict remains appropriate.","tokens_in":10429,"tokens_out":3540,"duration_ms":37321,"concrete_test":"At 1000 °C, measure the ND PUND switched polarization as a function of pulse width (e.g., 0.5, 2, and 10 µs) at fixed amplitude, and compare with a nominally identical control capacitor in which the AlScN layer is not switched or with an equivalent Al2O3/SiC stack. Ferroelectric switching should give a pulse-width-independent switched 2Pr with a sharp voltage threshold and a small D-pulse response; a leakage or trap mechanism would show an integrated charge that grows with pulse width and no clear saturation. In parallel, repeat the positive-direction measurement at 950 °C and 1000 °C with a higher current compliance source or smaller-area electrodes to determine whether a positive coercive peak exists at all, and explicitly report the result even if compliance still limits it.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim that the Al0.68Sc0.32N film exhibits 'stable and robust ferroelectric switching up to 1000 °C' requires that the current-density peaks and PUND responses at 900–1000 °C be ferroelectric polarization reversal rather than leakage, trap emission, or dielectric relaxation. At these temperatures the positive direction is compliance-limited: Figure 2(a) shows the 10 mA current compliance being reached, and Figure S6 explicitly states that the positive coercive field is not marked at 900 °C and 1000 °C. The only hysteresis evidence for a coercive field at 1000 °C is therefore the negative-direction value of −2.5 MV/cm, extracted by differentiating a J-E curve in which leakage is large and the applied voltage has been reduced to avoid breakdown. Likewise, Figure 2(d) and Figure S8 show that PU-sequence polarization is unreliable above 600 °C because of uncompensated leakage; the reported 1000 °C remanent polarization of −90.6 μC/cm² comes only from ND sequences. A one-direction current peak can arise from non-ferroelectric mechanisms, such as field-enhanced carrier injection at the AlScN/SiC interface or detrapping, which would produce a PUND asymmetry indistinguishable from single-polarity switching in this device geometry. Because no positive-direction switching is demonstrated above 800 °C, the claim that the film is fully switchable at 1000 °C is not yet established.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper reports Ni/Al0.68Sc0.32N/4H-SiC metal-ferroelectric-semiconductor capacitors and characterizes their switching, retention, and endurance from room temperature to 1000°C. The authors report a linear decrease of coercive field with temperature, stable remanent polarization from negative-down PUND sequences at high temperatures, and reliable retention and endurance at 600°C and 800°C. The central claim is that the 30-nm Al0.68Sc0.32N film exhibits stable and robust ferroelectric switching up to 1000°C, making these devices candidates for high-temperature non-volatile memory integrated with SiC.","tokens_in":10727,"tokens_out":6089,"duration_ms":52753,"significance":"If the 1000°C switching claim is fully substantiated, this would be an important advance: previous work on ferroelectric memory has rarely demonstrated bipolar switching above 600°C, and integration with SiC is technologically relevant. The paper's strengths include direct J-E and PUND measurements, benchmarking against prior AlScN results, and explicit reporting of measurement limitations (e.g., compliance, leakage subtraction artifacts). The reliability data at 600°C and 800°C are useful and appear internally consistent. However, the high-temperature evidence is asymmetric and not yet sufficient to support the abstract's unqualified claim.","major_comments":[{"comment":"At 900 °C and 1000 °C the positive-voltage current is compliance-limited (Fig. 2a) and the positive coercive field is not extracted (Fig. S6). Only the negative-direction J-E peak and the ND PUND data (Fig. 2c-d) are reported. Because a single-polarity current peak can originate from field-enhanced carrier injection or interface trapping in this MFeS stack, the data do not establish bipolar ferroelectric switching at these temperatures. The abstract's claim of 'stable and robust ferroelectric switching up to 1000 °C' should be tempered to 'negative-polarity switching evidence' unless positive-direction switching is demonstrated, e.g., with smaller-area devices, higher current compliance, or improved electrode design.","section":"High-temperature performance, Fig. 2(a) and Fig. S6"},{"comment":"The authors note that the compensated negative current response between 0 and 8 MV/cm at high temperatures is a mathematical artifact. This raises concern that the coercive-field extraction from derivatives at 900/1000 °C could be affected if the derivative is applied to compensated curves. The manuscript should state explicitly whether the EC values in Fig. 2b are derived from raw or compensated J-E data, and show raw dJ/dE traces at 900 and 1000 °C. In addition, a control experiment on a non-ferroelectric AlN/SiC or unipolar device under identical conditions would help exclude leakage or detrapping as the source of the negative peaks.","section":"Leakage-compensated J-E analysis, Fig. S2(d)"},{"comment":"The same-state retention measurement in the M-Polar direction produces negative depolarization at 700-800 °C, which the authors themselves call inconclusive. Yet this is the only retention evidence at 800 °C; the opposite-state retention test (Fig. 3c) was performed only at 600 °C. To support the statement that 'at 800 °C the devices retain data for at least 10,000 s', an opposite-state retention measurement at 800 °C (or a clear quantitative limit on polarization loss) is needed.","section":"Retention claims at 800 °C, Fig. 3(b)"},{"comment":"The endurance at 900/1000 °C is assessed from PUND 2Pr values, but the PU-sequence 2Pr is acknowledged to be unreliable above 600 °C due to uncompensated leakage (Fig. 2d). The manuscript should specify whether the endurance curves at 900/1000 °C are based on ND-only polarization, and if so, the claim of 'cycling endurance' should be restricted to single-polarity cycling until symmetric switching is demonstrated.","section":"Endurance at high temperature, Fig. 4(a)"}],"minor_comments":[{"comment":"The title in the submitted text reads 'operating up to ∘C' with the temperature value missing; it should read 'operating up to 1000 °C'.","section":"Title"},{"comment":"Many subscripts and superscripts are corrupted (e.g., 'μC cmିଶ', 'MV cm-1', '10ହ', 'P୰', 'Eେ'). Please ensure proper typesetting so that units such as μC cm⁻² and MV cm⁻¹ are unambiguous.","section":"Throughout"},{"comment":"The Figure 1(c) caption mentions a rise/fall time of 100 ns and pulse width of 2 μs, while the Experimental Section and Figure S9 state 200 ns rise/fall time. Please reconcile these values.","section":"Fig. 1(c) vs. Experimental Section"},{"comment":"The abstract claims 'stable and robust ferroelectric switching up to 1000 °C', but the conclusion more cautiously states 'partial EC measurements, full PUND measurement as well as cycling endurance measurements extending up to 1000 °C'. Please align the abstract with the evidential limitations.","section":"Abstract vs. Conclusion"},{"comment":"Reference 6 is a commercial website; consider citing a peer-reviewed source for commercial NVM temperature limits.","section":"References"},{"comment":"Polarization loss in the text is given as 'μC cm-1' but should be 'μC cm⁻²'; please ensure units are consistent throughout, including the axis labels.","section":"Fig. 3(a) and 3(c)"}],"recommendation":"major_revision","confidential_remarks":"The manuscript reports potentially significant high-temperature AlScN/SiC memory behavior, but the headline claim of robust ferroelectric switching up to 1000 °C rests on single-polarity evidence. This is fixable within the manuscript's scope either by providing additional bipolar high-temperature data or by appropriately qualifying the claim. The paper's reliability data at 600-800 °C are valuable and appear sound. I recommend major revision to address the asymmetry in the high-temperature evidence."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Two things to know about this paper. First, it's a genuine step beyond the group's earlier AlScN/SiC work: they show retention and endurance at 600°C (100 hours, ~2000 cycles) and 800°C (10,000 s, ~400 cycles), which is the kind of reliability data the field needs. Second, the abstract oversells the 1000°C result. The data at 900 and 1000°C consist of a negative-direction switching peak in the J-E loop and the ND branch of PUND. The positive direction is compliance-limited, the positive coercive field is not marked, and the PU polarization is dominated by leakage. So 'stable and robust ferroelectric switching up to 1000°C' is not established—what's established is that the negative-polarization state still switches at 1000°C. The stress-test note is right.\n\nWhat I like: the measurements are direct, the authors acknowledge the asymmetry in the SI (Figure S6 explicitly says positive EC is not marked at 900/1000°C; Figure S8 shows the PU leakage), and the 600°C reliability data come from multiple devices. The linear decrease of the negative EC with temperature is a clean trend worth reporting.\n\nSoft spots beyond the overclaim. The leakage-compensated subtraction in Figure S2(d) creates an artifact in the 0–8 MV/cm range, which makes me wary of the extracted peaks at high temperature. There are no error bars on EC above RT or on the 1000°C Pr. The breakdown field data show large device-to-device variation (Figure S7), and the authors mention the large pad size but don't mitigate it. None of this is fatal, but it means the 1000°C claim should be phrased as 'evidence of negative-direction switching' rather than full 'robust switching.'\n\nMy take: the paper is worth refereeing seriously, and I'd accept it after a major revision that aligns the abstract with the actual evidence. The 600°C/800°C reliability data are the real contribution; the 1000°C part becomes a preliminary observation. I'd cite this for the high-temperature endurance/retention numbers.","headline":"Real high-temperature reliability data, but the 1000°C switching claim is one-directional and the abstract overstates it.","tokens_in":11337,"tokens_out":2866,"would_cite":true,"duration_ms":27587,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":["77.80.-e","85.50.-n"],"model":"deepseek-v4-flash","headline":"Ferroelectric AlScN capacitors keep switching at 1000 °C.","keywords":["ferroelectrics","non-volatile memory","high-temperature electronics","aluminum scandium nitride","silicon carbide","metal-ferroelectric-semiconductor capacitors","PUND measurement","retention and endurance"],"falsifier":"Grow an identical 30-nm film with the ferroelectric phase suppressed (or a non-ferroelectric control such as AlN on the same SiC), run the same J-E and PUND protocol at 900–1000 °C, and check whether the negative-direction switching peaks and saturating ND-sequence polarization persist; if they do, the central claim needs re-assignment to a non-ferroelectric mechanism.","tokens_in":10258,"feed_emoji":"🔥","tokens_out":4306,"duration_ms":38765,"temperature":0.7,"pith_summary":"This paper reports that a 30-nm film of aluminum scandium nitride (Al0.68Sc0.32N) grown directly on silicon carbide remains ferroelectric, meaning it can reliably switch its polarization, up to 1000 °C, far beyond the roughly 300 °C limit of commercial non-volatile memory. The devices are Ni/Al0.68Sc0.32N/4H-SiC metal-ferroelectric-semiconductor capacitors. The authors measure the coercive field dropping linearly with temperature and remanent polarization staying near $-90.6\\ \\mu$C cm$^{-2}$ at 1000 °C in the negative-voltage direction. They also demonstrate retention of at least $10^4$ seconds and 467 write cycles at 800 °C. If correct, this establishes AlScN-on-SiC as a candidate memory technology for high-temperature SiC integrated circuits.","feed_headline":"AlScN memory cells keep switching at 1000 °C","feed_subtitle":"30-nm Al0.68Sc0.32N on SiC holds polarization for hours at 600–800 °C, a fit for SiC logic memory.","key_machinery":"The load-bearing structure is the metal-ferroelectric-semiconductor stack Ni/Al0.68Sc0.32N/4H-SiC, with a 30-nm Al0.68Sc0.32N film co-sputtered directly on heavily n-doped 4H-SiC. Two measurement techniques carry the argument: bipolar triangular J-E sweeps at 10 kHz with derivative-based extraction of the coercive field, and PUND ultrafast pulses used to separate switching polarization from leakage. A leakage-compensation scheme subtracts the second positive and negative triangular pulse responses from the first to expose the ferroelectric switching peaks.","core_discovery":"The paper's central claim is that a 30-nm-thick ferroelectric Al0.68Sc0.32N film sputtered directly on 4H-SiC shows stable ferroelectric switching from room temperature to 1000 °C. The evidence is current-density peaks in bipolar J-E hysteresis loops and PUND pulse responses: the coercive field decreases linearly from $-6.4/+11.9$ MV cm$^{-1}$ at 25 °C to $-3.1/+7.8$ MV cm$^{-1}$ at 800 °C, with the negative-direction value reaching $-2.5$ MV cm$^{-1}$ at 1000 °C. Remanent polarization from ND pulse sequences saturates near $-90.6\\ \\mu$C cm$^{-2}$ at 1000 °C, and at 600 °C the film loses only 3.4% of its N-polar polarization after 100 hours. The authors present this as a foundation for AlScN/SiC ferroelectric memory embedded in high-temperature SiC logic.","pith_inferences":["Because the positive-direction data at 900–1000 °C hit the measurement tool's current compliance, the 1000 °C claim currently rests on the negative-direction ND branch; a dedicated high-compliance measurement of positive switching would cement or revise the symmetric claim.","The linear coercive-field-versus-temperature trend suggests a single activation-energy picture for switching, which could be tested by checking whether switching time at fixed field follows one Arrhenius law across the whole temperature range.","If this stack is scaled into a ferroelectric field-effect transistor, the same 30-nm film's leakage and endurance will need to be re-evaluated under realistic gate-stack processing rather than on large capacitor pads."],"forward_implications":["AlScN/SiC metal-ferroelectric-semiconductor capacitors can serve as the memory element in SiC integrated circuits operating above 600 °C, a regime where no commercial non-volatile memory is available.","The linear decrease of coercive field with temperature means the write voltage can be scaled downward as operating temperature rises, easing power budgets in high-temperature systems.","At 600 °C the measured endurance of roughly 2000 cycles and 100-hour retention support read-intensive applications such as firmware storage and sensor data logging.","At 800 °C the demonstrated retention of at least 10,000 seconds and 467 write cycles exceed the requirements of low-write, long-retention tasks in extreme thermal environments."],"supporting_citations":[{"why":"Establishes that AlScN retains ferroelectricity after annealing at 1000 °C, the thermal-stability anchor for the present claim.","marker":"[15]"},{"why":"Prior demonstration of AlScN-based non-volatile memory operating at 600 °C, the reference point for high-temperature endurance and retention.","marker":"[16]"},{"why":"Earlier Al/AlScN/SiC ferroelectric capacitor work that this Ni/AlScN/SiC design builds on and improves.","marker":"[27]"},{"why":"SiC JFET integrated-circuit technology that motivates and sets the integration context for the high-temperature memory.","marker":"[22]"},{"why":"Supplies the sputtered AlScN ferroelectric behavior and thickness-scaling baselines used to interpret the 30-nm film.","marker":"[11]"},{"why":"Documents the >80% remanent-polarization loss in HfZrO2 at 400 °C, the degradation benchmark this work is contrasted against.","marker":"[7]"}],"fun_headline_variants":["AlScN memory cells switch at 1000 °C","Ferroelectric AlScN on SiC works to 1000 °C","AlScN capacitors endure 1000 °C for memory","High-temp memory: AlScN/SiC stable at 1000 °C","AlScN ferroelectric switches up to 1000 °C"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The high-temperature evidence rests on treating the current peaks seen at 900–1000 °C as ferroelectric switching rather than as thermally activated leakage or interface trapping.","fun_headline_variants_meta":{"raw":{"variants":["AlScN memory cells switch at 1000 °C","Ferroelectric AlScN on SiC works to 1000 °C","AlScN capacitors endure 1000 °C for memory","High-temp memory: AlScN/SiC stable at 1000 °C","AlScN ferroelectric switches up to 1000 °C"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000269,"raw_usage":{"total_tokens":1662,"prompt_tokens":1025,"completion_tokens":637,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":641,"completion_tokens_details":{"reasoning_tokens":539}},"tokens_in":641,"tokens_out":637,"duration_ms":5665,"temperature":1.0,"reasoning_tokens":539,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-12T12:51:47.580682+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Grow an identical 30-nm film with the ferroelectric phase suppressed (or a non-ferroelectric control such as AlN on the same SiC), run the same J-E and PUND protocol at 900–1000 °C, and check whether the negative-direction switching peaks and saturating ND-sequence polarization persist; if they do, the central claim needs re-assignment to a non-ferroelectric mechanism.","supporting_citations":[{"cited_title":"On the exceptional temperature stability of ferroelectric Al1- xScxN thin films","cited_arxiv_id":null,"evidence_quote":"Establishes that AlScN retains ferroelectricity after annealing at 1000 °C, the thermal-stability anchor for the present claim."},{"cited_title":"Metal-ferroelectric AlScN-semiconductor memory devices on SiC wafers","cited_arxiv_id":null,"evidence_quote":"Earlier Al/AlScN/SiC ferroelectric capacitor work that this Ni/AlScN/SiC design builds on and improves."},{"cited_title":"Extreme temperature 6H‐SiC JFET integrated circuit technology","cited_arxiv_id":null,"evidence_quote":"SiC JFET integrated-circuit technology that motivates and sets the integration context for the high-temperature memory."},{"cited_title":"Insights Into Curie‐Temperature and Phase Formation of Ferroelectric Hf1− xZrxO2 with Oxygen Defects from a Leveled Energy Landscape","cited_arxiv_id":null,"evidence_quote":"Documents the >80% remanent-polarization loss in HfZrO2 at 400 °C, the degradation benchmark this work is contrasted against."}],"review_version":1}