{"id":"39b2cc45-ef47-48b7-8f7c-79c786fadf6c","arxiv_id":"2501.06754","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":1,"one_line_summary":"Adding a CeO2-x oxygen-sponge cap and a symmetric LSMO electrode stack gives Hf0.5Zr0.5O2 capacitors fatigue-free behavior beyond 10^11 cycles and endurance beyond 10^12 cycles.","lead":"A thin cerium oxide layer between the electrode and a hafnia-zirconia ferroelectric film acts as an oxygen sponge, absorbing and releasing the defects that normally destroy memory performance during cycling. A symmetric electrode stack built around this layer keeps the capacitor working past 100 billion switching cycles with almost no polarization loss.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Endurance record lacks a defined cycling waveform and failure criterion; without these, the 10^11 fatigue-free and 10^12 lifetime claims cannot be compared to prior work.","rationale":"The paper contains a coherent mechanism story and a plausible material design; the XPS valence changes, micro-diffraction, and symmetric-architecture comparison all support the qualitative picture. None of those pieces, however, establishes the quantitative record. The claim 'fatigue-free ... <5% loss after 10^11' and 'lifetime surpassing 10^12' is an experimental result whose validity depends entirely on how the endurance test was run and how failure was defined. HZO capacitors are known to exhibit wake-up and fatigue that depend sensitively on field amplitude and pulse conditions; under-switched cycling can appear fatigue-free by construction. The reader's weakest assumption identified the same gap: the cycling protocol is incompletely specified. I see the missing failure criterion and missing quantitative 10^12 state as part of the same concern, not a separate one. If the authors supply the protocol and raw 2Pr-vs-cycles data, and the values hold under a standard threshold, the paper would be a strong ACCEPT. Without that, CONDITIONAL is the correct verdict; I do not see grounds to reject, because the evidence presented is internally consistent and no obvious contradiction exists. The DFT calculation has fixed layers and one U parameter, but it is not used to fit the endurance data, so it does not create circularity.","tokens_in":16555,"tokens_out":5967,"duration_ms":62784,"concrete_test":"Obtain the raw endurance log for the device in Fig. 4A and re-evaluate it with an explicit protocol table: cycling voltage, pulse shape, frequency, and PUND readout voltage for each decade. Re-plot normalized switchable polarization 2Pr using the same failure criterion as the comparison set (e.g., 20% loss from maximum 2Pr). If 2Pr at 10^11 is below 95% of its initial fully switched value, or if the device reaches a defined failure before 10^12, the fatigue-free and record claims are not supported. As an independent check, cycle a fresh Pt/LSMO/CeO2-x/HZO/LSMO capacitor with 3.5 V, 1 MHz square pulses and measure PUND at 3.5 V after 10^8, 10^9, 10^10, 10^11, and 10^12 cycles; report cycles to 20% polarization loss.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central quantitative claim is that Pt/LSMO/CeO2-x/HZO/LSMO loses less than 5% polarization at 10^11 cycles and remains ferroelectric at 10^12 cycles. This claim rests entirely on the endurance-measurement protocol, which is not specified: the main text reports only cycling frequency (Fig. 4A), while the voltage amplitude, pulse width, rise/fall time, and the definition of 'endurance lifetime' are absent from Methods. For a 6 nm HZO film with Ec=2.34 MV/cm, the coercive voltage is about 1.4 V; whether cycling was performed at 2 V, 3 V, or higher changes the conclusion qualitatively, because HZO fatigue is strongly field- and pulse-width-dependent. If the cycling waveform did not fully switch the film (too low voltage, too short pulse, or RC-limited at high frequency), the observed stability is trivial and the 'cycle count' overstates the number of switched states. Additionally, 'stable ferroelectricity after 10^12 cycles' has no quantitative threshold; if it means only that the P-E loop is not closed, it is not comparable with prior endurance reports that use defined failure criteria (e.g., 2Pr dropping below 80% of peak or hard breakdown). The mechanistic XPS/DFT evidence is supportive but cannot validate the record number without a defined metric.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The manuscript reports a Pt/LSMO/CeO2-x/HZO/LSMO planar ferroelectric capacitor that is claimed to be virtually fatigue-free for up to 10^11 switching cycles, with less than 5% polarization loss, and to remain ferroelectric after 10^12 cycles. The authors attribute this behavior to two design elements: a coherent CeO2-x/HZO interface that acts as a reversible oxygen-vacancy sponge, and a symmetric LSMO-based electrode architecture that suppresses imprint-driven directed defect drift. The evidence base includes DFT migration and switching-barrier calculations, STEM/EELS interface characterization, micro-XPS and synchrotron micro-XRD before/after cycling, and electrical measurements (PUND, leakage, retention, temperature stability, and endurance). The central claim is an experimentally measured endurance record, with the DFT and spectroscopy results serving as supporting mechanistic evidence.","tokens_in":16814,"tokens_out":2470,"duration_ms":27122,"significance":"If the endurance and fatigue results are robust, this would be a substantial advance for HfO2-based ferroelectric capacitors, which typically fatigue near 10^6 cycles in planar geometries. The paper is commendable for combining multiple independent probes: PUND-based polarization measurements, endurance cycling, valence-state analysis by XPS and EELS, phase analysis by micro-XRD, and first-principles calculations. The proposed oxygen-sponge mechanism is internally consistent and is supported by the observed Ce3+/Ce4+ interconversion and by the reduced VO migration barriers computed at the CeO2-x/HfO2 interface. The claim of record endurance is, however, only as strong as the cycling protocol used to generate it, and that protocol is not fully specified. The manuscript would be a valuable contribution after the measurement conditions, failure criteria, and statistical uncertainty are documented transparently.","major_comments":[{"comment":"The load-bearing endurance claim—fatigue-free for 10^11 cycles and ferroelectric after 10^12 cycles—is not accompanied by a complete cycling protocol. The main text reports only cycling frequency in Fig. 4A, while the voltage amplitude, pulse width, rise/fall time, and the definition of the 10^12 'lifetime' are absent from the Methods. For a 6 nm HZO film with Ec = 2.34 MV/cm, the coercive voltage is approximately 1.4 V, so whether cycling used 2 V, 3 V, or higher fields qualitatively changes the interpretation. If the waveform did not fully switch the film, the observed stability would be a trivial consequence of reduced switching stress rather than an intrinsic material improvement. Please specify the exact waveform, amplitude, pulse shape, and the quantitative failure criterion (e.g., 2Pr dropping below 80% of its initial value, or hard breakdown), and provide raw endurance traces at 10^11 and 10^12 cycles.","section":"Results, 'Behaviors of more symmetric capacitors'; Fig. 4A"},{"comment":"The claim of 'less than 5% polarization loss after 10^11 cycles' is presented without error bars or device-to-device statistics. The figures appear to show a single representative device, and the main text does not report how many capacitors were measured, what the spread was, or whether the 5% figure refers to the mean, median, or best device. Because record claims of this type are typically compared against many prior studies, the absence of statistical information prevents the reader from assessing whether the result is reproducible or a selected best case. Please provide statistics over at least several devices, and show the raw P-E or PUND traces at the relevant cycle counts.","section":"Fig. 4 and Supplementary figs. S14A-S14B"},{"comment":"The DFT calculations fix the bottom 6 HfO2 layers (half of the HfO2 slab) during structural optimization and NEB calculations, but no test is reported to show that this constraint does not qualitatively change the computed VO migration barriers (0.22-0.96 eV) or the polarization switching barriers in Fig. 1C. Since the key computational support for the oxygen-sponge mechanism is the reduced vacancy migration barrier near the CeO2-x/HfO2 interface, a convergence check with respect to the number of fixed layers, or a fully relaxed calculation for at least one representative path, is necessary to establish that the reported barrier reduction is not an artifact of the constraint.","section":"Methods, DFT calculations"},{"comment":"The Discussion states that the device's comprehensive reliability 'even exceed[s] those of Micron's very recent device with advanced 3D integration' and Fig. 4C claims the 'most stable HfO2-based planar capacitor reported to date.' These comparative statements are not accompanied by a table of test conditions (voltage, waveform, temperature, capacitor area, and failure criterion) for the cited prior work. Without matching protocols, such comparisons are not quantitatively meaningful. Please either provide a systematic comparison under defined conditions or temper the record language to what is directly demonstrated.","section":"Discussion; comparison claims"}],"minor_comments":[{"comment":"The phrase 'oxygen-voltammetry-generated chemical/energy fluctuations' is not standard and is not defined in the text; consider rephrasing to 'field-cycling-induced oxygen redistribution and associated chemical/energy changes'.","section":"Abstract; Introduction"},{"comment":"The caption contains the typo 'HADDF-STEM'; it should read 'HAADF-STEM'.","section":"Fig. 1D caption"},{"comment":"The y-axis label and units for the polarization switching barriers are not visible in the figure description; please ensure the barrier values and units are clearly labeled.","section":"Fig. 1C"},{"comment":"The term 'hexagonal warrior' is informal and unexplained; either define the set of six properties explicitly in the figure caption or remove the term.","section":"Discussion; Fig. 4F"},{"comment":"Reference 61 ('Positive Effect of Parasitic Monoclinic Phase...') appears incomplete, lacking a volume/page range; please verify and complete the citation.","section":"Reference list"}],"recommendation":"major_revision","confidential_remarks":"I believe the work is potentially publishable and the mechanistic story is attractive, but the record endurance claim needs to be placed on explicit experimental footing: specified cycling waveform, amplitude relative to coercive voltage, failure criterion, and device statistics. The DFT constraint issue and the comparative statements should also be addressed. The paper is a good fit for the journal if these concerns are resolved; I would not recommend rejection based on the current evidence."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Pal, here's the take. This paper is worth taking seriously. The central claim—Pt/LSMO/CeO2-x/HZO/LSMO showing <5% polarization loss after 10^11 cycles and still switching after 10^12—is, if true, a real advance for hafnia ferroelectrics. The novelty isn't any single ingredient: oxygen-scavenging layers and symmetric electrode stacks are both known. What's new is the specific combination, and the evidence that the CeO2-x/HZO interface reversibly absorbs vacancies while the symmetric stack suppresses imprint-driven drift. The XPS and EELS showing Ce3+/Ce4+ interconversion under field, and the micro-XRD showing the capped film keeps its polar phase while bare HZO turns monoclinic, are genuinely convincing supporting data. The DFT barriers are plausible, and the Hubbard U is taken from prior work rather than fit to the fatigue data, so there's no circularity problem.\n\nThe soft spot is exactly where the reader put it: the endurance protocol is underspecified. The paper reports cycling frequencies but never states the voltage amplitude, pulse width, rise/fall times, or the failure criterion for 'endurance lifetime.' For a 6 nm film with Ec around 2.3 MV/cm, that's the difference between cycling at 1.5 V (barely above coercive) and 3 V (full switching, harsher). Fatigue in HZO is strongly field- and pulse-width-dependent, so without this information the 10^11 and 10^12 claims cannot be compared to prior work. Also, no error bars or device-to-device statistics are shown, and the 'stable ferroelectricity after 10^12 cycles' lacks a quantitative threshold (is 2Pr above 80%? above 50%?).\n\nThese are reporting gaps, not evidence of a wrong result. The mechanism story is coherent and the experimental evidence is extensive. The comparison to prior endurance data comes from a supplementary note with many references, but selected comparisons need to be checked against a consistent protocol.\n\nMy verdict: the paper deserves a serious referee. The referee should demand the full cycling conditions, raw endurance traces, and a defined failure criterion, plus some statistics across devices. If that comes back clean, this is the most robust HZO planar capacitor I've seen. I'd cite it if I worked in hafnia reliability, and I'd bring it to the reading group to discuss what 'fatigue-free' should mean in this field.","headline":"Strong candidate for the most fatigue-free HZO planar capacitor to date, with a plausible oxygen-sponge mechanism—but the endurance headline needs a fully specified protocol before it can be trusted.","tokens_in":17397,"tokens_out":2629,"would_cite":true,"duration_ms":24468,"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 thin coherent CeO2-x layer that reversibly stores oxygen vacancies, combined with symmetric LSMO electrodes, makes Hf0.5Zr0.5O2 ferroelectric capacitors virtually fatigue-free through $10^{11}$ cycles and still ferroelectric after…","keywords":["ferroelectric Hf0.5Zr0.5O2","fatigue-free cycling","oxygen vacancies","oxygen sponge","CeO2-x buffer","imprint suppression","ferroelectric endurance","hafnia-based memory"],"falsifier":"Run the same $\\mathrm{Pt/LSMO/CeO}_{2-x}/\\mathrm{HZO/LSMO}$ capacitors under a fixed protocol, for example $\\pm 3$ MV/cm, 100 kHz bipolar square pulses with failure defined as 10% polarization loss; observing more than 5% loss before $10^{11}$ cycles, or a monotonically increasing $\\mathrm{Ce}^{4+}$ fraction before $10^{11}$ cycles, would show that the fatigue-free behavior is protocol-dependent rather than intrinsic.","tokens_in":16320,"feed_emoji":"⚡","tokens_out":12384,"duration_ms":105302,"temperature":0.7,"pith_summary":"This paper claims that the fatigue that normally limits hafnium-zirconium oxide (HZO) capacitors to roughly $10^6$ cycles can be removed by engineering both interfaces of a planar stack. The authors insert an ultrathin coherent $\\mathrm{CeO}_{2-x}$ layer that acts as an oxygen sponge, reversibly absorbing and releasing oxygen vacancies, and sandwich the HZO between two symmetric lanthanum strontium manganite (LSMO) electrodes to suppress imprint, the built-in field that drives defects in one direction during cycling. On the $\\mathrm{Pt/LSMO/CeO}_{2-x}/\\mathrm{HZO/LSMO}$ stack they report less than 5% polarization loss after $10^{11}$ bipolar cycles, stable ferroelectricity after $10^{12}$ cycles, and retention beyond ten years at 358 K. If true, this removes a central reliability barrier to commercial hafnia-based ferroelectric memory.","feed_headline":"Oxygen sponge lets hafnia capacitors last 10^12 cycles","feed_subtitle":"A thin ceria buffer and symmetric electrodes keep HZO polarization loss under 5 percent after 10^11 cycles.","key_machinery":"The load-bearing mechanism is the 'oxygen sponge' heterointerface: an ultrathin coherent $\\mathrm{CeO}_{2-x}$ layer on HZO whose multivalent cerium ions reversibly accept and release oxygen vacancies, lowering the vacancy diffusion barrier near the interface and keeping the ferroelectric phase intact by preventing oxygen stoichiometry drift. The second mechanism is the symmetric capacitor architecture $\\mathrm{Pt/LSMO/CeO}_{2-x}/\\mathrm{HZO/LSMO}$, which minimizes imprint, the horizontal shift of the polarization-voltage loop caused by a built-in field, and thereby suppresses net directional migration of defects under bipolar cycling. Together they reduce the coercive field to 2.34 MV/cm, raise the remnant polarization to 21 $\\mu\\mathrm{C}\\,\\mathrm{cm}^{-2}$, and preserve switchable polarization over more than $10^{12}$ cycles.","core_discovery":"The central claim is that a deliberately designed heterointerface, not a new ferroelectric material, is enough to make HZO-based planar capacitors as durable as the best perovskite ferroelectric devices. DFT calculations show that a coherent $\\mathrm{CeO}_{2-x}/\\mathrm{HfO}_{2}$ interface lowers the oxygen-vacancy migration barrier from about 2.3 eV in bulk HfO$_2$ to 0.22--0.96 eV near the interface, so vacancies can move instead of piling up at the electrode. X-ray photoelectron and electron-energy-loss spectra show Ce$^{3+}$/Ce$^{4+}$ interconversion under positive and negative poling, evidence that the ceria layer reversibly stores and releases vacancies and keeps the HZO oxygen content stable. Adding a symmetric top LSMO electrode reduces the built-in field (imprint) to 0.38 MV/cm and suppresses the directional defect drift that otherwise converts the metastable ferroelectric phase into a paraelectric monoclinic phase. The result is a capacitor that the authors report as virtually fatigue-free to $10^{11}$ cycles and still ferroelectric past $10^{12}$ cycles, without any rejuvenation step.","pith_inferences":["If the oxygen-sponge mechanism is general, other reducible oxides such as doped ceria or praseodymia could be tuned for lower diffusion barriers or better work-function matching, and the symmetric-electrode rule could transfer to 3D or transistor geometries.","The paper's imprint-versus-fatigue correlation suggests a fast screening rule: measure the built-in field of a candidate electrode stack and use it as a cheap predictor of endurance before committing to billion-cycle tests.","Because endurance reports across laboratories use different voltage amplitudes, pulse shapes, and failure criteria, a head-to-head measurement of this stack against a reference HZO capacitor under one common protocol would make the record claim fully quantitative.","The ceria layer's Ce3+ fraction still drifts over $10^8$ cycles in the symmetric device, so the sponge has finite capacity; tracking that drift during cycling could predict the eventual failure point and set the required buffer thickness."],"forward_implications":["HZO planar capacitors can reach endurance comparable to perovskite devices, above $10^{10}$ cycles, removing the main reliability gap that has blocked hafnia ferroelectric memory commercialization.","Oxygen-active buffer layers can protect the metastable ferroelectric phase during field cycling, so the design principle extends beyond the specific CeO2-x/HZO pair to other multivalent oxide and fluorite interfaces.","Symmetric electrode stacks reduce imprint without increasing leakage, giving a path to combine low coercive field, high polarization, and long endurance in one device.","The reported retention of more than ten years at 358 K and stable polarization across temperature mean the reliability improvement is not limited to room-temperature cycling.","No rejuvenation or wake-up scheme is needed, so the endurance gain is intrinsic to the device stack rather than a recoverable operating mode."],"supporting_citations":[{"why":"Established ferroelectricity in hafnium oxide thin films, the material system whose reliability this work targets.","marker":"[9]"},{"why":"Documents polarization fatigue in laminated HZO films, the failure mode and baseline the CeO2-x design is meant to overcome.","marker":"[17]"},{"why":"Demonstrated fatigue-free ferroelectric capacitors with platinum electrodes in perovskite oxides, the endurance benchmark planar hafnia devices are compared against.","marker":"[21]"},{"why":"Supplies the physical model of field-cycling degradation in HfO2-based capacitors that motivates the oxygen-sponge buffer.","marker":"[26]"},{"why":"Showed reversible oxygen migration and phase transitions in hafnia devices, the direct precedent for controlling oxygen stoichiometry to preserve the polar phase.","marker":"[28]"},{"why":"Identifies the electrode-ferroelectric interface as the primary constraint on endurance and retention, the interface this paper re-engineers.","marker":"[32]"},{"why":"Supplies the classic fatigue mechanism of defect migration under asymmetric potentials, the basis for the imprint-driven oriented-drift argument.","marker":"[42]"},{"why":"Reported high endurance above 10^12 cycles in HZO via optimized switching ratio, the benchmark this paper compares its fatigue-free result against.","marker":"[45]"}],"fun_headline_variants":["Interface design gives HZO capacitors 10^12 cycle endurance","Oxygen sponge interface stops HZO fatigue, hits 10^12 cycles","Ceria buffer pushes HZO endurance past 10^12 cycles","Symmetric electrodes and oxygen sponge give HZO 10^12-cycle life"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The fatigue-free counts assume the electrical cycling test used to reach $10^{11}$ and $10^{12}$ cycles was at least as demanding as the tests in earlier studies with which the device is compared; a gentler field or a loose failure threshold would make the endurance look better than it is.","fun_headline_variants_meta":{"raw":{"variants":["Interface design gives HZO capacitors 10^12 cycle endurance","Oxygen sponge interface stops HZO fatigue, hits 10^12 cycles","Ceria buffer pushes HZO endurance past 10^12 cycles","Symmetric electrodes and oxygen sponge give HZO 10^12-cycle life"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000866,"raw_usage":{"total_tokens":3800,"prompt_tokens":1035,"completion_tokens":2765,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":651,"completion_tokens_details":{"reasoning_tokens":2696}},"tokens_in":651,"tokens_out":2765,"duration_ms":18290,"temperature":1.0,"reasoning_tokens":2696,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-10T20:49:48.900987+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Run the same $\\mathrm{Pt/LSMO/CeO}_{2-x}/\\mathrm{HZO/LSMO}$ capacitors under a fixed protocol, for example $\\pm 3$ MV/cm, 100 kHz bipolar square pulses with failure defined as 10% polarization loss; observing more than 5% loss before $10^{11}$ cycles, or a monotonically increasing $\\mathrm{Ce}^{4+}$ fraction before $10^{11}$ cycles, would show that the fatigue-free behavior is protocol-dependent rather than intrinsic.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Established ferroelectricity in hafnium oxide thin films, the material system whose reliability this work targets."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Documents polarization fatigue in laminated HZO films, the failure mode and baseline the CeO2-x design is meant to overcome."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Demonstrated fatigue-free ferroelectric capacitors with platinum electrodes in perovskite oxides, the endurance benchmark planar hafnia devices are compared against."},{"cited_title":"Pešić, F","cited_arxiv_id":null,"evidence_quote":"Supplies the physical model of field-cycling degradation in HfO2-based capacitors that motivates the oxygen-sponge buffer."},{"cited_title":"Nukala, M","cited_arxiv_id":null,"evidence_quote":"Showed reversible oxygen migration and phase transitions in hafnia devices, the direct precedent for controlling oxygen stoichiometry to preserve the polar phase."},{"cited_title":"Alcala, M","cited_arxiv_id":null,"evidence_quote":"Identifies the electrode-ferroelectric interface as the primary constraint on endurance and retention, the interface this paper re-engineers."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the classic fatigue mechanism of defect migration under asymmetric potentials, the basis for the imprint-driven oriented-drift argument."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Reported high endurance above 10^12 cycles in HZO via optimized switching ratio, the benchmark this paper compares its fatigue-free result against."}],"review_version":1}