{"id":"309bbfbe-f5e0-443a-94a9-1e4ae46c1706","arxiv_id":"2508.16768","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":3,"one_line_summary":"Varying the vertical arrangement of Zr and Al dopants in hafnium oxide controls phase formation and improves FeFET gate stack endurance and remanent polarization.","lead":"It shows that placing zirconium and aluminum dopants in different vertical positions inside hafnium oxide films changes the crystal phases formed and improves the endurance and polarization of ferroelectric transistor gate stacks. Why read: This offers a practical recipe for making more reliable ferroelectric memories without changing materials.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"GIXRD-based phase comparison is not corrected for preferred orientation or thickness, so the claim that dopant arrangement primarily controls phase composition could reflect texture artifacts rather than true m-phase fraction changes.","rationale":"The paper is well-motivated, and the ALD-based spatial control of dopants is a sensible and interesting experimental design. The electrical observation that SiON/HZO interfaces degrade endurance is plausible and, if reproducible, practically useful. However, the novel mechanistic claim that phase composition is primarily controlled by the spatial arrangement of dopants is currently supported only by normalized GIXRD peak intensities from single samples. Preferred orientation is a well-known confound in grazing-incidence diffraction, and the paper itself reports texture-related changes in o{200}, so the absence of a texture correction is a genuine risk. A multi-angle or texture-corrected analysis would settle whether the m-phase fraction truly increases when HZO is placed at the center. The lack of replicate samples is a compounding issue but does not by itself invalidate the large qualitative differences; it mainly affects the electrical statistics. The reader's CONDITIONAL verdict already captures the need for additional validation, so my analysis does not move that verdict. I agree with the reader that the normalization to I_o(111) is the weakest point in the structural argument, and the proposed test directly targets that assumption.","tokens_in":8425,"tokens_out":9663,"duration_ms":110163,"concrete_test":"Re-measure the five samples by GIXRD at three incidence angles (e.g., 0.5°, 1.0°, 2.0°) and, if possible, at two azimuthal orientations, then compare the m(\\bar{1}11)/o(111) intensity ratio. If the ratio changes by more than the observed between-configuration differences, the normalization to o(111) is texture/absorption biased. Alternatively, perform Rietveld refinement with a March-Dollase preferred-orientation correction on the full 20°-40° patterns; if the refined m-phase weight fraction does not rank the HZO-center stack highest, the central structural claim fails.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central structural claim rests on relative intensities of m(\\bar{1}11) and o(111) in grazing-incidence XRD. The paper states that intensities were normalized to I_o(111) and that no texture correction was applied. In a 10 nm polycrystalline film measured by GIXRD, the intensity ratio depends on grain orientation distribution, incidence angle, and absorption. The paper itself observes that o{200} intensity is strongly influenced by HZO position and associates this with texture/templating, so texture is demonstrably non-uniform across the five stacks. If the o(111) texture coefficient differs, a sample with less favorably oriented o-phase grains will show a lower normalized o(111) intensity, making the m/o ratio larger even at constant m-phase fraction. Thus the pronounced m(\\bar{1}11) peak in the HZO-center stack does not by itself establish that more monoclinic phase is present. Because the m-phase is then invoked to explain the endurance improvement, a texture artifact would propagate into the electrical interpretation. Additionally, only one sample per configuration is reported, so run-to-run process variation is not separated from the intended dopant-arrangement effect; the electrical comparisons likewise lack error bars, making the significance of configuration differences difficult to judge.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"This manuscript reports an experimental study of ferroelectric HfO2-based MFIS gate stacks in which Zr and Al dopants are placed in different vertical arrangements by ALD. Three co-doped configurations (HZO at the bottom, center, and top of the film) are compared with mono-doped HZO and HAO controls. The authors use GIXRD to characterize phase composition, ToF-SIMS to verify dopant distributions, and PUND and fatigue measurements to evaluate remanent polarization, leakage, and endurance. They conclude that the spatial arrangement of dopants primarily determines the phase composition of the annealed HfO2 films, that positioning HZO in the center of the film improves endurance at moderate polarization, and that the SiON/HZO interface should be avoided because Zr diffusion into the SiON layer degrades the stack.","tokens_in":8670,"tokens_out":3168,"duration_ms":34015,"significance":"If the central claims are confirmed, the work offers a practical ALD-based design rule for tuning phase fractions and endurance in FeFET gate stacks and provides a useful comparison of five stack configurations on the same platform. The strengths of the paper include direct ToF-SIMS confirmation of the intended dopant gradients, PUND-based separation of leakage and switching contributions, and fatigue data that qualitatively show large differences between stacks. However, the main structural conclusion depends on an uncorrected GIXRD peak-intensity normalization, and the electrical comparisons lack replicate statistics. These issues are load-bearing because the proposed mechanism (m-phase stabilization and Zr diffusion) is tied to the structural and endurance interpretations. With additional analysis of texture and statistics, the significance for FeFET gate-stack engineering could be solid.","major_comments":[{"comment":"The phase-fraction comparison normalizes the m(\\bar{1}11) and o{200} peak intensities to I_o(111) without applying a texture or thickness correction. The paper itself attributes the variation of o{200} intensity to crystallographic texture or templating effects. Therefore the pronounced m(\\bar{1}11) peak in the HZO-center stack could reflect a change in the orientation distribution of the orthorhombic grains rather than an increased monoclinic phase fraction. This directly affects the central structural claim and the subsequent m-phase endurance argument. Please provide a texture-corrected analysis (for example, multiple reflections, pole figures, or whole-pattern fitting) or explicitly reframe the conclusion as qualitative.","section":"GIXRD quantitative comparison (Fig. 3)"},{"comment":"Only one sample is reported for each of the five configurations, and no error bars or device counts are given for the GIXRD intensities, fatigue curves, or the 'statistical measurements' of endurance in Fig. 4b. Without replicates, the observed differences between the three co-doped stacks cannot be separated from run-to-run process variation, which weakens the claim that the spatial arrangement of dopants is the controlling factor.","section":"Electrical and structural measurements (Figs. 2-6)"},{"comment":"The statement that 'Zr/Al heterogeneous co-doping significantly enhances the endurance and 2P_r of MFIS stacks, compared to mono-doped HZO and HAO films' overstates the polarization result. According to Fig. 4a, the mono-doped HZO stack has the highest 2P_r, while the co-doped stacks have comparable but not higher 2P_r. The data support an increase in endurance relative to mono-doped HZO and an increase in 2P_r relative to mono-doped HAO, but not a simultaneous enhancement over both mono-doped controls. Please revise the conclusion to match the data.","section":"Concluding paragraph"},{"comment":"The claim that Zr4+ diffusion into the SiON layer degrades the interface and causes early breakdown is inferred only from an overlap in ToF-SIMS profiles and is not quantitatively linked to the measured leakage or endurance. Similarly, the suggestion that an increased m-phase fraction improves cycling resistance is drawn from the literature (Ref. 38) without direct evidence in these stacks. Please soften the causal language or add supporting data, such as SIMS quantification correlated with electrical stress results.","section":"Endurance mechanism (text after Fig. 7)"}],"minor_comments":[{"comment":"The phrase 'The von Neumann bottleneck 1 of conventional memory devices have motivated' has a subject-verb agreement error; it should be 'has motivated.'","section":"Introduction, first sentence"},{"comment":"For reproducibility, please specify the X-ray wavelength, incidence angle, and instrument geometry used for the GIXRD measurements.","section":"GIXRD data processing section"},{"comment":"The caption says '(a) in positive direction and (b) in negative direction'; removing the redundant '(a)' and '(b)' would make the caption cleaner.","section":"Fig. 5 caption"},{"comment":"The abbreviation RTP is defined as rapid thermal process, but the acronym is used later without a reminder; adding '(RTP)' at the first occurrence is sufficient.","section":"Sample fabrication paragraph"}],"recommendation":"major_revision","confidential_remarks":"The paper is within the scope of cond-mat.mtrl-sci and presents a technically interesting extension of the authors' earlier MFM work. The major concerns are experimental rather than theoretical and could be addressed by reporting texture-corrected XRD analysis and statistics from replicate samples; no fundamental flaw in the measurement approach is apparent. If the authors can provide those missing elements, the paper would be a solid candidate for publication."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The paper is a straightforward, useful extension of the group's earlier heterogeneous co-doping work from MFM to MFIS stacks, and the most important finding is likely real: putting HZO directly on the SiON interfacial layer hurts endurance, and co-doped designs with SiON/HAO interfaces combine high remanent polarization with good cycling stability. The ToF-SIMS evidence of Zr diffusion into the SiON layer gives a concrete physical explanation for the interface degradation, and the PUND leakage data corroborate it. That alone makes the paper worth reading for anyone engineering FeFET gate stacks.\n\nWhat is genuinely new here is the system-level demonstration in MFIS, not the co-doping concept itself. The authors are transparent about building on Ref. 30, and the electrical characterization is direct measurement, not fitted prediction. The paper also does a decent job of comparing five stack configurations and connecting structure to electrical behavior.\n\nThe soft spots are mostly in the structural interpretation. The stress-test concern is on point: the GIXRD intensities are normalized to the o(111) peak with no texture correction, even though the paper itself shows that o{200} intensity varies strongly across samples—direct evidence that texture is not uniform. So the claim that dopant arrangement \"primarily determines\" phase composition is not actually established by the XRD ratio. It is a plausible hypothesis, and the m-phase argument from the prior MFM paper adds context, but the quantitative phase-fraction comparison is not rigorous as presented. Second, every stack is a single sample with no error bars on the electrical data. The endurance differences are large enough that they probably survive process variation, but the statistics are thinner than the conclusions suggest.\n\nThese problems are real but they do not sink the paper. The practical messages—avoid SiON/HZO, use heterogeneous co-doping to balance Pr and endurance—rest on the electrical and SIMS data, not on the XRD texture analysis. I would send this to peer review, ideally at an applied physics or device journal, with a request for a revised manuscript that either applies a texture-aware analysis (or at least shows raw intensity ratios across more than one sample) and adds replicate measurements or softens the phase-composition language.\n\nFor a researcher working on ferroelectric HfO2 process integration, this is a worthwhile read and cite; for a general audience, it is a competent but niche process study.","headline":"Solid extension of the same group's MFM co-doping work into MFIS stacks; the SiON/HZO interface finding is the strongest part, while the GIXRD-based phase claims need texture-aware reanalysis.","tokens_in":9205,"tokens_out":2032,"would_cite":true,"duration_ms":24519,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"This paper claims that the vertical arrangement of Zr and Al dopants in HfO2 gate stacks controls which crystal phases form after annealing, and that a deliberately layered Zr/Al co-doping can deliver both high remanent polarization and…","keywords":["ferroelectric HfO2","Zr/Al co-doping","MFIS gate stack","FeFET endurance","remanent polarization","atomic layer deposition","phase composition","dopant spatial arrangement"],"falsifier":"Perform texture-corrected diffraction measurements (e.g., pole figures or multiple tilt angles) on the five same stack layouts and compute absolute phase fractions; if the center-HZO stack no longer shows a significantly higher monoclinic fraction once orientation effects are removed, the claim that dopant arrangement alone controls phase composition would be falsified.","tokens_in":8237,"feed_emoji":"⚡","tokens_out":4290,"duration_ms":45957,"temperature":0.7,"pith_summary":"The paper reports that in HfO2-based ferroelectric memory gate stacks, the vertical position of zirconium and aluminum dopants matters as much as their concentration. By depositing separate Zr-doped and Al-doped layers in different sequences, the authors steer the film's crystal phase composition after a high-temperature anneal. Electrical measurements show that a stack with the Zr-doped layer at the film's center matches the polarization of pure Zr-doped HfO2 while lasting far longer, and that co-doped stacks with an Al-doped layer next to the silicon interface combine the strengths of both mono-doped films. The paper also identifies a specific interface, SiON/HZO, as a reliability weak point to be avoided.","feed_headline":"Zr/Al layering boosts FeFET endurance and polarization","feed_subtitle":"ALD layer order decides which crystal phase forms in HfO2 memory gates — and which interface fails first.","key_machinery":"The central mechanism is heterogeneous co-doping by atomic layer deposition (ALD), which creates a deliberate vertical compositional gradient in the HfO2 film with distinct HZO and HAO layers. The argument rests on the difference in crystallization temperatures: HZO nucleates at temperatures as low as 300 °C, whereas HAO crystallizes above 650 °C. When HZO sits at the film's center, crystallization starts without the constraint of top/bottom interfaces, yielding more monoclinic phase; when HZO is confined to a top or bottom interface, adjacent capping layers stabilize the orthorhombic ferroelectric phase. This spatial control of phase competition, together with the dopant-diffusion behavior at the SiON interface, is what the paper uses to explain the measured differences in polarization, leakage, and endurance.","core_discovery":"The authors establish that heterogeneous Zr/Al co-doping significantly enhances both the endurance and the remanent polarization (2Pr) of metal-ferroelectric-insulator-semiconductor (MFIS) gate stacks, compared to mono-doped HZO (HfZrO2) and HAO (HfAlO) films. Structurally, the phase composition of annealed HfO2 films is primarily determined by the spatial arrangement of dopants: placing the HZO layer at the center of the film promotes monoclinic phase formation, which improves cycling resistance but reduces polarization, while placing it at the top or bottom interface favors the ferroelectric orthorhombic phase through capping effects. Electrically, the co-doped stack with HZO at the center and HAO at the SiON interface shows high polarization and no degradation up to $10^{4}$ cycles, merging the advantages of the two mono-doped films. The paper further argues that a direct SiON/HZO interface should be avoided because Zr4+ diffuses into the SiON insulator, raising leakage and accelerating fatigue.","pith_inferences":["A natural extension would be to apply the same layering concept to other dopant pairs (e.g., Si, La, Gd) and to superlattices of more than two doped layers, potentially expanding the phase-control knob beyond the Zr/Al pair studied here.","The texturing effect observed on the o{200} reflection in HAO and top-HZO films could be exploited to engineer polarization switching dynamics, although the paper only reports the correlation, not a causal model for how the texture arises.","A testable prediction is that the endurance benefit of central HZO placement should scale with the m-phase fraction; if texture-corrected phase fractions are measured, the correlation could be quantified and possibly used as a predictive metric for other dopant arrangements.","Because the paper shows that the SiON/HZO interface degrades both mono- and co-doped stacks, an alternative interfacial engineering (e.g., an ultra-thin Al-doped buffer) might recover endurance in bottom-HZO configurations, but this goes beyond what is demonstrated here."],"forward_implications":["FeFET gate-stack reliability can be tuned by ALD sequence alone, without introducing new materials or changing total dopant concentrations.","Co-doped stacks with an SiON/HAO interface and HZO at the center achieve high remanent polarization and stable switching up to at least 10^4 cycles, suggesting a practical recipe for nonvolatile memory gates.","Avoiding a direct SiON/HZO interface becomes a design rule for both mono- and co-doped HfO2 films, since Zr diffusion into the insulator is linked to premature breakdown.","The trade-off between memory window and endurance can be adjusted continuously by choosing where the low-temperature-crystallizing layer is placed within the ferroelectric film.","The o{200} texture enhancement seen in some co-doped and HAO films implies that crystallographic orientation can be influenced even without a textured bottom electrode, which may matter for scaling."],"supporting_citations":[{"why":"Prior demonstration of heterogeneous co-doping in MFM stacks that this work extends to MFIS stacks, providing the initial concept.","marker":"[30]"},{"why":"Shows HZO nucleation at temperatures down to ~300 °C, establishing the low crystallization temperature used in the argument.","marker":"[31]"},{"why":"Supplies further evidence of low-temperature crystallization of HZO films.","marker":"[32]"},{"why":"Reports HAO crystallization above 650 °C, the high-temperature contrast that drives the phase-competition mechanism.","marker":"[33]"},{"why":"Provides the capping-effect rationale for how adjacent layers stabilize the orthorhombic phase near interfaces.","marker":"[34]"},{"why":"Indicates that a higher monoclinic fraction enhances resistance to polarization cycling, supporting the endurance explanation.","marker":"[38]"},{"why":"Attributes degraded HZO to high annealing temperature, used to explain the higher leakage of the mono-doped HZO stack.","marker":"[39]"},{"why":"Establishes the monoclinic phase as thermodynamically stable and the basis for phase-stability reasoning in HfO2 ferroelectrics.","marker":"[13]"},{"why":"Identifies the metastable orthorhombic phase (Pca21) as the source of ferroelectricity, making the o-phase fraction the central structural indicator.","marker":"[23]"}],"fun_headline_variants":["Zr/Al layering order boosts FeFET endurance and polarization","HfO2 co-doping: Zr/Al layering boosts FeFET endurance","Zr/Al spatial order in HfO2 improves FeFET reliability","Co-doped HfO2: Zr/Al layering enhances FeFET endurance","Zr/Al arrangement tunes HfO2 phase and FeFET endurance"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The five compared films are otherwise identical except for the vertical ordering of the Zr- and Al-doped layers, and the X-ray peak ratios normalized to the o(111) reflection faithfully measure how much of each crystal phase is present without being distorted by preferred grain orientation or film-thickness differences.","fun_headline_variants_meta":{"raw":{"variants":["Zr/Al layering order boosts FeFET endurance and polarization","HfO2 co-doping: Zr/Al layering boosts FeFET endurance","Zr/Al spatial order in HfO2 improves FeFET reliability","Co-doped HfO2: Zr/Al layering enhances FeFET endurance","Zr/Al arrangement tunes HfO2 phase and FeFET endurance"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000817,"raw_usage":{"total_tokens":3566,"prompt_tokens":921,"completion_tokens":2645,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":537,"completion_tokens_details":{"reasoning_tokens":2551}},"tokens_in":537,"tokens_out":2645,"duration_ms":20774,"temperature":1.0,"reasoning_tokens":2551,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-15T17:09:53.206011+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Perform texture-corrected diffraction measurements (e.g., pole figures or multiple tilt angles) on the five same stack layouts and compute absolute phase fractions; if the center-HZO stack no longer shows a significantly higher monoclinic fraction once orientation effects are removed, the claim that dopant arrangement alone controls phase composition would be falsified.","supporting_citations":[{"cited_title":"Lee , author R","cited_arxiv_id":null,"evidence_quote":"Prior demonstration of heterogeneous co-doping in MFM stacks that this work extends to MFIS stacks, providing the initial concept."},{"cited_title":"Lederer , author P","cited_arxiv_id":null,"evidence_quote":"Shows HZO nucleation at temperatures down to ~300 °C, establishing the low crystallization temperature used in the argument."},{"cited_title":"Mulaosmanovic , author P","cited_arxiv_id":null,"evidence_quote":"Supplies further evidence of low-temperature crystallization of HZO films."},{"cited_title":"u hnel , author K. Seidel , author T. K\\","cited_arxiv_id":null,"evidence_quote":"Reports HAO crystallization above 650 °C, the high-temperature contrast that drives the phase-competition mechanism."},{"cited_title":"Ma \\ and\\ author N","cited_arxiv_id":null,"evidence_quote":"Provides the capping-effect rationale for how adjacent layers stabilize the orthorhombic phase near interfaces."},{"cited_title":"Chernikova , author M","cited_arxiv_id":null,"evidence_quote":"Attributes degraded HZO to high annealing temperature, used to explain the higher leakage of the mono-doped HZO stack."},{"cited_title":"Thomann , author A","cited_arxiv_id":null,"evidence_quote":"Establishes the monoclinic phase as thermodynamically stable and the basis for phase-stability reasoning in HfO2 ferroelectrics."}],"review_version":1}