{"id":"7f738be0-376e-4bc2-a08b-0975678963bc","arxiv_id":"2411.14872","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":2,"one_line_summary":"Low-temperature plasma-assisted MBE yields crack-free, phase-pure, N-polar wurtzite Sc_xAl1-xN films on Si(111) for scandium contents up to x=0.3.","lead":"This paper reports a growth recipe for crack-free, single-polarity scandium aluminum nitride films on silicon wafers, up to 30% scandium, by using much lower growth temperatures than usual. It matters because scandium aluminum nitride is a promising piezoelectric and ferroelectric material, and growing it directly on silicon in this form could enable new acoustic and electrical functions in silicon devices.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The scandium content x is never measured on the films, and the stated N*/III=2 for x=0.3 is arithmetically inconsistent with N*/Al=3.3 (which implies 2.31); the quantitative claim of phase purity up to x=0.3 rests on unverified flux calibration.","rationale":"The reader's weakest_assumption identifies the same load-bearing issue: the scandium fraction is flux-calibrated, not measured. I agree with that assessment. The most important additional evidence is the arithmetical inconsistency in the reported N*/III ratio: for N*/Al=3.3, x=0.3 implies N*/III≈2.31, not the stated 2. This does not by itself falsify the paper, but it shows that the composition calibration is not merely an unverified detail—it contains a concrete numerical discrepancy that could indicate a systematic error in the flux-derived x. The central growth strategy (lowering growth temperature to avoid cracks and secondary phases) is physically plausible and well supported by the XRD, RHEED, and AFM data. The polarity assignment, though inferred from the AlN buffer rather than directly on ScAlN, is a secondary concern because the ScAlN is grown continuously on a verified N-polar template during the same run. The paper is appropriately scoped and openly acknowledges the composition uncertainty, so the CONDITIONAL verdict is justified. No change to the reader's verdict is needed; the requested direct composition measurement is precisely the condition that would convert the conditional acceptance to a full acceptance.","tokens_in":11967,"tokens_out":6262,"duration_ms":59270,"concrete_test":"Perform Rutherford backscattering spectrometry on the Sc0.3Al0.7N layer grown at 300°C and on at least one other composition (e.g., the x=0.25 layer) to determine the actual Sc/(Al+Sc) ratio. Compare with the flux-derived values. Simultaneously, re-derive the N*/III ratio from the stated fluxes: if N*/Al=3.3, then for x=0.3, N*/III=2.31, not 2. If the measured composition deviates from the stated x by more than 0.02, recompute the a(x) slope and re-locate the phase-purity and cracking boundaries in the temperature-composition map.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central result—crack-free, phase-pure wurtzite ScAlN on AlN-buffered Si(111) up to x=0.3 with on-/off-axis rocking-curve widths below 2°/3°—depends on the scandium fraction x being quantitatively correct. Yet x is never measured by a composition-sensitive technique; it is derived solely from the Sc flux while Al and N fluxes are held constant. The authors explicitly flag this in the text: 'our assessment of the Sc content may differ from previous ones in the literature.' A concrete internal inconsistency strengthens the concern: with N*/Al = 3.3, the N*/III ratio at x=0.3 should be 3.3×(1−0.3) = 2.31, not the stated value of 2. If the true x is higher (≈0.39 for N*/III=2), the claimed upper boundary for phase purity at x=0.3 is not actually demonstrated, and the fitted a(x)=0.7x+3.112 relationship would shift. If the true x is lower, the headline claim is overstated. Either way, the composition temperature boundaries (x=0.2, 0.25, 0.3) and the lattice-parameter slope are not securely tied to measured compositions. This is load-bearing because every quantitative statement in the paper is a function of x.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The manuscript reports a PAMBE growth study of N-polar Sc_xAl_1-xN layers (0 ≤ x ≤ 0.35, nominal thickness 340 nm) on AlN-buffered Si(111). The authors demonstrate that lowering the growth temperature with increasing Sc content prevents cracking and inhibits the formation of intermetallic Al-Sc inclusions, yielding crack-free and nominally phase-pure wurtzite layers with on-axis (off-axis) rocking-curve widths below 2° (3°) up to a claimed Sc content of x = 0.3. The scandium content is set by flux calibration only and is not measured by a composition-sensitive technique; the paper explicitly notes that its assessment of Sc content may differ from earlier reports. The lattice parameter a is reported to follow a(x) = (0.7x + 3.112) Å.","tokens_in":12167,"tokens_out":5779,"duration_ms":53504,"significance":"If the composition calibration is accurate, the paper offers a practically useful growth window for N-polar ScAlN on Si with application to piezoelectric and ferroelectric devices integrated on silicon. The study is systematic, and several aspects are well executed: the N-polar character of the AlN buffer is supported by RHEED and KOH etching, the authors carefully separate the AlN buffer contribution from the ScAlN XRC signal using two-component fits, and the strain state is examined with reciprocal space maps. However, the central quantitative claims—the phase-purity boundary at x ≈ 0.3, the crack-onset at x ≈ 0.2, and the fitted a(x) slope—are all indexed by x, and x is never directly measured. Because the paper itself flags the composition uncertainty and because the stated flux ratio contains an internal inconsistency, the headline results must be regarded as conditional on the flux calibration. The significance is therefore moderate and would be strengthened substantially by an independent composition measurement or an explicit uncertainty analysis.","major_comments":[{"comment":"The scandium content x is set only by the Sc flux while Al and N* fluxes are held constant; no composition-sensitive measurement (RBS, EDX, SIMS, or XRD-based composition calibrated against a standard) is reported. The manuscript itself states, 'our assessment of the Sc content may differ from previous ones in the literature' (paragraph discussing Fig. 4). This is load-bearing because the phase-purity and cracking boundaries (x = 0.2, 0.25, 0.3) and the fitted lattice-parameter relation a(x) = 0.7x + 3.112 in Fig. 5 are all quoted as functions of x. I also note an internal inconsistency in the stated flux ratios: with N*/Al = 3.3 and only Al and Sc as group-III metals, N*/III at nominal x = 0.3 should be 3.3 × 0.7 = 2.31, not 2 as written; conversely, N*/III = 2 would correspond to x ≈ 0.39. The authors should reconcile this arithmetic and provide either a direct composition measurement or a quantitative uncertainty budget for x. Without this, the claimed upper boundary for phase purity at x = 0.3 is not securely demonstrated.","section":"Experimental methods and Figure 5"},{"comment":"Phase purity is inferred from the absence of additional XRD reflections, but no detection limit is stated. The XRD scans are recorded with an open detector, which helps sensitivity, but the minimum detectable volume fraction of a secondary phase is not quantified. This matters particularly because the 500°C Sc0.3Al0.7N sample shows no 0002 or 0004 reflection from the ScAlN layer itself—only the AlN buffer reflections and AlSc/Al3Sc peaks are visible—indicating that the layer is either severely disordered, misoriented, or absent. The claim that the 300°C sample is 'phase-pure wurtzite' rests on the absence of secondary reflections in that sample, but without a detection limit, a small volume fraction of rocksalt ScN or intermetallic inclusions cannot be excluded. The authors should add an estimate of the detection sensitivity (e.g., from the signal-to-noise ratio and the structure factors of candidate phases) or support the phase-purity claim with a complementary technique such as TEM.","section":"Figure 3 and phase-purity discussion"},{"comment":"There is an inconsistency in the reported layer thickness. The abstract and the growth description state that the Sc_xAl_1-xN layers are 340 nm thick, but the caption of Fig. 4(a) says 'with a thickness of 240 nm.' Rocking-curve widths, strain relaxation, and cracking behavior are all thickness-dependent, so this discrepancy must be resolved. If the data in Fig. 4(a) were taken on 240-nm-thick layers, the text and abstract should be corrected accordingly, or the caption should be fixed.","section":"Abstract and Figure 4(a)"}],"minor_comments":[{"comment":"The phrase 'due the high thermal mismatch' should read 'due to the high thermal mismatch.'","section":"Abstract"},{"comment":"The word 'dependong' in the growth-temperature description should be 'depending.'","section":"Experimental section"},{"comment":"The sentence 'These values are is in good agreement' contains a grammatical error and should read 'These values are in good agreement.'","section":"Lattice-constant paragraph"},{"comment":"The text mentions 'intermetallic phases such as AlSc and Al3ScN', but the abstract and Fig. 3(a) identify AlSc and Al3Sc. The formula 'Al3ScN' appears to be a typo and should be corrected to 'Al3Sc.'","section":"Phase-separation paragraph"},{"comment":"Reference 34 contains a typo: 'second harmonic ceneration' should be 'second harmonic generation.'","section":"Reference 34"}],"recommendation":"major_revision","confidential_remarks":"The main blocker is the lack of a direct composition measurement. The internal flux-ratio inconsistency (N*/III = 2 vs. N*/Al = 3.3 implying 2.31) might be a simple typo, but it currently undermines confidence in the calibration. If the authors can provide RBS or EDX data for at least the x = 0.2 and x = 0.3 samples, or an explicit calibration uncertainty analysis, the paper would be suitable. The thickness inconsistency in Fig. 4(a) should also be resolved. I would not reject the paper; the growth study appears useful and the experimental methodology is otherwise careful, but the composition issue is load-bearing for the central claims."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"For someone working on ScAlN integration, this is a useful paper. The genuinely new result is the temperature-composition window: by dropping Tg from 740 to 300 °C as the Sc flux rises, they get crack-free, phase-pure wurtzite ScAlN on AlN-buffered Si(111) up to x=0.3, with on-/off-axis rocking curves under 2°/3°. That goes beyond Park’s x=0.12 on Si(111) and gives a practical recipe. The authors are also honest about the AlN buffer’s contribution to the XRC peaks, and the XRD/RHEED/AFM/KOH evidence is internally consistent. The lattice-constant fit a(x)=0.7x+3.112 is descriptive, not predictive, and they say so by comparing to their own strained-on-GaN data.\n\nThe soft spots are real but not fatal. The Sc content is flux-calibrated, never measured on the films; the authors even flag that their x may differ from literature values. That means the crack and phase boundaries at x=0.2, 0.25, 0.3 are pinned to an unverified scale. Also, the N*/III claim is arithmetically wrong: with N*/Al=3.3, the x=0.3 ratio should be 2.31, not 2. That looks like a typo, but it suggests the flux bookkeeping is not squeaky clean. Second, the N-polarity of the ScAlN itself is inferred from the AlN buffer, not confirmed on the alloy. Given that the buffer is grown under identical conditions to prior N-polar AlN work, the inference is reasonable, but it remains an inference.\n\nThe paper is scoped well and the central claim—that low-temperature growth suppresses both cracks and intermetallic phases—holds up as an empirical observation. It deserves a serious referee, and the requested edits are straightforward: measure composition on at least one film (RBS or EDX) and fix or explain the N*/III numbers. This is not a home run, but it is a solid, useful contribution for the nitride MBE community.\n\nRecommendation: send to peer review, with the composition and polarity points as the main asks.","headline":"A credible growth study that maps crack-free N-polar ScAlN on Si(111) up to x=0.3, but the unmeasured composition and a minor flux-ratio slip keep it from being fully load-bearing.","tokens_in":12830,"tokens_out":1407,"would_cite":true,"duration_ms":27720,"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":"By stepping the growth temperature down as scandium content rises, the authors obtain crack-free, phase-pure wurtzite Sc$_x$Al$_{1-x}$N layers on Si(111) up to $x=0.3$.","keywords":["scandium aluminum nitride","molecular beam epitaxy","wurtzite","N-polar","Si(111)","growth temperature","crack-free","intermetallic phases"],"falsifier":"Measure the actual scandium fraction in the 340-nm Sc$_{0.3}$Al$_{0.7}$N layer grown at 300 °C with Rutherford backscattering spectrometry; if the measured composition deviates from 30% by more than the usual few-percent uncertainty, the reported cracking and phase-purity boundaries and the fitted lattice-parameter slope are shifted from their stated values.","tokens_in":11670,"feed_emoji":"💎","tokens_out":7969,"duration_ms":68202,"temperature":0.7,"pith_summary":"Plasma-assisted molecular beam epitaxy of the piezoelectric alloy Sc$_x$Al$_{1-x}$N on silicon has been held back by two failures: thermal-mismatch cracking on cooling and the appearance of intermetallic phases at higher scandium contents. This paper claims that both can be avoided by lowering the growth temperature as the scandium fraction rises: 740 °C for $x\\le0.1$, 500 °C for $x\\approx0.2$–0.25, and 300 °C for $x\\approx0.3$. The resulting layers are crack-free, single-phase wurtzite, N-polar, with on-axis (off-axis) x-ray rocking curve widths below 2° (3°) up to a scandium content of 0.3. If this holds, it opens a practical route to integrating the strong piezoelectric and ferroelectric response of scandium aluminum nitride with silicon substrates.","feed_headline":"Crack-free ScAlN on Si(111) achieved up to 30% scandium","feed_subtitle":"Stepwise cooling from 740°C to 300°C suppresses cracks and intermetallic phases in N-polar wurtzite layers.","key_machinery":"The central mechanism is the reduction of the growth temperature itself, acting on two fronts. It lowers the tensile stress that accumulates during cooling from growth temperature to room temperature, which would otherwise crack the layer because the thermal-expansion mismatch between Sc$_x$Al$_{1-x}$N and Si increases with $x$. It also kinetically suppresses the formation of the intermetallic compounds AlSc and Al$_3$Sc, which appear under metal-stable conditions at higher temperature. The enabler is the single-domain N-polar AlN buffer grown on an Al-pre-deposited Si(111) surface, which fixes the polarity of the layers above it.","core_discovery":"The paper establishes that growth temperature is the decisive control parameter for the structural quality of Sc$_x$Al$_{1-x}$N on Si(111). With an AlN buffer layer that is N-polar, as shown by the (3$\\times$3) reflection high-energy electron diffraction pattern and confirmed by KOH etching, layers grown at a recipe that steps the temperature from 740 °C down to 500 °C and finally 300 °C with increasing scandium content are crack-free and phase-pure wurtzite up to $x=0.3$. For Sc$_{0.3}$Al$_{0.7}$N, growth at 500 °C produces AlSc and Al$_3$Sc inclusions, whereas growth at 300 °C suppresses them. The in-plane lattice constant $a$ grows linearly with composition as $a(x)=(0.7x+3.112)\\,\\AA$ while $c$ remains nearly constant, indicating essentially fully relaxed layers.","pith_inferences":["The same stepwise-temperature-lowering strategy may transfer to other wurtzite III-nitride alloys on Si(111) that crack from thermal mismatch and form metal-rich secondary phases at higher temperature.","X-ray diffraction sets only a coarse detection limit for secondary phases; transmission electron microscopy or atom-probe tomography of the 300 °C-grown Sc$_{0.3}$Al$_{0.7}$N layer could reveal nanoscale AlSc or Al$_3$Sc inclusions that the 2$\\theta$–$\\omega$ scans would miss.","If the flux-derived scandium fractions are systematically offset from true compositions, the fitted lattice-parameter slope $a(x)=0.7x+3.112$ and the claimed $x=0.3$ phase-purity boundary would shift accordingly; a direct composition measurement would tighten the recipe.","The strong broadening of rocking curves at 150–250 °C suggests the useful temperature window is narrow, and growth-rate or flux adjustments may widen it without sacrificing phase purity."],"forward_implications":["Crack-free wurtzite Sc$_x$Al$_{1-x}$N on Si(111) becomes available for surface-acoustic-wave, ferroelectric, and piezoelectric devices that need uninterrupted in-plane electrical transport and acoustic propagation.","The scandium fraction attainable on silicon with epitaxial-quality layers is extended from about 0.12 to 0.3 while keeping the on-axis rocking curve width below 2°.","Because the layers are single-domain N-polar, ferroelectric switching and electromechanical conversion should be more uniform than in the mixed-polarity textured films commonly produced by sputtering.","The measured $a(x)$ relation confirms that ScAlN is lattice-matched to GaN near $x\\approx0.1$, providing a quantitative basis for AlScN/GaN heterostructure design.","The abrupt broadening of the on-axis rocking curve for $x>0.3$, attributed to cubic precipitates, fixes the practical composition limit of this growth method near $x=0.3$."],"supporting_citations":[{"why":"Supplies the flux-ratio method for setting the scandium fraction and the earlier $a(x)$ slope of 0.85 for comparison.","marker":"[11]"},{"why":"Provides the previous state of the art for epitaxial ScAlN on Si(111) at $x=0.12$, including the rocking-curve baseline.","marker":"[4]"},{"why":"Contributes the aluminum pre-deposition method used to obtain N-polar AlN on Si(111).","marker":"[19]"},{"why":"Quantifies the thermal-expansion mismatch and elastic properties of ScAlN that drive the cracking problem.","marker":"[21]"},{"why":"Documents phase-purity control and secondary-phase formation in similarly grown ScAlN films.","marker":"[23]"},{"why":"Gives the Al-Sc phase diagram identifying AlSc and Al$_3$Sc as the intermetallic phases observed.","marker":"[24]"},{"why":"Establishes the ternary Sc-Al-N phase equilibria underlying the intermetallic precipitation behavior.","marker":"[25]"},{"why":"Provides the surface-reconstruction basis for assigning N-polarity from the (3$\\times$3) pattern.","marker":"[35]"},{"why":"Supplies reference lattice constants of relaxed AlN used to estimate residual strain.","marker":"[37]"}],"fun_headline_variants":["Crack-free ScAlN on Si up to 30% Sc via temperature steps","Cooling to 300°C suppresses cracks and intermetallic phases in ScAlN","ScAlN growth on Si: lower temperature for pure wurtzite layers","N-polar ScAlN on Si crack-free with 30% Sc using staged cooling"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The scandium fraction $x$ is set by the scandium flux relative to fixed aluminum and nitrogen fluxes, and is never verified by a composition-sensitive technique such as Rutherford backscattering or energy-dispersive x-ray spectroscopy, so every boundary reported as a function of $x$ inherits any error in that flux calibration.","fun_headline_variants_meta":{"raw":{"variants":["Crack-free ScAlN on Si up to 30% Sc via temperature steps","Cooling to 300°C suppresses cracks and intermetallic phases in ScAlN","ScAlN growth on Si: lower temperature for pure wurtzite layers","N-polar ScAlN on Si crack-free with 30% Sc using staged cooling"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000606,"raw_usage":{"total_tokens":2872,"prompt_tokens":1039,"completion_tokens":1833,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":655,"completion_tokens_details":{"reasoning_tokens":1742}},"tokens_in":655,"tokens_out":1833,"duration_ms":14402,"temperature":1.0,"reasoning_tokens":1742,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-12T14:46:19.590611+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Measure the actual scandium fraction in the 340-nm Sc$_{0.3}$Al$_{0.7}$N layer grown at 300 °C with Rutherford backscattering spectrometry; if the measured composition deviates from 30% by more than the usual few-percent uncertainty, the reported cracking and phase-purity boundaries and the fitted lattice-parameter slope are shifted from their stated values.","supporting_citations":[{"cited_title":"Dasgupta , author F","cited_arxiv_id":null,"evidence_quote":"Contributes the aluminum pre-deposition method used to obtain N-polar AlN on Si(111)."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Establishes the ternary Sc-Al-N phase equilibria underlying the intermetallic precipitation behavior."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provides the surface-reconstruction basis for assigning N-polarity from the (3$\\times$3) pattern."},{"cited_title":"Paszkowicz , author S","cited_arxiv_id":null,"evidence_quote":"Supplies reference lattice constants of relaxed AlN used to estimate residual strain."}],"review_version":1}