{"id":"be8a859a-cd07-4777-aab7-596c798a9371","arxiv_id":"2506.12173","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":2,"one_line_summary":"Band offsets at beta-Ga2O3/(AlxGa1-x)2O3 interfaces depend strongly on growth direction and strain, and TCAD models using these offsets reproduce forward-bias diode measurements.","lead":"This paper uses atomistic simulations and device modeling to show that the growth orientation and strain of aluminum-gallium-oxide layers on gallium-oxide crystals strongly change the electronic barriers at the interface. The results help explain why measured band offsets vary widely and offer a route for designing better power-electronics diodes.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Alloy band offsets rely on linearly interpolated interface potential lineups ΔV_AB with only x=0, 0.5, 1 computed; a nonlinear or configuration-dependent lineup would shift composition-dependent CBO/VBO values and could change offset type at near-zero-VBO orientations.","rationale":"The reader's weakest assumption identifies the same soft spot, and I agree. The paper's strongest independent evidence is the explicit interface calculations for the pure compounds and the ordered x=0.5 compound, plus the forward-bias TCAD comparisons against Refs. 37 and 39 with no current-fitting parameters. However, the scientific claim about alloying—'band offsets are computed from first principles for different Al concentrations'—is not fully supported at arbitrary x because the interface lineup is the only part of Eqs. (1)–(2) that is linearly interpolated rather than recomputed. The band-edge reference energies themselves include SQS-based bowing, so the final offsets mix a nonlinear band-edge term with a linear lineup term in a way that has no validation. The experimental TCAD comparisons use electron affinities from surface-slab calculations, so even a successful forward-bias match does not test the interpolated lineup; it tests a related but different set of calculations. This leaves a concrete, testable gap rather than a contradiction. The verdict should remain CONDITIONAL: the qualitative orientation trends for pure compounds and the forward-bias TCAD agreement stand, but the quantitative composition-dependent offset values should not be used until the interpolation is checked. Hence UNCHANGED relative to the reader's verdict.","tokens_in":11828,"tokens_out":12577,"duration_ms":138615,"concrete_test":"Compute explicit SQS-based interface superlattices for β-Ga2O3/(Al0.25Ga0.75)2O3 and β-Ga2O3/(Al0.75Ga0.25)2O3 for the (010) orientation using the same SCAN/HSE06 protocol, and obtain ΔV_AB directly from the macroscopic average electrostatic potential of each fully relaxed supercell. Compare the resulting VBO/CBO to the linearly interpolated values from x=0, 0.5, 1. Also rebuild the x=0.5 superlattice with an SQS random-alloy configuration rather than the fully ordered all-octahedral-Al GaAlO3, and check whether the lineup changes. If any VBO/CBO shifts by more than 0.1 eV, or the (-201) alloy VBO crosses zero, the linear-interpolation assumption is falsified and the composition-dependent offset claims require revision.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The most load-bearing weakness is the alloy interface potential lineup. The paper states (Section III): 'The values of the average electrostatic potential differences, ΔV_AB, for alloys are obtained by means of linear interpolation using Ga2O3, GaAlO3 and Al2O3.' Only x=0, the ordered x=0.5 GaAlO3 (all octahedral Al), and x=1 are explicit interface calculations; all intermediate alloy offsets inherit a linear interpolation of the interface dipole. This is an untested assumption: the lineup can depend nonlinearly on composition and on configurational disorder. The ordered x=0.5 anchor is not a representative SQS random alloy (in the random alloy, Al occupies octahedral sites for x<0.5 and progressively fills tetrahedral sites for x>0.5), so the midpoint of the interpolation may be biased for random alloys. If ΔV_AB(x) is nonlinear, the composition-dependent VBO/CBO values are systematically shifted. The error matters most for orientations with small VBO: at (-201) the pure-Al2O3 VBO is only -0.05 eV, so an interpolation error on the order of 0.1 eV at finite x would change the sign/type of the alloy band alignment. The TCAD forward-bias comparisons at x=0.21/0.22 use electron affinities from separate surface-slab calculations, so they do not directly test the interpolated lineup; the vulnerable quantities are the tabulated composition-dependent offsets.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The manuscript reports hybrid-DFT (HSE06) calculations of band offsets at pseudomorphic beta-Ga2O3/(AlxGa1-x)2O3 heterojunctions for four growth orientations, (100)B, (010), (001)B, and (-201), including the effect of strain on the Al2O3 and alloy layers. The authors use explicit superlattice calculations for the pure end members and an ordered x=0.5 GaAlO3 composition, with the alloy interface potential lineup obtained by linear interpolation. The resulting offsets and electron affinities are fed into TCAD simulations of Pt/(AlxGa1-x)2O3/Ga2O3 Schottky barrier diodes, and the simulated I-V curves are compared with published experimental data at x=0.21 and x=0.22. The central claims are that growth orientation and strain change the band offsets substantially, both in magnitude and type, and that TCAD simulations using the first-principles parameters reproduce the forward-bias experimental I-V characteristics.","tokens_in":12085,"tokens_out":4319,"duration_ms":150713,"significance":"If the central claims hold, the paper provides a useful systematic map of orientation- and strain-dependent band offsets for the important beta-Ga2O3/(AlxGa1-x)2O3 materials system and demonstrates a pathway for embedding first-principles offsets into TCAD device models. The strengths of the work include the use of explicit superlattice calculations for the pure beta-Ga2O3/theta-Al2O3 interfaces, the care taken to include strain in the film reference energies, the use of nonpolar symmetric terminations to avoid built-in fields, and quantitative comparison with published I-V data. The orientation dependence reported for the pure Al2O3 interface, including the near-zero VBO for the (-201) orientation, is physically plausible and defensible. However, the alloy-composition dependence, which is a central part of the paper's scope, rests on an interpolation assumption for the interface potential lineup that is not validated explicitly, and this limits the quantitative reliability of the composition-dependent offsets and of the validation claims built on them.","major_comments":[{"comment":"The paper states that 'The values of the average electrostatic potential differences, dV_AB, for alloys are obtained by means of linear interpolation using Ga2O3, GaAlO3 and Al2O3.' Only three compositions are actually computed for the interface lineup: x=0, the ordered x=0.5 GaAlO3, and x=1. This is a load-bearing assumption for the composition-dependent VBO and CBO values. The interface potential lineup can be nonlinear in x and can depend on the cation arrangement; the x=0.5 anchor is an ordered structure with all octahedral Al, which is not representative of the random SQS alloys used elsewhere in the paper, especially for x>0.5 where Al begins to occupy tetrahedral sites. Since the (-201) orientation has a pure-Al2O3 VBO of only -0.05 eV, an interpolation error of order 0.1 eV at finite x could change the sign of the offset and hence the alignment type. Please compute the potential lineup from explicit (AlxGa1-x)2O3/Ga2O3 superlattices for at least two or three additional compositions using SQS models, or otherwise provide a quantitative bound on the interpolation error and show that the main qualitative conclusions are insensitive to it.","section":"Section III, Eqs. (1)-(2) and text near Fig. 1"},{"comment":"The experimental validation in Figs. 4 and 5 is presented as supporting the calculated band offsets, but the TCAD comparisons at x=0.21 and x=0.22 use electron affinities from separate surface-slab calculations rather than the interface potential lineup that is the central quantity in Eqs. (1) and (2). The forward-bias current in these devices is controlled by the Ga2O3 electron affinity and the Pt/alloy Schottky barrier height, so the agreement with experiment does not directly test the interpolated dV_AB values or the orientation-dependent interface dipoles. The statement that the simulations validate the computed offsets should therefore be softened, and the fact that the validation is indirect should be acknowledged. In addition, the agreement is described only qualitatively; reporting the extracted barrier heights or ideality factors from the simulated and experimental curves would provide a more quantitative assessment.","section":"Section III, Figs. 4-5 and Table S2"},{"comment":"The band-gap bowing parameter b is fitted separately for each orientation, and the HSE06 mixing parameter is set to 0.32 rather than the standard 0.25. These are free parameters in the calculation, and the reported offset values inherit their uncertainty. This is not itself an error, but the paper should state explicitly how sensitive the final VBO/CBO values are to the choice of mixing parameter and to the bowing fit. In particular, the near-zero VBO for the (-201) orientation may be within the uncertainty of the HSE mixing parameter choice, and this possibility should be discussed when making claims about offset-type reversal.","section":"Section III, Eq. (3) and Table I"}],"minor_comments":[{"comment":"The text lists 'd10s2p1 for indium' although indium is not part of the systems studied here; this appears to be a typo for gallium, and the intended valence configurations for Ga, Al, and O should be stated unambiguously.","section":"Section II, pseudopotential description"},{"comment":"The text refers to the 'conduction band maximum' when describing CBM quantities; this should be the conduction band minimum throughout.","section":"Section II, Eqs. (1)-(2)"},{"comment":"The conclusion states 'Al concentrations of 0.22 % and 0.21 %' where the intended values are x=0.22 and x=0.21 (i.e., 22% and 21% Al content); please correct the percentages.","section":"Section IV, Conclusion"},{"comment":"Table I reports lattice parameters a, b, c for bulk materials and strained films but does not clearly indicate which axes are in-plane versus out-of-plane for each orientation; please clarify the axis conventions or refer more explicitly to the transformation matrix in the Supporting Information.","section":"Table I and associated text"},{"comment":"Reference [36] for the Ginestra TCAD software is incomplete; please provide a version number or a more specific citation so readers can identify the exact tool and version used.","section":"Reference [36]"},{"comment":"The abstract uses 'AlGaO' instead of the standard notation '(AlxGa1-x)2O3', and the title/abstract phrasing 'Alloying Effect' is slightly inconsistent with the stated scope 'on Electronic Properties'; please make the notation consistent.","section":"Abstract and title"}],"recommendation":"major_revision","confidential_remarks":"The manuscript is within the scope of cond-mat.mtrl-sci and presents a potentially useful methodology. The central concern is the untested linear interpolation of the alloy interface potential lineup, which affects the quantitative composition-dependent offsets and the strength of the claimed validation. This is fixable with additional explicit SQS interface calculations and a more cautious framing of the experimental comparison, so I do not recommend rejection; major revision is appropriate."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Short version: this is a credible paper with a clear qualitative result and one load-bearing approximation that the authors openly state but do not stress-test. The result that growth orientation and strain change the Ga2O3/Al2O3 valence and conduction band offsets by several tenths of an eV, including a sign change at (-201), is directly computed and convincing. The alloy composition dependence is less secure, because the interface potential lineup ΔV_AB is linearly interpolated from x=0, 0.5, 1, with the midpoint being an ordered GaAlO3 structure rather than a random SQS. For the x~0.2 values used in the TCAD validation this may be acceptable, but near (-201) where the pure Al2O3 VBO is only -0.05 eV, an interpolation error of order 0.1 eV would flip the offset type. The authors state this interpolation explicitly, so it is an honest limitation rather than a hidden one.\n\nWhat is genuinely new here is the systematic sweep: consistent HSE06 offsets for four orientations, strained films, alloy concentrations, and the feed-forward into Ginestra TCAD I-V/C-V predictions. Ref. 21, Ref. 22, and Ref. 27 cover pieces of this, but the full dataset plus the device-level comparison is not in the literature as a single result. The forward-bias agreement with the two experimental datasets, particularly Sundaram et al., is meaningful given that no parameter is fitted to the target current; the (-201) VBO also lands near the XPS value. The paper is transparent about the ideal Schottky and defect-free assumptions in the device model.\n\nThe soft spots are real but proportionate. The interpolated ΔV_AB is the main one; the TCAD validation actually tests the surface-slab affinities, not the interpolated lineup, so the vulnerable quantity is the tabulated composition-dependent offsets. The authors should either compute explicit alloy interfaces at one or two intermediate concentrations or bound the interpolation error with a few SQS configurations. The experimental validation is limited to the (010) orientation and forward bias, and the fig. 4 comparison shows noticeable magnitude deviations, so 'reasonable agreement' is doing some work. No error bars are given on the experimental curves. Minor: the novelty is incremental relative to the cited works, but the consolidation has practical value.\n\nThis paper is for people working on beta-Ga2O3 power electronics and band-offset engineering. It deserves a serious referee: the qualitative orientation/strain conclusion is solid, and the interpolation issue is fixable in revision. I would send it to review.","headline":"A credible, useful band-offset study for beta-Ga2O3/(AlxGa1-x)2O3 whose qualitative orientation/strain result is solid, but whose quantitative alloy offsets rest on an untested linear interpolation of the interface potential lineup.","tokens_in":12682,"tokens_out":4162,"would_cite":true,"duration_ms":43142,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":["71.15.Mb","71.20.Nr","73.40.-c"],"model":"deepseek-v4-flash","headline":"This paper establishes that growth orientation and interface strain alter beta-Ga2O3/(AlxGa1-x)2O3 band offsets by more than 2 eV, and that device simulations built on these offsets reproduce measured forward-bias diode currents.","keywords":["beta-Ga2O3","band offsets","heterojunctions","(AlxGa1-x)2O3 alloys","pseudomorphic strain","Schottky barrier diodes","growth orientation","TCAD device simulation"],"falsifier":"Compute explicit disordered (Al0.25Ga0.75)2O3/Ga2O3 superlattices for two orientations and compare CBO/VBO with values linearly interpolated from the endpoints; a deviation beyond numerical noise would falsify the quantitative alloy predictions. Alternatively, measure core-level photoemission band offsets on (100)- and (-201)-oriented films with the same aluminium fraction and check the roughly 0.5 eV-scale orientation split.","tokens_in":11573,"feed_emoji":"⚡","tokens_out":6959,"duration_ms":83028,"temperature":0.7,"pith_summary":"This paper tries to establish that band offsets at beta-Ga2O3/(AlxGa1-x)2O3 interfaces are not fixed material constants but move strongly with growth orientation and with the strain a pseudomorphic alloy film inherits from the substrate. Using hybrid-functional density functional theory, the authors compute valence- and conduction-band offsets for four orientations and find values ranging from 0.53 eV to -0.05 eV for the valence-band offset across (100)B and (-201). They then insert these first-principles offsets into Schottky diode device simulations and show that the forward-bias current-voltage characteristics reproduce published measurements for roughly 21-22% aluminium. If correct, the work makes growth orientation and strain practical design knobs for ultra-wide-bandgap power devices, and it explains part of the scatter in previously reported band-offset values.","feed_headline":"Growth direction swings Ga2O3 band offsets by 2.4 eV","feed_subtitle":"First-principles offsets fed into diode models match measured forward current, making orientation a design lever.","key_machinery":"The load-bearing object is the band-offset construction that combines bulk band-edge energies, referenced to the average electrostatic potential, with the interface potential lineup computed from fully relaxed superlattice structures: VBO = (bulk VBM energy difference) + interface potential lineup, and likewise for the CBO. The films are strained in plane to match the substrate lattice, then relaxed out of plane, and the lineup is extracted from the macroscopic average electrostatic potential of the formed interface. For alloy compositions, the lineup is obtained by linear interpolation using Ga2O3, ordered GaAlO3, and Al2O3 endpoint calculations rather than by explicit alloy interfaces at every concentration.","core_discovery":"The central claim is that both the magnitude and the type of band offset at beta-Ga2O3/(AlxGa1-x)2O3 interfaces depend so strongly on growth orientation and on the in-plane strain of the pseudomorphic film that previously reported offset values cannot be treated as unique. For the pure Ga2O3/Al2O3 interface, the (100)B orientation shows the largest offsets, VBO = 0.53 eV and CBO = 2.44 eV, while the (-201) orientation shows the smallest, with a VBO of only -0.05 eV. The strain matters because the aluminum oxide film's band edges shift noticeably under the in-plane tensile strain imposed by the beta-Ga2O3 substrate, and the interface potential lineup then determines the final offset. When the computed offsets are fed into Schottky barrier diode models, the simulated forward-bias I-V curves reproduce experimental data for (Al0.21Ga0.79)2O3 and (Al0.22Ga0.78)2O3 layers on beta-Ga2O3(010), with reverse-bias discrepancies attributed to the ideal, defect-free junction assumption. The conclusion is that orientation and strain engineering are first-order levers for tailoring electron confinement, built-in voltage, and diode switching behavior in this material system.","pith_inferences":["If the orientation dependence survives in experiment, growth-axis selection could be used to set the Schottky barrier height and the heterojunction band offset at the same time, reducing the need for composition or thickness grading in some diode designs.","The paper's linear interpolation of the interface lineup is the most exposed assumption; explicit alloy superlattice calculations at one intermediate concentration would be the quickest test and might refine the quantitative CBO values without changing the orientation ordering.","A similar strain- and orientation-aware treatment could apply to other low-symmetry oxide heterojunctions, where bulk band-edge shifts under pseudomorphic strain are large enough to change the sign of an offset."],"forward_implications":["On (100)B the computed offsets are largest (VBO = 0.53 eV, CBO = 2.44 eV), so that orientation gives the strongest electron confinement and the highest simulated I_ON/I_OFF ratio in Schottky diodes.","On (-201) the near-zero valence-band offset and low built-in voltage point to a fast-turn-on diode with higher forward current.","Raising aluminium content mainly raises the conduction band of the alloy, so alloy composition acts mostly on the conduction-band offset and electron confinement rather than on the valence band.","Forward-bias I-V curves built from first-principles offsets reproduce published measurements at 21-22% aluminium, and remaining reverse-bias discrepancies are attributable to an ideal, defect-free junction.","Orientation-dependent built-in voltage shifts the simulated C-V curves, which otherwise behave like a MOS-type deep-depletion capacitor, similar to structures reported to confine a two-dimensional electron gas."],"supporting_citations":[{"why":"Prior first-principles study showing strain changes both the magnitude and type of Ga2O3/Al2O3 offsets; the effect this paper systematizes across orientations and alloys.","marker":"[21]"},{"why":"Reference hybrid-functional band gaps and offsets for Ga2O3 and (AlxGa1-x)2O3 alloys used for bowing and offset comparisons.","marker":"[22]"},{"why":"Experimental XPS band alignment of Al2O3 with (-201) beta-Ga2O3 used to benchmark the computed (-201) valence-band offset.","marker":"[23]"},{"why":"Earlier orientation-dependent band-offset calculations for (AlxGa1-x)2O3/Ga2O3 that this work extends by explicit superlattice potential lineups.","marker":"[27]"},{"why":"Supplies the special quasirandom structure method used to build the disordered alloy supercells at each aluminium concentration.","marker":"[35]"},{"why":"Experimental I-V data for a beta-(Al0.22Ga0.78)2O3/Ga2O3(010) Schottky diode reproduced by the TCAD model in forward bias.","marker":"[37]"},{"why":"Experimental I-V data for an (Al0.21Ga0.79)2O3/Ga2O3(010) Schottky diode used as the second validation target.","marker":"[39]"}],"fun_headline_variants":["Orientation alone swings Ga2O3 band offsets by 2.4 eV","Strain and orientation tune beta-Ga2O3 band offsets","DFT offsets guide realistic Ga2O3 diode simulations","Orientation-dependent band alignment shapes Ga2O3 devices"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"For alloy compositions the interface potential-lineup term is linearly interpolated from three endpoint values rather than recomputed for explicit alloy interfaces at every concentration, so a strongly nonlinear lineup would shift all composition-dependent offsets.","fun_headline_variants_meta":{"raw":{"variants":["Orientation alone swings Ga2O3 band offsets by 2.4 eV","Strain and orientation tune beta-Ga2O3 band offsets","DFT offsets guide realistic Ga2O3 diode simulations","Orientation-dependent band alignment shapes Ga2O3 devices"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000566,"raw_usage":{"total_tokens":2719,"prompt_tokens":1020,"completion_tokens":1699,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":636,"completion_tokens_details":{"reasoning_tokens":1625}},"tokens_in":636,"tokens_out":1699,"duration_ms":16869,"temperature":1.0,"reasoning_tokens":1625,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-07T00:57:31.354473+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Compute explicit disordered (Al0.25Ga0.75)2O3/Ga2O3 superlattices for two orientations and compare CBO/VBO with values linearly interpolated from the endpoints; a deviation beyond numerical noise would falsify the quantitative alloy predictions. Alternatively, measure core-level photoemission band offsets on (100)- and (-201)-oriented films with the same aluminium fraction and check the roughly 0.5 eV-scale orientation split.","supporting_citations":[{"cited_title":"Lyu, Band offsets at the interfaces between β-ga 2 o 3 and al 2 o 3, Physical Review Materials 7, 014603 (2023)","cited_arxiv_id":null,"evidence_quote":"Prior first-principles study showing strain changes both the magnitude and type of Ga2O3/Al2O3 offsets; the effect this paper systematizes across orientations and alloys."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Reference hybrid-functional band gaps and offsets for Ga2O3 and (AlxGa1-x)2O3 alloys used for bowing and offset comparisons."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Experimental XPS band alignment of Al2O3 with (-201) beta-Ga2O3 used to benchmark the computed (-201) valence-band offset."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Earlier orientation-dependent band-offset calculations for (AlxGa1-x)2O3/Ga2O3 that this work extends by explicit superlattice potential lineups."},{"cited_title":"van de Walle, P","cited_arxiv_id":null,"evidence_quote":"Supplies the special quasirandom structure method used to build the disordered alloy supercells at each aluminium concentration."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Experimental I-V data for a beta-(Al0.22Ga0.78)2O3/Ga2O3(010) Schottky diode reproduced by the TCAD model in forward bias."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Experimental I-V data for an (Al0.21Ga0.79)2O3/Ga2O3(010) Schottky diode used as the second validation target."}],"review_version":1}