{"id":"112069ca-0e33-4bca-9b4a-e673f255ebb1","arxiv_id":"2504.15857","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"For nine oxide perovskites, DFT calculations show the AO surface forms more easily than the BO2 surface, but A-site vacancies at the AO surface are 1-2 eV more favorable, suggesting the AO surface degrades to expose a more durable BO2 layer.","lead":"This paper uses computer simulations to compare how easily the two possible surface finishes of nine oxide perovskites form, and what happens when atoms are missing from those surfaces. It finds that the easier-to-form surface is also the one that loses atoms more readily, so the other surface may last longer in applications.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"BO2-prevalence conclusion rests on single-vacancy energetics, but full AO-layer removal requires coupled A+O vacancies; authors concede in SI SIII G this cannot be captured by summing.","rationale":"The paper's computational results — AO surface formation energies lower than BO2, and A-site surface vacancies more favourable than B-site surface vacancies — are plausible and benchmarked against prior DFT literature, including a comparison to machine-learned potentials in SI SII G. The weakness is not in the individual numbers but in the inference from those numbers to the headline conclusion about long-term surface prevalence. To convert an AO-terminated surface into a BO2-terminated one, an entire AO monolayer must be removed, requiring both A and O atoms to leave the layer. The paper compares only isolated single-vacancy formation energies on the two different terminations. The authors themselves note in SI SIII G that simultaneous A- and O-site vacancies cannot be handled by summing isolated vacancy energies, which directly undermines the degradation mechanism they invoke. A proper test would be to compute the full-layer removal energy (or a coupled A+O vacancy pair in the same layer) and compare it with the single-vacancy proxy. Until such a calculation is performed or the conclusion is softened to a statement about relative susceptibility to isolated vacancy formation, the strong wording 'will lead to the BO2 surface being more prevalent' and 'better long term stability' is not supported. This matches the reader's weakest assumption, and the conditional verdict is appropriate: accept pending either a test of the coupled-vacancy pathway or a more cautious framing.","tokens_in":42472,"tokens_out":3960,"duration_ms":43691,"concrete_test":"For at least SrTiO3 and BaTiO3, compute the energy of removing the complete top AO layer from the AO-terminated 2x2 slab (i.e., relax the BO2-terminated slab and compare per surface area). Then compare this full-layer removal energy with the sum of the isolated A-vacancy and O-vacancy formation energies in the same AO layer from Eq. 9/SI Table SVI. If the pair sum deviates from the full-layer removal energy by more than ~0.5 eV, or if the full-layer removal is uphill while isolated vacancies are favourable, the single-vacancy comparison does not establish the AO-to-BO2 degradation pathway.","verdict_should_be":"CONDITIONAL","load_bearing_attack":"The central claim is that the AO (001) surface, though easier to form, degrades through A-site vacancy formation, making the BO2 surface more prevalent long-term (Conclusion, Section IV). The evidence for this is the 1–2 eV lower formation energy of an isolated A-site vacancy at the AO surface compared with an isolated B-site vacancy at the BO2 surface (Section III D, Table SVI). However, exposing a BO2 layer requires removing an entire AO layer — both A and O atoms — not just a single A vacancy. The energy of that coupled removal is not the sum of isolated vacancy energies; the authors state in SI SIII G that considering simultaneous A- and O-site vacancies 'cannot be considered by simply summing the relevant isolated vacancy energies.' Moreover, the clean-surface energetics in Section III A show the AO termination is more favourable than BO2 under the same chemical potentials, so a thermodynamic pathway for AO-to-BO2 conversion is not established by the single-vacancy comparison. The charged-defect analysis is performed only for SrSnO3, so the generalisation to all nine perovskites further extrapolates the degradation argument. Thus the load-bearing assumption — that isolated vacancy energies control full-layer degradation — is explicitly disavowed by the authors themselves and remains untested.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper presents a GGA-PBE DFT study of the (001) surfaces of nine oxide perovskites (ABO3 with A = Ca, Sr, Ba and B = Ti, Zr, Sn), comparing AO- versus BO2-terminated slabs, computing neutral vacancy formation energies as a function of depth, constructing chemical-potential surface phase diagrams, and performing charged-defect calculations for SrSnO3. The central claim is that the AO termination is energetically preferred on the clean surface, but A-site vacancies at the AO surface are 1-2 eV more favorable than B-site vacancies at the BO2 surface, so the AO surface degrades more readily and the BO2 termination may become more prevalent over long times. The study includes a systematic dataset across nine materials, comparison with literature titanate oxygen-vacancy energies and with Weston et al. for SrSnO3, and modern charged-defect correction schemes.","tokens_in":42750,"tokens_out":3876,"duration_ms":37217,"significance":"If the central conclusion were established, it would be practically relevant for perovskite surface manufacturing and for interpreting experimental observations of coexisting terminations. The paper has clear strengths: a consistent first-principles dataset across nine materials, careful treatment of charged defects with eFNV and 2D slab corrections, explicit benchmarking against known literature values, and a useful benchmark of machine-learned potentials against DFT for surface and vacancy energetics (Table SII). However, the central inference from single-vacancy energetics to full-layer degradation is not currently supported by the calculations, and the charged-defect part of the argument is demonstrated for only one of the nine materials. These issues affect the main conclusion and require substantial additional work or reframing.","major_comments":[{"comment":"The conclusion that BO2 surfaces become more prevalent through degradation of AO surfaces rests on comparing the isolated A-site vacancy at the AO surface with the isolated B-site vacancy at the BO2 surface. Exposing a BO2 layer requires removing an entire AO layer, i.e., both A and O atoms, and the authors state in SI SIII G that simultaneous A- and O-site vacancy formation \"cannot be considered by simply summing the relevant isolated vacancy energies.\" Therefore the 1-2 eV single-vacancy comparison in Table SVI does not by itself establish the proposed layer-removal pathway. I ask the authors to compute coupled A+O vacancy formation energies, at least for a subset of the nine materials, or to formulate an explicit thermodynamic model of AO-layer removal, before drawing the long-term stability conclusion.","section":"Section III D, Eq. (9), Table SVI"},{"comment":"The clean-surface and chemical-potential results show that the AO termination is more favorable than BO2 under essentially all conditions considered (with CaTiO3 and CaSnO3 noted as exceptions in the Conclusion). The proposed AO-to-BO2 conversion would have to overcome this thermodynamic preference, yet the manuscript provides no free-energy or kinetic argument for why vacancy formation would reverse the relative stability of the two terminations. At minimum, the conclusion should be reframed as a degradation hypothesis rather than a statement that BO2 \"will display better long term stability\" or \"will be more prevalent than initially expected.\"","section":"Section III A and Conclusion"},{"comment":"Charged-defect calculations are performed only for SrSnO3, both in bulk and in slabs, but the abstract and conclusion generalize the vacancy-stability argument to all nine perovskites, for example \"Charged vacancies only drive this further under oxygen-rich conditions\" and the recommendation for manufacturing BO2 surfaces of \"these materials.\" Since the charged-vacancy stabilization is part of the degradation mechanism, the generalization should either be supported by charged-defect calculations for additional representatives (for instance one titanate and one zirconate) or explicitly flagged as an assumption.","section":"Sections III C and III E"},{"comment":"Vacancy formation energies are reported without error bars or systematic convergence checks, at concentrations of roughly 0.021-0.042 vac/unit in 2x2 slabs, and the bulk vacancies are modeled in 2x2x2 supercells. The claim that surface defect densities are \"several orders of magnitude\" higher than in the bulk is quantitative and would be sensitive to these choices. A short convergence test (for example, a 3x3 surface cell or thicker slabs for one or two materials) would substantially strengthen the long-term stability discussion and is within the scope of the present study.","section":"Section II A and Section III D"}],"minor_comments":[{"comment":"The sentence \"These results indicates that...\" should read \"These results indicate that...\"","section":"Abstract"},{"comment":"The text uses \"VMB\" for the valence band maximum; it should be \"VBM\".","section":"Section III C"},{"comment":"There are typos in this section: \"BO2-termianted\" should be \"BO2-terminated\" and \"exhbits\" should be \"exhibits\".","section":"Section III E"},{"comment":"The sentence \"we can now simply Equation (S3)\" should read \"we can now simplify Equation (S3)\".","section":"SI SII C"},{"comment":"The caption lists \"MPACE-MPA-0\" as a model name; this should be \"MACE-MPA-0\" for consistency with the text.","section":"Table SII caption"},{"comment":"The sentence describing Fig. 2a is grammatically garbled: \"here, the values are given relative to composite binary oxides (AO and BO2) and the bulk perovskite (ABO3) are presented in Fig. 2a.\" It should be rephrased to clearly state that the formation energies are referenced to the binary oxides.","section":"Section III A"}],"recommendation":"major_revision","confidential_remarks":"The paper contains a useful and mostly careful DFT dataset, and the charged-defect implementation is commendable. The main issue is that the headline conclusion about long-term BO2 prevalence is not supported by the single-vacancy calculations, and the authors themselves concede in the SI that coupled vacancy formation cannot be treated by summation. This is fixable either by computing coupled A+O vacancy energies or by substantially softening the conclusion, so I would not reject the manuscript. I would encourage the authors to either add the coupled-vacancy calculations or explicitly present the degradation argument as a hypothesis with clear caveats."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Dear [Colleague],\n\nWorth a look if you care about perovskite (001) surface stability, but the headline claim outruns the evidence. The paper calculates surface formation energies and A/B/O vacancy formation energies as a function of depth for nine alkaline-earth perovskites, plus a charged-defect study for SrSnO3. The systematic cross-family comparison is the real contribution; single-material studies exist, but a consistent set across all nine on one methodology is genuinely useful.\n\nThe methods are solid. VASP with 700 eV cutoff, spin-polarized PBE, fixed-cell slab relaxations, and proper charged-defect corrections using doped with eFNV and qdef2d/sxdefectalign2d. Benchmarks against literature are reasonable: titanate oxygen-vacancy energies land in the known 5–6 eV range, and the SrSnO3 charged-defect transition levels agree decently with Weston et al. The machine-learning-potential comparison in the SI is a side analysis but well done.\n\nThe weak point is the inference from single vacancies to long-term termination stability. The authors find A-site vacancies at the AO surface are 1–2 eV cheaper than B-site vacancies at the BO2 surface, and conclude the AO surface will degrade, making BO2 more prevalent. But removing an AO layer requires removing A and O together, and the authors concede in SI SIII G that simultaneous A- and O-site vacancies cannot be captured by summing isolated vacancy energies. The clean-surface phase diagram also shows AO is thermodynamically preferred, so a conversion pathway is not demonstrated. The charged-defect analysis, which is the most detailed part, is only for SrSnO3; generalizing those conclusions to all nine is an extrapolation. There are also no error bars, and the 2x2 supercell vacancy concentrations are high enough to shift absolute numbers. Those are standard DFT-screening caveats, but they matter when the language is 'will lead to'.\n\nBottom line: the dataset and depth-resolved trends are worth having, and the paper deserves a serious referee. I would tell the authors to soften the degradation conclusion and ideally run a coupled A+O vacancy or an explicit layer-removal calculation to test the pathway. As is, the strong wording is not justified by the evidence.","headline":"Useful systematic DFT map of vacancy energetics across nine perovskites, but the BO2-prevalence claim overreaches what single-vacancy calculations can support.","tokens_in":43239,"tokens_out":3036,"would_cite":false,"duration_ms":28656,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":["68.35.-p","71.15.Mb"],"model":"deepseek-v4-flash","headline":"Across nine oxide perovskites, the $\\mathrm{AO}$ (001) surface is cheaper to form than the $\\mathrm{BO}_2$ surface but far more prone to A-site vacancy formation, so the $\\mathrm{BO}_2$ termination is the one that persists.","keywords":["oxide perovskites","(001) surface termination","vacancy formation energy","surface stability","density functional theory","AO and BO2 terminations","charged vacancies"],"falsifier":"Calculate the coupled removal energy of an A-site atom and a neighbouring oxygen at the $\\mathrm{AO}$ surface in the same $2\\times2$ slab geometry: if removing a whole AO unit costs far more than the isolated A-vacancy energy implies, the predicted selective degradation of the $\\mathrm{AO}$ termination is weaker than claimed. Experimentally, anneal an atomically defined $\\mathrm{AO}$-terminated $\\mathrm{SrTiO}_3$ surface under oxygen-rich conditions and track the termination fraction, since the mechanism predicts $\\mathrm{BO}_2$ coverage should rise as A-site vacancies form.","tokens_in":42294,"feed_emoji":"🕳️","tokens_out":13932,"duration_ms":98751,"temperature":0.7,"pith_summary":"This paper tries to settle which of the two competing (001) surfaces of an oxide perovskite survives in practice, and its answer is that they are not the same surface. For nine alkaline-earth perovskites ($\\mathrm{ABO}_3$ with A = Ca, Sr, Ba and B = Ti, Zr, Sn), density functional theory shows that the $\\mathrm{AO}$ termination is always cheaper to form than the $\\mathrm{BO}_2$ termination, and chemical-potential phase diagrams agree across nearly all conditions. But A-site vacancies at the $\\mathrm{AO}$ surface cost 1–2 eV less than B-site vacancies at the $\\mathrm{BO}_2$ surface, so the $\\mathrm{AO}$ layer sheds its cations readily and the $\\mathrm{BO}_2$ layer beneath is exposed. The significance is practical: the surface that is easiest to manufacture is the one least likely to persist, so real samples and devices should expect $\\mathrm{BO}_2$ terminations to dominate over time.","feed_headline":"Vacancies doom the easy perovskite surface first","feed_subtitle":"A-site vacancies deplete the cheap AO layer, so the BO2 termination is the one that survives.","key_machinery":"The load-bearing quantity is the bulk-referenced vacancy formation energy: the cost of removing an A-, B-, or O-site atom from a given layer of a $2\\times2$ six-layer slab of either termination, measured relative to removing the same atom from the bulk crystal. Referencing to the bulk removes the arbitrary chemical-potential choice, so the depth profile isolates purely surface effects, and it is this quantity that produces the decisive 1–2 eV asymmetry between A-site vacancies at the $\\mathrm{AO}$ surface and B-site vacancies at the $\\mathrm{BO}_2$ surface. Around this core sit the surface Gibbs free energies, built with the standard excess-atom ($\\Gamma$) formalism to map stability against growth conditions, and charged-vacancy transition-level calculations in $\\mathrm{SrSnO}_3$, corrected for image charges, which extend the argument to the Fermi-level dependence.","core_discovery":"For all nine $\\mathrm{ABO}_3$ perovskites studied, density functional theory calculations on two-termination (001) slabs show that the undefected $\\mathrm{AO}$ surface has the lower formation energy, and chemical-potential phase diagrams place $\\mathrm{AO}$ as the stable termination across nearly all growth conditions. Introducing vacancies reverses the ranking: A-site vacancies at the $\\mathrm{AO}$ surface are 1–2 eV cheaper than B-site vacancies at the $\\mathrm{BO}_2$ surface, and vacancy formation energies fall systematically toward the surface relative to bulk. For charged vacancies in $\\mathrm{SrSnO}_3$, the $\\mathrm{AO}$ surface develops negative-formation-energy cation vacancies across most of the band gap under oxygen-rich conditions, shrinking the Fermi-level window in which the $\\mathrm{AO}$ termination is stable. The paper concludes that $\\mathrm{AO}$ surfaces are easy to create but prone to degrade by losing A-site cations, thereby exposing the $\\mathrm{BO}_2$ layer below, so $\\mathrm{BO}_2$ should be the more prevalent termination on real, aged surfaces.","pith_inferences":["The paper's own caveat, that removing A and O together cannot be handled by summing isolated single-vacancy energies, implies the cleanest numerical test of its mechanism: compute the coupled (A + O) removal cost at the $\\mathrm{AO}$ surface, and a much higher coupled cost would slow the predicted AO-to-BO2 conversion considerably.","The mechanism carries a kinetic signature the paper does not develop: on an $\\mathrm{AO}$-terminated sample, $\\mathrm{BO}_2$ coverage should grow with annealing time wherever A-site vacancy formation is favourable, so time- and temperature-resolved termination measurements would test the claim directly.","Because A-site vacancy favourability traces to the ionic, weakly bonded nature of the A-site cage, the pattern that the easy termination is the vacancy-prone one is likely to extend beyond alkaline-earth perovskites to other polar oxide surfaces, where it could serve as a screening rule."],"forward_implications":["Real $\\mathrm{AO}$-terminated samples should convert toward $\\mathrm{BO}_2$ over time as A-site cations vacate the surface, so both terminations seen experimentally may reflect degradation rather than equilibrium formation.","A practical synthesis route follows directly: grow or prepare the $\\mathrm{AO}$ termination first, then treat the surface so its A-site depletion reveals a $\\mathrm{BO}_2$ termination, the strategy the paper explicitly endorses.","For applications requiring a chemically stable surface, such as catalysis, interfaces, and devices, the $\\mathrm{BO}_2$ termination is the safer choice, whereas $\\mathrm{AO}$ is only the easier starting point.","In $\\mathrm{SrSnO}_3$, the $\\mathrm{BO}_2$ surface keeps a wider Fermi-level window free of compensating native vacancies, so doping and electronic applications are more reliable on the $\\mathrm{BO}_2$ termination."],"supporting_citations":[{"why":"supplies the chemical-potential formalism used to build the surface stability phase diagrams","marker":"[25]"},{"why":"the generalized-gradient functional with which all slab and vacancy energies are computed","marker":"[52]"},{"why":"experimental observation of $\\mathrm{BO}_2$-dominated polished $\\mathrm{SrTiO}_3$ surfaces that the degradation argument must explain","marker":"[16]"},{"why":"experimental observation of total $\\mathrm{AO}$ coverage after prolonged annealing, the counter-case motivating the stability comparison","marker":"[19]"},{"why":"earlier first-principles evidence that cation vacancies are more favourable at titanate surfaces than in bulk, generalised here to nine perovskites","marker":"[84]"},{"why":"the reference charged-defect study in perovskite stannates against which the $\\mathrm{SrSnO}_3$ levels are benchmarked","marker":"[86]"},{"why":"the electrostatic finite-size correction method applied to all charged vacancy formation energies","marker":"[72]"}],"fun_headline_variants":["Easy perovskite surface is the one that fails first","The cheap AO surface degrades, leaving BO2 stable","Perovskite's easy surface is its weakest link","Vacancies flip perovskite surface stability"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The argument assumes that single-vacancy formation energies, computed at a fixed high vacancy concentration in $2\\times2$ slabs, control how the $\\mathrm{AO}$ layer actually disappears, even though the paper states that removing an A atom and an O atom together cannot be represented by summing the isolated single-vacancy energies.","fun_headline_variants_meta":{"raw":{"variants":["Easy perovskite surface is the one that fails first","The cheap AO surface degrades, leaving BO2 stable","Perovskite's easy surface is its weakest link","Vacancies flip perovskite surface stability"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000375,"raw_usage":{"total_tokens":2028,"prompt_tokens":1004,"completion_tokens":1024,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":620,"completion_tokens_details":{"reasoning_tokens":965}},"tokens_in":620,"tokens_out":1024,"duration_ms":10833,"temperature":1.0,"reasoning_tokens":965,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-16T11:15:38.523385+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Calculate the coupled removal energy of an A-site atom and a neighbouring oxygen at the $\\mathrm{AO}$ surface in the same $2\\times2$ slab geometry: if removing a whole AO unit costs far more than the isolated A-vacancy energy implies, the predicted selective degradation of the $\\mathrm{AO}$ termination is weaker than claimed. Experimentally, anneal an atomically defined $\\mathrm{AO}$-terminated $\\mathrm{SrTiO}_3$ surface under oxygen-rich conditions and track the termination fraction, since the mechanism predicts $\\mathrm{BO}_2$ coverage should rise as A-site vacancies form.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"supplies the chemical-potential formalism used to build the surface stability phase diagrams"},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"experimental observation of $\\mathrm{BO}_2$-dominated polished $\\mathrm{SrTiO}_3$ surfaces that the degradation argument must explain"},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"experimental observation of total $\\mathrm{AO}$ coverage after prolonged annealing, the counter-case motivating the stability comparison"}],"review_version":1}