{"id":"6475283d-83ab-4c15-8e67-d14d52ef7dc0","arxiv_id":"1908.03497","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"low","formal_verification":"none","parameter_count":0,"one_line_summary":"The Matsui-Akaogi and Luan-Huynh-Zhou empirical potentials of TiO2 yield significantly different pressure-temperature phase diagrams, and neither reproduces the experimental phase behaviour of titanium dioxide.","lead":"Using free-energy calculations, this paper maps out which crystal forms of titanium dioxide are stable under pressure and temperature for two widely used computer models, and shows the two models disagree strongly. It matters because these empirical potentials are common in simulations of nanoparticles, surfaces, and biomolecules, where the choice of model can change the conclusions.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The classical-nucleus caveat in the Conclusions is load-bearing for the unqualified experimental-consistency claim; a zero-point energy estimate is needed to see whether low-temperature phase regions in Fig. 4 survive.","rationale":"The reader's weakest-assumption field identifies the same issue, and the manuscript itself flags it prominently, so this is not a hidden flaw. I nevertheless treat it as the most load-bearing element because the abstract's experimental-consistency sentence is unqualified, and the low-temperature region is the area the author explicitly disclaims. If the zero-point check shows large shifts, the appropriate fix is not rejection of the paper but a narrowed claim: the two classical potentials have different phase diagrams, and the high-temperature inconsistencies with experiment are clear, while low-temperature agreement cannot be assessed classically. The reader's CONDITIONAL verdict already captures this need for qualification. The more specific condition I would attach is: quantify zero-point/quantum corrections at representative state points before using the low-temperature boundaries in the experimental comparison. The reported internal cross-checks (multiple Frenkel–Ladd pressures, Gibbs–Duhem integration, spread <0.01 kBT) give me confidence in the classical free-energy calculations themselves, so no deeper methodological objection emerged.","tokens_in":20837,"tokens_out":11774,"duration_ms":130472,"concrete_test":"Compute the harmonic phonon density of states and the resulting quasiharmonic Helmholtz free energy for rutile, columbite, pyrite, OI, and cotunnite at the equilibrium volumes of both potentials (Table II), and compare the zero-point-energy contribution at 0 K and the free-energy differences at 300 K and 700 K with the reported 0.01 kBT/particle precision. If the zero-point differences among the low-lying phases are larger than this precision and alter the low-temperature phase ordering in Fig. 4, the abstract's experimental-consistency statement should be restricted to temperatures where classical nuclei are adequate.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The paper's Conclusions state: 'we have used free-energy calculations and classical statistical mechanics without any (nuclear) quantum corrections to compute phase diagrams. This almost certainly means that the low-temperature behaviour that we see is not correct.' Yet the abstract's third sentence asserts, without this caveat, that 'neither potential results in phase behaviour that is fully consistent with the available experimental evidence.' The phase diagrams in Fig. 4 extend to 0 K, and some of the evidence used for the experimental comparison, including the rutile–columbite transition and the pyrite stability region, is at low temperature and high pressure. The reported free-energy precision is better than 0.01 kBT per particle (Section IV), about 0.26 meV/atom at 300 K. Zero-point energy differences between TiO2 polymorphs can be of the order of 1–10 meV per atom and vary from phase to phase, so a quasiharmonic or path-integral treatment could reorder the low-temperature phases and shift coexistence pressures by several GPa. The large MA-vs-LHZ difference, visible at 700 K in Fig. 3 (e.g., LHZ has no columbite stability; MA has a wide pyrite region), is unlikely to be reversed by NQE, so the central comparative message survives. What is not settled is the unqualified claim about consistency with experiment, because the low-temperature regime where the author says the classical result is wrong is exactly part of the comparison.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The manuscript presents pressure–temperature phase diagrams for two empirical TiO2 potentials: the widely used Matsui–Akaogi (MA) Buckingham/Coulomb potential and the Luan–Huynh–Zhou (LHZ) Lennard-Jones reparameterisation, which is taken as representative of a family of LJ-based variants. Phase stabilities are computed with Einstein-crystal/Frenkel–Ladd free-energy calculations at multiple pressures, supplemented by thermodynamic integration along isotherms and isobars and by Gibbs–Duhem integration used as a consistency check. The central results are that both potentials support many experimentally known TiO2 polymorphs, that the stable phase regions differ substantially between the two potentials, and that neither potential reproduces the experimental phase behaviour. The paper also compares densities and elastic constants of rutile between the two models and against experiment.","tokens_in":21080,"tokens_out":7787,"duration_ms":78891,"significance":"If correct, this is a valuable and rigorous comparison of two widely used empirical models: it shows that a seemingly minor change in the repulsive part of the potential leads to large changes in the predicted phase diagram, and it provides a benchmark set of free-energy data that is independent of any fitting to the experimental phase behaviour. The calculations are internally consistent (chemical-potential spreads below 0.01 kBT per particle, multiple cross-checks), the supporting data are publicly archived, and the paper makes no ad-hoc parameter adjustments. The main caveat is that the calculations are classical, and the author explicitly states that the low-temperature behaviour is almost certainly not correct; this limits the strength of the paper's conclusions about agreement with experiment, especially at low temperature and high pressure.","major_comments":[{"comment":"The abstract's third sentence asserts that 'neither potential results in phase behaviour that is fully consistent with the available experimental evidence,' but Section VI states that the phase diagrams were computed using 'classical statistical mechanics without any (nuclear) quantum corrections' and that 'This almost certainly means that the low-temperature behaviour that we see is not correct.' The phase diagrams in Fig. 4 extend to low temperature, and Section V's comparison with experiment includes the rutile–columbite boundary (Refs. 67 and 68), the MA columbite stability that exists only below about 20 K, and the columbite–baddeleyite transition (Refs. 69 and 70), all in the low-temperature/high-pressure regime that the author disclaims. Since the reported precision (better than 0.01 kBT per particle, Section IV) is far smaller than typical zero-point energy differences between TiO2 polymorphs, which can be on the order of 1–10 meV per atom, a quantitative estimate of nuclear quantum effects (e.g., quasiharmonic or path-integral calculations) is needed to assess which low-temperature regions of Fig. 4 survive. Without such an estimate, the unqualified experimental-consistency claim is not supported; the abstract and conclusions should be rephrased to restrict the claim to the temperature range in which the classical treatment is reliable, or supplemented with a quantum-correction analysis.","section":"Abstract and Section VI"},{"comment":"The statement that 'there may of course be other phases which we have not considered because they have not (yet) been reported for TiO2, and some of them may well have a lower free energy still for these empirical potentials' directly limits the conclusion that 'neither potential results in phase behaviour that is fully consistent with the available experimental evidence.' The phase diagrams in Fig. 4 are minima only over the enumerated set of polymorphs, and a phase outside that set could in principle be the global minimum for one of the potentials, altering the comparison with experiment. The manuscript's own text appropriately notes that 'we can only make comparisons between phases of which we are aware and not absolute predictions,' but this qualification is absent from the abstract. The abstract should be amended to say, for example, 'among the polymorphs considered, neither potential results in phase behaviour that is fully consistent with the experimental evidence.' I regard this as a matter of precision and scope in the central claim rather than as an error in the free-energy methodology.","section":"Section V"}],"minor_comments":[{"comment":"The paper uses the large stability region of the pyrite phase as evidence that the empirical potentials are poor for bulk phase behaviour, but the manuscript itself notes that pyrite 'has not been reported experimentally.' Absence of experimental observation is not strong evidence of inconsistency, since the phase might simply be kinetically inaccessible; the wording should distinguish 'not observed' from 'contradicted by experiment.'","section":"Abstract and Section V"},{"comment":"In the caption to Fig. 1, the space-group label 'C/2c' should read 'C2/c' to be consistent with the text and standard notation.","section":"Figure 1 caption"},{"comment":"The note to Table III says that values in brackets refer to the lowest temperature reported (298 K), but brackets appear for C33, C44 and C23; the text should explicitly list which entries are measured at 298 K rather than 4 K.","section":"Table III"},{"comment":"The description of truncation is asymmetric: the Buckingham part is truncated at 12 Å with no mention of a smoothing function, while the Lennard-Jones part uses a smoothing function between 10 and 12 Å; a sentence clarifying whether the Buckingham interaction is smoothed or abruptly truncated would remove ambiguity.","section":"Section IV"}],"recommendation":"major_revision","confidential_remarks":"The methodology is careful and the comparative result is likely to be of interest to the statistical-mechanics and empirical-potential communities. The main issue is that the central claim about experimental consistency is stated without the limitations that the author himself acknowledges in Section VI; the revision should either add a zero-point-energy estimate or qualify the claim. I see no indication of circular fitting or unsupported parametrization; the potentials are taken as fixed inputs from the literature."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The thing to know about this paper is that it is a genuinely solid piece of computational thermodynamics, and its main result is a useful negative one: the widely used Matsui–Akaogi potential and its Lennard-Jones reparameterisation by Luan–Huynh–Zhou produce equilibrium phase diagrams that are different from each other and neither matches the experimental picture. That is worth knowing before you trust either potential for titania phase studies.\n\nWhat is actually new is the finite-temperature phase diagram comparison itself. The Frenkel–Ladd Einstein crystal method is standard, but applying it systematically to these specific empirical potentials, with independent free-energy calculations at several pressures per phase, is a real piece of work. The internal cross-checks are impressive: thermodynamic integration along isotherms and isobars, Gibbs–Duhem integration as a check, and finite-size corrections estimated and shown to be small. The author is also honest about the limits of the method, which I respect.\n\nThe soft spots are proportionate. There are no propagated error bars on the coexistence lines; the author says small free-energy errors can shift coexistence pressures, and the reader's request to quantify that sensitivity is reasonable. The more substantive issue, flagged by the stress-test note, is the classical-nucleus treatment. The Conclusions explicitly say the low-temperature behaviour is \"almost certainly\" not correct, and the abstract's statement that neither potential is \"fully consistent\" with experiment would be better with that caveat attached. Some of the experimental comparisons used, like the rutile–columbite boundary, sit at low temperature where zero-point motion could matter. That said, the MA-vs-LHZ difference is large and visible at 700 K, so the comparative message is not in danger. The acknowledged possibility of undiscovered lower-free-energy polymorphs is a standard limitation, not a flaw.\n\nWho is this for? Anyone using empirical pair potentials for TiO2, and anyone wanting a good worked example of how to compute solid–solid phase boundaries by free-energy methods. It deserves a serious referee. I would recommend acceptance after minor revision, mainly asking for the coexistence-line sensitivity analysis and a softer abstract claim.","headline":"A careful, well-cross-checked free-energy study showing that two closely related TiO2 empirical potentials give different phase diagrams, with a real but acknowledged classical-nucleus caveat.","tokens_in":21597,"tokens_out":1629,"would_cite":true,"duration_ms":19248,"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":"Two TiO2 force fields each support many observed crystal forms, but their phase diagrams differ sharply and neither reproduces experiment.","keywords":["titanium dioxide","TiO2 polymorphs","empirical force fields","Coulomb-Buckingham potential","Lennard-Jones potential","free-energy calculations","phase diagrams","Einstein crystal method"],"falsifier":"Hold fine-grained TiO2 in a hydrostatic high-pressure cell at 700 K for long times at pressures from 5 to 60 GPa and identify the equilibrium phase by X-ray diffraction; if the observed stable-phase sequence matches either potential's predicted sequence, such as the original model's rutile–columbite–pyrite path or the reparameterised model's absence of any columbite stability, the claim that neither potential matches experiment would be directly contradicted.","tokens_in":20606,"feed_emoji":"💎","tokens_out":16188,"duration_ms":158027,"temperature":0.7,"pith_summary":"Titanium dioxide is usually simulated with one of two closely related empirical models: the original Coulomb-plus-Buckingham model, whose short-range term is an exponential repulsion plus $r^{-6}$ attraction, and a later reparameterisation that replaces that term with a 12–6 Lennard-Jones potential. This paper computes absolute free energies for twelve TiO2 polymorphs for both models and constructs their pressure–temperature phase diagrams. It establishes that the two models support nearly all experimentally observed polymorphs as metastable structures, but the thermodynamically stable regions are very different: the reparameterised model gives the OI phase a large stability field and no stable columbite phase, while the original model stabilises columbite only below about 20 K. It further establishes that neither model's equilibrium phase behaviour matches the available experimental evidence, with a broad predicted pyrite stability region that has never been seen experimentally. The practical upshot is that replacing one repulsive term in a composite potential, without considering where the full potential's minimum lies, changes density by 10–15% and qualitatively changes phase behaviour.","feed_headline":"Two TiO2 force fields disagree; neither matches experiment","feed_subtitle":"A switch in the repulsive part of the potential changes which TiO2 crystals are stable and shifts densities by up to 15%.","key_machinery":"The work is carried by the Einstein crystal method of free-energy calculation: each atom is harmonically tethered to its lattice site, the free energy of the resulting Einstein crystal is known analytically, and Hamiltonian thermodynamic integration switches on the real potential and then removes the springs, giving an absolute Helmholtz free energy for each polymorph. Chemical potentials $\\mu = G/N$ are then integrated along isotherms and isobars using the Gibbs–Duhem relation, with independent free-energy calculations at multiple pressures used as consistency checks and the Clapeyron equation used to trace coexistence lines. This machinery matters because enthalpy alone is not enough to order the phases: for the original potential, the OI analogue is enthalpically preferred up to 1600 K but is stabilised by entropy only below about 500 K. The object being compared is the full composite potential, in which the Coulomb attraction pulls opposite charges close enough that the steepness of the Lennard-Jones repulsion, rather than the shape near its own minimum, controls the equilibrium structure.","core_discovery":"The central discovery is that the two potentials are not interchangeable. When the Coulomb charges are kept identical and only the short-range repulsion is changed from a Buckingham form to a Lennard-Jones form, the minimum of the combined Ti–O potential moves into the steep $r^{-12}$ repulsive wall, so the solid phases of the reparameterised model are 10–15% less dense and are stabilised in a different order. Direct free-energy calculations show that the original MA potential makes rutile, columbite, pyrite, OI, and cotunnite the stable phases in different regions of the $P$–$T$ plane, whereas the LHZ potential never makes columbite thermodynamically stable and gives OI a much larger stability field; the two phase diagrams differ sharply. Both models can represent nearly all known TiO2 polymorphs as metastable or stable states, but the equilibrium diagrams are inconsistent with experiment: pyrite, which has only been considered theoretically, has a large stability region in both models, and neither model reproduces the experimental rutile–columbite or columbite–baddeleyite coexistence behaviour. The paper concludes that neither potential is fully consistent with available experimental evidence, and that potential parameterisation must be judged against the full composite potential, not against a single term in isolation.","pith_inferences":["The same free-energy pipeline could serve as a routine screening test for new TiO2 force fields: two models that match densities and elastic constants can still order the polymorphs differently, so coexistence calculations would catch errors that property matching misses.","The search space is not closed: the paper lists only polymorphs already reported for TiO2, and an unreported structure could be the true free-energy minimum for these potentials. A crystal-structure search coupled to these free-energy calculations could reveal such phases.","The sign of the OI–cotunnite coexistence slope differs between the empirical potentials and quasi-harmonic DFT, which suggests the two approaches disagree strongly on the entropy of the OI analogue; computing the vibrational density of states of the simulated OI structure would show whether the structural distortion seen in simulation is responsible.","The same comparison would be worth applying to machine-learned potentials, which are usually fitted to energies and forces rather than to free-energy differences; a learned potential could still misorder phase stability unless coexistence calculations are part of its validation."],"forward_implications":["Simulations run with the LJ-based model cannot be treated as equivalent to those run with the original model: at the same nominal state point the crystals differ in density by roughly 10–15% and the stable polymorph can be different.","Neither empirical model is a reliable predictor of TiO2 equilibrium phase behaviour; in particular, a large predicted pyrite stability region is suspect because pyrite has not been observed experimentally.","Entropy can reverse enthalpy-based stability rankings, so phase diagrams for empirical potentials need explicit free-energy calculations rather than energy minimisation alone.","Because the calculations are classical, the low-temperature parts of the reported phase diagrams are expected to be wrong; conclusions about which phases are stable near room temperature should be treated with caution.","Reparameterising a potential by fitting one term to another is not enough: a Lennard-Jones fit with the Ti–O minimum placed closer to the original yields polymorph densities within about 3%, showing the failure is in where the minimum sits, not in the Lennard-Jones form itself."],"supporting_citations":[{"why":"Supplies the original Coulomb-plus-Buckingham potential whose phase diagram is computed.","marker":"[1]"},{"why":"Supplies the representative Lennard-Jones reparameterisation whose phase diagram is compared with the original.","marker":"[4]"},{"why":"Introduces the Einstein crystal thermodynamic-integration method used to obtain absolute solid free energies.","marker":"[41]"},{"why":"Provides the accepted protocol for free-energy-based phase diagram calculation, including finite-size corrections and consistency checks.","marker":"[42]"},{"why":"Gives a DFT-based phase diagram of TiO2 against which the empirical results are compared.","marker":"[39]"},{"why":"Reports the experimental rutile–columbite coexistence boundary used to benchmark the empirical models.","marker":"[67]"},{"why":"Reports the experimental columbite–baddeleyite coexistence boundary used to test the high-pressure behaviour.","marker":"[69]"},{"why":"Predicts stability of the Pca21 phase from quasi-harmonic DFT, used to assess why the original potential lacks it.","marker":"[36]"},{"why":"Shows that similar empirical water models have very different phase diagrams, supporting the paper's interpretation of the TiO2 results.","marker":"[66]"}],"fun_headline_variants":["TiO2 potentials: same charges, different phases","Buckingham vs Lennard-Jones: TiO2 stability flips","Neither TiO2 force field matches real phase data","Repulsion term change redraws TiO2 phase diagram","TiO2 models diverge: no winner against experiment"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The whole comparison rests on treating the nuclei as classical particles and considering only the already-known crystal structures, even though the author notes that omitting nuclear quantum effects almost certainly makes the low-temperature behaviour wrong.","fun_headline_variants_meta":{"raw":{"variants":["TiO2 potentials: same charges, different phases","Buckingham vs Lennard-Jones: TiO2 stability flips","Neither TiO2 force field matches real phase data","Repulsion term change redraws TiO2 phase diagram","TiO2 models diverge: no winner against experiment"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000409,"raw_usage":{"total_tokens":2128,"prompt_tokens":958,"completion_tokens":1170,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":574,"completion_tokens_details":{"reasoning_tokens":1093}},"tokens_in":574,"tokens_out":1170,"duration_ms":9522,"temperature":1.0,"reasoning_tokens":1093,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T14:11:08.948794+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Hold fine-grained TiO2 in a hydrostatic high-pressure cell at 700 K for long times at pressures from 5 to 60 GPa and identify the equilibrium phase by X-ray diffraction; if the observed stable-phase sequence matches either potential's predicted sequence, such as the original model's rutile–columbite–pyrite path or the reparameterised model's absence of any columbite stability, the claim that neither potential matches experiment would be directly contradicted.","supporting_citations":[{"cited_title":"Luan , author T","cited_arxiv_id":null,"evidence_quote":"Supplies the representative Lennard-Jones reparameterisation whose phase diagram is compared with the original."},{"cited_title":"Vega , author E","cited_arxiv_id":null,"evidence_quote":"Provides the accepted protocol for free-energy-based phase diagram calculation, including finite-size corrections and consistency checks."},{"cited_title":"Vega , author J","cited_arxiv_id":null,"evidence_quote":"Reports the experimental rutile–columbite coexistence boundary used to benchmark the empirical models."},{"cited_title":"Akimoto , author Y","cited_arxiv_id":null,"evidence_quote":"Shows that similar empirical water models have very different phase diagrams, supporting the paper's interpretation of the TiO2 results."}],"review_version":1}