{"id":"ff54bcb3-061a-4dd9-9f1a-ca662079a782","arxiv_id":"2608.01501","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":4.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":3,"one_line_summary":"Using the 2018 Tkatchenko-Scheffler van der Waals radii corrects severe overbinding in alkali-containing solids, bringing lattice volumes close to many-body dispersion accuracy.","lead":"This paper benchmarks an updated version of the Tkatchenko-Scheffler dispersion correction, using new van der Waals radii, for alkali-containing materials. It finds the update fixes large structural errors from the original method and approaches the accuracy of a more expensive many-body dispersion method.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Reference volumes for alkali halides are room-temperature data minus a Debye zero-point correction, not true Born-Oppenheimer volumes; a biased or missing thermal correction could shift errors by 1–2% and change the TS_2018/MBD NL ranking.","rationale":"The reader's weakest assumption—that the zero-point-corrected experimental reference volumes may be biased—is essentially the same concern I identify, but I sharpen it: the problem is not just the Debye model's accuracy for the zero-point correction; it is that the alkali-halide reference volumes also omit thermal expansion entirely, because they are built from room-temperature lattice constants. This is a systematic, first-order bias in the reference that shifts all reported errors and can affect cross-material averages. The qualitative conclusion that TS_2018/TS_alkali fix the severe TS_2009 overbinding is robust, as TS_2009's errors (~20%) dwarf any plausible reference shift. However, the specific quantitative claim of 'comparable to MBD NL' depends on the reference definition, and the current definition is not a clean Born-Oppenheimer reference. I therefore partially agree with the reader, and I do not propose a different verdict—CONDITIONAL remains appropriate. The concrete test (using low-temperature experimental data or first-principles zero-point corrections) would resolve whether the comparability claim holds under a more accurate reference.","tokens_in":22683,"tokens_out":15648,"duration_ms":172865,"concrete_test":"Recompute the alkali-halide benchmark using low-temperature (4 K) experimental lattice constants (available in the literature for most of the 19 alkali halides) as reference volumes, instead of room-temperature lattice constants minus the Debye zero-point correction. Recalculate Table 1 error statistics and the ranking among PBE+TS_2018, PBE+TS_alkali, and PBE+MBD NL. If the ordering changes—for instance, if MBD NL is no longer more accurate than TS_2018, or if the average absolute error gap exceeds ~1 percentage point—the conclusion of comparability is not robust. For the perovskites, repeat the volume-error analysis using a zero-point-corrected reference (e.g., subtract the estimated 1.5% contraction, or use experimental data at lower temperature), and check whether PBE+TS_2018 remains competitive on volume as claimed.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The quantitative claim that PBE+TS_2018/TS_alkali are 'comparable' to PBE+MBD NL for alkali halides rests on volume errors relative to experimental references constructed in Appendix B. Table 2 states that alkali-halide lattice constants are room-temperature values, and the only correction applied is the Debye-model zero-point expansion Δv = (9/16)(B1−1)kBΘD/B0. This yields a reference of V_RT − Δv = V_e + V_th(300 K), i.e., it omits the thermal expansion from 0 K to room temperature. For typical alkali halides this omission is 1–2% of the volume, comparable to the ~1 percentage-point difference in average absolute error between PBE+TS_2018 (3.40%) and PBE+MBD NL (2.63%) in Table 1. The Debye/Dugdale-MacDonald estimate of Δv is itself approximate; for high-ΘD materials like LiF, Δv/v0 can exceed 4%, so even a 20–30% error in the model translates into a ~1% reference shift. The perovskite comparison is also affected: volumes are compared to 90–100 K experimental data without zero-point correction, and the paper itself notes a ~1.5% zero-point contraction for a related perovskite. If the true reference volumes are lower than those used, the reported TS_2018 volume errors for CsPbBr3 and CsPbI3 (which are already +3.8% and +4.5%, i.e., less accurate than TS_2009) become even larger, and the 'comparable to MBD NL' statement for volumes does not survive. Thus the central claim's quantitative component is sensitive to how the experimental reference is defined.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper benchmarks the Tkatchenko-Scheffler dispersion method with updated van der Waals radii (TS_2018), a more targeted alkali-only variant (TS_alkali), the original TS_2009, and the nonlocal many-body dispersion method (MBD NL) for structural predictions. The benchmark set comprises five alkali dimers against RPA@PBE reference curves, 45 inorganic solids against experimental volumes, and three CsPbX3 halide perovskites against low-temperature experimental structures. The central claim is that TS_2018 and TS_alkali correct the large alkali-related overbinding of TS_2009, reach accuracy comparable to PBE+MBD NL for alkali halides and binary semiconductors, and give better lattice anisotropies than TS_2009 for CsPbCl3, CsPbBr3, and CsPbI3.","tokens_in":23134,"tokens_out":6592,"duration_ms":58345,"significance":"If the conclusions hold, the paper provides a practical result: a simple pairwise dispersion correction with updated radii can replace a more expensive many-body method for alkali-containing solids, and the TS_alkali variant offers backward compatibility for non-alkali systems. The work is careful in several respects: RPA@PBE dimer references use counterpoise corrections and carefully converged basis sets; solid-state volumes are obtained from equation-of-state fits; and the reference volumes are corrected for zero-point motion. The data and implementations are promised to be publicly available, which strengthens reproducibility. However, the quantitative ranking among methods is sensitive to how the experimental reference volumes are defined, and the paper does not currently provide uncertainty estimates for its headline error statistics.","major_comments":[{"comment":"The reference volumes for alkali halides are not Born-Oppenheimer equilibrium volumes. Table 2 states that lattice constants and elastic moduli are room-temperature data, and Appendix B subtracts only the zero-point volume correction Δv. Thermal expansion from 0 K to 300 K is therefore left in the reference. For LiF the zero-point correction alone is ~4.4% of v0, and thermal expansion adds ~1–2% for typical alkali halides. Since Table 1 separates the methods by less than one percentage point (e.g., 2.63% for MBD NL vs 3.40% for TS_2018), the central claim that TS_2018/TS_alkali are 'comparable' to MBD NL is not robust until the thermal contribution is either subtracted using low-temperature lattice constants or explicitly estimated and propagated.","section":"Appendix B, Tables 2 and 4"},{"comment":"The perovskite volume errors are evaluated against 90–100 K experimental data without any zero-point correction, while the computational volumes are at the Born-Oppenheimer surface. The paper itself notes a ~1.5% zero-point contraction for a related perovskite. For TS_2018 the reported volume overestimates (+3.1%, +3.8%, +4.5% for Cl, Br, I) would increase by roughly that amount if the reference were corrected, making MBD NL's volume advantage larger than shown in Fig. 6c. The anisotropy claims are less affected, but the section should report corrected reference volumes or a sensitivity analysis for the volume comparison.","section":"Cs-containing 3D Metal Halide Perovskites, Table 6"},{"comment":"The aggregate errors ⟨R.E.⟩ and ⟨|R.E.|⟩ are reported without uncertainty estimates. The difference between the best-performing method (PBE+MBD NL, 2.63%) and PBE+TS_2018 (3.40%) for alkali halides is comparable to plausible systematic shifts in the experimental references (thermal expansion, Debye-model approximation). Without standard errors, bootstrap intervals, or per-material scatter, the 'comparable accuracy' conclusion is not quantitatively supported. Please add uncertainty measures and, if possible, a table of per-material errors.","section":"Table 1"}],"minor_comments":[{"comment":"Eq. (6) writes R0_A = C (α_eff_A)^{1/7} using the effective polarizability defined in Eq. (3), but Eq. (7) and the TS_2018 method as described use free-atom polarizabilities. This is confusing; clarify whether the updated radii are free-atom constants or environment-dependent.","section":"Section 2.2, Eq. (6)"},{"comment":"The GitHub URL 'https://github.com/aschankler/ts18benchmark data' contains a space and is not clickable. Fix the link.","section":"Data availability"},{"comment":"Several typographical errors: 'the the' in the abstract, 'compatability' near the TS_alkali definition, and 'calculated calculated' in the caption of Fig. 2. Please proofread.","section":"Throughout"},{"comment":"The statement that 'four methods ... perform similarly' at intermediate distances is based on visual inspection of <0.1 eV differences. A quantitative table of errors at the equilibrium distance or a measure of mean absolute deviation would make the claim more precise.","section":"Section 3.1, Fig. 3"}],"recommendation":"major_revision","confidential_remarks":"The paper's main quantitative claim depends on the construction of experimental reference volumes. The omission of thermal expansion from the 'zero-point-corrected' alkali-halide references is a load-bearing issue, not a cosmetic one, because the method differences in Table 1 are small. The perovskite volume comparison is also affected by the acknowledged zero-point caveat. I recommend major revision, requiring the authors to re-evaluate their statistics with properly corrected references or to explicitly re-frame the benchmark as room-temperature comparisons. The dimer RPA benchmarks and the anisotropy analysis are solid and will remain useful after revision."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The paper is what it says it is: a systematic benchmark of TS_2018 and TS_alkali against TS_2009 and MBD NL, with the alkali problem front and center. The core qualitative result is solid. TS_2009 badly overbinds alkali-containing systems because of underestimated vdW radii, and the 2018 update fixes that. The dimer curves, the 19 alkali halides, and the three perovskites all tell the same story. I believe the central conclusion.\n\nWhat is new here is the benchmark itself, not the methods. TS_2018 and TS_alkali both come from prior work, but the paper gives the first clean side-by-side test on a reasonable set of solids and makes the backwards-compatibility case for TS_alkali explicit. The calculations are careful: counterpoise-corrected RPA references, equation-of-state fits, and a genuine attempt to correct experimental references for zero-point motion. That is real work and it is honestly reported.\n\nThe soft spots are real but not fatal. The aggregate errors in Table 1 have no uncertainty estimates, which matters because the spreads between methods are only a few percent. More importantly, the stress-test note about the reference volumes is correct. The alkali-halide references are room-temperature lattice constants minus a Debye zero-point correction, not Born-Oppenheimer volumes. That means they still contain 1–2% thermal expansion, and the Debye correction itself is approximate. For LiF the zero-point correction alone is over 4%, so the reference could shift by around a percent. That is the same size as the difference between TS_2018 and MBD NL. The perovskite comparison is even less clean: low-temperature data without zero-point correction, and the paper itself notes a ~1.5% correction for a related perovskite. So the 'comparable to MBD NL' claim is reference-sensitive. The qualitative ordering—TS_2009 bad, TS_2018/TS_alkali much better—survives. The precise ranking between TS_2018 and MBD NL does not.\n\nI also want to flag the reproducibility gap: the GitHub link has no commit hash, and the libMBD implementations are 'forthcoming.' For a benchmark paper, that is a meaningful omission, though not a reason to reject.\n\nWho is this for? Anyone doing routine DFT structure prediction on alkali-containing solids or hybrid perovskites. It is a useful engineering paper, not a new-physics paper. It deserves a serious referee and likely publication after the reference-volume issue is either fixed or honestly framed as an uncertainty.\n\nRecommendation: send it to peer review, but ask the authors to add uncertainty estimates and to redo or caveat the comparison against a consistent set of Born-Oppenheimer reference volumes.","headline":"A solid, careful benchmark showing that updated TS radii fix the alkali overbinding; the qualitative conclusion holds, but the quantitative edge over MBD NL depends on reference-volume construction that still contains thermal expansion.","tokens_in":23638,"tokens_out":1305,"would_cite":true,"duration_ms":15002,"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":"A polarizability-based update to van der Waals radii fixes the systematic alkali overbinding in the Tkatchenko-Scheffler dispersion method, matching many-body accuracy for alkali halides and lead halide perovskites.","keywords":["Tkatchenko-Scheffler dispersion","van der Waals radii","alkali halides","halide perovskites","density functional theory","lattice constants","many-body dispersion","zero-point correction"],"falsifier":"An anharmonic phonon calculation of the zero-point equilibrium volumes of CsPbCl3, CsPbBr3, and CsPbI3 would settle the ranking: if the true 0 K references differ from the Appendix B values by more than about 1.5%, the reported ordering among TS 2018, TS alkali, and MBD NL in the perovskite benchmarks changes.","tokens_in":22617,"feed_emoji":"⚛️","tokens_out":14910,"duration_ms":123345,"temperature":0.7,"pith_summary":"The paper seeks to establish that the large overbinding error of the 2009 Tkatchenko-Scheffler pairwise dispersion correction in alkali-containing materials comes from underestimated free-atom van der Waals radii for alkali elements, and that the 2018 revision of those radii fixes it. Using alkali dimers, nineteen alkali halides, twenty-five non-alkali binary semiconductors, and three CsPbX3 perovskites, it shows that PBE+TS 2018 and the narrower PBE+TS alkali variant cut the mean absolute alkali-halide volume error from about 20% to roughly 3-4%, matching the much more expensive many-body dispersion method. For the perovskites, the updated pairwise methods avoid the artificial symmetry breaking and exaggerated lattice anisotropy of TS 2009, although the many-body method still gives the smallest low-temperature volume errors. This matters because the cheap pairwise correction can then be used with confidence for alkali-containing systems such as hybrid halide perovskites.","feed_headline":"Updated vdW radii cut alkali crystal errors from ~20% to ~3%","feed_subtitle":"A simple 2018 radius revision makes the cheap pairwise dispersion method match many-body accuracy for alkali solids.","key_machinery":"The load-bearing object is the polarizability-based van der Waals radius relation $R_0^A = C(\\alpha_A^{\\mathrm{eff}})^{1/7}$ ($C=2.54$ a.u.), introduced in the 2018 revision and paired with a revised heteronuclear distance $R_{AB}^0 = 2C((\\alpha_A+\\alpha_B)/2)^{1/7}$. It replaces the density-contour radii of the 2009 method. Because alkali atoms have large polarizabilities, the relation substantially increases their radii, which controls the onset of the damping function in the pairwise $E_{\\mathrm{vdW}} = -\\frac12\\sum f_{\\mathrm{damp}} C_6/R^6$ energy; the later onset removes the spurious extra binding. The Hirshfeld-volume scaling of $C_6$ and $\\alpha$ is unchanged, so the whole correction","core_discovery":"The central claim is that TS 2009's alkali overbinding stems from underestimated free-atom van der Waals radii, and that replacing them with polarizability-derived radii removes the error. The 2018 revision uses $R_0^A = C(\\alpha_A^{\\mathrm{eff}})^{1/7}$ with $C=2.54$ a.u., giving much larger radii for alkali atoms and a later damping onset for the $-C_6/R^6$ term. The paper shows that PBE+TS 2018 and PBE+TS alkali match RPA@PBE dimer curves within about 0.1 eV, cut alkali-halide volume error from 20% to about 3.4-4.0% (close to PBE+MBD NL's 2.6%), and preserve correct Pnma symmetry and better $b/a$ and $\\beta$ anisotropies in CsPbCl3, CsPbBr3, and CsPbI3, where TS 2009 artificially breaks s","pith_inferences":["Because the 2018 revision also changes radii substantially for lanthanides and actinides, the same fix may remove analogous overbinding in compounds of those elements; the paper sets those aside because of multireference complexity.","Other pairwise dispersion corrections that use element-dependent radii could harbor the same alkali error; transferring the polarizability-based radius update to those schemes is a natural test.","The better room-temperature lattice parameters reported for TS 2018 and TS alkali, despite being Born-Oppenheimer results, suggest error cancellation with missing thermal expansion; explicit finite-temperature phonon calculations could separate the two."],"forward_implications":["For alkali halides, PBE+TS 2018 and PBE+TS alkali lower the mean absolute cell-volume error from about 20% with TS 2009 to about 3.4-4.0%, closely matching PBE+MBD NL's 2.6%.","For non-alkali binary semiconductors, TS 2018 keeps errors around 3.0%, essentially unchanged from TS 2009's 2.3%, so the alkali fix does not hurt other materials.","For CsPbCl3, CsPbBr3, and CsPbI3, the updated methods remove TS 2009's artificial symmetry lowering and exaggerated b/a and beta distortions; MBD NL still wins on low-temperature volume but overshoots the beta angle.","TS alkali is exactly TS 2009 in alkali-free systems, so existing TS 2009-based workflows can be extended to alkali-containing systems without changing results elsewhere.","The cheap pairwise TS method becomes a practical option for hybrid organic-inorganic perovskites and other alkali-containing insulators."],"supporting_citations":[{"why":"Original TS pairwise dispersion scheme; provides the Hirshfeld-volume scaling and the alkali-overbinding baseline being corrected.","marker":"22"},{"why":"2018 update defining vdW radii from polarizability and revised heteronuclear distances; the method under test.","marker":"23"},{"why":"MBD NL nonlocal many-body dispersion method used as the higher-level accuracy reference.","marker":"28"},{"why":"PBE functional paired with all dispersion corrections in the benchmarked combinations.","marker":"67"},{"why":"Source of the free-atom C6 coefficients and polarizabilities used in the TS scaling.","marker":"72"},{"why":"Tabulated free-atom radii and reference values for both TS variants, including the alkali radius changes.","marker":"74"},{"why":"Earlier report of CsPbBr3 structural distortion and source of the alkali-only TS alkali construction.","marker":"11"},{"why":"Provides the equation-of-state framework for the zero-point volume correction applied to experimental references.","marker":"99"},{"why":"Derives the zero-point phonon correction procedure used in Appendix B for lattice-constant benchmarks.","marker":"100"},{"why":"Gives the Gruneisen-parameter approximation used in the zero-point expansion correction formula.","marker":"101"}],"fun_headline_variants":["New vdW radii cut alkali crystal errors from 20% to 3%","Simple radius fix rescues TS dispersion for alkali solids","Alkali overbinding solved by updated van der Waals radii","TS_2018: alkali errors drop to near many-body accuracy","Updated radii: 7x better alkali crystal structures"],"cache_read_input_tokens":2816,"weakest_assumption_plain":"The benchmark conclusions rest on the zero-point-corrected experimental reference volumes used in Appendix B; if that phonon-based correction is biased for any material class, the reported volume errors could shift by about 1.5%, enough to reorder the perovskite results.","fun_headline_variants_meta":{"raw":{"variants":["New vdW radii cut alkali crystal errors from 20% to 3%","Simple radius fix rescues TS dispersion for alkali solids","Alkali overbinding solved by updated van der Waals radii","TS_2018: alkali errors drop to near many-body accuracy","Updated radii: 7x better alkali crystal structures"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000208,"raw_usage":{"total_tokens":1350,"prompt_tokens":964,"completion_tokens":386,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":708,"completion_tokens_details":{"reasoning_tokens":297}},"tokens_in":708,"tokens_out":386,"duration_ms":4597,"temperature":1.0,"reasoning_tokens":297,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-06T00:04:27.745056+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"An anharmonic phonon calculation of the zero-point equilibrium volumes of CsPbCl3, CsPbBr3, and CsPbI3 would settle the ranking: if the true 0 K references differ from the Appendix B values by more than about 1.5%, the reported ordering among TS 2018, TS alkali, and MBD NL in the perovskite benchmarks changes.","supporting_citations":[{"cited_title":"P.; Burke, K.; Ernzerhof, M","cited_arxiv_id":null,"evidence_quote":"PBE functional paired with all dispersion corrections in the benchmarked combinations."},{"cited_title":"Linear response time-dependent density functional theory for van der Waals coefficients","cited_arxiv_id":null,"evidence_quote":"Source of the free-atom C6 coefficients and polarizabilities used in the TS scaling."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Tabulated free-atom radii and reference values for both TS variants, including the alkali radius changes."},{"cited_title":"B.; Perdew, J","cited_arxiv_id":null,"evidence_quote":"Provides the equation-of-state framework for the zero-point volume correction applied to experimental references."},{"cited_title":"I.; Philipsen, P","cited_arxiv_id":null,"evidence_quote":"Derives the zero-point phonon correction procedure used in Appendix B for lattice-constant benchmarks."},{"cited_title":"S.; MacDonald, D","cited_arxiv_id":null,"evidence_quote":"Gives the Gruneisen-parameter approximation used in the zero-point expansion correction formula."}],"review_version":1}