Pith. sign in

REVIEW 2 major objections 7 minor 11 references

A certified refinement and asymptotic analysis of the Kuznetsov-Sahinidis diameter bound for Lennard-Jones clusters

T0 review · 2 major / 7 minor · reviewed 2026-07-13 · grok-4.5

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read A certified one-layer tightening of the Kuznetsov–Sahinidis diameter bound for 92 Lennard–Jones sizes, plus the asymptotic form N − Θ(√N).

desk verdict Solid, self-contained theoretical note: certified one-layer tightening at 92 sizes plus a clean asymptotic resolution of an open KS question, with no solver impact claimed or shown. read the letter →

arxiv 2607.09555 v1 pith:Y3OMQW5Q submitted 2026-07-10 math.OC math-phmath.MP

Guillaume Lecomte This is my paper · ORCID
classification math.OCmath-phmath.MP MSC 90C2649M3765K05
keywords Lennard-Jonesclustersdiameterboundlayerglobaloptimizationrearrangementinequalityasymptoticanalysiscertifiedrefinement
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

Optimal Lennard–Jones clusters are hard to certify because the search box is large; an a priori diameter bound shrinks that box. Kuznetsov and Sahinidis obtained such a bound by slicing space into unit-width layers and charging every layer the loosest possible internal energy. This paper replaces that loose charge with a certified estimate built from the proven five- and six-atom minima and a subset inequality already present in their work, then minimizes the resulting energy lower bound over all population profiles. An arrangement-free relaxation that uses only the classical rearrangement inequality for sequences certifies the minimum over every profile, and every comparison is re-checked in directed-rounding arithmetic. The result is a rigorous one-layer improvement of the published bound at 92 sizes between 5 and 200. Independently, the original layer program is shown to satisfy N − ρ_KS = Θ(√N), settling an asymptotic question left open by the earlier paper, while the refined program gains an additional Θ(√N) layers with an explicit constant. The tightening does not yet resolve any open global-optimization case and is presented as a theoretical note: on the only sizes a solver can finish, the diameter box is not the binding resource.

What carries the argument

The arrangement-free lower bound of Lemma 2: after writing the layer energy as background plus surplus–base field plus surplus–surplus binding plus intra-layer terms, the central field is minimal, sorted distances of contiguous sites minimize the kernel, and the rearrangement inequality pairs the sorted products with those distances, closing a rigorous lower bound over every surplus multiset and every placement without enumerating arrangements.

What would settle it

Recompute the arrangement-free lower bound of Lemma 2 at D = ρ_KS(N) for any of the 92 certified sizes and check whether its directed-rounding lower endpoint still strictly exceeds the upper endpoint of V_put(N); a single size where the inequality fails or equals would falsify the claimed one-layer improvement.

Watch

Extended reading notes

Core claim

For each of 92 sizes N in 5 ≤ N ≤ 200, every N-atom configuration whose atoms occupy ρ_KS(N) consecutive nonempty layers has energy strictly greater than V_put(N) ≥ V*_N. By the structural property that a global minimizer occupies a block of consecutive nonempty layers, no global minimizer spans ρ_KS(N) layers, so the diameter bound improves to diam < ρ_KS(N) − 1. Independently, the original KS layer program satisfies N − ρ_KS(N) = Θ(√N), and the refined program gains an additional θ √(2(u_N − σ_0)N) + O(1) layers with explicit θ ≈ 0.026433.

Load-bearing premise

The structural claim that every global minimizer occupies a solid block of consecutive nonempty layers; without it the energy certificates for fully occupied profiles still hold, but they no longer imply a diameter bound on the minimizers themselves.

Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

2 major / 7 minor

Summary. The paper refines the Kuznetsov–Sahinidis (KS) a priori diameter bound for optimal Lennard–Jones clusters by replacing the trivial per-layer floor −binom(n,2) with a certified estimate f built from the KS subset inequality and the proven minima V*_5, V*_6. The resulting layer energy lower bound is minimized over population profiles; centred decreasing candidates supply numerical thresholds, while an arrangement-free lower bound (Lemma 2), relying only on the rearrangement inequality and elementary field/distance estimates, certifies every comparison over all profiles and placements. Directed-rounding arithmetic closes 92 one-layer improvements of ρ_KS for 5≤N≤200 (Theorem 1). Independently, the original KS layer program is shown to satisfy N−ρ_KS(N)=Θ(√N) (Theorem 2), resolving an open point in [1], and the refined program is proved to gain an additional θ√(2(u_N−σ_0)N)+O(1) layers with explicit θ≈0.026433 (Corollary 2). The authors report a negative downstream test on N≤6 (diameter box not binding) and present the work as a theoretical note that certifies no new global minimum.

Significance. If correct, the contribution is a rigorous, fully certified tightening of a published geometric bound together with a clean asymptotic analysis that replaces an empirical linear fit by N−Θ(√N) and quantifies the refined gain with an exact constant. Strengths that should be credited explicitly include: (i) the arrangement-free closure of Lemma 2, which makes the certificates independent of rearrangement completeness; (ii) directed-rounding re-verification with outward padding of V*_5, V*_6 and V_put, including an explicit worst margin (0.002708 at N=38); (iii) a self-contained reproducibility package that regenerates all 92 certificates and the asymptotic identities; and (iv) the authors’ transparent negative solver experiment and refusal to overclaim practical impact. For a global-optimization journal this is a legitimate theoretical note: it improves a tool used in deterministic LJ work and settles a structural question left open by KS, even though it does not advance the certified-N frontier.

major comments (2)
  1. Theorem 1 (and the last paragraph of its proof) uses the KS structural property that every global minimizer occupies a block of consecutive nonempty layers. The energy claim for fully occupied spans is unconditional and rests only on (5), (7), Lemma 2 and directed rounding; the passage to a diameter statement about minimizers, however, depends entirely on that external premise. The manuscript flags the dependence clearly, but for self-containment a short sketch of the argument in [1] (or an explicit theorem/lemma number) should be added in Section 3 so that a reader can verify the only load-bearing external structural input without retrieving the source paper.
  2. Section 6 and Theorem 1: the certificates are closed by showing that the arrangement-free bound (10) exceeds V_put at D=ρ_KS for every surplus multiset. The text states that a branch-and-bound over all p(e) partitions was used and that 92/92 close, with worst directed margin 0.002708 at N=38. Because the margin at N=38 is three orders of magnitude smaller than typical margins and rests on three copies of V*_5, the supplementary material (or a short appendix table) should list, for that single borderline size, the numerical lower endpoint of (10), the padded V_put, and the three-layer profile that realises the candidate minimum, so that the tightest certificate can be inspected without re-running the package.
minor comments (7)
  1. Section 3, display (4): the switch from the original KS kernel V_LJ(d) to the conservative V_LJ(d−1) is correctly described as a self-handicap, and the text asserts that integer ρ_KS is unchanged on 5≤N≤200. A one-line pointer to the script or table that records this identity would make the claim immediately checkable.
  2. Lemma 3 / Theorem 2: the closed form (12) and the prediction N−⌈e⋆⌉ are verified with zero discrepancy up to N=2000. The manuscript already states that this is a verified identity on that range rather than a theorem for all N; adding the same disclaimer next to the display of (19) would prevent a casual reader from over-reading the asymptotic claim.
  3. Proposition 1 (Property (R)) is proved by two-point polarization and is not used in the certificates of Sections 5–6. The exposition is clear, but a forward reference in Section 5 (“proved later as Proposition 1; not required for Theorem 1”) would help readers who stop after the main certificates.
  4. Table 1 is only a summary; the full list of 92 sizes is deferred to the supplementary material. For archival readability it would be useful to include at least the first and last few certified N together with their margins in the main text, or to print the full list as a short table in an appendix.
  5. Notation: κ(d):=−v̄(d) and the band forms R_t appear in Section 5 before the layer-cake identity is fully motivated. A single sentence recalling that κ is nonincreasing (hence Δ_t≥0) at the start of Lemma 1 would smooth the reading.
  6. References [3] and [4] both concern the N=5 certificate; a parenthetical note that V*_5 is taken from the Charibde computation as reported in [3] would avoid any ambiguity about the numerical source of the padded value −9.103853.
  7. Typographical: abstract and introduction use both “Kuznetsov–Sahinidis” and “Kuznetsov and Sahinidis”; pick one style. Also, “P´ olya” in the bibliography should be “Pólya”.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: certificates and asymptotics are derived from independent proven inputs, classical inequalities, and directed-rounding verification, not by construction from the claimed outputs.

full rationale

The derivation chain is self-contained and non-circular. The refined per-layer floor f(n) of (7) is assembled from the externally proven V*_5 and V*_6 together with the subset inequality already published by Kuznetsov–Sahinidis; V_put enters solely as an energy of an explicit configuration that upper-bounds V*_N. Lemma 2 closes every certificate by the classical rearrangement inequality for sequences and an elementary field/distance argument, without appealing to the later-proved Property (R). All 92 comparisons are re-verified in directed-rounding arithmetic against those independent upper bounds. The asymptotic form of the original KS program (Theorem 2) follows from the exact closed-form minimum (12) whose constants σ_0 and C_bg are zeta values of the kernel; the refined gain constant θ of Corollary 2 is an exact algebraic function of V*_6 alone. No parameter is fitted to data and then re-presented as a prediction, no uniqueness theorem is imported from the present author, and the single external structural premise (consecutive nonempty layers for minimizers) is flagged as an input from [1] rather than derived inside the paper. The work therefore contains no self-definitional step, no fitted-input-as-prediction, and no load-bearing self-citation chain.

Assumptions & free parameters 0 free parameters · 5 assumptions · 0 invented entities

The central certificates rest on three external pillars already present in the literature (the KS layer construction and consecutive-layer structural property, the proven numerical values V*_5 and V*_6, and the subset inequality) together with standard real analysis (rearrangement inequality, directed rounding). No free parameters are fitted to produce the certificates or the asymptotic constants; the only numerical tolerances are explicit safety paddings of 10^{-6} or 10^{-4}. No new physical entities are postulated.

assumptions (5)
  • domain assumption A global minimizer of the LJ energy occupies a block of consecutive nonempty unit-width layers (Kuznetsov–Sahinidis structural property).
    Invoked in Section 3 and in the final paragraph of the proof of Theorem 1 to pass from energy certificates on fully occupied profiles to a diameter statement about minimizers.
  • domain assumption The numerical values V*_5 = −9.103852 and V*_6 = −12.712062 are the true global minima for N=5 and N=6.
    Taken from Vanaret et al. (Charibde) and Kuznetsov–Sahinidis (Baron); entered with outward padding in the directed-rounding certificates of Section 6.
  • domain assumption Subset inequality V*_n ≥ n(n−1)/(M(M−1)) V*_M for M ≤ n (Kuznetsov–Sahinidis).
    Used at M=6 to define the refined per-layer floor f(n) for n ≥ 7 in equation (7)–(8).
  • standard math Rearrangement inequality for two finite sequences (Hardy–Littlewood–Pólya, Thm. 368).
    Sole classical ingredient of the arrangement-free lower bound of Lemma 2 (pairing step).
  • domain assumption V_put(N) is the energy of an explicit configuration, hence V*_N ≤ V_put(N).
    Used only as an upper bound in the comparison rule (6); sources [5–7,9] are tabulated putative minima.

how reviews work

0 comments
Cite this review

Pith. "Pith review of A certified refinement and asymptotic analysis of the Kuznetsov-Sahinidis diameter bound for Lennard-Jones clusters." pith.science (2026). https://pith.science/paper/Y3OMQW5Q

@misc{pith2026260709555,
  author       = {Pith},
  title        = {Pith review of: A certified refinement and asymptotic analysis of the Kuznetsov-Sahinidis diameter bound for Lennard-Jones clusters},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/Y3OMQW5Q}},
  note         = {Machine review of arXiv:2607.09555}
}
abstract

Kuznetsov and Sahinidis (J. Glob. Optim., 2025) prove distance bounds that confine optimal Lennard-Jones clusters and shrink the search region of deterministic solvers; their diameter bound charges each unit-width layer the loosest internal energy $-\binom{n}{2}$. We replace this by a certified estimate built from their own subset inequality and the proven minima $V_5^*$, $V_6^*$, and minimize the resulting layer bound over population profiles. Centred decreasing profiles supply candidate minima; an arrangement-free relaxation, whose only classical ingredient is the rearrangement inequality for sequences, closes every certificate over all profiles; and all comparisons are re-verified in directed-rounding arithmetic. For $5 \le N \le 200$ this certifies a one-layer improvement of the published bound at 92 sizes. The improvement resolves no open global-optimization case: it is a rigorous tightening of a published a priori bound, with a precise account of the mechanism. A direct downstream test on the only tractable sizes ($N \le 6$) finds the diameter box is not the binding resource for a deterministic solver there, and no solver runs at the sizes the refinement affects ($N \ge 38$); we therefore present the result as a theoretical note. We also derive the asymptotic form of the bound, $\rho_{KS} = N - \Theta(\sqrt{N})$, resolving a point left open by Kuznetsov and Sahinidis; the gain of the refinement itself grows like $\Theta(\sqrt{N})$ layers, with an exact asymptotic constant.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

11 extracted references · 1 canonical work pages

  1. [1]

    Kuznetsov and N

    A. Kuznetsov and N. V. Sahinidis. New bounds and formulations for the determinis- tic global optimization of Lennard–Jones clusters.Journal of Global Optimization, 2025. doi:10.1007/s10898-025-01476-7. Published online 11 March 2025

    work page doi:10.1007/s10898-025-01476-7 2025

  2. [2]

    L. T. Wille and J. Vennik. Computational complexity of the ground-state determination of atomic clusters.Journal of Physics A: Mathematical and General, 18(8):L419–L422, 1985

    1985

  3. [3]

    Vanaret, J.-B

    C. Vanaret, J.-B. Gotteland, N. Durand, and J.-M. Alliot. Certified global min- ima for a benchmark of difficult optimization problems. GECCO 2014 (Vancouver); arXiv:2003.09867; HAL hal-00996713. Contains the first numerical optimality proof for the five-atom Lennard–Jones cluster; see also [4]. 13

    arXiv 2014

  4. [4]

    C. Vanaret. Hybridization of interval methods and evolutionary algorithms for solving difficult optimization problems. arXiv:2001.11465, 2020

    arXiv 2001

  5. [5]

    R. H. Leary. Global optima of Lennard–Jones clusters.Journal of Global Optimization, 11(1):35–53, 1997

    1997

  6. [6]

    J. A. Northby. Structure and binding of Lennard–Jones clusters: 13≤N≤147.Journal of Chemical Physics, 87:6166–6177, 1987

    1987

  7. [7]

    D. J. Wales and J. P. K. Doye. Global optimization by basin-hopping and the lowest energy structures of Lennard–Jones clusters containing up to 110 atoms.Journal of Physical Chemistry A, 101:5111–5116, 1997

    1997

  8. [8]

    J. P. K. Doye, M. A. Miller, and D. J. Wales. The double-funnel energy landscape of the 38-atom Lennard–Jones cluster.Journal of Chemical Physics, 110(14):6896–6906, 1999

    1999

Show all 11 references
  1. [9]

    Romero, C

    D. Romero, C. Barr´ on, and S. G´ omez. The optimal geometry of Lennard–Jones clusters: 148–309.Computer Physics Communications, 123(1–3):87–96, 1999

    1999

  2. [10]

    G. H. Hardy, J. E. Littlewood, and G. P´ olya.Inequalities. Cambridge University Press, 2nd edition, 1952

    1952

  3. [11]

    M. K.-H. Kiessling and D. J. Wales. A note on the minimal pairwise dis- tance in optimal Lennard–JonesN-body clusters.Molecular Physics, 2025. doi:10.1080/00268976.2025.2590148; arXiv:2511.15008. 14

    2025 doi

Pith tools

Reviewed July 13, 2026 · model on record in the stance chip above.