{"id":"d07b9588-80d0-46d1-a91c-a409687dfa4a","arxiv_id":"2608.06513","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"For five non-overlapping unit disks, the minimum convex hull perimeter is 10+2π, attained exactly by three contact classes, two with perimeter-preserving flexes and one rigid.","lead":"This paper develops a variational calculus for clusters of non-overlapping disks and uses it to prove that the smallest possible perimeter of the convex hull of five unit disks is 10+2π, achieved by three contact geometries. The result is the first rigorous step beyond the four-disk case of an old packing problem.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Theorem 6.1 depends on the unpublished interval-arithmetic certificate [AHV26] for all non-leaf exclusions; until that certificate is released and independently reproduced, the central claim cannot be fully audited, so a conditional verdict is warranted.","rationale":"The reader's weakest_assumption identifies exactly the unpublished computational certificate [AHV26] as the main obstacle, and I agree. The certificate is load-bearing because the analytic part of the proof only excludes the circular classes X1/X7 and the generic non-parallel-leaf locus; the residual leaf-parallel loci in six classes are excluded solely by the certified interval-arithmetic negative eigenvalues. Without the certificate, a skeptical reader cannot distinguish a genuine proof from a plausible claim. The connected-graph reduction is a second gap, but it is less central because it is a structural statement that could be checked separately and is likely true; the certificate is the part that is actually missing from the submission. I therefore do not recommend changing the CONDITIONAL verdict: the paper's framework and analytic exclusions are coherent, but the finite verification must be made available and reproducible before the theorem can be accepted as proven. The proposed concrete test directly releases and validates the missing certificate, and an independent re-enumeration would confirm completeness of the 13-class reduction.","tokens_in":16729,"tokens_out":2356,"duration_ms":26353,"concrete_test":"Release the files cited as [AHV26] (supplementary_verification.py, 5disks.json, and full verification logs) and rerun the certificate with directed rounding and interval arithmetic. Independently re-enumerate all connected penny graphs on five vertices using plantri/nauty, re-derive the 13 contact classes, recompute all parallel-leaf realisations in X2, X3, X5, X6, X8, X9, and verify that each has a strictly negative intrinsic Hessian eigenvalue on the reduced rolling space via the stated LDL^T pivot test. Also prove the §6.1 connected-replacement claim for disconnected five-disk configurations. If any negative pivot fails to reproduce, or any leaf-parallel realisation is missed, Theorem 6.1 is unsupported.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The proof of Theorem 6.1 is not self-contained. Section 6.2 excludes the hull-leaf classes X2, X3, X5, X6, X8, X9 by citing [AHV26] for two facts: (i) the complete list of parallel-leaf realisations modulo rigid motions in each class, and (ii) a certified negative intrinsic Hessian eigenvalue for each such realisation, obtained by interval arithmetic via an LDL^T pivot test (Appendix B.2). The X4 exclusion also terminates in boundary contacts that belong to X8 and X9, so it inherits the same dependency. Since [AHV26] is 'available upon request' rather than provided, a referee cannot check whether the enumeration of parallel-leaf realisations is complete, whether the reduced rolling-space basis Z is computed correctly, or whether the negative-pivot certificates cover every relevant direction. The theorem's central value 10+2π is numerically plausible and consistent with prior work, but the proof's finite reduction is only as strong as this unaudited certificate. A second, independent gap is the unproved connected-graph reduction in §6.1: the assertion that any disconnected minimiser can be replaced by an incident connected configuration with no larger centre-hull perimeter is stated without proof and is load-bearing for restricting attention to the 13 connected penny graphs. Both gaps are addressable, but as published they make the theorem conditional.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper develops a variational calculus for the perimeter of the convex hull of hard disk clusters, based on fixed hull-contact classes, the rolling space of contact-preserving infinitesimal motions, a first-order hull-leaf obstruction, and an intrinsic Hessian on the reduced rolling space. As an application it proves (Theorem 6.1) that the minimum perimeter of a configuration of five unit disks is 10+2π, attained exactly in contact classes X10, X11, and X12, with local perimeter-preserving flexes of dimensions one and two in the first two classes and rigidity modulo rigid motions in the third. The proof combines analytic exclusions (X1 and X7 by a strict-concavity argument; hull-leaf classes by Theorem 4.2; X4 by an explicit one-parameter descent family) with computational certificates for the parallel-leaf realisations in classes X2, X3, X5, X6, X8, and X9, supplied by the referenced but unpublished file [AHV26].","tokens_in":17035,"tokens_out":9064,"duration_ms":81056,"significance":"If fully validated, this is a meaningful advance: it is the first exact determination of the minimum hull perimeter for n=5, and the proposed intrinsic framework (rolling space, first-order leaf obstruction, intrinsic Hessian diagnostics) may be reusable for larger n. The analytic parts I checked are sound: the X1/X7 concavity exclusion, the hull-leaf obstruction, and the X4 descent family are correct and clearly presented. The four-disk spectral prototypes in Appendix A usefully illustrate the distinction between rigidity, flat degeneracy, and second-order instability. The reported value 10+2π is consistent with the numerical literature. However, the proof of the main theorem is not self-contained as published, because a load-bearing reduction and the decisive computational certificate are not included.","major_comments":[{"comment":"The assertion that 'any disconnected minimising configuration can be replaced by an incident configuration with an additional contact and no larger centre hull perimeter' is load-bearing, because it restricts the search to the 13 connected penny graphs, but no proof or reference is given. Adding a contact can change the hull and may increase the perimeter; a rigorous argument (or a precise citation to a known lemma) is required. Without this, the finite enumeration is incomplete.","section":"§6.1"},{"comment":"The exclusions of the hull-leaf classes X2, X3, X5, X6, X8, and X9 rest entirely on the unpublished certificate [AHV26], which is listed as 'available upon request' rather than provided. The certificate is asserted to (i) identify all parallel-leaf realisations in each class, and (ii) provide a certified negative intrinsic Hessian eigenvalue via interval arithmetic. The X4 exclusion ends at boundary configurations belonging to X8 and X9, so it inherits the same dependency. A referee cannot audit the completeness of the enumeration, the construction of the reduced rolling-space basis Z, or the interval arithmetic bounds described in Appendix B.2. The Python script, the dataset, and the verification logs must be made publicly available, and the pipeline described in enough detail to be independently reproduced.","section":"§6.2 and Appendix B"},{"comment":"The sentence 'For X10 and X11, every hull edge is an active contact of length 2, so Per(P(c)) = 10' is incomplete as written. If the hull had only k<5 vertices, all of whose edges are contacts of length 2, the perimeter would be 2k, which is strictly less than 10 and would contradict the claimed lower bound. The proof should explicitly state that the unique hull class in X10 and X11 has all five centres as hull vertices, or cite the certificate for that fact. This is a small but real gap in the written proof of the central theorem.","section":"§6.2, proof of Theorem 6.1"}],"minor_comments":[{"comment":"There is a typo: 'depend stronlgy' should read 'depend strongly'.","section":"§1"},{"comment":"The description 'Integer-pivoted Gaussian elimination yields a basis Z for Roll(c)∩R(c)⊥' is terse; a reference to the specific algorithm or a brief derivation would improve reproducibility.","section":"Appendix B.2"},{"comment":"The eigenvalue interval for X5, [−12.204,−7.2100], is much wider than the other intervals; a footnote explaining whether this reflects multiple realisations or numerical issues would help the reader.","section":"Table 1"},{"comment":"The notation c0(t) is introduced, but the exact set of five active contacts is not listed; a short explicit statement of which contacts are preserved would make the verification easier to follow.","section":"§6.2, X4 paragraph"},{"comment":"The reference [AHV26] should be updated to include a repository URL or arXiv identifier, rather than only 'available upon request', so that the computational claims can be checked.","section":"References"}],"recommendation":"major_revision","confidential_remarks":"The decision should hinge on whether the authors can make [AHV26] publicly available and supply the missing proof of the connected-graph reduction. If those are provided and the certificate is consistent, the theorem is likely correct; I see no circularity in the analytic parts and no fitted parameters. The current dependence on an in-house, unpublished certificate is unusual for a theorem in a mathematics journal and should be resolved before publication. The fit with the journal's scope is appropriate."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The headline: this paper gives the first serious attack on the five-disk minimum perimeter problem and states the result as 10+2π, attained in exactly three contact classes. I think the result is probably correct, but as submitted it is not fully auditable.\n\nWhat is genuinely new: the variational framework built around the rolling space, hull leaf obstruction, and intrinsic Hessian is a clean and useful assembly of rigidity and optimization ideas. The hull leaf theorem (4.2) is a nice first-order exclusion tool, and the analytic exclusions for X1, X7, and X4 are solid. The classification of the surviving classes, with flat perimeter-preserving flexes in X10 and X11 and rigidity of X12, is exactly what the theory should produce. The paper is also honest about its computational dependence: Appendix B describes the interval-arithmetic certificate rather than hiding it.\n\nThe soft spots are precisely where the stress-test note puts them. The exclusions of X2, X3, X5, X6, X8, and X9 all rest on [AHV26], which is \"available upon request\" and not provided. A referee cannot check the completeness of the parallel-leaf enumeration or the negative-pivot certificates. The X4 boundary also passes into X8 and X9, so it inherits the same dependence. That is not necessarily fatal, but it is load-bearing.\n\nSecond, Section 6.1 asserts without proof that any disconnected minimizer can be replaced by an incident connected configuration with no larger center-hull perimeter. This is probably true and likely easy to prove, but as written it is an unproved reduction.\n\nThe X13 exclusion is terse but acceptable: the paper excludes the degenerate horizontal case and non-convex hulls via the hull leaf theorem.\n\nOverall, the mathematics is coherent and the gaps are addressable. I would send this to a serious referee rather than desk-reject. I would not cite the five-disk theorem in my own work until the certificate is public and the connected-graph reduction is proven, but I would bring it to a reading group focused on packing and rigidity.","headline":"Likely-true solution of the five-disk perimeter problem, but the written proof is conditional on an unpublished computational certificate and an unproved connected-graph reduction.","tokens_in":17508,"tokens_out":2019,"would_cite":false,"duration_ms":20291,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["52C15","52A40","49Q10","51M25"],"pacs":[],"model":"deepseek-v4-flash","headline":"The minimum perimeter of a cluster of five unit disks is 10+2π, attained exactly by contact classes X10, X11, and X12.","keywords":["disk packing","hard disks","perimeter minimisation","convex hull","contact graphs","rolling space","intrinsic Hessian","penny graphs"],"falsifier":"Take the exact symbolic coordinates recorded in Appendix B for the parallel-leaf realisations of classes \\(X_2, X_3, X_5, X_6, X_8, X_9\\), form the contact operator \\(A(c)\\) and the reduced rolling space \\(\\operatorname{Roll}(c)\\cap R(c)^\\perp\\), and compute the intrinsic Hessian \\(H_{\\mathrm{intr}}=Z^\\top H Z\\) in interval arithmetic. The paper certifies a negative eigenvalue for each; a single class for which the certified interval for the smallest eigenvalue contains zero or positive values would refute the exclusions and reopen the five-disk problem.","tokens_in":1866,"feed_emoji":"🔵","tokens_out":8792,"duration_ms":120886,"temperature":0.7,"pith_summary":"The minimum perimeter of a cluster of five unit disks is \\(10+2\\pi\\). No arrangement does better, and the proof shows that the bound is attained exactly by three realised contact classes. The paper builds a general calculus for hard disk clusters: configurations are grouped into fixed hull-and-contact classes, where the perimeter has a smooth local formula and the allowed infinitesimal motions form the rolling space. Within each class it derives first-order criticality conditions, a first-order descent test based on hull leaves, and a second-order intrinsic Hessian that distinguishes instability, flat degeneracy, and rigidity. Applying this calculus to five disks, it enumerates thirteen possible contact graphs, excludes ten of them, and identifies the three surviving classes as the complete set of minimizers. The result matters because the minimum perimeter problem was previously solved only up to four disks.","feed_headline":"Five unit disks have minimum hull perimeter 10+2π","feed_subtitle":"First solution beyond four disks: three contact classes attain the bound, two flex, one is rigid.","key_machinery":"The load-bearing object is the fixed hull and contact class \\(X=F\\cap C(G)\\), where \\(F\\) fixes the set and cyclic order of hull vertices and \\(C(G)\\) fixes the realised contact graph. At a configuration \\(c\\), each active contact \\(\\{i,j\\}\\) contributes the normal vector \\(u_{ij}=(c_j-c_i)/\\|c_j-c_i\\|\\), and the rolling space \\(\\operatorname{Roll}(c)=\\ker A(c)\\) consists of infinitesimal motions that preserve every contact to first order, with \\(A(c)\\) the contact operator. Class criticality is the multiplier equation \\(\\nabla\\widetilde{\\operatorname{Per}}(c)=A(c)^\\top\\$\\lambda$\\). The second variation is computed as the intrinsic Hessian \\(H_{\\mathrm{intr}}=Z^\\top(H_{\\mathrm{hull}}+H_{\\mathrm{cont}})Z\\), where \\(Z\\) spans \\(\\operatorname{Roll}(c)\\cap R(c)^\\perp\\), \\(H_{\\mathrm{hull}}\\) is assembled from transverse blocks \\(M_{pq}=\\|c_q-c_p\\|^{-1}(I-t_{pq}t_{pq}^\\top)\\) for hull edges, and \\(H_{\\mathrm{cont}}\\) from contact curvature blocks \\(K_{ij}=-\\lambda_{ij}(I-u_{ij}u_{ij}^\\top)/2\\). The sign of \\(H_{\\mathrm{intr}}\\) on the reduced rolling space decides whether a class-critical configuration is second-order unstable, flat degenerate, or rigid modulo rigid motions.","core_discovery":"The paper's central result is Theorem 6.1: among all configurations of five non-overlapping unit disks, the minimum of \\(\\operatorname{Per}(c)\\), the perimeter of the convex hull of the cluster, equals \\(10+2\\pi\\). The minimum is attained precisely by configurations in the realised contact classes \\(X_{10}\\), \\(X_{11}\\), and \\(X_{12}\\), and by no others. In \\(X_{10}\\) and \\(X_{11}\\) the minimisers are not isolated: they admit local perimeter-preserving admissible flexes of dimension one and two respectively, while the minimiser in \\(X_{12}\\) is rigid modulo rigid motions, meaning its rolling space consists only of infinitesimal rigid motions. On the way, the paper develops a variational calculus in which the realised contact normals and the cyclic hull order determine, respectively, the admissible first-order cone and the local form of the perimeter functional.","pith_inferences":["The same machine should extend to six disks: the connected penny-graph count grows to 46, and the paper's reductions—fixed hull classes, hull-leaf exclusion, and intrinsic Hessian diagnostics—are all finite, so the main new cost is computational rather than structural.","The flat-degeneracy criterion suggests a tighter link to rigidity theory of sticky disks: a class whose hull edges are all contacts is locally perimeter-constant, so perimeter minimality and rigidity become complementary rather than competing properties.","A direct testable extension is to run the interval-arithmetic certificate on the six-disk enumeration and look for a surviving class with positive semidefinite intrinsic Hessian but no hull leaf; such a class would be the first candidate for the next minimum that is neither trivially flat nor rigid.","The method's reproducibility hinges on the unpublished certificate [AHV26]; posting it would turn the five-disk theorem into a fully checkable proof."],"forward_implications":["Any five-disk minimizer has centre-hull perimeter exactly \\(10\\), since \\(\\operatorname{Per}(c)=\\operatorname{Per}(P(c))+2\\pi\\).","The three minimising classes are fully classified: \\(X_{10}\\) and \\(X_{11}\\) contain local one- and two-parameter perimeter-preserving families, while \\(X_{12}\\) is an isolated minimiser modulo rigid motions.","The first-order hull-leaf test and the second-order intrinsic-Hessian test give a general exclusion scheme that does not require solving a global optimization problem; the same scheme can be applied to any fixed hull and contact class for larger \\(n\\).","The five-disk problem reduces to checking thirteen connected penny graphs, of which ten are excluded; this confirms the enumeration count \\(a(5)=13\\) of connected penny graphs on five vertices.","Since every hull edge in the surviving classes \\(X_{10}\\) and \\(X_{11}\\) has length two, the perimeter is constant on those classes, so the minimum is achieved by whole flexing families rather than by a single shape."],"supporting_citations":[{"why":"Supplies the computational certificate enumerating the thirteen five-disk contact classes and certifying a negative intrinsic Hessian eigenvalue for each excluded hull-leaf class; the theorem's finite exclusions rely on it.","marker":"[AHV26]"},{"why":"Gives the edge bound \\(|E|\\le \\lfloor 3n-\\sqrt{12n-3}\\rfloor\\), bounding the contact-graph enumeration to \\(|E|\\le 7\\) for \\(n=5\\).","marker":"[Ha74]"},{"why":"Provides the graph generation tool used to enumerate the planar graphs from which the thirteen contact classes are filtered.","marker":"[BMc07]"},{"why":"Provides the graph isomorphism reduction tools used to remove duplicate contact graphs.","marker":"[McP14]"},{"why":"Records OEIS A085632, the connected penny-graph sequence, used to confirm that the filtered list of thirteen contact classes is complete.","marker":"[Sl26]"}],"fun_headline_variants":["Five disks: minimum hull perimeter 10+2π","Minimum perimeter for 5 disks is 10+2π","Five disk clusters: min hull perimeter 10+2π","Hard disk quintet: optimal hull perimeter 10+2π"],"cache_read_input_tokens":19712,"weakest_assumption_plain":"The theorem depends on the unpublished computational certificate [AHV26], which identifies all parallel-leaf realisations in classes X2, X3, X5, X6, X8, X9 and certifies a negative intrinsic Hessian eigenvalue for each; it also assumes, without proof, that any disconnected minimizer can be replaced by an incident connected configuration with no larger centre-hull perimeter.","fun_headline_variants_meta":{"raw":{"variants":["Five disks: minimum hull perimeter 10+2π","Minimum perimeter for 5 disks is 10+2π","Five disk clusters: min hull perimeter 10+2π","Hard disk quintet: optimal hull perimeter 10+2π"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000547,"raw_usage":{"total_tokens":2602,"prompt_tokens":921,"completion_tokens":1681,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":537,"completion_tokens_details":{"reasoning_tokens":1610}},"tokens_in":537,"tokens_out":1681,"duration_ms":10450,"temperature":1.0,"reasoning_tokens":1610,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-10T04:18:06.144787+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Take the exact symbolic coordinates recorded in Appendix B for the parallel-leaf realisations of classes \\(X_2, X_3, X_5, X_6, X_8, X_9\\), form the contact operator \\(A(c)\\) and the reduced rolling space \\(\\operatorname{Roll}(c)\\cap R(c)^\\perp\\), and compute the intrinsic Hessian \\(H_{\\mathrm{intr}}=Z^\\top H Z\\) in interval arithmetic. The paper certifies a negative eigenvalue for each; a single class for which the certified interval for the smallest eigenvalue contains zero or positive values would refute the exclusions and reopen the five-disk problem.","supporting_citations":[],"review_version":1}