{"id":"029b07f6-cf74-409c-a261-0f061030aff4","arxiv_id":"1908.08159","paper_version":3,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":1,"one_line_summary":"Nearby energy minima of 3D jammed soft-sphere packings are hierarchically clustered and approach an ultrametric metric in the thermodynamic limit, supporting Gardner physics in athermal jamming.","lead":"This paper measures how nearby stable packings, the energy minima of jammed soft-sphere systems, are organized and finds that they form a tree-like hierarchy known as an ultrametric structure. The result is offered as evidence that athermal jammed materials share the marginal glass physics predicted by mean-field theory.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"No null model or multi-parent averaging: the reported D≈2.7 plateau could be an artifact of sampling one basin, and the 'precisely ultrametric' conclusion is not actually established.","rationale":"The reader's conditional verdict is appropriate. The paper reports a careful pipeline: quad precision, a contact-network metric, and the MST-based subdominant ultrametric, and the D master curve is a nontrivial empirical finding. However, the central claim as worded overreaches. The sampled metrics all stem from one parent minimum per state point, so the observed hierarchy could characterize a single basin rather than the landscape; no null model or multi-parent averaging is provided. In addition, the thermodynamic-limit argument is a scaling argument: D≈2.7 is an O(1) absolute deviation, and distances scaling as √N imply only that D/√N→0, not that the metric is exactly ultrametric. Both gaps are addressable: repeat with multiple parents, vary εmax, and compare against a randomized-distance null. If those tests fail, the paper would still describe a basin-specific hierarchy, not the claimed Gardner-phase evidence. These concerns do not warrant rejection, but they reinforce the conditional verdict.","tokens_in":7776,"tokens_out":11988,"duration_ms":136490,"concrete_test":"Repeat the full protocol for N=4096 and p=10^{-3} using 10 independently created parent packings, collecting 500 distinct nearby minima from each parent, and compute D per parent and pooled. If the parent-to-parent spread in D covers the published value D≈2.7 and the pooled D is not significantly smaller, representative sampling of the local landscape is not established; if all parents give D well below 2.7, the single-parent concern is resolved.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The paper's quantitative case rests on sampling minima by perturbing a single arbitrary minimum and re-minimizing, with perturbation length chosen uniformly up to εmax=0.4/√N, a cutoff that is itself selected from the observed distance-vs-perturbation curves (Figure 2). All 500 or 5000 minima therefore come from one parent packing per state point, and no parent-to-parent averaging is reported. The subdominant ultrametric and D of Equations (2)-(3) are computed from that one local sample. Because the MST/single-linkage recipe can be applied to any distance matrix and always yields an ultrametric, the statement D≈2.7 has no baseline meaning without a null model or a comparison with uncorrelated minima. If the result is specific to a single basin, it cannot support the conclusion that the jamming energy landscape as a whole is hierarchical and ultrametric. Furthermore, the Conclusions claim the metric becomes 'precisely ultrametric' in the thermodynamic limit, but Figure 5 shows D≈2.7 independent of N while pairwise distances grow as √N. That only establishes a vanishing fractional excess, not D→0, so the exact wording is stronger than the measurement even under the representative-sampling assumption.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper claims that the energy landscape of three-dimensional athermal jammed packings is hierarchical and ultrametric, providing direct evidence for a marginal Gardner phase along the zero-temperature jamming line. The authors generate sets of nearby minima by repeatedly perturbing a single initial minimum with random displacements and re-minimizing, then define a metric based on differences of stable contact vectors. They construct the subdominant ultrametric from this metric via a minimum spanning tree and quantify the discrepancy with a quantity D. They report that D collapses as a function of N^2p and reaches a plateau of about 2.7 as the scaled pressure goes to zero. Since pairwise distances between minima scale as sqrt(N), they conclude that the fractional excess vanishes in the thermodynamic limit and that the metric becomes 'precisely ultrametric' for all pressures explored.","tokens_in":7983,"tokens_out":4234,"duration_ms":44971,"significance":"If established, this result would be a significant step: it would connect mean-field full replica symmetry breaking (Gardner) predictions to finite-dimensional athermal packings, where no thermal exploration is available. The paper's strengths include the high-precision minimization (force threshold 10^-20), the use of stable contact vectors to avoid rattler ambiguities, and the reported collapse of D across system sizes. However, the current evidence is not yet conclusive because the sampling protocol uses a single parent minimum per state point and lacks any null model or control comparison, so the measured D has no baseline meaning. The central observation is interesting and potentially important, but additional controls are needed before the strong 'precisely ultrametric' claim can be accepted.","major_comments":[{"comment":"The claim that the metric 'becomes precisely ultrametric' in the thermodynamic limit is stronger than the measurement. Figure 5 shows D reaching a plateau of about 2.7 as N^2p goes to zero, and this plateau is independent of N, while typical pairwise distances grow as sqrt(N). Since D itself does not decrease with N, the absolute violation of the ultrametric inequality does not vanish; only the relative violation D/d tends to zero. To support 'precisely ultrametric', the authors would need to show that D itself tends to zero, or reformulate the conclusion as asymptotic approximate ultrametricity in a relative sense.","section":"Figure 5 and Conclusions"},{"comment":"All nearby minima are generated by perturbing a single arbitrary initial minimum, and no averaging over independent parent packings is reported. The subdominant ultrametric and D are computed from that one local sample, so the results could reflect the structure of only a single basin rather than the landscape as a whole. The conclusion that the jamming energy landscape is hierarchical and ultrametric requires parent-to-parent averaging and a demonstration that D is independent of the chosen parent minimum.","section":"Sampling procedure, Figures 4 and 5"},{"comment":"The perturbation cutoff epsilon_max = 0.4/sqrt(N) is selected from the observed distance-versus-perturbation curves in Figure 2, which are obtained from the same systems that are later analyzed. This couples the sampling protocol to the measured outcome and leaves open the possibility that a different cutoff would change the value of D or the collapse. The authors should test the robustness of the plateau and the collapse to the choice of epsilon_max, or choose the cutoff from independent systems.","section":"Figure 2 and definition of epsilon_max"},{"comment":"No null model is provided for the quantity D. Because D compares the measured metric to the subdominant ultrametric constructed from the same data, a value of D about 2.7 has no baseline meaning without comparison to, for example, random distance matrices or samples of uncorrelated minima. Such a control is necessary to establish that the observed hierarchy is specific to jammed systems and not a generic property of any distance matrix processed through the same algorithm.","section":"Equations (2)-(3) and null model"}],"minor_comments":[{"comment":"In the sentence containing Equation (1), 'the distanced between' contains a typo; it should read 'the distance d between'.","section":"Equation (1) text"},{"comment":"The phrase 'labelled by square rooted numbers' is awkward; consider rewriting as 'the color scale labels are square roots of the metric distance' or similar.","section":"Figure 3 caption"},{"comment":"The perturbation epsilon is first defined with normalization by N^{-1/3}, but the cutoff is later written as epsilon_max = 0.4/sqrt(N). Please clarify the relationship between these normalizations, since the figures appear to use the sqrt(N) convention.","section":"Notation for epsilon_max"},{"comment":"The author names in references [31] and [32] contain garbled accented characters in the preprint text; the final version should use properly encoded names.","section":"References [31] and [32]"}],"recommendation":"major_revision","confidential_remarks":"The manuscript reports a potentially valuable measurement, but the missing null model and the absence of multi-parent averaging are load-bearing gaps that prevent acceptance in the current form. I do not recommend rejection because the central idea is defensible and the required controls appear feasible within the scope of the paper. The main concern is that the strong 'precisely ultrametric' conclusion goes beyond what the data show, even under the representative-sampling assumption."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Bottom line: this is a clean, readable numerical study that likely contains a real effect, but the main conclusion is stated more sharply than the measurements support. The genuinely new content is the first ultrametricity analysis of nearby minima in 3D athermal over-jammed packings, spanning pressures and system sizes, with a plausible collapse of D onto a master curve. That is worth having.\n\nWhat the paper does well: the contact-vector metric is sensible and avoids rattler and global-drift noise; the minimization is carefully done (quad precision, force tolerance 1e-20); the hierarchical structure visible in Figures 3 and 4 is a nice qualitative result. The construction of the subdominant ultrametric via MST is standard and correctly described. The pressure dependence of D and its finite-size collapse are the genuinely new data.\n\nThe soft spots are real and mostly addressable. First, all sampled minima come from perturbing a single parent minimum per state point; there is no parent-to-parent averaging and no null model. The MST procedure yields an ultrametric for any distance matrix, so D≈2.7 has no baseline meaning without comparing to uncorrelated minima. That is the biggest gap. Second, εmax=0.4/√N is chosen from the same systems that later produce the master curve (Figure 2), which couples the sampling window to the result and needs a sensitivity test. Third, the inference from D≈2.7 constant while pair distances grow as √N does not establish 'precisely ultrametric' in the thermodynamic limit. It gives vanishing fractional excess, but the absolute violation remains constant; whether that counts as precisely ultrametric is a substantive question, not a wording quibble.\n\nOverall, the qualitative claim—hierarchy that deepens on approaching jamming—is well supported. The stronger claim that this proves the marginal Gardner phase in 3D athermal jamming is not yet established. But the result is significant enough, and the path to strengthening it is clear: rerun with multiple independent parent packings, add a null model, and test εmax sensitivity. I would send this to peer review; with appropriate revision it could be a solid paper.","headline":"A clean numerical study that likely shows real hierarchical structure in the jamming landscape, but the headline claim of 'precisely ultrametric' in the thermodynamic limit is stronger than the data support.","tokens_in":8563,"tokens_out":2196,"would_cite":true,"duration_ms":22221,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"Jammed packings have a hierarchical energy landscape that becomes precisely ultrametric as system size grows.","keywords":["jamming","ultrametricity","energy landscape","Gardner transition","replica symmetry breaking","soft spheres","nearby minima"],"falsifier":"Run the same protocol with a perturbation cutoff an order of magnitude larger, $\\varepsilon_{\\max} = 4/\\sqrt{N}$, and compare the collapse of $D(N,p)$. If the plateau value of $D$ changes or no longer collapses onto the same master curve, the ultrametricity is an artifact of the chosen sampling radius rather than a structural property of the landscape. A complementary check: compute $D$ for a crystalline or very high-pressure packing, where the Gardner phase should be absent; if $D$ also vanishes, the measure does not distinguish marginal from non-marginal systems.","tokens_in":7554,"feed_emoji":"🫧","tokens_out":9118,"duration_ms":79243,"temperature":0.7,"pith_summary":"This paper argues that the energy landscape of athermal, three-dimensional jammed packings is hierarchical, and that in the thermodynamic limit the distance metric between local minima becomes exactly ultrametric. The authors reach this by generating hundreds to thousands of nearby minima through random perturbations of a single packing, then measuring pairwise distances based on differences in the stable contact network. They find that the measured metric approaches its subdominant ultrametric as pressure decreases and system size grows, for all pressures from $10^{-1}$ down to $10^{-5.5}$. If correct, this would put the jamming transition inside the marginal Gardner phase, meaning the fractal, nested-basin structure predicted by mean-field replica theory is a genuine feature of purely geometrical, athermal matter.","feed_headline":"Jammed packings become exactly ultrametric at large size","feed_subtitle":"Perturbing a 3D packing reveals nested basins of minima, evidence for the Gardner phase at zero temperature","key_machinery":"The key object is the subdominant ultrametric $d_<$, the closest ultrametric to a given metric, which the paper obtains by computing the minimum spanning tree of the pairwise-distance graph and setting $d_<(a,b)$ to the largest edge weight on the tree path from $a$ to $b$. The underlying metric $d$ counts differences in the stable contact vector network, so that $d(a,b)\\approx\\sqrt{\\text{(number of changed contacts)}}$ for distances below $\\sqrt{N}$. Comparing $d$ with $d_<$ through $D$ measures how close the landscape is to being ultrametric, and the scaling of $D$ with $N$ and pressure is what carries the argument.","core_discovery":"The central claim is that the energy landscape of jammed soft-sphere packings is asymptotically ultrametric: for any three nearby minima $a$, $b$, $c$, the contact-based distance satisfies $d(a,c) \\le \\max\\{d(a,b), d(b,c)\\}$ in the limit of infinite system size. The paper demonstrates this by constructing the distance metric from stable contact vector networks for sets of 500–5000 minima found by perturbing a single packing and re-minimizing, and comparing this metric to its subdominant ultrametric $d_<$, built from a minimum spanning tree. The average deviation $D = \\sqrt{\\langle (d-d_<)^2\\rangle}$ collapses onto a master curve as a function of $N^2p$ and approaches a plateau near 2.7, while typical distances grow as $\\sqrt{N}$; the fractional deviation therefore vanishes as $N\\to\\infty$. The authors take this as direct evidence for a marginal Gardner phase along the zero-temperature jamming line, arising from geometry alone rather than thermal fluctuations.","pith_inferences":["An extension not pursued here: polydisperse packings or Hertzian (non-harmonic) contacts may show the same ultrametricity, but the paper tests only monodisperse harmonic spheres, so the universality claim is still open.","A stricter validation would vary the perturbation cutoff $\\varepsilon_{\\max}$ well beyond $0.4/\\sqrt{N}$; if $D$ were to change appreciably, the hierarchy would be an artifact of the sampling radius rather than an intrinsic landscape property.","If the landscape is truly ultrametric, relaxation dynamics at low temperature should show hierarchical barrier crossing: a minimum in one sub-basin must pass through the parent basin before reaching a different sub-basin, a signature that could be sought in particle-tracking experiments."],"forward_implications":["The Gardner phase predicted by mean-field replica theory is not restricted to high dimensions: it appears in ordinary three-dimensional athermal packings.","Near jamming, the number of nearby minima proliferates and the region that can be densely sampled shrinks, consistent with a fractal basin structure.","The contact-network metric gives experimentalists a way to search for ultrametricity in granular or colloidal packings without any thermal sampling.","The finite-size scaling of $D$ means that larger simulated or experimental packings should show cleaner ultrametric signatures, making the prediction testable at accessible system sizes."],"supporting_citations":[{"why":"Mean-field prediction that jamming lies inside the marginal Gardner phase; the claim this paper tests.","marker":"[12]"},{"why":"Introduces the fractal free-energy landscape picture (full replica symmetry breaking) in structural glasses that motivates looking for nested basins.","marker":"[5]"},{"why":"Prior numerical observation of a hierarchical landscape in thermal hard-disk glasses; the athermal three-dimensional extension is contrasted with this.","marker":"[21]"},{"why":"Provides the pyCudaPacking software used to generate and minimize the jammed packings.","marker":"[28]"},{"why":"FIRE minimization algorithm used to re-minimize perturbed configurations and find distinct minima.","marker":"[32]"},{"why":"Kruskal's minimum spanning tree algorithm, the base of the subdominant ultrametric construction.","marker":"[41]"},{"why":"Defines the subdominant ultrametric and proves it is the closest ultrametric; the paper's $D$ compares $d$ with this $d_<$.","marker":"[42]"},{"why":"Supplies the finite-size scaling $N^2p$ used as the distance-to-jamming axis for the master curve collapse.","marker":"[44]"}],"fun_headline_variants":["Jammed packings exhibit ultrametric energy landscapes","Energy landscape of jammed spheres is ultrametric","Hierarchical structure found in jamming energy landscape","Jamming landscape is ultrametric, study finds","Ultrametricity emerges in jammed soft-sphere packings"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is that the set of nearby minima obtained by perturbing a single arbitrary packing with lengths up to $0.4/\\sqrt{N}$ fairly represents the local energy landscape, so that the observed ultrametricity reflects the landscape rather than the sampling method.","fun_headline_variants_meta":{"raw":{"variants":["Jammed packings exhibit ultrametric energy landscapes","Energy landscape of jammed spheres is ultrametric","Hierarchical structure found in jamming energy landscape","Jamming landscape is ultrametric, study finds","Ultrametricity emerges in jammed soft-sphere packings"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000143,"raw_usage":{"total_tokens":1118,"prompt_tokens":837,"completion_tokens":281,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":453,"completion_tokens_details":{"reasoning_tokens":203}},"tokens_in":453,"tokens_out":281,"duration_ms":2769,"temperature":1.0,"reasoning_tokens":203,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T11:48:14.447519+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Run the same protocol with a perturbation cutoff an order of magnitude larger, $\\varepsilon_{\\max} = 4/\\sqrt{N}$, and compare the collapse of $D(N,p)$. If the plateau value of $D$ changes or no longer collapses onto the same master curve, the ultrametricity is an artifact of the chosen sampling radius rather than a structural property of the landscape. A complementary check: compute $D$ for a crystalline or very high-pressure packing, where the Gardner phase should be absent; if $D$ also vanishes, the measure does not distinguish marginal from non-marginal systems.","supporting_citations":[{"cited_title":"Shear Yielding and Shear Jamming of Dense Hard Sphere Glasses,","cited_arxiv_id":null,"evidence_quote":"Mean-field prediction that jamming lies inside the marginal Gardner phase; the claim this paper tests."},{"cited_title":"Frac- tal free energy landscapes in structural glasses,","cited_arxiv_id":null,"evidence_quote":"Introduces the fractal free-energy landscape picture (full replica symmetry breaking) in structural glasses that motivates looking for nested basins."},{"cited_title":"Hierarchical Land- scape of Hard Disk Glasses,","cited_arxiv_id":null,"evidence_quote":"Prior numerical observation of a hierarchical landscape in thermal hard-disk glasses; the athermal three-dimensional extension is contrasted with this."},{"cited_title":"Geometric signa- tures of jamming in the mechanical vacuum,","cited_arxiv_id":null,"evidence_quote":"Provides the pyCudaPacking software used to generate and minimize the jammed packings."},{"cited_title":"On the shortest spanning subtree of a graph and the traveling salesman problem,","cited_arxiv_id":null,"evidence_quote":"Kruskal's minimum spanning tree algorithm, the base of the subdominant ultrametric construction."},{"cited_title":"On the degree of ultrametricity,","cited_arxiv_id":null,"evidence_quote":"Defines the subdominant ultrametric and proves it is the closest ultrametric; the paper's $D$ compares $d$ with this $d_<$."},{"cited_title":"Finite-Size Scaling at the Jamming Transition,","cited_arxiv_id":null,"evidence_quote":"Supplies the finite-size scaling $N^2p$ used as the distance-to-jamming axis for the master curve collapse."}],"review_version":1}