{"id":"454cc256-08f4-4793-89e1-86a86c592969","arxiv_id":"2506.04485","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"On a random geometric graph, the Yaldram-Khan model shows two second-order phase transitions at low average degree, whereas an Erdos-Renyi network preserves the original first- and second-order transitions.","lead":"This paper runs Monte Carlo simulations of the Yaldram-Khan catalytic surface reaction on two kinds of random networks. It reports that a random geometric graph turns the model's sharp phase transition into a smooth one at low average connectivity, while an Erdos-Renyi network preserves the original transition types.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"RGG order-of-transition claim rests on visual smoothness at a single system size; finite-size scaling is needed to rule out rounding of a first-order transition.","rationale":"The reader's conditional verdict targets exactly the right weakness: order is inferred from single-size density curves without finite-size diagnostics. That weakness is load-bearing because the paper's most striking new claim is a change in transition order, and order-of-transition determinations are known to be unreliable from steady-state density shapes alone unless accompanied by finite-size scaling, histograms, or cumulants. The paper otherwise does useful calibration (three lattice sizes for the hexagonal lattice, error bars, network-size snapshots) and its raw data are plausible, but those supports do not cover the RGG upper transition. I agree with the reader that the Conclusions overstate what the evidence shows. Since the evidence is insufficient but not contradictory, the appropriate disposition remains conditional confirmation pending the Binder/histogram finite-size test, matching the reader's verdict; no change is needed.","tokens_in":11555,"tokens_out":5121,"duration_ms":61648,"concrete_test":"For RGG with K≈4.5 and K≈6.06, perform steady-state simulations at N=32^2, 64^2, 128^2, and 256^2 for y in a narrow window around the upper transition y2 (e.g., y2±0.02 with Δy=0.002), with τ≥10^6 and sampling long enough to accumulate histograms. Compute the Binder cumulant U4=1−⟨ρV^4⟩/(3⟨ρV^2⟩^2) and the steady-state histogram of ρV. If U4 develops a minimum that deepens with N and histograms become bimodal, the transition is discontinuous and the central claim fails; if U4 curves cross at a single point and histograms remain single-peaked with width decreasing as N^-1/2, the second-order assignment is supported.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central conclusion — that RGG converts the YK upper transition from discontinuous to continuous for small K — is supported only by the visual smoothness of ρV(y) at N=128^2=16384 in Figs. 4(c) and 8(a). No Binder cumulant, order-parameter histogram, hysteresis scan, or finite-size scaling is reported. A first-order transition in a finite system is rounded over a window that shrinks with system size, so the smooth approach to y2 for K≈4.5–6.0 may be finite-size rounding rather than a genuine continuous transition. The same visual protocol is calibrated on the hexagonal lattice where the jump is clear, but that does not transfer to the RGG without quantitative analysis. The paper's own hedging in Sec. III D ('seems', 'suggests', 'need for further investigation') is converted in the Conclusions into an unqualified statement that 'the first-order phase transition is converted into second one for small values of the average degree, K.' The non-percolation at K≈3.8 is secondary: by K≈4.5 sc≈0.81, so the active-phase data are on a percolated network; the load-bearing weakness is the unverified order assignment.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies the Yaldram–Khan (YK) catalytic reaction model on three surfaces: the hexagonal lattice, Erdős–Rényi networks (ERN), and random geometric graphs (RGG), using steady-state Monte Carlo simulations at NO dissociation rate rNO=1. The authors report that the ERN preserves the qualitative phase diagram of the hexagonal lattice—one continuous and one discontinuous transition—for all average degrees considered, while the RGG appears to convert the upper, first-order transition into a second-order one for small average degree K, recovering a discontinuous transition only near K≈9. The paper also presents the reactive-window width as a function of average degree for both networks and discusses percolation of the RGG substrates.","tokens_in":11861,"tokens_out":5115,"duration_ms":55542,"significance":"If the central claim is correct, the RGG result is a genuinely interesting finding: local, spatially constrained randomness in the catalytic surface would change the order of a nonequilibrium phase transition, whereas long-range ERN randomness would not. The manuscript provides a useful dataset of phase boundaries and reactive-window widths for the YK model on two network families, and the authors are careful to report simulation parameters (N, τ, S) and to compare against known hexagonal-lattice transition points. However, the central order-of-transition claim is not yet established: it is inferred from the visual smoothness of density curves at a single system size, with no finite-size scaling, histogram analysis, or quantitative order-parameter diagnostic. The paper therefore needs substantial additional analysis before the central conclusion can be accepted.","major_comments":[{"comment":"The central claim that the RGG converts the upper YK transition from first to second order for small K rests on the visual smoothness of ρV(y) at a single system size N=128^2. In a finite system a first-order transition is rounded, so the smooth approach to y2 for K≈4.5–6.0 could be finite-size rounding rather than a genuine continuous transition. No finite-size scaling, Binder cumulant, order-parameter histogram, or hysteresis analysis is provided for the RGG. I ask for multi-N simulations at representative K values (e.g., K≈4.5, 6.06, and 9.1) with a quantitative order-parameter criterion—such as bimodality of the density distribution or crossing of a cumulant—before the order conversion is claimed.","section":"Sec. III.D, Figs. 4(c), 8(a)"},{"comment":"The Conclusions state without qualification that 'the first-order phase transition is converted into second one for small values of the average degree K' and that the discontinuity is recovered near K≈9.0. This conflicts with the hedged language in Sec. III.D ('seems', 'suggests', 'need for further investigation'). The conclusions should be explicitly conditioned on the finite-size and order-parameter analysis requested above; as written, the abstract and conclusions assert what the results only suggest.","section":"Sec. IV vs Sec. III.D"},{"comment":"The extraction of y1 and y2 is not defined, and the transition points in Figs. 5(b) and 8(b) are reported without error bars, even though the density curves in Fig. 8(a) carry standard errors from g=21 runs. The manuscript should state the criterion used to locate a transition on the ρV(y) curves and provide errors on y1 and y2 (e.g., from a threshold rule or an extrapolation procedure). Without this, the reported reactive-window widths are not quantitatively reproducible.","section":"Appendix A and Figs. 5(b), 8(b)"},{"comment":"The averaging procedure over disorder is unclear. The appendix says one network instance is generated per y value and that the quoted K for the RGG is an average over all y, while Fig. 8(a) reports standard errors from g=21 runs. It is not stated whether the 21 runs at fixed y use the same graph or independently generated graphs, so the error bars may reflect only time/seed fluctuations and not quenched topological disorder. This should be clarified and, ideally, observables should be averaged over several RGG realizations at each (y,K). The issue is particularly relevant at small K, where at K≈3.8 the largest component has sc≈0.05457 and the network is not percolated.","section":"Appendix A, Sec. III.D"}],"minor_comments":[{"comment":"The caption labels two panels as '(b)'; the RGG panel should be labeled '(c)'.","section":"Fig. 1 caption"},{"comment":"The caption writes 'N = 322, = 642 and 1282'; these should be typeset as 32^2, 64^2, and 128^2 for clarity.","section":"Fig. 2 caption"},{"comment":"The notation alternates between m, µ, and K for the average degree; a short explicit statement that m is the measured value, µ the control parameter for the ERN, and K the reported final average for both networks would avoid confusion.","section":"Sec. II.B and Sec. III.B"},{"comment":"In the definition of ρξ, the list ξ = V, O, CO, N2, CO2 mixes surface species with gas-phase products; this is not incorrect, but a sentence clarifying that N2 and CO2 are counted only as produced molecules, not as surface densities, would improve readability.","section":"Eq. (8) and surrounding text"},{"comment":"The snapshots present configurations for µ=2.0, 3.0, and 9.0, but the text says the active window begins at µ≈2.1; it would be helpful to state explicitly that the µ=2.0 case is an absorbing state, as the reader must infer this from ρV=0.","section":"Sec. III.C, Fig. 7"}],"recommendation":"major_revision","confidential_remarks":"The manuscript's main selling point is the order conversion on the RGG, and I believe this is potentially publishable if substantiated. The current evidence is too weak: a single-size visual smoothness argument cannot distinguish a continuous transition from rounding of a first-order transition. I would recommend major revision rather than rejection because the requested analysis (finite-size scaling, histograms, and explicit disorder averaging) is feasible within the scope of the paper and the authors have already assembled the necessary simulation machinery. The citation of the authors' own prior work is modest and not a concern."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Bottom line: this is the first time the Yaldram–Khan model has been simulated on random networks, and the ERN part of the paper is clean and believable. The RGG part, which carries the headline claim, is not yet supported: the order-of-transition change is inferred from smooth density curves at a single system size, with no finite-size scaling, histograms, or Binder cumulants.\n\nWhat is actually new and good. The application of YK dynamics to ERN and RGG surfaces is a legitimate gap-filling contribution, building sensibly on the group's earlier ZGB-on-networks work. The ERN results (Figs. 5, 6) are straightforward: transition order is preserved for μ between about 2 and 9, the reactive window widens with μ, and the onset of activity around μ≈2.1 is a clean observation. The calibration on the hexagonal lattice is standard and the algorithm appendix is detailed enough to reproduce. None of the self-citations are load-bearing; they provide context.\n\nWhere it gets soft. The central conclusion—that the RGG converts the discontinuous upper transition into a continuous one for small K—rests on the visual smoothness of ρ_V(y) at N=128² in Figs. 4(c) and 8(a). A first-order transition in a finite system is rounded over a window that shrinks with system size; the smooth approach to y2 for K≈4.5–6.0 could be exactly that rounding rather than a genuine second-order transition. The authors themselves hedge in Sec. III D ('seems,' 'suggests,' 'need for further investigation'), but the Conclusions restate the conversion as established fact. That is a mismatch. The non-percolation at K≈3.8 is a minor secondary issue: by K≈4.5 the largest component holds about 81% of sites, so the active-phase data are on a percolated network.\n\nIf the RGG claim survives a proper Binder-cumulant or histogram analysis (with two or three larger sizes), it is a nice result: topology as a controller of transition order in a canonical nonequilibrium model. As it stands, it is a well-framed suggestion, not a demonstration.\n\nWho this is for: the ZGB/YK and nonequilibrium-surface-reaction crowd, and people interested in how network topology alters absorbing-state phase transitions. The paper deserves a serious referee, but the referee should insist on finite-size scaling before the order-conversion claim can be accepted. I would not cite the order-conversion result until that is done.","headline":"YK model on random networks is new and the ERN results are solid, but the RGG order-conversion claim is visually inferred, not demonstrated.","tokens_in":12245,"tokens_out":2518,"would_cite":false,"duration_ms":24040,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["82C26","82C20","05C80"],"pacs":[],"model":"deepseek-v4-flash","headline":"The paper claims that the spatial structure of randomness on a catalytic surface controls whether the Yaldram–Khan model's discontinuous phase transition survives: random geometric graphs convert it to continuous at low connectivity…","keywords":["Yaldram–Khan model","catalytic surface reactions","phase transition order","random geometric graph","Erdős–Rényi network","nonequilibrium steady state","Monte Carlo simulation","reactive window"],"falsifier":"Compute Binder cumulants of the order parameter or the probability histogram of ρV for the RGG at K ≈ 4.5 across several system sizes, say N = $64^{2}$, $128^{2}$, and $256^{2}$: if the histogram is bimodal or the cumulant minimum deepens with N, the upper transition is discontinuous, contradicting the paper's claim; if the cumulant curves cross consistently, the second-order assignment holds.","tokens_in":11340,"feed_emoji":"🧪","tokens_out":7959,"duration_ms":63729,"temperature":0.7,"pith_summary":"The paper asks whether the kind of randomness used to model a real catalyst surface can change the phase diagram of the Yaldram–Khan (YK) reaction, a standard Monte Carlo model of NO + CO conversion. On a regular hexagonal lattice the model has a reactive window bracketed by one continuous and one discontinuous phase transition. Simulating the same reaction on two random network surfaces, the authors report that an Erdős–Rényi network preserves both transitions for every average degree studied, while a random geometric graph converts the discontinuous transition into a continuous one at small average degree, recovering the jump only near K ≈ 9. This matters because the order of the transition controls whether the catalytic window appears abruptly or gradually on disordered surfaces.","feed_headline":"Random geometric surfaces soften a catalytic reaction's abrupt switch","feed_subtitle":"Spatially constrained randomness turns the YK model's upper phase transition continuous at small average degree.","key_machinery":"The central objects are two random network ensembles used as catalytic surfaces: the Erdős–Rényi network, in which every pair of sites is connected with fixed probability, giving long-range randomness; and the random geometric graph, in which sites are placed uniformly in a square and connected when closer than a radius, giving spatially constrained randomness. The order parameter is the steady-state density of vacant sites ρV as a function of the CO adsorption rate y, and the reactive window y2 − y1 between the two transition points is used to quantify the active phase. The paper reads the order of each transition off the shape of ρV(y): a sharp drop marks a discontinuous transition, a smooth bend a continuous one.","core_discovery":"The paper's central claim is that how randomness is placed on the catalytic surface determines whether the reaction's phase transitions keep their character. On an Erdős–Rényi network, where edges connect sites without regard to spatial distance, the YK model retains one continuous and one discontinuous transition for all average degrees studied (μ from about 2 to 9), with the reactive window beginning near μ ≈ 2.1. On a random geometric graph, where edges only join sites within a fixed radius, the upper (originally discontinuous) transition appears to become continuous for small average degree (reactive window starting between K ≈ 3.8 and 4.5), and the discontinuous character returns only for K ≈ 9.0. The same steady-state Monte Carlo protocol, with NO dissociation rate rNO = 1 and density of vacant sites as the order parameter, is used throughout, so the contrast is attributed to topology.","pith_inferences":["Inference: The claim of a converted transition order would be much stronger with finite-size scaling; a Binder-cumulant or order-parameter-histogram study across N = 64^2, 128^2, and 256^2 at fixed K would confirm whether the smooth ρV curve is a true second-order transition or finite-size rounding.","Inference: Because the small-K RGG is not fully percolated (sc ≈ 0.05 at K ≈ 3.8), the effective catalytic surface is fragmented; the observed continuous upper transition may be an effect of disconnected reaction clusters rather than of local geometric randomness per se, and a study restricting dynamics to the giant component could separate the two.","Inference: Interpolating between the RGG and the ERN by gradually rewiring edges while keeping the average degree fixed could isolate whether it is spatial locality or network fragmentation that suppresses the discontinuous transition.","Inference: A direct experimental analogue would be a catalytic pellet whose pore connectivity is locally constrained rather than globally random; the model predicts the former should show smoother poisoning and ignition curves at low connectivity."],"forward_implications":["For an Erdős–Rényi surface, the YK model's two transitions remain first and second order for every average degree studied, so long-range randomness alone does not change the nature of the transitions.","For a random geometric surface at small average degree (K below about 9), the upper transition is claimed to become continuous, so the catalyst would enter and leave the reactive state without a discontinuous jump.","The reactive window widens with average connectivity on both networks, and for the RGG it begins to open between K ≈ 3.8 and 4.5.","If the two continuous transitions persist for rNO < 1, the model would exhibit two lines of second-order transitions in the (rNO, y) plane, a rare phase-diagram structure.","The contrast between the two networks implies that spatially localized randomness, rather than randomness in general, is what alters the order of the phase transition in this model."],"supporting_citations":[{"why":"Introduces the YK model on a regular hexagonal lattice, providing the baseline phase diagram that the random-network results are compared against.","marker":"[16]"},{"why":"Introduces the ZGB reaction model, the framework from which the YK model extends with NO dissociation and N2 formation.","marker":"[9]"},{"why":"Supplies the reference values y1 = 0.1725 and y2 = 0.3545 used to locate the continuous and discontinuous transition points.","marker":"[47]"},{"why":"Extends the ZGB model to complex networks, serving as the methodological precedent for simulating catalytic reactions on random topologies.","marker":"[30]"},{"why":"Defines the Erdős–Rényi random graph construction that provides the long-range randomness studied as one of the two surfaces.","marker":"[42]"},{"why":"Defines the random geometric graph, the spatially constrained surface whose transition order is reported to change.","marker":"[45]"},{"why":"Establishes the scaling convention for RGG distances used to set the radius r at fixed site density, so that network properties depend only on r.","marker":"[46]"},{"why":"Shows that inactive impurities can convert the discontinuous transition of the NO–CO reaction into a continuous one, the comparison invoked for the RGG result.","marker":"[26]"}],"fun_headline_variants":["Geometric randomness turns catalytic jump into smooth transition","Spatially constrained randomness flips catalytic transition to continuous","Network geometry, not randomness amount, dictates catalytic transition order","How randomness is placed on catalytic surfaces alters phase transitions"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The assignment of transition order is based on visually judging whether the density curves are smooth or sharp at a single system size (N = $128^{2}$), not on finite-size scaling, Binder cumulants, or order-parameter histograms.","fun_headline_variants_meta":{"raw":{"variants":["Geometric randomness turns catalytic jump into smooth transition","Spatially constrained randomness flips catalytic transition to continuous","Network geometry, not randomness amount, dictates catalytic transition order","How randomness is placed on catalytic surfaces alters phase transitions"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.001722,"raw_usage":{"total_tokens":6780,"prompt_tokens":882,"completion_tokens":5898,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":498,"completion_tokens_details":{"reasoning_tokens":5834}},"tokens_in":498,"tokens_out":5898,"duration_ms":39716,"temperature":1.0,"reasoning_tokens":5834,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-07T10:40:22.889513+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Compute Binder cumulants of the order parameter or the probability histogram of ρV for the RGG at K ≈ 4.5 across several system sizes, say N = $64^{2}$, $128^{2}$, and $256^{2}$: if the histogram is bimodal or the cumulant minimum deepens with N, the upper transition is discontinuous, contradicting the paper's claim; if the cumulant curves cross consistently, the second-order assignment holds.","supporting_citations":[{"cited_title":"The average degree K is calculated","cited_arxiv_id":null,"evidence_quote":"Introduces the ZGB reaction model, the framework from which the YK model extends with NO dissociation and N2 formation."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the reference values y1 = 0.1725 and y2 = 0.3545 used to locate the continuous and discontinuous transition points."},{"cited_title":"Loscar and E","cited_arxiv_id":null,"evidence_quote":"Extends the ZGB model to complex networks, serving as the methodological precedent for simulating catalytic reactions on random topologies."},{"cited_title":"Vilela, H.A","cited_arxiv_id":null,"evidence_quote":"Defines the Erdős–Rényi random graph construction that provides the long-range randomness studied as one of the two surfaces."},{"cited_title":"Seshadhri, T.G","cited_arxiv_id":null,"evidence_quote":"Establishes the scaling convention for RGG distances used to set the radius r at fixed site density, so that network properties depend only on r."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Shows that inactive impurities can convert the discontinuous transition of the NO–CO reaction into a continuous one, the comparison invoked for the RGG result."}],"review_version":1}