{"id":"c74a9051-00ae-4b61-941c-aff01d916b9d","arxiv_id":"1908.07334","paper_version":1,"verdict":"REJECT","confidence":"HIGH","novelty_score":4.0,"correctness_risk":"high","formal_verification":"none","parameter_count":2,"one_line_summary":"The paper proposes closed-form lower and tighter upper bounds for relative network delay in energy-harvesting networks, but the lower-bound derivation uses an invalid independence assumption and the upper bound relies on a simulation-fitted constant.","lead":"Micro-energy-harvesting wireless nodes switch on and off frequently, so sending a message can take minutes. This paper claims to derive bounds on that delay per unit distance, but the main lower bound rests on a geometric assumption that is false, and the upper bound is calibrated from the same simulations used to verify it.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Lemma 4.1's independence assumption is false; Eq. (10) is violated for n=3, so Theorem 5.1's lower bound is unsupported.","rationale":"The reader's weakest assumption correctly identifies the load-bearing flaw: Lemma 4.1's probabilistic independence of projected vertices is invalid for connected components of a square lattice. My independent analysis strengthens this from 'unproved' to 'false': for n=3, the exact small-p conditional probability of diameter 2 is 1/3, exceeding the claimed bound of 1/4. Since the lower bound theorem directly uses the expected-diameter bound derived from Lemma 4.1, the central advertised result is not supported. Lemma 3.2 also has unresolved inequalities, but the Lemma 4.1 failure is sufficient and more decisive. The upper-bound theorem (Theorem 5.2) may survive, but it is not the paper's main claimed contribution; the simulation-based kappa also introduces circularity for the numeric upper bound. No new concern changes the reader's verdict, so I recommend keeping REJECT.","tokens_in":13996,"tokens_out":7948,"duration_ms":85042,"concrete_test":"Analytically verify the n=3 counterexample: in the square lattice, enumerate the six pairs of incident edges at a vertex; two pairs are collinear (diameter 2) and four are perpendicular (diameter 1). Each pair is an exact size-3 component with probability p^2(1-p)^8 (two occupied edges, eight boundary edges absent). For p sufficiently small, the conditional probability that a size-3 component is straight is 2/(2+4)=1/3. Compare with Eq. (10): p_2(3) <= C(2,2)(1/2)^2 = 1/4. Since 1/3 > 1/4, Lemma 4.1 is false. If this enumeration is correct, the proof of Theorem 5.1 collapses and the lower bound needs a different argument.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The advertised lower bound (Theorem 5.1) depends on Lemma 4.1, which bounds P(D(C_n)=k). The proof assumes that, after randomly choosing one vertex of an n-vertex connected component, each of the other n-1 vertices independently has its horizontal projection equal to the chosen vertex's x-coordinate with probability 1/2. This conditional independence is false: the component's geometry fixes all projections. A straight horizontal component of n vertices has diameter n-1 with probability 1, not (1/2)^{n-1}, so Eq. (10) is not a valid upper bound. Concretely, for n=3 the bound gives p_2(3) <= C(2,2)(1/2)^2 = 1/4. But among all length-2 paths in the square lattice, two configurations per vertex are straight (diameter 2) and four are L-shaped (diameter 1); each exact size-3 component of this type occurs with probability p^2(1-p)^8, so for sufficiently small p the conditional probability of diameter 2 is 2/(2+4)=1/3 > 1/4. Thus Lemma 4.1 is false, and the expected-diameter upper bound used in Theorem 5.1 is unsupported. The lower-bound formula may be true for other reasons, but the proof given does not establish it.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies the relative network delay γ(λ), defined as the asymptotic ratio of end-to-end delay to Euclidean distance, in a micro-energy harvesting wireless network modeled as a random connection model with binary on/off energy states. The authors couple the network to a square lattice, derive upper bounds on connected-component size and diameter, and from these obtain a closed-form lower bound on γ(λ) (Theorem 5.1). They also propose an upper bound γ(λ) ≤ γ(λ_L) = κ E[T(e)] (Theorem 5.2), claiming it is tighter than the bound of Zhao et al. Simulation results for λ ∈ [1.4, 2.8] are presented to support the bounds.","tokens_in":14299,"tokens_out":5647,"duration_ms":57719,"significance":"If the main theorems were correct, the paper would supply a parameter-free closed-form lower bound on relative network delay, a quantity for which prior work gave only approximations, and it would improve the upper bound of prior art. The problem is timely and the lattice-coupling approach is a reasonable strategy. However, the central diameter lemma is demonstrably false, so the lower-bound theorem is not established, and the numerical upper bound depends on empirically measured constants that are also used for validation. The paper provides no machine-checked proofs or reproducible code, and the theoretical contribution is therefore not currently supported.","major_comments":[{"comment":"Lemma 4.1 is false. The proof assumes that, after randomly choosing one vertex of an n-vertex connected component, the horizontal projections of the other n−1 vertices are conditionally independent Bernoulli(0.5) relative to the chosen vertex's x-coordinate. In a lattice-connected component the geometry fixes these projections. For a straight horizontal component of n vertices, D(C_n(L)) = n−1 with probability 1, while Eq. (10) with k = n−1 evaluates to C(n−1, n−1)(1/2)^(n−1) = 2^(−(n−1)), which is less than 1 for n > 1. Thus Eq. (10) is not a valid upper bound on the conditional probability of a given diameter.","section":"Section IV, Lemma 4.1, Eq. (10)"},{"comment":"The proof of Theorem 5.1 relies directly on Eq. (15), which is obtained by summing the false bound of Lemma 4.1. Since Lemma 4.1 is invalid, Eq. (15) is not an upper bound on E(D_g(λ)), and consequently the closed-form lower bound in Eq. (14) is unsupported. The lower-bound formula may be true for other reasons, but the proof given in the manuscript does not establish it.","section":"Section V.A, Theorem 5.1, Eq. (14)"},{"comment":"The derivation of the component-size bound is not rigorous. In the proof, the quantities s_i^(n−1) differ from group to group, yet the argument factors a single s_i^(n−1) out of the sum and later replaces it by its maximum 2n. The assertion that an n-size component generated from C_i^(n−1) must contain at least (s_i^(n−1) + 1) neighboring vertices is also stated without proof. Since the resulting bound ¯p_n is used in Theorem 5.1, this gap is load-bearing.","section":"Section III, Lemma 3.2"},{"comment":"The numerical upper bound γ(λ) ≤ γ(λ_L) = κ E[T(e)] uses the values κ ≈ 1.7 and λ_L = 1.44, which are obtained from simulations, and the same simulations are then used to validate the bound. The paper does not derive κ or λ_L theoretically. Consequently, the claimed tightness of the upper bound in Fig. 3(b) is a partly circular empirical comparison rather than a parameter-free theoretical prediction.","section":"Section VI.B, Theorem 5.2"}],"minor_comments":[{"comment":"The phrase 'alternative renewal process' should be 'alternating renewal process'.","section":"Section II.B"},{"comment":"The text 'we tested 103 pairs of nodes in total' should read '10^3 pairs' (or '1000 pairs'), since 100 pairs repeated 10 times gives 1000 trials.","section":"Section VI"},{"comment":"The combination symbol C^a_{n−1} is used without a definition; it should be defined explicitly as the number of ways to choose a objects from n−1.","section":"Section IV"},{"comment":"The statement that each vertex has 'zero to three' non-connected neighboring vertices appears inconsistent with a square lattice, where a vertex has four neighbors; please clarify whether oriented or boundary effects are intended.","section":"Section III, Definition 7"},{"comment":"In Fig. 3(b), the legends 'Lower bound [20]' and 'Upper bound [20]' appear to refer to the bounds from reference [14] rather than from Liggett's ergodic theorem [20]; please correct the citations.","section":"Figs. 2 and 3"}],"recommendation":"reject","confidential_remarks":"The paper addresses a relevant problem and is clearly written, but the central lower-bound theorem rests on a demonstrably false combinatorial lemma, and the upper-bound tightness is calibrated on the validating simulations. Replacing the component-diameter argument would require a substantially new proof, which is beyond the scope of a revision. I therefore recommend rejection."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Read this one for the upper bound, not the lower bound. The new inequality gamma(lambda) <= gamma(lambda_L) genuinely improves the earlier bound from [14] (kappa*sqrt(lambda/lambda_L)*E[T(e)]), and the authors deserve credit for identifying that a closed-form expected cluster size was missing in [14]. The lattice-coupling idea is sensible, and the simulations, though minimal, do show the claimed monotonic behavior.\n\nBut the main advertised result—Theorem 5.1's closed-form lower bound—is not supported. The proof of Lemma 4.1 assumes that after picking one vertex of an n-vertex connected component, the other n-1 vertices project onto the chosen vertex's x-coordinate independently with probability 1/2. That is false. The component's geometry fixes the projections. The concrete counterexample from the stress-test note is correct: for n=3, the lemma's upper bound on P(D=2) is C(2,2)(1/2)^2 = 1/4, but among the two straight and four L-shaped configurations, exactly two give diameter 2, so the conditional probability is 1/3. Equation (10) is therefore not a bound, and the lower-bound theorem inherits the error.\n\nOther soft spots are secondary. The numeric upper bound uses kappa ≈ 1.7 estimated from the same simulations that are used to validate it, which makes the tightness claim partly circular. The infinite sums in (14) are not proved to converge. Lemma 3.2's recurrence also leans on an unproved inequality about neighboring-vertex counts. None of these matter as much as the false lemma.\n\nThe citation pattern is clean; the paper builds directly on [14] and [22], and the relation of the lower bound to Lemma 5.1 of [14] is disclosed.\n\nWho this is for: readers working on percolation-inspired delay bounds for energy-harvesting networks, and even they should treat Theorem 5.1 as unproven until a correct diameter bound appears. The upper-bound theorem might be salvageable and deserves a closer look in a revision.\n\nFor peer review: I would not send this out as is. The counterexample to Lemma 4.1 is decisive. Desk reject with an invitation to resubmit after fixing the diameter argument would be the proportionate editorial decision.","headline":"Lemma 4.1's false independence assumption sinks the lower-bound theorem; only the upper bound is worth a second look.","tokens_in":14805,"tokens_out":3101,"would_cite":false,"duration_ms":31926,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":false},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"The paper derives a closed-form lower bound and a tighter upper bound on the delay-to-distance ratio in micro-energy harvesting wireless networks.","keywords":["micro-energy harvesting wireless networks","relative network delay","random connection model","percolation theory","first-passage time","connected component diameter","energy harvesting on-off model","delay bounds"],"falsifier":"For a straight horizontal connected component of $n$ vertices, the actual diameter is $n-1$ with probability 1. Lemma 4.1's right-hand side for $n=3$ and $k=2$ is $1/4$, so Eq. (10) would assert $1\\le 1/4$ for that component; a reader can check whether the inequality holds for such straight components, and if it fails, the closed-form lower bound in Theorem 5.1 is not established. A simulation comparing empirical diameter distributions of lattice components with Eq. (10) would give the same answer.","tokens_in":13777,"feed_emoji":"🔋","tokens_out":8164,"duration_ms":77131,"temperature":0.7,"pith_summary":"The paper studies micro-energy harvesting wireless networks, in which each node is often off while it harvests energy, so a packet can wait many time slots at each hop before a usable link appears. It defines the relative network delay as the limit of end-to-end delay divided by Euclidean distance, and argues that this limit lies between a closed-form lower bound and a tighter upper bound than previously available. The lower bound is built by coupling the wireless network to a square lattice, bounding the probability that a lattice connected component has a given size and diameter, and converting the expected diameter into a delay floor. The value of the result, if correct, is that delay becomes a predictable function of node density, harvesting rate, and per-hop waiting time rather than an object that needs simulation.","feed_headline":"Delay in energy-harvesting wireless nets gets closed-form bounds","feed_subtitle":"A formula ties per-hop waiting time to node density, giving designers a way to predict end-to-end latency.","key_machinery":"The machinery is a square-lattice overlay: each lattice edge is occupied with probability $p=1-e^{-\\lambda\\sqrt g \\pi r_0^2/4}$, and occupied-edge components stand in for wireless clusters. Lemma 3.2 bounds the size distribution $\\Pr\\{S=n\\}$ through a recurrence in the number of neighboring vertices, producing $\\bar p_n$; Lemma 4.1 bounds the probability that an $n$-vertex component has horizontal diameter $k$ by treating the projections of the component's vertices as independent fair coin flips. Combining those two bounds yields the expected diameter that appears in the lower bound, while Theorem 5.2 obtains the upper bound by thinning the network to density $\\lambda_L$ and routing through the long-term giant component.","core_discovery":"On the paper's own terms, the discovery is a closed-form expression for how much delay per unit distance a packet must suffer in an energy-harvesting network. For node density $\\lambda$ between the long-term critical density $\\lambda_L$ and the instantaneous critical density $\\lambda_I$, the relative network delay $\\gamma(\\lambda)=\\lim_{d(u,v)\\to\\infty}T_\\lambda(u,v)/d(u,v)$ satisfies the closed-form lower bound $$\\gamma(\\$\\lambda$) \\ge \\frac{1}{\\bigl(\\sum_{n=2}^\\infty \\bar p_n \\sum_{k=1}^{n-1} k \\sum_{a=k}^{n-1} \\binom{n-1}{a}(1/2)^{n-1}(1/k)^{a-k}+1\\bigr) r_0},$$ where $\\bar p_n$ bounds the probability that a lattice connected component has $n$ vertices, and the upper bound $\\gamma(\\lambda)\\le\\gamma(\\lambda_L)=\\kappa E[T(e)]$, with $\\kappa\\approx1.7$ and $E[T(e)]=1/g-1$. The lower bound is the closed-form replacement for the approximated cluster size used in earlier work, and the upper bound removes a factor $\\sqrt{\\lambda/\\lambda_L}$ from the previous bound. The paper claims these bounds hold in the random connection model with binary on/off energy harvesting and reports simulations consistent with them.","pith_inferences":["A direct extension would replace Lemma 4.1's coin-flip independence with exact enumeration of connected lattice-animal projections; that would either repair the lower bound or show that only its constant changes.","The lattice-diameter method could transfer to other duty-cycled or intermittently available wireless networks, but only if the projection independence is verified for those topologies.","The binary on/off energy model ignores correlation in a node's availability across time slots; under a queued energy-storage model the per-hop waiting time would become history-dependent and the bounds would need a queueing correction.","Even if the closed-form constant is not tight, the qualitative inverse relation between relative delay and $r_0$ times expected component diameter may survive, so the paper's structure is a plausible template for stronger bounds."],"forward_implications":["Network designers can compute a guaranteed delay floor for a given node density and harvesting rate directly from the closed-form sum, without simulating the network.","The upper bound says that above the long-term critical density, relative delay is no larger than the delay at that critical density, so density increases beyond $\\lambda_L$ only help until the network percolates instantaneously.","The bounds separate the energy-harvesting contribution, the per-hop waiting time $E[T(e)]=1/g-1$, from the topological contribution encoded in the geometric constants.","At high harvesting probability the upper bound collapses, matching the simulation observation that the network percolates in the first time slot and relative delay approaches zero."],"supporting_citations":[{"why":"Supplies the cluster-based lower-bound framework and the previous approximated bound that this paper tightens.","marker":"[14]"},{"why":"Supplies percolation theory, defining giant components and critical densities used throughout.","marker":"[8]"},{"why":"Supplies first-passage time as the model for end-to-end delay.","marker":"[9]"},{"why":"Supplies the Poisson point process model for node placement.","marker":"[15]"},{"why":"Supplies the thinning theorem used to build the sparser network and the upper-bound argument.","marker":"[21]"},{"why":"Supplies the limit result used to turn averaged cluster distances into an expectation in Lemma 5.1.","marker":"[22]"},{"why":"Supplies the existence of a finite-distance connection from any node to the giant component of the thinned network.","marker":"[23]"},{"why":"Supplies the result that expected hop count between finite-distance nodes in a percolated network is finite.","marker":"[24]"}],"fun_headline_variants":["Closed-form bounds set for energy-harvesting wireless delay","Per-hop delay bounds now tight for harvest-powered networks","Energy-harvesting nets: delay per distance bounds found","Tight delay bounds for self-powered wireless without batteries","Closed-form delay bounds derived for micro-energy harvesting"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The central derivation depends on treating the horizontal positions of the vertices in a connected component as independent fair coin flips relative to one selected vertex; in a valid lattice-connected component those positions are not independent, so the diameter bound feeding the lower-bound formula is not guaranteed to be a true upper bound.","fun_headline_variants_meta":{"raw":{"variants":["Closed-form bounds set for energy-harvesting wireless delay","Per-hop delay bounds now tight for harvest-powered networks","Energy-harvesting nets: delay per distance bounds found","Tight delay bounds for self-powered wireless without batteries","Closed-form delay bounds derived for micro-energy harvesting"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000789,"raw_usage":{"total_tokens":3514,"prompt_tokens":1015,"completion_tokens":2499,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":631,"completion_tokens_details":{"reasoning_tokens":2420}},"tokens_in":631,"tokens_out":2499,"duration_ms":22207,"temperature":1.0,"reasoning_tokens":2420,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T14:02:21.490586+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"For a straight horizontal connected component of $n$ vertices, the actual diameter is $n-1$ with probability 1. Lemma 4.1's right-hand side for $n=3$ and $k=2$ is $1/4$, so Eq. (10) would assert $1\\le 1/4$ for that component; a reader can check whether the inequality holds for such straight components, and if it fails, the closed-form lower bound in Theorem 5.1 is not established. A simulation comparing empirical diameter distributions of lattice components with Eq. (10) would give the same answer.","supporting_citations":[{"cited_title":"Fundamental relat ionship between nodedensity and delay in wireless ad hoc networks wi th unreliable links,","cited_arxiv_id":null,"evidence_quote":"Supplies the cluster-based lower-bound framework and the previous approximated bound that this paper tightens."},{"cited_title":"Meester, Continuum percolation","cited_arxiv_id":null,"evidence_quote":"Supplies percolation theory, defining giant components and critical densities used throughout."},{"cited_title":"First-passage percolati on, sub- additive processes, stochastic networks, and generalized renewal theory,","cited_arxiv_id":null,"evidence_quote":"Supplies first-passage time as the model for end-to-end delay."},{"cited_title":"The poisson point process,","cited_arxiv_id":null,"evidence_quote":"Supplies the Poisson point process model for node placement."},{"cited_title":"Penrose, Random geometric graphs","cited_arxiv_id":null,"evidence_quote":"Supplies the thinning theorem used to build the sparser network and the upper-bound argument."},{"cited_title":"Node density and delay in large-sca le wireless networks with unreliable links,","cited_arxiv_id":null,"evidence_quote":"Supplies the limit result used to turn averaged cluster distances into an expectation in Lemma 5.1."},{"cited_title":"Distributed energy management al gorithm for large-scale wireless sensor networks,","cited_arxiv_id":null,"evidence_quote":"Supplies the existence of a finite-distance connection from any node to the giant component of the thinned network."},{"cited_title":"Latency of wir eless sensor networks with uncoordinated power saving mechanism s,","cited_arxiv_id":null,"evidence_quote":"Supplies the result that expected hop count between finite-distance nodes in a percolated network is finite."}],"review_version":1}