{"id":"4ce56554-55a9-4c68-9048-e8096b28d6e5","arxiv_id":"1908.01726","paper_version":1,"verdict":"REJECT","confidence":"MODERATE","novelty_score":4.0,"correctness_risk":"high","formal_verification":"none","parameter_count":0,"one_line_summary":"The paper derives an effective capacity formula for an energy-harvesting transmitter, but the formula is not supported because the service process is not a standard Markov-modulated process.","lead":"This paper develops analytical formulas for energy overflow, energy outage, and throughput in an energy-harvesting wireless transmitter with a data buffer. The headline result, an effective capacity formula under QoS constraints, does not follow from the stated Markov-modulated model because the outage service depends on the previous battery state.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Theorem 1 applies the Markov-modulated spectral-radius formula to a battery state process that is not Markov, and the state-0 service MGF is history-dependent; the central effective-capacity result lacks support.","rationale":"The reader's weakest assumption correctly identified that the service MGFs are not state-homogeneous, but the more fundamental defect is that the modulating process W itself is not a Markov chain: the transition out of state m depends on the residual energy, which is not encoded in W. The paper's q_m in (10) is a survival ratio from an empty battery, not a Markov transition probability, so the entire Leslie-matrix spectral-radius machinery in Appendix C lacks a valid modulating Markov process. Even if one attempted to repair this by augmenting the state with residual energy, the service distribution in the outage state would depend on the previous state, so the simple state-indexed MGFs φ0 and φj would still be invalid. The absence of any simulation-based validation of the effective capacity in Section V means the paper's main advertised contribution is not independently supported. The energy-overflow approximation and average-service-rate derivations may have merit, but the central Theorem 1 is load-bearing and its proof does not hold together. The reader's REJECT verdict is therefore appropriate; I do not see a need to change it.","tokens_in":22628,"tokens_out":6360,"duration_ms":70744,"concrete_test":"Monte Carlo simulate the exact system of Section II with the Section V parameters (i.i.d. Weibull energy arrivals, constant demand p, AWGN or Rayleigh channel) for a long horizon T. First estimate P(W(i+1)=m+1 | W(i)=m) for m=0,...,M and compare with q_{m+1} from (10); if the estimates differ, or if they depend on the residual-energy level within state m, W is not Markov. As a direct end-to-end check, estimate the effective capacity from the simulated service process as −1/(Nθ) ln E[e^{−θS(t)}] for large t and compare with the root of (27); a mismatch beyond Monte Carlo error falsifies Theorem 1.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim is Theorem 1, which computes CE from the spectral radius of the Leslie-type matrix M Φ(−θ) via [36, Example 7.2.7]. That example requires the modulating process to be Markov and requires E[e^{θs(i)} | W(i)=x] to depend only on the current state x. Neither condition is met. First, W(i), the 'consecutive number of times energy demand was met since last outage,' is not Markov: from state m, surviving one more frame requires residual stored energy E(i)+u(i+1) ≥ p(i+1), and E(i) is not fixed by W(i)=m; different paths into state m carry different reserves. Equation (10) computes the retrospective ratio Pr{survive m frames}/Pr{survive m−1 frames} starting from an empty battery, not the transition probability Pr{W(i+1)=m | W(i)=m−1} for an arbitrary visit to state m−1. Hence the transition matrix M in (11) and the steady-state π do not describe the true W process. Second, even granting W Markov, φ0(θ) in (28) is a function of the previous state, not just the current outage state: when W(i)=0, s(i)=g(e(i−1)+u(i)), whose distribution depends on W(i−1); the paper's own expression averages over m. The Markov-modulated MGF must be history-independent. Finally, Section V plots only analytical effective-capacity curves; no simulation of the actual EH system is used to validate (26)–(29). Thus the main advertised formula is unsupported.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies an energy-harvesting (EH) transmitter with a battery and a data buffer. It develops three main results: an exponential approximation for the energy overflow probability characterized by an energy decay rate µ; a Markov-chain model for energy outages based on the number of consecutive successful energy-supply frames, together with an outage-probability expression; and throughput analysis consisting of an average data service rate and an effective-capacity formula. The advertised central contribution is Theorem 1, which expresses the effective capacity as CE(θ,µ) = −1/(Nθ) ln(χ*), where χ* is the unique positive root of the characteristic function z(χ) in Eq. (27), obtained by applying the Markov-modulated spectral-radius formula from [36, Example 7.2.7] to a Leslie-type matrix.","tokens_in":22967,"tokens_out":16758,"duration_ms":188062,"significance":"If Theorem 1 were valid, the paper would provide a useful design tool connecting battery management (energy overflow and outage constraints) with QoS-constrained throughput in EH systems. The overflow approximation and the outage Markov-chain construction are simple and potentially useful, and the average service rate expression in Eq. (21) is derived transparently from the model. Numerical results for the overflow and outage probabilities and for the average service rate are also provided. However, the central effective-capacity result is not supported, because the service process does not satisfy the state-homogeneous Markov-modulation condition required by the cited spectral-radius formula. The paper's main advertised contribution is therefore not established, and its significance is correspondingly reduced.","major_comments":[{"comment":"The application of [36, Example 7.2.7] is not justified. In the model, when the battery enters state 0 at time i, the service is s(i) = g(ξ(i)) (when the channel supports transmission), where ξ(i) = e(i−1) + u(i) includes the residual energy accumulated during the preceding run of successful frames. The distribution of ξ(i) therefore depends on the length m of that run, i.e., on the previous state W(i−1) = m, not only on the current state W(i) = 0. The paper's φ0(θ) in Eq. (28) is an average over m, and replacing every transition into state 0 by this common average in the matrix M Φ(−θ) does not produce the correct Markov additive kernel A_{m0}(θ) = (1 − q_{m+1}) E[e^{θs(i)} | W(i−1)=m, W(i)=0]. The spectral radius of M Φ(−θ) need not equal the asymptotic log-moment generating function of the cumulative service process S(t). Consequently, the effective-capacity expression in Eq. (26) is unsupported; the proof must either derive the transition-dependent spectral radius or prove that the averaged φ0 yields the same spectral radius, which is not done.","section":"Appendix C / Theorem 1 (Eqs. (26)–(30))"},{"comment":"The effective-capacity curves in Figs. 5 and 6 are computed from the very formula in Theorem 1 that is in question, and no independent simulation of the actual EH system is provided to validate Eqs. (26)–(29). Given the concern in the previous major comment, these figures cannot serve as confirmation of the result. The authors should simulate the EH transmitter with a finite data buffer, measure the empirical buffer overflow decay rate or the maximum sustainable constant arrival rate, and compare those data with the theoretical CE(θ,µ).","section":"Section V-C / Figs. 5 and 6"}],"minor_comments":[{"comment":"The paper discusses two different stability regimes: E[u] < E[p] is used for the overflow analysis, while E[u] > E[p] is mentioned for avoiding battery depletion. The text should state explicitly which stability condition is assumed for which result, since the outage Markov chain's positive recurrence and the expressions for π0 depend on this choice.","section":"Definition 1 and footnote 3"},{"comment":"The upper bound on the outage probability relies on the monotonicity q1 ≤ q2 ≤ q3 ≤ ···, which is asserted via an inductive method but not proved. A short proof or a reference establishing this monotonicity for i.i.d. energy arrivals and demands would make the bound self-contained.","section":"Eq. (12)"},{"comment":"There is a typographical error in the display of φj(θ): an extraneous closing brace appears in E[exp(θr(i)1[r(i)≤I(i)]}|w(i)=j]. This should be corrected to E[exp(θr(i)1[r(i)≤I(i)])|w(i)=j].","section":"Eq. (38)"},{"comment":"The caption refers to 'different data transmission delay constraints, i.e., θp', but the notation θp is not defined anywhere; the figure actually varies the QoS exponent θ. Please correct the caption and define the parameter consistently.","section":"Figure 5 caption"}],"recommendation":"reject","confidential_remarks":"The overflow, outage, and average-service-rate parts contain useful material and could form the basis of a revised submission. The reason for rejection is the flawed proof of Theorem 1, which is the paper's main advertised contribution; this is a technical error rather than a disagreement with consensus. If the authors can re-derive the effective capacity using a correct Markov additive model (or prove that the averaged φ0 is legitimate) and support the result with independent simulation, a resubmission would be worth considering."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The paper has a solid core in the energy overflow part: the exponential approximation via large deviations is standard and cleanly presented, and the examples in Propositions 1 and 2 are instructive. The related work is thorough, the writing is clear, and there is no circular fitting or invented data. The average service rate expression (21) follows from the model as a conditional-expectation exercise, though it inherits the problems below.\n\nThe central advertised result, Theorem 1, is not supported. The battery state w(i) — consecutive successful energy supplies since last outage — is not Markov. Knowing w(i)=m does not fix the residual stored energy e(i), only that the last m demands were met. The transition to state m+1 depends on e(i), which varies across visits to state m. The paper's q_m in (10) computes the ratio of survival probabilities starting from an empty battery, not the transition probability from an arbitrary state m-1. So the transition matrix M and the steady-state probabilities π do not describe the actual process. This invalidates the outage probability formula and the whole Markov-modulated setup for the effective capacity.\n\nEven if the state process were Markov, the service MGF φ0(θ) in (28) depends on the previous state m, not just the current state 0 — the paper's own formula averages over m. That violates the state-homogeneity condition of Chang's Example 7.2.7, which Theorem 1 invokes. The spectral-radius formula for effective capacity therefore does not apply.\n\nThe upper bound on outage probability also relies on q1 ≤ q2 ≤ q3 ≤ …, which is false in general. Simple example: i.i.d. energy arrivals with u=2 w.p. 1/2 and u=0 otherwise, with constant demand p=1, gives q2 = 1.5 > q1 = 0.5 but q3 = 2/3 < q2. So the inequality used in (12) is unsupported.\n\nThe numerical section plots only analytical effective capacity curves; there is no system simulation validating (26)–(29). Given that the derivation is flawed, that absence is a significant omission.\n\nWho is this for? People working on energy harvesting links with QoS constraints might find the overflow approximation and the general modeling approach useful as a starting point, but they should not rely on the effective capacity formula as derived. I would send this to a serious referee — the flaws are subtle and worth catching — but I would expect a rejection or major revision. The authors could potentially repair the model by enlarging the state to include a quantized energy level, but that is a substantially different paper.\n\nBottom line: cite the overflow part if you need that standard result, but not the effective capacity; and treat the outage probability as heuristic pending a real Markov state description.","headline":"The overflow analysis is fine, but the main effective capacity result rests on a battery-state Markov assumption that doesn't hold; worth refereeing, not worth accepting as is.","tokens_in":23432,"tokens_out":4407,"would_cite":false,"duration_ms":48039,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"This paper derives the maximum constant data arrival rate for an energy-harvesting wireless transmitter as $-\\frac{1}{N\\theta}\\ln(\\chi^\\ast)$, with $\\chi^\\ast$ the unique positive root of a polynomial built from the battery-state Markov…","keywords":["energy harvesting","energy overflow","energy outage","effective capacity","queueing theory","large deviation theory","Markov process","QoS constraints"],"falsifier":"Simulate the exact two-queue recursion with a finite battery, i.i.d. Weibull energy arrivals with shape $k=5$ and scale chosen so the average energy arrival is below the constant demand, Rayleigh fading, and a fixed transmission-rate policy; estimate the data-buffer overflow probability versus buffer threshold, fit its exponential decay rate, and compare the implied QoS exponent with the $\\theta$ obtained from the unique positive root of $z(\\chi)$ in (27). If the two disagree whenever outage histories leave different pre-outage energy levels, most visibly for small batteries, the Markov-modulated spectral-radius formula fails.","tokens_in":22432,"feed_emoji":"🔋","tokens_out":9402,"duration_ms":90088,"temperature":0.7,"pith_summary":"An energy-harvesting transmitter must balance two failure modes: a battery that overflows and wastes harvested energy, and a battery that runs dry and interrupts data transmission. This paper establishes a single analytical bridge between the battery's storage dynamics and the data service process, by representing the battery through a Markov chain whose state counts the frames since the last energy outage and by approximating the energy overflow probability as $\\Pr\\{e \\ge e_{\\mathrm{th}}\\} \\approx \\exp(-\\mu e_{\\mathrm{th}})$ with $\\mu$ the energy decay rate. Its main claim is a closed-form effective capacity, the largest constant data arrival rate that the data buffer can support under a QoS constraint on buffer overflow, given by $-\\frac{1}{N\\theta}\\ln(\\chi^\\ast)$, with $\\chi^\\ast$ the unique positive root of a polynomial built from the battery-state transition probabilities and per-state service statistics. If correct, this gives system designers an analytic way to translate battery size, energy arrival statistics, and outage constraints into a maximum sustainable data rate, instead of relying on simulation of a coupled energy-data queue.","feed_headline":"Battery-state formula sets max data rate for energy harvesters","feed_subtitle":"The maximum sustainable data rate follows from a polynomial root built on the battery's outage history.","key_machinery":"The central object is the battery-state Markov chain $\\mathcal{W}=\\{0,1,\\dots\\}$, in which state $m$ means $m$ consecutive frames have passed since the last energy outage and the transition probability $q_m$ is the probability that the battery satisfies the energy demand for the $m$-th straight frame. The load-bearing identity is the Markov-modulated spectral-radius formula, $\\Lambda(\\theta)/\\theta = \\frac{1}{\\theta}\\ln \\operatorname{sp}\\{M\\Phi(\\theta)\\}$, applied to the data service process, with $M$ the battery-chain transition matrix and $\\Phi(\\theta)$ the diagonal matrix of per-state service moment generating functions $\\varphi_0,\\varphi_1,\\dots$. Because $M\\Phi(-\\theta)$ is a Leslie matrix, its characteristic polynomial $z(\\chi)$ has a unique positive root $\\chi^\\ast$, and the effective capacity is $-\\ln(\\chi^\\ast)/(N\\theta)$. This machinery collapses a two-queue energy-and-data system into a one-dimensional spectral computation.","core_discovery":"The paper claims that, for a point-to-point energy-harvesting link with a battery and a data buffer, the effective capacity under a data QoS exponent $\\theta$ and an energy decay rate $\\mu$ is $C_E(\\theta,\\mu) = -\\frac{1}{N\\theta}\\ln(\\chi^\\ast)$, where $\\chi^\\ast$ is the unique real positive root of the polynomial $z(\\chi)$ in (27). That polynomial is assembled from the transition probabilities $q_n$ of a Markov chain whose state counts the number of frames since the last energy outage, and from per-state moment generating functions $\\varphi_0$ and $\\varphi_j$ of the service process, which describe data served during outages and during stretches of successful energy supply. The derivation models the service process as Markov-modulated and uses the spectral-radius formula for its asymptotic log-moment generating function; the Leslie-matrix form of the resulting matrix guarantees a unique positive root. The paper presents this root as determining the maximum constant data arrival rate that the data buffer can support while keeping its overflow probability near $\\exp(-\\theta d_{\\mathrm{th}})$.","pith_inferences":["If the formula is right, the same Markov-modulated spectral-radius machinery should extend to adaptive transmission policies that react to instantaneous channel state, provided the per-state moment generating functions average over both energy and fading; the paper only treats a fixed transmission rate.","The exponential overflow approximation suggests an explicit Pareto frontier between the probability of wasting energy and the probability of running dry, parameterized by $\\mu$; the paper demonstrates the tradeoff numerically but does not write it as a closed-form curve.","A natural stress test is to replace the constant energy demand with a battery-state-dependent threshold policy and check whether the effective capacity still equals the spectral radius of a Leslie-like matrix; if the service distribution during outages depends on pre-outage energy, the polynomial in (27) may need conditioning on that history."],"forward_implications":["A designer can compute the maximum sustainable data arrival rate directly from the energy arrival statistics, the constant energy demand rate, and the QoS exponent, without running a full battery-and-buffer simulation.","The energy decay rate $\\mu$ becomes a tunable design parameter: smaller $\\mu$ lowers energy outage probability while raising overflow probability, and the effective capacity formula locates the best operating point.","In an AWGN channel the effective capacity is largest when $\\mu$ is as small as possible, whereas in Rayleigh fading there is an interior optimum, because transmission outages and energy outages push in opposite directions.","Truncating the Leslie matrix to $\\alpha$ states gives an upper bound on effective capacity that tightens as $\\alpha$ grows, yielding a finite computable approximation to the infinite-state formula."],"supporting_citations":[{"why":"Supplies the Markov-modulated spectral-radius formula, $\\Lambda(\\theta)/\\theta = \\frac{1}{\\theta}\\ln \\operatorname{sp}\\{M\\Phi(\\theta)\\}$, which is the basis for the effective capacity derivation.","marker":"[36]"},{"why":"Defines the effective capacity as the maximum constant arrival rate under a QoS exponent, the quantity Theorem 1 computes.","marker":"[47]"},{"why":"Provides the Leslie-matrix structure used to write the characteristic polynomial $z(\\chi)$ and to justify the truncation-based upper bound.","marker":"[60]"},{"why":"Supplies Cauchy's theorem guaranteeing that $z(\\chi)$ has a unique positive root, which becomes $\\chi^\\ast$ in the effective capacity formula.","marker":"[61]"},{"why":"Supports the exponential approximation for the overflow probability of a queue with a characteristic decay rate, used for the energy overflow approximation.","marker":"[41]"}],"fun_headline_variants":["Battery outage history yields max data rate","Polynomial root sets max data rate for harvesters","Energy harvester data rate from outage Markov root","Effective capacity via battery outage polynomial","Max data rate linked to energy outage dynamics"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The whole calculation depends on assuming that the data service distribution is fixed by the battery's current state alone, even though after an outage the energy available to serve data can vary with how the battery reached that state.","fun_headline_variants_meta":{"raw":{"variants":["Battery outage history yields max data rate","Polynomial root sets max data rate for harvesters","Energy harvester data rate from outage Markov root","Effective capacity via battery outage polynomial","Max data rate linked to energy outage dynamics"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000226,"raw_usage":{"total_tokens":1508,"prompt_tokens":1026,"completion_tokens":482,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":642,"completion_tokens_details":{"reasoning_tokens":415}},"tokens_in":642,"tokens_out":482,"duration_ms":5368,"temperature":1.0,"reasoning_tokens":415,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T15:08:38.692933+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Simulate the exact two-queue recursion with a finite battery, i.i.d. Weibull energy arrivals with shape $k=5$ and scale chosen so the average energy arrival is below the constant demand, Rayleigh fading, and a fixed transmission-rate policy; estimate the data-buffer overflow probability versus buffer threshold, fit its exponential decay rate, and compare the implied QoS exponent with the $\\theta$ obtained from the unique positive root of $z(\\chi)$ in (27). If the two disagree whenever outage histories leave different pre-outage energy levels, most visibly for small batteries, the Markov-modulated spectral-radius formula fails.","supporting_citations":[{"cited_title":"Chang, Performance Guarantees in Communication Networks","cited_arxiv_id":null,"evidence_quote":"Supplies the Markov-modulated spectral-radius formula, $\\Lambda(\\theta)/\\theta = \\frac{1}{\\theta}\\ln \\operatorname{sp}\\{M\\Phi(\\theta)\\}$, which is the basis for the effective capacity derivation."},{"cited_title":"Effective capacity: A wireless link model for support of quality of service,","cited_arxiv_id":null,"evidence_quote":"Defines the effective capacity as the maximum constant arrival rate under a QoS exponent, the quantity Theorem 1 computes."},{"cited_title":"Leslie matrix models,","cited_arxiv_id":null,"evidence_quote":"Provides the Leslie-matrix structure used to write the characteristic polynomial $z(\\chi)$ and to justify the truncation-based upper bound."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies Cauchy's theorem guaranteeing that $z(\\chi)$ has a unique positive root, which becomes $\\chi^\\ast$ in the effective capacity formula."},{"cited_title":"Effective bandwidths of departure processes from queues with time varying capacities,","cited_arxiv_id":null,"evidence_quote":"Supports the exponential approximation for the overflow probability of a queue with a characteristic decay rate, used for the energy overflow approximation."}],"review_version":1}