{"id":"b9aa028d-c324-4e82-9ba7-a682967d4e8a","arxiv_id":"1908.06388","paper_version":2,"verdict":"REJECT","confidence":"MODERATE","novelty_score":5.0,"correctness_risk":"high","formal_verification":"none","parameter_count":4,"one_line_summary":"A BER analysis and four-case optimization for TDMA molecular communication are presented, claiming joint optimization of molecule counts and slot durations lowers error rates, but the derivation has algebraic and modeling errors.","lead":"This paper proposes a time-division scheme for molecular communication in which tiny transmitters release drug molecules, and it optimizes the number of molecules and the length of each transmitter's time slot to reduce bit errors. A general reader might look here to see whether communication theory can guide drug dosing, but the paper's central derivation has mathematical and modeling errors.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The IUI model in Eq. (7) uses interval arrival differences for a passive counting receiver, which is internally inconsistent and invalidates the derived BER and optimization results.","rationale":"The reader's weakest assumption identifies exactly the same load-bearing concern: Eq. (7) models interference as an interval arrival probability for a receiver that is explicitly passive and counts molecules occupying its volume at the sampling instant. The paper's own Eq. (4) defines Ph as an integral of the position PDF over the receiver volume, so Ph(t) is an occupancy probability. Consequently, the correct contribution of a previous release to the current sample is Ph(elapsed time), not the difference of occupancy probabilities used in Eq. (7). This is not a matter of disagreement with a modeling convention; it is an internal inconsistency between the stated receiver model and the interference statistics. Since the mean, variance, and BER expressions in Eqs. (13)-(21), and therefore all optimization objectives and the reported BER values, are derived from these IUI statistics, the central quantitative claim is not supported. The reader's verdict of REJECT remains appropriate. I also note the variance derivation in Appendix A appears algebraically suspect, but the Eq. (7) modeling error is the primary independent reason the analysis fails.","tokens_in":21292,"tokens_out":3790,"duration_ms":41877,"concrete_test":"Run a Brownian particle simulation of the passive receiver exactly as described in Section II: no absorption, spherical receiver, count molecules inside the receiver at the end of the TX-s slot. Using the Table III SDE parameters, release A_j molecules in an earlier slot and record the empirical mean and variance of the count at the TX-s sampling instant. Compare these with the Eq. (7) prediction A_j[Ph(lambda) - Ph(lambda - t_s)] and with the occupancy prediction A_j Ph(lambda). If the simulated IUI mean tracks Ph(lambda) rather than the interval difference, Eq. (7) is wrong and the claimed BER values cannot be reproduced.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim is that Eqs. (9)-(21) provide the correct mean, variance, and BER for the TDMA-based MCvD system, leading to DTDN achieving BER 3.45e-8. The load-bearing premise is Eq. (7), where interference from a previous release is modeled as Binomial(A_j, Y) with Y = Ph(lambda) - Ph(lambda - t_s). This is inconsistent with the receiver model stated in Section II: node D is a passive receiver that counts molecules inside its volume at the sampling instant, and Ph from Eq. (4) is the integral of the position PDF over the receiver volume, i.e., the occupancy probability at time t. For a molecule released lambda seconds before the sampling instant, the probability that it is inside the receiver at the sampling time is Ph(lambda), not Ph(lambda) - Ph(lambda - t_s). The subtracted form would describe an absorbing receiver that records arrivals during the current slot and removes molecules, but that is not the model. Moreover, Ph is an occupancy probability and is not necessarily monotone in time, so the difference can be negative and cannot serve as a binomial probability. Because every objective in Section III and every BER value in Section VI is built on these IUI statistics, the numerical claims do not follow from the stated model.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The manuscript studies a TDMA-based molecular communication via diffusion system with multiple transmitter nanomachines and one passive receiver, modeling the receiver count as Gaussian and deriving mean, variance, and BER expressions that include inter-user and inter-symbol interference. On this basis it defines four resource-allocation cases (STSN, STDN, DTSN, DTDN) and formulates multi-objective optimization problems solved by weighted-sum scalarization and alternating search with CVX. Numerical evaluations across three diffusion environments report that the dynamic-time dynamic-number (DTDN) approach achieves BER near 3.45e-8 in the severe diffusive environment, while the static approach remains near 1e-2.","tokens_in":21589,"tokens_out":10318,"duration_ms":103293,"significance":"The problem is practically motivated, and the four-case framework is clearly organized; if the analytical channel statistics were correct, the paper would provide a useful quantitative design tool for drug-release management in a TDMA molecular communication system with drift. The paper also contains useful sensitivity checks, such as the IUI-length saturation in Fig. 5 and the real-versus-integer molecule count comparison in Fig. 6(b). However, the central quantitative claims depend entirely on the correctness of the interference statistics and the convexity arguments, and the errors identified below mean that the numerical results in Section VI do not follow from the stated model as written.","major_comments":[{"comment":"The variance of the interference mixture is computed incorrectly. For a zero/one Bernoulli mixture of Binomial(A_j, Y), the unconditional variance is 0.5 A_j Y(1-Y) + 0.25 A_j^2 Y^2. Eq. (44) and hence Eq. (15) give the A_j^2 Y^2 coefficient as 1.25 instead of 0.25, and Eq. (42) contains a sign error in the conditional second moment. Because sigma_{0s}^2 and sigma_{1s}^2 enter the BER expression in Eq. (21) and every objective function in Section III, the numerical results reported in Section VI are not supported by the stated statistics.","section":"§II-B and Appendix A, Eqs. (7), (42)-(45)"},{"comment":"The interference model is inconsistent with the passive receiver described in Section II. The receiver counts molecules inside its volume at the sampling instant, so the probability that a molecule released from TX-j lambda seconds before the sampling instant is counted is the occupancy probability Ph(X_j, lambda), not the interval arrival difference Y = Ph(X_j, lambda) - Ph(X_s, lambda - t_s). The subtracted form would correspond to counting arrivals during the current slot under an absorbing receiver, which is not the stated model. Moreover, Ph is an occupancy CDF and is not guaranteed to be monotone, so the difference can be negative and is not a valid binomial probability. Because every BER and optimization result is built on these IUI statistics, this inconsistency invalidates the quantitative claims of the paper.","section":"§II-B, Eq. (7)"},{"comment":"The convexity arguments supporting the CVX-based solutions are not established. Appendix C's Eq. (54) gives only part of the second derivative of P_e^s with respect to A_s; it omits the second derivatives of the mean and variance terms that arise from the full chain rule, and the sign conclusion depends on an assertion about the ordering mu_{0s} < tau_s < mu_{1s} rather than a proof. Appendix B similarly asserts positivity of the second derivative 'after some manipulations' and contains a derivative g(t_s) in Eq. (50) that is not a correct derivative of Eq. (4). Since claimed convexity is the justification for solving (29), (31), and (33) with CVX, the optimality of the reported solutions is unsupported.","section":"Appendices B and C; §IV-A/B/C"}],"minor_comments":[{"comment":"Eq. (19) uses mu_{0s} and sigma_{0s}^2 in the expression for Pr(x_s[n]=0 | x_s[n]=1); this should be mu_{1s} and sigma_{1s}^2 to be consistent with Eq. (21).","section":"Eq. (19)"},{"comment":"The definition of erf in the Notation paragraph is nonstandard: the standard error function is 2/sqrt(pi) times the integral from 0 to y of exp(-x^2) dx, not 1/sqrt(2 pi) times the integral.","section":"Notation, Section II"},{"comment":"The expression for g(t_s) contains duplicated denominator terms and appears dimensionally inconsistent; if it is intended to support the convexity proof, it should be corrected.","section":"Eq. (50)"}],"recommendation":"reject","confidential_remarks":"The two core derivation errors, Eqs. (7) and (44), are independent and each is sufficient to invalidate the numerical claims as printed. A corrected reanalysis would require re-deriving the interference statistics and repeating the optimization and all numerical experiments, which goes beyond a routine revision. I recommend rejection, though a substantially revised version with corrected statistics could be considered as a new submission."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"You can skip the numerical results in this one. The paper's real contribution—if you can call it that—is a clean taxonomy of TDMA scheduling for molecular communication (static/dynamic time, static/dynamic molecule count) and a sensible formulation of the optimization problems. The idea of jointly choosing slot durations and release counts for multiple transmitters against IUI/ISI is natural and, as far as the reference list shows, new. I give credit for that, and for using external diffusion constants rather than fitting free parameters.\n\nBut the interference statistics, which everything else is built on, are not right. The reader's two concrete catches are valid, and I found a third. First, Eq. (7) models interference as molecules arriving in the current slot, using differences like Ph(λ)−Ph(λ−t_s). The paper explicitly says the receiver is passive and counts molecules inside its volume at the sampling instant. For that receiver, the probability a molecule from an earlier release is counted is Ph(λ), not an interval difference. The difference describes an absorbing receiver. Also, Ph is not monotone in time, so the difference can be negative and cannot be a binomial probability. Second, even within that flawed model, Y^u_{j,s} and H_{j,s} mix the transmitter index: the first term uses Ph(X_j,·), the second uses Ph(X_s,·). That is a straightforward typo, but it means interference from other transmitters is computed with the wrong distance. Third, the variance in Eq. (44) has the wrong coefficient. For a 0/1 mixture of Binomial(A_j,Y), the A_j^2 Y^2 term enters with coefficient +0.25, not +1.25. The derivation in Appendix A drops a sign. I'll add a fourth, smaller issue: Eq. (19) conditions on x=1 but uses μ0,σ0, which contradicts Eq. (21). Should be a typo, but it is in the paper.\n\nBecause every BER and every optimized value depends on these statistics, the numerical claims—DTDN reaching 3.45e-8 and so on—do not follow from the model. The qualitative ranking of the four schemes may be plausible, but the paper offers no independent simulation or event-based verification that would rescue the numbers. The ML threshold is also never specified; the authors point to [27] and move on, so the BER curves are not reproducible.\n\nThe convexity arguments in Appendices B and C are more assertion than proof. The condition μ0<τ<μ1 in Appendix C is stated to hold 'always' under ML, but τ is never defined, so that is not checkable.\n\nNet: this is a framework paper with a wrong engine. It deserves a serious referee because the problem is real and the structure is a sensible way to think about drug-release scheduling, but in current form the central results are not trustworthy. I would reject, and encourage resubmission with a corrected interference model and a Monte Carlo check.","headline":"Useful TDMA drug-release framework, but the interference statistics have load-bearing errors that invalidate the BER and optimization results.","tokens_in":22126,"tokens_out":6080,"would_cite":false,"duration_ms":51096,"reading_group":"maybe","serious_thinker":"no","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"The paper claims that the bit error rate of each transmitter in a TDMA molecular-communication drug-delivery system can be expressed through Gaussian interference statistics, and that jointly optimizing slot lengths and released molecule…","keywords":["molecular communication","molecular communication via diffusion","TDMA","drug delivery","bit error rate","multi-objective optimization","inter-symbol interference","resource allocation"],"falsifier":"Run a particle-based Brownian simulation of the same three-transmitter geometry with a passive receiver that samples the molecule count inside its volume at the end of each slot, and compare the empirical mean, variance, and BER to Eqs. (13)-(16) and (21). If the simulated interference follows an occupancy-probability model rather than the arrival-difference model, or if the BER ordering of the four strategies changes, the central claim is rejected.","tokens_in":21111,"feed_emoji":"💊","tokens_out":11592,"duration_ms":111007,"temperature":0.7,"pith_summary":"This paper designs a drug-release scheduler for a molecular communication link in which several transmitter nanomachines share one receiver by time division. Its central aim is to show that the bit error rate of each transmitter can be computed from Gaussian statistics of the received molecule count, including interference from earlier symbols and other users, and then minimized by choosing how many molecules each transmitter releases and how long its slot lasts. The paper derives the mean and variance of the received count and an explicit BER for each transmitter, and constructs four management strategies: static or dynamic slot lengths crossed with static or dynamic molecule counts. Simulated results place the fully dynamic strategy's BER near $3.45 \\times 10^{-8}$ in a severe diffusive environment, while the simplest fixed strategy stays near $10^{-2}$. If correct, this gives a quantitative tool for deciding drug dosage and release timing in molecular drug delivery.","feed_headline":"Dynamic drug scheduling cuts molecular bit errors to 3.45e-8","feed_subtitle":"A TDMA drug-delivery plan that tunes both slot length and molecule count beats fixed designs by orders of magnitude.","key_machinery":"The central object is the arrival-probability function $P_h(X_s,t_s)$, the probability that a molecule released by transmitter $s$ is received by time $t_s$, obtained by integrating a 3D Gaussian drift-diffusion density over the receiver volume. The argument is carried by separating the total received count into a wanted Gaussian signal plus an interference Gaussian built from the difference probabilities $Y_{j,s}^u$ and $H_{j,s}$, and then writing the ML BER as an error-function expression in the conditional mean and variance pairs. The same expressions are reused as objectives in four optimization variants, solved by scalarizing weighted sums and alternating between time-slot and molecule-count blocks.","core_discovery":"The load-bearing discovery is the interference-aware statistics of the TDMA slot. Received molecules from the intended transmitter are modeled as a Gaussian count $M_s[n]\\sim\\mathcal{N}(x_s[n]A_s P_h(X_s,t_s), x_s[n]A_sP_h(X_s,t_s)(1-P_h(X_s,t_s)))$, while interference from previous frames and other transmitters is accumulated as sums of Gaussian terms whose means and variances depend on two probability differences: $Y_{j,s}^u=P_h(\\lambda_{j,s}^u)-P_h(\\lambda_{j,s}^u-t_s)$ for past-frame leakage and $H_{j,s}=P_h(\\sum_{q=j}^s t_q)-P_h(\\sum_{q=j}^{s-1} t_q)$ for same-frame leakage. The ML threshold decision then yields a per-transmitter BER $P_s^e[n]$ expressed through error functions of the two conditional means and variances, Eqs. (11)-(21). These expressions are the objective functions of weighted-sum multi-objective problems over symbol durations and molecule budgets. The paper reports that jointly optimizing both quantities (DTDN) reaches BER $3.45\\times10^{-8}$ in a severe diffusive environment, compared with about $10^{-2}$ for uniform allocation.","pith_inferences":["If the receiver is modeled as occupancy-counting rather than arrival-counting, as the paper's passive-receiver description implies, the interference probabilities would change from interval arrival differences to occupancy probabilities; re-running the same optimizations under that model is a direct test of the reported BER gains.","The crossover between dose optimization and slot optimization is tied to the drift and diffusion parameters; under different vessel flow or molecule sizes the crossover frame lengths would shift, so a deployed system should recompute the operating regime rather than reuse the reported thresholds.","The same weighted-sum machinery could be applied to transmitter-specific weights, unequal molecule types, or absorption-based receivers; any of these variations changes the interference structure and likely alters which of the four strategies is preferred.","A particle-based Brownian simulation of the same geometry would settle whether the Gaussian-sum interference approximation holds at the small molecule budgets near the lower bound of 100 molecules, where the normal approximation is least reliable."],"forward_implications":["Jointly optimizing slot duration and molecule count (DTDN) gives BER as low as $3.45\\times10^{-8}$ in severe diffusion, while static-uniform scheduling (STSN) stalls near $10^{-2}$, so the gap is orders of magnitude.","For small frame budgets, allocating molecule counts matters more than shaping slots; for larger frames, shaping slots matters more, with a crossover that depends on diffusion severity (around 3.29 ms in MDE and 2.97 ms in SDE).","Only about three previous frames contribute meaningful interference; beyond that the BER is flat, so truncating the interference sum at $U=3$ is safe.","The stated complexity ranking (STSN lowest, DTDN highest) means that in moderate and severe diffusion with larger frames, DTSN offers most of DTDN's BER at lower complexity.","When the optimized molecule count is rounded from a real to an integer, the BER penalty is small for typical budgets and shrinks as the lower bound grows."],"supporting_citations":[{"why":"supplies the 3D diffusive-drift channel model whose arrival CDF defines the probability $P_h(X_s,t_s)$ used for wanted and interfering molecules.","marker":"[14]"},{"why":"establishes the passive receiver model: a sphere that counts molecules inside its volume and does not absorb them.","marker":"[10]"},{"why":"justifies modeling the per-release received count as binomial, the starting distribution in Eq. (5).","marker":"[20]"},{"why":"provides the condition under which the binomial count is approximated by a Gaussian, used throughout the analysis.","marker":"[23]"},{"why":"supplies the ML threshold detection rule and synchronization framework used to set $\\tau_s$.","marker":"[27]"},{"why":"gives the BER expression for molecular-communication links that Eq. (21) adapts to the TDMA interference setting.","marker":"[18]"},{"why":"supplies the weighted-sum scalarization that turns each multi-objective BER problem into a single-objective problem.","marker":"[35]"},{"why":"provides the alternating search method used to solve the coupled slot-duration and molecule-count problems.","marker":"[40]"}],"fun_headline_variants":["Joint TDMA scheduling and molecule count push BER to 3.45e-8","Dynamic drug release cuts molecular bit errors million-fold","Optimizing slots and molecule numbers achieves BER 3.45e-8","TDMA drug delivery with dynamic slots and amounts hits 3.45e-8 BER","Interference-aware TDMA drug release reaches BER 3.45e-8"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is that interference from an earlier release can be treated as molecules that arrive at the receiver during the current slot with probability equal to a difference of two arrival probabilities; for the passive receiver the paper describes, which counts molecules inside its volume at the sampling instant, the correct weighting would instead be the probability that the molecule is still inside the receiver volume at that instant.","fun_headline_variants_meta":{"raw":{"variants":["Joint TDMA scheduling and molecule count push BER to 3.45e-8","Dynamic drug release cuts molecular bit errors million-fold","Optimizing slots and molecule numbers achieves BER 3.45e-8","TDMA drug delivery with dynamic slots and amounts hits 3.45e-8 BER","Interference-aware TDMA drug release reaches BER 3.45e-8"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000459,"raw_usage":{"total_tokens":2371,"prompt_tokens":1089,"completion_tokens":1282,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":705,"completion_tokens_details":{"reasoning_tokens":1181}},"tokens_in":705,"tokens_out":1282,"duration_ms":10065,"temperature":1.0,"reasoning_tokens":1181,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T12:47:57.712271+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Run a particle-based Brownian simulation of the same three-transmitter geometry with a passive receiver that samples the molecule count inside its volume at the end of each slot, and compare the empirical mean, variance, and BER to Eqs. (13)-(16) and (21). If the simulated interference follows an occupancy-probability model rather than the arrival-difference model, or if the BER ordering of the four strategies changes, the central claim is rejected.","supporting_citations":[{"cited_title":"3-D diffusive-drift molecu lar channel characterization for active and passive receiv ers,","cited_arxiv_id":null,"evidence_quote":"supplies the 3D diffusive-drift channel model whose arrival CDF defines the probability $P_h(X_s,t_s)$ used for wanted and interfering molecules."},{"cited_title":"Active versu s passive: Receiver model transforms for diffusive molecul ar communication,","cited_arxiv_id":null,"evidence_quote":"establishes the passive receiver model: a sphere that counts molecules inside its volume and does not absorb them."},{"cited_title":"Performance anal ysis of amplitude modulation schemes for diffusion-based molecular communication,","cited_arxiv_id":null,"evidence_quote":"justifies modeling the per-release received count as binomial, the starting distribution in Eq. (5)."},{"cited_title":"Papoulis and S","cited_arxiv_id":null,"evidence_quote":"provides the condition under which the binomial count is approximated by a Gaussian, used throughout the analysis."},{"cited_title":"Receiver design for molecular communication,","cited_arxiv_id":null,"evidence_quote":"supplies the ML threshold detection rule and synchronization framework used to set $\\tau_s$."},{"cited_title":"Non-Uniform BCSK Modulation in Nutrient-Limited Relay-Assisted Molecular Communication System: Optimization and Performance Evaluation","cited_arxiv_id":"1903.04749","evidence_quote":"gives the BER expression for molecular-communication links that Eq. (21) adapts to the TDMA interference setting."},{"cited_title":"The weighted sum method for multi-objective optimization: new insights,","cited_arxiv_id":null,"evidence_quote":"supplies the weighted-sum scalarization that turns each multi-objective BER problem into a single-objective problem."},{"cited_title":"Joint access and fronthaul radio resource allocation in pd-noma-based 5g networks enabling dual connectivity and c omp,","cited_arxiv_id":null,"evidence_quote":"provides the alternating search method used to solve the coupled slot-duration and molecule-count problems."}],"review_version":1}