{"id":"a5e79bec-7792-4a71-a445-d7fde673cff1","arxiv_id":"1908.04863","paper_version":4,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"Joint optimization of BS precoding and IRS phase shifts for weighted sum rate maximization in SWIPT MIMO systems is solved by a convergent BCD algorithm, with simulations showing IRS expands the energy-harvesting range.","lead":"This paper designs an algorithm that jointly controls a base station's transmit beams and an intelligent reflecting surface's phase shifts to maximize data rates while meeting energy harvesting targets for remote sensors. It shows by simulation that using the reflecting surface can extend the operating range of energy-harvesting devices.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Theorem 4's proof in Appendix D concatenates block KKT conditions with distinct dual variables for the coupled EH constraint; the original KKT requires one multiplier, and no argument shows they coincide, so BCD may converge to a point that is not jointly stationary.","rationale":"The reader's conditional verdict is based on omitted proofs, the Algorithm 3 typo, and the acknowledged perfect-CSI limitation. The perfect-CSI assumption is indeed a limitation, but the authors explicitly frame the results as an upper bound, so it weakens the practical range claims rather than the internal mathematical claim. The more load-bearing problem is that Theorem 4's proof is internally incomplete: it never shows that the two block subproblems produce the same dual variable for the shared EH constraint. The toy counterexample shows that exact block coordinate maximization with a coupled inequality constraint can converge to a coordinatewise stationary point that is not a KKT point of the joint problem. Since the central advertised contribution is the rigorous KKT guarantee, this gap undermines the main theoretical claim. The paper still contains a plausible algorithm and self-consistent simulations, and the missing argument might be repairable under additional structural assumptions, but as written the central claim is not established. I therefore recommend moving from CONDITIONAL to REJECT unless the authors supply a proof of a common multiplier or explicitly weaken the convergence claim to a coordinatewise stationary point.","tokens_in":23151,"tokens_out":14764,"duration_ms":163716,"concrete_test":"Run Algorithm 5 on one random instance from Section V. At the converged point, solve for the EH-constraint multiplier mu* from the F-block stationarity equation (D.3) and nu* from the phase-block stationarity equation (C.8). Then compute the joint KKT residual r(xi) = || grad_F R + xi grad_F Q ||_F^2 + || grad_Phi R + xi grad_Phi Q ||^2, minimizing over xi >= 0. If min r(xi) is comparable to the numerical solver tolerance (e.g., below 1e-6 relative to the gradient norms), the point is jointly KKT and the concern is refuted. If the minimal residual is nonzero at machine precision, the converged point violates the KKT conditions of Problem (8), contradicting Theorem 4 as stated.","verdict_should_be":"REJECT","load_bearing_attack":"The central claim is Theorem 4: the sequence generated by Algorithm 5 converges and the final solution satisfies the KKT conditions of Problem (8). The proof in Appendix D derives stationarity for the F block with a dual variable mu* for the energy-harvesting constraint (D.3), and stationarity for the phase block with a different dual variable nu* for the same constraint (C.8). The final paragraph then declares these equations to be exactly the KKT conditions of Problem (8). This is not justified: the KKT system of Problem (8) has a single multiplier xi for the EH constraint, and that same xi must appear in both the F-stationarity and the phase-stationarity equations. The proof never establishes mu* = nu*, and without this equality the concatenated equations do not form a valid KKT system. This is not a minor technicality. For block coordinate methods with coupled constraints, coordinatewise stationary points need not be jointly stationary. A minimal example is max x + 2y subject to x^2 + y^2 >= 1, x,y >= 0; exact block maximization converges to (1/sqrt(2), 1/sqrt(2)), which satisfies the two block KKT conditions with different multipliers (1/sqrt(2) and 2/sqrt(2)) but violates the joint KKT condition. Nothing in Section III or Appendix D rules out this failure mode for Problem (8). The inverted termination condition in Algorithm 3 (Step 1, 'If J(0) <= Qhat, terminate') is also a real defect, but even correcting it does not repair the KKT gap.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"This paper considers an intelligent reflecting surface (IRS) aided multiuser MIMO simultaneous wireless information and power transfer (SWIPT) system. A multi-antenna base station serves multiple multi-antenna information receivers (IRs) while guaranteeing a weighted sum harvested-power constraint for energy receivers (ERs), with the IRS deployed in the vicinity of the ERs. The authors formulate the weighted sum rate (WSR) maximization problem over transmit precoding matrices and the IRS phase-shift matrix, subject to the transmit power budget, the energy-harvesting (EH) constraint, and unit-modulus phase constraints. They propose a block coordinate descent (BCD) algorithm based on the WMMSE reformulation: the precoding subproblem is solved by successive convex approximation with a bisection-search dual method, and the phase-shift subproblem by an MM method combined with a price-based bisection method. They prove convergence of the subroutines and claim that the BCD algorithm converges to a KKT point of the original problem (Theorem 4). A feasibility-checking algorithm is also presented. Simulations over a Rician/Rayleigh fading scenario with four ERs and two IRs demonstrate that the IRS significantly improves the harvested power and WSR relative to no-IRS and fixed-phase benchmarks, and that the BCD algorithm converges quickly.","tokens_in":23443,"tokens_out":13724,"duration_ms":127723,"significance":"The paper addresses a timely problem and provides a plausible algorithmic framework that appears to work numerically. The WMMSE transformation, the SCA lower bound for the nonconvex EH constraint, and the MM bounding of the phase objective are sensible, and the simulation results support the qualitative claim that IRS deployment extends the ER operating range and improves WSR. The claimed guarantees, especially Theorem 4's assertion of KKT convergence for the BCD algorithm, are central to the paper's theoretical contribution and are not established by the arguments given. The paper also acknowledges the perfect-CSI assumption as a limitation, which is appropriate. If the convergence claims are corrected or appropriately weakened, the algorithmic and numerical contributions would be of interest to the SWIPT and IRS communities. The provided proofs for the subproblem algorithms are mostly standard when given; the omitted proofs and the flawed concatenation in Appendix D are the main obstacles.","major_comments":[{"comment":"The proof of Theorem 4 concatenates the KKT stationarity and complementary slackness equations of the F-block (D.3)-(D.5), with dual variable μ* for the energy-harvesting constraint, and those of the phase block (C.5), (C.6), (C.8), with dual variable ν* for the same constraint. The KKT conditions of Problem (8) require a single multiplier ξ for constraint (8c), and that same ξ must multiply the gradient of the constraint in both stationarity equations. The authors do not prove μ* = ν*, and the concatenated equations are therefore not a valid KKT system. Block-coordinate methods with coupled constraints can converge to points that satisfy each block's KKT conditions with different multipliers but do not satisfy the joint KKT conditions, so this gap is a load-bearing error in the central convergence claim.","section":"Appendix D, Theorem 4"},{"comment":"Step 1 of Algorithm 3 states: 'Calculate J(0). If J(0) ≤ Qhat, terminate; otherwise go to step 2.' Since the linearized EH constraint (43) requires J(p) ≥ Qhat, the termination condition is inverted: when J(0) < Qhat, the candidate φ(0) is infeasible and the bisection step should be invoked; when J(0) ≥ Qhat, the solution should be returned. As written, the algorithm may terminate with an infeasible phase vector, which propagates to Algorithm 4 and undermines the feasibility guarantee in Theorem 3.","section":"Section III-C, Algorithm 3"},{"comment":"The monotonicity of J(p) is load-bearing because Algorithm 3 relies on it for the bisection search, yet the proof is omitted with the statement 'similar to Lemma 1 and thus omitted.' Lemma 1's proof concerns the value function of a convex optimization problem with a linear perturbation; J(p) is the value of a linear function evaluated at the global maximizer of a nonconvex unit-modulus problem, so the argument does not transfer directly. The claim is not obviously true for this nonconvex problem, and a complete proof is required.","section":"Section III-C, Lemma 3"},{"comment":"The proof of Theorem 1 is omitted ('similar to [44]'). Because Algorithm 2's SCA treatment of the nonconvex EH constraint differs from the setting of [44], and because Theorem 1 is used as a premise in the proof of Theorem 4, the convergence of the F-block to a KKT point needs a self-contained proof or a precise argument explaining why the proof of [44] covers the present problem.","section":"Section III-B, Theorem 1"}],"minor_comments":[{"comment":"The left-hand side uses the subscript i in Q_i, although the quantity is defined for the l-th ER (Q_l); please unify the notation.","section":"Equation (7)"},{"comment":"The initialization statement 'initialize the precoding matrices F(0) from Section 2' should read 'Section IV' (the feasibility-check section).","section":"Algorithm 2, Step 1"},{"comment":"The text describing Fig. 3 uses 'xEH' whereas the figure and the rest of the text use 'xER'; please use one notation consistently.","section":"Section V, Fig. 3 description"},{"comment":"The caption contains a typo ('Havested Power'); correct it to 'Harvested Power'.","section":"Caption of Fig. 3"},{"comment":"The notation 'A*' and 'A⋆' for the conjugate operator and converged solution, respectively, may be confusing; consider using more standard notation such as A^H for conjugate transpose and A^* for converged value, or a clearer distinction.","section":"Notation, Section I"}],"recommendation":"major_revision","confidential_remarks":"The manuscript's heavy self-citation pattern is not in itself disqualifying, but the omitted proofs are frequently deferred to these self-citations, which compounds the reviewer's burden. The central KKT-proof gap in Appendix D is the main substantive concern; if the authors can close it or appropriately weaken the claim to block-wise stationarity, the paper would be publishable. The simulation study is extensive and the problem is timely for the SWIPT and IRS communities. I recommend major revision rather than rejection, as the algorithmic framework and numerical findings are valuable and the theoretical gaps appear addressable in a revision."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Short version: the paper has a real problem instance — WSR maximization for IRS-assisted SWIPT MIMO with EH constraints — and the algorithm is a sensible assembly of WMMSE, SCA, MM, and pricing. The subproblem derivations are mostly careful, and the simulations are internally consistent. But the sharpest claim, Theorem 4, is not proven as written.\n\nThe stress-test note is right. Appendix D takes a KKT point for the F block with multiplier mu* on the EH constraint and a KKT point for the phase block with multiplier nu* on the same constraint, then declares the concatenation to be the KKT conditions of the original problem. In the original KKT system, a single multiplier must appear in both stationarity equations. Nothing shows mu* = nu*. For block coordinate methods with coupled constraints, coordinatewise stationarity does not imply joint stationarity; the example in the note (max x + 2y s.t. x^2 + y^2 >= 1, x,y >= 0) demonstrates the failure mode cleanly. So the convergence-to-KKT claim is unsupported.\n\nOther soft spots are smaller. Algorithm 3's step 1 terminates when J(0) <= Qhat, which is the wrong side of the inequality; that looks like a straightforward typo. Theorem 1 and Lemma 3 are deferred to 'similar to' references; acceptable if the referenced arguments truly carry over, but Lemma 3's monotonicity underlies the bisection and should be stated. The paper also assumes perfect CSI and a linear EH model, which the authors acknowledge gives an upper bound; that limits the practical reach of the range-extension numbers but does not undercut the algorithmic framework.\n\nCredit where due: the WMMSE reformulation, the SCA lower bound on the EH constraint, the MM bounding for the phase objective, and the price-based global optimality argument for the subproblem are all standard tools applied carefully. The simulations use sensible benchmarks and the complexity accounting is explicit. The self-citations are heavy but the cited prior work is genuinely relevant.\n\nThe audience is researchers in IRS/RIS and SWIPT resource allocation. They get a useful algorithm and a clean problem formulation. The paper deserves serious refereeing, but a referee should push for a repaired proof of Theorem 4 — either a true joint-KKT argument or a weaker but honest convergence statement.","headline":"The paper's real contribution is a new problem instance and a plausible BCD-style algorithm, but Theorem 4's proof is not valid as written because it concatenates block KKT systems with different dual multipliers for the same constraint.","tokens_in":24029,"tokens_out":2155,"would_cite":true,"duration_ms":24074,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"A passive reflecting surface can nearly double the operating range of wireless energy harvesting.","keywords":["intelligent reflecting surface","simultaneous wireless information and power transfer","MIMO broadcasting","weighted sum rate maximization","block coordinate descent","phase shift optimization","energy harvesting","passive beamforming"],"falsifier":"Run the same scenario with channel estimation errors of the magnitude produced by current two-stage IRS channel estimators; if the weighted-sum-rate or harvested-power advantage over no-IRS shrinks toward zero, or the 5.5-to-9 m range extension disappears, the central claim is falsified.","tokens_in":22943,"feed_emoji":"📡","tokens_out":5424,"duration_ms":50878,"temperature":0.7,"pith_summary":"This paper argues that placing an intelligent reflecting surface (a reconfigurable passive mirror) near energy-harvesting receivers lets a base station deliver both data and power more effectively than broadcasting without it. It formulates the design as maximizing the weighted sum rate of information receivers subject to a minimum harvested-power requirement, and solves it by alternating between optimizing base-station precoding and IRS phase shifts. The paper proves each subproblem converges to a Karush-Kuhn-Tucker point and that the overall alternating algorithm converges to a KKT point of the original non-convex problem. Simulations show the harvested-power target can be met at much larger distances with an IRS, with the weighted sum rate also improving by up to about 10 bit/s/Hz.","feed_headline":"A passive reflector nearly doubles wireless power range","feed_subtitle":"With 40 reflecting elements, energy harvesters work at 9 m versus 5.5 m without the surface.","key_machinery":"The engine is the WMMSE reformulation of weighted sum rate, which turns the rate objective into a block-friendly form, followed by block coordinate descent. With phase shifts fixed, the precoding subproblem is handled by successive convex approximation of the energy-harvesting constraint plus a Lagrangian dual/bisection search; with precoders fixed, the phase-shift subproblem is recast as a quadratic program under a unit-modulus constraint and solved by majorization-minimization with a price-based bisection for the energy constraint. The unit-modulus phase vector $\\varphi$ and the equivalent channels $H_{b,k}+H_{r,k}\\Phi Z$ carry the passive beamforming effect.","core_discovery":"The paper's central claim is that jointly optimizing the base station's transmit precoding matrices and the IRS's passive phase shift matrix can satisfy a fixed harvested-power requirement at energy receivers while maximizing the weighted sum rate of information receivers, and that doing so expands the energy receivers' operating range. The design problem is solved by a block coordinate descent algorithm that alternates between a convexified precoding update and a phase-shift update; each subproblem provably reaches a KKT point, and the overall algorithm provably converges to a KKT point of the original problem. In simulation, the IRS lifts the maximum energy-receiver distance from about 5.5 m (no IRS) to about 9 m with 40 reflecting elements, and yields up to about 10 bit/s/Hz higher weighted sum rate than a no-IRS baseline.","pith_inferences":["If the perfect-CSI assumption is relaxed with a realistic channel-estimation error model, the range-extension numbers should be treated as upper bounds; the authors themselves flag robust design as future work.","The same alternating pattern—convexify the energy-harvesting constraint, optimize phase shifts by majorization-minimization, and price the constraint—applies to other IRS design problems with non-convex unit-modulus and coupling constraints, such as secrecy-rate or latency-constrained designs.","A testable next step is to replace the linear energy-harvesting model with a nonlinear rectifier model; the optimal precoding and phase-shift split may shift toward concentrating power in the rectifier's sensitive region.","Because the block coordinate descent loop converges to a KKT point rather than a global optimum, comparing it to a global solver on small instances would show how much performance is left on the table."],"forward_implications":["Placing the IRS near the energy receivers, rather than midway between base station and receivers, is the configuration that yields the large range extension reported in the simulations.","The number of reflecting elements $M$ is a direct performance knob: with 40 elements the energy-receiver working distance grows from 5.5 m to 9 m, and the weighted sum rate grows with $M$.","IRS location matters through the path-loss exponent of the IRS links; when that exponent rises to 3, most of the weighted-sum-rate gain over no-IRS disappears, so obstacle-free siting is part of the design.","A quick feasibility check can tell an operator whether an energy-harvesting target can be met at all under the transmit power limit before running the full optimization."],"supporting_citations":[{"why":"Gives the WMMSE equivalence that converts the weighted sum rate objective into a tractable block-wise form.","marker":"[39]"},{"why":"Supplies the BCD convergence argument pattern and the alternating update structure.","marker":"[33]"},{"why":"Establishes the MIMO broadcasting SWIPT model and the linear harvested-power constraint used here.","marker":"[29]"},{"why":"The prior IRS-SWIPT weighted sum power maximization that this paper extends to WSR and multi-antenna users.","marker":"[30]"},{"why":"The authors' multicell IRS-MIMO framework without the energy-harvesting constraint, whose WSR problem is the starting point here.","marker":"[24]"},{"why":"Provides the successive convex approximation technique used to linearize the non-convex constraints.","marker":"[40]"},{"why":"Provides the majorization-minimization machinery for the phase-shift subproblem.","marker":"[31]"},{"why":"Supplies the matrix identity turning the phase-shift objective into a quadratic form in the phase vector.","marker":"[45]"}],"fun_headline_variants":["Passive reflector doubles wireless power reach","IRS boosts SWIPT range and rate in one design","Joint precoding and phase shifts extend power distance","Algorithm proves optimal for IRS-aided SWIPT","Reflecting surface widens energy harvesting zone"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The base station knows all channels perfectly, and the authors state that this makes the results an upper bound.","fun_headline_variants_meta":{"raw":{"variants":["Passive reflector doubles wireless power reach","IRS boosts SWIPT range and rate in one design","Joint precoding and phase shifts extend power distance","Algorithm proves optimal for IRS-aided SWIPT","Reflecting surface widens energy harvesting zone"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000164,"raw_usage":{"total_tokens":1250,"prompt_tokens":951,"completion_tokens":299,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":567,"completion_tokens_details":{"reasoning_tokens":230}},"tokens_in":567,"tokens_out":299,"duration_ms":3691,"temperature":1.0,"reasoning_tokens":230,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T13:31:02.710371+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Run the same scenario with channel estimation errors of the magnitude produced by current two-stage IRS channel estimators; if the weighted-sum-rate or harvested-power advantage over no-IRS shrinks toward zero, or the 5.5-to-9 m range extension disappears, the central claim is falsified.","supporting_citations":[{"cited_title":"Joint precoding and RRH selection for user-centric green MIMO C-RAN,","cited_arxiv_id":null,"evidence_quote":"Supplies the BCD convergence argument pattern and the alternating update structure."},{"cited_title":"MIMO broadcasting for simultaneous wireless information and power transfer,","cited_arxiv_id":null,"evidence_quote":"Establishes the MIMO broadcasting SWIPT model and the linear harvested-power constraint used here."},{"cited_title":"Weighted Sum Power Maximization for Intelligent Reflecting Surface Aided SWIPT","cited_arxiv_id":"1907.05558","evidence_quote":"The prior IRS-SWIPT weighted sum power maximization that this paper extends to WSR and multi-antenna users."},{"cited_title":"Multicell MIMO Communications Relying on Intelligent Reflecting Surface","cited_arxiv_id":"1907.10864","evidence_quote":"The authors' multicell IRS-MIMO framework without the energy-harvesting constraint, whose WSR problem is the starting point here."},{"cited_title":"The non-coherent ultra-dense C-RAN is capable of outperforming its coherent counterpart at a limited fronthaul capacity,","cited_arxiv_id":null,"evidence_quote":"Provides the successive convex approximation technique used to linearize the non-convex constraints."}],"review_version":1}