{"id":"2f2a47d0-85e2-4c83-a237-d64ca9625d89","arxiv_id":"2505.02254","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":4.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":2,"one_line_summary":"Tri-hybrid precoding with spherical-harmonics-optimized antenna patterns shows large simulated sum-rate gains, but the gains largely vanish after projection onto 64 experimentally validated patterns from a real ERA array.","lead":"This paper proposes a way to jointly tune digital, analog, and antenna-pattern precoding for multi-user MIMO using spherical-harmonic descriptions of reconfigurable antenna patterns. The open question is whether the theoretical pattern optimization survives contact with real hardware, and the simulations say the gain mostly does not.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Projected-pattern evaluation likely uses precoders optimized for unconstrained patterns, so the reported realizability collapse may be an artifact.","rationale":"The reader's formal weakest_assumption focuses on the channel model (3) and the SH representation (12), but the reader's rationale also flags an 'ambiguous post-projection evaluation protocol.' My stress-test compresses that ambiguity into a specific, testable flaw: the projected-pattern comparison likely keeps F_RF and F_BB at their unconstrained-optimum values, which biases the projected curves downward. This is the single most load-bearing issue because the paper's headline message is the post-projection collapse, not the SH optimization itself. The fix is straightforward and does not require new hardware or new theory: re-optimize the digital and analog precoders after fixing the projected patterns. If the collapse persists after re-optimization, the central negative claim is supported; if it disappears, the paper's main practical message is an artifact. The other concerns noted by the reader -- no error bars, no Monte Carlo count, no code/data, and an unquantified positivity assumption -- are real but secondary: they affect confidence in the quantitative curves, whereas the missing re-optimization affects the validity of the comparison itself. Given that the paper is honestly written and the mathematical derivations appear internally consistent, the appropriate verdict remains CONDITIONAL pending this re-evaluation, so I do not move the reader's verdict.","tokens_in":15273,"tokens_out":5187,"duration_ms":68792,"concrete_test":"Re-run the Fig. 2(b) experiment with the projected F_EM fixed but recompute the digital and analog precoders: for each channel realization and each P_max, re-solve the tri-hybrid problem over F_D (and then decompose to F_RF, F_BB) using the WMMSE updates (15)-(17) with the projected patterns, before evaluating sum rate. Compare this re-optimized projected curve against (i) the unconstrained SH curve and (ii) conventional hybrid precoding, using identical re-optimization for all methods. If the re-optimized projected curve climbs close to the unconstrained SH curve or above conventional hybrid, the reported realizability collapse is an artifact of post-hoc projection; if it remains below conventional hybrid at high P_max, the paper's negative conclusion survives.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The paper's central negative result is the post-projection collapse in Fig. 2(b): after replacing the optimized spherical-harmonic patterns with realizable ERA patterns, the sum rate drops and can even fall below conventional hybrid precoding at P_max = 20 dBm. The load-bearing question is whether this collapse reflects the cost of realizability or the cost of a mismatched evaluation protocol. In Section IV-C, the projection step (28)-(29) replaces each optimized coefficient vector c_opt^(n) with a realizable pattern selected by Euclidean distance at the sampled path angles, and nothing in Section IV-C or V states that F_RF and F_BB are re-optimized after this replacement. Since F_RF and F_BB were solved jointly with the unconstrained F_EM in Algorithm 1, reusing them with the projected F_EM can only lower the achievable sum rate, even when the realizable patterns are equally good for the problem. The projected curves therefore measure the cost of not re-solving the tri-hybrid optimization for the chosen realizable patterns, not the intrinsic cost of physical realizability. This is directly load-bearing because the abstract and conclusion use exactly this comparison to claim that 'the performance gain is notably reduced or even negligible' for real ERA hardware. If the evaluation is unfair, that conclusion is not established.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper considers a tri-hybrid multi-user precoding architecture in which a base station with electromagnetically reconfigurable antennas (ERAs) jointly optimizes baseband, RF, and EM-domain precoders. Each antenna's radiation pattern is represented by a truncated spherical-harmonic expansion, and the weighted sum-rate problem is solved via an alternating WMMSE-type algorithm with closed-form updates. After optimization, the coefficient vectors are projected onto a set of 64 experimentally validated radiation patterns taken from the authors' earlier hardware work. Simulations show large sum-rate gains for the unconstrained spherical-harmonic optimization, but after projection the gains shrink and can become negative at high transmit power, leading the authors to conclude that relaxed pattern optimization overstates the benefit of ERAs.","tokens_in":15529,"tokens_out":5892,"duration_ms":69920,"significance":"If the conclusion were established, the paper would be a useful cautionary result for the tri-hybrid MIMO literature, which often relies on relaxed or discrete radiation-pattern models. The algorithmic contribution has merit: the WMMSE reformulation is standard, the per-antenna EM update in Theorem 1 is a plausible closed-form solution with a practical bisection procedure, and the use of a real hardware pattern set is a strength. However, the load-bearing comparison in Fig. 2(b) is compromised by the evaluation protocol, so the main message is not yet convincing. With a fair comparison and full simulation details, the paper could become a solid contribution.","major_comments":[{"comment":"After replacing each optimized coefficient vector c_opt^(n) with a realizable pattern via (28)-(29), the manuscript does not state that F_RF and F_BB are re-optimized. Since Algorithm 1 solves F_RF, F_BB, and F_EM jointly, reusing the F_RF and F_BB obtained for the unconstrained F_EM with a different, projected F_EM yields a suboptimal combination; the resulting sum-rate drop measures the cost of not re-solving the tri-hybrid problem for the realizable patterns, not the intrinsic cost of physical realizability. The abstract and conclusion rely on exactly this comparison to claim the gain is \"notably reduced or even negligible.\" Please either re-optimize F_RF and F_BB (for example, by running Algorithm 1 with F_EM fixed to the selected realizable patterns) or explicitly present the curves as the performance of the relaxed solution after a one-shot pattern replacement.","section":"Section IV-C, Eqs. (28)-(29); Section V, Fig. 2(b)"},{"comment":"The claim that fixing c_DC^(n) = eta in (0, sqrt(4*pi)) with a \"sufficiently large\" eta ensures strictly positive radiation patterns is not quantified. For the truncated expansion G^(n)(theta,phi) = Y_0^0(theta,phi)*eta + b_AC(theta,phi)^T c_AC^(n) with ||c_AC^(n)|| = sqrt(4*pi - eta^2), the minimal value over the sphere depends on the sup-norm of the truncated harmonics. The chosen eta = sqrt(2*pi) in Section V is not justified by any analytic bound or by a numerical positivity check. Without this, the unconstrained upper bound may be physically meaningless even before projection. Please provide a lower bound on eta for the given U or verify positivity of the optimized patterns on a dense angular grid.","section":"Section IV-B, Eqs. (18)-(19); Section V"},{"comment":"The simulation setup is insufficiently specified for the quantitative claims. The paper does not report the number of Monte Carlo realizations, the distribution of the complex gains alpha_{k,ell}, the path angles and their distributions, the array geometry (element spacing, near-field versus far-field assumption), or the radiation pattern used for the conventional hybrid baseline. Consequently, the magnitudes in Fig. 2(b), including the reported crossover at P_max = 20 dBm, cannot be reproduced or independently assessed. Please provide complete parameters, run counts, and error bars.","section":"Section V"}],"minor_comments":[{"comment":"There are several typographical errors, including \"preconding\" in Section II, \"recorders\" instead of \"precoders\" in Section I, and the heading \"CONSLUSION\" in Section VI.","section":"Throughout"},{"comment":"The axis label reads \"Decomposition (Step 5 in Algo. 1)\", but the text describes the plot as sum rate versus iteration number of Algorithm 1; the label should be corrected.","section":"Figure 2(a)"},{"comment":"The truncation order U is not stated explicitly; please state that T = 25 corresponds to U = 4, and clarify the choice of truncation length.","section":"Section III-B.1"},{"comment":"The projection criterion in (29) is based on Euclidean distance at the sampled path angles only; a pattern close at those angles may differ substantially elsewhere. A brief justification of this metric, or a discussion of its limitations, would strengthen the paper.","section":"Section IV-C"}],"recommendation":"major_revision","confidential_remarks":"The central negative conclusion of the paper appears to be driven by the evaluation protocol: if F_RF and F_BB are not re-optimized after projecting F_EM, the reported collapse in Fig. 2(b) may largely disappear under a fair comparison. The authors should be asked to provide re-optimized results and full simulation details before the paper can be considered further. The algorithmic core is plausible, so I do not recommend rejection, but the main claim needs to be re-established."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Quick take: the genuinely new thing is taking the spherical-harmonics EM-domain channel model, solving the tri-hybrid WMMSE in closed form, then projecting onto the 64 measured patterns from the authors' own ERA hardware and showing the gain shrinks. That last comparison is the paper's punchline, and it is where the trouble is.\n\nWhat is good: the spherical-harmonics truncation is sensible; the alternating WMMSE derivation is standard but clean, and Theorem 1 gives a practical bisection for the two local optima of each EM-precoder update. Using a real measured pattern set as the projection target is a real step beyond the discretized angular-domain models in [8]. The paper says plainly that relaxed pattern optimization overstates gains, and that honest statement deserves credit.\n\nThe load-bearing flaw: after projecting the optimized patterns onto the realizable set, the paper does not re-solve for F_RF and F_BB. Algorithm 1 optimizes all three jointly. If you swap in the nearest realizable patterns and keep the old analog/digital precoders, sum rate can only drop, even when the realizable patterns are perfectly good for the problem. So Fig. 2(b)'s collapse at high power, which the abstract and conclusion lean on, measures the cost of not re-optimizing rather than the intrinsic cost of realizability. This is not a minor detail; it is the central empirical claim. The authors call the projection \"suboptimal\" in the conclusion, but they present the result as evidence that real ERA hardware erases the gain. That inference is not established by this comparison.\n\nOther soft spots are secondary: no Monte Carlo runs or error bars; channel generation details are thin; the positivity guarantee via a fixed eta is asserted without quantification. Both local optima are evaluated, which is fine.\n\nBottom line: the optimization machinery is competent and the projection idea is worth publishing, but the negative result needs to be re-tested by re-running Algorithm 1 with projected patterns, or at least by matching benchmarks to the same realizable set. A serious referee should ask for that. I'd send it to review, conditional on fixing the evaluation protocol.","headline":"Useful cautionary study of tri-hybrid ERA precoding, but the central negative result likely conflates physical realizability with a mismatched evaluation protocol.","tokens_in":16055,"tokens_out":1671,"would_cite":false,"duration_ms":20047,"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":"Optimizing radiation patterns in a spherical-harmonics basis boosts tri-hybrid precoding sum rate, but projecting onto realizable ERA patterns erases most of the gain.","keywords":["tri-hybrid precoding","electromagnetically reconfigurable antennas","radiation pattern optimization","spherical harmonics","weighted MMSE","pattern reconfigurable antennas","MISO downlink","sum rate"],"falsifier":"Directly comparing measured or full-wave-simulated sum rates for the 3×3 ERA array under the unconstrained spherical-harmonic patterns and under the projected discrete patterns would settle the claim. If the unconstrained patterns also fail to deliver the modeled gain, the spherical-harmonic channel model is the source of the overestimate; if they match the model but the projected patterns lose the gain, the projection library is the bottleneck.","tokens_in":15089,"feed_emoji":"📡","tokens_out":11793,"duration_ms":129280,"temperature":0.7,"pith_summary":"The paper tries to establish how much of the tri-hybrid precoding gain promised by electromagnetically reconfigurable antennas survives contact with real hardware. It models each antenna's radiation pattern with a truncated real spherical-harmonic expansion, jointly optimizes the EM-domain, analog, and digital precoders through an alternating WMMSE solver with closed-form updates, and then projects the optimized patterns onto a set of 64 realizable patterns from an ERA prototype. Unconstrained spherical-harmonic optimization produces large sum-rate gains over conventional hybrid precoding, but after projection the gain shrinks markedly and at the highest transmit power tested the projected scheme falls below the conventional baseline. This matters because ERA-based tri-hybrid MIMO is being proposed as a way to increase data rates without extra spectrum, and the result suggests that realized gains depend on hardware expressiveness and projection quality, not just on the optimization algorithm.","feed_headline":"Tri-hybrid precoding gains shrink once real antennas are used","feed_subtitle":"Once the optimized patterns are projected onto real antenna hardware patterns, most of the gain disappears","key_machinery":"The load-bearing object is the truncated real spherical-harmonic representation of each antenna's radiation pattern, $G^{(n)}(\\theta,\\phi)\\approx b(\\theta,\\phi)^{\\mathsf T}c^{(n)}$, collected into the block-diagonal EM precoder $F_{\\mathrm{EM}}$. That representation turns each antenna's pattern into a vector of coefficients, so the channel factorizes as $h_k=F_{\\mathrm{EM}}^{\\mathsf T}h_k^{\\mathrm{EM}}$ and the radiation pattern can be optimized jointly with the analog and digital precoders. The solution alternates WMMSE updates for auxiliary variables, a fictitious fully digital precoder $F_D$, and each coefficient vector $c^{(n)}$, with Theorem 1 guaranteeing two locally optimal Lagrange multipliers that can be found by bisection. A final projection step replaces each optimized pattern by the nearest physically realizable pattern from a 64-pattern hardware library, which is what exposes the gap between relaxed and realizable performance.","core_discovery":"The paper's central discovery is that the apparent benefit of reconfiguring each antenna's radiation pattern in tri-hybrid precoding depends almost entirely on how much freedom the pattern model has. When the radiation pattern is optimized over a truncated spherical-harmonic basis, the sum-rate gain over conventional hybrid precoding is large. But the paper then projects each optimized pattern onto the closest of 64 radiation patterns measured from a real ERA prototype, and most of that gain disappears; at the highest power simulated ($P_{\\max}=20$ dBm), the projected tri-hybrid scheme falls below the conventional hybrid baseline. The authors read this as evidence that relaxed pattern optimization overstates ERA benefits, and that both better projection methods and richer reconfigurable-antenna hardware are needed before the tri-hybrid architecture delivers its theoretical promise.","pith_inferences":["A testable extension would be to optimize directly over the convex hull of the 64 realizable patterns instead of projecting after the fact; the negative result suggests this would recover much of the lost gain if the bottleneck is the projection, and would not if the bottleneck is the limited pattern library.","If the same relaxed-then-projected pattern holds for other reconfigurable-antenna architectures, the lesson generalizes: theoretical pattern degrees of freedom are not a reliable proxy for achievable rate until hardware expressiveness is priced in.","The 64-pattern library comes from one antenna prototype; interpolating or learning a continuous manifold between those patterns would separate hardware limits from projection losses and could guide hardware designers on which patterns to add."],"forward_implications":["With the 64-pattern realizable library used in the paper, the unconstrained spherical-harmonic gains are mostly out of reach, so ERA hardware that offers richer or more application-tuned pattern sets would be needed to realize them.","At high transmit power (such as $P_{\\max}=20$ dBm), the projected tri-hybrid scheme can fall below the conventional hybrid baseline, so realizability should enter the optimization as a constraint rather than as a post-hoc repair.","Because the channel model covers both far-field and near-field paths, the same alternating solver applies to near-field XL-MIMO scenarios without needing a separate formulation.","The monotonicity result for the two Lagrange-multiplier intervals makes each coefficient-vector update computationally cheap via bisection, so the method scales to arrays larger than the 3×3 test case."],"supporting_citations":[{"why":"Supplies the 64 experimentally validated radiation patterns used in the projection step; the realizable set that erases most of the gain.","marker":"[3]"},{"why":"Introduces the spherical-harmonics representation and the 4π gain normalization behind the EM precoder constraints.","marker":"[7]"},{"why":"Provides the prior use of spherical-harmonics decomposition for reconfigurable-antenna radiation patterns in the electromagnetic domain.","marker":"[9]"},{"why":"Supplies the WMMSE equivalence and alternating convexity that the paper's closed-form updates inherit.","marker":"[17]"},{"why":"Gives the iterative decomposition algorithm used in Step 5 to factor the optimized fully digital precoder into analog and baseband precoders.","marker":"[18]"},{"why":"Serves as the hybrid digital/analog beamforming baseline whose sum-rate performance is compared in the simulations.","marker":"[10]"}],"fun_headline_variants":["Ideal antenna patterns overstate tri-hybrid gain","Real antennas erase tri-hybrid precoding boost","Simulated ERA gain vanishes on real hardware","Tri-hybrid gain depends on pattern fantasy"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The whole comparison rests on the channel model in (3) and the truncated spherical-harmonic representation (12), which assume each antenna's far-field gain pattern stays valid when mounted in the 3×3 array and that mutual coupling and polarization effects can be ignored.","fun_headline_variants_meta":{"raw":{"variants":["Ideal antenna patterns overstate tri-hybrid gain","Real antennas erase tri-hybrid precoding boost","Simulated ERA gain vanishes on real hardware","Tri-hybrid gain depends on pattern fantasy"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000271,"raw_usage":{"total_tokens":1591,"prompt_tokens":869,"completion_tokens":722,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":485,"completion_tokens_details":{"reasoning_tokens":661}},"tokens_in":485,"tokens_out":722,"duration_ms":8694,"temperature":1.0,"reasoning_tokens":661,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-16T00:57:29.699773+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Directly comparing measured or full-wave-simulated sum rates for the 3×3 ERA array under the unconstrained spherical-harmonic patterns and under the projected discrete patterns would settle the claim. If the unconstrained patterns also fail to deliver the modeled gain, the spherical-harmonic channel model is the source of the overestimate; if they match the model but the projected patterns lose the gain, the projection library is the bottleneck.","supporting_citations":[{"cited_title":"Reconfigurable massive MIMO: Precoding design and channel estimation in the electromagnetic domain,","cited_arxiv_id":null,"evidence_quote":"Provides the prior use of spherical-harmonics decomposition for reconfigurable-antenna radiation patterns in the electromagnetic domain."},{"cited_title":"An iteratively weighted MMSE approach to distributed sum-utility maximization for a MIMO interfering broadcast channel,","cited_arxiv_id":null,"evidence_quote":"Supplies the WMMSE equivalence and alternating convexity that the paper's closed-form updates inherit."},{"cited_title":"Hybrid precoder design for angle- of-departure estimation with limited-resolution phase shifters,","cited_arxiv_id":null,"evidence_quote":"Gives the iterative decomposition algorithm used in Step 5 to factor the optimized fully digital precoder into analog and baseband precoders."}],"review_version":1}