{"id":"a86b5b92-6e29-495b-b8e8-038907f11634","arxiv_id":"2501.00909","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":3,"one_line_summary":"A dual-polarized RIS-aided ISAC system is optimized with WMMSE, alternating optimization, majorization-minimization, and penalty methods, and is shown by simulation to outperform single-polarized baselines.","lead":"This paper designs transmit and reflection algorithms for a communication-and-radar system in which the base station, the reflecting surface, and the users all use dual-polarized antennas. Simulations show the dual-polarized setup roughly doubles the data rate of a same-size single-polarized system while still detecting two targets.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The DP RIS constraint (24e) lets all four per-element polarization-path coefficients be unit-modulus independently, which no passive element can satisfy and which allows up to 4x (6 dB) amplification; since the RIS link dominates, part of the DP-vs-SP sum-rate gap in Fig.","rationale":"The paper's central claim is quantitative: at equal physical aperture, a DP BS plus DP RIS roughly doubles the ISAC sum rate (Fig. 5: 14->26 vs 11->15 nat/s/Hz) while preserving dual-target sensing. For this to hold, the DP and SP systems must be compared fairly, with physically realizable hardware and a common power budget. The weakest link I find is the DP RIS constraint (24e): the four polarization-path reflection coefficients per element are each constrained to unit modulus independently. No passive, lossless dual-polarized element can satisfy that jointly; the per-element scattering matrix is bounded by passivity (sigma_max <= 1) or unitarity, and the paper's relaxation effectively grants the RIS up to 6 dB of active gain per element. Because the RIS-assisted link has much lower path loss (2.25+2.25 vs 4.75) and therefore dominates received power, this relaxation can materially inflate the DP sum rate, and the SP baselines (standard passive |phi| = 1) receive no such benefit. This is an internal inconsistency (power is conserved everywhere else in the model, but not in the RIS), not a dispute with the literature; it also has a clear direction of bias, favoring the DP system. The reader's flagged assumption, perfect polarization isolation in the sensing model (eq. (16)), is real but secondary: it affects the dual-target sensing feasibility part of the claim, while the central sum-rate advantage would plausibly survive depolarization. Both concerns are checkable by re-simulation. Since the DP aperture advantage is independently supported (e.g., [17]) and the fix is a constraint correction plus re-run, the appropriate outcome remains the reader's CONDITIONAL verdict rather than rejection; I therefore keep the verdict unchanged.","tokens_in":18987,"tokens_out":18746,"duration_ms":165702,"concrete_test":"Re-run the Fig. 5 experiment (Nt = 4 to 10, L = 10, K = 3, P0 = 1, same channels and parameters) with (24e) replaced by a per-element passivity constraint: S_l = [[phi^vv_l, phi^vh_l],[phi^hv_l, phi^hh_l]] with sigma_max(S_l) <= 1, or S_l^H S_l = I_2 for the lossless case, and compare DP against SP 1x and SP 2x. If the DP advantage over SP 1x retains at least about 70% of the original gap, the headline claim stands; if the DP curve drops substantially (e.g., from about 26 to about 20 nat/s/Hz or below at Nt = 10), the gain is partly an artifact of the non-passive model.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The headline claim requires the DP and SP systems in Fig. 5 to be compared under a common, physically realizable power budget. That condition is violated by the DP RIS model. Problem (24) imposes (24e), |phi_{n,pq}| = 1 for each of the four polarization paths per element, with Phi_pq = diag(phi^pq_1,...,phi^pq_L) in (12). A physical DP RIS element is a 2x2 scattering matrix S_l = [[phi^vv_l, phi^vh_l],[phi^hv_l, phi^hh_l]] that must satisfy sigma_max(S_l) <= 1 for passivity (S_l^H S_l = I_2 for lossless reciprocal), which couples the four coefficients and bounds the reflected power by the incident power per element. The paper's constraint set instead allows, e.g., S_l = [[1,1],[1,1]]; for an incident wave with equal-power, in-phase v and h components, this yields 4x (6 dB) reflected-power amplification per element. Because the MM update (45)-(46) coherently aligns phases to maximize the reflected-channel gain, the optimized Phi exploits this non-physical gain. The bias is first-order: with path-loss exponents 2.25+2.25 (BS-RIS-UE) versus 4.75 (BS-UE), the RIS-reflected link dominates received power, and the SP 1x/SP 2x baselines use the standard passive |phi| = 1 model, so the comparison is asymmetric. Internally, power is conserved in the DP channel model (eq. (6)) and at the BS through (24d), but not through the DP RIS, so the model is inconsistent within its own power budget. A material part of the reported gap (DP 14->26 vs SP 1x 11->15 nat/s/Hz) could therefore arise from the infeasible relaxation rather than from the DP aperture itself.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"This paper studies a dual-polarized RIS-aided ISAC downlink in which a DP BS illuminates two radar targets with orthogonal polarizations while serving K DP users through a direct and a RIS-reflected path. The authors maximize the achievable sum rate subject to two sensing SNR constraints and a unit-modulus constraint on the DP RIS phase matrices. The solution uses a WMMSE reformulation, alternating optimization, an MM algorithm for the RIS phases, and a penalty-based closed-form beamforming update; simulations compare DP with SP baselines, quantized phase shifts, XPD, and sensing-SNR trade-offs. The algorithmic machinery is standard and appears internally consistent, but the physical model of the DP RIS and the target scattering model contain idealizations that directly affect the headline DP-vs-SP conclusions.","tokens_in":19410,"tokens_out":12226,"duration_ms":120010,"significance":"If the idealizations were removed or justified, the paper would provide a useful complete framework for a DP ISAC design. It correctly identifies the polarization dimension as a way to double aperture capacity without increasing footprint, and it gives a transparent optimization pipeline with convergence checks and quantitative comparisons, including quantization and XPD sensitivity. The penalty/MM derivations are standard and the empirical convergence plots support the internal algorithmic claims. However, contribution 3 ('substantially outperforms traditional SP systems') is not established under a physically consistent model: the DP RIS constraint permits 6 dB per-element amplification, and the sensing model assumes perfect polarization selectivity at targets. These are not cosmetic: both are embedded in the problem formulation and the simulations. The paper is therefore promising but requires a substantive revision to the model before the central claims can be accepted.","major_comments":[{"comment":"The DP RIS model violates passivity. Because the four block-diagonal phase matrices in (12) are optimized independently under |phi^pq_n|=1 for every {p,q}, the per-element 2x2 reflection matrix S_l is not constrained; S_l = [[1,1],[1,1]] is feasible and has sigma_max=2, so an incident wave with equal-power, in-phase v and h components is reflected with 4x (6 dB) the incident power. The MM update in Eq. (46) coherently aligns the four polarization paths, so the optimized Phi exploits this non-physical gain. With alpha_BS-RIS=alpha_RIS-UE=2.25 dominating the alpha_BS-UE=4.75 direct link, and with the SP baselines constrained to the passive single-coefficient |phi|=1 model, a material part of the DP-vs-SP sum-rate gap in Fig. 5 may be an artifact of the RIS model rather than of dual polarization. The authors should replace (24e) with a physically realizable per-element constraint (e.g., sigma_max(S_l)<=1 for passive surfaces, or a unitary/reciprocal scattering matrix for lossless reciprocal surfaces) and re-run the simulations.","section":"Section II-C/IV-B, Eqs. (12), (24e), (46)"},{"comment":"The sensing model assumes perfect polarization selectivity at the targets. Eq. (16) sets y1 = A1 x1 + z1 and y2 = A2 x2 + z2 with x1 = E1 x and x2 = E2 x, so target 1 is illuminated only by the vertical component and target 2 only by the horizontal component; the rank-one response matrices in (18) contain no cross-polar entries. Real targets are characterized by a 2x2 polarimetric scattering matrix and depolarize the incident wave, so the v-pol and h-pol sensing beams will couple. Since the sensing SNR constraints (24b)-(24c) and the dual-target beampattern claim in Fig. 9 are built directly on this assumption, the claimed polarization-domain multi-target sensing advantage is not established for general targets. The authors should either explicitly restrict the claim to ideal polarization-selective point targets and discuss the restriction, or extend the model to include target cross-polar scattering.","section":"Section III-B, Eqs. (16)-(18) and constraints (24b)-(24c)"}],"minor_comments":[{"comment":"U_k is declared as C^{2x1}, but it must be C^{2x2} to estimate the 2x1 data vector d_k in (25); correct the dimension and the subsequent trace expressions.","section":"Section IV-A, Eq. (25)"},{"comment":"F and C are 2Lx2L matrices, so their 2x2 blocks are LxL, not N_t x N_t as written; the same symbol F is also used for the transmit covariance matrix sum F_k F_k^H, which is confusing.","section":"Section IV-B, Eq. (38)"},{"comment":"Eq. (15) uses log2, while Section V and Fig. 5 report rates in nat/s/Hz; make the log base and rate units consistent throughout.","section":"Eq. (15) vs Section V"},{"comment":"Eq. (35) has a typo in the summation index ('k=i'), and Eq. (36) introduces N_k and H_{ru,k} without definition; please clean up the notation in the Phi-subproblem derivation.","section":"Eqs. (35)-(36)"},{"comment":"The outer-loop continuation condition is described as the 'fractional increase' of (51), but (51) is a minimization objective; the criterion should be a decrease.","section":"Algorithm 2 and surrounding text"},{"comment":"The SP 1x baseline uses N_t single-polarized elements while the DP system uses N_t dual-polarized elements, i.e., twice the number of RF/antenna ports in the same aperture; state explicitly that the 'same hardware scale' comparison is on physical aperture/element count, not number of ports, and note the power-per-antenna implications.","section":"Section V-B, Fig. 5"}],"recommendation":"major_revision","confidential_remarks":"I recommend a major revision rather than rejection because the algorithmic framework is sound and the DP RIS constraint can in principle be corrected. The main risk is that the headline comparison in Fig. 5 will shrink or disappear once a passive per-element constraint is imposed; the authors should be asked to re-validate the claims under a physically consistent DP RIS model and to include a sensitivity analysis for target depolarization. I also suggest asking them to clarify the relation of their DP RIS model to the conventions used in the cited DP RIS papers [38]-[42]."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"First paper I've seen that combines dual-polarized BS, RIS, and users in ISAC, and the WMMSE/AO/MM/penalty machinery is competently applied. The problem is the DP RIS model. Constraint (24e) makes the four polarization-path coefficients per element independently unit-modulus. A physical passive element has a 2x2 scattering matrix bounded by passivity; letting all four entries be 1 gives a largest singular value of 2, i.e., 4x reflected power. The MM update coherently aligns phases to exploit this, and because the RIS link dominates (path-loss exponents 2.25+2.25 vs 4.75 direct), part of the Fig. 5 gap (DP 14-26 vs SP 1x 11-15) is an artifact. The SP baselines use the standard |phi|=1 model, so the comparison is asymmetric. This is load-bearing, not a nit. The sensing model is also idealized: (16)-(18) give target 1 only the vertical component and target 2 only the horizontal, with no cross-polarization scattering. That assumption is structural to the dual-target sensing claim, and real targets depolarize.\n\nCredit where due: the system model is new relative to the cited literature, the derivations are standard but carefully done, and the simulations show the expected qualitative trends for XPD, sensing thresholds, and quantization. Complexity analysis is reasonable. There are a few typos and sloppy definitions (e.g., 'KX k=i' in Eq. 35, F versus Fk in Section IV.D), but those are minor.\n\nBottom line: the idea is worth engaging and the algorithmic framework is reusable, but the performance claims need major revision. I'd send this to peer review with a request to impose a passivity constraint on the DP RIS element scattering matrix, relax or justify the polarization-isolated sensing model, and add Monte Carlo details or code. It's a candidate for a reading group discussion on model idealizations in RIS/ISAC.","headline":"First DP-RIS-ISAC design, but the headline polarization gain is inflated by an unphysical RIS model that lets each element amplify by up to 6 dB.","tokens_in":19967,"tokens_out":4129,"would_cite":true,"duration_ms":39279,"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":"Using dual-polarized transmit arrays and a dual-polarized RIS can roughly double the sum rate of an integrated sensing and communication system at the same physical array size, while exploiting the two polarizations to track two targets…","keywords":["integrated sensing and communication","dual-polarized channels","reconfigurable intelligent surface","sum rate maximization","dual-polarized beamforming","majorization-minimization","weighted minimum mean-square error","polarization diversity"],"falsifier":"Re-run the optimization with a target response matrix containing non-zero cross-polarization entries, then check whether both sensing constraints $\\gamma_1 \\geq \\gamma_{1,\\mathrm{th}}$ and $\\gamma_2 \\geq \\gamma_{2,\\mathrm{th}}$ can still be met while the sum-rate gain over single polarization persists; if the cross-polarized echo terms cause the two sensing constraints to interfere severely, the claimed dual-target advantage shrinks or disappears.","tokens_in":18767,"feed_emoji":"📡","tokens_out":9533,"duration_ms":80976,"temperature":0.7,"pith_summary":"This paper proposes an integrated sensing and communication (ISAC) system in which the base station, the reconfigurable intelligent surface (RIS), and the users all use dual-polarized (DP) antennas, and argues that exploiting the polarization domain materially improves performance. The system transmits vertical and horizontal signals simultaneously, using the two polarizations both to serve users and to illuminate two different radar targets, and jointly optimizes the DP transmit beamforming matrix and the DP RIS phase-shift matrix to maximize the sum rate subject to per-target sensing signal-to-noise-ratio constraints. The authors solve the non-convex problem by a WMMSE reformulation, alternating optimization, a majorization-minimization step for the unit-modulus RIS phases, and a penalty-based closed-form beamforming update. Simulations show the DP configuration achieves roughly double the sum rate of a single-polarized system with the same physical array size (about 26 vs 15 nat/s/Hz at ten transmit elements) while maintaining both sensing beams, and they quantify a trade-off between sensing quality and communication rate. The paper's central claim is that polarization is a free and substantial resource for ISAC: it doubles the effective array size without doubling the footprint.","feed_headline":"Two polarizations nearly double radar-plus-communication rates","feed_subtitle":"Two polarizations let one array roughly double ISAC data rates while tracking two radar targets.","key_machinery":"The load-bearing object is the polarization-domain decomposition of the transmit signal: the DP signal $x = \\sum_{k=1}^K F_k s_k$ is split by the selection matrices $E_1 = [I_{N_t} \\, 0_{N_t}]$ and $E_2 = [0_{N_t} \\, I_{N_t}]$ into vertical and horizontal parts, so that target 1 is illuminated only by $x_1 = E_1 x$ and target 2 only by $x_2 = E_2 x$. The DP channel is parameterised by an XPD factor $\\beta$ that sets the variance ratio between co-polar and cross-polar components, and the DP RIS is represented by a $2L \\times 2L$ phase-shift matrix $\\Phi$ whose diagonal blocks $\\Phi_{vv}$, $\\Phi_{hh}$ and off-diagonal blocks $\\Phi_{vh}$, $\\Phi_{hv}$ model polarization-preserving and polarization-converting reflection. The optimization machinery is a WMMSE reformulation of sum rate into a weighted MSE objective, an alternating optimization that separates the beamforming update from the RIS update, a majorization-minimization step with closed-form phase update $\\phi^{[t+1]} = e^{j \\arg(q^{[t]})}$ for the unit-modulus constraint, and a penalty-based beamforming algorithm whose inner subproblems have closed-form solutions $F_k = A_k^{-1} b_k$, $X_k^{\\mathrm{opt}} = F_k/(1+\\tau)$, and $Y_k^{\\mathrm{opt}} = V_1 F_k/(1-\\mu_1)$. This combination makes the joint non-convex problem tractable and lets the DP gain appear in simulation.","core_discovery":"The paper's central claim is that a dual-polarized architecture—DP base station, DP RIS, and DP user antennas—substantially outperforms an otherwise identical single-polarized ISAC system in achievable sum rate, while enabling simultaneous dual-target sensing by assigning one polarization to each target. The sensing model treats vertical and horizontal transmit components independently: target 1 is illuminated only by the vertical part $x_1 = E_1 x$ and target 2 only by the horizontal part $x_2 = E_2 x$, with rank-one target response matrices $A_1$ and $A_2$. Under this model, the joint optimization of the beamforming matrices $\\{F_k\\}$ and the RIS phase-shift matrix $\\Phi$ yields a beampattern in which the vertical and horizontal beams point at the two targets while the communication sum rate is maximized. The numerical results show the DP system scaling from 14 to 26 nat/s/Hz as the number of transmit elements grows from 4 to 10, versus 11 to 15 for a single-polarized system of the same size; an SP system with twice the physical size reaches 23 to 31 nat/s/Hz, slightly higher, confirming that DP recovers most of the gain of doubling the array without the extra footprint.","pith_inferences":["The paper leaves implicit that a depolarizing target would couple the two sensing constraints; a direct extension is to add cross-polarized echo terms to $\\gamma_1$ and $\\gamma_2$ and optimize a joint sensing constraint instead of two decoupled ones.","Because the DP system nearly matches an SP system with twice the physical aperture, the results imply a cost argument the paper does not state: at high frequencies where array footprint is tight, DP hardware may be a cheaper way to buy degrees of freedom than doubling the element count.","The polarization split suggests a sensing multiple-access scheme: with additional orthogonal bases such as $\\pm45^\\circ$ slanted or circular polarizations, more than two targets could be illuminated simultaneously, although the rank-one target model would need to be generalized.","The penalty-based closed-form beamforming update is largely independent of the specific sensing SNR thresholds, so the same algorithm could be adapted to other ISAC metrics, for example maximizing sensing mutual information under a sum-rate constraint, by swapping only the penalty terms."],"forward_implications":["At fixed physical array size, switching from single polarization to dual polarization roughly doubles the achievable sum rate (about 14 to 26 nat/s/Hz versus 11 to 15 nat/s/Hz as the transmit array grows from 4 to 10 elements) while preserving dual-target sensing.","The two orthogonal polarizations can act as separate sensing beams pointed at two different targets, giving concurrent multi-target detection and polarimetric target information without adding antennas.","Higher XPD, meaning cleaner polarization preservation, raises the sum rate up to a saturation point, and the gain is larger when the RIS has more elements.","There is a tunable trade-off between sensing and communication: raising the required sensing SNR lowers the communication sum rate, and more transmit antennas make the system less sensitive to that trade-off.","With only a few phase-quantization bits, roughly four or more, the DP system already nearly reaches its continuous-phase sum rate, so the dual-polarization gain survives practical RIS hardware."],"supporting_citations":[{"why":"Supplies the dual-polarized massive MIMO channel model and the SP 1x/SP 2x comparison setup used to benchmark the DP gain.","marker":"[17]"},{"why":"Provides the polarized MIMO channel model and the XPD-based variance split that defines co-polar and cross-polar channel components.","marker":"[14]"},{"why":"Supplies the majorization-minimization framework used to handle the unit-modulus DP RIS phase constraints.","marker":"[44]"},{"why":"Provides the measured statistical independence of orthogonal polarization components that justifies the zero cross-polar correlation assumption.","marker":"[43]"},{"why":"Supplies the polarization beampattern synthesis formula used to show that vertical and horizontal beams point at the two targets.","marker":"[45]"}],"fun_headline_variants":["Dual-polarized RIS nearly doubles ISAC sum rate","Two polarizations: one array, dual-target radar, higher rates","Polarization diversity boosts ISAC sum rate almost twofold","DP RIS: double the rate, two radar targets, same footprint"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The sensing model assumes each target reflects only the polarization aimed at it and never mixes vertical with horizontal; real targets depolarize, so this assumption carries the dual-target advantage.","fun_headline_variants_meta":{"raw":{"variants":["Dual-polarized RIS nearly doubles ISAC sum rate","Two polarizations: one array, dual-target radar, higher rates","Polarization diversity boosts ISAC sum rate almost twofold","DP RIS: double the rate, two radar targets, same footprint"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000199,"raw_usage":{"total_tokens":1399,"prompt_tokens":1000,"completion_tokens":399,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":616,"completion_tokens_details":{"reasoning_tokens":328}},"tokens_in":616,"tokens_out":399,"duration_ms":4575,"temperature":1.0,"reasoning_tokens":328,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-10T22:39:51.467541+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Re-run the optimization with a target response matrix containing non-zero cross-polarization entries, then check whether both sensing constraints $\\gamma_1 \\geq \\gamma_{1,\\mathrm{th}}$ and $\\gamma_2 \\geq \\gamma_{2,\\mathrm{th}}$ can still be met while the sum-rate gain over single polarization persists; if the cross-polarized echo terms cause the two sensing constraints to interfere severely, the claimed dual-target advantage shrinks or disappears.","supporting_citations":[{"cited_title":"Massive MIMO with dual-polarized antennas,","cited_arxiv_id":null,"evidence_quote":"Supplies the dual-polarized massive MIMO channel model and the SP 1x/SP 2x comparison setup used to benchmark the DP gain."},{"cited_title":"Modeling and capacity of polarized MIMO channels,","cited_arxiv_id":null,"evidence_quote":"Provides the polarized MIMO channel model and the XPD-based variance split that defines co-polar and cross-polar channel components."},{"cited_title":"Majorization-minimization algo- rithms in signal processing, communications, and machine learning,","cited_arxiv_id":null,"evidence_quote":"Supplies the majorization-minimization framework used to handle the unit-modulus DP RIS phase constraints."},{"cited_title":"Propagation characteristics of polarized radio waves in cellular commu- nications,","cited_arxiv_id":null,"evidence_quote":"Provides the measured statistical independence of orthogonal polarization components that justifies the zero cross-polar correlation assumption."},{"cited_title":"Optimal polarization synthesis of arbitrary arrays with focused power pattern,","cited_arxiv_id":null,"evidence_quote":"Supplies the polarization beampattern synthesis formula used to show that vertical and horizontal beams point at the two targets."}],"review_version":1}