{"id":"9bb5ffdc-cd1a-4a66-aa9b-650eb9f6c572","arxiv_id":"2412.14657","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":2,"one_line_summary":"In isotropic scattering, antenna directivity makes wavenumber-domain channel gains uneven, so XL-MIMO effective degrees of freedom depend on antenna type and element spacing.","lead":"This paper shows that the directional radiation pattern of antenna elements changes the number of usable communication modes in extremely large MIMO systems, even when radio waves arrive equally from all directions. It gives a formula and an EM-simulation-based method to compute these modes for different antenna types and spacings, which is useful for designing 6G antenna surfaces.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Theorem 1 integrates over wavenumber cells shifted half a cell from the sampling grid; this biases coupling coefficients and EDoF.","rationale":"The reader's weakest-assumption pick (the unit-cell EMCC boundary condition) is plausible but secondary. The paper's central contribution is Theorem 1, the closed-form directivity-aware coupling coefficient. The load-bearing defect I identify is in that theorem: the integration region K_T defined in (13) is shifted half a cell from the sampling grid in (3). In the stationary wavenumber-domain model of [8], each Fourier coefficient's variance is the integral of the angular power spectrum over the cell centered at the corresponding wavenumber; using a corner-aligned cell changes which portion of the spectrum is attributed to each mode. Because the spectrum in (15) is even, this is not a harmless relabeling: it redistributes power among adjacent modes, and for directive antennas with a sharp normal-incidence peak it significantly reduces the largest coefficient. Since EDoF and capacity are functionals of the sorted coefficient distribution, the computed figures shift. The EMCC estimator in (23) is a least-squares projection and therefore has a centered response, so it cannot validate the offset formula for non-uniform spectra. The fix is routine, but it changes the theorem and all numerical results. Thus the verdict should remain CONDITIONAL with a newly required, well-specified condition: re-derive Theorem 1 using centered cells and regenerate Figs. 3-6. I disagree with the reader's claim that the analytic derivation is correct.","tokens_in":9249,"tokens_out":13274,"duration_ms":98201,"concrete_test":"Recompute σ²_T(mx,my) from Eq. (15) with the corrected centered cell K'_T(mx,my) = [(mx−1/2)λ/Lx, (mx+1/2)λ/Lx] × [(my−1/2)λ/Ly, (my+1/2)λ/Ly] ∩ {(kx,ky): kx²+ky²≤1}, and regenerate Figs. 3-6. Compare the resulting EDoF and capacity curves with the SVD benchmarks. In addition, run the EMCC procedure of Section III-C for a directive pattern such as G(θ,φ)=cos²θ and compare the LS-estimated coefficient for mode (0,0) against the offset-cell integral (13) and the centered-cell integral; if the EMCC matches the centered-cell prediction and not (13), Theorem 1 and all downstream numerical results require correction.","verdict_should_be":"CONDITIONAL","load_bearing_attack":"Equation (13) assigns to the mode (mx,my) the wavenumber cell K_T(mx,my) = [mxλ/Lx, (mx+1)λ/Lx] × [myλ/Ly, (my+1)λ/Ly], whose lower-left corner is the sample point (mxλ/Lx, myλ/Ly) from Eq. (3). In the wavenumber-domain representation of a spatially stationary channel on an aperture of size L, the variance of the coefficient at sample (mx,my) is the integral of the angular power spectrum over the cell centered at that sample, i.e. [(mx−1/2)λ/Lx, (mx+1/2)λ/Lx] × [(my−1/2)λ/Ly, (my+1/2)λ/Ly]. The cell in (13) is therefore shifted by half a cell in each dimension. For the even spectrum in Theorem 1 this is not a harmless relabeling: the power physically belonging to the origin mode is spread across four adjacent offset cells. For directive antennas (m>1), where the spectrum is sharply peaked at kx=ky=0, the coefficient assigned to mode (0,0) is a small fraction of the true value, while neighboring modes are inflated. Since EDoF in (19)-(20) is computed from the sorted coefficients and capacity in (10) uses their distribution, the shift changes both. The EMCC estimator in (23) is a least-squares projection onto the Fourier basis; its main lobe is centered on the sampled wavenumbers, so it estimates centered-cell variances. Thus (15) and the EMCC cannot agree except for the special uniform case m=1. This puts in doubt the Fig. 3 validation and the Figs. 4-6 results for directive antennas.","agreement_with_reader":"disagree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper analyzes the effect of antenna element directivity on the effective number of degrees of freedom (EDoF) and ergodic capacity of extremely large-scale MIMO (XL-MIMO) in isotropic Rayleigh fading. It introduces a wavenumber-domain coupling coefficient that incorporates the antenna radiation pattern, derives a closed-form expression (Theorem 1) for a cos^m(theta) pattern, defines EDoF from the cumulative distribution of sorted coupling coefficients, and proposes a full-wave-simulation-based numerical method (EMCC) to extract these coefficients. Numerical results are given for patch, dipole, RIS, and hypothetical antennas, showing how coupling coefficients, EDoF, and capacity vary with element spacing.","tokens_in":9629,"tokens_out":12960,"duration_ms":98351,"significance":"If the results are correct, the paper extends the wavenumber-domain framework for spatially stationary XL-MIMO channels to practical directional antenna elements and provides a numerical pipeline based on full-wave simulation. The derivation of Theorem 1 is self-contained and transparent, and the idea of using coupling-coefficient distributions to compute EDoF is natural. However, the paper currently contains a load-bearing definitional issue in the wavenumber-cell assignment and an unspecified threshold in the EDoF definition, both of which affect the numerical claims. The contribution is potentially useful, but the manuscript needs substantive correction and re-validation before it can be accepted.","major_comments":[{"comment":"The cell K_T(m_x,m_y) is defined as [m_x lambda/L_x, (m_x+1)lambda/L_x] x [m_y lambda/L_y, (m_y+1)lambda/L_y], so its lower-left corner is the sampled wavenumber from Eq. (3). In the wavenumber-domain representation of a spatially stationary field on a finite aperture, the variance associated with sample (m_x,m_y) is the integral of the angular power spectrum over the cell centered at that sample, namely [(m_x-1/2)lambda/L_x, (m_x+1/2)lambda/L_x] x [(m_y-1/2)lambda/L_y, (m_y+1/2)lambda/L_y]. The definition in (13) is therefore shifted by half a cell in each dimension. For the even spectrum in Theorem 1 this is not a harmless relabeling: for m>1, the spectral peak at k_x=k_y=0 is split among four adjacent cells, so sigma_T^2(0,0) is undercounted and the neighboring coefficients are inflated. Since Eqs. (19)-(20) sort these coefficients and Eq. (10) uses their distribution, the EDoF and capacity results change. In addition, the LS estimator in Eq. (23) projects onto Fourier basis functions whose main lobes are centered on the sampled wavenumbers, so the EMCC method estimates centered-cell variances; the agreement between Eq. (15) and EMCC shown in Fig. 3 cannot be expected to hold for non-constant spectra. Please correct the cell definition (or provide a reference that justifies the lower-left cell) and recompute the affected figures.","section":"Section III-A, Eq. (13)"},{"comment":"The EDoF is defined as the number of largest coupling coefficients whose cumulative power reaches a fraction gamma, but no value of gamma is reported in Section IV. The SVD-based dotted curves in Fig. 5 also count 'dominant singular values' without a stated threshold criterion. Without these thresholds, the close agreement between the solid and dotted curves in Fig. 5 is not a falsifiable validation, because gamma can be chosen to match the SVD count. Please specify gamma, specify the SVD dominance criterion, and report sensitivity of EDoF and capacity to gamma.","section":"Section III-B, Eqs. (19)-(20), and Fig. 5"},{"comment":"The EMCC method obtains a single realized gain pattern from a unit-cell periodic full-wave simulation and then simulates the spatial channel by weighting each multipath by sqrt(G(theta,phi)) independently for every element. For a finite XL-MIMO array with subwavelength spacing, edge effects and element-dependent mutual coupling can make the realized pattern position-dependent; the paper provides no comparison against a finite-array full-wave simulation or a measured pattern. The coupling coefficients in Fig. 4 and the EDoF/capacity curves in Figs. 5-6 therefore rely on an unverified approximation. Please either provide such a validation or explicitly state this approximation as a limitation and discuss its expected impact.","section":"Section III-C, Eq. (21), and Section IV"}],"minor_comments":[{"comment":"The y-axis label 'EffecitveDoF' should be 'Effective DoF'.","section":"Fig. 5"},{"comment":"Please specify the number of multipaths S, the number of LS runs I, and clarify the y-axis formatting ('2 10-3' should be 2 x 10^-3).","section":"Fig. 3 and Section III-C"},{"comment":"The value of gamma in (19) and the SNR or noise power mu^2 in (10) are not given; both are needed to reproduce Fig. 6.","section":"Section III-B and Fig. 6"},{"comment":"Eq. (22) states that ha is CN(0, 2 sigma_T^2), while Eq. (7) writes Ha ~ CN(0, sigma^2); please reconcile the factor of 2 and state whether the plotted coupling coefficients are sigma_T^2 or 2 sigma_T^2.","section":"Eq. (22) and Section III-C"},{"comment":"The notation diag(sigma_T ⊙ sigma_T) should be defined explicitly, since sigma_T ⊙ sigma_T is a vector of variances and the diagonal operator is overloaded.","section":"Eq. (10)"},{"comment":"The caption does not explain which curves are solid and which are dotted; the text should clarify that solid curves use eta_e and dotted curves use eta_u.","section":"Fig. 6"},{"comment":"Please state that G is a power pattern and that large-scale path loss is omitted; otherwise the normalization by 1/sqrt(S) and the units of h(x,y) are ambiguous.","section":"Eq. (21)"},{"comment":"There are several typographical artifacts ('Extreme ly', 'influenced', and similar spacing issues) throughout the manuscript; please proofread carefully.","section":"Title/Abstract"}],"recommendation":"major_revision","confidential_remarks":"This is a borderline case. The central idea is useful and likely salvageable, but the half-cell shift in Eq. (13) directly affects the reported coupling coefficients, EDoF, and capacity, and the comparison with SVD-based EDoF in Fig. 5 is not a validation without stated thresholds. Both issues are fixable within the scope of the manuscript, so I recommend major revision rather than rejection. The novelty is moderate, but the topic is timely for XL-MIMO analysis."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The real value of this paper is the EMCC pipeline: using unit-cell full-wave simulation to get realized gain, simulating wavenumber-domain channels with that gain, and estimating coupling coefficients by least-squares projection and Gaussian fitting. That is a practical extension of the wavenumber-domain framework in [8], which only handled hypothetical isotropic elements. The paper also shows convincingly that element directivity and element spacing change the distribution of coupling coefficients, EDoF, and ergodic capacity. The qualitative conclusions about RIS versus patch versus dipole are plausible and worth having in the literature.\n\nThe soft spot is in the theory. In Eq. (3) the sampled wavenumbers are at (m_x λ/L_x, m_y λ/L_y), but the integration cell in Eq. (13) runs from m_x λ/L_x to (m_x+1)λ/L_x, i.e., a half-cell shift. For a stationary spatial process expanded on a finite aperture, the variance of the coefficient at the sample belongs to the cell centered on that sample, not the cell starting there. So Theorem 1's Eq. (15) is not the variance of the actual wavenumber-domain coefficients in Eq. (7). This matters most for directive antennas (m>1), where the spectrum is peaked at the origin: the (0,0) coefficient gets a small fraction of its true power, and neighbors are inflated. It also weakens the validation in Fig. 3: if EMCC estimates centered-cell variances and Eq. (15) computes shifted-cell integrals, agreement is not a clean check. The stress-test note that flagged this is correct on the arithmetic; I checked the bin boundaries against the sampling grid.\n\nThe other concerns are less severe but real: the EDoF threshold gamma in Eqs. (19)-(20) is never specified, so Fig. 5's 'match' between statistical EDoF and SVD counts is not an independent benchmark; there are no error bars or realization counts for the simulated curves; and no code or data are provided. The unit-cell mutual-coupling assumption is a reasonable approximation but unvalidated for finite arrays with edge effects.\n\nDespite these problems, the EMCC approach itself is useful and likely salvageable. I would send this to peer review with a request to fix the bin alignment, specify gamma, and provide code/data. The paper deserves referee time, but the theoretical formula in Theorem 1 should not be accepted as-is.","headline":"Useful EMCC pipeline for directivity-aware wavenumber-domain coupling, but Theorem 1's integration cell is shifted half a cell from the sampling grid and the EDoF threshold is unspecified, so the validation needs a closer look.","tokens_in":10121,"tokens_out":4518,"would_cite":false,"duration_ms":52636,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["94A05","78A50"],"pacs":[],"model":"deepseek-v4-flash","headline":"Antenna directivity remaps the spatial modes an XL-MIMO array can use.","keywords":["XL-MIMO","effective degrees of freedom","wavenumber-domain channel","antenna directivity","mutual coupling","ergodic channel capacity","reconfigurable intelligent surfaces"],"falsifier":"Measure the spatial channel of a fabricated $10\\lambda \\times 10\\lambda$ patch or dipole array in an isotropic scattering chamber, compute the coupling coefficients from separately measured element patterns, and count dominant singular values; if the measured EDoF disagrees with the prediction of Eqs. (19)--(20) for any antenna type or spacing, the central claim fails.","tokens_in":9039,"feed_emoji":"📡","tokens_out":9150,"duration_ms":74946,"temperature":0.7,"pith_summary":"The paper asks whether the directional radiation pattern of real antenna elements changes how many independent spatial channels an extremely large-scale MIMO (XL-MIMO) surface can support, and answers yes. Working in an isotropic Rayleigh-fading channel and using the wavenumber-domain channel representation, it shows that directivity reshapes the coupling coefficients that weight each plane-wave direction, and that the effective degrees of freedom (EDoF) and ergodic capacity are governed by the distribution of these coefficients rather than by aperture alone. The paper supplies a closed-form coupling coefficient for a general $\\cos^m(\\theta)$ radiation pattern, a numerical simulation pipeline for arbitrary antenna structures, and simulations showing that different antenna types and element spacings produce different EDoF and capacity. This matters because EDoF is the standard measure of how many data streams a large antenna surface can carry, so the result turns antenna-element design and spacing into first-order determinants of XL-MIMO performance.","feed_headline":"Antenna pattern, not array size alone, sets XL-MIMO capacity","feed_subtitle":"Wavenumber-domain analysis shows element radiation shape and spacing govern effective degrees of freedom.","key_machinery":"The wavenumber-domain channel representation: the spatial channel is expanded as a finite sum of plane waves whose wavevectors are sampled on a grid set by the array aperture, so channel gain toward each direction becomes a scalar coupling coefficient. The paper's load-bearing identity is Theorem 1's integral, which converts the angular radiation pattern and the isotropic angle distribution into a per-cell weight $(1-k_x^2-k_y^2)^{(m-1)/2}/(2\\pi)$. Threshold-based EDoF in Eqs. (19)--(20) then counts how many of these weights are dominant, and the EMCC procedure obtains the weights numerically from full-wave element patterns by least-squares projection and variance fitting.","core_discovery":"The central claim is Theorem 1: for an element with radiation pattern $G(\\theta_T,\\phi_T)=\\cos^m(\\theta_T)$ in an isotropic scattering Rayleigh-fading channel, the wavenumber-domain coupling coefficient equals $\\sigma_T^2(m_x,m_y) = \\frac{1}{2\\pi}\\iint_{K_T(m_x,m_y)} (1-k_x^2-k_y^2)^{(m-1)/2}\\,dk_x\\,dk_y$, with the integral over the wavenumber cell $K_T$ defined by the array aperture. This makes the coefficient distribution pattern-dependent: $m=1$ gives uniform coefficients, $m>1$ concentrates energy at broadside, and $m=0$ gives a bowl-shaped distribution. The paper then defines EDoF as the smallest number of largest coupling coefficients at each side that together contain a fraction $\\gamma$ of the total coupling power, and verifies that this statistical definition matches deterministic singular-value counting in simulation. It also demonstrates that the EMCC numerical method recovers the same coefficients from full-wave simulated patterns, and that EDoF and capacity vary with both element type and spacing, with the considered RIS array reaching higher EDoF than patch or dipole arrays.","pith_inferences":["Beyond the paper: because the coupling-coefficient formula is written for a generic angle distribution $p(\\theta,\\phi)$, the same EDoF criterion should apply to measured or ray-traced non-isotropic scattering, though the paper only demonstrates the isotropic case.","Beyond the paper: the unit-cell assumption can be stress-tested by comparing arrays of very different physical size, since edge elements in a finite panel are expected to deviate from the periodic-boundary pattern; measuring per-element patterns would show how much EDoF shifts.","Beyond the paper: the evenness of the wavenumber coupling distribution, not realized gain alone, is the quantity that tracks EDoF, suggesting a design rule for element spacing that the paper does not state explicitly."],"forward_implications":["For ideal elements with $\\cos^m$ patterns, only $m=1$ yields a uniform coupling distribution; larger $m$ concentrates gain at broadside and reduces the number of EDoF for a fixed aperture.","Replacing the upper-bound degree of freedom $\\eta_u$ with the threshold-based $\\eta_e$ in the capacity expression reproduces the simulated capacity, so weak coupling coefficients can safely be dropped from the capacity sum.","Element spacing changes the realized pattern through mutual coupling, so EDoF and capacity are not monotonic in spacing; the best spacing depends on the element type, with the considered RIS optimum at about $0.4375\\lambda$."],"supporting_citations":[{"why":"supplies the wavenumber-domain channel representation, the finite plane-wave sampling, and the separable coupling-coefficient model used throughout.","marker":"[8]"},{"why":"gives the cos^m directivity model and the unit-cell treatment of mutual coupling between elements.","marker":"[10]"},{"why":"provides the isotropic half-space angle distribution p(theta,phi)=sin(theta)/(2 pi) used in Theorem 1.","marker":"[12]"},{"why":"defines EDoF as the number of dominant eigenvalues, the concept the paper's threshold-based definition extends.","marker":"[5]"},{"why":"provides the deterministic SVD-based EDoF counting that the simulations compare against.","marker":"[7]"},{"why":"supplies the RIS antenna array example whose 0.434 lambda element spacing is close to the paper's found optimum.","marker":"[4]"},{"why":"explains how element spacing modifies the radiation pattern through equivalent RLC circuit parameters.","marker":"[13]"}],"fun_headline_variants":["Antenna shape, not size, dictates XL-MIMO capacity","Radiation pattern sets XL-MIMO degrees of freedom","Directivity drives XL-MIMO capacity, not just array size","XL-MIMO capacity hinges on antenna radiation pattern","Wavenumber view: antenna directivity sets XL-MIMO EDoF"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing assumption is that a unit-cell full-wave simulation with periodic boundaries gives the radiation pattern of every element in the finite XL-MIMO array, so that mutual coupling and edge effects enter only through one average pattern.","fun_headline_variants_meta":{"raw":{"variants":["Antenna shape, not size, dictates XL-MIMO capacity","Radiation pattern sets XL-MIMO degrees of freedom","Directivity drives XL-MIMO capacity, not just array size","XL-MIMO capacity hinges on antenna radiation pattern","Wavenumber view: antenna directivity sets XL-MIMO EDoF"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000196,"raw_usage":{"total_tokens":1376,"prompt_tokens":977,"completion_tokens":399,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":593,"completion_tokens_details":{"reasoning_tokens":329}},"tokens_in":593,"tokens_out":399,"duration_ms":4020,"temperature":1.0,"reasoning_tokens":329,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-11T12:03:07.586602+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Measure the spatial channel of a fabricated $10\\lambda \\times 10\\lambda$ patch or dipole array in an isotropic scattering chamber, compute the coupling coefficients from separately measured element patterns, and count dominant singular values; if the measured EDoF disagrees with the prediction of Eqs. (19)--(20) for any antenna type or spacing, the central claim fails.","supporting_citations":[{"cited_title":"Spatially -stationary model for holographic MIMO small-scale fading,","cited_arxiv_id":null,"evidence_quote":"supplies the wavenumber-domain channel representation, the finite plane-wave sampling, and the separable coupling-coefficient model used throughout."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"gives the cos^m directivity model and the unit-cell treatment of mutual coupling between elements."},{"cited_title":"Small-scale spatial-temporal corre lation and de- grees of freedom for reconfigurable intelligent surfaces,","cited_arxiv_id":null,"evidence_quote":"provides the isotropic half-space angle distribution p(theta,phi)=sin(theta)/(2 pi) used in Theorem 1."},{"cited_title":"On Landau’s eigenvalue theorem f or line-of- sight MIMO channels,","cited_arxiv_id":null,"evidence_quote":"defines EDoF as the number of dominant eigenvalues, the concept the paper's threshold-based definition extends."},{"cited_title":"Communicating with large intelligent surf aces: Fundamental limits and models,","cited_arxiv_id":null,"evidence_quote":"provides the deterministic SVD-based EDoF counting that the simulations compare against."},{"cited_title":"RIS-based IMT-2030 testbed for MmWave multi-stream ultra-massive MIMO communications,","cited_arxiv_id":null,"evidence_quote":"supplies the RIS antenna array example whose 0.434 lambda element spacing is close to the paper's found optimum."},{"cited_title":"An overview of eq uivalent circuit modeling techniques of frequency selective surfac es and metasur- faces,","cited_arxiv_id":null,"evidence_quote":"explains how element spacing modifies the radiation pattern through equivalent RLC circuit parameters."}],"review_version":1}