{"id":"caa21b89-996d-463e-b02d-df5743885601","arxiv_id":"2502.04819","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"IQI imposes a severe rate penalty on power-limited, ultra-wideband THz array-of-subarrays systems, captured by new wideband slope expressions.","lead":"This paper analyzes how I/Q imbalance (IQI) affects terahertz (THz) communication systems that use large antenna arrays and ultra-wide bandwidths. It derives theoretical limits on spectral efficiency and shows that IQI can sharply limit data rates in THz links, more so than at lower frequencies.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The central quantitative claim rests on wideband-slope formulas (24)-(25), which are asserted by analogy with [15] rather than derived from Verdu's definition; given the mirror-subcarrier coupling and the inconsistent noise covariance (16), the IQI impact is not yet rigorously established.","rationale":"The reader's weakest-assumption analysis identifies the same load-bearing concern: Eqs. (24)-(25) are used as the technical basis for the abstract's claim, but they are never derived from the definition of wideband slope or from the actual signal model. My independent reading of the manuscript confirms this gap. In addition, the noise covariance inconsistency in (16) is not merely a typo; it affects the SINR expression (23) and hence the very inputs to (24)-(25), because the IQI-modified noise variance is a different quantity than sigma^2. The qualitative claim that IQI imposes a rate ceiling is almost certainly correct--as transmit power grows, the IQI interference term in the denominator of (23) makes the SINR saturate--but the quantitative wideband-slope result is not established. The paper does have independent value: it models IQI for the THz AoSA architecture, and the simulation comparison of IQI versus IUI is informative. Thus the appropriate disposition is unchanged from the reader's conditional acceptance: the derivation gap should be closed in revision, but the overall direction and practical message are credible. My concern does not move the verdict because the qualitative conclusion is robust and the paper already warrants conditional acceptance for revisions; a full rejection would be disproportionate given that the main phenomenon is real and the simulation evidence supports the qualitative behavior.","tokens_in":7661,"tokens_out":6042,"duration_ms":72525,"concrete_test":"Build a minimal instance of the model in (3): two subcarriers k and -k, one or two users, scalar IQI coefficients G1, G2, K1, K2, and a diagonal frequency-selective channel Hc[k], Hc[-k]. Compute the sum rate R[k] + R[-k] as an explicit function of total transmit power P with equal power allocation. Then numerically or symbolically compute the wideband slope from the definition, S0 = 2 lim_{P->0} C(P) / (P/N0)^2 / log2(e)? (or equivalently the second derivative of the spectral-efficiency expansion at P=0). Compare this value with Eq. (25). If they differ, the substitution of the SINR into [15] is invalid, and the paper's quantitative claim needs revision. As a secondary check, recompute the noise covariance from (3) and verify whether (16) matches; it will not unless K1 and K2 are identity and the receiver combiner is ignored.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The paper's central quantitative claim that IQI strictly limits achievable rate in the wideband, power-limited regime depends on Eqs. (24)-(25). These are obtained by inserting the IQI-affected SINR (23) into the interference-channel formulas (21)-(22) of Shen and Høst-Madsen [15]. This substitution is not a theorem: [15] derives the wideband slope for an interference channel where each interferer is an independent same-band user. In the present OFDM system, the IQI-induced interference comes from the mirror subcarrier s[-k], which is statistically independent but spectrally coupled to s[k]; the low-SNR capacity expansion must be taken over the full 2K-subcarrier system, and it is not shown that [15]'s scalar SINR substitution preserves the second derivative of the sum-rate. There is also a concrete internal inconsistency: the noise covariance in (15)-(16) is written as sigma^2/2 (I + G_T G_T^H), but the noise term in (3) is W_R (K1 z[k] + K2 z*[-k]), whose covariance involves W_R, K1, K2, and receiver-side imbalance matrices, not the transmitter amplitude-error matrix G_T. Since (23) uses the same sigma^2 in the denominator, the SINR itself is not the true SINR of the model. These issues do not necessarily overturn the qualitative conclusion--IQI self-interference does impose a rate ceiling--but they mean the paper's headline quantitative contribution, the wideband slope with IQI, is not currently supported by a valid derivation. The simulations in Section V do not close this gap because they operate at moderate SNR and do not extract the low-SNR wideband slope from the underlying capacity expression.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper investigates in-phase/quadrature imbalance (IQI) in a multi-user MIMO OFDM terahertz system with an array-of-subarrays architecture and hybrid beamforming. It models transmitter and receiver IQI, expresses the SINR in terms of desired and image channels H_d and H_i, and then applies wideband-regime metrics from Verdú and from Shen and Høst-Madsen to obtain the minimum energy per bit and wideband slope in Eqs. (24)-(25). Numerical results show rate saturation with IQI, compare THz with lower-frequency behavior, and propose subcarrier nulling as a mitigation. The paper's headline claim is that IQI imposes a strict limit on achievable rate in power-limited ultra-wideband THz links.","tokens_in":7979,"tokens_out":6183,"duration_ms":64868,"significance":"If the wideband-slope formulas were rigorously derived, the paper would make a useful contribution: it would quantify how IQI-induced self-interference enters the denominator of the wideband slope and raises the minimum energy per bit, and it would support the practical suggestion of subcarrier nulling. The paper also grounds the THz-versus-low-frequency comparison in measured image rejection ratios, so the qualitative point that IQI is more severe at THz is plausible. However, the central quantitative claim currently rests on an unproved analogy with an interference-channel result, and the noise model has an internal inconsistency, so the result is conditional rather than established. No code or reproducibility artifacts are provided.","major_comments":[{"comment":"The derivation of the IQI wideband slope is not given. Equations (24) and (25) are obtained by inserting the SINR gamma_m[k] from (23) into the interference-channel expressions (21) and (22) of Shen and Høst-Madsen [15], but the paper never computes the second derivative of spectral efficiency at zero rate that defines S0 in Verdú's framework. In the present OFDM system the IQI interference originates from the mirror subcarrier of the same user, so the sum rate couples s[k] and s[-k]; it is not established that the formulas of [15], derived for independent same-band interferers, apply after this substitution. Without a derivation from the definition of wideband slope, or a theorem showing that the mapping preserves S0 and Eb/N0_min, Eqs. (24)-(25) remain an assumption, not a proven result.","section":"Section IV-B, Eqs. (24)-(25)"},{"comment":"The noise treatment is internally inconsistent. In (1), z[k] is MQ-dimensional array noise, but after analog combining the noise entering y[k] should involve F_R^H z[k] and its image term, not W_R[k](K1[k]z[k]+K2[k]z*[-k]) as written in (3), because K1 and K2 act on the baseband signal dimension after combining. Correspondingly, the covariance ar Z in (16), written as sigma^2/2 (I + G_T G_T^H), is not the covariance of the noise term in (3): it lacks F_R, W_R, K1, and K2, and it uses the transmitter amplitude-error matrix G_T even though the text says the matrix is drawn for the receiver. Since this same sigma^2 appears in the denominator of the SINR (23), the numerical results in Section V are based on an incorrect noise covariance.","section":"Section III-B, Eqs. (3), (16), (23)"},{"comment":"The simulations do not test the low-SNR asymptotic regime in which Eqs. (18), (24), and (25) apply. Figures 5-8 plot rate against transmit power budget over an SNR range of 0 to 9 dB, whereas the wideband slope is an asymptotic quantity evaluated as Eb/N0 approaches Eb/N0_min, which typically corresponds to negative SNR in dB. The figures therefore do not validate the wideband-slope computation, and the claim that IQI imposes a strict rate limit is illustrated only at moderate SNR values, not at the operating point where the wideband analysis is performed.","section":"Section V, Figs. 5-8"}],"minor_comments":[{"comment":"The steering-vector expression in (10) appears to have a typo: the x and y contributions are both written as cos(phi_0) sin(theta_0), so the formula cannot describe a general 3D array response; the y term likely should contain sin(phi_0) sin(theta_0).","section":"Eq. (10)"},{"comment":"The definitions in (5) are missing commas inside the diag arguments: they should read diag{g_x,1, g_x,2, ..., g_x,N_x} and diag{Phi_x,1, Phi_x,2, ..., Phi_x,N_x}.","section":"Eq. (5)"},{"comment":"The sentence 'the transmit symbol vector is denoted by' is incomplete; the symbol vector should be defined explicitly with its distribution and dimension.","section":"Section III-B, after Eq. (3)"},{"comment":"The abbreviation 'SC nulling' in the figure captions is not defined; it should be spelled out as 'subcarrier nulling' at first use.","section":"Figures 7-8"},{"comment":"There are numerous typos and grammatical errors, including 'perforamance', 'equiped', 'spatioal', 'communication' in the introduction, and the fragment 'and the same issue occurs during upconversion at the transmitter' at the end of Section II; a thorough language edit is needed.","section":"Throughout"}],"recommendation":"major_revision","confidential_remarks":"The paper is within the scope of the journal and addresses a relevant THz hardware-imperfection problem, but the main theoretical contribution needs a genuine derivation rather than an analogy-based substitution, and the noise model needs to be fixed. If the authors can derive Eqs. (24)-(25) from Verdú's definition and correct the noise covariance, the paper could become acceptable; otherwise the quantitative wideband claims should be removed or explicitly presented as heuristic."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"First thing to know: this paper takes a real practical problem—I/Q imbalance in THz arrays—and gives it a wideband, power-limited framing that hasn't been done before. The qualitative conclusion, that IQI is a first-order bottleneck in THz links, is likely right and worth saying. But the central quantitative machinery, the wideband slope with IQI in Eqs. (24)–(25), is asserted via substitution from [15], not derived, and there's an internal inconsistency in the noise covariance. So treat the numbers as not yet established.\n\nWhat's genuinely new is the combination of the AoSA THz architecture with Verdu's wideband metrics under IQI. The system model in Sec. III is coherent: the IQI coupling between mirror subcarriers is correctly written, and the authors identify the right interference terms. The subcarrier-nulling mitigation is sensible and, in the simulations, clearly outperforms full bandwidth at THz. The comparison with lower frequencies, using measured image rejection ratios from [3] and [4], makes a useful point: IQI is worse at THz because the hardware is less mature. That's a design-level insight that will matter to people building 6G transceivers.\n\nThe soft spots are the load-bearing ones. Eqs. (24)–(25) are obtained by replacing the SINR in the interference-channel formulas of [15] with (23). But [15] derives the wideband slope for interferers that are independent same-band users; here the interferer is the mirror subcarrier, which is spectrally coupled through the IQI model. The paper never shows that the second derivative of the sum spectral efficiency at zero SNR reduces to that substitution. It may be true, but it needs proof. Also, the noise covariance in (16) is sigma^2/2 (I + G_T G_T^H), which references the transmitter amplitude error while the text says receiver noise; the actual noise in (3) is WR (K1 z + K2 z*[-k]), so the SINR in (23) is not the true SINR of the model. The simulations don't rescue this: they run at SNR 0–9 dB, not in the low-SNR asymptotic regime where the wideband slope is defined.\n\nThis paper is for THz system engineers wanting a first-order view of the IQI penalty. It deserves a serious referee—the problem is important and the direction is sensible—but the referee should insist on a derivation of the wideband slope under IQI, a corrected noise model, and low-SNR simulations. If those come, the paper could be a useful reference.","headline":"The qualitative IQI-in-THz point is sensible and practically relevant, but the wideband-slope formulas (24)–(25) are asserted by analogy rather than derived, and the noise covariance has an internal inconsistency.","tokens_in":8539,"tokens_out":3910,"would_cite":false,"duration_ms":37275,"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":"I/Q imbalance puts a strict cap on terahertz data rates.","keywords":["Terahertz communication","I/Q imbalance","wideband slope","minimum energy per bit","power-limited regime","MU-MIMO OFDM","array-of-subarrays","hardware impairments"],"falsifier":"Simulate the 300 GHz IQI system model of the paper and compute the spectral efficiency at low SNR directly, then evaluate the wideband slope as the second derivative of spectral efficiency with respect to $E_b/N_0$ at the minimum bit energy and compare it with (25); a mismatch would mean the substitution that produces (25) is invalid and the strict limit is not established.","tokens_in":7417,"feed_emoji":"📡","tokens_out":9440,"duration_ms":88739,"temperature":0.7,"pith_summary":"The paper sets out to show that I/Q imbalance, a hardware mismatch between the in-phase and quadrature branches of the local oscillator, is not a minor distortion in terahertz links but a central obstacle to their promised data rates. It models IQI in an array-of-subarrays MU-MIMO OFDM downlink and studies the power-limited ultra-wideband regime, where the relevant measures are the wideband slope and the minimum energy per bit. Its central claim is that THz systems with IQI have a strict limit on achievable rate even though the band offers enormous spectrum, because IQI turns the mirror-image subcarrier into self-interference that grows with transmit power. The paper further claims that at THz frequencies this self-interference is more damaging than inter-user interference, the opposite of the usual low-frequency ordering. If correct, this means harvesting the THz spectrum requires dealing with IQI directly, not just adding power or bandwidth.","feed_headline":"IQ imbalance caps terahertz data rates","feed_subtitle":"Mirror-image interference from I/Q mismatch grows with power, so rate saturates despite huge bandwidth.","key_machinery":"The central object is the wideband slope $S_0$ together with the minimum bit energy $E_b/N_{0,\\min}$, the two parameters of the low-SNR spectral-efficiency line $SE \\approx S_0\\,(E_b/N_0 - E_b/N_{0,\\min})/3\\,\\mathrm{dB}$. The wideband slope measures how much spectral efficiency each additional unit of power buys in the energy-efficient regime, and the argument is that IQI shrinks it by adding the interference term $\\zeta_m$ to the denominator of (25). The mechanism is a substitution: the IQI-affected SINR, built from the desired channel $H_d$ and the image channel $H_i$, is fed into the interference-channel formulas of [15] that give (21)-(22). Around this sit the THz-specific ingredients: the array-of-subarrays antenna structure, analog beamforming aligned to the dominant line-of-sight path, and the multicarrier mirror-subcarrier model in which the image band acts as the interferer.","core_discovery":"The paper's central claim is that in the power-limited wideband regime, in-phase/quadrature imbalance adds a mirror-image self-interference term $\\zeta_m$ to the denominator of the wideband slope $S_0$ and raises the minimum energy per bit $E_b/N_0$, so an IQI-affected THz MU-MIMO OFDM link has a hard ceiling on spectral efficiency even with unlimited power or bandwidth. The model uses transmitter and receiver imbalance matrices $G_1$, $G_2$, $K_1$, $K_2$ to split the effective channel into the desired channel $H_d$ and the IQI interference channel $H_i$, giving the per-subcarrier SINR of equation (23). The minimum bit energy (24) and wideband slope (25) are then obtained by inserting that SINR into the interference-channel wideband formulas of [15]; equation (25) has the extra term $\\zeta_m$ in its denominator, which is exactly the IQI contribution. Simulations at 300 GHz and 10 GHz bandwidth show the rate-versus-power curve saturating when IQI is present, and show that IQI, rather than inter-user interference, is the dominant limiter at THz frequencies.","pith_inferences":["The saturation mechanism is a power-proportional self-interference floor: because the IQI interference covariance depends on the signal at the mirror subcarrier, the asymptotic rate limit should be set by the image rejection ratio rather than by noise. A quick check would be to sweep the image rejection ratio and see whether the saturated rate shifts by the same number of decibels.","The same SINR-to-wideband-slope substitution could be applied to other power-proportional hardware impairments such as phase noise or nonlinear power-amplifier distortion; if the substitution holds for any additive self-interference term, the strict-limit conclusion would generalize beyond IQI, though the paper only demonstrates it for IQI.","Subcarrier nulling is a bandwidth-for-linearity trade: it removes the mirror-image interference at the cost of half the spectrum. The results imply that full-bandwidth THz designs should instead use IQI calibration or widely linear receivers to keep both the bandwidth and the rate."],"forward_implications":["With IQI present, increasing transmit power stops increasing the achievable rate because the image-band self-interference grows with the signal power.","The wideband slope falls as amplitude imbalance worsens, so each extra unit of power or bandwidth buys less spectral efficiency in an IQI-limited THz link.","At THz frequencies IQI is the dominant impairment compared with inter-user interference, while at conventional frequencies the ordering is reversed.","Nulling the mirror subcarrier halves the active bandwidth but can give a higher rate than using the full bandwidth when IQI is present.","Stronger IQI raises the bit energy needed to reach a desired spectral efficiency, making power-limited THz operation even harder."],"supporting_citations":[{"why":"Provides the interference-channel wideband-slope and minimum-bit-energy formulas that the paper adapts to the IQI case.","marker":"[15]"},{"why":"Provides the low-SNR spectral-efficiency approximation that defines the wideband slope and minimum bit energy.","marker":"[14]"},{"why":"Supplies the transmitter and receiver IQI matrices and the mirror-subcarrier interference model used in the system equations.","marker":"[11]"},{"why":"Supplies the THz channel and antenna steering-vector model used to build the array-of-subarrays link.","marker":"[13]"},{"why":"Provides the measured 300 GHz I/Q attenuation level used to set IQI severity and the image rejection ratio in the simulations.","marker":"[3]"},{"why":"Introduces the array-of-subarrays architecture on which the antenna structure and beamforming are based.","marker":"[12]"},{"why":"Reports a 219-266 GHz IQ receiver, supporting the premise that IQI is severe at THz frequencies.","marker":"[4]"}],"fun_headline_variants":["IQ mismatch bottlenecks THz links","Terahertz rates hit ceiling from IQ imbalance","I/Q imbalance sets hard cap on THz throughput","Mirror-image interference caps THz data rates"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing step is a substitution: the paper takes the signal-to-interference-plus-noise expression that includes IQI and plugs it into previously derived wideband-slope formulas for interference channels, rather than deriving the slope from the definition; if that substitution is not valid, the claimed rate limit does not follow.","fun_headline_variants_meta":{"raw":{"variants":["IQ mismatch bottlenecks THz links","Terahertz rates hit ceiling from IQ imbalance","I/Q imbalance sets hard cap on THz throughput","Mirror-image interference caps THz data rates"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000531,"raw_usage":{"total_tokens":2596,"prompt_tokens":1021,"completion_tokens":1575,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":637,"completion_tokens_details":{"reasoning_tokens":1518}},"tokens_in":637,"tokens_out":1575,"duration_ms":13443,"temperature":1.0,"reasoning_tokens":1518,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-08T21:21:20.730463+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Simulate the 300 GHz IQI system model of the paper and compute the spectral efficiency at low SNR directly, then evaluate the wideband slope as the second derivative of spectral efficiency with respect to $E_b/N_0$ at the minimum bit energy and compare it with (25); a mismatch would mean the substitution that produces (25) is invalid and the strict limit is not established.","supporting_citations":[{"cited_title":"The wideband slope of interference channels: The small bandwidth case,","cited_arxiv_id":null,"evidence_quote":"Provides the interference-channel wideband-slope and minimum-bit-energy formulas that the paper adapts to the IQI case."},{"cited_title":"Spectral efficiency in the wideband regime,","cited_arxiv_id":null,"evidence_quote":"Provides the low-SNR spectral-efficiency approximation that defines the wideband slope and minimum bit energy."},{"cited_title":"Estimation and compensation of TX and RX IQ imbalance in OFDM-based MIMO systems,","cited_arxiv_id":null,"evidence_quote":"Supplies the transmitter and receiver IQI matrices and the mirror-subcarrier interference model used in the system equations."},{"cited_title":"An overview of signal processing techniques for terahertz communications,","cited_arxiv_id":null,"evidence_quote":"Supplies the THz channel and antenna steering-vector model used to build the array-of-subarrays link."},{"cited_title":"50- gb/s direct conversion QPSK modulator and demodulator MMICs for terahertz communications at 300 GHz,","cited_arxiv_id":null,"evidence_quote":"Provides the measured 300 GHz I/Q attenuation level used to set IQI severity and the image rejection ratio in the simulations."},{"cited_title":"Terahertz communications: An array-of- subarrays solution,","cited_arxiv_id":null,"evidence_quote":"Introduces the array-of-subarrays architecture on which the antenna structure and beamforming are based."},{"cited_title":"A 219–266 GHz fully-integrated direct-conversion IQ receiver module in a SiGe HBT technology,","cited_arxiv_id":null,"evidence_quote":"Reports a 219-266 GHz IQ receiver, supporting the premise that IQI is severe at THz frequencies."}],"review_version":1}