Source-linked AI summary
Terahertz-Band Ultra-Massive Spatial Modulation MIMO
Hadi Sarieddeen, Mohamed-Slim Alouini, Tareq Y. Al-Naffouri
TL;DR
Severe propagation losses and power limits constrain THz communication range, while compact arrays face a trade-off between beamforming and spatial multiplexing. This paper investigates spatial modulation for THz UM-MIMO, using channel-aware antenna configurations and analytical and numerical evaluation. Proper spacing optimization improves the reported performance of the proposed modes over untuned Region 2 operation.
Problem
Severe THz-band propagation losses and power limitations make communication distances very short, while compact AoSA designs must trade beamforming against multiplexing and spatial diversity.
Source: §I · p. 2Method
The paper analyzes THz channel conditions and proposes two spatial-modulation modes for graphene-based AoSA arrays, with antenna spacing tuned to support favorable channel conditions.
Source: §III · p. 4, §VIII · p. 11Results
With proper spacing tuning, the Region 2 optimized mode performs best, and both proposed modes significantly outperform operation in Region 2 without optimized antenna spacing.
Source: §VI · p. 8Takeaways & Limitations
The authors present THz-band spatial modulation as a research direction and propose extensions including frequency-interleaved antenna maps and generalized hierarchical detection and coding.
Source: §I · p. 2
Abstract
from arXiv · showhide
The prospect of ultra-massive multiple-input multiple-output (UM-MIMO) technology to combat the distance problem at the Terahertz (THz)-band is considered. It is well-known that the very large available bandwidths at THz frequencies come at the cost of severe propagation losses and power limitations, which result in very short communication distances. Recently, graphene-based plasmonic nano-antenna arrays that can accommodate hundreds of antenna elements in a few millimeters have been proposed. While such arrays enable efficient beamforming that can increase the communication range, they fail to provide sufficient spatial degrees of freedom for spatial multiplexing. In this paper, we examine spatial modulation (SM) techniques that can leverage the properties of densely packed configurable arrays of subarrays of nano-antennas, to increase capacity and spectral efficiency, while maintaining acceptable beamforming performance. Depending on the communication distance and the frequency of operation, a specific SM configuration that ensures good channel conditions is recommended. We analyze the performance of the proposed schemes theoretically and numerically in terms of symbol and bit error rates, where significant gains are observed compared to conventional SM. We demonstrate that SM at very high frequencies is a feasible paradigm, and we motivate several extensions that can make THz-band SM a future research trend.
I. INTRODUCTION
THz communication is constrained by severe propagation losses and power limits, while compact arrays must balance beamforming, spatial multiplexing, and antenna spacing. The paper investigates spatial modulation for THz UM-MIMO and proposes channel-aware configurations and analytical and numerical evaluation.
I. INTRODUCTION
- Severe THz propagation losses and power limitations yield very short communication distances, motivating dense nano-antenna arrays for beamforming gain.The AoSA architecture also offers a trade-off between beamforming and multiplexing.Source: §I · p. 2
- Larger antenna separations can improve spatial diversity and sampling, but compact designs are attractive and frequency-interleaved antenna maps are not spectrum-efficient.The paper frames this as a trade-off between system performance and antenna compactness.Source: §I · p. 2
- The paper introduces spatial modulation as a spectrum- and power-efficient paradigm for THz UM-MIMO, where LoS dominance makes SM challenging.The authors present it as a future research direction.Source: §I · p. 2
- The authors analyze channel conditions across frequency, communication range, and antenna separation to identify settings with sufficient channel diversity.They propose adaptive hierarchical SM that maps bits to subarray or antenna-element locations, and configurable graphene sheets whose dimensions adapt to target bit rate and range.Source: §I · p. 2
- The proposed schemes receive closed-form SER analysis and extensive simulations, with analytical and numerical results reported to closely match.The paper also motivates frequency-interleaved antenna maps, generic index modulation, pulse-based THz SM, and generalized hierarchical detection and coding.Source: §I · p. 2
II. SYSTEM MODEL
The system model represents graphene-based plasmonic nano-antennas as transmit and receive arrays of subarrays. It specifies the baseband signal, channel, precoding and combining, and assumptions used to describe the link.
II. SYSTEM MODEL
- The transmit and receive AoSAs contain M_t×N_t and M_r×N_r subarrays, respectively, each comprising a Q×Q set of nano-antennas.For simplicity, the model assumes M_t=M_r=N_t=N_r=M.Source: §II · pp. 2–3
- The model assumes single-carrier LoS transmission with frequency-flat fading and perfect beamforming-angle alignment.It neglects mutual coupling when adjacent nano-antenna spacing is at least the SPP wavelength; individually addressing nano-antennas is possible for Q=1 but makes the architecture much more complex.Source: §II · pp. 3–4
- The baseband model relates the received vector to the channel, transmitted symbols, precoder, combiner, and additive white Gaussian noise.The precoder and combiner define subarray utilization; the model describes a large MIMO system by vectorizing antenna indices.Source: §II · p. 3
- Each subarray generates one beam, and the channel matrix entries describe frequency responses between transmit and receive subarrays.The channel expression uses path gain, steering vectors, antenna gains, and transmit and receive angles.Source: §II · p. 3
III. CHANNEL CONDITION AT THE THZ-BAND
The section models THz propagation as a dominant line-of-sight path and examines how frequency, range, and antenna spacing shape channel conditions. It identifies spacing choices that produce orthogonal channel columns and well-conditioned channels, while noting that absorption affects available spectrum and noise.
III. CHANNEL CONDITION AT THE THZ-BAND
- THz propagation is modeled with a dominant LoS path, with scattered and refracted rays neglected.The channel gain includes frequency-dependent absorption and distance-dependent path loss.Source: §III · p. 4
- At 1 and 3 THz, channel condition varies with spacing ∆ and range D; Region 1 stays well-conditioned, while Region 2 can have highly correlated paths.The low-condition-number curves mark equal-singular-value configurations.Source: Figure 2 · p. 4, Figure 3 · p. 4, §III · p. 5
- Orthogonal channel columns maximize capacity; at their locations, equal singular values form eigenchannels for spatial multiplexing or spatial modulation.For uniformly spaced arrays, optimal spacing depends on wavelength and communication range.Source: §III · pp. 4–5
- Molecular absorption creates path-loss spikes and, beyond 1 m at frequencies above 1 THz, divides available spectrum into smaller sub-bands.Unless stated otherwise, subsequent analysis assumes absorption-free windows and Gaussian noise.Source: §III · p. 5, Figure 4 · p. 5
IV. PROPOSED SM SCHEMES
The proposed design uses two operating modes selected according to antenna spacing and communication range. One adapts antenna activation, while the other dynamically adjusts graphene-sheet dimensions to satisfy the channel-spacing condition.
IV. PROPOSED SM SCHEMES
- The scheme selects between two operating modes according to inter-antenna spacing and communication range.The modes correspond to Region 1 and Region 2 channel conditions.Source: §IV · p. 5
- In Region 1, adaptive two-level spatial modulation assigns activation at the subarray or antenna-element level.Physical antenna-element separations and dimensions remain fixed for the target chip frequency.Source: §IV · p. 5
- In Region 2, a configurable graphene sheet dynamically adjusts inter-antenna separations to satisfy condition (4).This mode applies when D ≫ ∆; the design avoids the yellow peak curves in Fig. 2.Source: §IV · p. 5
A. Adaptive Hierarchical SM
Adaptive hierarchical spatial modulation assigns information bits to subarray and antenna-element locations as well as a QAM symbol. The activation level depends on channel conditions, and the receiver selects an antenna index and slices the detected symbol.
A. Adaptive Hierarchical SM
- For short-range links with D ≤ 1 m, antenna-element-level spatial modulation is used when δ is sufficiently large for well-conditioned channels.When D < 20δ at 1 THz, the hierarchical design activates at subarray and antenna-element levels; otherwise it retains modulation at the subarray level.Source: A. Adaptive Hierarchical SM · p. 5
- The adaptive mapping assigns bits to a selected subarray, a selected antenna element, and a QAM symbol.The number of subarrays, antenna elements, and constellation size can be traded for a target bit rate.Source: A. Adaptive Hierarchical SM · p. 5, Figure 5 · p. 6
- The receiver chooses the transformed received-vector element with maximum magnitude, then slices the detected symbol to the nearest constellation point.The demodulator then performs bit demapping.Source: A. Adaptive Hierarchical SM · p. 6
B. Configurable Graphene Sheet
The configurable graphene sheet adapts active antenna and subarray settings to operating conditions, while the performance analysis derives an SER approximation for Region 2 using antenna-index and symbol errors.
B. Configurable Graphene Sheet
- A reprogrammable graphene sheet uses densely spaced antenna elements, with settings adjusted to system parameters; SM adds spectral efficiency logarithmically in antenna count, unlike SMX’s multiplicative increase.The discussion assumes matched transmitter and receiver sheets and limits its parameter calculations to the transmit array.Source: B. Configurable Graphene Sheet · p. 6
- The Region 2 configuration selects the number of antenna elements per subarray to meet the required path-loss compensation at a given frequency and communication distance.The required array gain determines the number of beamforming elements Q.Source: B. Configurable Graphene Sheet · p. 6
- Region 2 retains δ = λ/2 within each subarray for beamforming and adapts the inter-subarray spacing ∆ to the communication distance.The target spacing is approximated by activating subarrays at intervals of ¯∆.Source: B. Configurable Graphene Sheet · p. 6
- Individually controlling contiguous or non-contiguous antenna elements depends on excitation: optical pumping is impractical, while electrical pumping faces low source power and short SPP propagation.The paper notes that attaching small plasmonic sources directly to each antenna element has been argued to yield high radiated power.Source: B. Configurable Graphene Sheet · p. 6
V. PERFORMANCE ANALYSIS
- The Region 2 SER analysis separates antenna-index errors from symbol errors and approximates the overall error using their probabilities under MRRC detection.The bound is based on pairwise error probabilities; the analysis uses a deterministic line-of-sight channel and does not average over fading.Source: §V · p. 7
- Near-orthogonality in Region 2 lets the analysis model most channel entries as zero-mean, then use folded-normal distributions and order statistics to estimate antenna-selection error.The selected channel entry has a distinct mean; the resulting antenna-error probability is substituted into the SER expression.Source: §V · p. 7
VI. NUMERICAL RESULTS
The numerical results show that tuning antenna spacing improves BER, and that SM is less sensitive than SMX to unfavorable channel conditions. Large subarrays shift performance to a lower SNR range through array gain.
VI. NUMERICAL RESULTS
- At 1 m, increasing operation from 1 THz to 3 THz results in a 10 dB SNR loss, and analytical and simulated SER results closely agree.The SER comparison uses M = 4 and 16-QAM without beamforming.Source: §VI · p. 8
- Optimized Region 2 achieves the best BER, while Region 2 without spacing optimization becomes severely distorted and diverges at high SNR.The reported relative performance holds for the tested 1 m and 5 m distances.Source: Figure 8 · p. 8, §VI · p. 8
- SMX slightly outperforms SM under optimal channel conditions, but its performance deteriorates severely otherwise.The comparison is for 64 × 64 MIMO with 1024-element subarrays and Q = 32.Source: Figure 9 · p. 8, §VI · pp. 8–9
- A 30 dB array gain from 1024-element subarrays shifts the same relative performance to a much lower SNR range.The plotted setting uses a 64 × 64 MIMO system.Source: §VI · p. 8
VII. EXTENSIONS
The paper presents possible future research directions aimed at overcoming limitations of the proposed solutions and efficiently using spatial degrees of freedom at very high frequencies.
VII. EXTENSIONS
- The paper frames future research as a way to address limitations of its proposed solutions.Source: §VII · p. 9
- The proposed directions are intended to support efficient use of spatial degrees of freedom at very high frequencies.Source: §VII · p. 9
- The section introduces possible extensions rather than reporting evaluated results for them.Source: §VII · p. 9
A. Frequency-Interleaved Antenna Maps
Frequency-interleaved antenna maps combine compact packaging with larger same-frequency separations for beamforming or multiplexing. Their implementation relies on tuning antenna-element resonance, which is difficult at frequencies below 1 THz.
A. Frequency-Interleaved Antenna Maps
- Frequency interleaving places antennas tuned to different frequencies between non-contiguous same-frequency antennas, retaining δ = λ/2 separation for beamforming.The approach addresses limited beamforming gain from densely packed elements and the larger separations desired for multiplexing.Source: A. Frequency-Interleaved Antenna Maps · p. 9
- Interleaving can be applied at the antenna-element or subarray level, or at both levels with higher complexity.At the subarray level, same-frequency subarrays are activated at a sufficiently large optimized spacing ∆.Source: A. Frequency-Interleaved Antenna Maps · p. 9, Figure 11 · p. 9
- The maps require dynamic resonance tuning through graphene conductivity, but this tunability is difficult to realize below 1 THz.The passage identifies software-defined plasmonic metamaterials as an alternative solution.Source: A. Frequency-Interleaved Antenna Maps · p. 9
B. Relaxing Ideal Assumptions
The section considers how relaxing ideal channel and modulation assumptions changes the proposed THz spatial-modulation system. It identifies practical channel effects and pulse-based alternatives, while noting that carrier-based modulation is assumed in the analysis.
B. Relaxing Ideal Assumptions
- The model assumes a LoS component but can include reflected rays; multipath lowers the path-loss threshold, and the channel model can use Rician fading.The additional reflected-ray path gain depends on frequency and the distances from reflector to transmitter and receiver.Source: B. Relaxing Ideal Assumptions · pp. 9–10
- Other stated assumptions include unmodeled channel correlation, RF impairments, imperfect channel estimation, and blockage; obtaining channel state information at very high frequencies is challenging.The text points to channel tracking and SM channel-estimation work, as possible ingredients for future extensions.Source: B. Relaxing Ideal Assumptions · p. 10
- Carrier-based modulation is assumed, although graphene at room temperature is described as supporting only short pulses, making pulse-based modulation relevant to nanoscale applications.The text discusses TS-OOK and adapting the proposed solutions to pulse-based modulation using optical SM results.Source: B. Relaxing Ideal Assumptions · p. 10
C. Generalized Spatial and Index Modulation
The section outlines extensions that encode information through combinations of active antennas or through frequency choices. It also identifies detection complexity and frequency-transition speed as constraints.
C. Generalized Spatial and Index Modulation
- Generalized spatial modulation (GSM) would select combinations of active antennas, but large-scale systems still need low-complexity detection algorithms.The number of simultaneously active antenna elements is denoted by S.Source: C. Generalized Spatial and Index… · p. 10
- Generalized index modulation (GIM) assigns information bits to interleaved antenna maps and frequency choices, leveraging THz spectral windows created by the absence of molecular absorption.The text also describes assigning different frequencies to the same antenna element across channel uses.Source: C. Generalized Spatial and Index… · p. 10
- Joint frequency and spatial maps could increase GIM's bit count and use of THz spectrum, but frequency-transition speed and detection across many combinations constrain applicability.The text suggests artificial intelligence and machine learning as possible approaches to receiver-side index detection.Source: C. Generalized Spatial and Index… · p. 10
D. Enhanced Detection and Coding Schemes
The paper identifies hierarchical detection and coding, optimization, and array-design questions for THz spatial modulation. Its conclusion reports that optimizing active antenna-element positions significantly enhances performance in the studied system.
D. Enhanced Detection and Coding Schemes
- Hierarchical detection can estimate the subarray index with MRRC, then jointly detect the antenna-element index and symbol with ML; stronger coding can protect antenna-element-index bits.When Δ > δ, the text says errors are more likely at the antenna-element level.Source: D. Enhanced Detection and Coding Schemes · p. 10
- Generalized spatial modulation calls for near-optimal detection at reasonable complexity, with candidate approaches including lattice reduction, message passing, sampling, and search methods –[87].The text also mentions sub-space decompositions.Source: D. Enhanced Detection and Coding Schemes · p. 11
E. Optimization Problems
- Antenna and resource-allocation optimization can be divided into selected subproblems, with existing frequency-allocation, pulse-waveform, and beamforming formulations adaptable to spatial-modulation constraints,,,.The cited examples optimize capacity or adapt waveforms and beamforming to communication distance.Source: E. Optimization Problems · p. 11
- Finite antenna-array dimensions limit near-orthogonal channels and high multiplexing gains beyond the Rayleigh distance; spatial oversampling below λ_spp is discussed as a possible design avenue.The text notes that oversampling is argued to reduce plane waves' spatio-temporal frequency-domain support, with potential noise-figure and linearity benefits.Source: E. Optimization Problems · p. 11
VIII. CONCLUSION
- Optimizing active antenna-element positions significantly enhances performance in the proposed THz spatial-modulation schemes, following analytical SER derivations and channel-spacing analysis.The conclusion also proposes future research directions for realizing THz-band spatial modulation.Source: §VIII · p. 11