Dual-Polarization SIMs (DPSIM)

• Dual-Polarization SIMs are techniques that use horizontal and vertical polarization channels to enable multidimensional (4-D) modulation with improved spectral and energy efficiency.
• Advanced constellation designs, such as D4-lattice and constant-modulus schemes, yield notable SNR gains and reduced BER compared to conventional dual-polarization methods.
• DPSIM enhances synchronization, link adaptation, and interference suppression, facilitating efficient implementations in satellite, holographic MIMO, and dual-polarized antenna systems.

Dual-Polarization SIMs (DPSIM) employ two orthogonal polarization states to realize advanced multidimensional (typically four-dimensional) signaling, electromagnetic wave manipulation, or antenna radiation. Architectures based on DPSIM have been developed for satellite modulation, spatial index modulation with dual-polarized RF chains, stacked intelligent metasurfaces for MIMO and holographic communications, and double-asymmetric leaky-wave antennas. Leveraging the polarization degree of freedom, DPSIM enables higher-dimensional constellations, increased spatial multiplexing, improved link adaptation, and efficient, low-complexity transceiver designs.

1. Fundamental Principles and Signal Models

DPSIM refers to schemes that simultaneously utilize two independent polarization channels—typically the horizontal (H) and vertical (V) linear polarizations—to transmit or manipulate information-carrying signals. The DPSIM signal, at any instant, is represented as a pair of complex amplitudes, one per polarization: $c_i = [c_{x,i}, c_{y,i}]$ where $c_{x,i}$ and $c_{y,i}$ are the complex envelopes for the two orthogonal polarizations, and the total symbol energy is $E_s = \mathbb{E}[|c_{x,i}|^2 + |c_{y,i}|^2]$ (Arend et al., 2015).

In spatial index modulation frameworks, bits are mapped both to a constellation symbol and a polarization index. If $s \in \mathcal{S}$ is an $M$-ary QAM/PSK symbol and $b$ the index bit, then the DPSIM transmit vector is

$\mathbf x = s \cdot \mathbf e_b$

where $\mathbf e_1 = [1,0]^T$ selects the V polarization, and $\mathbf e_2 = [0,1]^T$ selects H (Tato et al., 2018).

Intelligent metasurface-based DPSIM architectures extend this formulation by incorporating dual-polarized electromagnetic meta-units (DPEMs) on each layer, supporting independent phase (and sometimes amplitude) control for each polarization. The complete end-to-end channel stack—including free-space coupling, metasurface phase shifts, and multipath dual-polarized channels—appears as a series of block-diagonal or nearly block-diagonal transformations (Zhang et al., 15 Sep 2025, Zhang et al., 27 May 2025).

2. Dual-Polarization Constellation Designs and Efficiency

DPSIM architectures enable multidimensional modulation exploiting both polarizations. Four major 4-D constellation families have been formulated (Arend et al., 2015):

• D4-lattice amplitude modulation (LAM): Based on the densest 4-D sphere packing, LAM constellations carved from the D4 lattice with a spherical boundary achieve up to 1.5 dB SNR gain over conventional dual-polarization QAM or APSK.
• Sphere-based 4-D PSK: Constant-energy points on the complex 3-sphere maximize minimum distance at constant total power.
• Cylinder-based (hexagonal) 4-D PSK: Constant-modulus per polarization with optimized 2D hexagonal phase-plane packing.
• 4-D Bi-Orthogonal signaling: Eight orthogonal ±1 symbols per axis (constant modulus per pol), giving a 1.6 dB power-efficiency gain over dual QPSK.

Typical SNR or energy efficiency improvements are summarized below:

Constellation	Efficiency Gain vs. Dual-Pol Reference
88-LAM vs. 8-PSK	2.2 dB vs. 8-PSK; 0.7 dB vs. 8-QAM
256-LAM vs. 16-QAM	1.3–1.5 dB
64-4D-PSK vs. 8-PSK	0.7 dB
Bi-orthogonal vs. QPSK	1.6 dB

Power efficiency gains are most pronounced at high spectral efficiencies. However, the back-off required for high peak-to-average power ratio (PAPR) constellations (e.g., LAM, sphere-based) in nonlinear channels can partially offset these benefits. Constant-modulus schemes (bi-orthogonal, cylindrical PSK) retain their advantage without additional back-off (Arend et al., 2015).

3. Synchronization, Capacity, and Link Adaptation in DPSIM

Dual-polarization processing leads to theoretical and practical improvements in synchronization and estimation accuracy. Since both polarization signals are subjected to the same oscillator and channel phase noise processes, combining both in maximum-likelihood or Cramér–Rao lower bound (CRLB) estimators doubles the effective observation energy, halving timing and carrier recovery jitter variance relative to single-polarization implementations. Simulations with independent error detectors per pol validate the √2 reduction in jitter variance across SNRs (Arend et al., 2015).

Capacity analyses show that index-modulated DPSIM increases the available channel capacity by exploiting the polarization state as an additional index. The ergodic capacity formula for two equally likely polarization states is

$$C_\mathrm{DPSIM} = \frac{1}{2}\mathbb{E}[\log_2(1+\gamma_1)] + \frac{1}{2}\mathbb{E}[\log_2(1+\gamma_2)]$$

where $\gamma_i$ is the instantaneous SNR per polarization (Tato et al., 2018).

Practical link adaptation algorithms utilize instantaneous channel estimates on both polarization branches and select the optimal modulation and coding scheme (MCS) per frame using an effective SNR mapping. The adaptation rule is: $$\gamma_\mathrm{eff} = -\beta\ln\left[\frac{1}{2}(e^{-\gamma_1/\beta} + e^{-\gamma_2/\beta})\right]$$ where $\beta$ is calibrated per MCS (Tato et al., 2018). This approach provides spectral efficiency gains of up to 1 b/s/Hz over both single-polarization QAM and conventional V-BLAST MIMO under mid-SNR conditions.

4. DPSIM in Stacked Metasurfaces and Holographic MIMO

DPSIM is central to emerging wave-domain signal processing, particularly in the context of stacked intelligent metasurfaces and reconfigurable holographic MIMO (HMIMO) arrays. A DPSIM array consists of multiple cascaded layers of DPEM cells per metasurface, enabling independent manipulation of two polarizations at subwavelength granularity. Each phase-shift matrix $\Phi_p^l$ is diagonal, acting on all DPEM elements for polarization $p$ in layer $l$. The overall metasurface transfer is a product of layer-wise propagation matrices and block-diagonal phase shifts (Zhang et al., 15 Sep 2025, Zhang et al., 27 May 2025).

End-to-end DPSIM architectures jointly execute modulation, precoding, combining, and demodulation in the electromagnetic domain. The electromagnetic neural network (EMNN) framework abstracts the DPSIM stack as a hybrid DNN where metasurface phase weights $\theta_{p,m}^l$ are learned, while channel and propagation blocks remain fixed. Training involves statistical and, subsequently, instantaneous CSI-aware fine-tuning for robust end-to-end performance (Zhang et al., 15 Sep 2025).

Key performance outcomes include:

• ~3–5 dB lower BER for DPSIM-E2E OFDM over single-polarization at fixed power and SNR.
• At BER = 10⁻³, a 16T/9R DPSIM link matches a 256T/49R massive MIMO system.
• Total trainable parameters scale as $$\mathcal O(2M^{\rm dp}L^{\rm dp}+2N^{\rm dp}K^{\rm dp}J+D^2HJ)$$; DPSIM doubles parameter count over single-polarization configurations (Zhang et al., 15 Sep 2025).

5. Optimization, Interference Suppression, and Algorithmic Advances

DPSIM architectures in HMIMO leverage doubled processing dimensionality in the wave domain for enhanced spatial multiplexing and interference suppression. The received signals are modeled as

$$y_{s,p} = H_{pp}(s,s)x_{s,p} + \sum_{t\neq s} H_{pp}(s,t)x_{t,p} + \sum_t H_{p\bar p}(s,t)x_{t,\bar p} + n_{s,p}$$

where the first sum denotes inter-stream interference (ISI) within a polarization and the second sum is polarization cross-interference (PCI).

Joint optimization of the metasurface phase shifts and per-stream power allocation is formulated to maximize total spectral efficiency: $$\max_{\theta,\,\xi,\,p_{s,p}} \sum_{p,s} \log_2(1+\mathrm{SINR}_{s,p}) \;\text{s.t.}\; \sum_{p,s} p_{s,p} \leq P_t$$ The Layer-by-layer Gradient Descent with Water-Filling (LGD–WF) algorithm alternates between phase update steps—driven by SVD matching error—and standard water-filling power allocation (Zhang et al., 27 May 2025).

Simulations show DPSIM HMIMO supports more ISI-free streams and achieves spectral and energy efficiencies closer to the theoretical upper bound than single-polarization SIMs, with robustness to polarimetric imperfections (e.g., cross-polar ratio $\epsilon = 0.4$). Algorithmic complexity is compatible with real-time FPGA/ASIC deployment.

6. Antenna Implementations: Double-Asymmetric Leaky-Wave Antennas

DPSIM concepts are implemented at the antenna level via double-asymmetric periodic leaky-wave antennas (DA P-LWA). Double asymmetry is achieved by breaking both longitudinal symmetry (to equalize series and shunt radiation) and transverse symmetry (to decouple excitation ports and pol modes). Each port excites a distinct orthogonal polarization channel at broadside, with cross-polarization discrimination controlled by transformer ratio $T$ in the equivalent circuit model (Al-Bassam et al., 2017).

Prototype DA P-LWA (e.g., SF-LCP structures) deliver dual orthogonal beams (XPD = 14–19 dB), ~9 dBi gain, and excellent impedance match at 24 GHz. Key extensions include wideband frequency scanning, dynamic LA/TA reconfiguration, and integration with massive MIMO for simultaneous dual-pol spatial multiplexing.

7. Implementation Issues and Practical Considerations

DPSIM design is subject to several practical engineering concerns:

• Nonlinear channel constraints: Each pol is usually amplified separately, enforcing individual peak/average power constraints. Large PAPR constellations incur SNR losses due to required amplifier back-off. Only constant-modulus schemes (e.g., bi-orthogonal, cylindrical PSK) are immune (Arend et al., 2015).
• Calibration and fabrication tolerances: Metasurface phase errors ($\Delta\theta$), physical misalignments, and polarization leakage degrade interference suppression—mitigated by calibration loops and robust phase quantization (Zhang et al., 27 May 2025).
• Complexity and scalability: EMNN/DPSIM architectures scale moderately in trainable parameters (factor of two over single-pol), but hardware and memory requirements remain tractable for moderate array sizes (Zhang et al., 15 Sep 2025).
• Integration with 5G/6G: 5G NR supports dual-polarization antennas and port mapping; DPSIM enhances spectral efficiency through index modulation without adding RF hardware (Tato et al., 2018).

In summary, Dual-Polarization SIMs enable the full exploitation of polarization as an independent signaling and processing dimension, leading to high-power efficiency, spatial multiplexing, capacity gains, and ultra-compact, low-complexity implementations across RF, baseband, and EM-domain architectures. These advantages are validated in both hardware and simulated deployments over satellite, MIMO, and holographic communications platforms (Arend et al., 2015, Tato et al., 2018, Zhang et al., 15 Sep 2025, Zhang et al., 27 May 2025, Al-Bassam et al., 2017).