A unified resource-pool architecture for high-dimensional direct-detection optical communication

Technical Summary: A Unified Resource-Pool Architecture for High-Dimensional Direct-Detection Optical Communication

Problem Statement
The continued growth of data traffic demands higher capacity in optical communication systems, yet scaling capacity often incurs prohibitive increases in receiver complexity, footprint, and energy consumption. Conventional direct-detection (IMDD) architectures are fundamentally "dimension-partitioned," treating wavelength, polarization, and intensity as isolated channels that require separate demultiplexing, polarization handling, and dedicated receiver branches. This approach ties capacity expansion directly to hardware escalation, creating a mismatch between the richness of light's physical degrees of freedom and the limited efficiency of current recovery architectures. While high-dimensional photodetection has advanced in sensing and imaging, its translation into fiber-transmission communication—where symbols are defined by composite states and evaluated by bit error rate (BER)—remains largely unexplored.

Methodology
The authors propose a unified resource-pool architecture that reorganizes wavelength, polarization, and intensity into a single, composite optical symbol space. Instead of separating these dimensions at the receiver, the system employs optical-domain joint projection to recover the composite state directly.

• Core Mechanism: The receiver utilizes an integrated disordered photonic processor fabricated on a silicon-on-insulator platform. This device acts as a fixed optical projection engine that deliberately mixes the incoming wavelength, polarization, and intensity components through disorder-induced multipath interference.
• Signal Transformation: The processor transforms each composite optical state into a reproducible, multi-output electrical "fingerprint" (a feature vector) via direct photodetection. This converts multidimensional detection from explicit channel separation into passive feature generation.
• Recovery: A lightweight decision layer (classifier) recovers the transmitted symbol from the projected feature vectors without decomposing the signal into its constituent dimensions.
• Experimental Setup: The system was validated using a dual-wavelength configuration (1550 nm and 1551 nm). Information was encoded across 4 intensity levels (PAM-4) and 16 polarization states per wavelength, creating a 64-state sub-symbol. The joint two-wavelength state formed a 4096-state composite alphabet (12 bits per symbol). The receiver employed a monolithic chip with a disordered scattering region coupled to a 16-channel photodetector array, though only 4 outputs were used for the primary 4096-state demonstration.

Key Contributions

1. Architectural Paradigm Shift: The paper introduces a framework where existing optical variables are jointly organized into a composite alphabet, decoupling symbol cardinality growth from linear hardware scaling.
2. Disorder-Enabled Joint Projection: It demonstrates that integrated disorder, typically viewed as a source of noise, can be engineered as a resource for high-dimensional state projection, enabling hardware-efficient detection of complex optical states.
3. Hardware Efficiency: The architecture achieves significant reductions in Physical Resource Scale (PRS) and Hardware Complexity (HC) compared to conventional dimension-partitioned baselines. For a 12-bit capacity (4096 states), the proposed system reduces PRS by 50% and HC by 83% relative to a traditional PDM-PAM4 baseline.
4. Experimental Validation: The system successfully resolved 4096 composite symbols with high accuracy over fiber links, validating the feasibility of dense polarization alphabets and wavelength-indexed state-space expansion in a direct-detection context.

Results

• Symbol Resolution: In a dual-wavelength experiment over 10 km of standard single-mode fiber (SMF), the system achieved a joint recovery accuracy of 99.49% for the 4096-state alphabet. This corresponds to a Bit Error Rate (BER) of 4.25 × 10⁻⁴, well below the 7% hard-decision Forward Error Correction (HD-FEC) threshold.
• Dimensional Performance: Post-recovery analysis confirmed that wavelength, intensity, and polarization were each identified with accuracies exceeding 99%, proving that the joint projection preserves the distinguishability of constituent variables.
• Transmission Robustness:
  • Reach: The system maintained error-correctable performance (BER < 7% HD-FEC limit) over 30 km of SMF, with an accuracy of 97.12% (BER ≈ 2.4 × 10⁻³).
  • Power Sensitivity: In standard SMF, performance degraded at high launch powers (>12 dBm) due to nonlinearities and back-scattering. However, additional experiments demonstrated successful operation over hollow-core fiber in high-launch-power regimes, mitigating these nonlinear impairments.
• Component Performance: The integrated photodetectors exhibited a 3-dB optoelectronic bandwidth of 22.6 GHz, and the processor demonstrated a polarization sensing precision of 1.2 degrees and a spectral decorrelation bandwidth of 0.4 nm.

Significance
The paper establishes that disorder-enabled joint projection in an integrated photonic processor offers a viable route to hardware-efficient, high-dimensional direct-detection communication. By moving beyond the conventional "dimension-partitioned" receiver architecture, this approach allows for the exploitation of multidimensional optical resources (wavelength, polarization, intensity) without the proportional increase in demultiplexing hardware and electronic processing typically required. The results suggest that reorganizing existing optical variables into a composite alphabet, rather than introducing new physical degrees of freedom, can significantly expand usable symbol alphabets while maintaining a compact receiver footprint. This work bridges the gap between high-dimensional light-field detection and practical fiber-optic communication, offering a path toward higher capacity interconnects under stringent constraints on complexity and energy.