Intensity modulation with direct detection, or IM/DD, has long been the workhorse of cost-sensitive optical links. A laser is switched or modulated in brightness, and a simple photodiode at the far end converts the light back into an electrical signal. It is compact, energy-efficient and inexpensive, which is precisely why it dominates short-reach systems. The trouble begins as symbol rates climb. At 200 gigabaud, the spectral width of the signal becomes so large that chromatic dispersion, the tendency of optical fibre to smear different wavelengths of light at slightly different speeds, wreaks havoc. The signal develops frequency-dependent power fading, deep notches in its spectrum that can obliterate information before it ever reaches the receiver.

Conventional fixes are unattractive at this scale. Digital dispersion compensation demands enormous signal-processing power and bandwidth-hungry analogue-to-digital converters. Coherent detection, which measures both the amplitude and phase of light, solves the problem elegantly but at a price in complexity, power consumption and cost that is hard to justify for the short links that connect data centres or serve homes through passive optical networks. The research consortium, spanning the University of Southampton, the University of West Attica, the Technical University of Denmark, CEA-Leti and Ghent University working with imec, took a different route: fix the problem in the optical domain before the signal is ever detected.

The team’s solution is an architecture they call recurrent optical spectrum slicing, or ROSS, implemented as a photonic integrated circuit. The concept is a form of optical preprocessing that provides frequency-diversity detection. Instead of letting a single photodiode see the whole, dispersion-distorted spectrum, the ROSS receiver splits the incoming signal into multiple spectral slices and recombines them in a recurrent configuration, generating several complementary intensity-domain observations of the same data stream. Because the power-fading notches induced by dispersion fall at different points in each sliced observation, information lost in one slice survives in another, allowing the original data to be reconstructed.

The elegance of the approach lies in what it removes from the receiver’s electronic burden. By performing the heavy lifting optically, the ROSS receiver allows the system to use low-bandwidth electro-optic components, which are cheaper, more readily available and consume less power than the ultra-wideband devices that a 200-gigabaud signal would otherwise require. Recovery of the data then relies on simple feed-forward equalization, a lightweight digital technique, rather than the computationally intensive compensation schemes that would be needed to undo dispersion after detection.

The experimental results are striking. The researchers demonstrated intensity-modulated, direct-detected transmission of 4-level pulse-amplitude modulation, or PAM-4, at 160 gigabaud over 50 kilometres of uncompensated standard single-mode fibre, and at 175 gigabaud over 25 kilometres. Pushing further, they transmitted 200-gigabaud on-off keying, the simplest modulation format of all, over 75 kilometres of the same uncompensated fibre. These figures translate into a measured dispersion tolerance of up to 2.4 times 10 to the power of 5 gigabaud-picoseconds per nanometre, a metric that quantifies how much dispersion a signal of a given symbol rate can endure.

To put those numbers in perspective, uncompensated standard single-mode fibre accumulates roughly 17 picoseconds per nanometre per kilometre of dispersion at telecommunication wavelengths. A 200-gigabaud signal over 75 kilometres therefore accumulates a dispersion-length product that would ordinarily produce catastrophic power fading in a conventional direct-detection receiver. That the ROSS architecture handles this with simple components and modest equalization suggests a practical path to scaling short-reach links without migrating to coherent technology.

The implications reach well beyond the laboratory. Passive optical networks, the fibre-to-the-home infrastructure serving hundreds of millions of subscribers, are under pressure to deliver ever higher bit rates while keeping per-user costs and power consumption low. Data centre operators face a similar squeeze, as artificial intelligence workloads drive explosive growth in the traffic flowing between servers and switches over links of a few hundred metres to a few tens of kilometres. A receiver that keeps the cost and simplicity of direct detection while tolerating dispersion at 200 gigabaud could reshape the economics of both markets, reducing the energy footprint of the optical layer at exactly the moment when its power draw has become a strategic concern.

The photonic integration aspect is equally significant. By fabricating the spectrum-slicing receiver as an integrated chip, the team moves the concept from a bench-top arrangement of discrete components toward a manufacturable device that could be mass-produced using established foundry processes. Integration shrinks size, improves stability against environmental drift, and lowers packaging cost, all prerequisites for adoption in the high-volume, cost-sensitive segments where IM/DD reigns. The work was supported by the European Union’s Horizon Europe PROMETHEUS project, the H2020 NEoteRIC project, the Villum Foundations OPTIC-AI project, and the UK’s EPSRC HASC Telecommunications Hub and DSIT REASON programmes, with experiments conducted at the University of Southampton’s X-band laboratories.

Challenges remain before such receivers appear in commercial transceivers, including refining the equalization algorithms and demonstrating performance across the temperature and manufacturing variations of real deployments. But the demonstration marks a clear milestone: a photonic chip that lets the humblest optical modulation format carry some of the highest symbol rates yet achieved over dispersive fibre, using detection hardware that a network operator could actually afford. As data demand continues its relentless climb, solutions that trade silicon-area optical ingenuity for expensive electronics are likely to become increasingly central to the internet’s physical backbone.

Subject of Research: Photonic integrated recurrent optical spectrum slicing receivers for dispersion-tolerant high-speed direct-detection optical communications

Article Title: Photonic integrated recurrent spectrum slicer enabling low complexity direct detection over 200 GBaud dispersive links

Article References: Liu, H., Sozos, K., Teofilovic, I., Wantee, S., Bottrill, K. R. H., Malhouitre, S., Garcia, S., Sarantoglou, G., Bienstman, P., Charbonnier, B., Mesaritakis, C., Vigliar, C., Da Ros, F., Bogris, A., & Petropoulos, P. (2026). Photonic integrated recurrent spectrum slicer enabling low complexity direct detection over 200 GBaud dispersive links. Communications Engineering. https://doi.org/10.1038/s44172-026-00801-6

Image Credits: AI Generated

DOI: 10.1038/s44172-026-00801-6

Keywords: photonic integration, optical communications, direct detection, chromatic dispersion, spectrum slicing, PAM-4, on-off keying, passive optical networks, data centre interconnects, feed-forward equalization, IM/DD, fibre optics

The challenge stems from the very technique that makes optical wireless so spectrally efficient: high-order quadrature amplitude modulation, or QAM. Schemes such as 256-QAM and 512-QAM pack enormous amounts of data into each symbol by using extremely dense constellations of amplitude and phase combinations. But as the constellation density grows, the decision boundaries between neighboring symbols shrink to razor-thin margins, making the signal exquisitely sensitive to noise, phase fluctuations, and the nonlinear distortions introduced by optical components such as light-emitting diodes and laser diodes. In intensity-modulated direct-detection systems, where light intensity cannot go negative, engineers must also contend with a high peak-to-average power ratio, which further degrades signal quality. Traditional receivers that rely on analytical models and linear approximations simply cannot capture these nonlinearities, and their performance collapses precisely where 6G needs it most.

The research team built their evaluation around a two-user optical NOMA system, in which both users share the same time and frequency resources through power-domain superposition coding. The user with the weaker channel is allocated 80 percent of the transmit power while the stronger user receives 20 percent, and at the receiver a successive interference cancellation procedure separates the overlapping signals. The channel model incorporated a realistic battery of impairments: distance-dependent path loss with an exponent of 2.2, Rayleigh fading, additive white Gaussian noise, optical nonlinear distortion from the light source, and hardware impairments including synchronization errors. This deliberately harsh environment was designed to emulate the conditions a real 6G optical link would face, with independent channel realizations generated for each user to reflect heterogeneous propagation conditions.

Against this backdrop, the authors evaluated six detectors within a single unified framework. The minimum mean square error (MMSE) detector offers a favorable complexity-performance trade-off but remains fundamentally linear. The QR-decomposition with M-algorithm maximum likelihood detector (QRM-MLD) approaches optimal performance through tree search but becomes computationally prohibitive at high modulation orders. Convolutional neural network (CNN) detectors learn spatial signal patterns efficiently but cannot model temporal dependencies, while recurrent neural network (RNN) detectors capture channel memory and time correlations at the cost of slower convergence. Two hybrids, CNN-MMSE and the newly proposed RNN-MMSE, combine statistical estimation with deep learning refinement. Prior studies had examined these techniques only in isolation; the unified comparison is itself a significant contribution.

The proposed RNN-MMSE detector operates in three stages. First, an RNN learns the channel characteristics directly from pilot symbols, replacing conventional pilot-based estimation and capturing temporal variations in the process. Second, the MMSE stage applies a statistically optimal linear estimate that suppresses noise and linear interference, producing a structured initial symbol estimate. Third, a second RNN refines that estimate by learning only the residual nonlinear distortions and temporal dependencies that the linear stage could not remove. The authors provide a theoretical justification rooted in the orthogonality principle of linear estimation: because the MMSE output extracts all linearly available information, the residual presented to the recurrent network has a substantially smaller variance than the raw signal, which smooths the optimization landscape, lowers gradient variance, and accelerates training convergence.

The simulation results are striking. At 64-QAM, conventional optical NOMA detection required approximately 13.5 dB of signal-to-noise ratio to achieve a bit error rate of 10 to the minus 3, while the proposed RNN-MMSE detector reached the same target at just 6.5 to 7 dB. As the modulation order climbed, the gap widened. At 128-QAM the hybrid detector delivered a 7 to 8 dB improvement, at 256-QAM an 8 to 9 dB gain, and at 512-QAM, the most demanding scheme tested, it achieved a 9 to 10 dB advantage, reaching the target error rate at roughly 12.5 to 13 dB where the conventional baseline needed 21 to 22 dB. Standalone CNN and RNN detectors and the CNN-MMSE hybrid improved on the classical methods but consistently fell short of the proposed architecture.

Spectral behavior told a similar story. Using Welch power spectral density estimation with Hamming windowing, the team measured out-of-band leakage across all detectors. Conventional optical NOMA exhibited in-band power spectral density around minus 40 dB/Hz with poor suppression near minus 60 dB/Hz, while the RNN-MMSE detector pushed spectral leakage below minus 130 dB/Hz at 512-QAM, indicating excellent sidelobe suppression and minimal adjacent-channel interference. Training convergence favored the hybrid as well: over 20 epochs the RNN-MMSE model began at roughly 85 percent accuracy and converged to about 90 percent, outperforming standalone models by 18 to 32 percentage points. Statistical validation across 20 independent Monte Carlo runs confirmed the results were reproducible, with a standard deviation of only 0.22 dB on the key SNR metric and a 95 percent confidence interval of plus or minus 0.10 dB.

Practicality was a central concern. The entire framework was implemented in MATLAB R2024a on a workstation with an Intel Core i7 processor, 32 GB of RAM, and an NVIDIA RTX 3060 GPU. Training on 100,000 generated signal samples took only 14 to 16 minutes, and inference required approximately 3.8 milliseconds per OFDM frame, fast enough for near-real-time deployment. Memory consumption stayed below 800 MB during inference. The authors acknowledge that the hybrid architecture carries more computational overhead than a standalone linear detector, with complexity of the order of N cubed for the MMSE stage plus N squared terms for the recurrent stage, but they argue the performance gains justify the cost, particularly since the heavy training can be performed offline once while deployment requires only lightweight forward inference that GPUs, FPGAs, or edge accelerators can handle.

The study is candid about its limits. All results derive from Monte Carlo simulation rather than a physical testbed, and the authors identify experimental validation on an LED- or laser-based intensity-modulation link with software-defined-radio baseband processing as an immediate priority. Their comparison with Transformer-based detectors draws on literature benchmarks obtained under different channel models rather than a controlled head-to-head experiment, and their robustness analysis of imperfect channel estimation, imperfect interference cancellation, and synchronization errors is mechanistic rather than simulated. The two-user configuration, while modular and scalable in principle, will need extension to denser multi-user networks with adaptive grouping and dynamic power allocation. Even so, the findings make a compelling case that hybrid intelligent detection is not an incremental refinement but a necessary evolution for 6G optical networks, and the roadmap toward testbed validation, attention-based comparisons, and large-scale multi-user trials is already clearly drawn.

Subject of Research: Neural network-based signal detection for 6G optical non-orthogonal multiple access systems with high-order QAM modulation

Article Title: Comprehensive analysis of 6G optical NOMA waveforms using neural network-based detection for high-order QAM modulation

Article References: Kumar, A., Revathi, V., Gaur, N., & Nanthaamornphong, A. (2026). Comprehensive analysis of 6G optical NOMA waveforms using neural network-based detection for high-order QAM modulation. Results in Optics, 25, Article 101159. https://doi.org/10.1016/j.rio.2026.101159

Image Credits: AI Generated

DOI: 10.1016/j.rio.2026.101159

Keywords: 6G, optical wireless communication, NOMA, QAM, recurrent neural network, MMSE detection, deep learning, bit error rate, power spectral density, IM/DD, OFDM, signal detection