From Feasibility to Convergence: The Innovation Timeline

Photonic signal processing has progressed through three distinct phases between 2009 and 2026, each marked by a shift in what researchers could demonstrate on chip. The dataset of patent and literature records surveyed here spans that full period, revealing a field that moved from proving on-chip optical integration was possible to deploying it in AI, wireless, and quantum workloads.

The foundational phase (2009–2014) established that photonic functions — signal conversion, quantum detection, ultra-wideband pulse generation, and nanophotonic CMOS interconnects — could be realised on chip. Osaka University’s 2010 review of optical signal form conversion, NIST’s 2011 on-chip photon-number-resolving detectors at telecom wavelengths, the Technical University of Denmark’s 2009 UWB pulse generator, and Intel’s 2013 nanophotonic CMOS interconnect scaling framework collectively set the engineering baseline.

The development and scaling phase (2015–2019) saw significant investment in silicon photonics telecom platforms and programmable PICs. NTT’s Si-Ge-silica integration platform (2014) and the PLAT4M European project (Thales, 2015) consolidated telecom-oriented PIC development. Lightmatter’s 2018 review of programmable nanophotonic processors and Harvard University’s 2019 demonstration of ultra-low-loss thin-film lithium niobate — achieving 6 dB/m propagation loss at visible wavelengths and resonators with Q-factors of 11 million — signalled rapid capability building beyond silicon.

The convergence and AI-integration phase (2020–2023) is the most densely populated in the dataset. Huawei Technologies Canada’s photonic MIMO accelerator (2021), RMIT University’s Kerr microcomb video processor (2021), Ghent University-IMEC’s programmable PIC ecosystem paper (2020), and the University of Strathclyde’s GHz-rate VCSEL spiking neural network (2023) all appeared within this window. The EPFL 2022 Roadmap on integrated quantum photonics — identifying up to 650 integrated optical and electrical components on a single chip — marks the period’s integration benchmark.

Four Technology Clusters Shaping the Field

Photonic signal processing is not a single technology but a family of overlapping approaches, each with distinct performance characteristics, fabrication requirements, and application fits. Four clusters account for the majority of innovation activity in the dataset.

Silicon Photonics and PIC Integration Platforms

Silicon photonics exploits CMOS-compatible fabrication to integrate modulators, detectors, switches, and arrayed waveguide gratings on silicon-on-insulator (SOI) substrates. Key performance drivers include insertion loss reduction, high-speed modulation above 40 Gb/s, and wavelength-division multiplexing (WDM) scalability. InP platforms complement silicon by enabling monolithic integration of active and passive components, critical for telecom and 5G front-haul. RISE AB demonstrated 200 Gbps/lane interconnects for short-reach applications in 2020, while the University of Bordeaux’s 2021 work targeted beyond-5G THz signal generation using InP-based OEICs. Standards bodies including IEEE have been instrumental in defining interoperability requirements for these platforms.

Programmable and Reconfigurable Photonic Processors

Programmable photonic integrated circuits use waveguide mesh architectures of tunable couplers and phase shifters that can be reconfigured in software to implement arbitrary signal processing functions. These “generic programmable PICs” reduce development cycles by bypassing design-fabrication iterations. A key advance is the ability to implement universal linear optical transformations applicable to both classical signal processing and quantum information tasks. Ghent University-IMEC (2020) and Lightmatter (2018) are identified as early movers in this architectural direction; Huazhong University of Science and Technology demonstrated a self-configuring and reconfigurable silicon photonic signal processor in 2020.

Neuromorphic Photonics and Photonic Neural Networks

Neuromorphic photonic architectures exploit the nonlinear dynamics of semiconductor lasers — particularly VCSELs and feedback-coupled distributed feedback (DFB) lasers — to emulate neuron-like spiking behaviour, enabling reservoir computing and spiking neural network implementations at sub-nanosecond rates. Classification tasks including spoken digit recognition and chaotic time-series prediction have been demonstrated at greater than 1 Gbyte/s data rates by IFISC (UIB-CSIC) using transient states of semiconductor lasers. The University of Strathclyde’s 2023 experimental GHz-rate spiking neural network, built with a single VCSEL, demonstrated sub-nanosecond spike processing for pattern recognition and image classification.

“Classification tasks including spoken digit recognition and chaotic time-series prediction have been demonstrated at greater than 1 Gbyte/s data rates using photonic reservoir computing — a speed regime entirely inaccessible to conventional electronic neural network hardware.”

Photonic Computing for Broadband Signal and Image Processing

Wideband optical processing exploits the terahertz bandwidth of photonic carriers for real-time image processing, anti-jamming signal separation, LiDAR signal handling, and massive-MIMO acceleration. Optical frequency comb sources — specifically Kerr soliton microcombs — enable massively parallel channel processing. RMIT University demonstrated real-time video image processing with Kerr microcombs in 2021. Huawei Technologies Canada proposed massive-MIMO matrix computation using photonic processors in the same year, while Rowan University demonstrated anti-jamming wideband signal cancellation at zero latency using photonic analog processing.

Where Photonic Signal Processing Is Being Deployed

Photonic signal processing is being applied across six distinct domains, each demanding different performance trade-offs and creating distinct IP opportunities.

Telecommunications and Data Centers

The largest cluster of applications in the dataset centres on high-speed optical interconnects and coherent transmission systems. Silicon photonics supports 200 Gbps/lane interconnects for short-reach optical interconnects, while InP-based OEICs target beyond-5G THz signal generation. Optical time-division multiplexed (OTDM) packet networking addresses ultra-high-speed TDM, and photonic integrated access network transceivers for passive optical networks (PON/NGPON2) have been demonstrated on InP platforms by PICadvanced in 2020. The ITU has published standards for coherent optical transmission that underpin many of these deployments.

Wireless and RF Signal Processing

Photonic technologies address millimetre-wave and THz signal generation and processing for radar, satellite, and 5G/6G radio access. Anti-jamming wideband signal cancellation at zero latency has been demonstrated using photonic analog processing (Rowan University, 2021). Massive-MIMO matrix computation using photonic processors was proposed by Huawei Technologies Canada in 2021. The Polytechnic University of Valencia contributed foundational work on photonic technologies for millimetre- and submillimetre-wave signals as early as 2012.

LiDAR and Autonomous Sensing

Silicon photonics optical phased arrays (OPAs) have been reviewed for automotive LiDAR applications by the Chinese Academy of Sciences Xi’an Institute of Optics in 2019. Photonic signal processing architectures using acousto-optic modulators reduce detector bandwidth constraints in phase-coded LiDAR, as demonstrated by Harbin Institute of Technology in 2023 — promising higher-resolution automotive and environmental sensing without requiring wider-bandwidth electronic detectors.

Quantum Information Processing

Integrated quantum photonic circuits (QPICs) combining up to 650 optical and electrical components on a single chip are targeted for programmable quantum computing, chip-to-chip networking, and quantum communications, according to the EPFL 2022 Roadmap. On-chip photon-number-resolving detectors at telecom wavelengths — demonstrated by NIST in 2011 — represent an enabling component for scalable quantum photonic information processing. Sapienza University of Rome’s 2018 comprehensive review covers entanglement, linear optical circuits, and boson sampling as the core primitives.

On-Chip Optical Interconnects

Photonic networks-on-chip (PNoC) are proposed to replace energy-limited electronic interconnects in multi-core processors. The University of Kentucky contributed both the PROTEUS adaptive laser power management system for PNoCs and a silicon-on-sapphire platform targeting greater than 1 Tb/s aggregated data rates, both in 2020. Colorado State University developed hybrid photonic NoC synthesis tools in 2016.

Optical Computing and AI Acceleration

The emergence of optical computing for general-purpose AI inference is documented in an IPSI RAS status review published in 2022. Femtosecond laser fabrication is used to produce three-dimensional integrated photonic devices for multi-dimensional multiplexing and optical amplification, as demonstrated by Huazhong University of Science and Technology in 2022. Research published in Nature and affiliated journals has tracked the progression of optical matrix processors as practical AI inference accelerators.

Geographic and Assignee Landscape

Innovation in photonic signal processing is distributed across a wide range of academic and industrial assignees in multiple jurisdictions, with no single dominant commercial actor. Among retrieved results, academic institutions outnumber commercial assignees approximately 3:1, with commercial activity concentrated in the US and emerging players such as Huawei Technologies Canada and RMIT-associated industry.

United States: Intel Corp. (nanophotonic CMOS interconnects, 2013), NIST (on-chip quantum photon detectors, 2011), Lightmatter (programmable nanophotonic processors, 2018), and Harvard University (thin-film lithium niobate visible photonics, 2019) represent a strong US presence spanning commercial, federal lab, and academic contributors.

China: Multiple Chinese Academy of Sciences institutes, Huazhong University of Science and Technology (self-configuring silicon photonic processor, 2020; femtosecond laser chips, 2022), Fudan University (optical injection locking for VLC, 2023), Harbin Institute of Technology (phase-coded LiDAR, 2023), and Huawei Technologies Canada (photonic MIMO, 2021) demonstrate deep Chinese investment, particularly in silicon photonics, LiDAR, and wireless acceleration.

Europe: Strong academic institutions dominate: Ghent University-IMEC (Belgium, programmable PICs, 2020), EPFL (Switzerland, quantum photonics roadmap, 2022), University of Strathclyde (UK, VCSEL neuromorphic, 2023), IFISC UIB-CSIC (Spain, reservoir computing, 2013/2020), Polytechnic University of Valencia (Spain, microwave photonics), University of Bordeaux (France, InP for beyond-5G), and the University of the Aegean (Greece, EP patent, 2025).

Japan: NTT Corporation (Si-Ge-silica photonics platform, 2014), Keio University (silicon photonics review, 2020), and Osaka University (optical signal processing, 2010) contribute foundational work in telecom-oriented platforms.

Six Emerging Directions to Watch Through 2026

Publications and patents from 2020 onward point to six directions that are transitioning from research prototypes to patentable system architectures or early commercial products.

1. Transient Optical Computing on Integrated Chips (2025). The University of the Aegean’s EP patent (2025) introduces a photonic integrated circuit performing optical transient computing on RF-modulated carriers using filter-delay feedback loops with group delays of 1–100 ps and offset centre frequencies of 1–100 GHz. This directly enables trainable readout for in-chip machine learning inference on modulated optical signals.

2. GHz-Rate VCSEL-Based Photonic Spiking Neural Networks (2023). The University of Strathclyde’s experimental GHz-rate spiking neural network, built with a single VCSEL, demonstrates sub-nanosecond spike processing for pattern recognition and image classification, pointing toward ultrafast spike-based inference engines.

3. Photonic Processing for LiDAR Resolution Enhancement (2023). Acousto-optic modulator-based photonic signal processing architectures for phase-coded LiDAR (Harbin Institute of Technology, 2023) address the bandwidth limitations of electronic detectors, promising higher-resolution automotive and environmental sensing.

4. On-Chip Visible-Band Optical Parametric Oscillation (2020). The Joint Quantum Institute and NIST demonstrated on-chip optical parametric oscillation across the full visible spectrum — red through green — in Si₃N₄ microresonators in 2020, enabling compact on-chip visible light sources for spectroscopy, quantum science, and sensing.

5. Orbital Angular Momentum (OAM) On-Chip Integration (2020). Vortex microlasers with tunable OAM states and direct photocurrent detection of topological charges are proposed for on-chip multiplexing by the National University of Singapore (2020), representing a new physical degree of freedom for photonic signal processing multiplexing beyond wavelength and polarisation.

6. Sub-1 nm Spectral Resolution for Semiconductor Process Monitoring (2026). Verity Instruments’ 2026 KR patent combines narrow-pass filters and optical etalons to achieve sub-1.0 nm optical bandwidth for semiconductor process monitoring, indicating photonic signal processing expanding into precision metrology and fabrication control.

Strategic Implications for R&D and IP Teams

Five strategic observations emerge from the dataset for teams working in or entering the photonic signal processing space.

Programmable PIC platforms are becoming the dominant integration paradigm. Teams entering the silicon photonics space should consider software-defined, reconfigurable architectures — waveguide mesh PICs — over application-specific designs to reduce development cycles and enable multi-function deployment. Ghent University-IMEC and Lightmatter represent early movers in this architectural direction.

Neuromorphic photonics is transitioning from proof-of-concept to hardware platforms. The convergence of VCSEL-based spiking neural networks with GHz-rate operation (Strathclyde, 2023) and transient computing patents (University of the Aegean, EP 2025) signals an imminent commercialisation window for photonic inference accelerators. IP strategists should monitor EP and US filings in the integrated optical neural computing space.

LiDAR, autonomous systems, and 5G/6G wireless are the highest-growth application verticals in this dataset, with Chinese academic institutions — including the Chinese Academy of Sciences, Harbin Institute of Technology, Fudan University, and HUST — filing heavily across sensing and communications, suggesting competitive IP pressure in these verticals for non-Chinese entrants.

Thin-film lithium niobate (TFLN) and non-silicon platforms represent underexploited IP space. Harvard’s 2019 TFLN demonstration is foundational; relatively few follow-on commercial patents are visible in this dataset, suggesting opportunity for IP capture in TFLN-based modulators, switches, and visible-band processors.

Quantum photonic integration is on the cusp of scaling. The EPFL 2022 Roadmap identifies 650-component QPICs as a demonstrated benchmark, with chip-to-chip networking and hybrid quantum-classical integration as near-term engineering targets. R&D teams in quantum computing and secure communications should align quantum photonic signal processing architectures with the platform roadmaps emerging from EPFL, NIST, and European consortia. Organisations such as WIPO track patent activity in quantum technologies that can help teams identify white space and freedom-to-operate risks in this rapidly evolving area.

Teams monitoring this space can use PatSnap’s innovation intelligence platform to track real-time filings across these clusters, map assignee activity by jurisdiction, and identify white space in TFLN, OAM multiplexing, and neuromorphic photonic architectures before the commercialisation window closes.