The need for alternatives to conventional memory is becoming more urgent as data-intensive computing expands. Electronic memory has benefited from decades of engineering, but shrinking components further creates difficult trade-offs involving power consumption, switching speed, heat dissipation and physical scaling. Conventional computer architectures also separate memory from processing, a design that forces data to travel back and forth between storage and logic. This so-called von Neumann bottleneck can dominate the energy and time required for machine-learning workloads. Optical computing offers a different route: photons can carry information at high bandwidth and with low propagation delay, while multiple signals may be processed in parallel. Yet optical systems still require memory-like elements capable of retaining information, changing their response in controlled increments and being read without destroying the stored state. The new plasmonic proposal addresses these requirements by using a nanoscale material transition to encode optical weights.

At the heart of the device is GST, a phase-change material whose atomic arrangement can be reversibly altered by short optical pulses. In its amorphous state, the atoms lack the long-range order found in a crystal. A suitable heating pulse can induce crystallization, changing the material’s electrical and optical properties. A stronger, shorter pulse can then melt and rapidly quench the material, returning it to an amorphous configuration. These transformations are non-volatile: after the optical stimulus disappears, the material remains in its new state until another programming pulse is applied. In the proposed memory, crystallization is associated with a higher refractive index and lower electrical resistance, while amorphization produces the opposite trend. Because the refractive index determines how light interacts with the nanostructure, the physical phase of GST can be translated into a measurable transmission level.

The programming process depends on the different thermal requirements of the two transitions. The study estimates that crystallizing GST requires about 7.5 picojoules, delivered by a 150-milliwatt pulse lasting 150 nanoseconds. Amorphization requires approximately 2.2 picojoules from a more powerful 110-milliwatt pulse lasting 20 nanoseconds. These figures describe the energy and pulse conditions used for the proposed operating scheme, rather than proving that a complete commercial device has already been fabricated. The distinction matters because phase-change memories face a familiar engineering compromise: higher pulse energies can accelerate switching but increase thermal stress, while repeated cycling can gradually degrade the material or surrounding structure. Precise control is particularly important when a memory is expected to hold many intermediate states instead of simply switching between binary zero and one.

The optical confinement comes from a metal–insulator–metal, or MIM, plasmonic waveguide. In this geometry, light interacts with conducting metal layers separated by a dielectric region, allowing electromagnetic fields to be compressed far below the scale possible in ordinary dielectric waveguides. The proposed design adds two coupled rings containing GST and uses silver waveguides to shape the resonant response. When light at the device’s operating wavelength—1814 nanometers—enters the structure, the local electromagnetic field is strongly influenced by the phase and refractive index of the GST. Small changes in the material can therefore produce comparatively large changes in transmission. This is the central advantage of plasmonics for memory: it can concentrate light into extremely small volumes, enabling compact devices and strong light–matter interaction. The cost is that metals introduce optical absorption, fabrication becomes demanding and heat must be carefully managed.

Rather than limiting the cell to two states, the researchers map GST conditions onto quantized transmission levels. An 8-bit memory can, in principle, distinguish 256 states, allowing one physical cell to represent a finely adjustable synaptic weight. In an optical neural network, such a weight determines how strongly one signal contributes to another, much as the strength of a biological synapse influences the transmission of information between neurons. A multi-level photonic element could therefore perform more computation in place, reducing the number of separate components needed for multiplication and accumulation operations. The device is not described as a biological neuron, nor does it reproduce the full complexity of learning in the brain. Instead, it supplies a programmable optical transfer function that can be assigned a numerical weight. The non-volatile nature of GST would allow those weights to remain available when the programming light is removed.

The reported simulated performance is unusually strong compared with many earlier plasmonic-memory concepts. The optimized structure produces an optical contrast of 87 percent between relevant states and an extinction ratio of 44.04 decibels. Extinction ratio measures how effectively a device distinguishes high- and low-transmission conditions; a larger value generally indicates cleaner separation during readout. The reported insertion loss is 0.60 decibels for logic state one and 45.60 decibels for logic state zero, although the latter value reflects the strongly attenuated state rather than a low-loss transmission path. The design also predicts a readout time of 62 femtoseconds. Such a response is associated with the optical resonance and propagation dynamics of the modeled structure, not necessarily with the slower thermal process used to rewrite GST. Writing and reading are therefore distinct operations: the material may require nanosecond-scale energy pulses to change phase, while a stored state can be interrogated optically on a much shorter timescale.

The proposed cell also includes features intended to make it more practical for integrated photonics. The researchers outline a five-stage back-end-of-line CMOS-compatible fabrication route with a maximum process temperature of 200 degrees Celsius. Keeping the thermal budget low is important because photonic memory elements may eventually need to be fabricated alongside electronic circuits and existing interconnects. The analysis further indicates that dimensional deviations of up to plus or minus 5 nanometers cause only minimal changes in performance. That tolerance could be valuable because nanoscale fabrication inevitably introduces variations in ring dimensions, gaps, layer thicknesses and alignment. Still, tolerance in a numerical design does not eliminate the challenges of real manufacturing. Silver can be chemically and thermally vulnerable, nanoscale GST layers must be deposited uniformly, and the optical response of coupled resonators can be sensitive to roughness and defects. Experimental fabrication and cycling tests will be needed to determine whether the predicted characteristics survive outside the simulation environment.

The study places its design within a rapidly developing field of non-volatile optical memories. Earlier concepts have used GST nanoantennas, ring resonators, plasmonic chains, photonic-crystal waveguides and even photochromic molecules. Reported devices have demonstrated different combinations of optical contrast, switching energy, footprint and state density, but no single architecture has solved every problem. Some offer fast switching but suffer from loss or demanding fabrication; others provide strong contrast but occupy larger areas or require complex thermal control. The double-ring MIM design attempts to combine several desirable properties in one cell: small size, multi-bit storage, non-destructive optical readout, external optical programmability and compatibility with neuromorphic weighting. Its stated density of 4.56 bits per square micrometer is a particularly eye-catching feature, but practical system performance will also depend on how cells are connected, how heat spreads through dense arrays, how often states can be rewritten and how reliably adjacent transmission levels can be distinguished in the presence of noise.

The immediate significance of the work is therefore less a finished optical computer than a blueprint for a compact photonic memory element. If experimental devices confirm the predicted contrast, speed and fabrication tolerance, arrays of these cells could act as programmable weight banks for optical neural networks, allowing computation and storage to occur in the same physical platform. Such systems might eventually process high-bandwidth signals for machine learning, communications or sensing without converting every operation into the electronic domain. Major obstacles remain, including fabrication at scale, thermal crosstalk, material fatigue, calibration of 256 analog-like states and the integration of efficient optical sources and detectors. Even so, the proposal highlights why phase-change plasmonics has become a prominent candidate for next-generation neuromorphic hardware: it links a persistent nanoscale material state to a controllable optical response, potentially turning memory from a passive data store into an active computational component.

At the heart of the approach is “all-optical” processing, where light pulses propagate through carefully engineered optical components to execute transformations such as modulation, routing, and logic-like behavior. By reducing the need to repeatedly convert signals back and forth between optical and electronic forms, researchers aim to cut latency and energy losses that typically grow as frequencies rise.

The paper reports progress toward clocked operation at extreme rates. Operating on a near-100-GHz timescale means that the system must reliably manage ultrafast optical waveforms, maintain timing synchronization, and preserve signal integrity over rapid cycles. Achieving this demands not only fast components, but also stable schemes for controlling and extracting the relevant optical states as each clock tick occurs.

Technically, such performance depends on ultrafast material and device responses, as well as optical interference and nonlinear effects that can occur on very short timescales. Nonlinearities can enable effective switching and signal reshaping—key ingredients for implementing computational primitives without electronic gating. The study emphasizes that careful photonic design can turn these fast physical effects into a repeatable, clocked computation framework.

Importantly, the results are framed for “real computing” rather than only passive demonstrations. The research positions all-optical architectures as candidates for high-throughput processing, especially for tasks that benefit from parallelism and rapid, deterministic timing. If clocked operation at these rates can be made robust, it may open a pathway toward photonic accelerators for sensing, communications, and data-heavy signal processing.

The broader challenge for the field remains scalability: generating, controlling, and detecting many synchronized optical channels at once while keeping losses low. Still, advances toward 100-GHz operation suggest that the speed ceiling for photonic logic is moving outward, bringing optical computing closer to practical deployment.

From a viral-science-news perspective, the headline is clear: computing with light is not just faster in principle—it is edging toward the kind of rhythmic, clock-driven operation that modern computing demands.

Subject of Research: All-optical computing toward 100-GHz clock rates
Article Title: All-optical computing towards 100-GHz clock rates.
Article References: Li, G.H.Y., Parto, M., Ge, J. et al. All-optical computing towards 100-GHz clock rates. Light Sci Appl 15, 321 (2026). https://doi.org/10.1038/s41377-026-02314-5
Image Credits: AI Generated
DOI: 10.1038/s41377-026-02314-5

At the heart of this technology lies the exploitation of microcombs—broadband optical frequency combs generated within microresonators. These microcombs act as a hugely multiplexed light source, supplying a dense array of coherent wavelengths across a broad spectrum. This capability is vital for enabling parallel processing of massive data streams through optical convolution, a core operation in many neural network architectures used for image, signal, and pattern recognition tasks. Unlike traditional electronic processors, the optical convolution processor circumvents electronic bottlenecks by performing operations directly in the photonic domain.

What sets this processor apart is its integrated self-calibration mechanism, a critical innovation addressing the inherent complexity and sensitivity of photonic systems. Optical components often suffer from fabrication imperfections, thermal drift, and environmental fluctuations, which can degrade operational accuracy over time. The proposed self-calibration method dynamically compensates for these discrepancies without interrupting ongoing computations, ensuring consistent precision and stable performance. This feature drastically reduces maintenance overhead and enhances device reliability when deployed in real-world scenarios.

The parallel architecture is meticulously designed to harness the microcomb’s multi-wavelength output, enabling simultaneous convolutional operations across numerous channels. This architectural design delivers a streaming workflow whereby input data undergo convolutional transformation in real time without the need for sequential processing. Such capability is particularly advantageous for applications requiring ultra-fast data analysis, including advanced image recognition, autonomous vehicle sensing, and high-throughput scientific instrumentation. The streaming nature of computation exemplifies the processor’s capacity to handle large-scale, continuous data flows seamlessly.

To validate their approach, the research team constructed a prototype integrating state-of-the-art photonic components such as microresonator-based microcombs, programmable optical delay lines, and balanced photodetectors. Through rigorous experimentation, the processor demonstrated exceptional performance metrics, notably surpassing traditional electronic counterparts in both speed and energy efficiency. Their results also highlighted the robustness of the self-calibration scheme under various environmental stressors, including temperature variations and component aging, affirming the design’s practical resilience.

One of the underlying technological cornerstones of this processor is the sophisticated optical convolution algorithm optimized for hardware implementation. Unlike conventional digital algorithms that operate through binary computation, this optical convolution exploits light intensity modulation and interference in the frequency domain, effectively implementing mathematical operations with photons. This approach not only reduces latency significantly but also minimizes heat dissipation—a notorious limitation in electronic processors—thereby lowering the overall power consumption.

Furthermore, the system’s scalability is a critical facet that future-proofs its utility in ever-evolving computational landscapes. By expanding the microcomb’s spectral bandwidth and enhancing wavelength channel density, the processor can accommodate increasingly complex neural network models and larger datasets. This scalability paves the way for integration into next-generation optical computing platforms, potentially becoming a backbone technology in data centers and artificial intelligence accelerators specialized for intensive convolutional workloads.

The engineering challenges addressed by the research extend beyond fundamental photonics. The interplay between microcomb stability, calibration feedback loops, and real-time data handling required the development of innovative control algorithms and hardware-software co-design. This holistic systems engineering approach ensured that the processor maintains operational stability and high fidelity throughout extended runs, crucial for deployment outside laboratory environments where fluctuating conditions could otherwise degrade performance.

Broadly speaking, the significance of microcomb-enabled optical convolution processors reverberates across multiple scientific and industrial domains. In medical imaging, for instance, faster and more accurate convolution computations enable enhanced real-time diagnostic imaging and processing of massive health data. In autonomous driving systems, rapid image analysis performed on low-power optical hardware increases safety by enabling quicker response times and reducing system latency. Meanwhile, telecommunications could leverage this technology to implement faster signal processing, enhancing the throughput and reliability of optical networks.

The integration of self-calibration functionality marks an important stride toward fully autonomous photonic processors. While traditional optical systems often require manual calibration and frequent recalibration cycles to maintain function, this research demonstrates how adaptive feedback mechanisms—powered by real-time error detection and correction—can sustain optimal operation autonomously. This capability not only simplifies user interaction but also broadens deployment opportunities, facilitating incorporation into consumer electronics and industrial automation systems.

Looking ahead, the researchers envision further enhancements by combining the processor with emerging quantum photonic technologies. Incorporating quantum frequency comb sources and leveraging quantum algorithms could exponentially boost computational efficiency and security, opening paths toward truly transformative computing paradigms. Additionally, miniaturizing the current system through advanced silicon photonics fabrication techniques could yield compact, integrated chips suitable for widespread commercial use.

The publication of this research marks a key milestone in the optical computing revolution. By overcoming fundamental barriers related to calibration, parallelism, and streaming data processing, it charts a clear course for future innovation and application. As the demand for high-performance computing continues to escalate, particularly driven by artificial intelligence and big data analytics, microcomb-enabled optical processors offer a scalable, efficient, and powerful alternative to electronic processors.

In conclusion, the unveiling of the microcomb-enabled parallel self-calibration optical convolution streaming processor represents not just an incremental advance but a paradigm shift. Its combination of high-speed optical convolution, intrinsic self-calibration capabilities, and scalable parallelism addresses critical challenges limiting past photonic computing efforts. With continued development and practical deployments on the horizon, this technology is well-poised to enable the next generation of intelligent, ultra-fast computational systems that meet the demands of tomorrow’s data-driven world.

Subject of Research: Optical computing and convolutional processing using microcomb technology with integrated self-calibration.

Article Title: Microcomb-enabled parallel self-calibration optical convolution streaming processor.

Article References:
Wang, J., Xu, X., Zhu, X. et al. Microcomb-enabled parallel self-calibration optical convolution streaming processor. Light Sci Appl 15, 149 (2026). https://doi.org/10.1038/s41377-025-02093-5

Image Credits: AI Generated

DOI: 10.1038/s41377-025-02093-5 (05 March 2026)

Optical neural networks have long been heralded as the future of artificial intelligence hardware, with their potential to surpass the speed and energy limits of electronic systems. However, their widespread adoption has been hampered by difficulties in achieving nonlinearity with low power consumption, particularly at timescales compatible with ultrafast optical pulses. The newly devised nonlinear activator addresses these challenges by harnessing a reconfigurable mechanism that operates below a femto-joule threshold, enabling stable, chip-integrated nonlinear responses within picosecond regimes.

The team’s approach leverages a sophisticated optical design that integrates nonlinear materials and waveguide architectures in a way that drastically reduces input energy requirements without sacrificing operational speed. This is a crucial aspect, as lower energy thresholds translate directly into reduced heat dissipation and enhanced device scalability—two factors that are critically important for practical deployment in large-scale optical computing systems. By reconfiguring the device’s operational parameters dynamically, this nonlinear activator can adapt to different neural network architectures and signal characteristics on-the-fly, vastly improving its versatility.

At the heart of this development lies the precise control of nonlinear optical effects such as saturable absorption or Kerr nonlinearity, engineered to respond instantly to femtojoule-level inputs delivered in ultrafast picosecond pulses. The nonlinear response effectively performs an all-optical activation function—akin to the nonlinear transformations in biological neurons and essential in artificial neural networks for enabling the computation of complex functions. Prior implementations often required much higher power inputs or were limited by slower electronic control schemes; thus, this advancement marks a pivotal shift towards efficient all-photonic neural computation.

The implications of this technology extend far beyond just energy savings. Optical neural networks equipped with such ultra-low threshold nonlinear activators promise data processing speeds orders of magnitude faster than current electronic counterparts. Leveraging light’s inherent speed, they can conduct massively parallel computations in real-time, making them ideal for applications requiring high throughput and low latency such as real-time image recognition, autonomous vehicle navigation, and advanced signal processing.

Moreover, the reconfigurability of this activator allows for flexible design adaptation post-fabrication, facilitating the tailoring of neural network functions without the need for redesigning or recreating hardware—a significant advantage in research and industrial settings. This flexibility also opens doors to more complex network architectures, including recurrent and convolutional optical neural networks, which demand nuanced nonlinear activations for their operation.

The experimental validation, performed with picosecond pulses at femtojoule energy levels, demonstrates consistent nonlinear activation with rapid recovery times. This ensures that the device can sustain high repetition rates without degradation, an essential feature for real-time data streams. The research team achieved this through careful selection and engineering of nonlinear materials, as well as micro-resonator structures that enhance light-matter interaction, enabling the strong nonlinear response at ultra-low energy thresholds.

In addition to hardware design, the study underscores advancements in the theoretical modeling of optical activation functions. By characterizing the nonlinear device response in detail, the researchers optimized the activator’s behavior, ensuring sharp thresholding and high contrast in signal modulation. These characteristics enhance the robustness and reliability of optical neural network inference, addressing concerns related to noise and error propagation in photonic systems.

Beyond the immediate impact on optical neural network architectures, this innovation showcases the potential of nanoscale photonic components to transform how information processing systems are designed. Utilizing femtojoule-level nonlinearities paves the way for integrating more complex functionalities within compact photonic chips, potentially accelerating the advent of all-optical signal processing platforms that are both fast and energy efficient.

The researchers also highlight the significance of picosecond pulsed inputs in their system, which represent a sweet spot between speed and system complexity. These ultrafast pulses help minimize distortions and timing mismatches, facilitating synchronous operation of network layers and improving overall system coherence. The use of picosecond regimes contrasts with slower electronic or continuous-wave optical signals, which often introduce latency or increase device footprint.

Looking ahead, the reconfigurable nature of the nonlinear activator lends itself to dynamic neural network reprogramming, enabling on-demand modification of network parameters to suit different tasks or environmental conditions. This capability is particularly valuable in adaptive systems or edge computing, where hardware needs to operate under varying loads and input characteristics in real time.

Beyond neural networks and computing, the all-optical nonlinear activation demonstrated here could impact other domains such as optical communications, where nonlinear signal processing is essential for managing channel distortions and enabling advanced modulation schemes. The ability to achieve such effects at femtojoule powers and picosecond timescales could lead to more compact, energy-efficient transceivers and switches.

The study, published in the prestigious journal Light: Science & Applications, represents a significant milestone in photonic integration, marrying fundamental physics with engineering innovation. The convergence of ultra-low power nonlinear optics and flexible device design charts a promising future not only for optical neural networks but also for a broader class of photonic technologies demanding high speed, low power nonlinear control.

As the field continues to evolve, this research sets a benchmark for the performance and versatility of optical nonlinear components, encouraging further exploration into novel materials, device architectures, and integration techniques. The advent of femtojoule-threshold nonlinear activators promises to accelerate the timeline for practical all-optical computing systems to move from laboratory demonstrations to commercial viability.

By pushing the limits of how efficiently light can be manipulated for computation, this work lays the groundwork for a new generation of photonic processors capable of tackling the exponential growth in data and computation demand faced by modern technology landscapes. The synergy between ultrafast nonlinear optics and artificial intelligence hardware holds immense transformative potential, heralding a paradigm shift in the future of computing.

In summary, this innovative work provides a blueprint for harnessing nonlinear optics at previously inaccessible low energy scales, enabling reprogrammable and ultrafast optical neural networks. The intersection of coupled resonator designs with meticulously engineered nonlinear materials creates a versatile platform for advancing optical AI accelerators, potentially reshaping the future of machine learning and data processing technologies globally.

Article Title: Femto-joule threshold reconfigurable all-optical nonlinear activators for picosecond pulsed optical neural networks

Article References:
Liu, R., Wang, Z., Zhong, C. et al. Femto-joule threshold reconfigurable all-optical nonlinear activators for picosecond pulsed optical neural networks. Light Sci Appl 15, 128 (2026). https://doi.org/10.1038/s41377-025-02175-4

Image Credits: AI Generated

DOI: 27 February 2026