A New Way to Control Light for Faster Future Computers

Scientists have created a new type of metamaterial that can offer comprehensive light-blocking functionality for photonic computing.

A metamaterial is an engineered material whose properties do not arise from the chemical composition of its base components but from their carefully designed internal structure. As such, these materials can exhibit unusual properties that are not found in naturally occurring materials.

These materials are typically composed of multiple materials, such as metals and plastics,  and arranged in repeating, sub-wavelength structures. The shape, size, geometry, orientation, and arrangement give them their properties, enabling them to manipulate electromagnetic, acoustic, or seismic waves by absorbing, bending, enhancing, or blocking waves to achieve benefits not possible with conventional materials.

The new metamaterial engineered¹ by scientists at New York University combines features that are typically associated with liquids and crystals but surpass both of them in its ability to block incoming light from all angles.

Termed gyromorphs, the new class of functionally correlated disordered materials merges liquid-like randomness with large-scale structural patterns to block light from every direction. The study stated:

“We generate gyromorphs in 2D and 3D by spectral optimization methods, verifying that they display strong discrete rotational order but no long-range translational order, while maintaining rotational isotropy at short range for sufficiently large 𝐺.” 

With this innovation, the researchers have solved limitations in quasicrystal-based designs that have long been bothering scientists. It can also help drive progress in photonic computing.

From Quasicrystals to Gyromorphs in Photonic Computing

In photonic computing, photons instead of electrical currents are used for performing computations. This new generation of computers, once realized, can be far more efficient and faster than traditional conventional machines.

With data processing at the speed of light, it holds promise for high-performance tasks like AI, but the technology currently faces challenges in miniaturization and cost. 

Advancements in the field have led to the development of functional photonic chips for integration into high-performance computing servers. But light-driven computing is still at an early stage, with researchers struggling to control microscopic streams of light traveling through a chip. 

Carefully designed materials are what we need to successfully reroute these tiny optical signals without weakening their strength. Keeping these signals strong requires a specialized, lightweight substance in the hardware that prevents stray light from entering from any direction. 

A crucial component to achieving this is incorporating the isotropic bandgap material. This material blocks light or other waves from propagating in all directions, as long as the frequencies are within its bandgap. Such material can be disordered yet hyperuniform, meaning it lacks long-range translational order but possesses a specific, controlled type of randomness.

When engineering isotropic bandgap materials, researchers have long focused on quasicrystals.

These structures that follow mathematical rules but do not repeat like traditional crystals were first discovered by scientist Dan Shechtman back in the early 1980s, for which he won the Nobel Prize in Chemistry in 2011.

The discovery was made while researching aluminum and manganese. When the two metals were melted together and rapidly cooled to form an alloy, they exhibited tenfold symmetry under an electron microscope, a property that doesn’t occur in crystalline structures such as metals.

Quasicrystals have properties of crystalline structures, like diamonds, which means they are organized into patterns, as well as amorphous structures like glass, which means those patterns do not repeat. Their unique properties make quasicrystals both durable and brittle.

In a study from the University of Michigan earlier this year, researchers found that quasicrystals are fundamentally stable materials² despite sharing similarity with disordered solids.

“We need to know how to arrange atoms into specific structures if we want to design materials with desired properties,” noted the study’s co-author, Wenhao Sun, the Dow Early Career Assistant Professor of Materials Science and Engineering. “Quasicrystals have forced us to rethink how and why certain materials can form.”

To provide the answers to just why quasicrystals exist or how they are formed, the researchers had to first understand just what makes them stable. For this, they had to determine if quasicrystals are enthalpy- or entropy-stabilized, so the researchers took smaller nanoparticles from a larger simulated block of quasicrystal, then calculated the total energy in each nanoparticle.

The researchers discovered that both the well-explored quasicrystals, an alloy of scandium and zinc, and an alloy of ytterbium and cadmium, are enthalpy-stabilized.

For the calculation, the team utilized quantum-mechanical simulations of quasicrystals, and to solve the computing bottleneck, they had only the neighboring processors communicate rather than every computer processor communicate with one another, which made their algorithm up to 100 times faster.

“We can now simulate glass and amorphous materials, interfaces between different crystals, as well as crystal defects that can enable quantum computing bits.”

– Vikram Gavini, a U-M professor of mechanical engineering and materials science and engineering

In another research, the National Institute of Standards and Technology (NIST) scientists found quasicrystals in a new aluminum-zirconium alloy³, which was formed under the extreme conditions of 3D metal printing.

While adding zirconium to aluminum powder enables printing high-strength aluminum alloys, the NIST team wanted to understand what makes this metal so strong, so it can be used in critical components like military aircraft parts. 

And they found that quasicrystals are responsible for that. Breaking up the regular pattern of aluminum crystals strengthens the alloy. When viewed from just the right angle, the team found the “very rare” fivefold rotational symmetry, in addition to twofold and threefold symmetries, from two different angles.

This, according to NIST physicist and co-author, Fan Zhang, “will open up a new approach to alloy design. With the research showing that “quasicrystals can make aluminum stronger. Now people might try to create them intentionally in future alloys,” he added.

Inside the Gyromorph Revolution: Isotropic Bandgap Materials

Quasicrystals have a lot of promise. They even have the capability to completely block light. But only from limited directions. And while they can weaken light from all directions, they can’t stop it entirely.

To overcome this limitation, scientists have been searching for alternatives that can block signal-degrading light more effectively. This has led to the development of gyromorphs, which can help build materials that prevent stray light from entering from any direction more effectively. According to the study’s senior author, Stefano Martiniani, who is an assistant professor of physics, chemistry, mathematics, and neural science:

“Gyromorphs are unlike any known structure in that their unique makeup gives rise to better isotropic bandgap materials than is possible with current approaches.” 

However, a major obstacle in engineering these materials, whose properties depend on their architecture, is the arrangement required to achieve the desired physical properties.

Published in Physical Review Letters, the New York University researchers detail a novel strategy⁴ to tune the optical behavior.

The team has developed an algorithm that can produce functional structures with built-in disorder. The new form of “correlated disorder” revealed by the team sits between the two extremes: fully ordered and fully random.

“Think of trees in a forest – they grow at random positions, but not completely random because they’re usually a certain distance from one another. This new pattern, gyromorphs, combines properties that we believed to be incompatible and displays a function that outperforms all ordered alternatives, including quasicrystals.”

– Martiniani

During their research, the team observed that all isotropic bandgap materials displayed the same structural signature. So, they focused on making it “as pronounced as possible,” leading to the creation of gyromorphs.

The resulting new class of materials, lead author Mathias Casiulis, a postdoctoral fellow in NYU’s Department of Physics, stated, “reconcile seemingly incompatible features,” because they don’t have a crystal-like, fixed, repeating structure, which gives them a liquid-like disorder. At the same time, though, when looking from a distance, they form regular patterns.

“These properties work together to create bandgaps that lightwaves can’t penetrate from any direction.”

– Casiulis 

The team also introduced “polygyromorphs” with multiple rotational symmetries at various length scales to enable the formation of multiple band gaps in a single structure, thus opening the doors to achieving fine control over optical properties.
Swipe to scroll →

AI and Next-Generation Quantum Materials in Discovery

As researchers continue to push deeper into next-generation materials, entirely new classes of materials are emerging.

Recently, a research team led by the Department of Energy’s Berkeley Lab reported the discovery⁵ of “berkelocene,” an organometallic molecule that contains the synthetic, heavy, radioactive chemical element berkelium.

The molecules consist of a metal ion surrounded by a carbon-based framework, and while relatively common for early actinide elements, they are scarcely known for later ones.

“This is the first time that evidence for the formation of a chemical bond between berkelium and carbon has been obtained. The discovery provides new understanding of how berkelium and other actinides behave relative to their peers in the periodic table,” said co-author Stefan Minasian, a scientist in Berkeley Lab’s Chemical Sciences Division, which has been working on preparing actinides’ organometallic compounds as they allow them to observe actinides’ distinct electronic structures.

Actinides are a series of 15 radioactive metallic elements on the periodic table, located in the f-block. Uranium and plutonium are examples of actinides. They are known for their radioactive properties and are used in nuclear reactors and other technologies.

Last year, a partnership between researchers from Uppsala University, Sweden, and Columbia University, US, led to the discovery of a 2D quantum material called CeSiI⁶, with a crystal structure of cerium, silicon, and iodine. Its crystal structure resembles a two-dimensional arrangement of distinct, atom-thin layers. 

The electrons of CeSil behave as heavy fermions, with an effective mass up to 100 times greater than in ordinary materials. This effective mass is anisotropic; thus, it depends on the direction in which the electrons move in the atomic layers.

“With this discovery, we now have a significantly improved material platform for investigating correlated electron structures. 2D materials are like a construction kit with LEGO pieces. Our partners are already working on adding layers from other 2D materials to create a new material with customized quantum properties.”

– Chin Shen Ong from the Department of Physics and Astronomy at Uppsala

In materials science, there are countless possibilities, and selecting the right material is a key hurdle to making new discoveries. While theory-driven predictions and experiment-based validations help inform selection, it has remained fragmented.

This is where AI-driven materials informatics is taking over, integrating quantum-scale insights with large datasets to rapidly screen, model, and optimize new materials that would be impossible to discover through conventional trial-and-error.

A team of researchers at Tohoku University built an AI-built materials map⁷ to unify all the experimental data with representative first-principles computational data, with the aim of helping researchers find the right material for a given situation.

The map is a large graph with axes for structural similarity and thermoelectric performance (zT), with each datapoint representing a material. Similar materials here appear in close proximity. As these materials are usually synthesized and evaluated using similar methods and devices, the map allows experimentalists to rapidly spot analogs of unknown high-performance materials and to repurpose existing synthesis protocols as next steps.

This way, the tool can help reduce development costs and accelerate innovation and its real-world deployment. In the future, the team plans to extend their framework beyond thermoelectrics to include topological and magnetic materials and to incorporate additional descriptors to create a comprehensive, AI-assisted materials design support platform.

“By providing an intuitive, bird’s-eye view over many candidates, the map helps researchers to select promising targets at a glance; therefore, it is expected to substantially shorten development timelines for new functional materials.”

– Associate Professor Yusuke Hashimoto

Meanwhile, a study from the University of Gothenburg developed an AI model to efficiently determine the strength and durability⁸ of woven composite materials.

Performing physical tests and detailed computer simulations to design new high-quality composite materials, is “particularly difficult when the composite is created as a woven textile fiber material, where the fibers are wrapped around each other and behave differently depending on the forces the material is subjected to,” noted Ehsan Ghane, a Ph.D. student at the Department of Physics at the University of Gothenburg.

While computers can already simulate realistic microstructures based on a material’s interactions and influences, woven composite materials still require substantial computational resources. Neural networks offer an alternative, but they require large amounts of training data and struggle to extrapolate, so the team developed a generalized AI model that doesn’t require as much data.

The model has been trained on existing simulation and test data for the composite’s constituent materials, enabling it to predict the durability of the new composite.

While the Gothenburg study investigated methods to integrate material laws into the AI model, a team of researchers from KAIST has combined physical laws with its AI model to allow for the rapid exploration of new materials even when data is noisy or limited.

Property identification is one of the key steps in developing new materials, but it requires massive amounts of experimental data and expensive equipment, which limits research efficiency. The KAIST team overcame that need by integrating the laws that govern the deformation and interaction of materials and energy.

The researchers reported a physics-informed neural network (PINN) technique⁹ to detect material properties and deformation behavior using only a small amount of data from a single experiment. They then introduced an AI model, the Physics-Informed Neural Operator (PINO), that understands the laws of physics and can generalize to unseen materials.

MIT researchers took it even further by developing a method that incorporates information from multiple sources¹⁰: literature, chemical compositions, microstructural images, and more. 

It is part of the new Copilot platform for Real-world Experimental Scientists (CRESt). Their method uses robotic equipment to enable high-throughput testing of materials, then feeds the results back into large multimodal models to improve their recipes.

The researchers used this “assistant, not a replacement, for human researchers,” to explore over 900 chemistries and conduct 3,500 electrochemical tests that led to the discovery of a catalyst material that delivered record power density in a fuel cell to generate electricity.

Investing in Materials Science Advancement

In the world of materials science, ATI Inc. (ATI -1.25%) is known for its technically advanced specialty materials and complex components. The company produces high-performance materials for aerospace, defense, medical, electronics, and energy markets.

ATI’s products are made from nickel-based alloys and superalloys, titanium and titanium-based alloys, and specialty alloys. It operates through two segments:

• High Performance Materials & Components (HPMC)
• Advanced Alloys & Solutions (AA&S)

With a market cap of $13.5 billion, ATI shares are trading at $99.37, up 80.5% this year. It has an EPS (TTM) of 3.10 and a P/E (TTM) of 32.09. The company pays a dividend yield of 0.32%.

Recently, ATI reported a 7% YoY increase in sales to $1.13 billion for Q3 2025. Its GAAP net income was $110 million, and its adjusted net income was $119 million, both up 30% year over year.

Earnings per share for the quarter were $0.78, and adjusted earnings per share were $0.85.

“We exceeded our guidance in the third quarter, delivering strong adjusted earnings and operating cash flow performance. We continue to see positive demand signals in our core markets, as our customers ramp to achieve their growth targets. We are well-positioned to grow our defense-related business through an expanding mix of highly differentiated products critical to the US and our allies.”

– CEO Kimberly A. Fields

Latest ATI Inc. (ATI) Stock News

Conclusion: Why Gyromorphs Matter for Photonics and Investors

Material science is key to technological advancement, and it is being accelerated by AI. The convergence of experimentalists, theorists, and machine-learning systems is drastically reducing the timeline from discovery to application, enabling the rapid identification, simulation, and optimization of materials that once required decades of trial-and-error research. 

Amidst this, the discovery of gyromorphs represents a striking milestone. By merging disorder with pattern, this new class of material enables unprecedented control over light and carries the promise of unlocking the next major leap in photonic computing.

References

1. Casiulis, M., Shih, A., & Martiniani, S. “Gyromorphs: A New Class of Functional Disordered Materials.” Physical Review Letters 135(19) (2025). https://doi.org/10.1103/gqrx-7mn2
2. Baek, W., Das, S., Tan, S., Gavini, V., & Sun, W. Quasicrystal stability and nucleation kinetics from density functional theory. Nature Physics 21, 980–987 (2025). https://doi.org/10.1038/s41567-025-02925-6
3. Liu, Y., Zhang, H., Zhou, X., & Chen, Y. “Catalytic performance of novel mixed-metal oxides for CO₂ hydrogenation.” Applied Catalysis B: Environmental 330 (2025). https://doi.org/10.1016/S0925-8388(25)01839-0
4. Casiulis, M., Shih, A., & Martiniani, S. Gyromorphs: A New Class of Functional Disordered Materials. Physical Review Letters 135, 196101 (2025). https://doi.org/10.1103/gqrx-7mn2
5. Russo, D. R., Gaiser, A. N., Price, A. N., Sergentu, D-C., Wacker, J. N., Katzer, N., Peterson, A. A., Branson, J. A., Yu, X., Kelly, S. N., Ouellette, E. T., Arnold, J., Long, J. R., Lukens, W. W., Jr., Teat, S. J., Abergel, R. J., Arnold, P. L., Autschbach, J., & Minasian, S. G. Berkelium–carbon bonding in a tetravalent berkelocene. Science 387(6737), 974–978 (2025). https://doi.org/10.1126/science.adr3346
6. Posey, V. A., Turkel, S., Rezaee, M., Devarakonda, A., Kundu, A. K., Ong, C. S., Thinel, M., Chica, D. G., Vitalone, R. A., Jing, R., Xu, S., Needell, D. R., Meirzadeh, E., Feuer, M. L., Jindal, A., Cui, X., Valla, T., Thunström, P., Yilmaz, T., Vescovo, E., Graf, D., Zhu, X., Scheie, A., May, A. F., Eriksson, O., Basov, D. N., Dean, C. R., Rubio, A., Kim, P., Ziebel, M. E., Millis, A. J., Pasupathy, A. N. & Roy, X. Two-dimensional heavy fermions in the van der Waals metal CeSiI. Nature 625, 483-488 (2024). https://doi.org/10.1038/s41586-023-06868-x
7. Hashimoto, Y., Jia, X., Li, H., & Toma, T. “A materials map integrating experimental and computational data via graph-based machine learning for enhanced materials discovery.” APL Machine Learning 3(3), 036104 (2025). https://doi.org/10.1063/5.0274812
8. Ghane, E. Learning from Data and Physics for Multiscale Modeling of Woven Composites. PhD thesis, University of Gothenburg, 2025. https://hdl.handle.net/2077/85666
9. Moon, H., Park, D., Cho, H.-K., Noh, H.-K., & Ryu, S. Physics-informed neural network-based discovery of hyperelastic constitutive models from extremely scarce data. Computer Methods in Applied Mechanics and Engineering 446, 118258 (2025). https://doi.org/10.1016/S0045782525005304
10. Zhang, Z., Ren, Z., Hsu, C.-W., Chen, W., Hong, Z.-W., Lee, C.-F., Penn, A., Xu, H., Zheng, D. J., Miao, S., Huang, Y., Gao, Y., Chen, W., Smith, H., Niu, Y., Tian, Y., Lu, Y.-R., Shao, Y.-C., Li, S., Wang, H.-T., Abate, I. I., Agrawal, P., Shao-Horn, Y., & Li, J., A multimodal robotic platform for multi-element electrocatalyst discovery. Nature 647, 390–396 (2025). https://doi.org/10.1038/s41586-025-09640-5