Extremely cold and highly efficient: Scientists present new light fiber

Max Planck Institute for the Science of Light

Facebook

XLinkedIn
MessageWhatsAppEmail

 

image: 

Artist impression of a frozen optical fiber core in a glass capillary, which guides and couples light and sound waves efficiently

view more 

Credit: @ Philipp Denghel

When volcanoes erupt, one can observe the liquid streams of lava cool down and solidify into rock formations at the bottom of the volcanoes. The same physical process – a liquid changing into the solid phase when cooling – can be observed when lakes start freezing over cold winters. These phase changes always come with changes in the physical properties of the material, for example the density or the refractive index that govern how sound and light move through it. This fundamental physical process is also used during the melting of glass preforms to loosen their structure while pulling optical fiber. These fibers then guide light through their cores, allowing the transmission of information via light over long distances very quickly, which is why they are used widely for telecommunication applications. For more specialized applications, for example fiber lasers, fiber endoscopes or fiber sensors, other types of optical fibers have been developed. Hollow core fibers, for instance, can be filled with different gases or liquids and measure temperature distributions, or act as microscopic chemistry labs.

In a collaboration, researchers at the Max Planck Institute of the Science of Light (MPL) in Erlangen, the Leibniz University Hannover (LUH) and the Leibniz Institute for Photonic Technologies (IPHT) in Jena have developed a new type of optical fiber by freezing such liquid core optical fibers (LiCOF) in nitrogen at -196 °C, leading to a phase change in the fiber core from liquid to solid. “The key point is, that the frozen section of the LiCOF retains its ability to guide light. Not only that, but both the liquid and the frozen section of the fiber also guide hypersonic sound waves,” says Simon Seiderer, one of the three lead authors of the article and a researcher in the “Quantum Optoacoustics” research group of Prof. Dr. Birgit Stiller, who leads the project.

The researchers utilize the extremely efficient coupling between light and sound in their new fiber, an effect known as Brillouin-Mandelstam scattering. The effect is already well known in traditional optical fibers, however, with the phase transition to the frozen LiCOF, the researchers create an extreme, highly confined and dense environment. Here, the optoacoustic coupling becomes more than 1000 times stronger than in standard optical fibers. By harnessing this efficient coupling, the researchers demonstrated  optoacoustic memory. This fundamental building block for photonic neuromorphic computing in fibers works by utilizing the drastic differences in velocities between light and sound waves. Information is transferred from the fast light wave to the much slower sound waves, and later converted back into light. The efficient optoacoustic coupling in the frozen LiCOF opens up new avenues for drastically reducing the energy consumption of photonic computing architectures.

The project was possible thanks to the established cooperation with Prof. Markus Schmidt and Prof. Mario Chemnitz from the IPHT Jena, who pioneered the research with liquid core optical fibers. With this additional step of freezing the LiCOFs core, higher nonlinearities became possible. “By freezing the liquid core, we have created an entirely new physical platform that provides extreme nonlinearities while being easy to handle,” says Stiller. “While demonstrating a highly efficient optoacoustic memory is a fantastic first step, this level of light-sound coupling not only opens up exciting new possibilities for neuromorphic computing, but also for quantum information processing, microwave photonics and high-precision sensing.”

---

Journal

Optica

DOI

10.1364/OPTICA.600056 

Method of Research

Experimental study

Subject of Research

Not applicable

Article Title

Giant Brillouin gain in frozen CS2 capillaries

Article Publication Date

19-Jul-2026

 

New encryption method uses mathematical "chaos" to lock down sensitive medical images

Why medical images need special protection

Bentham Science Publishers

Facebook

XLinkedInWhatsAppEmail

The article by Mangal Deep Gupta is published in Wireless and Communication Letters (In Press, available online)

The study tests a chaos-based encryption scheme on real patient scans and report categories, aiming to keep e-healthcare data safe as it disseminates across the internet. Hospitals and clinics increasingly store and share patient information digitally in the form of X-rays, CT scans, and other medical images that deeply reveal personal health details. This convenience comes with risk: once this data moves across the internet, it becomes a target for interception or tampering. A new study, published ahead of print in Wireless and Communication Letters, proposes a way to protect these images using the mathematics of chaos.

Why Medical Images Need Special Protection

Encrypting medical images is not quite the same problem as encrypting a text message or a password. Images contain large amounts of structured data, and neighboring pixels tend to look similar to each other — a property that traditional encryption methods do not always account for well. If an encryption scheme leaves patterns intact, it becomes easier for an attacker to guess the original image even without fully breaking the encryption. This is especially concerning for medical data, where the images in question may include sensitive material such as scans revealing a patient's gender-related medical information, brain imaging, kidney-related studies, or lung X-rays.

Borrowing From Chaos Theory

The method described in the study relies on what's known as a chaotic system — a mathematical system that behaves in ways that look random and unpredictable, even though it follows precise underlying rules. Because tiny differences in starting conditions can lead to wildly different outcomes, chaotic systems are well-suited to generating the kind of unpredictable, hard-to-reverse-engineer sequences that strong encryption depends on. Specifically, the study uses Chen's chaotic system to drive a pseudorandom number generator, which in turn scrambles the pixel values of the original medical image into what appears to be visual noise.

How the Method Was Tested

The encryption process was simulated using MATLAB, a widely used engineering and scientific computing tool, rather than tested on physical hardware. To evaluate how well the encryption worked, the study relied on several standard measures used in image security research. Histogram analysis, essentially a chart showing how pixel brightness values are distributed across an image, was used to compare the original, encrypted, and decrypted versions of each image. A well-encrypted image should show a flat, uniform histogram, meaning there is no leftover pattern for an attacker to exploit, in contrast to the uneven, structured histogram typical of an unencrypted photo or scan.

The study also measured the correlation coefficient for both the original and encrypted images, which indicates how similar neighboring pixels are to each other, along with two additional standard security metrics: the Number of Pixel Change Rate (NPCR) and the Unified Average Changing Intensity (UACI). Both of these assess how sensitive the encryption is to small changes in the original image — a property that matters because it makes the encryption harder to crack through trial-and-error.

What the Results Showed

According to the study, the encrypted versions of the medical images showed consistently flat, uniform pixel intensity distributions, which the author interprets as a sign of strong resistance to common image-analysis attacks. In other words, the encrypted images revealed very little about the structure of the original scan or report.

Practical Implications

The author frames the technique as a candidate building block for real-world hardware, describing its potential use in designing an image crypto-processor — a dedicated chip or circuit that could handle image encryption directly on a device. That kind of hardware-level implementation is relevant for real-time applications where speed and device-level security both matter, such as secure transmission of medical images between healthcare facilities.

About the Corresponding Author

Mangal Deep Gupta (corresponding author, marked with an asterisk in the original publication) is an Assistant Professor in the Department of Electronics and Communication Engineering at Babasaheb Bhimrao Ambedkar University (BBAU), Lucknow, Uttar Pradesh, India, where his work centers on VLSI design, FPGA implementation, and cryptographic hardware.

---

Article title: Securing E-healthcare Data on the Internet Using Chaotic System-based Image Encryption.

Read the article here: https://bit.ly/44BprP9

---

Journal

Wireless and Communication Letters

DOI

10.2174/0126669463393519251127152243 

Article Title

New Encryption Method Uses Mathematical "Chaos" to Lock Down Sensitive Medical Images

New computer graphics research: From volcanic lava to robot vacuum room exploration

Researchers from the Institute of Science and Technology Austria (ISTA) present two computer graphics papers at SIGGRAPH 2026

Institute of Science and Technology Austria

LinkedInWeChatBlueskyMessageWhatsAppEmail

 

image: 

 A gooey blob is smeared, and its texture map gets distorted (black & white). Naïve deformation (red) compared to the ISTA team’s spectrum-rescaling algorithm (magenta) and isotropy-preserving method (blue). © Kalinov et al., ACM Transactions on Graphics

view more 

Credit: © Kalinov et al., ACM Transactions on Graphics

Representing Extremely Deformable Surfaces

To create realistic movies, producers often use 3D modeling to build reusable digital representations of objects. The artistic challenge is to make these digital models look lifelike. In animated models, objects without a distinct shape at rest are particularly challenging to represent.

In a study presented at this year’s SIGGRAPH, PhD student and first author Aleksei Kalinov and colleagues from Professor Chris Wojtan’s group at the Institute of Science and Technology Austria (ISTA) sought to represent extremely deformable surfaces more realistically while cutting computational costs. To do so, they took inspiration from nature.

“If a feature is preserved when natural systems such as lava or cake batter are deformed, we consider that nature ‘likes’ it and that it’s an ‘allowed’ texture feature,” Wojtan explains. “We then try to come up with nature-inspired rules that are cheap to compute.”

Ultimately, the team aims to achieve flicker-free, realistic representations with seamless details when modeling deformable surfaces by preserving or restoring ‘allowed’ texture features.

From cake batter to computer graphics

While mixing cake batter, air bubbles appear. But instead of an air bubble stretching infinitely—as an image might when one keeps zooming in—the batter’s surface eventually ruptures and forms new features as mixing continues.

To visually represent this process, the team modeled the textures using physical parameters such as frequency and amplitude rather than in terms of pixels.

In fact, stretching does not actually erase surface features. Rather, only the high-frequency information is lost and replaced with low-frequency information as the surface is stretched. In other words, the stretched surface ‘regenerates’ the information it carries, but at a lower frequency. This process can be reversed, bringing back the initial high-frequency surface information.

Compression has the opposite effect: high-frequency detail replaces the lost low-frequency information of the surface’s resting state. However, compressing a surface infinitely can make high-frequency information finer than the pixel size, leading to flickering—a phenomenon that computer scientists call aliasing.

“To overcome this risk of causing aliasing, our graphical models must feature an automatic process that prevents the details from reaching excessively high frequencies,” says Wojtan. “Taking inspiration from nature, we see what features are found in similar systems and add them back. Our method becomes more natural by acting as a restoring force in frequency space.”

Tackling persistent graphics problems

The team developed two modeling approaches that overcome the limitations of existing surface deformation models in two different ways. Their “spectrum-rescaling algorithm” restores high‑frequency details consistent with the deformation. On the other hand, the “isotropy-preserving method” aims to counteract deformation and preserve the initial texture. Both are fully procedural algorithms that generate information automatically, without physics simulations or stored deformation history. As a result, they considerably reduce computational costs.

Ultimately, these techniques could be applied in visual effects, video games, and simulation software. However, the team’s research focuses on more fundamental principles.

“We seek to develop fundamentally different ways of modeling nature through a wave-based rather than a picture-based approach,” Wojtan sums up. “The models we present in this work might not yet be fully polished for concrete applications. However, these approaches tackle long‑standing graphics problems from a new angle while keeping computational costs low.”

 

Like a Robot Vacuum Exploring a New Room

How does a computer perceive and interpret a shape? Much like a robotic vacuum scanning a new room, a computer exploring a shape from a single reference point will seek to map all the edges it can ‘see’ to understand whether it is enclosed within the shape or if there is a way out. This approach evaluates a mathematical quantity called the ‘winding number,’ which determines whether a point lies inside or outside a shape.

In another study presented at SIGGRAPH 2026, PhD student and first author Peiyuan Xie and colleagues from the Wojtan group at ISTA developed a method to compute complex shape properties far more efficiently—reducing computational cost by an order of magnitude.

Instead of scaling with the entire surface area, the computation depends only on the boundaries—the points where surfaces end or connect.

“If the computer recognizes that the surface around the reference point is closed, calculating the shape becomes straightforward and inexpensive,” says Wojtan.

A key feature of this method is that the winding number of two adjacent surfaces that share a boundary is additive. Ultimately, the winding number of an ‘open’ shape, which would normally be expensive computationally, can therefore be calculated cheaply and quickly. To do so, the researchers construct a simple auxiliary shape that connects to the boundaries of the unknown open shape, resulting in an easy-to-compute closed shape. Subtracting the known winding number of the auxiliary shape gives the winding number of the open shape.

Generalized Winding Number algorithm developed by ISTA researchers to determine inside-outside relationships when computing shapes

 

Computing an increasingly sophisticated cat head model as fast as possible. Best performance among the exact methods: The ISTA team compares their method (violet) against others on the subdivided cat head model.

Credit

© Xie et al., ACM Transactions on Graphics

---

Journal

ACM Transactions on Graphics

DOI

10.1145/3811353 

Method of Research

Computational simulation/modeling

Subject of Research

Not applicable

Article Title

Physics-Inspired Procedural Texturing of Extremely Deformable Surfaces

Article Publication Date

23-Jul-2026

ELTE ethologists’ service robot sparks curiosity rather than fear

Do people behave differently around a robot than around a person performing the same task in everyday situations outside the laboratory? ELTE ethologists are about to find it out in a new study.

Eötvös Loránd University

Facebook

XLinkedInWeChatBlueskyMessageWhatsAppEmail

 

image: 

When a robot offers you candy, it is hard not to be curious. In an experiment conducted by ethologists at ELTE, visitors at public events noticed an autonomous service robot more often than a human server, and the candy on the robot’s tray disappeared faster. The study also suggests that the presence of robots may influence the way people behave in social situations. 

view more 

Credit: Photo: Márta Gácsi / ELTE Eötvös Loránd University

When a robot offers you candy, it is hard not to be curious. In an experiment conducted by ethologists at ELTE, visitors at public events noticed an autonomous service robot more often than a human server, and the candy on the robot’s tray disappeared faster. The study also suggests that the presence of robots may influence the way people behave in social situations. 

Imagine walking through a crowded event when a robot with free candies on its tray rolls up beside you. Would you stop? Would you smile? Or would you rather step out of its way? According to a new study published in the journal Computers in Human Behavior Reports, most people seem to behave more curiously than cautiously around certain social robots.

Researchers at ELTE Eötvös Loránd University (Hungary) and HUN-REN–ELTE Comparative Ethology Research Group investigated how visitors reacted to an autonomous service robot, Biscee, and a human, both offering candies at real-life events. The goal was simple: to find out whether people behave differently around a robot than around a person performing the same task in everyday situations outside the laboratory.

The experiment took place at public events, where the mobile non-humanoid robot, developed by the research group, moved fully autonomously through the crowd while carrying candies. At the same time, a human server (a young woman) performed the same task in the same environment, allowing the researchers to make a direct comparison.

The robot turned heads. “Visitors were more likely to observe the robot than the human server. Yet despite attracting extra attention, the robot did not make people more likely to move away or avoid it. In other words, people noticed the robot, but based on their behaviour, they seemed to be just as comfortable sharing space with it as they were with another person,” explains Melitta Csepregi, first author of the study.

Biscee remained just as interesting when it wasn't offering any candies and simply walked around with an empty tray. People looked at it, pointed at it, waved to it and touched it just as often, suggesting that the robot itself—not the sweets—was the main attraction.

The robot proved remarkably effective at its job. Its tray was emptied faster than the human server’s in every trial. Interestingly, this was not strictly due to more people stopping by. Instead, visitors often grabbed two or more candies from the robot at once—something they never did when taking candy from the human server. The researchers suggest that people may have felt less pressure to follow social norms, such as politely taking "just one," when interacting with a robot. Nonetheless, the results show that even relatively simple social robots can successfully perform routine service tasks in busy public spaces.

The researchers did not just observe people's behaviour—they also asked visitors how they felt. Here, an interesting difference emerged. Although people stepped away from Biscee just as rarely as they did from the human server, some visitors said they felt slightly uncomfortable when the robot came very close, perceiving it as somewhat pushy or intrusive. 

“This discrepancy highlights the importance of combining different methods to assess the intended users’ attitudes towards a social robot in more depth during the early development of the agent. This step is often missing, creating a disparity between users’ preferences and the marketed robotic agent,” adds Márta Gácsi, senior researcher of the project.

The study offers a glimpse into a future in which robots and humans work and live side by side. Although the transition to this future is often portrayed as—and thus, widely believed to be—a seamless process, this assumption is often far from reality. Achieving truly successful and trouble-free human–robot coexistence will require rigorous research to aid the development of our future robotic helpers.

---

Journal

Computers in Human Behavior Reports

DOI

10.1016/j.chbr.2026.101204 

Article Title

Taking candy from a robot: Attitudes toward an autonomous non-humanoid service robot at real-life events

 

Machine learning accelerates search for longer-lasting materials for solar cells

University of California - Davis

Facebook

XLinkedInWeChatBlueskyMessageWhatsAppEmail

 

image: 

Marina Leite, a UC Davis professor of Materials Science and Engineering, stands in front of rows of solar panels. Her new study focuses on perovskite solar cells, which are cheaper and more efficient but less stable than silicon-based ones.

view more 

Credit: Mario Rodriguez / UC Davis

Perovskite solar cells have gained huge momentum in the search for cheaper and more efficient solar energy. However, the materials degrade over time, limiting their widespread commercial use. 

To overcome one of the biggest barriers to commercializing perovskite solar cells, Marina Leite, a professor of materials science and engineering at the University of California, Davis, and an interdisciplinary team of researchers are harnessing AI.

In a paper published in Advanced Materials, the researchers demonstrated how AI can dramatically accelerate research advancements. Instead of relying solely on trial-and-error experimentation, the team used AI to learn from thousands of automated experiments and accurately predict how new material compositions will respond to heat, a significant environmental stressor, to identify the most promising materials more quickly. 

The search for stable perovskites 

Compared with conventional silicon solar cells, perovskites are lighter, more flexible, less expensive to manufacture and highly efficient. However, they are unstable in response to environmental stressors such as heat, moisture and light, which limits their scalability. 

“Our scientific community is very interested in understanding the chemical and physical processes that drive the stability, or lack thereof, within these materials,” Leite said. 

Testing every possible perovskite material composition under every single environmental condition would be nearly impossible. Instead, Leite’s team asked whether AI could be taught to recognize patterns and predict behavior from a carefully selected subset of experiments. 

The team tested 10 perovskite compositions by exposing them to repeated temperature cycles, resulting in 137,000 unique measurements. The measurements were used to train machine learning models that could accurately predict how previously untested material compositions would respond to repeated heating. 

From these predictions, the researchers identified which compositions were more thermally stable. Compositions with lower levels of cesium, an extremely reactive alkali metal, generally recovered after repeated heating, while compositions with higher cesium content were more likely to degrade permanently. 

AI: A research partner

The findings give researchers a roadmap for developing more durable perovskite solar cells. Instead of experimentally testing thousands of possible recipes, they can now focus their efforts on the candidates most likely to withstand real-world operating conditions.

“AI does not replace experiments in our research,” Leite said. “Instead, it can be used to increase the efficiency of scientific discovery.” 

Leite and her collaborators used AI to learn relationships from a limited set of high-throughput experiments. Using the appropriate algorithms, AI could accurately forecast the stress-response behavior of halide perovskites under previously unmeasured conditions.

Eventually, Leite hopes the same approach can be used to predict how perovskite materials will perform in different climates around the world, helping researchers design solar cells tailored to real operating conditions and paving the way for more efficient solar energy. 

“Our successful implementation of machine learning models to analyze experiments is an important demonstration of how AI can be truly helpful.”

Co-authors on the paper include Abigail Hering and Mansha Dubey, Ph.D. students in materials science and engineering at UC Davis; materials science and engineering alum Meghna Srivastava; Ph.D. student Elahe Hosseini and Professor Houman Homayoun of electrical and computer engineering at UC Davis; and Yu An and Juan-Pablo Correa-Baena from Georgia Institute of Technology. 

The project received financial support from several entities, including the Defense Advanced Research Projects Agency, National Science Foundation and the Department of Energy. 

Perovskite solar cells are cheaper and more efficient but less stable than silicon-based ones.

Credit

Mario Rodriguez/UC Davis

Journal

Advanced Materials

DOI

10.1002/adma.73636 

Article Title

High-Accuracy Machine Learning Projections of Composition-Dependent Thermal Stability in Halide Perovskites