Researchers in Sweden have developed a machine-learning approach that embeds the laws of physics directly into neural networks.
A new study from Chalmers University of Technology in Sweden shows that machine learning can become far more efficient when it starts with a built-in understanding of the laws of physics. Researchers found that giving an AI system this foundational knowledge dramatically reduced the time needed to develop advanced optical components used in technologies ranging from quantum computers to camera and eyeglass lenses.
“When we fed the super-brain information about the laws of physics, it immediately got much smarter. Our calculations now take one tenth of the time previously required,” said Philippe Tassin, a professor in the Department of Physics and Astronomy at Chalmers University of Technology.
Tassin’s team works in nanophotonics, a field focused on controlling light at extremely small scales. When light interacts with structures smaller than its wavelength, it can behave very differently than it does on larger scales. However, natural optical materials have limits that restrict how light can be manipulated. To overcome those constraints, the researchers use computer simulations to design artificial optical materials.
These engineered materials could lead to lighter, thinner, and more effective camera and eyeglass lenses. The research may also support future quantum computing technologies. Working with scientists from Chalmers’ Department of Microtechnology and Nanoscience, where Sweden’s first large-scale quantum computer is under development, the team is exploring nanostructured materials that can precisely control the movement of light.
One potential application involves transmitting information between quantum computers, or across longer distances, using optical frequencies and mechanically compliant photonic crystals. These specially designed crystals can reflect light with extremely high efficiency.
Simulations show how to design the material optimally
The researchers rely entirely on supercomputer simulations, using machine learning and neural networks to analyze how different materials behave. These tools help identify material properties and guide the design process.
“I know electromagnetism’s equations inside out and I teach them, but I still can’t draw all the conclusions that the neural network can. The physics is so complex that I don’t understand the properties of a material just by looking at it – but the computer does,” says Philippe Tassin.
Time-consuming to feed data into neural networks
Training neural networks for these simulations has traditionally required enormous amounts of data. Creating a single data point can take anywhere from ten minutes to an hour, and researchers may need as many as 40,000 simulations.
“It might take us a whole month to generate enough data to train the neural network. Then if you realize that you need to add more things, it can take another month,” said Viktor Lilja, a doctoral student in the Department of Physics and Astronomy at Chalmers University of Technology.
The team has now cut that process to about one tenth of the original time. Tasks that once required 30 days can now be completed in roughly three days because the neural network already understands key physical principles before training begins.
Teaching the neural network the laws of physics
The researchers recognized that optical components must always follow the laws of physics and electromagnetism. Instead of forcing the neural network to discover those rules from training data alone, they incorporated the laws directly into the system.
As a result, the AI no longer has to relearn the same physical relationships from scratch each time. The approach emerged while the researchers were trying to make the network’s predictions easier for humans to interpret by embedding familiar equations into the model. During testing, they found that the network also became significantly more capable and required much less training data. The work was described in the journal Laser & Photonics Reviews.
“Once we’d trained the network, we could ask it to examine any structure at all and get the optical properties in a millisecond. With these new networks, we get better estimates and avoid obvious errors,” Lilja said.
For Tassin, the greatest advantage is the time saved.
“Now that we can work so much faster, we can speed up design development for optical components.”
Reference: “A General Framework for Knowledge Integration in Machine Learning for Electromagnetic Scattering Using Quasinormal Modes” by Viktor A. Lilja, Albin J. Svärdsby, Timo Gahlmann and Philippe Tassin, 17 March 2026, Laser & Photonics Reviews.
DOI: 10.1002/lpor.202502769
The research was funded by the Chalmers Nano Area of Advance, the Swedish Research Council, and the Knut and Alice Wallenberg Foundation. The training of the neural network was carried out using resources provided by the Swedish National Infrastructure for Computing (NAISS) at Chalmers/C3SE and KTH/PDC, in part with funding from the Swedish Research Council. The work was carried out in part within the META-PIX competence centre at Chalmers.
Never miss a breakthrough: Join the SciTechDaily newsletter.
Follow us on Google and Google News.
7 Comments
I already have AI compatible with scientific relativity laws built in to its base of information , In fact I can ask it anything about anything and its information well has it all , I call it Watson because there can be more than one AI named Watson , you can’t patent a name that is used world wide .
thanks for this
The first nuclei imagined have devolved into utter hear say – where every kid’s repeated misunderstandings have been amplified by 10 million and where that is what AI has prodigiously sorted out?
I mean, exactly where is that crafty electron hiding?
hint2: accelerator output destroyed rest-state input, 100%, and all figures attributed are annihilation break-points; and not viable working parameters. You know, that place where the item in question was no longer in service. No longer the animal sought – and endlessly written about
VERY GOOD! Now that we can work so much faster, we can speed up design development for optical components.
However, it is ridiculous that many so-called peer-reviewed publications are still stuck in primitive societies.
The so-called peer-reviewed publications—including Physical Review Letters, Nature, Science, and Nature Physics—have never earnestly reflected on a fundamental physical question: Where do the things in space come from? Do they arise from the dynamic evolution of space itself, or are they placed there from the outside by God, devils, or angels? This qualitative inquiry determines the very starting point of all cosmology, yet it has been systematically suspended by these outlets. In practice, these publications echo and shield one another. They stubbornly cling to and loudly trumpet that two sets of cobalt-60 artificially rotated in opposite directions, regardless of whether the procedure is truly symmetric, are unproblematically treated as two objects that are mirror images of each other. In the physical world they have constructed, a topological vortex and its twin anti-vortex can even be defined as two vortices possessing entirely different spacetime manifolds (as shown in https://pic2.zhimg.com/v2-4127b0b58fe8b88feb27c189fb705029_1440w.jpg?source=172ae18b ), conveniently ignoring that such a definition already presupposes the qualitative arbitrariness that spacetime can be segmented at will [32, 33]. Moreover, they brazenly presuppose pseudoscientific premises (such as CP violation) as self-evident axioms, and on this basis forcibly define two manifestly different particles as one and the same particle—witness the historical conundrum of the θ and τ particles—using post-hoc quantitative patches to conceal a fundamental qualitative fallacy. God, devils, angels, and their pet cats have thus come to preside, in an invisible yet all-pervasive manner, over the much-celebrated physical world of these so-called peer-reviewed publications.
—— https://zhuanlan.zhihu.com/p/2050535928066838797.
The Fluidized Absolute Space Theory (FAST) and generalized Topological Vortex Theory (TVT) constitute a systematic and grand picture of physical reality. Its core contribution lies not in proposing new quantum mechanical equations, but in the profound “supplement” it provides to the foundations of quantum mechanics. By directly confronting the chronic problems of traditional interpretations and by reducing abstract mathematical symbols to intuitive topological processes of spacetime—the phase, winding, and stability of vortices—it fills the long-missing physical-ontological gap for core concepts such as wave functions, operators, spin, entanglement, and measurement collapse [18, 19]. This endeavor pulls the discussion of quantum mechanics back into the classical realist tradition [20, 21], striving to offer a self-consistent picture of the universe independent of the observer.
Despite the mainstream physics community and its captive so-called “peer-reviewed” publications stubbornly entrenching the foundations of quantum mechanics as an insoluble mystery, clinging to the dogma of CP violation, blatantly spreading that topological vortices and their twin anti-vortices are inherently asymmetric [16], and absurdly equating two artificially prepared, counter-rotating cobalt-60 systems as perfect mirror-image objects—regardless of their actual preparation processes [17]—thereby obstinately refusing to give any serious scrutiny to the paradigm-shattering generalized Topological Vortex Theory, the formidable explanatory power and undeniable physical reality of the TVT have already become manifest.
Throughout the history of physics, every epoch-making paradigm revolution has never been born within the halls of the so-called “mainstream,” let alone been midwived or sanctioned by the “peer-reviewed” publications controlled by academic gatekeepers. On the contrary, these revolutions are invariably a complete overthrow of mainstream authority—isolated insights piercing through the barriers of collective conformity.
The generalized Topological Vortex Theory unequivocally responds to the deepest call of physics: to find a comprehensible physical essence for a powerful computational tool, and to construct a unified stage for the four fundamental forces of nature. From the vortices of Descartes to Lord Kelvin’s vortex atoms [9], and onward to contemporary topological order [8], the approach of using vortices and topology to understand matter and interactions has always shone throughout the history of thought. The generalized Topological Vortex Theory inherits this grand tradition and systematically modernizes it in an unprecedented way [10, 11, 12], injecting new thought into solving the century-long problem of quantum gravity [13, 14, 15] and providing an immensely inspiring blueprint for our understanding of the ultimate weave of the universe.
This paper has argued that starting from the three postulates of fluidized absolute space—“zero viscosity, zero compressibility, and zero anisotropy”—the entire empirical content of general relativity and the Standard Model can be derived rigorously and uniquely via topological vortex structures. Spacetime curvature becomes an apparent effect of fluid inhomogeneities, fundamental particles become quantized excitations of topological vortices, gauge interactions reflect the topological rules of vortex reconnection, and masses and mixing angles are determined by the geometry of knots and rings. In this way, known physical laws are no longer fundamental axioms, but the necessary emergent behavior of a deeper fluid ontology in the low-energy limit. FAST thus not only bridges the rift between relativity and quantum theory, but also ushers physics into a new era of parsimony in which “What is space?” stands as the sole central question.
—— https://zhuanlan.zhihu.com/p/2050535928066838797.
Why doesn’t science use ‘back and forth, up and down, right and left instead of xyz – just like when mommy first told them when they were kids?
Because scientists can’t get a ridiculous idea out of their minds.
People try to think about something using ‘ideas,’ little picto-grams and squiggy drawings in their head – when those are not the thing they are trying to understand. And never will be. But does that give anyone pause? no.