Tech & AI

KTH har utvecklat en ny litografisk teknik för att mönstra tunna konfforma skikt direkt på komplexa 3D-strukturer, från 100 μm ned till under 30 nm. Metoden använder en scaffoldarkkitekturerad liftoff-mask (SALO) som möjliggör högupplöst patterning på böjda ytor och vinklar upp till 90 grader — en utmaning befintliga tekniker misslyckats lösa. Tekniken är agnostisk mot depositionsmetod och fungerar med atomlagrdeposition (ALD), något konventionell shadowing-baserad liftoff inte klarar. Detta öppnar väsentliga möjligheter för tillverkning av 3D MEMS-anordningar, optiska komponenter och mjukrobotik i större skala. För teknik- och produktchefer är detta relevant eftersom det löser ett fabrikationsbottleneck för strukturerade 3D-enheter. Leverantörer av depositionsutrustning kan integrera SALO-processen, och hårdvaruföretag targeting mikro- och nanoskala 3D-applikationer får tillgång till tidigare otillgängliga designmöjligheter. Kommersialisering från denna KTH-forskning kan påskynda tidslinjen för nästa generation strukturerade funktionskomponenter.

Scientists have created the first fully programmable photonic circuit that can precisely control quantum light absorption and transmission, adjusting from complete blockage to full passage with a single input. The breakthrough could accelerate development of quantum sensors, secure communication systems, and next-generation optical computers by providing a tunable building block for quantum information processing.EN

Researchers have developed a method to reverse-engineer the true costs built into automated control systems—a capability that could reshape how companies audit AI decision-making and optimize autonomous operations. The technique reconstructs entire cost structures from observable system behavior, offering transparency for industries from robotics to finance where understanding hidden incentives matters.EN

Researchers have developed simulation models that can infer clamp force in bolted joints by analyzing torque and angle data—eliminating the need for costly direct measurement during manufacturing. The breakthrough could reduce defects, improve product reliability, and cut waste across industries relying on threaded fasteners, from automotive to aerospace.EN

A new study reveals that large construction companies test digital technologies repeatedly—BIM, AI, IoT—but rarely turn pilot projects into company-wide change. The result: billions in unrealized efficiency gains and missed opportunities to compete in an increasingly digital industry.EN

Scientists have developed an optimization framework that allows autonomous systems to plan complex tasks involving multiple robots and moving obstacles more efficiently. The breakthrough could accelerate deployment of autonomous warehouses, construction sites, and delivery operations where coordinating robot movements and energy use is critical to profitability.EN

New research exposes widespread vulnerabilities in next-generation aircraft communication networks, leaving planes exposed to hacking and message tampering during takeoff, landing, and handover between ground stations. As air traffic intensifies globally, airlines and regulators face pressure to deploy unified security standards before attackers exploit the gap.EN

Researchers have developed a fully 3D-printed device that measures sweat rate noninvasively, enabling continuous hydration monitoring. The breakthrough uses novel "floating electrodes" to detect moisture levels, opening commercial potential for sports performance, workplace safety, and clinical applications where dehydration poses risks.EN

Researchers have identified viable strategies for capturing and reusing CO₂ from biomethane production—currently vented as waste—rather than releasing it to the atmosphere. The findings could unlock new revenue streams for waste-to-energy operators and accelerate Europe's shift away from fossil fuels while improving project economics.EN

Researchers have compared four competing AI systems that automatically generate mathematical equations by analyzing multiple datasets simultaneously—a technique that cuts down on overfitting and the need for huge amounts of data. The finding matters because it gives scientists and engineers a practical toolkit for reverse-engineering the rules governing physical systems, from pipeline corrosion to material behavior, faster and with greater reliability than traditional methods.EN

Researchers discovered that shared entanglement between parties doesn't automatically improve energy-constrained quantum communication unless combined with specific encoding techniques. The finding reshapes expectations for quantum networks and could accelerate deployment of practical quantum security systems that don't require expensive entanglement infrastructure.EN

Researchers have developed OASIS, a machine learning method that can extract hidden details from heavily obscured objects in scientific imaging—a capability that could accelerate breakthrough physics experiments and improve industrial quality control. By training AI to focus on the trickiest parts of overlapping images rather than treating all pixels equally, the technique enables scientists to capture faint signals previously lost to clutter.EN

A new study reveals that color—especially glowing effects—is far more effective than motion or shape at capturing player attention in 3D game environments, and players actively prefer it. For game designers and interactive software makers, the finding offers a concrete formula for guiding user behavior without relying on jarring animations or complex visual designs.EN

Researchers have extended the shelf life of gelatin-based inks used to 3D print personalized veterinary tablets, allowing the same batch to be stored and reused over two weeks. The breakthrough slashes waste and production time for custom dosing—a gap in pet medicine where one-size-fits-all tablets often don't match individual animal needs.EN

Researchers have developed a machine learning system that dramatically improves how computers reason about complex relationships in plant databases—a capability with clear applications in agriculture, pharmaceutical discovery, and climate research. The breakthrough addresses a fundamental limitation in existing AI: the inability to trace long chains of botanical connections without losing accuracy.EN

Researchers have developed a new mathematical framework that allows computers to reason about intricate networks of interconnected logical structures — a capability with implications for formal verification, cryptography, and AI safety. The advance extends a foundational language used in programming and mathematics, enabling it to handle dramatically more complex computational relationships than before.EN

Researchers have developed a method to recover valuable copper, indium, and gallium from thin-film solar cells—materials worth recovering as demand for next-generation tandem panels surges. The breakthrough uses controlled chemical leaching to separate recyclable layers, potentially creating a circular supply chain for critical elements currently scarce and expensive.EN

Researchers engineered cyanobacteria to nearly double acetate output using light and CO2—reaching 2.3 grams per liter. The breakthrough could unlock a cheaper, greener route to making industrial chemicals by replacing fossil fuel-based production with photosynthetic microbes.EN

A new study reveals that the way AI systems filter out background noise in pathology scans dramatically affects cancer diagnosis accuracy—yet this critical step is almost never reported or validated. For hospitals deploying AI tools to grade prostate cancer, the choice of filtering method could mean missed diagnoses, highlighting a gap in clinical AI oversight.EN

Researchers discovered that white blood cells naturally degrade graphene oxide—a promising material for delivering drugs inside the body—raising hopes that the technology won't accumulate as toxic waste. The finding removes a major regulatory hurdle for companies developing graphene-based medical devices and therapies.EN

Researchers have developed an artificial intelligence system that combines optical and radar satellite imagery to spot land changes with 85% accuracy, outperforming existing methods. The breakthrough matters for infrastructure monitoring, disaster response, and urban planning—sectors that depend on rapid, precise detection of environmental shifts across large areas.EN

Half of the nickel needed to power the global shift to electric vehicles and renewable energy over the next 25 years will come from mines in areas critical for biodiversity and carbon storage, a new analysis shows. The dilemma presents a stark choice for policymakers and investors: secure climate targets or preserve nature—without a clear way to do both.EN

Researchers discovered that catalytic bed material in commercial-scale biomass pyrolysis reactors degrades significantly over time, causing oil yields to drop by up to 8%. The finding matters because biomass-to-fuel plants operate continuously—understanding how quickly equipment loses productivity could reshape maintenance schedules and production planning across the emerging bioenergy sector.EN

Researchers have developed ConvReflex, a method that reduces energy consumption in AI inference tasks by predicting and clamping neural network outputs. The technique could extend battery life in edge devices and lower operating costs for data centers running machine learning models at scale.EN

Researchers developed a machine learning system that identifies building structures from visible features like shape and age, filling critical gaps in material inventory data. The breakthrough could help cities and construction firms track embodied resources in existing buildings, essential for waste reduction and circular economy strategies.EN