How AI is Changing Materials Design

Crashing metals with AI looks like entertainment. It’s actually a glimpse into the future of engineering. Fascinating? Every collision is data. In materials science, traditional testing is slow and expensive. Physical crash tests can cost thousands per run and explore only a narrow set of scenarios. AI flips that model: • Millions of virtual crash scenarios simulated in hours • Up to 70–90% reduction in physical prototyping cycles (industry benchmarks) • Design iteration speeds improved by 10–100x in early-stage engineering What does this teach us? 1. Failure is the fastest teacher AI doesn’t avoid failure—it hunts for it. Microfractures, stress concentrations, fatigue points… mapped instantly. 2. Strength is no longer about mass Using generative design, AI creates lighter structures that outperform heavier ones—often cutting weight by 20–50% while maintaining or improving strength. 3. Edge cases become the norm Humans test what they expect. AI tests what we miss. Rare, high-impact failure scenarios become visible before they happen in the real world. 4. Simulation becomes reality’s twin With high-fidelity physics engines, virtual models now predict real-world behavior with increasing accuracy—closing the gap between digital and physical. 5. Speed is the real disruption The winner isn’t who builds best. It’s who learns fastest from failure. This isn’t about watching metal break. It’s about teaching machines the physics of the world—and using that to build safer cars, stronger infrastructure, and entirely new classes of products. The companies that master this won’t just design better things. They’ll compress decades of learning into months. #AI #Engineering #DeepTech #Innovation #Simulation #FutureOfWork

How do materials fail, and how can we design stronger, tougher, and more resilient ones? Published in #PNAS, our physics-aware AI model integrates advanced reasoning, rational thinking, and strategic planning capabilities models with the ability to write and execute code, perform atomistic simulations to solicit new physics data from “first principles”, and conduct visual analysis of graphed results and molecular mechanisms. By employing a multiagent strategy, these capabilities are combined into an intelligent system designed to solve complex scientific analysis and design tasks, as applied here to alloy design and discovery. This is significant because our model overcomes the limitations of traditional data-driven approaches by integrating diverse AI capabilities—reasoning, simulations, and multimodal analysis—into a collaborative system, enabling autonomous, adaptive, and efficient solutions to complex, multiobjective materials design problems that were previously slow, expert-dependent, and domain-specific. Wonderful work by my postdoc Alireza Ghafarollahi! Background: The design of new alloys is a multiscale problem that requires a holistic approach that involves retrieving relevant knowledge, applying advanced computational methods, conducting experimental validations, and analyzing the results, a process that is typically slow and reserved for human experts. Machine learning can help accelerate this process, for instance, through the use of deep surrogate models that connect structural and chemical features to material properties, or vice versa. However, existing data-driven models often target specific material objectives, offering limited flexibility to integrate out-of-domain knowledge and cannot adapt to new, unforeseen challenges. Our model overcomes these limitations by leveraging the distinct capabilities of multiple AI agents that collaborate autonomously within a dynamic environment to solve complex materials design tasks. The proposed physics-aware generative AI platform, AtomAgents, synergizes the intelligence of LLMs and the dynamic collaboration among AI agents with expertise in various domains, incl. knowledge retrieval, multimodal data integration, physics-based simulations, and comprehensive results analysis across modalities. The concerted effort of the multiagent system allows for addressing complex materials design problems, as demonstrated by examples that include autonomously designing metallic alloys with enhanced properties compared to their pure counterparts. We demonstrate accurate prediction of key characteristics across alloys and highlight the crucial role of solid solution alloying to steer the development of alloys. Paper: https://lnkd.in/enusweMf Code: https://lnkd.in/eWv2eKwS MIT Schwarzman College of Computing MIT Civil and Environmental Engineering MIT Department of Mechanical Engineering (MechE) MIT Industrial Liaison Program MIT School of Engineering

AI is inventing new materials. Lots of people are sharing this here on LinkedIn. That framing is seductive. It’s also misleading. Profoundly misleading. AI is not “discovering” matter in the way we discovered electricity or antibiotics. What it’s really doing is exposing how narrow human intuition has been in materials science. For decades, progress in materials followed a familiar loop: theory → small experiments → incremental improvement → scale (if we’re lucky). That wasn’t because nature was exhausted. It was because human imagination was. AI changes this not by being creative, but by being exhaustive. It doesn’t ask, “What material makes sense?” It asks, “What configurations are mathematically and physically allowed?” That distinction matters. Most of the “stronger than steel, lighter than foam” headlines aren’t breakthroughs in chemistry. They’re breakthroughs in structure — lattices, periodic geometries, hierarchical designs — things humans historically ignored because they were impossible to reason about, manufacture, or even visualize at scale. AI didn’t invent these regions of the design space. They were always there. We just couldn’t reach them. And here’s the less comfortable implication: AI is quietly removing human taste from engineering. When a material is selected because it performs better — but no engineer can explain why it looks the way it does — authority shifts. From intuition to validation. From explanation to measurement. From “this makes sense” to “this works.” That’s not just a technical change. It’s a cultural one. Entire industries are built on tacit knowledge: senior engineers, rules of thumb, aesthetic judgments about what feels “robust” or “elegant.” AI doesn’t care. It proposes designs that look fragile, alien, even wrong — and then dares humans to disprove them experimentally. So the disruption isn’t that AI will replace materials scientists. It’s that it will redefine expertise. The best engineers of the next decade won’t be the ones who know the materials. They’ll be the ones who know how to interrogate AI-generated designs, stress them, break them, and decide when not to trust them. AI isn’t inventing matter. It’s forcing humans to confront the limits of their own intuition about the physical world. And that may be the more profound shift.

Earlier I talked about AI exposing that batteries fail because plastics lose order... not because metals degrade. But now this is the next, more uncomfortable step. AI-designed polymers are now breaking the “explainability contract” in materials science. In polymer electrolytes and separators, AI models are generating materials that outperform human-designed systems, and no one can fully explain why. Researchers can verify the results: 📍ionic conductivity improves, mechanical stability holds, cycle life extends. But the usual structure–property logic breaks down. What’s actually emerging: 📍Non-intuitive side-chain interactions that enable ion hopping 📍AI-discovered free-volume networks no polymer chemist would sketch 📍Emergent ion-transport pathways that don’t map cleanly to textbook models Now, this flips a century-old norm: materials science assumes understanding precedes scaling. AI is doing the opposite... scaling first, understanding later. That creates real tension for senior scientists: 📍How do you peer-review something you can’t rationalize? 📍How do you protect IP you can’t explain? 📍What happens to “expert intuition” when the model consistently beats it? AI is designing materials... yes, but a lot more is also happening. It’s quietly rewriting how we decide what counts as scientific knowledge, and we’re not ready for that shift yet. #MaterialScience #Innovations #AIinMaterialScience

Chemists + AI = Smarter Materials Discovery A fascinating advancement in the field of materials science is emerging from a collaboration between University of North Carolina at Chapel Hill and Carnegie Mellon University, where researchers have developed a human-in-the-loop artificial intelligence system that significantly accelerates the discovery of new materials — particularly polymers with complex, high-performance properties. Traditionally, designing new materials has required a time- and resource-intensive process of trial and error. Chemists often face trade-offs: enhancing one property (like elasticity) may compromise another (like durability). The new approach, detailed in the article below, leverages AI not as a replacement for human insight, but as a collaborative partner. This human-AI co-design loop works as follows: 1️⃣ A machine learning model analyzes data and proposes promising material candidates. 2️⃣ Chemists synthesize and test these materials in the lab. 3️⃣ Results are fed back into the model, refining future predictions. This tight integration enables the system to efficiently search vast chemical design spaces, identifying materials that meet multiple, sometimes competing performance criteria — such as strength, flexibility, and thermal resistance — all at once. Key highlights: ✅ The AI model rapidly eliminates low-potential candidates before synthesis, reducing experimental burden. ✅ The framework is open-source and adaptable, making it accessible to research labs and industry partners globally. ✅ Early applications include advanced consumer products, medical devices, aerospace materials, and sustainable polymers. This work illustrates the power of interdisciplinary innovation, where chemistry, data science, and machine learning converge to solve some of the most complex challenges in materials design. https://lnkd.in/eFhs76Qy Frank Leibfarth, Olexandr Isayev, Kirsten Heuring, Ralph House, David DeFusco, Amie Solosky, UNC Research

It's Manuscript Monday! What: This week we highlight a review paper surveying how AI is being applied to material development for additive manufacturing. The scope covers metals, polymers, and bioinks/biomaterial inks, organized around two themes: AI-driven material design (composition optimization, phase prediction, metamaterial architecture) and AI-driven performance optimization (linking process parameters to microstructure and properties). The review also lists online materials databases, data preprocessing strategies, and commonly deployed AI methods. Most coverage on AI in the literature has focused on topics of in-situ monitoring and defect detection. The authors here have provided a great roundup of how AI accelerates the discovery and tailoring of materials themselves. Why this is important: For those in the ICME community, it is worth considering how these AI approaches differ from traditional integrated computational materials engineering. ICME chains physics-based models across length scales with each link mechanistically grounded. Many studies highlighted here use ML to learn direct mappings from composition and process parameters to final properties, bypassing intermediate microstructural modeling entirely. This sacrifices interpretability but gains tremendous speed and, critically, enables inverse design as a native capability. This is one of those headaches of AI, it can be a black box. Rather than running the ICME chain forward and iterating, Bayesian optimization and generative models let you define target properties and search backward through the design space. The most compelling examples in the review are hybrid. One highlighted study used CALPHAD-generated thermodynamic data as ML training input for maraging steel composition optimization. Another integrated electronic structure calculations with data-driven fatigue models. These sit at the intersection of ICME and AI rather than replacing one with the other. The future likely demands this convergence, and contextualizing both AI methods and ICME would strengthen the community dialogue about where each approach excels and where they complement each other. The review is also a useful resource for its compilation of open-access and licensed materials databases, though the authors rightly note most were built for conventional manufacturing and lack AM-specific process data (c'mon standards, this is your time to shine). Closing that infrastructure gap remains essential for the field to mature. Excellent work by Peijie Shangguan, @Huifei Zhou, Xi Huang, Jinlong Su, Wai Yee Yeong, and Swee Leong Sing. At the time of publication, researchers affiliations were either at National University of Singapore or Nanyang Technological University Singapore. https://lnkd.in/gV4iNAkj #additivemanufacturing #AI #materials #artificialintelligence #machinelearning #ICME As always, help pointing to the profiles of authors for proper attribution is always much appreciated!

Designing next-generation materials often runs into the same obstacle! 𝗪𝗲 𝗻𝗲𝗲𝗱 𝗮𝘁𝗼𝗺𝗶𝘀𝘁𝗶𝗰 𝗮𝗰𝗰𝘂𝗿𝗮𝗰𝘆 𝗮𝘁 𝘀𝗰𝗮𝗹𝗲𝘀 𝘄𝗵𝗲𝗿𝗲 𝘁𝗿𝗮𝗱𝗶𝘁𝗶𝗼𝗻𝗮𝗹 𝘀𝗶𝗺𝘂𝗹𝗮𝘁𝗶𝗼𝗻𝘀 𝘀𝗶𝗺𝗽𝗹𝘆 𝗰𝗮𝗻𝗻𝗼𝘁 𝗿𝗲𝗮𝗰𝗵. That’s the challenge researchers faced when studying complex materials systems such as molten salts and catalytic environments. Understanding 𝘀𝘁𝗿𝘂𝗰𝘁𝘂𝗿𝗲, 𝘁𝗵𝗲𝗿𝗺𝗼𝗱𝘆𝗻𝗮𝗺𝗶𝗰𝘀, 𝗮𝗻𝗱 𝘁𝗿𝗮𝗻𝘀𝗽𝗼𝗿𝘁 𝗽𝗿𝗼𝗽𝗲𝗿𝘁𝗶𝗲𝘀 𝗮𝘁 𝗵𝗶𝗴𝗵 𝘁𝗲𝗺𝗽𝗲𝗿𝗮𝘁𝘂𝗿𝗲𝘀 requires simulations spanning thousands of atoms and long time scales - well beyond the reach of conventional DFT. In this recent study, the team set out to solve that problem. The approach combined 𝗗𝗙𝗧-𝘁𝗿𝗮𝗶𝗻𝗲𝗱 𝗻𝗲𝘂𝗿𝗮𝗹 𝗻𝗲𝘁𝘄𝗼𝗿𝗸 𝗽𝗼𝘁𝗲𝗻𝘁𝗶𝗮𝗹𝘀 𝘄𝗶𝘁𝗵 𝗹𝗮𝗿𝗴𝗲-𝘀𝗰𝗮𝗹𝗲 𝗺𝗼𝗹𝗲𝗰𝘂𝗹𝗮𝗿 𝗱𝘆𝗻𝗮𝗺𝗶𝗰𝘀, enabling simulations that capture atomic interactions while scaling to realistic system sizes. But even more importantly, the researchers demonstrated that 𝗠𝗮𝘁𝗹𝗮𝗻𝘁𝗶𝘀, powered by the 𝗣𝗙𝗣 𝘂𝗻𝗶𝘃𝗲𝗿𝘀𝗮𝗹 𝗺𝗮𝗰𝗵𝗶𝗻𝗲-𝗹𝗲𝗮𝗿𝗻𝗶𝗻𝗴 𝗶𝗻𝘁𝗲𝗿𝗮𝘁𝗼𝗺𝗶𝗰 𝗽𝗼𝘁𝗲𝗻𝘁𝗶𝗮𝗹, can serve as an efficient alternative for exploring complex chemical systems without the need to develop system-specific potentials. With Matlantis, they were able to reproduce the 𝘀𝘁𝗿𝘂𝗰𝘁𝘂𝗿𝗮𝗹 𝗰𝗼𝗿𝗿𝗲𝗹𝗮𝘁𝗶𝗼𝗻𝘀, 𝘁𝗵𝗲𝗿𝗺𝗼𝗱𝘆𝗻𝗮𝗺𝗶𝗰 𝗯𝗲𝗵𝗮𝘃𝗶𝗼𝗿, 𝗮𝗻𝗱 𝘁𝗿𝗮𝗻𝘀𝗽𝗼𝗿𝘁 𝗽𝗿𝗼𝗽𝗲𝗿𝘁𝗶𝗲𝘀 of high-temperature materials with strong agreement with ab initio simulations. The result is more than a simulation benchmark. It shows how 𝗺𝗮𝗰𝗵𝗶𝗻𝗲-𝗹𝗲𝗮𝗿𝗻𝗶𝗻𝗴-𝗯𝗮𝘀𝗲𝗱 𝗮𝘁𝗼𝗺𝗶𝘀𝘁𝗶𝗰 𝗺𝗼𝗱𝗲𝗹𝗶𝗻𝗴 𝗰𝗮𝗻 𝗯𝗿𝗶𝗱𝗴𝗲 𝘁𝗵𝗲 𝗴𝗮𝗽 𝗯𝗲𝘁𝘄𝗲𝗲𝗻 𝘁𝗵𝗲 𝗾𝘂𝗮𝗻𝘁𝘂𝗺-𝗹𝗲𝘃𝗲𝗹 𝗼𝗳 𝗮𝗰𝗰𝘂𝗿𝗮𝗰𝘆 𝗮𝗻𝗱 𝗶𝗻𝗱𝘂𝘀𝘁𝗿𝗶𝗮𝗹𝗹𝘆 𝗿𝗲𝗹𝗲𝘃𝗮𝗻𝘁 𝘀𝗰𝗮𝗹𝗲𝘀. Turning atomistic insight into practical guidance for designing new electrolytes, catalysts, and functional materials. For researchers and scientists working on complex materials, this is where 𝗔𝗜-𝗱𝗿𝗶𝘃𝗲𝗻 𝘀𝗶𝗺𝘂𝗹𝗮𝘁𝗶𝗼𝗻 𝗯𝗲𝗴𝗶𝗻𝘀 𝘁𝗼 𝗿𝗲𝘀𝗵𝗮𝗽𝗲 𝗺𝗮𝘁𝗲𝗿𝗶𝗮𝗹𝘀 𝗱𝗶𝘀𝗰𝗼𝘃𝗲𝗿𝘆. Read the full paper: https://lnkd.in/dqgxjiyv

Headline: AI Is Creating the Next Great Materials—Not Finding Them Introduction: A revolution is underway in materials science. Instead of searching the natural world for new substances, scientists are computing them. Artificial intelligence is transforming how materials are conceived, tested, and brought to life—accelerating discoveries once thought to take decades into just weeks. From Prediction to Production: Massive Discovery Leap: Google DeepMind’s GNoME AI predicted 2.2 million new crystal structures, including 380,000 stable compounds, among them thousands of potential graphene-like materials and battery electrolytes. Autonomous Labs in Action: In one demonstration, a self-driving lab synthesized 41 new compounds in 17 days, all first identified by AI—condensing years of trial-and-error into days. Shift in Workflow: Traditional R&D relied on luck and labor. Now, algorithms preselect the most promising candidates, while human engineers validate, refine, and scale what the computer imagines. Tools Powering the New Alchemy: Bayesian Optimization: AI actively decides which experiment to run next, dramatically cutting time and cost. One alloy study found breakthroughs in 36 trials instead of 800,000. Graph Neural Networks (GNNs): These AIs map atoms and bonds as networks, predicting material properties with striking accuracy. DeepMind’s models used GNNs to identify hundreds of lab-verified crystals. Generative Models: IBM and others now deploy foundation AIs trained on billions of molecular structures, capable of inventing entirely new materials for semiconductors, batteries, and composites. In Practice: Citrine Informatics helped aerospace engineers create AL 7A77, the first 3D-printable aluminum alloy certified for flight. Argonne National Lab’s “Polybot” autonomously produced defect-free conductive polymers using AI-guided experimentation—achieving world-class results. Berkeley Lab and DeepMind have already synthesized hundreds of GNoME’s AI-predicted materials, confirming theory with reality. The Next Frontier: While AI accelerates discovery, scaling up from micrograms to manufacturable tons remains the bottleneck. The future lies in AI-designed materials “born ready” for industrial production, integrating manufacturability, sustainability, and performance into their digital blueprints. Why It Matters: The fusion of AI, robotics, and materials science marks a turning point in human innovation. We are compressing millennia of discovery into months, opening the door to breakthroughs in energy storage, electronics, aerospace, and quantum computing. The next wonder material won’t be mined—it will be computed, tested by robots, and engineered for the world. I share daily insights with 29,000+ followers and 10,000+ professional contacts across defense, tech, and policy. If this topic resonates, I invite you to connect and continue the conversation. Keith King https://lnkd.in/gHPvUttw

Turns out, the AI models that turn text into images, are also useful for generating new materials. After 10 years of searching for quantum spin liquids, researchers had found only 12 candidates. Then MIT's team built SCIGEN. 10 million new material candidates with exotic quantum properties. Two already synthesized and confirmed in the lab. "We don't need 10 million new materials to change the world. We just need one really good material," says MIT's Mingda Li. Every AI model brags about scale. More parameters. More data. More compute. MIT tried a different direction. They added constraints to force AI models to ignore stable, boring materials. To focus only on structures with quantum potential. Kagome lattices. Archimedean patterns. The geometric shapes that create superconductivity and magnetic states we've never seen before. These new materials can mimic rare earth elements without the supply chain risks. Without the geopolitical tensions. Without the environmental damage. China controls 90% of rare earth processing. Every tech company is scrambling for alternatives. MIT just generated a million of them. The traditional materials discovery process takes decades. Testing. Failure. Incremental progress. Academic papers that lead nowhere. SCIGEN gives scientists 26,000 materials to test instead of 12. 41% showed magnetic properties the model predicted correctly. Two have already been synthesized: TiPdBi and TiPbSb. Names only a physicist could love. But properties that could unlock quantum computing at scale. This is what AI is best at. Large scale simulations and testing. *Image: Jose-Luis Olivares, MIT; iStock