{"683111":{"#nid":"683111","#data":{"type":"news","title":"Study: New AI Tool Deciphers Mysteries of Nanoparticle Motion in Liquid Environments ","body":[{"value":"\u003Cdiv\u003E\u003Cdiv\u003E\u003Cdiv\u003E\u003Cdiv\u003E\u003Cdiv\u003E\u003Cp\u003ENanoparticles \u2013 the tiniest building blocks of our world \u2013 are constantly in motion, bouncing, shifting, and drifting in unpredictable paths shaped by invisible forces and random environmental fluctuations.\u0026nbsp;\u003C\/p\u003E\u003Cp\u003EBetter understanding their movements is key to developing better medicines, materials, and sensors. But observing and interpreting their motion at the atomic scale has presented scientists with major challenges.\u003C\/p\u003E\u003Cp\u003EHowever, researchers in Georgia Tech\u2019s School of Chemical and Biomolecular Engineering (ChBE) have developed an artificial intelligence (AI) model that learns the underlying physics governing those movements.\u0026nbsp;\u003C\/p\u003E\u003Cp\u003EThe team\u2019s research, \u003Ca href=\u0022https:\/\/www.nature.com\/articles\/s41467-025-61632-1\u0022\u003Epublished\u003C\/a\u003E in \u003Cem\u003ENature Communications\u003C\/em\u003E, enables scientists to not only analyze, but also generate realistic nanoparticle motion trajectories that are indistinguishable from real experiments, based on thousands of experimental recordings.\u003C\/p\u003E\u003C\/div\u003E\u003C\/div\u003E\u003C\/div\u003E\u003C\/div\u003E\u003C\/div\u003E\u003Cdiv\u003E\u003Cdiv\u003E\u003Cdiv\u003E\u003Cdiv\u003E\u003Cdiv\u003E\u003Cp\u003E\u003Cstrong\u003EA Clearer Window into the Nanoworld\u003C\/strong\u003E\u003C\/p\u003E\u003Cp\u003EConventional microscopes, even extremely powerful ones, struggle to observe moving nanoparticles in fluids. And traditional physics-based models, such as Brownian motion, often fail to fully capture the complexity of unpredictable nanoparticle movements, which can be influenced by factors such as viscoelastic fluids, energy barriers, or surface interactions.\u003C\/p\u003E\u003Cp\u003ETo overcome these obstacles, the researchers developed a deep generative model (called LEONARDO) that can analyze and simulate the motion of nanoparticles captured by liquid-phase transmission electron microscopy (LPTEM), allowing scientists to better understand nanoscale interactions invisible to the naked eye. Unlike traditional imaging, LPTEM can observe particles as they move naturally within a microfluidic chamber, capturing motion down to the nanometer and millisecond.\u003C\/p\u003E\u003Cp\u003E\u201cLEONARDO allows us to move beyond observation to simulation,\u201d said \u003Ca href=\u0022https:\/\/vidajamali.github.io\/\u0022\u003EVida Jamali\u003C\/a\u003E, assistant professor and Daniel B. Mowrey Faculty Fellow in ChBE@GT. \u201cWe can now generate high-fidelity models of nanoscale motion that reflect the actual physical forces at play.\u0026nbsp;LEONARDO helps us not only see what is happening at the nanoscale but also understand why.\u201d\u003C\/p\u003E\u003Cp\u003ETo train and test LEONARDO, the researchers used a model system of gold nanorods diffusing in water. They collected more than 38,000 short trajectories under various experimental conditions, including different particle sizes, frame rates, and electron beam settings. This diversity allowed the model to generalize across a broad range of behaviors and conditions.\u0026nbsp;\u003C\/p\u003E\u003C\/div\u003E\u003C\/div\u003E\u003C\/div\u003E\u003C\/div\u003E\u003C\/div\u003E\u003Cdiv\u003E\u003Cdiv\u003E\u003Cdiv\u003E\u003Cp\u003E\u003Cstrong\u003EThe Power of LEONARDO\u2019s Generative AI\u003C\/strong\u003E\u003C\/p\u003E\u003Cp\u003EWhat distinguishes LEONARDO is its ability to learn from experimental data while being guided by physical principles, said study lead author Zain Shabeeb, a PhD student in ChBE@GT. LEONARDO uses a specialized \u201closs function\u201d based on known laws of physics to ensure that its predictions remain grounded in reality, even when the observed behavior is highly complex or random.\u003C\/p\u003E\u003Cp\u003E\u201cMany machine learning models are like black boxes in that they make predictions, but we don\u2019t always know why,\u201d Shabeeb said. \u201cWith LEONARDO, we integrated physical laws directly into the learning process so that the model\u2019s outputs remain interpretable and physically meaningful.\u201d\u003C\/p\u003E\u003Cp\u003ELEONARDO uses a transformer-based architecture, which is the same kind of model behind many modern language AI applications. Like how a language model learns grammar and syntax, LEONARDO learns the \u0022grammar\u0022 of nanoparticle movement, identifying hidden reasons for the ways nanoparticles interact with their environment.\u003C\/p\u003E\u003Cp\u003E\u003Cstrong\u003EFuture Impact\u003C\/strong\u003E\u003C\/p\u003E\u003Cp\u003EBy simulating vast libraries of possible nanoparticle motions, LEONARDO could help train AI systems that automatically control and adjust electron microscopes for optimal imaging, paving the way for \u201csmart\u201d microscopes that adapt in real time, the researchers said.\u003C\/p\u003E\u003Cp\u003E\u201cUnderstanding nanoscale motion is of growing importance to many fields, including drug delivery, nanomedicine, polymer science, and quantum technologies,\u201d Jamali said. \u201cBy making it easier to interpret particle behavior, LEONARDO could help scientists design better materials, improve targeted therapies, and uncover new fundamental insights into how matter behaves at small scales.\u0022\u003C\/p\u003E\u003Cp\u003ECITATION: Zain Shabeeb , Naisargi Goyal, Pagnaa Attah Nantogmah, and Vida Jamali, \u201c\u003Ca href=\u0022https:\/\/www.nature.com\/articles\/s41467-025-61632-1\u0022\u003ELearning the diffusion of nanoparticles in liquid phase TEM via physics-informed generative AI\u003C\/a\u003E,\u201d \u003Cem\u003ENature Communications\u003C\/em\u003E, 2025.\u003C\/p\u003E\u003C\/div\u003E\u003C\/div\u003E\u003C\/div\u003E","summary":"","format":"limited_html"}],"field_subtitle":"","field_summary":[{"value":"\u003Cp\u003EResearchers at Georgia Tech\u2019s School of Chemical and Biomolecular Engineering have developed an AI model that uncovers the hidden physics behind the motion of nanoparticles\u2014tiny particles constantly influenced by random forces. Understanding their movement is critical for advancing drug delivery, materials, and sensing technologies\u003C\/p\u003E","format":"limited_html"}],"field_summary_sentence":[{"value":"Researchers have developed an AI model that learns the underlying physics governing movements of nanoparticles"}],"uid":"27271","created_gmt":"2025-07-11 20:06:00","changed_gmt":"2025-07-14 19:30:46","author":"Brad Dixon","boilerplate_text":"","field_publication":"","field_article_url":"","location":"Atlanta, GA","dateline":{"date":"2025-07-11T00:00:00-04:00","iso_date":"2025-07-11T00:00:00-04:00","tz":"America\/New_York"},"extras":[],"hg_media":{"677402":{"id":"677402","type":"image","title":"nanoparticles.jpeg","body":"\u003Cp\u003ESchematic showing nanoparticles in the microfluidic chamber of liquid-phase transmission electron microscopy\u003C\/p\u003E","created":"1752264372","gmt_created":"2025-07-11 20:06:12","changed":"1752264372","gmt_changed":"2025-07-11 20:06:12","alt":"Schematic showing nanoparticles in the microfluidic chamber of liquid-phase transmission electron microscopy","file":{"fid":"261297","name":"nanoparticles.jpeg","image_path":"\/sites\/default\/files\/2025\/07\/11\/nanoparticles_0.jpeg","image_full_path":"http:\/\/hg.gatech.edu\/\/sites\/default\/files\/2025\/07\/11\/nanoparticles_0.jpeg","mime":"image\/jpeg","size":165079,"path_740":"http:\/\/hg.gatech.edu\/sites\/default\/files\/styles\/740xx_scale\/public\/2025\/07\/11\/nanoparticles_0.jpeg?itok=OckkBENh"}},"677412":{"id":"677412","type":"image","title":"vida_image.jpg","body":"\u003Cp\u003EVida Jamali, assistant professor in Georgia Tech\u0027s School of Chemical and Biomolecular Engineering\u003C\/p\u003E","created":"1752521358","gmt_created":"2025-07-14 19:29:18","changed":"1752521358","gmt_changed":"2025-07-14 19:29:18","alt":"Vida Jamali, assistant professor in Georgia Tech\u0027s School of Chemical and Biomolecular Engineering","file":{"fid":"261308","name":"vida_image.jpg","image_path":"\/sites\/default\/files\/2025\/07\/14\/vida_image.jpg","image_full_path":"http:\/\/hg.gatech.edu\/\/sites\/default\/files\/2025\/07\/14\/vida_image.jpg","mime":"image\/jpeg","size":1498575,"path_740":"http:\/\/hg.gatech.edu\/sites\/default\/files\/styles\/740xx_scale\/public\/2025\/07\/14\/vida_image.jpg?itok=pZ-nW65a"}}},"media_ids":["677402","677412"],"groups":[{"id":"1188","name":"Research Horizons"},{"id":"1240","name":"School of Chemical and Biomolecular Engineering"}],"categories":[{"id":"141","name":"Chemistry and Chemical Engineering"}],"keywords":[{"id":"187915","name":"go-researchnews"},{"id":"192863","name":"go-ai"}],"core_research_areas":[{"id":"193655","name":"Artificial Intelligence at Georgia Tech"}],"news_room_topics":[],"event_categories":[],"invited_audience":[],"affiliations":[],"classification":[],"areas_of_expertise":[],"news_and_recent_appearances":[],"phone":[],"contact":[{"value":"\u003Cp\u003EBrad Dixon, \u003Ca href=\u0022mailto:braddixon@gatech.edu\u0022\u003Ebraddixon@gatech.edu\u003C\/a\u003E\u003C\/p\u003E","format":"limited_html"}],"email":["braddixon@gatech.edu"],"slides":[],"orientation":[],"userdata":""}}}