Ultrafast 3D Microfabrication Technology Produces Complex Structures in Just 60 Seconds

Abstract Despite the advantages of additive manufacturing (AM) in creating customized 3D shapes, conventional layer-by-layer approaches are limited by low production rates, restricting their broader applications. Volumetric additive manufacturing (VAM) has emerged as a promising technique, enabling the simultaneous photopolymerization of entire volumes, which significantly reduces fabrication time. However, current computed axial lithography requires manual operations per print cycle, such as loading resin into a vial, physically placing and aligning the vial (with or without an index-matching medium), and removing the printed object, limiting continuous, high-throughput production of multiple parts. Here, we propose a dispensing volumetric additive manufacturing (DVAM) method that prints and dispenses each part within a droplet in less than a minute. The printing process occurs within a single droplet dispensed from a glass pipette, enabling simultaneous printed object removal and resin replenishment in a second. Light pattern distortion caused by the absence of the index-matching fluid was corrected through real-time droplet profile estimation and inverse ray-tracing within the optical system. We demonstrate rapid serial VAM of 10 different objects within 10 min. This approach establishes a practical pathway toward scalable, high-throughput volumetric manufacturing, enabling rapid production of complex 3D structures without the operational bottlenecks of conventional VAM workflows. A research team affiliated with UNIST has achieved a major breakthrough in 3D microfabrication, enabling the rapid production of intricate 3D structures within a single minute. The new technology promises to significantly accelerate manufacturing processes in fields, ranging from biomedical devices to microelectronics. Led by Professor Im Doo Jung from the Department of Mechanical Engineering at UNIST, the research team developed a cutting-edge dispensing volumetric additive manufacturing (DVAM) system.The new approach moves beyond traditional layer-by-layer methods, allowing for the swift, continuous fabrication of diverse three-dimensional shapes within a matter of minutes by employing a volumetric printing process. DVAM utilizes a fine resin droplet dispensed from a glass pipette as the entire build volume. Light is projected onto the droplet to cure the desired shape in real-time. Once a structure is solidified, compressed air expels the finished object, and a new droplet is immediately dispensed for the next cycle. This seamless process enables high-speed, continuous production without the need for post-processing or layer-by-layer assembly. Unlike conventional volumetric printing methods—such as Computed Axial Lithography (CAL)—which require filling a resin container, matching refractive indices with immersion fluids, and manually removing finished parts, the UNIST team’s approach eliminates these constraints. By performing printing within a single resin droplet that acts as the entire volume, they achieve rapid, uninterrupted manufacturing. One of the main challenges was optical distortion caused by the droplet’s curved surface. To overcome this, the researchers integrated artificial intelligence with inverse ray-tracing optical calculations. A deep learning-based AI accurately detects the droplet’s profile in real-time, enabling the system to mathematically correct for distortions and project precise light patterns for uniform curing. This advanced system successfully produced intricate structures, such as the Eiffel Tower and The Thinker, within just 10 minutes—demonstrating the ability to fabricate ten different objects with an average time of around 60 seconds each. Hongryung Jeon, the first author of the study, explained, “Unlike traditional 3D printing that builds objects layer by layer, our method cures the entire volume at once and continuously dispenses resin without additional post-processing, increasing speed by over 100 times. This opens up exciting possibilities for large-scale, rapid production of micro-scale components.” Professor Jung, the corresponding author, emphasized, “Speed has long been a limiting factor in 3D printing, especially for customized manufacturing. By moving away from conventional photopolymerization and employing artificial intelligence to compensate for optical distortions, we have significantly advanced the potential of ultrafast volumetric fabrication. Now, complex shapes can be produced in seconds—eliminating long wait times.” The findings of this research have been published online in Advanced Functional Materials (Impact Factor: 19.0, top 5% in JCR) on March 21, 2026. The project was supported by the National Research Foundation of Korea (NRF), the Ministry of Science and ICT (MSIT), the Institute for Information & communications Technology Planning & Evaluation (IITP), and the Ministry of Trade, Industry & Energy (MOTIE). Journal Reference Hongryung Jeon, Yunsoo Lee, Seobin Park, et al., "Dispensing Volumetric Additive Manufacturing," Adv. Funct. Mater., (2026).

2026.05.12

New AI Algorithm to Enhance Accuracy of Thermal and Stress Predictions in Semiconductors

Abstract PDE surrogate models such as FNO and PINN struggle to predict solutions across inputs with diverse physical units and scales, limiting their out-of-distribution (OOD) generalization. We propose a π-invariant test-time projection that aligns test inputs with the training distribution by solving a log-space least squares problem that preserves Buckingham π-invariants. For PDEs with multidimensional spatial fields, we use geometric representative π-values to compute distances and project inputs, overcoming degeneracy and singular points that limit prior π methods. To accelerate projection, we cluster the training set into K clusters, reducing the complexity from O(MN) to O(KN) for the M training and N test samples. Across wide input scale ranges, tests on 2D thermal conduction and linear elasticity achieve MAE reduction of up to ≈91% with minimal overhead. This training-free, model-agnostic method is expected to apply to more diverse PDE-based simulations. A research team affiliated with UNIST has introduced a novel AI-based algorithm that enhances the accuracy of thermal and mechanical predictions across various scales, from microchips to large pipelines. Led by Professor Changwook Jeong from the Graduate School of Semiconductor Materials and Devices Engineering, their π-invariant test-time projection method realigns input data to conform with physical laws, addressing a crucial challenge in AI modeling—accurate predictions when faced with unfamiliar or out-of-distribution data. The algorithm identifies the most physically similar data within existing training sets based on a dimensionless ratio derived from Buckingham's π theorem. It then transforms new inputs into familiar, physically consistent forms without retraining the model, operating in log space to preserve physical ratios. This approach is computationally efficient, reducing processing costs by approximately 99% compared to traditional methods. Applied to 2D thermal conduction and linear elasticity problems, the technique achieved up to a 91% reduction in prediction error, even under conditions outside the original training range. It also demonstrated promising results in fluid dynamics, improving the accuracy of Navier–Stokes equation predictions in complex scenarios. This advancement is expected to accelerate and economize simulations in semiconductor design, packaging reliability, battery management, and structural safety analysis—fields where varying sizes and conditions demand both precision and efficiency. The study has been supported by the National Research Foundation of Korea (NRF) and the Institute of Information & Communications Technology Planning & Evaluation (IITP). Journal Reference Seokki Lee, Min-Chul Park, Giyong Hong, and Changwook Jeong, "Buckingham π-Invariant Test‑Time Projection for Robust PDE Surrogate Modeling," ICLR 2026 .

2026.05.11

New Study Unveils Viral DNA ‘Barcode’ that Activates Innate Immunity, Opening New Therapeutic Avenues

Abstract Herpes simplex virus type 1 (HSV-1) infection remains a major global health challenge, yet the mechanisms underlying strain-specific innate immune responses are poorly understood. Here, we show that distinct HSV-1 strains differentially activate the absent in melanoma 2 (AIM2) inflammasome. The HF strain robustly induces AIM2-dependent inflammasome activation, whereas the F and KOS strains elicit minimal responses despite comparable infection efficiency. We demonstrate that this difference is driven by viral genomic features rather than replication capacity. Genomic analyses identify a poly(T) DNA sequence within the UL25-UL26 intergenic region that is enriched in the HF strain. Deletion of a 14-mer poly(T) sequence markedly impairs inflammasome activation, cytokine release, and host protection in vivo, whereas introduction of a poly(T) tract into the F strain is sufficient to confer AIM2 activation and enhanced host defense. Furthermore, poly(T)-mediated AIM2 activation is length-dependent, conserved in human macrophages, and requires a cGAS-STING-IRF1 licensing axis. Together, these findings identify viral poly(T) DNA as a key determinant of strain-specific AIM2 inflammasome activation and reveal how viral genomic variation shapes innate immune recognition. A research team, led by Professor SangJoon Lee from the Department of Biological Sciences at UNIST has identified a specific viral DNA sequence that triggers the body’s innate immune system, offering new insights into viral detection mechanisms and potential avenues for immunomodulatory therapies. In collaboration with Professor Joo Sang Lee from Sungkyunkwan University, Professor Eui Tae Kim from Jeju National University, and Director Young Ki Choi from the Korea Virus Research Institute at the Institute for Basic Science (IBS), the team demonstrated that a repetitive poly(T) sequence within the genome of herpes simplex virus type 1 (HSV-1) functions as a molecular barcode recognized by the innate immune sensor AIM2. HSV-1 infects about 67% of the global population and remains dormant in nerve cells, evading immune detection. Upon reactivation, it infects epithelial cells, provoking inflammation as part of the host’s defense. The study revealed that AIM2, a pivotal immune sensor in macrophages, detects the virus by recognizing poly(T) tracts—regions of repeated thymine (T) nucleotides—in its DNA. Importantly, immune activation was shown to depend on both the presence and length of this poly(T) sequence: longer repeats elicited stronger inflammatory responses and induced apoptosis in infected cells. Strains lacking the sequence or with fewer than 20 thymine bases elicited minimal immune activity. Introducing the poly(T) sequence into less responsive strains restored immune detection, with increased length correlating to heightened responses. These findings were confirmed in animal models, where viruses containing the poly(T) motif triggered inflammation and suppressed viral replication, while those without it led to unchecked proliferation and more severe disease outcomes. Further genomic analysis revealed that similar poly(T) repeats are conserved across various pathogenic viruses—including orthopoxviruses, adenoviruses, and coronaviruses—indicating a shared mechanism of immune recognition. “This study demonstrates that innate immune sensors recognize viral genomes with greater specificity than previously thought,” said Professor Lee. “Targeting these sensing pathways, rather than the viruses themselves, may offer innovative therapeutic options. Our research provides a foundation for developing immunomodulatory treatments for infectious diseases.” He added, “Given the conservation of this sequence across multiple viruses associated with severe illnesses, understanding its role could help link viral genomic features to disease severity and inform the design of targeted therapies.” Published in Nature Communications on April 13, the study was supported by multiple funding agencies, including the National Research Foundation of Korea (NRF), the Korea Health Industry Development Institute (KHIDI), the Ministry of Health and Welfare (MOHW), the Institute for Basic Science (IBS), the Circle Foundation, and Yuhan Corporation. Journal Reference SuHyeon Oh, Jueun Oh, Kyeongchan Im, et al., "Herpes simplex virus 1 harboring poly(T) DNA sequences as a key ligand for AIM2 inflammasome activation and host defense," Nat. Commun., (2026).

2026.05.04

Re‑Engineering the Future: For A More Resilience World

《 Editor's Note: Across fields that range from energy engineering to quantum physics, UNIST researchers share a common aim: redesigning the systems that sustain life and technology. Their latest work—converting solar‑panel waste into clean hydrogen, restoring vision through nanomedicine, and decoding the feedback loops driving global wildfires—positions science not merely as discovery, but as a blueprint for resilience.》 Sensing the Future From visions that steer robots to displays sharper than reality, UNIST researchers are redefining how machines interpret the human world. A smart contact lens equipped with AI‑driven sensors enables vision‑based robotic control, while a new quantum‑dot technique achieves over 4,000 pixels per inch—paving the way for ultra‑compact extended‑reality glasses. On a different frontier, three papers accepted to ICLR 2026 show the university's rising influence in reinforcement learning, advancing physical AI systems that learn directly from their environments. Together, these studies represent technology that doesn't just emulate perception—it extends it. • Smart Contact Lens for Vision-Based Robotic Control (Adv. Funct. Mater. l Mar., 2026) • Breakthrough QD Display with 4,000 PPI Resolution (Nat. Commun. l Mar., 2026) • Three Research Papers Accepted to ICLR 2026 Circulating Energy What if waste could power the next energy revolution? Two separate breakthroughs at UNIST bring that vision closer. One process converts end‑of‑life solar panels into high‑purity hydrogen and silica using mechanochemical reactions; Another turns industrial glycerol into hydrogen and value‑added chemicals through a scalable electrochemical system. Both replace extraction with regeneration, showing how circular design can meet industrial scale without environmental cost. This emerging “loop thinking” is central to UNIST's broader push toward a sustainable hydrogen economy. • Solar Panel Waste into H2 and High-Value Silica (Joule l Mar., 2026) • Turning Waste Glycerol into H2 and High-Value Chemicals (Joule l Mar., 2026) Healing Systems At the intersection of biology and environment, UNIST scientists are uncovering how systems heal themselves. A nanomedicine platform that selectively removes aging retinal cells restores vision in degenerative models, while another study reveals how alcohol and immune signaling interact to amplify liver damage—insights that could guide future therapies. On a planetary scale, researchers have identified a climate “regime shift” behind Australia's increasing wildfire intensity and developed AI frameworks for real‑time multilevel air‑pollution prediction. From cells to climate, the common thread is understanding how complex networks can recover balance. • Targeted Nanoparticles Aging Retinal Cells (Nat. Commun. l Mar., 2026) • Alcohol Amplifies Liver Damage During Illness (Sci. Adv. l Apr., 2026) • Climate Prediction (Nat. Commun. l Mar., 2026 ) • Wildfire Prediction (Agric. For. Meteorol. l Apr., 2026 ) • Deep Learning for Multiple Air Pollutant Analysis (Environ. Sci. Technol. l Mar., 2026) Across its laboratories, UNIST demonstrates that resilience is not a separate discipline—it is a design principle. By approaching technology, energy, and biology as interconnected systems, these efforts move beyond repair. They point instead toward a more fundamental question: How Can the World Be Designed to Endure—and Renew?

2026.05.01

Breakthrough QD Display Technology Achieves 4,000 PPI Resolution — Ideal for XR Glasses

Abstract Colloidal quantum dots (QDs) are leading candidates for next-generation optoelectronics owing to their tuneable bandgaps, narrow emission linewidths, and high luminescence quantum yields. For virtual-, augmented-, and mixed-reality display applications of these materials, patterning full-color QDs at μm-length scales is essential. However, existing photolithographic approaches often degrade QD luminance characteristics by exposing them to harsh processing conditions, or they compromise the structural fidelity of the resulting patterns. Here we report a photoresist-guided indirect (PIN) photopatterning strategy that includes (i) lithographic formation of sacrificial PR patterns, (ii) deposition of a crosslinked QD film on top, and (iii) PR stripping that removes the sacrificial PR, leaving behind crosslinked QD patterns on the substrate. QD crosslinking is mediated by a diazo-based ligand thermocrosslinker, Diazo-4-LiXer. Leveraging low-temperature (110–120 °C)-activated carbene chemistry, Diazo-4-LiXer bridges neighbouring QDs while maintaining their intrinsic photoluminescence and electroluminescence through repeated processing. Moreover, Diazo-4-LiXer enables thermocrosslinking without affecting the underlying photoresist pre-patterns, which serve as structural templates determining the thickness and fidelity of the QD patterns. Using PIN photopatterning, we realize high-fidelity RGB patterns exceeding 4,000 pixels per inch resolution and demonstrate integration-level scalability by fabricating a 10 × 10 passive-matrix full-colour RGB QD–LED array. Researchers, affiliated with UNIST have announced a significant advancement in quantum dot (QD) display technology, enabling ultra-fine, high-resolution patterning of QDs suitable for next-generation XR glasses. This innovation promises brighter, more vivid images even in outdoor environments. Professor BongSoo Kim from the Department of Chemistry at UNIST, in collaboration with Professor Moon Sung Kang of Sogang University and Dr. Chan-mo Kang at Electronics and Telecommunications Research Institute (ETRI), have developed a novel photopatterning technique that maintains the integrity of quantum dots while achieving micron-scale precision. To meet the demanding resolution required for XR displays—exceeding 3,000 pixels per inch (PPI)—the team’s method patterns quantum dots into 2-micrometer (μm) pixels. This enables over 4,000 PPI, meaning more than 10 million pixels can be packed into a space about the size of a coin, ensuring sharp, immersive visuals. The process employs a custom-developed additive, Diazo-4-LiXer, which facilitates thermocrosslinking at low temperatures (110–120°C) via carbene chemistry, preserving the quantum dots’ luminescent properties. It involves creating a sacrificial photoresist (PR) template, depositing a crosslinked quantum dot film, then removing the PR to leave behind precisely patterned quantum dots—without damage or shape distortion. This technique not only achieves high fidelity and density but also enables the fabrication of full-color RGB quantum dot LED arrays, demonstrating its potential for commercial application. Professor Kim commented, “Our method allows for ultra-high resolution patterning of quantum dots while maintaining their exceptional optical properties. It opens new avenues for advanced XR glasses and microdisplays, where brightness, color purity, and resolution are crucial.” The findings of this research have been published in Nature Communications on March 19, 2026. The study has been supported by the Samsung Research Funding & Incubation Center of Samsung Electronics, the National Research Foundation of Korea (NRF), and the Ministry of Science and ICT (MSIT). Journal Reference Hyeokjun Kim, Hyobin Ham, Chang Hyeok Lim, et al., "Photoresist-guided indirect photopatterning of quantum dots via carbene-mediated ligand thermocrosslinking," Nat. Commun., (2026).

2026.04.27

UNIST Demonstrates Global Competitiveness in Reinforcement Learning with Three Papers Accepted to ICLR 2026

UNIST has demonstrated strong international competitiveness in reinforcement learning, a core technology for physical AI, with three papers by Professor Seungyul Han’s research group from the Graduate School of Artificial Intelligence accepted to the International Conference on Learning Representations (ICLR 2026), held in Rio de Janeiro, Brazil, from April 23 to 27, 2026. ICLR, alongside NeurIPS and ICML, is widely regarded as one of the world’s leading artificial intelligence conferences. This year, approximately 5,300 papers—about 27% of more than 19,000 submissions—were accepted, making the selection of three papers from a single research group a notable achievement. The accepted studies address key challenges in reinforcement learning, where AI systems learn optimal actions through interaction with their environment—an approach essential for applications, such as robotics and autonomous systems operating in complex, real-world conditions. The first study proposes Self-Improving Skill Learning (SISL), a method designed to enable robust learning from noisy offline data. By decomposing long-horizon tasks into reusable skills and refining them through prioritized updates, SISL mitigates the impact of imperfect data and supports stable adaptation across complex tasks. The second study introduces Strict Subgoal Execution (SSE), which improves long-horizon planning by distinguishing feasible subgoals from unreachable ones. By leveraging past failures and partial successes, the method enhances planning efficiency and increases overall task reliability in goal-conditioned environments. The third study presents Successive Sub-value Q-learning (S2Q), a framework for multi-agent reinforcement learning (MARL) that retains multiple high-value action candidates. This approach enables agents to adapt more effectively in dynamic environments where optimal strategies shift over time, improving both coordination and overall performance. The research was led by Sanghyun Lee, Jaebak Hwang, and Yonghyeon Cho as first authors, respectively, and supported by programs funded by the Ministry of Science and ICT (MSIT) and the National Research Foundation of Korea (NRF). Professor Han said, “Our research demonstrates that reinforcement learning can be applied more reliably in environments with limited data and uncertainty, with strong potential for applications in autonomous driving, robotics, and smart manufacturing.” Journal Reference [1] Sanghyeon Lee, Sangjun Bae, Yisak Park, Seungyul Han, "Self-Improving Skill Learning for Robust Skill-based Meta-Reinforcement Learning," ICLR 2026. [2] Jaebak Hwang, Sanghyeon Lee, Jeongmo Kim, Seungyul Han, "Strict Subgoal Execution: Reliable Long-Horizon Planning in Hierarchical Reinforcement Learning," ICLR 2026. [3] Yonghyeon Jo, Sunwoo Lee, Seungyul Han, "Retaining Suboptimal Actions to Follow Shifting Optima in Multi-Agent Reinforcement Learning," ICLR 2026.

2026.04.27