Yongmin Baek
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Yongmin Baek has a total of 99 co-authors. Here are the top co-authors by number of shared publications:
- Kyusang Lee
Department of Electrical and Computer Engineering University of Virginia Charlottesville Virginia USA
8 shared publications - Hyunseok Kim
Department of Materials Science and Engineering, University of Illinois Urbana-Champaign, Urbana, USA
4 shared publications - Jeehwan Kim
Department of Mechanical Engineering, Massachusetts Institute of Technology, Cambridge, USA
4 shared publications - Byungjoon Bae
Department of Electrical and Computer Engineering, University of Virginia, Charlottesville, USA
3 shared publications - Chanyeol Choi
LinqAlpha, New York, USA
3 shared publications - Junseok Heo
Ajou University
3 shared publications
Latest Publications
- 19 August 2026
- Published by Springer Nature in Nature Electronics
- Vol. 9 (8) , 853-867
Co-packaged optics for high-performance computing and artificial intelligence
- 19 August 2026
- Published by Springer Nature in Nature Electronics
- Vol. 9 (8) , 853-867
Co-packaged optics for high-performance computing and artificial intelligence
- Open Access
- 4 June 2026
- Published by Springer Nature in Nature Communications
- Vol. 17 (1)
AI-Assisted segmentation and volumetric reconstruction of radiographs through multi-angular scintillation imaging
X-ray imaging serves as a fundamental tool for non-destructive inspection. Although conventional radiography is well suited for two-dimensional imaging, it cannot provide volumetric structure. Computed tomography provides three-dimensional reconstruction but remains constrained by bulky instrumentation, high radiation exposure, and cost. Here we demonstrate a patch-type scintillator integrated with multi-stage neural network that segments and reconstructs three-dimensional volumes from sparse angular two-dimensional radiographs. The scintillator is fabricated by electrospraying cellulose nanocrystals onto a bulk cellulose matrix, followed by dip-coating of perovskite, yielding a composite with enhanced radioluminescence under X-ray excitation. This flexible film conforms to complex geometries, enabling distortion-free and multi-angle imaging. Neural networks are trained on synthetic datasets and validated on experimentally acquired avian tibiotarsus radiographs, accurately reconstructing volumetric bone structures. This approach serves as a proof-of-concept for low-dose, accessible artificial intelligence-enabled three-dimensional X-ray imaging, demonstrating the feasibility of recovering macroscopic three-dimensional morphology from as few as three sparse projections.
- 4 June 2026
- Published by Springer Nature in Nature Communications
- Vol. 17 (1)
Open AccessAI-Assisted segmentation and volumetric reconstruction of radiographs through multi-angular scintillation imaging
X-ray imaging serves as a fundamental tool for non-destructive inspection. Although conventional radiography is well suited for two-dimensional imaging, it cannot provide volumetric structure. Computed tomography provides three-dimensional reconstruction but remains constrained by bulky instrumentation, high radiation exposure, and cost. Here we demonstrate a patch-type scintillator integrated with multi-stage neural network that segments and reconstructs three-dimensional volumes from sparse angular two-dimensional radiographs. The scintillator is fabricated by electrospraying cellulose nanocrystals onto a bulk cellulose matrix, followed by dip-coating of perovskite, yielding a composite with enhanced radioluminescence under X-ray excitation. This flexible film conforms to complex geometries, enabling distortion-free and multi-angle imaging. Neural networks are trained on synthetic datasets and validated on experimentally acquired avian tibiotarsus radiographs, accurately reconstructing volumetric bone structures. This approach serves as a proof-of-concept for low-dose, accessible artificial intelligence-enabled three-dimensional X-ray imaging, demonstrating the feasibility of recovering macroscopic three-dimensional morphology from as few as three sparse projections.
- 25 November 2025
- Published by AIP Publishing in Applied Physics Reviews
- Vol. 12 (4)
Organic active waveguides
As the demand for high bandwidth and long-distance data transmission escalates in modern computing, optical interconnects via waveguides have attracted significant attention. While conventional inorganic materials-based waveguide necessitates complex components such as grating couplers and optical amplifiers, organic semiconductor-based waveguides offer simplified systems with unique functionalities stemming from their inherent radiative properties that facilitate efficient light–matter interactions, such as exciton–polariton formation and Förster resonance energy transfer. These interactions enable active light modulation, encompassing intensity control, wavelength shift, exciton–polariton lasing, and nonlinear optical effects. Moreover, their optical properties and structural geometries can be precisely tuned through molecular design and controlled synthesis techniques. As a result, organic waveguides have been explored for a range of applications including optical-logic operations, bio-chemical sensing, and advanced photonic integration systems. In this review, we delineate the fundamental principles of organic semiconductor waveguides, as well as their fabrication and potential impact on various photonic applications.
- 25 November 2025
- Published by AIP Publishing in Applied Physics Reviews
- Vol. 12 (4)
Organic active waveguides
As the demand for high bandwidth and long-distance data transmission escalates in modern computing, optical interconnects via waveguides have attracted significant attention. While conventional inorganic materials-based waveguide necessitates complex components such as grating couplers and optical amplifiers, organic semiconductor-based waveguides offer simplified systems with unique functionalities stemming from their inherent radiative properties that facilitate efficient light–matter interactions, such as exciton–polariton formation and Förster resonance energy transfer. These interactions enable active light modulation, encompassing intensity control, wavelength shift, exciton–polariton lasing, and nonlinear optical effects. Moreover, their optical properties and structural geometries can be precisely tuned through molecular design and controlled synthesis techniques. As a result, organic waveguides have been explored for a range of applications including optical-logic operations, bio-chemical sensing, and advanced photonic integration systems. In this review, we delineate the fundamental principles of organic semiconductor waveguides, as well as their fabrication and potential impact on various photonic applications.
- Open Access
- 18 October 2024
- Published by American Association for the Advancement of Science (AAAS) in Science Advances
- Vol. 10 (42)
Network of artificial olfactory receptors for spatiotemporal monitoring of toxic gas
Excessive human exposure to toxic gases can lead to chronic lung and cardiovascular diseases. Thus, precise in situ monitoring of toxic gases in the atmosphere is crucial. Here, we present an artificial olfactory system for spatiotemporal recognition of NO 2 gas flow by integrating a network of chemical receptors with a near-sensor computing. The artificial olfactory receptor features nano-islands of metal-based catalysts that cover the graphene surface on the heterostructure of an AlGaN/GaN two-dimensional electron gas (2DEG) channel. Catalytically dissociated NO 2 molecules bind to graphene, thereby modulating the conductivity of the 2DEG channel. For the energy/resource-efficient gas flow monitoring, trust-region Bayesian optimization algorithm allocates many sensors optimally in a complex space. Integrated artificial neural networks on a compact microprocessor with a network of sensors provide in situ gas flow predictions. This system enhances protective measures against toxic environments through spatiotemporal monitoring of toxic gases.
- 18 October 2024
- Published by American Association for the Advancement of Science (AAAS) in Science Advances
- Vol. 10 (42)
Open AccessNetwork of artificial olfactory receptors for spatiotemporal monitoring of toxic gas
Excessive human exposure to toxic gases can lead to chronic lung and cardiovascular diseases. Thus, precise in situ monitoring of toxic gases in the atmosphere is crucial. Here, we present an artificial olfactory system for spatiotemporal recognition of NO 2 gas flow by integrating a network of chemical receptors with a near-sensor computing. The artificial olfactory receptor features nano-islands of metal-based catalysts that cover the graphene surface on the heterostructure of an AlGaN/GaN two-dimensional electron gas (2DEG) channel. Catalytically dissociated NO 2 molecules bind to graphene, thereby modulating the conductivity of the 2DEG channel. For the energy/resource-efficient gas flow monitoring, trust-region Bayesian optimization algorithm allocates many sensors optimally in a complex space. Integrated artificial neural networks on a compact microprocessor with a network of sensors provide in situ gas flow predictions. This system enhances protective measures against toxic environments through spatiotemporal monitoring of toxic gases.
- 17 November 2023
- Published by Springer Nature in Nature Electronics
- Vol. 6 (11) , 799-800
Reconfigurable heterojunction transistors for off-grid medical devices
- 17 November 2023
- Published by Springer Nature in Nature Electronics
- Vol. 6 (11) , 799-800
Reconfigurable heterojunction transistors for off-grid medical devices
- 20 March 2023
- Published by Springer Nature in Nature Nanotechnology
- Vol. 18 (5) , 464-470
High-throughput manufacturing of epitaxial membranes from a single wafer by 2D materials-based layer transfer process
Layer transfer techniques have been extensively explored for semiconductor device fabrication as a path to reduce costs and to form heterogeneously integrated devices. These techniques entail isolating epitaxial layers from an expensive donor wafer to form freestanding membranes. However, current layer transfer processes are still low-throughput and too expensive to be commercially suitable. Here we report a high-throughput layer transfer technique that can produce multiple compound semiconductor membranes from a single wafer. We directly grow two-dimensional (2D) materials on III–N and III–V substrates using epitaxy tools, which enables a scheme comprised of multiple alternating layers of 2D materials and epilayers that can be formed by a single growth run. Each epilayer in the multistack structure is then harvested by layer-by-layer mechanical exfoliation, producing multiple freestanding membranes from a single wafer without involving time-consuming processes such as sacrificial layer etching or wafer polishing. Moreover, atomic-precision exfoliation at the 2D interface allows for the recycling of the wafers for subsequent membrane production, with the potential for greatly reducing the manufacturing cost.
- 20 March 2023
- Published by Springer Nature in Nature Nanotechnology
- Vol. 18 (5) , 464-470
High-throughput manufacturing of epitaxial membranes from a single wafer by 2D materials-based layer transfer process
Layer transfer techniques have been extensively explored for semiconductor device fabrication as a path to reduce costs and to form heterogeneously integrated devices. These techniques entail isolating epitaxial layers from an expensive donor wafer to form freestanding membranes. However, current layer transfer processes are still low-throughput and too expensive to be commercially suitable. Here we report a high-throughput layer transfer technique that can produce multiple compound semiconductor membranes from a single wafer. We directly grow two-dimensional (2D) materials on III–N and III–V substrates using epitaxy tools, which enables a scheme comprised of multiple alternating layers of 2D materials and epilayers that can be formed by a single growth run. Each epilayer in the multistack structure is then harvested by layer-by-layer mechanical exfoliation, producing multiple freestanding membranes from a single wafer without involving time-consuming processes such as sacrificial layer etching or wafer polishing. Moreover, atomic-precision exfoliation at the 2D interface allows for the recycling of the wafers for subsequent membrane production, with the potential for greatly reducing the manufacturing cost.
- 22 September 2022
- Published by Springer Nature in Nature Nanotechnology
- Vol. 17 (10) , 1054-1059
Graphene nanopattern as a universal epitaxy platform for single-crystal membrane production and defect reduction
Heterogeneous integration of single-crystal materials offers great opportunities for advanced device platforms and functional systems1. Although substantial efforts have been made to co-integrate active device layers by heteroepitaxy, the mismatch in lattice polarity and lattice constants has been limiting the quality of the grown materials2. Layer transfer methods as an alternative approach, on the other hand, suffer from the limited availability of transferrable materials and transfer-process-related obstacles3. Here, we introduce graphene nanopatterns as an advanced heterointegration platform that allows the creation of a broad spectrum of freestanding single-crystalline membranes with substantially reduced defects, ranging from non-polar materials to polar materials and from low-bandgap to high-bandgap semiconductors. Additionally, we unveil unique mechanisms to substantially reduce crystallographic defects such as misfit dislocations, threading dislocations and antiphase boundaries in lattice- and polarity-mismatched heteroepitaxial systems, owing to the flexibility and chemical inertness of graphene nanopatterns. More importantly, we develop a comprehensive mechanics theory to precisely guide cracks through the graphene layer, and demonstrate the successful exfoliation of any epitaxial overlayers grown on the graphene nanopatterns. Thus, this approach has the potential to revolutionize the heterogeneous integration of dissimilar materials by widening the choice of materials and offering flexibility in designing heterointegrated systems.
- 22 September 2022
- Published by Springer Nature in Nature Nanotechnology
- Vol. 17 (10) , 1054-1059
Graphene nanopattern as a universal epitaxy platform for single-crystal membrane production and defect reduction
Heterogeneous integration of single-crystal materials offers great opportunities for advanced device platforms and functional systems1. Although substantial efforts have been made to co-integrate active device layers by heteroepitaxy, the mismatch in lattice polarity and lattice constants has been limiting the quality of the grown materials2. Layer transfer methods as an alternative approach, on the other hand, suffer from the limited availability of transferrable materials and transfer-process-related obstacles3. Here, we introduce graphene nanopatterns as an advanced heterointegration platform that allows the creation of a broad spectrum of freestanding single-crystalline membranes with substantially reduced defects, ranging from non-polar materials to polar materials and from low-bandgap to high-bandgap semiconductors. Additionally, we unveil unique mechanisms to substantially reduce crystallographic defects such as misfit dislocations, threading dislocations and antiphase boundaries in lattice- and polarity-mismatched heteroepitaxial systems, owing to the flexibility and chemical inertness of graphene nanopatterns. More importantly, we develop a comprehensive mechanics theory to precisely guide cracks through the graphene layer, and demonstrate the successful exfoliation of any epitaxial overlayers grown on the graphene nanopatterns. Thus, this approach has the potential to revolutionize the heterogeneous integration of dissimilar materials by widening the choice of materials and offering flexibility in designing heterointegrated systems.
- Open Access
- 5 September 2022
- Published by Springer Nature in Nature Communications
- Vol. 13 (1) , 1-9
In-sensor image memorization and encoding via optical neurons for bio-stimulus domain reduction toward visual cognitive processing
As machine vision technology generates large amounts of data from sensors, it requires efficient computational systems for visual cognitive processing. Recently, in-sensor computing systems have emerged as a potential solution for reducing unnecessary data transfer and realizing fast and energy-efficient visual cognitive processing. However, they still lack the capability to process stored images directly within the sensor. Here, we demonstrate a heterogeneously integrated 1-photodiode and 1 memristor (1P-1R) crossbar for in-sensor visual cognitive processing, emulating a mammalian image encoding process to extract features from the input images. Unlike other neuromorphic vision processes, the trained weight values are applied as an input voltage to the image-saved crossbar array instead of storing the weight value in the memristors, realizing the in-sensor computing paradigm. We believe the heterogeneously integrated in-sensor computing platform provides an advanced architecture for real-time and data-intensive machine-vision applications via bio-stimulus domain reduction.
- 5 September 2022
- Published by Springer Nature in Nature Communications
- Vol. 13 (1) , 1-9
Open AccessIn-sensor image memorization and encoding via optical neurons for bio-stimulus domain reduction toward visual cognitive processing
As machine vision technology generates large amounts of data from sensors, it requires efficient computational systems for visual cognitive processing. Recently, in-sensor computing systems have emerged as a potential solution for reducing unnecessary data transfer and realizing fast and energy-efficient visual cognitive processing. However, they still lack the capability to process stored images directly within the sensor. Here, we demonstrate a heterogeneously integrated 1-photodiode and 1 memristor (1P-1R) crossbar for in-sensor visual cognitive processing, emulating a mammalian image encoding process to extract features from the input images. Unlike other neuromorphic vision processes, the trained weight values are applied as an input voltage to the image-saved crossbar array instead of storing the weight value in the memristors, realizing the in-sensor computing paradigm. We believe the heterogeneously integrated in-sensor computing platform provides an advanced architecture for real-time and data-intensive machine-vision applications via bio-stimulus domain reduction.
- Open Access
- 6 April 2022
- Published by Springer Nature in Research Square
In-sensor image memorization and encoding via optical neurons for bio-stimulus domain reduction towards visual cognitive processing
As machine vision technology generates large amounts of data from sensors, it requires efficient computational systems for visual cognitive processing. However, the spatially separated sensor/processor architecture suffers from data transportation bottleneck issues due to repeated and redundant data exchange between them. Recently, in-sensor computing systems have emerged as a potential solution for reducing unnecessary data transfer and realizing fast and energy-efficient visual cognitive processing. However, they still lack the capability to process stored images directly within the sensor. Here, we demonstrate an advanced machine vision system for in-sensor visual cognitive processing, emulating a mammalian image encoding process to extract features from the input images, to achieve a domain reduction from the sensor. Heterogeneous integration of a compound semiconductor-based 1-photodiode and oxide-based 1-memristor array enables to memorize images in pixels, and directly conducts a multiply–accumulate operation within the sensor. Since trained weight values are applied as an input to the image-saved crossbar array instead of storing the weight value in it as using the vectorized image as an input for multiply–accumulate operations in our system. Therefore, the input data towards the artificial neural network (ANN) is significantly reduced, and vectorization of the image is no longer needed, thereby minimizing data transportation bottlenecks. We believe the heterogeneously integrated in-sensor computing platform provides an advanced architecture for real-time and data-intensive machine-vision applications via bio-stimulus domain reduction.
- 6 April 2022
- Published by Springer Nature in Research Square
Open AccessIn-sensor image memorization and encoding via optical neurons for bio-stimulus domain reduction towards visual cognitive processing
As machine vision technology generates large amounts of data from sensors, it requires efficient computational systems for visual cognitive processing. However, the spatially separated sensor/processor architecture suffers from data transportation bottleneck issues due to repeated and redundant data exchange between them. Recently, in-sensor computing systems have emerged as a potential solution for reducing unnecessary data transfer and realizing fast and energy-efficient visual cognitive processing. However, they still lack the capability to process stored images directly within the sensor. Here, we demonstrate an advanced machine vision system for in-sensor visual cognitive processing, emulating a mammalian image encoding process to extract features from the input images, to achieve a domain reduction from the sensor. Heterogeneous integration of a compound semiconductor-based 1-photodiode and oxide-based 1-memristor array enables to memorize images in pixels, and directly conducts a multiply–accumulate operation within the sensor. Since trained weight values are applied as an input to the image-saved crossbar array instead of storing the weight value in it as using the vectorized image as an input for multiply–accumulate operations in our system. Therefore, the input data towards the artificial neural network (ANN) is significantly reduced, and vectorization of the image is no longer needed, thereby minimizing data transportation bottlenecks. We believe the heterogeneously integrated in-sensor computing platform provides an advanced architecture for real-time and data-intensive machine-vision applications via bio-stimulus domain reduction.
- 10 February 2020
- Published by Springer Nature in Nature Nanotechnology
- Vol. 15 (4) , 272-276
Graphene-assisted spontaneous relaxation towards dislocation-free heteroepitaxy
Although conventional homoepitaxy forms high-quality epitaxial layers1,2,3,4,5, the limited set of material systems for commercially available wafers restricts the range of materials that can be grown homoepitaxially. At the same time, conventional heteroepitaxy of lattice-mismatched systems produces dislocations above a critical strain energy to release the accumulated strain energy as the film thickness increases. The formation of dislocations, which severely degrade electronic/photonic device performances6,7,8, is fundamentally unavoidable in highly lattice-mismatched epitaxy9,10,11. Here, we introduce a unique mechanism of relaxing misfit strain in heteroepitaxial films that can enable effective lattice engineering. We have observed that heteroepitaxy on graphene-coated substrates allows for spontaneous relaxation of misfit strain owing to the slippery graphene surface while achieving single-crystalline films by reading the atomic potential from the substrate. This spontaneous relaxation technique could transform the monolithic integration of largely lattice-mismatched systems by covering a wide range of the misfit spectrum to enhance and broaden the functionality of semiconductor devices for advanced electronics and photonics.
- 10 February 2020
- Published by Springer Nature in Nature Nanotechnology
- Vol. 15 (4) , 272-276
Graphene-assisted spontaneous relaxation towards dislocation-free heteroepitaxy
Although conventional homoepitaxy forms high-quality epitaxial layers1,2,3,4,5, the limited set of material systems for commercially available wafers restricts the range of materials that can be grown homoepitaxially. At the same time, conventional heteroepitaxy of lattice-mismatched systems produces dislocations above a critical strain energy to release the accumulated strain energy as the film thickness increases. The formation of dislocations, which severely degrade electronic/photonic device performances6,7,8, is fundamentally unavoidable in highly lattice-mismatched epitaxy9,10,11. Here, we introduce a unique mechanism of relaxing misfit strain in heteroepitaxial films that can enable effective lattice engineering. We have observed that heteroepitaxy on graphene-coated substrates allows for spontaneous relaxation of misfit strain owing to the slippery graphene surface while achieving single-crystalline films by reading the atomic potential from the substrate. This spontaneous relaxation technique could transform the monolithic integration of largely lattice-mismatched systems by covering a wide range of the misfit spectrum to enhance and broaden the functionality of semiconductor devices for advanced electronics and photonics.