Embedded Files
• Nano-optics, Nanophotonic physics, Metasurfaces
• Nano optical phenomena and their applications for industrial use
• AI-based design methods for physical systems
Toward next-generation displays and imaging devices
1) Super high-resolved pixel technology
1) Super high-resolved pixel technology
Structural coloration with nano-resonance
Interestingly, 'nanostructures' can represent color without any coloring materials (such as pigments or dyes). Compared to the material-based colors, structural color shows better stability from heat, light, and chemicals. In addition, the design method is quite simple. There needs to be more progress for practical industrial use as follows.
- Super high-resolution pixels and its optical issues (ex. crosstalk between pixels, pixel size, emission patterns, etc..)
2) High efficient & Narrowband emission device by optical design
2) High efficient & Narrowband emission device by optical design
Super-Luminescent Diode (SLD): High efficiency and High spectral-precision of electroluminescent devices
Light sources in industrial devices generally adopt 'Electroluminescence(EL)' emission, such as OLEDs, micro-LEDs, and QLEDs. In general, they are made up of multi-layered thin films of organic or inorganic materials sandwiched between two electrodes and are used in a variety of applications such as lighting, displays, and sensors. Superluminescent diodes (SLDs) occupy a unique regime between laser emission and electroluminescence, enabling optical properties unattainable by either technology alone. Investigating this transition region is essential for developing next-generation photonic platforms and expanding new application domains. Further research is required to realize the ultimate optical uses of the future, as follows.
- Super-high spectral precision emission: full width half maximum (FHWM) under 10nm by EL devices- High directionality of beam emission pattern: reducing complexity of optic system in XR devices
- Simple and industrial-friendly fabrication methods: without nano patterning process
- Angular independence microcavity structure
AI-driven design methodology for physical devices
With the rapid advancement of nanophotonic fields, photonic device research is becoming increasingly restricted by design and optimization strategy rather than fabrication methods. The extensive parameter space defined by geometry and material properties is challenging to explore with existing methodologies, including simple parameter sweep simulations or data-intensive deep learning. Effectively exploiting this dense parameter domain requires deep understanding of intricate theoretical background, making it difficult for early-career researchers to enter this field. These obstacles impede innovation in structural colors, metasurfaces, and other functional nanophotonic devices that require precise optical control.
Recently, artificial neural network (ANN)-assisted designs have been employed in photonic structures as a rapid prototyping method. Since an ANN autonomously learns input-output relationships while concurrently exploring a broader design space, this approach can serve as a simple alternative without domain expertise. Given that photonic structures comprise intricate geometries, the volume of each training dataset becomes huge, and computations of ANNs get increasingly sophisticated. Recent advancements in DNN, including solutions to the vanishing gradient problem (e.g., ReLU, residual connections), improved weight initialization data normalization, and regularization, have allowed ANNs to train effectively on large datasets with deeper architectures. Given these advancements, ANN-assisted design has been utilized in various photonic devices.
1) Inverse design by extrapolative optimization
1) Inverse design by extrapolative optimization
Algorithm: Domain knowledge extraction & Self-adaptive learning
In the inverse network, the deep neural network (DNN) predicts design parameters from target performances provided as input. Such inverse designs often lack unique solutions and may even have no valid solutions. Moreover, DNN-based inverse design models have limitations in predicting design parameters for targets that deviate from the distribution of training data. Despite eliminating the need for expertise during the design process, substantial theoretical knowledge is still required to construct appropriate training datasets.
We aim to ... TBD
2) Adaptive learning for practical applications
2) Adaptive learning for practical applications
Learning with non-ideal experimental environments
TBD
3) Fewshot learning algorithm for physical systems
3) Fewshot learning algorithm for physical systems
Physics-informed AI design: overcome data paradox
Unlike purely computational systems, physical devices inherently rely on scarce experimental data due to the complexity and cost of fabrication and characterization. Generally, a supervised regime of deep learning-based design requires considerable prior calculations to prepare training datasets to achieve high accuracy.
> Physics knowledge ....
TBD
Optical and photonic engineering of optoelectronic devices
Electro-luminescent (EL) device, Photovoltaic (PV) device
A solar cell is a device that converts light into electricity. Solar cells with various absorber layers, such as silicon, compound semiconductors, and organic materials, are available, and numerous studies are being conducted to achieve high power conversion efficiency at a low cost. For example, if the thickness of the absorber layer is reduced to save money, the efficiency of the solar cell suffers as a result. Internal light reflection can be used to increase the efficiency of a solar cell by sandwiching a passivation layer between the absorption layer and the interface. Using photoelectric modeling to adjust the passivation layer's local contact opening width and pitch, our group is investigating the relationship between the passivation layer's performance and structure, as well as structural parameters on the nano-micro scale. We may also propose a method for fabricating high-efficiency thin-film solar cells for PV generators and sensors, and we are developing optical and photovoltaic cells for color or transparent PV using transparent substrates and electrodes.
Research Environment
Google Sites
Report abuse