Source author record

Kevin Lee

Kevin Lee appears in the imported research catalog. Authorship, coauthor and topic links are available while profile ownership is still unclaimed.

ResearcherUnclaimed source record

physics.app-ph cond-mat.mtrl-sci Machine Learning Methodology

Catalog footprint

What is connected

3works

4topics

4close collaborators

Actions

Connect this record

Open graph Browse works

Inspect adjacent papers, topics, institutions and collaborators without losing the researcher page.

Building this map preview

BZPEER is loading the nearby papers, people, topics and institutions for this page.

preprint2020arXiv

Crystal orientation dictated epitaxy of ultrawide bandgap 5.4-8.6 eV $α$-(AlGa)$_2$O$_3$ on m-plane sapphire

Ultra-wide bandgap semiconductors are ushering in the next generation of high power electronics. The correct crystal orientation can make or break successful epitaxy of such semiconductors. Here it is discovered that single-crystalline layers of $α$-(AlGa)$_2$O$_3$ alloys spanning bandgaps of 5.4 - 8.6 eV can be grown by molecular beam epitaxy. The key step is found to be the use of m-plane sapphire crystal. The phase transition of the epitaxial layers from the $α$- to the narrower bandgap $β$-phase is catalyzed by the c-plane of the crystal. Because the c-plane is orthogonal to the growth front of the m-plane surface of the crystal, the narrower bandgap pathways are eliminated, revealing a route to much wider bandgap materials with structural purity. The resulting energy bandgaps of the epitaxial layers span a range beyond the reach of all other semiconductor families, heralding the successful epitaxial stabilization of the largest bandgap materials family to date.

preprint2019arXiv

Enhanced injection efficiency and light output in bottom tunnel-junction light-emitting diodes using UID GaN spacers

Recently, the use of bottom-TJ geometry in LEDs, which achieves N-polar-like alignment of polarization fields in conventional metal-polar orientations, has enabled enhancements in LED performance due to improved injection efficiency. Here, we elucidate the root causes behind the enhanced injection efficiency by employing mature laser diode structures with optimized heterojunction GaN/In$_{0.17}$Ga$_{0.83}$N/GaN TJs and UID GaN spacers to separate the optical mode from the heavily doped absorbing p-cladding regions. In such laser structures, polarization offsets at the electron blocking layer, spacer, and quantum barrier interfaces play discernable roles in carrier transport. By comparing a top-TJ structure to a bottom-TJ structure, and correlating features in the electroluminescence, capacitance-voltage, and current-voltage characteristics to unique signatures of the N- and Ga-polar polarization heterointerfaces in energy band diagram simulations, we identify that improved hole injection at low currents, and improved electron blocking at high currents, leads to higher injection efficiency and higher output power for the bottom-TJ device throughout 5 orders of current density (0.015 - 1000 A/cm$^2$). Moreover, even with the addition of a UID GaN spacer, differential resistances are state-of-the-art, below 7x10-4 $Ω$cm$^2$. These results highlight the virtues of the bottom-TJ geometry for use in high-efficiency laser diodes.

preprint2015arXiv

Nonparametric mixture of Gaussian graphical models

Graphical model has been widely used to investigate the complex dependence structure of high-dimensional data, and it is common to assume that observed data follow a homogeneous graphical model. However, observations usually come from different resources and have heterogeneous hidden commonality in real-world applications. Thus, it is of great importance to estimate heterogeneous dependencies and discover subpopulation with certain commonality across the whole population. In this work, we introduce a novel regularized estimation scheme for learning nonparametric mixture of Gaussian graphical models, which extends the methodology and applicability of Gaussian graphical models and mixture models. We propose a unified penalized likelihood approach to effectively estimate nonparametric functional parameters and heterogeneous graphical parameters. We further design an efficient generalized effective EM algorithm to address three significant challenges: high-dimensionality, non-convexity, and label switching. Theoretically, we study both the algorithmic convergence of our proposed algorithm and the asymptotic properties of our proposed estimators. Numerically, we demonstrate the performance of our method in simulation studies and a real application to estimate human brain functional connectivity from ADHD imaging data, where two heterogeneous conditional dependencies are explained through profiling demographic variables and supported by existing scientific findings.