Journal Articles#

† indicates corresponding author, * indicates equal contribution

Computational X-ray Microscopy

Based on structured illumination, we develop a single-pixel X-ray imaging approach coupled with a generative image reconstruction model for mapping the compositional heterogeneity with nanoscale resolvability.

We develop an unmixing framework to retrieve material chemical states in X-ray microspectroscopy, which is robust to noise and spectral variability. Extensive experimental results on simulated and real datasets demonstrate its effectiveness and reliability.

Data-driven Battery Science Synchrotron X-rays

Utilizing hard x-ray holotomography and a morphology-informed neural network, we identified how particle behaviors correlate with performance and deterioration. Our insights reveal that both individual particles and their neighbors influence damage over time, guiding better electrode designs.

We measured the attributes of each NMC76 particle in terms of size, sphericity, SOC, SOC variation, and anisotropic polarization. Using machine learning, we identified the morphology most affected by the interphase. Our findings indicate that LiDFP more significantly suppresses SOC variation and anisotropic polarization in smaller, more spherical particles.

We investigate the structural deformation and mechanical damage of LIB composite LiNixMnyCozO2 (NMC, x + y + z = 1) cathode upon exposure to low-temperature conditions. Our results suggest that, in order to design batteries for use in a wide temperature range, it is critical to develop electrode components that are structurally and morphologically robust when the cell is switched between different temperatures.

By merging X-ray nanoprobe diffractive imaging with advanced machine learning, we effectively probe meso-scale heterogeneity and the evolution of lattice defects down to atomic details. This provides a clearer understanding of how crystalline battery materials react under external stimuli, offering empirical insights for defect-engineering strategies to enhance cathode resilience during intense battery use.

Using a modified Mask-RCNN model, we automatically identified and quantified over 650 NMC particles in high-resolution X-ray nano-tomography. Analysis shows particle detachment increases with charging rate, and smaller particles demonstrate more detachment variability from the carbon/binder matrix.

Computational Optical Microscopy

We introduce a non-iterative image deconvolution algorithm tailored for Poisson or mixed Poisson-Gaussian noise. Unlike current methods, ours simply involves solving a linear system, ensuring quick and precise solutions. Simulations across varied convolution kernels and noise intensities show our approach surpasses existing techniques in restoration quality and computational efficiency.

We introduce a method to estimate a microscope's spherically aberrated PSF directly from observed samples. By expressing the PSF with four basis functions and using a unique criterion based on microscope noise statistics, we derive the PSF from the acquired image. This principle is adaptable for non-spherical aberrations and various microscope modalities.

An accurate PSF model enhances performance in deconvolution microscopy and precision in single-molecule microscopy. We offer a rapid and precise approximation to the Gibson-Lanni model, representing its integral as a combination of rescaled Bessel functions. This eliminates integral-based calculations. Our method notably reduces computation time compared to quadrature methods and is adaptable to other microscopy PSF models.

See the full list here.