Discipline: Technology & Engineering
Subcategory: STEM Research
Hacene Boukari - Delaware State University
Co-Author(s): Noureddine Melikechi, University of Massachusetts, Lowell Gour Pati and Gary Holness, Delaware State University
The second phase of The Center for Research and Education in Optical Science and Applications (CREOSA) started in 2012 with three scientific sub-projects- (i) spectroscopy and imaging of biomacromolecules (BM’s) in crowded and complex media, (ii) spin polarization spectroscopy in nanodiamond for nanoscale sensing and imaging, (iii) interactive data mining in experimental optics. Faculty, staff and students from the department of physics and engineering, computer information sciences, chemistry and mathematical sciences had been working to fulfill the goals these sub-projects. During 2015-2016 cycle, the research team from sub-project 1, has investigated the rotational and translational diffusion properties of diverse nanoprobes (e.g. Alexa488, Fluorescein, BSA proteins, and Phycoerythrin) dispersed in Ficoll solutions, Polyethylene Glycol solutions, and Poly(Vinyl) alcohol (PVA) solutions and gels, using fluorescence confocal microscopy, fluorescence correlation spectroscopy (FCS), fluorescence anisotropy (FA), and laser induced breakdown spectroscopy (LIBS). The research team from sub-project 2, have performed a series of spectroscopic measurements to characterize individual fluorescent nanodiamonds (f-NDs) by identifying their stable, bright, broadband and anti-bunched fluorescence properties under the optical excitation. Fluorescence spectra of f-NDs containing a few nitrogen-vacancy (NV) defects have also been obtained to study two existing charge states (i.e. NV- and NV0) of these defects. Our recent experiments have also demonstrated efficient binding of protein biomolecules with nanodiamonds (NDs) and f-NDs which is necessary developing microbioassays for performing molecular-level detection. The research team from sub-project 3 investigated the efficacy of non-linear manifold methods for LIBS amino acid data analysis. In this work, suitable manifold neighborhood size for distinguishing inter and intra class variance was discovered. A new method, entropy density, was developed that uses information structure for the visualization of pattern phenomena in LIBS spectra. The team developed a new information theoretic metric, Chisini Jensen Shannon Divergences (CJSDs) that are effective with LIBS spectra as well as data domains characterized by stochasticity and subtlety among the classes. CJSD’s were used in the development of new SVM Kernels that achieve superior classification results on LIBS Amino Acid spectra. This resulted in a number of important discoveries on how distance measures are transformed through an approach called dilation. This work was validated on LIBS data, a synthetic data set controlling for subtlety, and an image data-set. Subproject 3 also contributed an approach for GPU acceleration of dimensionality reduction.
Funder Acknowledgement(s): This project was supported by the National Science Foundation's CREST Program - grant #HRD-1242067
Faculty Advisor: None Listed,