Scientific visualization tools are currently not optimized to create cinematic, production-quality representations of numerical data for the purpose of science communication. In our pipeline \texttt{Estra}, we outline a step-by-step process from a raw simulation into a finished render as a way to teach non-experts in the field of visualization how to achieve production-quality outputs on their own. We demonstrate feasibility of using the visual effects software Houdini for cinematic astrophysical data visualization, informed by machine learning clustering algorithms. To demonstrate the capabilities of this pipeline, we used a post-impact, thermally-equilibrated Moon-forming synestia from \cite{Lock18}. Our approach aims to identify "physically interpretable" clusters, where clusters identified in an appropriate phase space (e.g. here we use a temperature-entropy phase-space) correspond to physically meaningful structures within the simulation data. Clustering results can then be used to highlight these structures by informing the color-mapping process in a simplified Houdini software shading network, where dissimilar phase-space clusters are mapped to different color values for easier visual identification. Cluster information can also be used in 3D position space, via Houdini's Scene View, to aid in physical cluster finding, simulation prototyping, and data exploration. Our clustering-based renders are compared to those created by the Advanced Visualization Lab (AVL) team for the full dome show "Imagine the Moon" as proof of concept. With \texttt{Estra}, scientists have a tool to create their own production-quality, data-driven visualizations.

We apply a new deep learning technique to detect, classify, and deblend sources in multi-band astronomical images. We train and evaluate the performance of an artificial neural network built on the Mask R-CNN image processing framework, a general code for efficient object detection, classification, and instance segmentation. After evaluating the performance of our network against simulated ground truth images for star and galaxy classes, we find a precision of 92% at 80% recall for stars and a precision of 98% at 80% recall for galaxies in a typical field with $\sim30$ galaxies/arcmin$^2$. We investigate the deblending capability of our code, and find that clean deblends are handled robustly during object masking, even for significantly blended sources. This technique, or extensions using similar network architectures, may be applied to current and future deep imaging surveys such as LSST and WFIRST. Our code, Astro R-CNN, is publicly available at https://github.com/burke86/astro_rcnn.

We demonstrate physics-blind and purely data-based nucleosynthetic archetype discovery from heterogeneous stellar elemental abundance measurements. Archetype extraction is performed via non-negative matrix factorization (NMF). Because stellar elemental abundance measurements can be incomplete and sparsely sampled, we perform combined low-rank NMF and matrix completion on the Hypatia Catalog and JINAbase to extract archetypal elemental abundance patterns, the so-called endmembers, and impute the missing measurements. We qualitatively compare the discovered nucleosynthetic archetypes to patterns anticipated from direct astrophysical nucleosynthetic calculations. Our tentative results indicate that the physics-blind analysis relying entirely on the incomplete chemical mixing in star formation can discern nucleosynthetic archetypes associated with asymptotic giant branch (AGB) stellar sources of the s-process, thermonuclear supernovae, and non-thermonuclear supernovae with an r-process.

Using standard 1D-LTE model atmosphere analysis, we provide an in-depth investigation of iron abundance as derived from neutral and singly ionization iron lines (Fe {\scriptsize{I, II}}) in nearby star clusters. Specifically, we replicate the discrepancy regarding $\Delta$[Fe/H], wherein the difference of Fe {\scriptsize{II}} - Fe {\scriptsize{I}} increases for stars of the same cluster with decreasing $T_\mathrm{eff}$, reaching an astonishing 1.0 dex at $T_\mathrm{eff}$ $\thicksim$ 4000 K. Previous studies have investigated this anomaly in the Pleiades and Hyades clusters with no concrete solution. In this analysis, we probe two samples: 63 wide binary field stars where the primary star is of sun-like temperatures and the secondary is a K-dwarf, ranging from 4231 K $\leq$ $T_\mathrm{eff}$ $\leq$ 6453 K, and 33 Hyades stars of temperatures 4268 K $\leq$ $T_\mathrm{eff}$ $\leq$ 6072 K. Previous studies have found discrepancies on the order of 1.0 dex. However, we find that these studies have neglected line-blending effects of certain Fe {\scriptsize{II}} lines, namely $\lambda$ = \{4508.29 \AA, 4993.34 \AA, 5197.58 \AA, 5325.55 \AA, 5425.26 \AA, 6456.38 \AA\}. When these lines are removed from the line-list, we find $\Delta$[Fe/H] decreases to $\thicksim$ 0.6 dex in the field binaries and $\thicksim$ 0.3 dex in the Hyades. The reason for this remaining trend is investigated by probing NLTE effects, as well as age and activity considerations using Ca {\scriptsize{II}} H+K emission and Li absorption, but these results appear to be small to negligible.

Using high-quality spectra of the twin stars in the XO-2 binary system, we have detected significant differences in the chemical composition of their photospheres. The differences correlate strongly with the elements' dust condensation temperature. In XO-2N, volatiles are enhanced by about 0.015 dex and refractories are overabundant by up to 0.090 dex. On average, our error bar in relative abundance is 0.012 dex. We present an early metal-depletion scenario in which the formation of the gas giant planets known to exist around these stars is responsible for a 0.015 dex offset in the abundances of all elements while 20 M_Earth of non-detected rocky objects that formed around XO-2S explain the additional refractory-element difference. An alternative explanation involves the late accretion of at least 20 M_Earth of planet-like material by XO-2N, allegedly as a result of the migration of the hot Jupiter detected around that star. Dust cleansing by a nearby hot star as well as age or Galactic birthplace effects can be ruled out as valid explanations for this phenomenon.