Events on Friday, August 8th, 2025
- Characterization of Tearing Mode Dynamics with the Radial Interferometer-Polarimeter on DIII-D
- Time: 10:00 am - 12:00 pm
- Place: B343 Sterling or
- Speaker: Rachel Myers, Physics PhD Graduate Student
- Abstract: The Faraday-effect Radial Interferometer-Polarimeter (RIP) diagnostic on DIII-D measures line-integrated magnetohydrodynamic fluctuations, such as tearing modes, at all radii, including modes not detected with external sensing coils. RIP results are compared to a synthetic line-integrated eigenfunction, corresponding to the mode amplitude on coils, using the linear code TJ. Even modes are finite on RIP, where the cylindrical approximation shows them vanishing. It is confirmed that RIP measures odd sidebands generated by linear coupling to toroidicity, which scales with inverse aspect ratio, and shaping. Experimentally measured shaping effects on RIP and tearing stability in inner-wall limited plasmas, which have more shape flexibility, are presented. Stability is calculated with DCON, which relies on equilibrium reconstructions with kineticEFIT. Modes grow as they approach the ideal-wall stability limit, which is near the classical tearing limit. These low-safety factor plasmas were dominated by sawteeth, which did not trigger tearing modes as expected. Tearing modes were driven as the plasma approached the ideal-wall limit, as opposed to the triggered, pressure-driven neoclassical tearing modes found in higher-performance plasmas. Even modes were less attenuated than previously measured because low safety factor induces a strong, detectable odd sideband to the even mode. RIP amplitudes varied over time due to nonlinear effects. RIP also experimentally confirms nonlinear coupling between edge harmonic oscillation (EHO) components in quiescent H-mode (QH-mode), previously predicted by nonlinear MHD simulations. We observe with RIP, but not sensing coils, nonlinear coupling of the EHO to core-resonant tearing modes, which interrupts the EHO and prevents access to long-lived QH-mode. This is concerning because QH-mode has an H-mode pedestal without edge localized modes (ELMs), which is attractive for ITER or a fusion reactor. Bicoherence values above 99% significance reveal that the first three EHO components are coupled throughout the QH phase, reflecting the cascade that sustains the EHO. In a case where QH-mode ceases and ELMy H-mode arises, we detect with RIP, but not with the coils, an intermittently returning EHO nonlinearly coupled to two core-resonant tearing modes. Coupling to tearing modes has caused the coupling between EHO components to vanish. The EHO becomes intermittent, while ELMs persist.
- Host: John Sarff/Brett Chapman
- Discovery and Characterization of Strong Gravitational Lenses in the Dark Energy Survey with Machine Learning
- Time: 11:00 am - 1:00 pm
- Place: 1310 Sterling or
- Speaker: Jimena Gonzalez, Physics PhD Graduate Student
- Abstract: Strong gravitational lensing, where a massive galaxy bends and magnifies the light from a more distant source, offers a unique window into the underlying cosmology of the universe and serves as a powerful tool for studying dark energy. Yet identifying and analyzing these rare systems within the vast datasets produced by modern surveys remains a major challenge. In this talk, I will present my thesis work developing new machine learning and hybrid techniques to make strong lens discovery and modeling more efficient. First, I introduce a machine learning–based search that uncovered hundreds of strong lensing candidates in the Dark Energy Survey. Next, I compare three independent machine learning searches applied to the same dataset and show how combining them improves performance and reduces missed discoveries. Finally, I describe a new pipeline that integrates machine learning with traditional parametric modeling, dramatically reducing the time required to model individual systems while maintaining accuracy. Together, these projects provide scalable, automated approaches for discovering and characterizing strong gravitational lenses, maximizing their potential as cosmological probes in the new era of large-scale surveys such as LSST and Euclid.
- Host: Keith Bechtol
- Synthetic source injection in photometric surveys for cosmology
- Time: 2:00 pm - 4:00 pm
- Place: B343 Sterling or
- Speaker: Julian Beas-Gonzalez, Physics PhD Graduate Student
- Abstract: Synthetic source injection (SSI) is an extremely useful technique for the validation of photometric survey data. It consists in artificially inserting objects, such as galaxies and stars, with well-known properties into real astronomical images. By processing images with injected sources, we can obtain measurements of the observed properties of these sources and compare them to their true properties. The transfer function between observed and true values allows us to quantify the accuracy of image processing pipelines in recovering the real properties of targets in a survey. Given the high level of precision that current and future large-scale structure surveys are expected to achieve in constraining cosmological parameters, a robust characterization of uncertainties is necessary, in which SSI plays a critical role. Furthermore, SSI is helpful in simulating weak-lensing magnification effects, as well as in providing a way to infer the photometric redshifts of wide-field objects by extrapolating a color-redshift relationship from injected deep-field objects. The quantities derived from these applications are paramount to the cosmological analysis, making SSI an important part of the process. In this talk, I will describe Balrog, the SSI effort in the Dark Energy Survey (DES) Y6 cosmological analysis, in which we injected real sources from the deep fields over the entire 5000 deg^2 wide-field survey footprint. I will explain the process of injecting the sources and validating the injected sample, as well as the applications in the DES Y6 weak-lensing magnification and photometric redshift calibration projects. Finally, I will highlight ongoing work on SSI for the Legacy Survey of Space and Time (LSST), particularly on the recently released Data Preview 1 (DP1), as well as future potential applications.
- Host: Keith Bechtol