Project:	Helping Euclid to see the Dark Matter halos of galaxies
Disciplines:	Astrophysics, Computer Science
Mentor:	Andreas Faisst, Dr., (PMA), afaisst@caltech.edu
AO Contact:	Narot Piric, nbahar@caltech.edu
Background:	Dark Matter has a significant impact on galaxy evolution. On it largest scales (the Dark Matter cosmic web), it determines the large scale structure of galaxies and helps funneling gas into the most massive galaxy clusters. On smaller scales (Dark Matter halos), it may helps reducing or shutting down the formation of stars in massive galaxies. As the name states, Dark Matter is invisible and can only be observed indirectly through gravity. A special case that allows us to measure the Dark Matter halo masses of galaxies is galaxy-galaxy strong lensing. In this case, a massive foreground galaxy bends the light of a (usually less massive) galaxy behind it. In the case of perfect line-of-sight alignment, this leads to a so-called "Einstein ring", in which the background sources galaxy is bend in a perfect circle. Off-axis alignments lead to arcs or isolated images of the background source. Importantly, the radius of the Einstein ring allows us to measure the enclosed total mass, which includes stars and Dark Matter. By subtracting the stellar mass component, we can measure the Dark Matter halo mass of the foreground lens galaxy. The new Euclid space telescope, is expected to identify more than 100,000 such galaxy-galaxy strong-lens systems. However, its spatial resolution is limiting the analysis of these strong lenses. This is because the foreground lens and the source (in the form of the Einstein ring) are blended together and therefore difficult to separate. It is therefore difficult for low-mass Dark Matter halos (forming a smaller Einstein ring) to disentangle the light of the massive lens and the background source. To resolve this issue there are several tricks that can be applied. First, by modeling the strong lens from first principles we can produce a spatial prior that will allow us to separate the light of the lens and the source in some cases spatially. Second, we can predict the brightness as a function of wavelength of the foreground lens from large statistical samples of unrelated galaxies, which allows us a better image subtraction of the lens. The student as part of this SURF will contribute to establishing a framework for the latter.
Description:	The student will use machine learning methods to predict the brightness as a function of wavelength (the so-called spectrum) of lens galaxies from their 3-band photometric points (meaning light measured at three different wavelengths). Specifically, we will use the photometry in the u, g, r, i, and z bands (spanning 0.35 to 0.8 micro-meters) as well as the spectra (continuously covering 0.35 to 0.8 micro-meters) from the Sloan Digital Sky Survey (SDSS), which covers large parts of the northern sky. Using unsupervised machine learning clustering methods (for example Self-Organizing Maps [SOM]), we will then define a mapping from these four colors to the spectra. For example, a red wavelength slope may be mapped to a smooth spectrum without additional emission lines. A bluer wavelength slope may be mapped to the spectrum of a galaxy that is forming stars showing strong emission lines. Having this mapping in hand, we can now use the Euclid photometric bands to directly predict a spectral prior of the lens. As part of this project, the student will specifically be responsible for gathering and downloading the spectra and 5-band photometry from SDSS and then for setting up the unsupervised machine learning framework.
References:	- Euclid Space Telescope: https://science.nasa.gov/mission/euclid/ - SDSS general information: https://www.sdss.org/ - SDSS Sky Server: https://skyserver.sdss.org/dr19 - Galaxy classification with Self Organizing Maps (SOM): https://ui.adsabs.harvard.edu/abs/2015ApJ...813...53M/abstract - Application of the SOM to galaxy spectra: https://ui.adsabs.harvard.edu/abs/2024ApJ...967...60J/abstract - Gravitational strong lensing: https://en.wikipedia.org/wiki/Strong_gravitational_lensing
Student Requirements:	- Interest in physics or astrophysics and knowledge of basic concepts is preferred. - Experience with Python programming and basic statistical concepts either in a class or independent projects is preferred. - Some experience with machine learning is preferred.
Programs:	This AO can be done under the following programs:   Program    Available To        SURF    both Caltech and non-Caltech students  Click on a program name for program info and application requirements.