Contents
Large-scale Structures
Mulguisin Cluster-finding Algorithm
As galaxy survey have been expanded enormously, it broadens our horizon of the universe. From the galaxy as the small scale of the universe to the Large-Scale Structure (LSS) as the largest complex system, we are trying to explore constituents of universe or an interaction between them. Among the cosmic structures, a galaxy cluster has important role to understand the universe. A galaxy cluster is based on matter distribution of the universe, in consequence we can track of nature of the universe. Despite a galaxy cluster has crucial capacity for the universe, a distinct definition of the cluster is ambiguous. There are lots of method to get galaxy clusters from a galaxy catalogue, in addition a criteria for parameters in the same method is varied by researcher.
For that reason, we have explored the possibility of developing an algorithm that creates galaxy clusters in a way that more closely resembles how the human eye and brain identify patterns. One approach we considered was to adapt jet-finding software used in high-energy particle physics research, with a particular focus on the MulGuisin (MGS) algorithm as a potentially suitable software for galaxy clustering. MulGuisin (물귀신) is a Korean word for a ghost that lives in water and is a figure that often appears in old Korean stories. The MGS algorithm started with the idea that the ghosts hiding in the water could be found in the order of height by simply draining the water from the lake.
The MulGuisin (MGS) algorithm is a powerful technique for identifying clusters in data from astrophysical simulations and observations. It consistently produces results closer to those inferred from human visual inspection. MGS also provides auxiliary topological information such as the number and length of connections for each galaxy. With this tool we can explore the use of this enhanced information in testing or constraining cosmological models.
Co-evolution of Galaxies and Large-scale structures
Galaxies and galaxy halos evolve while they are interact with the background large-scale structures. However, current semi-analytical galaxy formation models assume that galaxies form and evolve solely, which makes it important to study the interaction and co-evolutionary nature of galaxies and structures. In particular, filament structures, as transporters of continuous matter flows to nodes(clusters), could be interesting because they are unstable and many properties are still not known.
To study the effects of filaments on galaxies and halos, we run DisPerSE on the cosmological simulation data to extract the filaments. We examine their effects on halo evolution by analyzing them in phase-space. Currently, we are also trying to construct a "filament merger tree" in order to involve the evolution of structure itself.
Dark matter density map(color scheme), filament structures extracted by DisPerSE(red lines) and massive and less massive halos(red and blue circles each) around the most massive cluster in N-Cluster Run.
CLML(Cosmology with Large scale structure using Machine Learning)
With Large Scale Structure (LSS) data, we can extract information for predicting cosmological parameters. One statistical method is calculating two-point correlation function (2pcf) but we want to compare this method by using Machine Learning algorithm especially for 3d computer vision task. For computer vision task, Convolution Neural Network (CNN) is popular way to predict cosmological parameters but we want to introduce new algorithm in this field, Vision Transformer (ViT). We first used ViT algorithm for estimating cosmological parameters, Om, sigma8, w0 and S8 with lightcone data.
(Figure) red : CNN, blue : ViT, green : 2pcf
arXiv:2304.08192
with Higher-order statistics
In large scale structure, the distribution of galaxies can be expressed in the form of N - point correlation function(Npcf). It is the form of higher order statistic function which is calculated from the distance, angle and other measurements derived from the distribution of galaxies. In the plot there appears a shape called Baryon Acoustic Oscillation(BAO), which works as the standard ruler of the universe. Through the process of analyzing BAO, we can extract the background information and estimate the cosmological parameters such as matter density, amplitude of density fluctuations.
In oure study we calculate and use 2, 3, and 4 point correlation function. We have studied the Npcf by calculating it from simulations and comparing it with the SDSS CMASS observational result, used simple machine learning to train with different cosmological datas to estimate cosmological parameters and comparing it with deep learning result. Both resulted significant level of parameter estimation, and now studying about whether cosmological parameters can be predicted from interpolated Npcf.
Black Hawk
Black Hole Accretion and Wind Kinetic feedback galaxy simulations
AGN Triggering Mechanisms
One of the galaxies in the hydrodynamical zoom-in simulations (Choi et al. 2017), processed by Powderday with NIRCam Filters of James Webb Space Telescope.
Active Galactic Nuclei(AGN) are one of the most brightest objects in the sky. AGNs are known to be powered by the accretion of matter onto a supermassive black hole(SMBH) at the center of a galaxy. Galaxy Merger is considered as one of the possible triggers of BH gas accretion, but observations
SN Wind and Satellite Galaxies
Supernova is a powerful explosion of star, emitting it's mass and energy towards nearby space. During this emitting process it affects nearby interstellar medium.
We study the effect of SN winds in simulated satellite galaxies. In two simulations of experimental and control, later use fiducial conditions including AGN feedback, stellar feedback, snowplow SN feedback, metal cooling and others. The experimental simulation has the same condition with control, except in this simulation SN feedback is greatly reduced. By comparing the result of these 2 kind of simulations we're trying to see how the SN wind affects the mass distribution in satellites.
Fiducial_MrAGN model m0215, from the zoom - in simulations(Choi et al. 2017), processed by Pygad
Fiducial_weakSNwind model m0215, from the zoom - in simulations(Choi et al. 2017), processed by Pygad
Ultra-Light Dark Matter
Recently, (ΛCDM)-based cosmology explains large-scale(Mpc) structure of our universe well. However, there are still many tensions of ΛCDM for small-scale(kpc), such as Missing Satellite Problem (Fig.1), Cusp Core Problem (Fig.2). Much of them could be solved with adding baryonic effect, but some of them couldn't, which we call "Strong Tension".
To resolve them, for one of alternatives, the Ultra-Light Dark Matter(ULDM) of bosonic DM particle mass 1e-22eV/c^2 is suggested. This stabilizes to a Bose-Einstein-Condensate(BEC) solitonic system, which we regard as a ULDM halo. Also, this suppresses the linear power spectrum of high-k mode, which we can visually know in cosmological large-scale structure simulation (Fig.3).
We are only interested in ULDM without self-interaction, to focus on the wave-nature due to ultra-light mass, called "Fuzzy Dark Matter (FDM)".
Figure 1. Missing Satellite Problem
(B.Moore et al. 9907411)
The number of satellite galaxies near Milky Way-like galaxy in ΛCDM simulation is much more than observed.
Figure 2. Cusp-Core Problem
(S.H.Oh et al. 1011.2777)
The halo for ΛCDM model follows NFW profile which has a sharper central density than observed by rotational velocity of stars.
Figure 3. Comparison of Cosmological Large-Scale Structures
(H.Schieve et al. 1406.6586)
Left : Structure created by evolving a wave-nature dark matter.
Right : Structure created by cold dark matter run with GADGET2 simulation.
Head-on Collision of FDM/CDM Halos
KooHM_2023KPS_Spring.pdf