CPLUOS Astrophysics team meeting

Friday 8 Apr 2022, 14:00 → 15:10 Asia/Seoul

물리학과

https://uos-ac-kr.zoom.us/j/8264902652

14:00 → 15:00 COSMOLOGY

14:00 Hyeonmo's Report - Head-on Collsion of FDM/CDM Halos(POSTER)

Speaker: Hyeonmo Koo (University of Seoul)

Head-on Collision of Fuzzy/Cold Dark Matter Halos

• Abstract
• Collision of Fuzzy Dark Matter Halos
• Collision of Cold Dark Matter Halos
• Results & Theoretical Interpretation
  • The appearance of some bumps in Δv_CDM is due to the interference as head-on collision of two halos which has the characteristic time scale, the period for oscillating r_1/2.
• -34 ~ 0 ~ 34 → -20 ~ 0 ~ 20 because of wide [kpc]
• FDM profile spreads over, while CDM profile is concentrated. Of course, it might be just a visual misconception, because the density profile from CDM cannot reconstruct low-dense region due to the lack of mass particle. This is one of the main reason why he should draw only sufficiently dense region for both FDM and CDM.
• As a result, FDM profile and CDM profile should be calculated with high density in dense area.

14:10 Young's report. FoF and MST algorithm

Speaker: Young Ju

Mulguisin Clustering Algorithm I. Comparison of Clustering Algorithms for Study of Cosmic Structure Finding

• What is the cluster? → We need Cluster finding algorithm
• MulGuiSin algorithm
  • We propose a new cluster finding algorithm, MulGuiSin(MGS)
• Schematic idea of MGS
  • 1. MGS with different height hide into the water tank
  • 2. We drain water from the water tank
  • 3. The highest MGS first appears
  • 4. As draining consecutively, the next highest MGS appears
  • 4. Keep draining until there is no remained MGS
  • → The MGS trace galaxies top-bottom approach
• MGS algorithm
  • 1. FInd galaxy with the highest density
  • 2. The galaxy becomes a new cluster, "MGS"
  • 3. FInd the next galaxy with second highest density and estimate the distance between "MGS" and next galaxy
  • 4. If the distance is less than our criteria, the second galaxy is merged with "MGS"
  • 5. This process continues until there is no galaxy
• Features of MGS algorithm
  • 1. MGS can find structure from galaxy data in detail
  • 2. Provide topological information
    • Number of nodes
    • Number of branches
    • Number of children
    • Link length
    • Average node generation
    • ...
• What is the difference between MGS and other algorithm?
  • Traditional algorithm : FoF (Friends of Friends)
  • Graph based algorithm : MST (Minimum spanning tree)
  • Machine learning (ML) algorithm :  DBSCAN (Density-based spatial clustering of applications with noise)
• Performance test
  • We test performance of MGS :
  • Test : How many cluster are returned
  • 1. 2 controlled data
    • a. simple gaussian model
    • b. power-law model & noise
  • 2. Comparing with other algorithms - FoF, MST, DBSCAN
  • Controlled data 1 : simple gaussian model
    • Check the number of clusters with changing linking-length
  • Controlled data 2 : power-law model
    • 50 halos ; 7041 galaxies
    • MGS find well number of cluster when linking length > 4
    • MST find less number of cluster than MGS
    • FoF and DBSCAN is similar with MST
    • Several cluster is very close to each other, other algorithms find some cluster to one giant cluster
  • Controlled data 2 : power-law model & noise
    • Add the noise into the controlled data 2
    • We check number of clusters changing with linking-length
    • The minimum member of cluster is 50
    • MGS find well number of cluster when linking length > 4
    • MST, FoF and DBSCAN find less number of cluster than MGS
    • Several cluster are very close to each other, other algorithms find some cluster to one giant cluster
    • For linking length > 12, the number of cluster in larger than 50 for MGS
    • Lots of noise become cluster, while the other algorithm make one giant cluster
    • In the high noise, the more noise become new clusters
    • In contrast, the other algorithms make one giant cluster
    • This shows that the MGS is very different algorithm comparing with other 3 algorithms
    • In the beginning, we introduce that MGS find cluster in detail
    • The other algorithms find cluster as one giant cluster when linking length is large
    • In contrast, the MGS always trace structure of cluster. They do not make one giant cluster
    • It is very different feature comparing with other algorithms
• Conclusion
  • We propose a new cluster finding algorithm MGS
  • The MGS shows good performance in the controlled data comparing with other algorithms
  • MGS can provide lots of topological information, it can give us other perspective of research for large-scale structure

14:20 Hannah's Report

See the slides from 8p

Speaker: Hannah Jhee (University of Seoul)

한국천문학회_구두_지한나 복사본.pdf

Tracking Halo Orbits and Their Mass Evolution around the Large-scale Filaments

• 1. Motivation
• 2. Data and Method
• 3. Results
  • 3.1. Trajectories in the Phase-space
  • 3.2. Virialization of Halos
  • 3.3. Mass Evolution of Halos
  • 3.4. Mass segregation
    • Massive halos arrive earlier, less massive later
      • The fraction of massive halos is lower when farther from the filaments
      • Massive crosser halos lose their kinetic energy and sink in(consistent with observation)
        • 1. Due to mass evol. in filaments?
        • 2. Due to arrivals of less-massive halos?
• 4. Summary
  • 1. Halos show a similar trajectory in perpendicular phase-space.
  • 2. Halos are virialized in filament environments after al least 6 Gyr since the first pericenter crossing.
  • 3. Halos grow in mass as they approach filaments, and will lose mass if the environment is harsh enough.
  • 4. Mass segregation of halos around the filaments in mostly caused by massive halos approaching faster than less massive ones, and dynamical friction plays a role for crossers.

14:30 CHOA(Cosmology of High-Order Statistics) (Poster with CLML, NPCF)

Speaker: Sumi Kim (University of Seoul)

14:40 CLML

Cosmology with Large scale structure using Machine Learning

Speaker: Se Yeon Hwang

Comparing Cosmological Information from Deep Learning and Higher Order Statistics

• Abstract
• Data
• CNN(Convolutional Neural Networks) model
• Deep Learning Result
• Correlation Function Result
• Comparison : DL vs N-PCF

14:50 Dr.Sabiu's Report - N-PCF

Speaker: Dr Cristiano Sabiu (University of Seoul)

15:00 → 15:10 Discussion