CPLUOS Astrophysics team meeting

Monday 6 Dec 2021, 16:30 → 17:55 Asia/Seoul

물리학과

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

COSMOLOGY

1 Dr.Sangnam Park's Report - FDM_Offset

Speaker: Sangnam Park

2 Hyeonmo's report - FDM vs CDM Halo Collision

Speaker: Hyeonmo Koo (University of Seoul)

• FDM vs CDM Halo Collision
  • Some Bumps in CDM Simulation
  • In Initial Velocity [28.184km/s] - Velocity Decrease [28.184km/s] graph,
    • A strange thing is seen in the velocity decrease curve of CDM, that as the initial velocity v_0 increases, Δv is not slight and decreases with a few bumps.
  • Oscillation Period of Particle
    • T_(1/2)=0.06τ_gadget
    • This value of 0.06 corresponds to the oscillation period according to time of r_(1/2) of the individual halo.
    • Since the phenomenon of two halos colliding is quite catastrophic, a perturbative interpretation is not possible.

3 Young's report. FoF and MST algorithm

Speaker: Young Ju

• FoF and MST algorithm
  • The density of 50 clusters and 5k, 10k, 50k noise level
  • The algorithms should find 49 cluster at least
  • Controlled random data 1 : mass function
    • 1. make halo mass function using colossus package
    • 2. using HOD convert from mass to number of galaxies
    • 3. calculate cumulative distribution and pick up number of galaxies using uniform random number
    • 4. make cluster distribution
  • Controlled random data 1 : noise data
    • total 4542 galaxies : 5k, 10k, 50k
  • Controlled random data 1 : 5k noise data
    • member > 50 , (MST, FoF,  MGS, DBSCAN, Hierarchical)
      • (17, 3, 47, 39, 17)
  • Controlled random data 1 : 50k noise data
    • member > 50 , (MST, FoF,  MGS, DBSCAN, Hierarchical)
      • (1, 3, 331, 1, 1)
  • Controlled random data 1 : 5k noise data, linking length = 0.3 x mean separation = 2.83
    • member > 50 , (MST, FoF,  MGS, DBSCAN, Hierarchical)
      • (38, 47, 19, 14, 38)
  • Controlled random data 1 : 50k noise data
    • member > 50 , (MST, FoF,  MGS, DBSCAN, Hierarchical)
      • (1, 3, 331, 1, 1)

4 CHOA(Cosmology of High-Order Statistics)

Speaker: Se Yeon Hwang

• CHOA(Cosmology of High-Order Statistics)
  • Running code using Random pairs > this should be go to 0
  • We check the 10 results
  • Estimator problem
    • In practice, for the N-point correlation function, we use estimator using pairs(DD, DR, RR)
    • There are many kinds of estimator
    • i.e. DD = (count pair)/N(N-1)
    • For getting N point correlation function, We are using estimator like this :
      • 2pcf = (DD/RR)-1
      • 3pcf = (DDD/RRR)-1
      • 4pcf = (DDDD/RRRR)-1
    • We are changing like this : using Landy&Szalay(1993)
      • 2pcf = (D-R)^2/RR
      • 3pcf = (D-R)^3/RRR
      • 4pcf = (D-R)^4/RRRR
  • What changes does the estimator make? (schematic examples on 2pcf
    • In many literatures, they said  LS estimator is more accurate
  • Methods
    • Preparing data pairs like this(referenced by Philcox(2021)
    • New catalog N which combined data and random points(α is normalized part), N = D-αR
    • New normalized random catalog, new R = αR
    • Because data catalog is already normalized, we turn off the normalized part in the code
    • So, when we calculate N pairs and R pairs,  we can get LS estimator
      • 2pcf = NN/RR = (D-αR)^2/(ααRR) = (DD - 2αDR+ααRR)/(ααRR)
      • 3pcf = NNN/RRR
      • 4pcf = NNNN/RRRR
  • When we checked through code
    • Ngram_nonorm is using N and normalized R data points. NN/RR
    • kstat(Ls) is using D and R. (DD-2DR+RR)/RR

5 Dr.Sabiu's Report - Density Reconstruction

Speaker: Dr Cristiano Sabiu (University of Seoul)

Density Reconstruction using ML

• Density Reconstruction
  • 21cm Part 2!
    • We will update the previous work with a more rigorous approach
    • Use realistic foreground maps AND recover input cosmology
    • Add warm DM (Neutrinos) and add flexibility to fiducial model ie (± N-σ around Planck values)
  • Determining the underlying density field from the Distribution and 1-D velocities of Galaxies in Redshift-space via Machine Learning

6 Discussion