1. Clustering, Proximity, and Balrog
Eric Suchyta
CCAPP Summer Series
01 July 2014

2. Outline Brief introduction to galaxy clustering
New software called Balrog, particularly as it relates to clustering
What happens to clustering when we include proximity effects?
Simplified demonstration of the problem, connected with real data, which we hope to solve with Balrog

3. The Universe’s matter is clustered
Credit: Millenium Simulation Project
Credit: SDSS

4. Clustering is often measured with correlation functions
Guo et al. 2013
e.g. the correlation function of galaxy number density has been measured

5. Anderson et al. 2013
Clustering measurements contribute to almost uncountable Astronomy analyses
Parejko et al. 2013
Morrison et al. 2012

6. Galaxies are a biased tracer of the dark matter distribution
Kaiser (1984) showed that objects which occupy only the highest density fluctuations are more strongly clustered compared to the overall matter distribution, by a ~ constant multiplicative bias factor
Guo et al. 2013

7. Science results are not robust until systematics are understood
Science results are not robust until systematics are understood. This is where Balrog enters.

8. The Balrog methodology is a very simple idea
Start with real imaging data
Draw GalSim (Jarvis et al. in prep) galaxy atop the image
Simulated galaxy inherits the possibly hard to model image properties by construction
Original
Balrog

9. The Balrog methodology is a very simple idea
Iterate many times to build up statistics
For any quantity, do we we get back what we put in? (Balrog is quite general.)
In today’s scope: are randomly placed Balrog galaxies harder to detect around the real galaxies? i.e. How do proximity effects influence clustering?

10. Proximity effects are not negligible in real data
DES - Preliminary

11. The usual way of thinking about galaxy clustering

12. But other galaxies also influence detection
y encodes the proximity effects,
e.g.Galaxies missed because of excess sky subtraction

13. All the terms in the observed galaxy-galaxy correlation function

14. All the terms in the observed galaxy-galaxy correlation function

15. All the terms in the observed galaxy-galaxy correlation function

16. Let us insert Balrog galaxies (B) at random points (i. e
Let us insert Balrog galaxies (B) at random points (i.e. no clustering to B)

17. Difficult, we can’t solve for all of these exactly
Let us insert Balrog galaxies (B) at random points (i.e. no clustering to B)
Difficult, we can’t solve for all of these exactly

18. Math aside, we know what the answer must look on scales >> proximity scale
Proximity effects are pronounced in the densest regions, reminiscent of how galaxies form in the densest regions.
Kaiser (1984) showed this leads to a bias on large scales between and
Drawing analogy, we expect a ~ constant bias factor between and on scales significantly larger than the proximity effect scale
Smaller scales…???

19. Test with simple simulations
Generate log normal density field
Proximity effects modeled by subtracting a portion of adjacent galaxies’ fluxes (improper background subtraction)
Making a flux cut causes some galaxies to now not be detected because of the proximity effects

20. Test with simple simulations
Recover a nearly constant large scale bias on large scales

21. Test with simple simulations
Recover a nearly constant large scale bias on large scales
But the bias between true and observed is much larger than you might naively guess from measuring Balrog-Balrog autocorrelation or Balrog-Observed cross correlation

22. Test with simple simulations
Recover a nearly constant large scale bias on large scales
But the bias between true and observed is much larger than you might naively guess from Balrog-Balrog autocorrelation or Balrog-Observed cross correlation

23. Connect with real data DES - Preliminary Simulation
We’ve used Balrog on real data
Balrog x Observed simulation has been tuned to same order of magnitude we measure in DES data
DES - Preliminary
Simulation

24. Wrapping up…
Proximity effects can introduce difficult to measure, non- negligible bias on clustering
The bias should be constant on large scales
We need to nail this down in the math, with suitable approximations, and then solve it with Balrog ? ?

26. Backup Slides

27. Detailed Balrog algorithm

28. Building a realistic simulation source population
We sample our galaxy properties from the COSMOS mock catalog (Jouvel et al., 2009)
COSMOS is very deep, but intentionally devoid of large, bright objects
For now one-component Sérsic models. No intrinsic shapes yet.

29. Test with simple simulations
The bias appears to very nearly be the fraction of galaxy’s missing
This is what one might naively guess for a constant bias factor: