galaxy color matching in catalogs

Galaxy Color Matchingin CatalogsBryce Kalmbach

University of Washington

What are we doing?

• Finding best fit model SEDs for galactic catalog objects

• Need SEDs to provide observational catalogs

• Link between cosmological simulations and working science groups

Matching Algorithm• Calculate colors for model SEDs we want to

match– Use tools in sims_photUtils

• Find best least-squares fit across all colors for each catalog object– See readGalfast in sims_photUtils for example

Sample Matching Result

Current SED Models• Bruzual and Charlot (2003) with Chabrier (2003) IMF

• 4 different Star Formation Histories:• Burst• Constant• Exp• Instant

• Age grid from 1.585 Myr to 12.5 Gyr

• Metallicity from .5% to 250% Z_Solar using Padova (1994) isochrones

B & C Model Coverage

B & C Model Coverage

B & C Model Coverage

Galacticus Catalog

• Currently working with galacticus catalogs– Developed by Andrew Benson (see Benson 2010)

• Does not seem to match well with B&C SEDs

Comparing Galacticus

Comparing Galacticus

Comparing Galacticus

Need Better Coverage

• Should we get new SEDs?– FSPS (Conroy, Gunn & White 2009)

Comparing with FSPS

Comparing with FSPS

Comparing with FSPS

Need Better Coverage

• Should we get new SEDs?– FSPS (Conroy, Gunn & White 2009)

• Refine the coverage of our grid?

Changing Grid Coverage

Current Issues

• Bluer catalog objects than can currently match to SEDs– Single Star Populations?

Individual Stars 10Myr

(J. Dalcanton)

Current Issues

• Bluer catalog objects than can currently match to SEDs– Single Star Populations?

• Need more statistics from galaxy catalog– Will be provided in next run

Future Work

• PCA (Principal Component Analysis)– Determine axes of maximum variance and use

these as new basis vectors– Reduce Dimensionality• Storage Savings

Capture Information in Few Components

Capture Information in Few Components

Capture Information in Few Components

99.8% Information in 10 Principal Components…but…

Now 99.99999%, unfortunately with 2x components, but good color match

Future Work

• PCA (Principal Component Analysis)– Determine axes of maximum variance and use

these as new basis functions– Reduce Dimensionality• Storage Savings• Challenge: What is the minimum number of

components we can get the maximum amount of accuracy from?

Future Work

• PCA (Principal Component Analysis)– Continuous coverage of sample space rather than

grid

PCA will provide continuum

Future Work

• PCA (Principal Component Analysis)– Continuous coverage of sample space rather than

grid• Challenge: How do we sample to get the best set of

eigenspectra? • Challenge: How do we find the eigenvalues that

generate an SED that best matches the object color?

– Will need new methods that can combine linear combinations of eigenspectra

Thank You!

Contact: brycek@uw.edu