Target Selection for DESpec

Filipe Abdalla, Stephanie Jouvel, With input from many from the DESpec target selection team.

A DESspec target selection?

1)100% spectroscopic completeness of DES galaxies to r=21stmagnitude with resolution 50 km/s.

2)Case 1 plus 50% completeness to r=22.5 magnitude evenly distributed over all redshift bins

3)~300 km/s redshift precision with 100% completeness to r=22ndmagnitude

Q: on what parameters give the most improvement on FoM and advice as to what new techniques are opened up.

Largest FOM’s are always with a constant number density.

1- Constant density 0.2<z<1.7, 10^7galaxies

2- Constant density 0.2<z<0.5, plus I<22.5 for 0.5<z<1.7 @65% eff’y. Total 10^7galaxies. Note redshiftcut-off (right)

3- Constant density 0.2<z<0.7, plus emission line galaxies for 0.7<z<1.7. Total 10^7galaxies.

Is FOM the best way to asses these

Large difference between different target selection techniques.

Best is to reach a magic number of n=~2.10^-4Mpc^-2

This magic number ranges from a few hundred to a few thousand galaxies per sqdeg depending on z range

LRG target selection

Strawman LRG target selection from Jim...

- Sensitivities for 30 mins exposure time on 2 arcsec ø fibers (70% light fraction)- Resolution R=2500 at 0.9um.- Calculation include DES throughput + sky noise and atmospheric lines from Hanuschik (2003)

Line sensitivities from Huan Lin

Line sensitivities from Huan Lin 0.6-1um

- Sensitivities for 30 mins exposure time on 2 arcsec ø fibers (70% light fraction)- Resolution R=2500 at 0.9um.- Calculation include DES throughput + sky noise and atmospheric lines from Hanuschik (2003)

Redshift distribution / em lines gal

Redshift distribution / em lines gal

Redshift distribution from em linesRedshift distribution of emission line galaxies selected with :- I (DES) < 23- 1 < photoz < 2- at least one em line detected at 5 sigma

Photoz with Le Phare (template fitting method) u DES+VISTA

Can we select the emission line galaxies based on DES+VISTA photometry only ?

Total of roughly1200 gal/deg2

Can we do color-color cut ?g-r vs r-i r-i vs i-z

It looks difficult to select the OII line in a color-color cut. We need a higher dimensional kind of cut such as photoz codes can provide

Multidimensional method based on NN selection

Photo-z’s, target selection and Neural networks:

Has an architecture: defined by a number of inputs/ outputs and nodes in hidden layers

Collister & Lahav 2004

http://www.star.ucl.ac.uk/~lahav/annz.html

Internally values range from 0 to 1 roughly

Use of NN to select em line galaxies

We assign a value of :- 1 for targetted em line galaxies- 0 for non targetted galaxies

We then use ANNz and train a NN on a sample of 10000 galaxies from the CMC (DES+VISTA)

Results for i(DES) < 24

Using Huan Lin sensitivities cut between 0.6-1um

Which ANNz selection criterion ?

Fig : Cumulated nbr galaxies/deg2 fct ANNz target probability

We suggest a selection criterion of 0.8 yielding to : -7500 emission line galaxies / deg2 i(DES)<24- and a contamination of 800 gal/deg2 i(DES)<24

0.8Using sensitivities cut between 0.6-1um

Which ANNz selection criterion ?

Solid : percentage of galaxies for which we will measure a redshift/total nbr gal & Dashed : SSR fct ANNz target probability

We suggest a selection criterion of 0.8 yielding to : -7500 emission line galaxies / deg2 i(DES)<24- and a contamination of 800 gal/deg2 i(DES)<24

Using sensitivities cut between 0.6-1um

Redshift distribution of ANNz sel 0.8Using sensitivities cut between 0.6-1um

Redshift distribution / em lines gal

Target selection

Target selection in wordsOut of the high redsfhit EL

galaxies, use the photo-z to obtain a nP=1 survey out to z=1.7 use photo-z from DES to “shape that distribution” ~1000 sq deg?

Obtain a high nP sample at low z as it overlaps the lensing kernel. ~1500 sq deg?

Some could be EL galaxies to have low bias for RSDs ~750 deg sq? Can be Ha or OIII lines

Some could be LRG’s & RG’s ~ 750 deg sq? so that we probe down the luminosity function of groups and relate lenses and lensing galaxies.

ConclusionWe simulated a DES+VISTA catalogue using the CMC

which include emission line fluxes.Used derived a fiber sensitivities including all instruments

troughput and sky noise including astmospheric lines for a DESpec instrument (Huan Lin).

Using these sensitivities we can select emission line galaxies from DES+VISTA photometry using ANNz. They are complementary to the LRG selection.

We suggest a selection criterion of 0.8 with a yield of 8000gal/deg2 for i(DES)<24 and a success rate of 93% or 5000gal/deg2 i(DES)<23 and a success rate of 95% ~ 1000 of which at z>1.

To do this would need a pilot survey to identify the successful galaxies or trust the mocks.

A color-color cut does not seem the optimal solution to select target for DESpec.