Comparing RAVE with theoretical models of the Milky Way

Sanjib Sharma

Joss Bland HawthornUniversity of Sydney

Outline

• Galaxia a code for generating a synthetic/mock survey according to a given galaxy evolution model

• A mock RAVE survey with Galaxia

• What matches and what does not.

• What can we learn with RAVE, constraining models with observations.

Motivation• A framework to compare theoretical

models of our Galaxy with observations.

Theoretical model

Observed Catalog

Theory of Stellar Evolution (Isochrones) Galaxia

Synthetic Catalog

Analytical N-body

Comparison

Monte Carlo Markov Chain,

Chi square, etc

(Extinction, Measurement Errors)

),,vx,,( mZf

Observational Space

l, b, r, μl, μb, vr, B, V, log(g)

(Age,Pos,Vel,Metallicity,Mass)

Drawbacks of current schemes

• Besancon Model- state of the art (Robin et al 2003)• Also Trilegal, (Girardi et al, Padova group)• Designed for simulating a particular line of sight

– at max 25 line of sights• Discrete (l,b,r) step sizes to be supplied by user• Not suitable for wide area surveys, or large catalog of stars

– takes too much time • No possibility to simulate substructures or incorporate N-

body models– Sagittarius dwarf galaxy, simulation of tidally disrupted

galaxies

Theoretical Model-Analytical Models

),,vx,,( mZf

),x,()vx,,()()(

Zffmm Zxv

Star Formation Rate SFR

Initial Mass Function IMF

2)(log

)(/))(log(log log

Z

ZZ Ze

Age Metallicity Relation AMR

Phase space distribution

Sampling Analytical Model(Von Neumann rejection sampling)

Adaptive Mesh (Barnes Hut Tree)

Optimization• To generate a patch do not need to

generate the full galaxy– If a survey is not all sky, first check

if a node intersects with survey geometry.

• Faint stars which dominate in number are visible only for nearby nodes.

• For far away nodes there is a minimum mass above which stars are visible– Sort nodes according to distance.

Calculate appropriate m’ – Generate only those stars that are

visible.

r

x

yz

mmin m’ mmax

r

Galaxia summary

• Analytical model for disc, with warp– Robin et al 2003 (Besancon model)

• Padova Isochrones – m >0.15 , Marigo et al 2008, Bertilli et al 1994– 0.07<m<0.15 Chabrier et al 2000

• 3d extinction model – double exponential disc with warp and flare, hR=4.4

kpc, hz=0.088 kpc –

E(B-V) at infinity match Schlegel et al 1998 or– 0.54 mag/kpc in solar neighborhood

Computational Performance

• Run time nearly linear with mass of the galaxy being simulated– Due to the use of adaptive mesh or node

• Speed- 0.16 million stars per second (2.44 GHz proc)– For shallower surveys a factor of 3 less

• V<20, 10,000 sq degrees towards NGP, 35 ×106 stars, 220 secs

• V<20 GAIA like survey 4 billion stars can be generated in 6 hours on a single CPU

A synthetic RAVE survey

• RAVE_internalA (S/N>20 , July, VDR3)• For each RAVE field match the number

distribution of stars in IDENIS magnitude (9-12).– Assumption RAVE is a magnitude limited survey– Number counts were matched per 0.25 mag bin– Proper sampling required about 50 million stars in

9<IDENIS<12• Add extinction, add observational errors

– Photometric errors for the time being only added for 2MASS J and K not IDENIS

– Stellar parameter errors taken from Siebert et al 2011• σ= σ(log(g),Teff)

All stars

Dwarfs

Giants

Red Clumps (Siebert et al 2011 criteria)

(NRAVE-Nmodel)/√(NRAVE+Nmodel)

• Log(g) Vs Teff

Checking the kinematics

• Let us assume we have faithfully reproduced the RAVE catalog in color magnitude space.

• For the time being let us take radial velocities only.

• Convert to U V W components and check the distribution.

• Contains information about USun, VSun,

WSun,, σv(R,τ), τ being age

X

y

•Values of USun VSun sensitive to weights used, range used to fit. Inner and outer regions cannot be matched simultaneously

This is a good match

which means σvz(τ) is approximately correct.

What needs to be done

• Need to move beyond simple one dimensional fits. Multidimensional parameter search. A model fitting machinery .

• Need to use distance information also.

• Simultaneous constraints from different surveys SDSS, 2MASS, Hipparcos etc.

Hunting for structures

• Multi-dimensional group finder EnLinK. – Density based hierarchical group finder – Uses nearest neighbor links to search for peaks and

valleys in density distribution.– Gives a statistical significance of a group

• Develop a similar scheme for finding groups where we look for differences between a given data and an expected smooth model.

• Exploit all the multidimensional information that RAVE provides.