New Insight Into the Dust Content of Galaxies Based on the

Analysis of the Optical Attenuation Curve

DustDust is an important component of a galaxy

• In situ, it affects the evolution of the galaxy by the role it plays in the process of star formation

• Observationally, it shapes the spectra through absorption and scattering of ultraviolet and optical photons and emission in the mid- and far-infrared

In a typical star-forming galaxy, almost 1/3 of the total luminosity is processed by dust into the IR

How to quantify dust attenuation?

In the case of a single point source (star, AGN) behind a dust Screen, attenuation is termed “extinction”

λA = 1.086

In the case of a whole galaxy, the term attenuation refers to the overall absorption (scattering) of photons emitted in all directions by stars in all location within the galaxy. It can always be written as:

a oλA =m - m

a om , m are attenuated (observed) and unattenuated magnitudes

To determine the dust attenuation of SDSSgalaxies, we followed the prescription of

Calzetti et al. 1994

- analyzed 39 starburst galaxies - as a measure of dust attenuation she used Balmer optical depth in H and H lines defined as:

- divided spectra into 6 lB bins

- averaged spectra in each bin (to average out the effect of different stellar population ages) - divided all spectra with spectrum that

corresponds to the smallest lB (pair method)

- at the end, averaged all spectra to get

attenuation curve

l l lB β α

/τ =τ -τ ln

2.88

H HL L

Derivation of the Attenuation Law of Starburst Galaxies(Calzetti et al. 1994)

This Work

Similar method (Calzetti et al., 1994) was applied in this work. It was done with real spectra and modeled spectra

1. Observed spectra (Christy Tremonti & Tim Heckman) - used ~10000 high quality (S/N>5) galaxy spectra from DR2

database of the SDSS catalog - AGNs were selected out 2. Modeled spectra (myself) - used ~ 5*105 stochastically generated galaxy spectra

Method of Selection

It is known that dust has the same effect on spectra as1. Age/SFH2. Metallicity3. IMF

Since “pair method” is based on comparison of same-type objectswe must break down (or at least to reduce) all these degeneracies withcareful selection of galaxies

Method of Selection IDn4000 measures the 4000 A – break good tracer of the mean stellar age

HA measure the absorption in Balmer line good tracer of recent burst of star formation

The selection was performed in the Dn4000-HA space. Spectra were divided into 10 boxes in the Dn4000-HA space (Same scales for comparison)

Binning & Averaging

Real spectra Modeled spectra

Spectra IN EACH BIN were divided into 7 Balmer optical depth bins in line ( ) and averaged.

Example for Bin 2

lBτ

Rescaled Optical Depth (Qn) is calculated

Fn and F1 are spectra in nth and 1st binH and H are luminosities in lines

n

n 1n line line

α β nBn B1

α β 1

F (λ)-ln( )

τ (λ) F (λ)Q = =...

(H /H )τ -τln( )

(H /H )

Interesting to note this bump in both(observation & model) cases near 4000AIn higher Balmer optical depths

lBτ

Dividing & Smoothing

All spectra were divided with the spectrum that corresponds to smallest Balmer optical depth (nearly dust-free) and smoothed on 200 A scale

Real spectra Modeled spectra

Effective Attenuation Law

The effective attenuation curve (Qeff) i.e. the weighted (with ) average of Qn for each box is calculated.

line lineBn B1τ -τ

Real spectra Modeled spectra

Reddening

The reddening experienced by old stars relative to young stars i.ec

B / lB for all 10 boxes.

Modeled spectraReal spectra

Attenuation curve

The Qeff/(Qeff(H)-Qeff(H) is calculated

observation model

Conclusion

1. Dividing spectra into boxes in the Dn4000-HA space enable us to derive the corresponding attenuation curves and to correct the individual spectra for dust attenuation

2. The remarkable similarity between real and modeled spectra enable us to analyze dust properties of galaxies in more detail

Future work

This analysis, in the case of observed spectra, was Performed using DR2 SDSS database. We plan to do it with DR4 Database also. We expect to improve the analysis in two way:

- boxes in the Dn4000-HA space will be more populated with objects even for large Dn4000 values

- if the number of objects in each box is large enough, we plan to bin them according to metallicities also