Studying the Interstellar Medium with Massive Spectroscopic … · 2014. 12. 14. · D. Baron (TAU) D. Watson (Dark) J. X. Prochaska (UCSC) P. E. Nugent (UCB) Y. Yao (NERSC) Needles

November, 2014

Dovi PoznanskiTel Aviv University

With:D. Baron (TAU)D. Watson (Dark)J. X. Prochaska (UCSC)P. E. Nugent (UCB)Y. Yao (NERSC)

Needles in Astronomical Haystacks

Studying the Interstellar Medium with Massive Spectroscopic Surveys

The Data - SDSS

Galactic

180

0 2084Counts

number of spectra per 3.35 deg2

Low spectral resolution

Low S/N

> 1,500,000 spectra

Cover 2/3 of sky

The basic idea

Over 400 bands seen.Carrier(s) unknown. Since 1922.

The Diffuse Interstellar Bands

Image: P. Jenniskens & F.-X. Desert

Method - “brute force”

“Clean” spectra

Divide the spectra according to rule

@ every λ calculate median

Done within SciDB

Compare resulting stacked spectra (measure lines etc.)

Let there be (less) light

5700 5720 5740 5760 5780 5800 5820 5840Wavelength (A)

Nor

mal

ised

flux

+co

nsta

nt

QSO � 0.1mr = 21mag

Stacked 70,000Stacked allSimulated

Correlate λ with amount of dust

Tile the Sky

Galactic

180

250 patches

Mostly 20–40 deg2

λ - λ

Clustering

dib index8.8

7.8

6.8

5.8

4.8

3.8

2.8

1.8

0.8

Num

ber

of�

from

0co

rrel

atio

n

DIB clustering - all DIBs

Questions:

Thanks!

Smarter stacking? Super-res?

Smarter classification methods?

Test robustness?

Open to new ideas.

[email protected]