Transcript ppt

Spectroscopic Surveys: Present
Ching-Wa Yip
The Johns Hopkins University
March 21st, 2007
Astronomical Spectroscopy & The Virtual Observatory, Workshop at ESAC 2007
Present Spectroscopic Surveys & Science Highlights
SDSS & Science
Needs and Challenges
On-going efforts at JHU
Spectroscopic Science & VO
Motivations: stellar populations and galaxy evolution
379,000 yr -- recombination; 1 Myr -- formation of first stars (population III);
1 Myr to 100 Myr -- reionization; 1 Gyr -- formation of galaxy and star (population II & I);
13.7 Gyr -- present day
• In the hierarchical galaxy formation, galaxies interact or merge with each other
through dark matter halo interactions.
• Star form and die along the way, resulting in the galaxy spectra we see.
• How well do the properties of galaxies locally and at high-redshift fit in the
scenario of hierarchical galaxy formation?
Stellar populations and galaxy evolution: e.g. K+A galaxies
+ A star absorption features?
The ``K+A'' (or post-starburst) galaxies show a mixed populations of K stars
and A stars (Dressler & Gunn 1983).
Attract many recent interests in studies of star formation rates in different
density environments, showing that high density environment suppresses
star formation. (Balogh et al. 2001; Martinez et al. 2002, Lewis et al. 2002; Gomez et al. 2003;
Blanton et al. 2003b; Hogg et al. 2003, 2004; Kauffmann et al. 2004; Blanton et al. 2005)
(The studies of) stellar populations in galaxies are intimately linked to large
scale structures.
Moore’s Law in Surveys
•The spectra in redshift surveys increase
rapidly per year, following nearly the
Moore's Law.
•The large number of spectra requires new
methods to consider the higher-dimensional
data sets objectively and automatically, and to
reduce the error rate, e.g., in classification.
Spectroscopic Surveys
• The surveys have recently starting to go deeper instead of more redshifts.
• Magnitude selections are used instead of color selections.
Magnitude Limit
0 to 0.3
r < 17.77 (for galaxies)
0 to 6
r < 19.1 (for quasars)
0.7 to 1.5
I < 24
0.2 to 4.0
I < 22.5
0.2 to 4.0
I < 24
0.2 to 4.0
I < 26
0.1 to 1.2
I < 22.5
1.4 to 3.0
color selection
Spectroscopic Surveys - Stars
SEGUE – the Sloan Extension for Galactic Understanding and Exploration
• It will map structure and stellar content of the Milky Way with
240,000 stars ( accuracy ~10 km/s, [Fe/H] accuracy ~0.3 dex)
by 2008.
RAVE – the Radial Velocity Experiment
• It will provide 1 million radial velocity and spectra (R ~ 7500)
by 2010.
Spectroscopic Surveys – Discovery Highlights
– High-z quasars (z ~ 6, Fan et al.)
– New Milky Way Stellar Streams and Satellites (SDSS collaboration)
– High-z galaxy populations (z = 1.4 – 5, 2-6 times larger then previous
Lyman-break galaxies using color-selection) (Le Fèvre et al.)
– Luminosity function evolution (Faber et al.): the number density of red
galaxies increased from z = 1 to now, where that for the blue galaxies is
nearly constant.
– Stellar dynamics of galaxy agree with dark matter model (Wyse et al.)
– Mass of the Milky Way (0.7 - 2.0 x 1012 solar) from high-velocity stars
• etc…
Spectroscopic Surveys – Discovery Highlights
VVDS found high-redshift (z = 1.4-5, look-back time 9-12 Gyr) galaxy populations
• 2-6 times larger then previous sample of Lyman-break galaxies.
• Composite spectra do not show strong Ly, implying that the galaxies may be dusty.
(Le Fèvre et al.)
Spectroscopic Surveys – Discovery Highlights
SDSS found high-redshift quasars (Fan et al.)
• Gunn-Peterson absorption trough is found – reionization time scales can be deduced.
• Combine with other studies, this discovery is also a strong evidence for the
important role of supermassive black hole (106 – 109 solar) in galaxy evolution.
• z = 6.42 QSO (SDSS J114816.64+525150.3)
• Carbon-monoxide (CO) image suggests
a bulge mass of 109 solar, inconsistent with
local bulge mass.
• Black hole forms before galaxy bulge?
(Walter et al.)
Spectroscopic Surveys – Discovery Highlights
SDSS found new stellar streams and satellites of the Milky Way Galaxy
• The Milky Way Galaxy has undergone a rich interaction/merging history with its satellites
during its formation.
† The spatial density of SDSS stars
around the North Galactic Cap. Most
distant stars (red), nearest stars (blue)
(Belokurov et al.)
The Sloan Digital Sky Survey
• The SDSS have produced the biggest map
of the nearby universe (200,000,000
celestial objects in a quarter of the sky).
• The SDSS and SDSS-II will together
provide spectra of approximately
1,000,000 galaxies,
100,000 QSOs,
500,000 stars.
• Fiber spectroscopy with R = 1800, 3800 –
9200 Å in the observed frame.
• It has produced 2000 publications.
• It has excellent user interfaces (e.g.
CasJobs, Thakar et al.) and active public
outreach (e.g. Roddick et al.).
(2.5m telescope in APO.)
(SDSS fiber plug plate.)
Using SDSS Spectra
Flux-limited survey:
– Need a volume limited sample to infer statistical properties on galaxy redshifts.
Fiber-fed survey:
– Aperture effects in physical parameters e.g. spectral type, line equivalent width.
– Spectrophotometry at present (DR5) it is 10% accurate.
– Spectra were not corrected against foreground extinction (e.g. using SFD map)
from the pipeline (DR5).
– Stellar absorptions: needed to be corrected if use pipeline values (e.g. constant
increment of 0.7 Å, Hopkins et al. 2003).
† The central 3 arcsec of the galaxy (green circle).
MPA/Garching galaxy value-added catalog (DR4)
Mass-metallicity relationship
– Galaxy evolution is not closed box, galactic wind
may play a role in driving out metal
Stellar mass of galaxies
– (Dn(4000), HA) as a stellar mass estimator
– Average galaxy stellar mass ~ 5x1010 solar
– Most galaxies lie on locus of smooth star
formation history
(Kauffmann et al. 2002)
(Tremonti et al. 2004)
Galaxy/Narrow Line AGN Host Classifications: Baldwin-Phillips-Terlevich
(BPT) Diagrams
(Kewley et al. 2006)
• BPT diagram is shown to be a good
way to classify ~85000 SDSS galaxies.
• Seyfert 2 and LINERs seem to form a
continuous sequence of decreasing
black hole accretion rate.
(Groves 2006)
Galaxy Classification by Eigenspectra: the eClass
• Decompose each observed galaxy spectrum into an orthogonal set of eigenfunction {ei}:
f = a0 e0 + a1 e1 + a2 e2 + ···
(Yip et al. 2004)
KL transform (or called PCA) on
observed galaxy spectra to get
eigenspectra (Connolly et al. 1995).
The Main Galaxy Sample (Strauss
2002), ~170,000 spectra
1st order: mean galaxy spectrum
(continuum is similar to stellar
populations usually found in early-type
2nd order: modulation of the
continuum slope
3rd order: discriminate post-starburst
activities in galaxies (Connolly et al.
Removing host galaxy component from QSO spectra
• To apply the eigenspectrum in host-galaxy removal, we construct the cleaned
QSO spectrum as follows
f (QSO; cleaned) = f (QSO; observed) – a2 e2(host galaxy)
† QSO spectrum (top black) is
decomposed into AGN (blue) and
host galaxy (green).
† QSO 2nd eigenspectrum (black) vs.
galaxy mean spectrum (red). (Yip et al. 2004)
(Vanden Berk et al. 2006)
Analyzing spectra
• Model based
– e.g. stellar population
synthesis models for galaxy
continua, Iron templates for
• Pro
• Empirical approach
– e.g. composite spectra to get
average properties,
eigenspectra to get
correlations and diversities.
• Pro
– easier to extract physical
parameters as described by
the models
• Con
– models are imperfect
• e.g. ingredient
– may not be the best way to
make discovery about the data
– does not rely on models
– may be easier to make
• Con
– need more interpretations to
relate to real physics
– may need to go back to
modeling later on
Present Challenges
• Model Galaxy Spectra
• Spectrophotometry is not perfect in observed or/and synthetic integrated
spectra of stellar populations.
• Model ingredients are incomplete, e.g. no AGB stars, metallicity
evolution is mostly neglected (e.g. PÉGASE, Bruzual & Charlot).
• Model QSO spectra
• Line widths can be different from line to line, and object to object.
• Line blue-shift (may be due to outflows) (Richards et al.).
• Pure emission/absorption lines in QSO spectra are difficult to be
isolated because of iron emissions.
• Multi-wavelength, -epoch, -catalog
• Cross-calibration can be tricky, e.g. how to separate systematical from
statistical errors? (Talks on the 2nd day)
Stellar Population Synthesis on Galaxy Spectra
• It is important to sample the full probability distribution of model parameters.
† Estimated parameters:
MCMC vs. chisq minimization.
Synthetic spectra are
generated from PEGASE.
Mean parameter errors from
chisq-min can be a factor of
couple inferior.
† True vs. estimated
parameters using MCMC,
S/N = 30 for the synthetic
(Yip et al. 2007 in prep)
Stellar Population Synthesis on Galaxy Spectra:
Parameter Degeneracy
† Degeneracy involving intrinsic dust
extinction of some galaxies can be more
dominating than the infamous age-metallicity
degeneracy of old stellar population.
† 1-sigma confidence contour between
the estimated stellar population
parameters of a galaxy, in age vs. Z,
age vs. e-folding time, and age vs. E(B-V).
Template Fitting: QSO & Iron contamination
• Iron has many excitation levels making the theoretical modeling complicated.
• (23,000 FeII transitions in over 800 fine-structure levels, Sigut and Pradhan 1995)
• One approach is to fit iron templates to QSO spectra.
• Iron templates (derived from Zwicky 1 spectra)
• in the optical, Veron-Cetty and Veron 2004
• in the UV, Vestergaard and Wilkes 2001
• Iron templates (through radiative transfer)
• in the UV-optical, Sigut and Pradhan 2003
† Observed QSO (black)
Best fit iron template + continuum (orange)
Masking in fitting (blue)
† QSO spectrum with iron lines removed (black)
Refining spectrophotometry in SDSS galaxies
• We adopt 2 methods to remove dependent systematics between the
spectra at two epochs:
- Wilhite et al. (2005) to remove dependent systematics between the
spectra at two epochs
- similar to Filipenkko (1982) approach to
remove atmospheric differential refraction
of stars, but using observed light profiles
of galaxies instead of the King profile
• Both the absolute continuum band
scaling and the [OIII] absolute
calibration (Peterson et al. 1998) are
tested to remove -independent
normalization between the spectra at
two epochs.
† The calibration spectrum for plate 390, SDSS.
Refining spectrophotometry in SDSS: correct against
atmospheric differential refraction using realistic galaxy light profiles
refractive index of the Earth atmosphere
n = n (air temp, air pressure)
Earth Atmosphere
R = R() – R(5000)
= 206265 [n() – n(5000)] tan(z)
Refining spectrophotometry in SDSS galaxies
† Flux S/N vs. fractional difference between
twice observed spectra.
The SDSS provides on average 10% spectrophotometry.
DR6 will have improved spectrophotometry (Schlegel et al.).
Spectrophotometry refinement on DR5 galaxies (JHU/Pitt) – a factor of 5
improvements in repeatability (before: 3.2% to after: 0.5%) is achieved.
Refining spectrophotometry in SDSS galaxies
3% difference
† Observed spectra at 2 epochs (top).
Difference spectrum between them (bottom).
0% difference
Set the standard. Remove discrepancies among astronomers.
Help in error checking: converge faster to the right solutions.
Fast searching and cross matching multiple catalogs.
Nice tools have already been available, so we can focus on science:
– VO Services (Budavári)
• Searching tools
• Spectrum Services (Dobos et al. 2004, talk this afternoon)
• Filter Services
• Cosmological Services
• Footprint services
– Specview (Busko)
– Your work here…
• On-going efforts at JHU
– Spectrophotometry in the SDSS galaxies have been improved
by a factor of 5 for the repeated observations.
– Modeling of QSO continuum and line kinematics.
– We demonstrate the importance of sampling the full probability
distribution of parameters in the stellar population synthesis
modeling of galaxy spectra (i.e. MCMC in general give different
results from chisq minimization).
– Degeneracies between intrinsic dust extinction and other
parameters estimated from some galaxy spectra are found to be
more dominating than the infamous age-metallicity degeneracy.
• Spectroscopic surveys are essential for understanding galaxy
• They provide a platform for science discovery.
• Surveys like SDSS and before have followed Moore’s Law.
• Now they tend to go deeper for higher-redshifts objects.
• Many interesting science have been done using spectra, but more
challenges will likely call for more innovative approaches.