Observations of galaxies

• Except for the nearest galaxies (how near?), we are limited to studies of integrated light from component stars.
  • Studying resolved stars in nearby galaxies can provide important clues for understanding observations of other galaxies
  • Stellar populations
  • Surface brightness from star counts
• Photometric measurements: surface and/or total brightnesses
• Spectroscopic measurements: line (or feature) strengths and/or locations/profiles (for velocities)
• Morphology

Photometric

• Basic measured property is surface brightness ( $ergs/cm^2/s/sterradian, mag/arcsec^2$) since galaxies are resolved: note SB independent of distance until geometry of universe becomes important ( $SB\propto (1+z)^{-4}$).
• In principle, surface brightness measured directly from 2D detector as arbitrary function of location
  • However, galaxies are relatively faint. Typical spiral central SB is $V\sim 20-21$ mag/square arcsec, not too much brighter than typical dark sky; elliptical central SBs are higher (16-17). More than 1/2 light from galaxies comes from regions with $SB<sky SB$, so S/N is difficult to obtain.
  • usually employ averaging technique with some assumption about shape of galaxies, e.g., that they are elliptical. Note that ellipses appear to represent elliptical galaxies fairly well (but not perfectly), and also that ellipses are what circular disks look like when viewed at arbitrary inclination angle (but also it is obvious that spiral galaxies are not axisymmetric). Fit elliptical isophotes (e.g., Kent, ApJ 266, 562 (1983), Lauer, ApJ 311, 34).
  • Problems of seeing and sky determination for determination of SB distribution in center and outer parts of galaxies. Note sky problem worse with small detectors.

• Basic structure of galaxies: two general types of luminous objects, spheroids and disks.
• Spheroids generally characterized by
  • deVaucouleurs profile
    
    $$\Sigma(r) = \Sigma_e \exp({-7.67[({r\over r_e})^{1/4} - 1]})$$
    
    
    ($r_e$ is half-light radius if model is extrapolated to infinity, $\Sigma_e$ is surface brightness at effective radius, $\Sigma_0 \sim 2000\Sigma_e$)
  • King model (truncated Gaussian for velocity distribution),
  • Hubble profile:
    
    $$\Sigma(r) = {\Sigma_s \over (1+{r\over r_s})^2}$$
    
    
    ( $\Sigma_s = 0.25\Sigma_0$).
  • Note that King model is only profile with theoretical basis: others are empirical - Hubble law doesn't converge.
  • None are perfect matches to data over all scales. ( See Burkert (1993) for comparison with deVaucouleurs law).
  • not all spheroids are necessarily best represented by these laws: e.g., spiral bulges, low luminosity Es may be better represented by exponentials.
  • at small radii, ellipticals are not homogeneous in the profiles; to be discussed later

• Disks usually well represented on average by exponential:
  
  $$\Sigma(r) = \Sigma_s \exp{-{r\over r_s}}$$
  
  
  • Most spiral galaxies are a combination of a disk plus a spheroid
  • Different families of disk profiles exist, e.g., Freeman type I and type II profiles, may be related to a bar or dust

• Note that both components have distributions of the form:
  
  $$\Sigma(r) = \Sigma(0) \exp{-({r\over r_s})^{1/\beta})}$$
  
  
  where $\beta=4$ for ellipticals and $\beta=1$ for disks. This is sometimes known as a Sersic law.
• Of course, these profiles are implicitly one-dimensional, and in reality, galaxies have more complex features, e.g. asymmetries, departures from elliptical isophotes, bars, spiral arms, etc.
  • Often described by amplitudes of Fourier moments of intensity distribution along circular/elliptical contours, e.g. $a_1$, $a_2$, $a_4$,
    
    $$I(r,\theta) = \sum c_m \cos m\theta + sum s_m \sin m\theta$$
    
    
    
    
    $$a_m = \sqrt{ (c_m^2 + s_m^2)}/I_0$$
    
    

• Integrated brightness:
  • metric mags (brightness within an aperture of fixed angular or physical size)
  • isophotal mags (brightness within a specified surface brightness contour, e.g., Holmberg mags, which is the brightness within $\mu_{pg}=26.5$)
  • total mags (require assumption about outer SB profile)
  • Petrosian mags (mag within some fixed number of Petrosian radii (SDSS uses 2), where latter is defined as radius at which SB (in an annulus, SDSS used 0.8-1.25r) drops to some fraction of light (SDSS uses 0.2) within this radius). Advantages vis a vis surface brightness (distance) and seeing effects. Note relation of Petrosian mag to model mag depends on SB profile.

• galaxy SB and total mags depend on distance/redshift: surface brightness ($(1+z)^{-4}$) and integrated magnitudes (depends on cosmology)
• absolute mags ($M_V$) range from -22 to -8
• dependence of mags on wavelength and thus
  • redshift (K-correction, e.g. Hogg et al. astro-ph/0210394, Kinney et al, ApJ 467, 38 (1996), Coleman, Wu, Weedman, ApJS 43, 393 (1980))
  • foreground extinction (Burstein & Heiles ,ApJS 54,33 (1984) Schlegel, Finkbeiner, & Davis, ApJ 500, 525 (1998)),
  • internal extinction/inclination which can be very significant (up to 1.5 mag in RC3 prescription!).

• Sizes of galaxies:
  • isophotal ($D_{25}$ in RC is B=25.0)
  • model dependent sizes (e.g., $R_e$, $r_s$)
  • Petrosian radius
  • typical angular sizes: arcminutes for nearby galaxies, arcsec for distant galaxies.
  • get linear sizes with distance estimate (depends on cosmology)
  • typical physical size (of luminous component!): 1-30 kpc

Spectroscopic

• typical optical spectra in Kennicut 1992, ApJS 79,255 (ellipticals, spirals, irregulars , annotated by Vogt), ASTR505 summary.
  • spectra are composite - line strengths at different wavelengths measure different types of stars, and continuum almost always comes from a range of stars of different temperatures.
    • note that lines are significantly broader than for stars because of stellar velocity dispersion; high resolution spectroscopy to isolate individual lines generally isn't possible!
    • Measuring line strengths using indices (e.g., Lick indices): more discussion later about extracting stellar population information (distributions of masses, ages, compositions) from these.
  • No direct abundance measurements possible except in gas, because stellar spectra are composite
    • Even for gas, abundances are not trivial to measure
    • Underlying stellar absorption, esp. for H lines
    • Need gas temperature to get abundance. Lines required for this, e.g. [NII]5755, [OIII]4363, are generally very faint
    • Abundance often inferred from $R_{23}$ parameter (see, e.g. Kewley & Dopita, ApJS 142, 35, 2002):
      
      $$R_{23} = {F([OII]3727) + F([OIII]4959) + F([OIII]5007)\over F(H\beta 4861)}$$
      
      
      However, $R_23$ is double-value for some values!
    • Reddening is also an issue; often derived from $H\alpha$ to $H\beta$ ratio
  • Measure line positions (often done using cross correlation, especially for absorption line spectra, works at surprisingly low S/N, e.g., Tonry and Davis AJ 84, 1511 (1979)) to get redshifts/rotation curves.
  • Measuring velocity dispersions (e.g., Dalle Ore et al, ApJ 366, 38).
    • At high S/N, can consider measuring higher order moments of line as well as probes of stellar velocity anisotropies
  • Recall difficulty of measurements outside of nuclear regions due to low surface brightness; also remember that there are likely spectral/population gradients!

• HI measurements: 21 cm line
  • HI flux at a given location gives column density (atoms/surface area)
  • Total HI flux (integrated over galaxy), along with distance gives total HI mass: $M_{HI}/M_{sun} = 2.36\times 10^5 ({D\over Mpc})^2 F_{HI}$
  • Measuring galaxy rotation - optically and using HI (both spatially resolved, e.g. VLA, and unresolved data) (some rotation data from Vogt el al AJ 127, 3300 (2004)
    • Rotation often characterized by $v_{max}$: maximum of rotation curve (sometimes hard to measure)
    • From unresolved HI profiles, rotation characterized by line widths, e.g. $W_{50}$, $W_{20}$ (width at 50, 20 percent of peak line flux)

• Multiwavelength observations: X-ray, UV, optical, IR, mm, radio
  • UV-submm figure from Dale et al., astro-ph/0610688
  • X-ray to radio: Helou figure
  • opt/IR
  • What contributes to the light at different wavelengths?

Morphological

• Traditional visual classification.
  • want to illuminate some physical differences, want to be unambiguous, and unique.
  • since galaxies have low SB and contain morphological features at high angular resolution, classification depends on observational material available, so any large classification must be done on uniform set of material.

• Morphological systems - good reference Sandage in Galaxies and the Universe, 1975. Some pictures:
  • ellipticals
  • S0s
  • spirals
  • Tuning fork diagram)

• Hubble classification:
  • Ellipticals given as En with $n=10(1-b/a)$, no distinction between dwarf ellipticals and dwarf spheroidals;
  • spirals as barred or unbarred, classified abc depending on B/D, tightness of arms, degree to which arms are resolved into individual HII regions (note latter two are essentially impossible to judge for edge-ons)
  • S0's are intermediate, no spiral structure, but have disk system, split into $S0_1, S0_2, S0_3$ depending on amount of dust
  • irregulars are IrrI (Magellenic irregulars with lots of distinct HII regions), and IrrII which lack the resolution into distinct HII regions.

• deVaucouleurs/RC3 classification:
  • extends Hubble to later spiral types Sd, Sm, finally Im.
  • Extra classes around S0 (SO$^-$, SO$^+$).
  • Allows for intermediate between barred and unbarred. Normal spirals become SA, barred SB, transition SAB.
  • Adds extra distinction for ringed vs. s shaped.
  • Note that location along spiral sequence may not be the same as in ``Hubble classification'' (more based on B/D).
  • Numerical galaxy types (T).

• DDO/van den Bergh classification
  • has 3 spiral sequences: lenticulars, anemics, and spirals, where classification along each is purely on B/D
  • luminosity class depending on length and coherence of spiral arms.

• Morgan (Yerkes) classification.
  • Based on degree of central concentration (afgk), plus associated form family of each galaxy ( S, B, E, I, R, D); also involved integrated spectra (??). Note D galaxies are ellipticals with extended envelopes (cD nomenclature still survices for central cluster galaxies). N galaxies have bright nucleii superposed on much fainter background.
  • Also involves integrated spectra (??)

• Note that peculiar galaxies exist! Are they fundamental? Beware of ignoring them!

• Dependence of morphology on wavelength/redshift (O'Connell 9609101, http://nedwww.ipac.caltech.edu/level5/Kuchinski/frames.html). Note possible evolution of morphology. Also strong non-uniformity of plate material. Beware of galaxy classification at high redshift!

• Problems with traditional morphological classification: subjective, requires uniform classification material, inclination dependent (no face-on S0's), evolution
  • Size of errors in classification. Lahav (Science 267, 859) have 6 experts classify APM galaxy images, get rms differences with RC3 of 2 T-units, slightly better agreement between classifiers, but still substantial dispersion. RC3 quotes 0.3-1 T classes depending on sources).

• Non-subjective morphological classification: Lahav 1995 gets rms dispersion of 2-2.5 T-units with neural net, Andreon and Davoust 9610235.

• Quantitative morphological classification: parametric measurements
  • bulge-disk ratio, using BD decomposition (e.g. MacArthur, Courteau, & Holtzman, ApJ 582, 689 (2003)). Issues: covariance between parameters, 1D vs. 2D, validity of models
  • global Sersic index

• Quantitative morphological classification: non-parametric measurements
  • concentration, e.g. SDSS $r_{90}/r_{50}$, $r_{80}/r_{20}$ using circular apertures; can also consider elliptical apertures, e.g., based on second order moments. Related to B/D. Sensitivity to seeing/distance.
  • asymmetry (e.g. Abraham et al, ApJS 107, 1, 1996 : rotate about center, self-subtract, $A = 0.5 \vert subtracted\vert/total$). Possible indicator of mergers.
  • clumpiness, e.g. S (Conselice ApJS 147, 1 (2003)): subtract smoothed version of image from original image, ratio of flux in subtracted image to flux in original images gives S. Possible indicator of star formation.
  • Gini coefficient, (Abraham et al, ApJ 588, 218 (2003), Lotz et al., AJ 128, 163 (2004) ) applied to sorted list of pixel values. May be more stable at low resolution, S/N.
  • M20: second order moment of brightest 20% of galaxy light (Lotz et al., AJ 128, 163 (2004) ) May be more stable at low resolution, S/N.

• Important question: what about morphology is fundamental?

Dark matter ``observations''

• Galaxies contain dark matter, which may dominate total mass. Evidence:
  • galaxy rotation curves
  • satellite galaxy orbits (e.g. Prada, Vitvitska, Klypin, Holtzman et al)
  • hot (X-ray) gas in clusters
  • motions in clusters, e.g. cluster velocity dispersions
  • gravitational lensing
  • large scale structure
• Quantity: $\Omega_m \sim 0.3$
• Composition still unknown (particle experiments in progress!). Structure formation models favor massive particles (cold dark matter).
• Big Bang nucleosynthesis implies most dark matter is non-baryonic
• typical galaxies like the Milky Way have M/L around 20-100, but as one goes to both larger and smaller systems, M/L ratios likely increase.
• question: does dark matter/baryonic ratio vary between galaxies, and if so, how and why?

Measuring distances to galaxies

• As discussed below, many of the classical galaxy catalogs do not include measured distances. Obviously, knowing the distance to a galaxy allows galaxy parameters (and their correlations) to be studied in a much more physically meaningful way, because distances are needed to determine luminosities and physical sizes
• Galaxy distances from redshifts
  • Galaxies expand with everything else in the cosmic flow according to $v=Hd$.
  • However, galaxies move around relative to the uniform expansion because of local gravitational attraction from concentrations of mass - the deviation of the velocity from the smooth flow is called the peculiar velocity
  • Typical peculiar velocities are on the order of several hundred km/s, but depend on environoment; c.f., Milky Way and galaxy clusters
  • At redshift $z\sim.1$, peculiar velocities lead to relatively small distance errors
  • Determining spectroscopic redshift for distant (faint) galaxies is observationally time consuming
  • Lower accuracy distances are possible from so-called photometric redshifts.
    • Use low resolution spectral information, i.e., multiband photometry to constrain redshift
    • Need to separate different intrinsic SEDs of galaxies, internal reddening, and redshift
    • Accuracy of results obviously depends on quality of photometry and number of bandpasses
    • Typical accuracies (checked against spectroscopic redshifts) give ${\Delta z \over (1+z)} \sim 0.1$ (sometimes better); note typical $\sim$ 10% outlier fraction (varies for different types of galaxies)
• Distances to nearer galaxies are trickier (and note, distances to nearer galaxies were required to determine the Hubble constant, used in the redshift distance indicators!)
  • For nearest galaxies in which individual bright stars can be observed, use Cepheids; issues are basic calibration of Cepheids and variation of Cepheid properties with metallicity
  • Use relation of internal dynamical properties of galaxies to intrinsic physical properties
    • Spiral galaxies: Tully-Fisher relation between maximum circular velocity and luminosity
    • Elliptical galaxies: fundamental plane relation between velocity dispersion, surface brightness, and physical size ($D_n-\sigma$ relation)
    • We'll talk about both of these relations in more detail later, as they provide important constraints on models of galaxy formation/evolution, but here they're just introduced as a distance indicator
  • Estimate distances using model of mass distribution and velocity field; likely not super well determined

Galaxy catalogs and selection effects

• No catalogs contain ``all galaxies'', all have selection functions
• Sometimes selection functions are well defined, sometimes not. Catalog is ``complete'' if it contains all objects satisfying some selection criteria; complete doesn't mean contains all galaxies!
• Understanding of selection function critical for any study of systematic variations of galaxy properties
• $V/V_{max}$ : one test for completeness (van der Kruit, Milky way as a galaxy, p 258).
  • For each object in a catalog, calculate ratio of volume at current position over volume to position at which object would just fall out of the sample.
  • should be equal number of objects at all values of this ratio between 0 and 1.
  • the mean value of the ratio (which is better determined than the histogram of individual values unless $n$ is very large), should be 0.5; the rms of this mean is $(12n)^{-1/2}$.
  • only works as a test of completeness if the objects being studied are uniformly distributed in space (or if volume considered is significantly larger than inhomogeneities); often not the case!
  • can use $V_{max}$ values for objects of different values to correct numbers of objects to get relative numbers of objects at each value in a volume

• Since galaxies (and most astronomical objects) come in range of brightnesses, obvious selection effect in a magnitude limited sample is that as one goes out in distance, only the brighter portion of the population is sampled. Of course, this tells you nothing about whether the distribution within the population is changing, until you model the selection effects.

• For galaxies, beware that methods of finding galaxies optically may be biased against large, faint, objects (e.g., LSB) and small, bright objects which would be confused with stars. MB Fig 5-47.

• Surface brightness selection
  • Obviously, need to see a galaxy before it will be included. But even if central surface brightness renders a galaxy visible, if one uses a sample with an isophotal (magnitude or size) selection cut, one is biased against low surface brightness galaxies:
  • Consider disk galaxies with exponential profiles. Using isophotal mags and sizes, galaxies with low CSB will be measured smaller if size is the same as HSB counterpart; they will be fainter than galaxies than HSB galaxies of same total mag ( McGaugh Fig 1 ). Consequently, they are biased against in either diameter or mag limited samples.
    • What is the amplitude of the effect?
    • Consider galaxy with profile:
      
      $$\Sigma(r) = \Sigma_0 \exp{-({r\over h})}$$
      
      
      
      
      $$\Sigma(\theta) = \Sigma_0 \exp{-({\theta\over \alpha})}$$
      
      
      where $r$ is the linear radius, $\theta = {r\over d}$ the angular radius for distance $d$; $h$ is the linear scalelength, and $\alpha = h/d$ is the angular scalelength. In magnitudes:
      
      $$\mu(\theta) = \mu_0 + 1.086 {\theta\over \alpha}$$
      
      

    • Consider a diameter-limited sample. The isophotal diameter at some limiting magnitude, $\mu_l$, is
      
      $$D_{\mu_l} = 2 \theta_l = 1.84\alpha (\mu_l - \mu_0)$$
      
      
      The maximum distance at which a galaxy can lie and have $D>\theta_l$ is
      
      $$d = 1.84 {h\over \theta_l} (\mu_l - \mu_0)$$
      
      
      The volume within which such a galaxy can lie is
      
      $$V = {4\pi\over 3} d^3 \Omega$$
      
      
      for a survey which covers solid angle $\Omega$. Consequently, we have
      
      $$V\propto [h(\mu_l-\mu_0)]^3$$
      
      
      Bigger galaxies and high $\mu_0$ galaxies will be significantly overrepresented. McGaugh Fig 2.

    • Now consider a magnitude-limited sample where magnitudes are isophotal mags. Galaxies with brighter CSB will emit larger fraction of total flux above any given isophote, even for galaxies of the same luminosity. So LSB galaxies appear fainter; increasing problem when limiting magnitude is brighter.

      Brightness within some isophote for an exponential profile is (by integration):
      

      $$L(r<r_{max}) = 2\pi h^2 \Sigma_0 [1-(1+x)\exp{-x}]$$
      
      
      where
      
      $$x\equiv {r_{max}\over h}$$
      
      
      is the radius expressed in units of the scalelength. In angular units, we have
      
      $$x = {\theta_l\over \alpha}$$
      
      
      . The apparent brightness is
      
      $$\L _{app} \propto {L\over d^2} \propto {h^2\over d^2} \Sigma_0 [1-(1+x)\exp{-x}]$$
      
      
      The maximum distance at which a galaxy can lie and have $L>L_l$ is
      
      $$d\propto L^{1/2}$$
      
      
      and the corresponding volume is:
      
      $$V\propto L^{3/2}\propto h^3 \Sigma_0^{3/2} \propto h^3 10^{-0.6\mu_0}$$
      
      
      McGaugh Fig 3.

  • Turns out that LSB galaxies can in fact be found, and volume corrections suggest that they may be as numerous as HSB galaxies - but not in any of the catalogs! The example of Malin 1: LSB with bright nucleus in direction of Virgo, but at z=0.08 (Virgo is at 0.004!). Galaxy has CSB=26 and a scale length of 55 kpc! At Virgo distance, it would be a degree across, and at Andromeda distance, it would be 20 degrees across!

    McGaugh Fig 5 gives an estimate of number of LSB's with the assumption that $\mu$ and $\alpha$ are uncorrelated - which appears to be the case (McGaugh, Schombert and Bothun Fig 1).

  • However, subsequent observations, e.g. SDSS (see below), suggest that LSB galaxies in general, tend to be fainter, and do not contribute a dominant fraction of the total luminosity of galaxies.

• Selection effects can also bias derived parameters, e.g. luminosity distributions, size distributions, SB distributions, etc., for a given sample. Several types of bias exist:
  • surface brightness bias. As discussed above, one may miss LSB galaxies. If these have systematically different properties, our sample will be biased.

  • distance dependent, or Malmquist, bias. This occurs in any sample which is selected on the basis of some distance-dependent parameter, when the quantity in question has a dispersion either from intrinsic dispersion or observed errors. These occur such that more galaxies are scattered into the sample from larger distances. Since larger distances contain more volume, will get more galaxies scattered in from larger distances than scattered out from smaller ones, leading to a bias to brighter (or larger) galaxies. Amplitude of effect depends on distribution of galaxies in space and on dispersion of quantity (actually proportional to square of dispersion).
    • Note that different people use the term Malmquist bias in different ways. Some examples:
      • Survey of type Ia supernovae, which come in range of absolute luminosities: only brightest types are seen at larger distances in a magnitude limited sample. This is just simple selection effect.
      • At given SnIa light curve shape, there's a dispersion of absolute luminosities. At largest distances of a magnitude limited sample, only the most luminous (of the most luminous) SN will be detected. This leads to a mean overestimate of the brightness of this class of SN.
      • mag-limited survey only sees bright galaxies at larger distances, so mean brightness will be overestimated. To first order, this is removed by correcting for selection function, but Malmquist bias comes in because there are dispersion/errors in the observed mags and/or distances. Given perfect distances, it's clear that in the most distant bin, you'll preferentially select galaxies which are observationally measured too bright from random photometric error. Given imperfect distances, you'll include more galaxies which are really at larger distances than those which really are at smaller distances. All ways, you estimate brightnesses to be too bright.

    • A common example these days is turned around: given some estimator of intrinsic brightness (with scatter), determine distances: these will be biased.

    • It is possible to correct for Malmquist bias if you understand the spread of the observed quantity and you know (or can make a good assumption) about the spatial distribution of your objects.

  • wavelength dependent bias. Selecting galaxies optically may bias, e.g., against dusty or gas dominated galaxies. Galaxies selected in blue may be biases towards galaxies with younger stellar popualations.

  • inclination dependent bias. If selection property depends on inclination, then you'll have a sample with properties biased as a function of inclination.

• If you are looking for trends within a sample, it's critical to understand how the selection function will affect possible trends. Just because a sample is well-defined, complete and/or large does not guarantee that there are no biases and that observed trends are real!! No complete volume-limited sample exists.

• Some galaxy catalogs
  • NGC (Dreyer, Sulentic&Tifft RNGC): selected from 19th century scans (mostly Herschel). 8640 galaxies. Includes star clusters, etc., as well.

  • IC (Dreyer 1895, 1908).

  • Revised Shapley Ames (RSA): originally compiled by SA 1932 to contain all galaxies with $m(pg)<13.2$. However, it's really only complete to $\sim 12$. And as we'll discuss shortly, this is only for high surface brightness galaxies!

  • RC3 (deVaucouleurs): heterogeneous selection from a variety of sources. "Attempts to be reasonably complete for galaxies having apparent diameters larger than 1 arcmin at D(25) and total B-band magnitude $B_T$ brighter than about 15.5, with redshift not in excess of 15,000 km/s", apart from LSB galaxies.

  • UGC (Nilsson): visually selected from Palomar sky survey. Complete for galaxies with $D>1$ arcmin on blue plates. Note size is at SB limit of POSS-I ($\sim 25.3$).

  • CGCG: Zwicky catalog. Mag limit 15.5.

  • ESO/B, Lauberts (1982), diameters greater than 1 arcmin

  • ESO/LV, with surface photometry

  • MCG (Vorontsov-Velyaminov+ 1962-68), from POSS, I don't know selection function

  • Nearby Galaxies Catalog (Tully): galaxies within 3000 km/s. Mostly RSA galaxies plus galaxies from neutral hydrogen survey.

  • Markarian: strong UV excess

  • Arp, Arp/Madore. Peculiar and Interacting Galaxies.

  • APM, magnitude limited sample from UK Schmidt plates

  • HIPASS, HI selected catalogs

  • SDSS: northern galactic cap survey (+ few southern strips)
    • Main galaxy sample: photometric objects with $r<17.77$ that are (conservatively) distinguished from stars, and satisfy a surface brightness $\mu_{50}<23$ are targetted for spectroscopy
      • From spectra, redshift can be determined and object can be fairly conclusively classified as galaxy, quasar, or star
    • LRG (luminous red galaxy) sample: fainter, but with color cuts to target red galaxies

  • 2DF: magnitude limited redshift survey from APM catalogs (southern galactic cap)

  • other surveys