Although classifying with Galaxy Zoo doesn’t rely on it explicitly, many participants soon learn about Edwin Hubble’s system of classifying galaxies. More precisely, his system of classifying blue-filtered images of galaxies, because that’s what was available when he did his work. His system is often summarized in the tuning-fork diagram, showing how the properties of central bulges and spiral arms change systematically along a sequence from ellipticals through S0 galaxies (which he hypothesized, but could only be identified later), and both barred and ordinary spirals of types Sa and SBa through Sc and SBc. What has given Hubble’s classification such staying power is that (to our enormous good fortune) many galaxy characteristics of physical importance – star-formation history, gas and dust content, stellar dynamics – all correlate well with something that is, after all an eyeball estimate based on how much a galaxy looks like one of a set of “standard” galaxies.

Might we have selected other features to key on? This is more than a rhetorical question – additional systems of galaxy classification have been suggested over the years, either parallel to Hubble’s or incorporating refinements to describe additional features. Some of these have largely been added to the Hubble types in wide use, some have given us a few additional descriptions such as N and cD galaxies, and some remain islands of specialist interest.

Probably the most influential of these refinements came from Gerard de Vaucouleurs, who made several important additions. First, he recognized a continuous sequence right past the Sc galaxies to irregular systems – to Sd, and then to disk galaxies organized as well (or as poorly) as the Large Magellanic Cloud Sm, and finally to completely disorganized irregulars I. Each of these came in both barred and nonbarred forms. Speaking of bars, de Vaucouleurs looked at the central regions of galaxies and found that there were strong bars, nonbarred galaxies, and small or weak bars. One more parameter. Also, spiral arms can either spring out from the core like an S or from a ring, or a combination. He denoted nonbarred spirals as SA, barred spirals as SB, and intermediate (weakly barred) galaxies as SAB. For the core, an inner ring would be (r), core spiral (s), and combination (rs). Then he added a stage letter like Hubble’s, but now extended past Sc as far as Sm – so, as an example, his estimate for the Milky Way’s type is SB(rs)bc.

These distinctions are easiest to recognize in the middle of the spiral sequence, so de Vaucouleurs introduced the idea of a 3-dimensional classification volume, of which Hubble’s tuning fork is a sort of cross section parallel to one axis. The Reference Catalog of Bright galaxies by him and colleagues showed what was described as an artist’s depiction of a slice across the classification volume for Sb galaxies. (Years later, de Vaucouleurs’ one-time student Ron Buta found that the artist had been de Vaucouleurs himself, on a cloudy night at McDonald Observatory.)

Ron also told me that each of these was recognizably based on a bright galaxy, a challenge I could hardly let pass. Here’s what I came up with from memory (not all are in the SDSS).

SA(rs) NGC 3486
SA(s) NGC 1566
SB(s) NGC 1300
SB(r) NGC 5921
SAB(rs) NGC 4303

(I’m still waiting to get my test score on those).

Rings also occur in the outer parts of galaxies – de Vaucouleurs noted these and designated them by an initial R or (R) – the latter is a pseudoring, actually formed by tightly wound spiral arms. Rings are important, wherever they occur, because they show us where there is a resonance between the orbital speeds of stars in the disk and global pattern, which can move stars (and interstellar matter) radially within the galaxy. They are often associated with bars; one working idea for strong rings without a bar is that the bar itself was finally dissolved as the ring strengthened over time.

de Vaucouleurs also noted that not all elliptical and lenticular galaxies are identical. Some ellipticals have more or less extended outer starlight distributions than typical, and S0s can be grouped depending on how prominent the disk of stars is, and how much dust (and by inference gas) remains in the disk – from the most elliptical like S0^– syses through, S0 to S0⁺, which are most like Sa spirals. Examples are (respectively) NGC 3115, 5866, and 4710.

Almost by accident, de Vaucouleurs introduced a numerical coding of his types for his First Reference Catalog, which was (noteworthy for its time) typeset from computer-printer output, For easy sorting and legibility, he added a coding of each type in easy-to-read form for a computer, including a type index T – T=-4 for ellipticals, 0 for a S0⁺, 1 for Sa, 3 for Sb, and so on to 10 for irregular. This has proven to be a very useful variable in innumerable data plots ever since.

Some of these refinements, especially the spiral/ring/bar distinctions, are so informative that many people add them to Hubble types. This isn’t exactly right – the two sets of types are not defined in quite the same way. In polishing Hubble’s original system, Allan Sandage often based the spiral types Sa-Sb-Sc on how much resolved detail showed up in high-quality photographs from Mt. Wilson or Palomar, while de Vaucouleurs, often working from material of lower quality for galaxies father away, used more global quantities such as the pitch angle of spiral arms. It’s too late to stop the hybridization, just because it has proven so useful.

Sidney van den Bergh not only considered the structures of galaxies, but their distances and thus luminosities, to add additional wrinkles to the classification puzzle. Close examination of galaxies in clusters led him to suggest that perhaps the tuning fork should be a trident, with a separate tine for disk galaxies with weak, ghostly spiral structure and low rates of star formation (anemic galaxies, which could be denoted Aa, Ab, Ac by analogy with the S and SB classes in Hubble’s system). This leads into a whole slew of arguments about S0 galaxies, how many kinds there are, and how many pathways for a spiral to end up that way.

More lasting has been van den Bergh’s finding that there is an overall (albeit rough) correlation between the brightness and continuity of the arms and the overall luminosity of spiral galaxies. He distinguished various luminosity classes, from the most luminous (I) to least luminous (V). Despite having the name “luminosity”, these classes are assigned strictly from examination of the galaxy structure (like other morphological classes). If these classes were very strongly correlated with luminosity, you could even use them as a kind of distance estimator. And up to a point, this works; its most lasting impact has been the recognition of Sc I galaxies, particularly large and luminous spirals with prominent grand-design spiral patterns, which are indeed as a class very luminous. Some familiar examples of Sc I galaxies are M100 and M101. Vera Rubin used this category in an early (and controversial) study of bulk motions in the local Universe. The luminosity classes proved useful enough that Allan Sandage incorporated them in the classifications in the Revised Shapley-Ames Catalog and the giant NASA galaxy atlas in the mid-1980s. Revisiting this identification of the prominence of spiral structure as a facet of spiral galaxies (which is independent, in principle, of whether it’s Sa, Sb, barred or whatever), Debra and Bruce Elmegreen assigned arm classes to galaxies – from 1 for a galaxy with only a few disconnected bits of spiral, to 12 for galaxies with two long, dominant spiral arms.

W.W. Morgan, better known as the M in the MK classification of stellar spectra, proposed a galaxy-classification system (the Yerkes classes) which explicitly recognized the usual correlation between galaxy structures and their colors (or at least characteristic spectral lines). He considered form families – ordinary spirals S, barred spirals B, ellipticals E, other kinds of symmetric galaxies not quite sharing the brightness concentration of ellipticals D, and galaxies strongly dominated by their nucleus N. In place of a stage a,b,c… he used the equivalent stellar spectral type of most galaxies sharing a given degree of central concentration – the most concentrated galaxies, typically ellipticals, were k, the least concentrated (ratty spirals and irregulars) were type a. Finally he added a shape class, which for spirals tells us the inclination – 7 is edge-on, 1 is face-on. So a galaxy like M86, a giant elliptical in the Virgo cluster, would be kE2, while the edge-on and rather disorganized spiral NGC 4631 would be afS7 (af denoting something in between classes a and f). Despite looking a bit like stellar spectral types, here again, the types are assigned purely from the galaxy’s appearance on images. The most-used pieces of the Yerkes classification have been N galaxies, for many AGN hosts where the nucleus outshines the galaxy, and cD galaxies, the largest normal galaxies (c comes from an old designation for the sharp spectral lines which indicate that a star is a supergiant). One of the best-known cD galaxies is IC 1101 at the center of the cluster Abell 2029, seen here in the SDSS:

In spirit, some of Morgan’s approach has lived in in much of the recent work in quantitative classification, in which a galaxy is described straightforwardly only by various numerical parameters of its image – light concentration, characteristic brightness, shape, and so on. Sure, computers can do that very well – but is that enough?

Boris Vorontsov-Velyaminov (alias VV) used a rather iconoclastic classification scheme, based on close and extensive inspection of the prints of the Palomar Sky Survey. He noted such exotica as gamma forms (where one spiral arm appears to wind opposite the others), whether the two “handles” of an edge-on galaxy matched or not, whether a ring was broken or complete, and so on. His types are highly descriptive, but can’t easily fall into continuous sequences. And with the arrival of digitized catalogs, it didn’t help that he availed himself of not only the Roman alphabet, but mirror-reversed Roman letters and italics (not even getting into Cyrillic characters, oddly enough). On top of all that, his catalog listed galaxies with running numbers according to which plate of the sky survey their image was found on. For these reasons, his classification has found much less wide use than has the VV catalog of interacting galaxies, for which he did the first reasonably complete sky survey from the Palomar pictures.

All these different ways of sorting galaxies face us with a key question – what are the important features of galaxies that we ought to catalog? Some details of spiral patterns surely come, go, and change form, while the forms of ellipticals and spiral bulges are very long-lasting. Where should we draw the line? And all these classification systems are set up to work for images take in the optical window of the spectrum. What would we have come up with if we looked at galaxies in the infrared or ultraviolet, or bands even more exotic? That’s one of the questions that enter into Hubble Zoo, and deserves its own discussion…

As a follow-up to my post describing the science behind simulating AGN for classification of their host galaxies, here is an example of one of the 116 galaxies that we used to create the 1740 simulated AGN in the sample.

Each galaxy was made into 15 different simulated AGN using 3 colors and 5 luminosity ratios.

If you’d like to see the full-size version, click on the image.

As some of you have pointed out, the cases where the simulated AGN is as bright as or brighter than the host galaxy do look a little like, well, simulations. But hopefully that doesn’t significantly affect the main science question: how does its presence affect the classification of the underlying galaxy (if at all)?

As Carie mentioned in her recent post on this topic, the results from the classification of these simulated AGN will be posted shortly.