ABSTRACT

We present a representative sample of near-infrared (1 - 2 µm) images spanning the Hubble galaxy morphology classification sequence. The images come from the Two Micron All Sky Survey, ranging in brightness between 7 < K_s < 13.5 mag, and the morphological classifications from the NASA/IPAC Extragalactic Database (NED). We differentiate between elliptical, normal and barred spirals, compact, dwarf, peculiar and AGN/Seyfert-type galaxies. This image atlas serves as a qualitative guide to the appearance of galaxies at wavelengths that penetrate the obscuring veil of dust, as well as provide a window to the older population of stars that dominate the observed light at 2 µm, including large-scale bar features. We show that central surface brightness and integrated flux may be used to discriminate spheroid and early-type spirals from late-type spirals. The optical and near-infrared properties are compared.

Key words: galaxies: statistics --- galaxies: general --- galaxies: fundamental parameters--- astronomical databases: surveys --- astronomical databases: atlases
 

For well over a century astronomers have differentiated or classified galaxies based on their visual appearance (of course, in the time of Herschel, the true nature of galaxies was yet to be deciphered, but nonetheless these "nebulae" exhibited repeating characteristic patterns). The science of galaxy "morphology" matured with the work of Hubble (1926 & 1936), whose famous "Hubble Sequence" remains today a standard technique by which to study galaxies. This sequence of normal spiral and elliptical galaxies was further expanded and delineated into subclasses by de Vaucouleurs in his classic 1959 paper, also described in the Hubble Atlas (Sandage 1961; see also Sandage 1975). More recent reviews are given in Buta (1992), Roberts & Haynes (1994) and Van Den Bergh (1998). Even though the morphology/classification scheme has grown into a rather large and complex set of designations, the basic observables are simple: relative size and brightness of the nucleus with respect to the disk or outer envelope (quantified as the bulge-to-disk ratio), the shape of the disk (e.g., arms, bars, rings, or asymmetries), and other factors, such as the surface brightness, color, and the presence of dust.

The Hubble sequence, at first glance, appears to be simply a qualitative "roadmap" giving convenient names to the wide variety of galaxies observed in the Universe. But as Sandage (1986) points out, the sequence does in fact separate galaxies into physically related classes, representing an "evolutionary" sequence. For example, the mean disk age of spirals progresses along the sequence (early to late-type), as does the present star-formation rate (e.g., late-type galaxies are forming stars at a much higher rate than early-type galaxies; see also Ferrini & Galli, 1988). Hubble & Humason (1931) understood that the morphological sequence was correlated to the galaxy density environment: galaxy clusters contain more lenticular and elliptical galaxies than the typical field density.

Morphological segregation appears to be a property of large clusters, including those of Virgo and Coma. Further, the luminosity functions of spirals is different from that of elliptical-type galaxies (cf. Binggeli et al. 1988). Although the origin and formation of elliptical galaxies is still largely not understood, more recent work has pointed to a scenario in which elliptical galaxies may be the creation of spiral galaxy mergers in the dense cores of galaxy clusters (e.g., Dressler 1980; Postman & Geller 1984; but also see Mamon 1992; Whitmore et al. 1993; Julian et al. 1997). There is evidence that widely separated morphological types (in the Hubble Sequence) have systematically different mean colors (e.g., B-V), presumably from different populations of stars, both old and newly formed, that dominate the light (e.g., Holmberg 1958; de Vaucouleurs 1977; Giovanelli & Haynes 1983; see the review by Roberts & Haynes 1994; Buta et al. 1994; Odewahn et al. 1996). Another class of galaxy, the dwarf spheroidal, also appears to be associated with the dense environments of clusters, although their origin, evolution, age and sustained existence (from the disruptive gravitational tidal forces) remains a mystery (cf. Sandage 1990; also discussed by Impey & Bothun 1997).

Spheroidal galaxies have the simplest morphology, exhibiting a smooth tri-axial (in two dimensions, elliptical) distribution. Disk galaxies, on the other hand, have a variety of morphological features, comprising the bulk of the Hubble Sequence. Spiral galaxies, for example, may exhibit large-scale "ring"-like or "barred" features, both delineated in the revised sequence of de Vaucouleurs (1959). Finally, there are galaxies that defy classification, they do not fit neatly into any category of the Hubble sequence other than generic "irregular," "peculiar" or otherwise "unknown" classification. These galaxies are relatively few in number, implying either a short-lived transition phase (evolution into something more recognizable) or their formation requires unusual circumstances (e.g., galaxy-galaxy interactions). A galaxy of this "type" does not fall under the broad definition of a "normal" galaxy. In this paper we present results for what are mostly "normal" spiral and elliptical galaxies, as based on luminosity, color and symmetry. For comparison, we include galaxies that are not "normal," including active, irregular and peculiar types. Morphologies are generally derived from optical images, primarily due to the superior resolution, sensitivity and areal coverage gained from photographic and CCD detectors. At these wavelengths, the light of a spiral galaxy disk is dominated by intrinsically bright and massive young stars, providing a clear contrast between the core/bulge (whose light is dominated by the older stellar population) and disk/spiral-arm components. Elliptical galaxies, mostly devoid of gas and dust (with very little star formation) and other discriminating features, are instead classified by their spheroidal shape.

The spiral "evolution" is clearly separated into early-type ("Sa"), intermediate-type ("Sb"), late-type ("Sc"), and very-late-type ("Sd") classifications. Some spirals clearly show a "bar" feature, while others appear to be intermediary between "normal" and "barred" spirals. It may be that most if not all spirals contain some form of a linear "bar," but this issue remains controversial (see Roberts & Haynes 1994). Although this issue remains unresolved, the near-infrared images presented in this paper include these three closely related types (normal, barred, intermediary) for comparison.
Galaxies may be "classified" using alternative wavelength imagery, for example at infrared wavelengths, but the contrast between morphological classes is not as clearly defined as that seen in the optical. Nonetheless, these alternative- data can provide vital supplementary information toward more reliable classification and, in particular, for those systems with significant dust absorption (e.g., barred galaxies). In the near-infrared (NIR), for example ~2 µm, the emitted light from galaxies is dominated by the older population of stars-which represents the major integrated mass component. Moreover, this light is mainly free of the obscuring effects of dust, a particularly vexing problem in late-type spirals, rendering an optically-thin view of the galaxies. It is possible to construct a large database of morphologically-ordered galaxies, as seen in the near-infrared window, using data coming from modern large areal-coverage surveys.

The Two Micron All Sky Survey (2MASS; see Skrutskie et al 1997) is now publicly releasing catalog and imaging data that covers a large fraction of the sky at J (1.2 µm), H (1.6 µm) and K_s (2.2. µm) bands. Galaxies as faint as J = 15.0 and K_s = 13.5 mag, or 1.7 mJy and 3.1 mJy, respectively, will comprise a complete and reliable extended source catalog (Jarrett et al. 1997; Jarrett et al. 2000). 2MASS provides a large enough data set to construct an infrared version of the Hubble sequence. However, as we show in §5, the sensitivity and resolution (~2") of the NIR image data is not adequate to derive independent sub-classifications with the exquisite fidelity that optical imaging allows. Consequently, we utilized optically-derived classifications coming from previously identified galaxies in the NASA/IPAC Extragalactic Database (NED) to separate our infrared galaxy images into a Hubble sequence of elliptical, spiral and other types. The "Atlas" presented here functions as a "first step" toward deriving a classification sequence based solely upon infrared imaging, or most promising, in combination with optical imaging. A true infrared/optical "sequence" would have the advantage of being minimally affected from dust extinction-particularly important for understanding bar features-and would be more representative of the stellar populations that dominate the system mass.

The basic data are introduced in §2. A definition and brief review of the Hubble types is given in §3. In §4 we present an "atlas" of galaxy images, coming from 2MASS, separated according to the previously-derived optical- Hubble type. A "composite" infrared Hubble sequence is first presented showing the seven categories of galaxies types: elliptical, spiral, compact, dwarf, irregular/peculiar, and Seyfert/AGN, with the main atlas images to follow. Irregular, peculiar and active galaxies are not easily classified, but we include them here for completeness. Spirals are further delineated into subclasses, including normal and barred versions, while ellipticals are separated according to their ellipticity or spheroidal flattening. In §5 we consider some of the basic image parameters, including surface brightness, flux and color, to evaluate the degree with which the different Hubble types separate at near-infrared wavelengths. Finally, a comparison with optical properties is given. In the future, it may yet be possible to derive an infrared Hubble sequence from 2MASS and optical data using, for example, the atlas presented in this paper as an initial solution for an iterative approach.

The NIR data come from the 2MASS Second Incremental Release, covering over 47% of the sky, including the northern and southern hemispheres. 2MASS extended sources (e.g., galaxies) are identified and characterized with a "pipeline" software package called GALWORKS, developed specifically for 2MASS images. The 2MASS Extended Source Catalog includes the coordinate positions, integrated brightness, radial and azimuthal parametric measurements, and images. An overview and description of the algorithms and GALWORKS can be found in Jarrett et al. (2000). For a more general overview of the 2MASS "pipeline" reduction and resultant data products, see the 2MASS Explanatory Supplement (Cutri et al. 2000).

2.1 Galaxy Size and Photometry

The morphology Atlas presented in this paper is sorted by the optically-derived morphological type and the 2.2 µm integrated flux. 2MASS computes a variety of integrated flux measurements for a given extended source. In this study we use the integrated flux based on the elliptical aperture best fit to the K_s-band isophote corresponding to 20.0 mag arcsec^-2 surface brightness (roughly equivalent to 1-sigma of the RMS background noise in the K_s-band image). We show in §5.4 that the equivalent surface brightness in the B band is ~24 mag arcsec^-2 for early types and ~23 mag arcsec^-2 for late types; hence, for late-type spirals the 2MASS images are sensitive to only the brighter morphological features. The same "fiducial" aperture (position, size and orientation) is applied to each band, allowing cross-comparison between bands (i.e., colors). It is this aperture diameter, D₂₀, that represents the size of a galaxy. The elliptical orientation (position angle and axis ratio) of the galaxy is determined from the K_s (2.2 µm) image at roughly the isophote corresponding to three × the background RMS "noise" or 3-sigma.

2.2 Image Data

Each extended source in the 2MASS database includes a small image data cube, representing the J, H and K_s "postage stamp" images (planes 1, 2 and 3, respectively), extracted from the full resolution 8.5´ × 17´ images which comprise a 2MASS "scan" or "tile" (total size: 8.5´ × 6°). The size of the cube is dynamically selected, scaled according to the measured size of the extended source (see above), with min/max limits of 21 and 101´´ pixels, where each pixel is 1´´ in width. The dynamic image size was needed for practical reasons; namely, limited disk storage. For the largest galaxies (typically K_s < 9th mag in brightness) the image size is not large enough and the outer parts of the galaxy are clipped, resulting in an underestimated total flux; galaxies larger than ~1´ in diameter have integrated fluxes that are systematically too small (some examples can be found in the images presented in this paper). Note that the J-K_s color is only minimally affected since the flux loss occurs nearly equally for all bands.

 2.3 Atlas Images

We have elected to present the galaxy "postage stamp" images as gray-scale halftones of the K_s (2.2 µ) image, overlaid with contours corresponding to the J (1.3µm) image. The gray-scale ranges from a minimum value of 1-sigma of the background noise to a maximum value corresponding to the peak pixel value for the galaxy (but no smaller than 5-sigma). We then take the base-10 logarithm of the 2.2 µm image, which enhances the contrast in the low SNR galaxy emission. The J-band contours correspond to the 1%, 2%, 4%, 8%, 16% and 32% 1.3 µm peak intensity levels. The typical surface brightness of the lowest contour is 20.4 mag arcsec^-2 in J (1-sigma ~ 21.6 mag arcsec^-2). A figure caption accompanies each image, including the catalog or NED name, morphological or Hubble type, 2MASS name, K_s mag and J-K_s color. The 2MASS name maps into the J2000 equatorial position of the source (Jhhmmsss+-ddmmss). The angular size of the image is noted in the lower right hand corner of the image, ranging from 21 to 101´´.

We employ the revised system of de Vaucouleurs (1959), with brief descriptions of the basic categories and classes given below. We will combine some subclasses for convenience. Galaxy morphology and name or designation are taken from NED, a large heterogeneous set derived from predominantly visual- catalogs and surveys. The bulk of NED's types are from the RC3 (de Vaucouleurs et al. 1991), with a smaller number coming from Sandage (1961) and Buta (1995).

3.1 Elliptical Types

Elliptical galaxies are characterized by a bright nucleus and a featureless envelope (i.e., lacking a disk). They are, more precisely, spheroidal in shape. They typically have a characteristic r^1/4 surface brightness profile and usually are redder in B-V color than spirals (indicative of older more evolved stars). These are further sub-classified according to their ellipticity: E(n), where n = 10 ( 1- b/a), b/a is the axis ratio. Hence, classifications progress from E0, E1,..., E7.
 
3.2 Lenticular Types

Lenticular galaxies, S0 and SA0, are probably some kind of "transition" class between an elliptical and spiral galaxy, although this relationship is not well understood. They have a bright nucleus and a smooth envelope and disk, resembling a "lens" shape or profile. In general they do not have delineated arms, although some lenticulars do have dust lanes. It is not clear whether lenticulars represent an evolution from the ellipticals, or the reverse, an evolution from the spiral class (see Roberts & Haynes 1994; see also van den Bergh 1994 for a different perspective). Qualitatively, they represent an "early" type in comparison to most other spiral types.
 

3.3 Spiral Types

Normal spiral galaxies range from early type, Sa, intermediate, Sb, late-type, Sc and very late-type, Sd. Qualitatively early-type spirals have prominent cores, while late-type spirals have prominent disks.
Barred spirals exhibit a large-scale linear "bar" feature crossing the nucleus. The bar to disk/arm association may be closed, forming a ring (or "r" type), or may be open, forming a spiral (or "s" type) shape. Early type galaxies, SB0 and SBa, tend to have closed structures, while late-type galaxies, SBb and SBc, tend to have more open spiral structures. A "transition" class between normal and barred spirals is designated as SAB. It should be emphasized that aside from the bar itself, there is little (if any) difference in optical properties between SA and SB galaxies of similar age and evolution.

3.4 Other Types

Compact galaxies are not obviously elliptical or spiral. They include the "blue compact dwarf" (BCD) galaxy (cf. Sandage & Binggeli 1984; Hoffman et al. 1989). As their name implies, their cores are much more prominent than their envelope. They are almost exclusively round in shape. Compact galaxies will not usually be resolved with 2MASS and, hence, will only be sporadically included in the extended source catalog.

Dwarf galaxies include spheroidal/spiral (dS) and elliptical (dE) versions (cf. Sandage & Binggeli 1984). They tend to have a low surface brightness (LSB), with weak or absent central cores (e.g., Kormendy 1977). They may also be compact, with a high surface brightness in the core, or they may be "irregular" with asymmetric profiles (see below). See Sandage (1990) for a comprehensive look at Virgo & Fornax dwarf galaxies.

Seyfert, LINER, and AGN-type galaxies exhibit strong emission lines in their cores (e.g., Osterbrock 1993). These galaxies may be spirals, peculiar (e.g., interacting pairs), or part of starbursts or very dusty galaxies, and many are associated with quasars. They tend to have red colors in the near-infrared. They also tend to be distant (z > 0.3), making them difficult to detect and resolve with 2MASS.

Finally, irregular galaxies (Irr) may include dwarfs (although we make a distinction, see above), peculiar (Pec), interacting, merger and "magellanic cloud" (SBm; Im) versions (e.g., de Vaucouleur & Freeman 1972). Irregulars are usually barred, IB, and "s" shaped, I(s). In general, galaxies that are not classified as elliptical or spiral type may be lumped into this disparate group, even though some may be physically unrelated.

The NIR images are presented in "plates" of 16 images each (see Table 1), comprising 864 total galaxies. The sequence order is: ellipticals, spirals, compacts, dwarfs, peculiars and AGN/Seyferts. A handful of the elliptical and spiral galaxy types include LINER or Seyfert features (although qualitatively they are indistinguishable from their "normal" cousins). The order within the plate is given by the integrated K_s flux of the galaxy (denoted with the K_s mag in the figure caption), ranging between 7 and 13.5 mag (300 and 2 mJy, respectively). The size (in arcsec) for each image is noted in the lower right-hand corner of the gray-scale image. The Atlas does not represent a complete sample, nor does it represent a flux-limited sample (although we do avoid sources fainter than an SNR ~ 10, or K_s ~ 13.5 mag). However, the atlas does represent a mostly unbiased sample of 2MASS galaxies that have catalogued morphological classifications-it is biased only in the sense that NED is biased. Plate 0 represents a composite near-infrared sequence, directly analogous to the Hubble sequence, including the elliptical sequence (E0 - E6) and the spiral sequence (S0, Sa, Sb, Sc, Sd), delineated between barred, transition and non-barred types.
 

In order to make sense of the Hubble sequence with NIR imaging, we investigate the key basic image parameters that 2MASS derives, including the integrated flux, central surface brightness, axial ratio and colors. As we show below, just a few simple image parameters coming from the survey can be used to delineate early-type spirals and ellipticals from disk-dominated galaxies. These measurements are only a subset of what can be potentially derived from the image data, although they do represent the most important measurements extracted by 2MASS. There are undoubtedly other parameters (and their corresponding phase space) which may prove to be even more powerful at discriminating between spiral-types as well as between normal, transition and barred spirals. More sophisticated methods, for example the implementation of supervised neural networks and the addition of image information coming from visual-based surveys (e.g., using the POSS II), are a promising avenue toward effective morphological classification.

The integrated flux is represented by the K_s-band fiducial 20 mag/arcsec² elliptical isophote aperture (see §2.1), which approximates the "total" flux of the galaxy. In actuality, this measure underestimates the total flux between 10 and 20%, depending on the radial distribution (i.e., Hubble type). The surface brightness is computed from the peak-pixel flux (which we call the "peak" surface brightness) and from the mean of the pixels located inward of 5´´ from the nucleus (which we call the "core" or "bulge" surface brightness). Consequently, the surface brightness is coupled to the galaxy size. One must be careful how to interpret this non-optimized mean surface brightness (see §5.4 for a discussion of the half-light radius). The axial ratio corresponds to the minor-to-major axis ratio computed from the K_s image at the 3-sigma contour (roughly 18.8 mag arcsec^-2; see Jarrett et al. 2000). The infrared colors correspond to J-H, H-K_s and J-K_s (of course, only two colors are independent.) In §5.4 we compare the optical and NIR size and photometry of galaxies.

5.1 Colors

The infrared color distribution is shown in Figures 1 - 5, corresponding to ellipticals (Fig. 1), normal spirals (Fig. 2), normal-barred transition spirals (Fig. 3), barred spirals (Fig 4.) and other types: compact, dwarf, irr/pec and AGN/Seyfert; Fig 5. Early to late-types are ordered from bottom to top. For the ensemble average, ellipticals are slightly, ~few%, redder in color than lenticulars (Fig. 1). For normal and transition spirals the color differences are not statistically significant (Fig 2 & 3). Note, however, there is a clear increase, factor of ~two, in the histogram dispersion going from early-type (S0) to late-type (Sc). This early to late-type dispersion gradient is undoubtedly the same phenomenon seen in at visual wavelengths: the effects of star formation, a younger population of stars and an increase in extinction all lead to an increase in the color dispersion for late-type galaxies. The trend is even apparent between lenticulars and their close-cousin early-type Sa galaxies. Barred spirals (Fig. 4) also show little color difference between Hubble types and between non-barred spirals, except for the addition of a "blue" tail for late-types. A population of relatively blue barred spirals is consistent with a younger (and bluer) population of stars associated with star-formation induced by large-scale bar structures. All told, the NIR colors in comparison to optical colors (e.g., Odewahn & Aldering, 1995; Odewahn et al. 1996) show only small contrast between morphological types.

These statistical trends are not powerful enough to reliably place individual galaxies on the Hubble sequence-color can only be used statistically or with an ensemble of galaxies to distinguish between early and late-type. On the other hand, AGN and Seyfert-type galaxies (Fig 5) exhibit a prominent "red" bias as both H-K_s and J-K_s colors are typically ~0.3 to 0.4 mag redder than, for example, compact galaxies. Active galaxies are known to be intrinsically red in color, partially due to severe dust extinction, in addition to cosmic reddening due to the expansion of the Universe (i.e., their "K-correction" is significant). Extremely red, J-K > 1.3, 2MASS extended objects are good candidates for active galaxies.
 

5.2 Axis Ratios

The minor-to-major, b/a, axis ratio distribution is shown in Figs. 6-8, corresponding to ellipticals, normal spirals and barred spirals, respectively. We do not show figures for small angular-sized galaxies (e.g., compact & AGN) due to the uncertainty in the axis ratio estimate. For ellipticals (Fig. 6), the axis ratio represents the degree of flattening in the spheroids (from circular E0 to elongated E7; see §3.1). Not surprisingly, the near-infrared axis ratios are consistent with the Hubble classifications, where the axis ratios for E0-E1 peak at 0.85, while for E2-E4 they peak at 0.75 and for E5-6 they peak at 0.55. Some intrinsic scatter is to be expected due to the tri-axial nature of spheroids.

For spirals (Fig 7 & 8) there is significant trend from early to late-types: the b/a distribution shifts from a peak of ~0.7 to 0.8 (i.e., the orientation is closer to face-on) for early-types to a more edge-on orientation, ~0.2, for late-types. This trend is probably related to a discrepancy in detection completeness. Face-on spirals are harder to detect and resolve due to their lower surface brightness, which is exaggerated for late-types due to their small bulge-to-disk ratios (see below). We should caution, however, that our NED galaxy sample was not selected to minimize biases or other statistical effects. There are a number of effects that potentially complicate the axis ratio distribution (c.f. Odewahn et al. 1997). It is beyond the scope of this paper to accurately assess the completeness of different morphological types in the 2MASS catalog.

We are able to compare the 2MASS K-band axis ratios with those in the RC3, as given by the R₂₅ parameter (ratio of the major-to-minor 25 B mag arcsec^-2 isophotal diameters). We find good agreement (~10%) for early types, and increasingly poorer agreement (>20%) for late-types (see §5.4 and Table 2). Interestingly, the 2MASS axis ratios are systematically larger (i.e., more face-on) than those of RC3 for normal spirals, while for barred spirals the trend is reversed: the NIR axis ratios are considerably more elongated than their optical counterparts. This result is consistent with the NIR being more sensitive to the (rounder) bulge component and the (linear) barred components of spirals.

The near-infrared is most sensitive to older (redder) populations of stars, which are generally not the dominant stellar source of light coming from late-type spirals (nor of many dwarf or LSB-type galaxies). The inclination is strongly correlated with surface brightness such that blue low surface brightness galaxies (e.g., SAc types) are normally undetected by 2MASS except when viewed edge-on or nearly so. Not only does the disk surface brightness increase with inclination, bars would also be most easily detected edge-on. Since the near-infrared is much less affected by dust, bar features may dominate the light coming from a late-type galaxies at 2 µm (assuming the bar is composed of an older population of stars)-2MASS is ideally suited for studying bars in nearby spirals. We should again point out that the respective axis ratios are determined at different surface brightness levels (18.8 vs. 25 mag arcsec^-2) and thus subject to different morphological features, moderating our overall interpretation of the results.

5.3 Surface Brightness and Integrated Flux

The K_s elliptical isophotal photometry includes the core, bulge and disk components, representing a "total" integrated flux. In contrast, the central surface brightness represents the flux coming from the core or bulge component of a galaxy. By comparing these two measures, we approximate the bulge-to-disk ratio that is traditionally used to construct the Hubble sequence.

The central surface brightness is measured both from the peak pixel flux (peak flux coming from one 1´´ × 1´´ pixel) and from the mean value within 5´´ of the nuclear peak. Due to the pixel undersampling and resolution (~2´´), the peak surface brightness is of only limited use. For the 2MASS catalog, the mean central surface brightness (r <= 5´´) is the most meaningful measure of the central surface brightness. In Figs. 9-13 we show the resultant J and K_s-band surface brightness as a function of the K_s integrated flux for ellipticals (Fig. 9), normal spirals (Fig. 10), barred-transition spirals (Fig. 11), barred spirals (Fig 12) and other types (Fig. 13). The 1.3 µm and 2.2 µm central surface brightness is labeled J₅ and K_s5, respectively. The J-band surface brightness is denoted with filled circle symbols and the K_s-band with cross symbols (note: J is about 1 mag fainter than K_s). Here we have fit a line to the points corresponding to SA0/S0 type galaxies, denoted with dashed lines; also see solutions below. Here lenticulars represent the earliest spiral type. The same S0 fit is shown for each morphological type for comparison. The panels are in sequence of early to late-types (from bottom to top).

Elliptical and lenticular spirals have similar bulge surface brightness to total flux ratios (Fig. 9), with only a slightly higher surface brightness for E0 in comparison to S0. Although small, this result is consistent with Elliptical-S0 evolution sequence. The ratios are more dramatic when comparing early-type spirals to late-type spirals (Fig. 10 - 12). In comparison to late-type spirals, early-type spirals exhibit both a higher surface brightness (i.e., more prominent bulges) and less dispersion in their distribution. The surface brightness difference between E0/S0 and Sc, for example, is large enough (~0.3 mag arcsec^-2) that it is possible to discriminate between individual lenticular or elliptical galaxies and disk-dominated spirals, at least for high SNR (>10) that are clearly resolved (apparent size > 20´´). This can be understood by plotting the residuals relative to the best-fit line to the SA0/S0 distribution:

Fig 14 shows the S0/SA0 histogram distribution corresponding to the difference between the measured surface brightness (at K_s mag) and the best fit solutions (Eq. 1). The J-band surface brightness residual is shown with a solid line and the K_s with a dotted line. Ignoring the low surface brightness tail, the residuals shows a dispersion of only ~0.12 mag arcsec^-2 (Fig 14), a value smaller than the average surface brightness difference between early and late-type spirals, as compared in Fig. 15. At the most extreme, late-type spirals are systematically ~0.4 to 0.5 mag arcsec² fainter than elliptical/lenticular galaxies, although the dispersion is also greater due to the lower SNR for late-type disk galaxies (in addition to their more complex morphology; see Figs 2-4). Intermediate-type spirals are systematically fainter than lenticulars by ~0.2 mag arcsec^-2, while early-type spirals (e.g., Sa) are ~0.1 mag arcsec^-2 fainter than lenticulars in their central cores. Even though these are statistical trends, it is clear that S0/E -type galaxies can be delineated from disk-dominated galaxies using only near-infrared measurements. This result is consistent with that of Grauer & Rieke (1998), who acquired deep 1-2 µm infrared imaging of many nearby UGC galaxies, finding that central surface brightness is strongly correlated with Hubble type. Interestingly, this surface brightness effect is much less prominent at optical wavelengths for some studies (cf. Odewahn & Aldering 1995).

Even with a crude measure of the mean central surface brightness, the Hubble galaxy types can be differentiated statistically. A cleaner separation between types should be possible using a better representation of the surface brightness, including those derived from the "effective" or half-light radius (see below). Similarly, by combining physical information such as disk scale length, isophotal size and more precise (i.e., decoupled) bulge-to-disk surface brightness and color ratios as derived from infrared and optical imaging, more robust taxonomy of spiral and barred-types should be possible.

5.4 Comparison with Optical Properties

Compared to visual-band imaging, ground-based infrared imaging (e.g., 2MASS) is not as sensitive to low surface brightness emission from galaxies due to the high backgrounds inherent to the 1-2 µm atmospheric windows. On the other hand, the NIR is less affected by dust extinction, both within our Galaxy and within the targeted galaxies. A detailed comparison between the bands provides the first step toward understanding the properties of the stellar light coming from the different large-scale components (cf. Grauer & Rieke 1998). Here we compare the NIR properties of 2MASS galaxies with the corresponding measures in the RC3 (de Vaucouleurs et al. 1991), including the size, total flux, central surface brightness and elongation. To help define the context, we first begin with a qualitative comparison between optical and NIR images of bright galaxies.

In Figure 16 we show 5 different cases: spheroidal E2 galaxy (ESO 507-G046), early-type Sa spiral (NGC 4305), barred SBb spiral (NGC 3485), late-type Sd spiral (NGC 5892) and Seyfert-type active galaxy (MRK 463E). The Ks-band flux for each galaxy is ~10^th mag (~75 mJy), they are relatively bright and large galaxies for 2MASS. The J-K_s colors range from 0.8 to 1.0 mag, except for MRK 463E, which has a very red color of 2.2 mag (see also Plate 51). The galaxies were also chosen due to their mostly face-on orientation, which aids comparison of large-scale features. The left column represents visual imaging, where we display the Digitized Sky Survey (DSS) image using a gray-scale halftone with contours overlaid. The right column represents the 2MASS K_s-band, shown with gray-scale halftone, overlaid with J-band contours. For both sets of data the gray-scale stretches and contours were chosen to highlight similar features, allowing for qualitative comparison. Moreover, the optical contours show the surface brightness difference between the optical and infrared for the Hubble sequence: the lowest level contour ranges between 24.5 and 23.0 B mag arcsec^-2, from early to late-type.

The DSS images are more sensitive to low surface brightness emission, particularly for late-type ("blue") spirals, owing to their lower backgrounds and favorable colors. Nevertheless, the 2 µm images do show the same features as that seen in the 0.7 µm images, including spiral arms, bars, and asymmetries. For the active galaxy, MRK 463E, the "double" nucleus feature is distinctly different between data sets, probably due to dust extinction. Although 2MASS imaging alone is not sufficiently sensitive to provide unambiguous classification between spiral sub-types, in combination with optical data it should be possible to clearly distinguish between the classic Hubble types of galaxies.

A detailed comparison is made with the RC3, including ~900 galaxies with surface brightness measurements, delineated into 50 to 100 galaxies per Hubble type. Here we compute the ratio between the 2MASS isophotal size, D₂₀, corresponding to the 20 K mag arcsec^-2 isophote, with the RC3 D₂₅ isophotal size, corresponding to the 25 B mag arcsec^-2 isophote. For the total flux comparison, we take the difference between the 2MASS integrated flux (from D₂₀) and the RC3 B_T mag (derived from D₂₅). For the axis ratio comparison, we multiply the K-band axis minor-to-major axis ratio (K_b/a; derived from the 3-sigma contour, roughly equal to 18.8 K mag arcsec^-2) with the RC3 major-to-minor axis ratio, R₂₅. Finally, for the central surface brightness, µ_e (B mag arcsec^-2), we use the RC3 "effective" aperture, A_e, roughly corresponding to the half-light radius. Since the current (first release) version of the 2MASS galaxy catalog does not include a half-light or "effective" aperture, it was necessary to perform additional photometry on the 2MASS images using an aperture matched to the RC3 effective aperture, A_e. In order to minimize source contamination in the photometry, stars were carefully subtracted from the 2MASS images using the appropriate PSF image. To avoid the effects of undersampling, galaxies with an A_e diameter < 14´´ were not included in the analysis (but representing only a tiny fraction of the total number of sources). The absolute calibration uncertainty is estimated to be less than 5%. The resultant surface brightness, µ_Ke (K mag arcsec^-2), can be directly compared with corresponding optical version.

The results are shown in Table 2. For early types, the NIR isophotes extend down to about 75% of the corresponding optical size. Galaxies appear even smaller in the NIR for the later Sc/d types, ~56% of the optical size. For barred spirals the trend is similar from early to late-type, but the amplitude is smaller (76 vs. 67%), while the dispersion is significantly larger for late-types. The dispersion might be indicative of the lower signal-to-noise in the K band images relative to the optical (see below). The results are also consistent with a more prominent bar, as seen in the NIR. The bar is increasing the apparent size and elongation, K_b/a*R₂₅, of the galaxy (see also §5.2). The effective (half-light) diameter of the optical images, A_e, is about 45 to 60% in size of the K-band isophotal diameter (recall that 20 mag arcsec^-2 is about 1-sigma of the background noise), again showing the same trend from early to late types. For the total flux comparison, B_T-K_s, the average color is about 3.9 to 4.0 for early types, decreasing to 3.1 for late-types (even bluer, 2.9 mag, for barred spirals). The NIR sensitivity is decreasing for the later-type (i.e., "blue") galaxies.

This color effect is most clearly seen in the central surface brightness comparison, µ_e-µ_Ke, where we have matched the apertures. Now the early types are much redder, with colors between 4.6 and 4.7 mag arcsec^-2, while for the late-types the color drops to ~3.8 to 4.1 mag arcsec^-2 (Fig. 17). Compared to the total B_T-K_s, the central surface brightness color is redder from to two effects: dust extinction (absorbing mostly blue light) and the presence of older (redder) stars in the core and halo of galaxies. The color gradient from early to late-types (Fig 17) maps the change in stellar populations. The gradient is strongest for the barred spirals, which suggests a younger massive population of stars (i.e. starburst) associated with the bar feature (see also Fig 18). It is interesting to note that irregular and peculiar galaxies, including mergers, tidal interactions, and starbursts, also have very blue colors similar to barred spirals (but note the large dispersion). As the late-type galaxies evolve, their cores will gradually redden and brighten into early-type disks or spheroids.

We present an image Atlas of a sample of galaxies observed with the Two Micron All Sky Survey. The images are separated according to the traditionally-based Hubble type or morphological classification. In addition to the main elliptical/spheroidal and spiral-type classes, we also include examples of compact, dwarf, irregular/peculiar and AGN/Seyfert types of extragalactic objects. A simple analysis of the Atlas shows that the different classes are not as well separated with infrared colors (the exception being the extremely red class of AGN galaxies), but are clearly distinguished when comparing central surface brightness and total flux. A detailed comparison between optical and infrared properties, including sizes, elongation, colors and surface brightness, reveals clear differences between spiral types and between barred and normal disk galaxies. With more sophisticated measurements (e.g., decoupled bulge-to-disk ratios) and methods (e.g., supervised neural networks), and in combination with optical imaging, it should be possible to derive morphological classifications for a data set that is numbered in the millions.
 
Acknowledgements
We thank Harold Corwin for his wise council and RC3 data sets. We thank Tom Chester, Joe Mazzerella, Barry Madore, Marion Schmitz and George Helou for their insightful comments. We thank Steve Odewahn for a careful review of this paper. We thank Geraldine Jarrett for proofreading the final manuscript. This publication makes use of data products of the 2MASS, which is a joint project of the Univ. of Massachusetts and the Infrared Processing and Analysis Center, funded by the NASA and NSF. This research has made use of the NASA/IPAC Extragalactic Database (NED) which is operated by the Jet Propulsion Laboratory, California Institute of Technology, under contract with NASA. The Digitized Sky Surveys were produced at the Space Telescope Science Institute. The images of these surveys are based on photographic data obtained using the Oschin Schmidt Telescope on Palomar Mountain and the UK Schmidt Telescope.

Binggeli, B., Sandage, A. & Tammann, G. 1988, ARA&A, 26, 509.
Buta, R. 1992, in Morphological and Physical Classification of Galaxies, ed. Longo, G., Capaccioli, M, & Busarello, G. (Dordrecht: Kluwar), 1.
Buta, R., Mitra, S., de Vaucouleurs, G., & Corwin, H.G. 1994, AJ, 107, 118.
Buta, R. 1995, ApJS, 96, 39.
Cutri, R.M., et al., 2000, The 2MASS Explanatory Supplement.
de Vaucouleurs, G. 1959, Handbuch der Physik, Vol. III, 275.
de Vaucouleurs, G. 1977, in The Evolution of Galaxies and Stellar Populations, ed. B. Tinsley & R. Larson, (New Haven: Yale Univ. Observatory), p 43.
de Vaucouleurs, G. & Freeman, K.C. 1972, Vistas in Astronomy, 14, 163.
de Vaucouleurs, G., de Vaucouleurs, A., Corwin, H.G., Buta, R,. Paturel, G., & Fouque, P., 1991, Third Reference Catalogue of Bright Galaxies (RC3).
Dressler, A. 1980, ApJ, 236, 351.
Ferrini, f. & Galli, D. 1988, A&A, 195, 27.
Giovanelli, R. & Haynes, M.P. 1983, AJ, 88, 881.
Grauer, A.D. & Rieke, M. 1998, ApJS, 116, 29.
Hoffman, G.L, Helou, G., Salpeter, E.E. & Lewis, B.M. 1989, ApJ, 339, 812.
Holmberg, E. 1958, Medd. Lund. Ast. Obs. Ser. II, 136.
Hubble, E. 1926, ApJ, 64, 321.
Hubble, E. 1936, The Realm of the Nebulae. Yale Univ. Press (New Haven).
Hubble, E., & Humason, M. 1931, ApJ, 74, 43.
Impey, C. and Bothun, G. 1997, ARA&A, 35, 267.
Jarrett, T.H, Chester, T., Cutri, R., Schneider, S., & Huchra, J. 2000, AJ, in press.
Jarrett, T.H, Chester, Schneider, S., & Huchra, J. 1997, in The Impact of Large-Scale Near-IR Sky Surveys, ed. F. Garzon et al. (Dordrecht: Kluwar), 213.
Julian, W.H., Kooiman, B.L., & Sanders, W.L. 1997, PASP, 108, 297.
Kormendy, J.A. 1977, ApJ, 218, 333.
Mamon, G. 1992, ApJ, 401, L3.
Odewahn, S.C. & Aldering, G. 1995, AJ, 110, 2009.
Odewahn, S.C., Burstein, D., & Windhorst, R. 1997, AJ, 114, 2219.
Odewahn, S.C., Windhorst, R., Driver, S.P. & Keel, W.C. 1996, ApJL, 472, L13.
Osterbrock, D.E. 1993, ApJ, 404, 551.
Postman, M. & Geller, M.J. 1984, ApJ, 281, 95.
Roberts, M.S. & Haynes, M.P. 1994, ARAA, 32, 115.
Sandage, A. 1975, in Stars & Stellar Systems, vol. 9, ed. Sandage, A., Sandage, M. & Kristian, J. (Chicago: Univ. of Chicago Press), 1.
Sandage, A. 1961, the Hubble Atlas, Washington, D.C.: Carnegie Institution of Washington.
.Sandage, A., & Binggeli, B., 1984, AJ, 89, 919.
Sandage, A. 1986, A&A, 161, 89.
Sandage, A. 1990, in Clusters of Galaxies, ed. Oegerle, W., Fitchett, M. & Danly, L., (New York: Cambridge Univ. Press), 201.
Skrutskie, M., et al. 1997, in The Impact of Large-Scale Near-IR Sky Surveys, ed. F. Garzon et al. (Dordrecht: Kluwar), 25.
van den Bergh, S., 1994, AJ, 107, 153.
van den Bergh, S. 1998, Galaxy Morphology and Classification, (New York: Cambridge Univ. Press).
Whitmore, B.C., Gilmore, D., & Jones, C.1993, ApJ, 407, 489.

Plate 0.--The 2MASS near-infrared galaxy morphology sequence "trident," analogous to the Hubble sequence "tuning fork." The images were created by combining the J, H and K_s image data with an RGB color scale: J is assigned blue intensity levels, H green and K_s red intensity levels. Also shown are J-band contours, corresponding to the 10%, 20%, 30%, ..., etc, of the J-band peak flux. The integrated brightness ranges between 10.0 and 11.0 mag at 2.2 µm (30 to 70 mJy, respectively), and angular sizes range from 50 to 100´´. The 2MASS beam size, in comparison, is ~3´´.

Plates 1-54.--2MASS Galaxy Morphology Atlas. Each plate consists of 16 different galaxy images, separated according to morphology and integrated flux (see TABLE 1). The images are presented as gray-scale halftones of the K_s (2.2 µm) image, overlaid with contours corresponding to the J-band (1.3 µm) image. The gray-scale ranges from 1sigma (~20 mag arcsec^-2) of the background noise to the peak pixel value for the galaxy (but with a lower threshold of 5sigma). The gray-scale stretch includes a base-10 logarithm transformation, which enhances the contrast in the underlying galaxy emission. The J-band contours correspond to the 1%, 2%, 4%, 8%, 16% and 32% peak-intensity levels, with a minimum value corresponding to 2sigma (~20.8 mag arcsec^-2) of the J-band background. Also included are the catalog or NED name, morphological or Hubble type, 2MASS name, K_s mag and J-K_s color. The 2MASS name maps into the J2000 equatorial position of the source (Jhhmmsss+-ddmmss). The angular size of the image is noted in the lower right hand corner of the image, ranging from 21 to 101´´ (the beam size, in comparison, is ~3´´).

Figure 1.--Near-infrared color distribution for elliptical and lenticular type galaxies. Three different color combinations are shown: J-H (left column), H-K_s (middle column) and J-K_s (right column).

Figure 2.--Near-infrared color distribution for normal spiral galaxies, starting with lenticulars (bottom row panels) and sequentially later types moving upward: SAa (second row), SAb (third row) and SAc/SAd (top row).

Figure 3.--Near-infrared color distribution for normal-barred transition spiral galaxies, starting with early-type SAB0/SABa (bottom row panels), next with SABb (middle row), and finally, SABc/SABd (top row).

Figure 4.--Near-infrared color distribution for barred spiral galaxies, starting with early-type SB0/SBa (bottom row panels), next with SBb (middle row), and finally, SBc/SBd (top row).

Figure 5.--Near-infrared color distribution for compact (bottom row), dwarf (2^nd row), irregular/peculiar (3^rd row) and AGN/Seyfert (top row) types of galaxies.

Figure 6.--Axis ratio distribution for elliptical galaxies. The axis ratio, or the ratio between the minor-to-major axis, is derived from the combined J+H+K_s image (referred to as the "super" coadd; see Jarrett et al. 2000), and corresponds to the best-fit ellipse to the 3-sigma isophote for each galaxy (this is typically at a surface brightness of ~20.3, 19.4 and 18.8 mag arcsec^-2 at 1.3, 1.6 & 2.2 µm, respectively).

Figure 7.--Axis ratio distribution for normal spiral galaxies. See Figure 6 caption for details.

Figure 8.--Axis ratio distribution for barred spiral galaxies. See Figure 6 caption for details.

Figure 9.-- Near-infrared central surface brightness as a function of the K_s integrated flux for elliptical and lenticular galaxies. The 1.3 µm and 2.2 µm mean (radius <= 5´´) surface brightness is labeled as J₅ and K_s5, respectively. The J-band surface brightness (in mag arcsec^-2) is denoted with filled circle symbols and the K_s-band with cross symbols. The J-band points generally lie above (i.e., fainter) the K_s points. A line (denoted with dashed lines) has been fit to the points corresponding to the lenticular or SA0/S0 type galaxies (see Eq. 1). For comparison, the same lines are drawn for the later-type galaxies.

Figure 10.--Near-infrared central surface brightness as a function of the K_s integrated flux for normal spiral galaxies. See Figure 9 caption for details.

Figure 11.--Near-infrared central surface brightness as a function of the K_s integrated flux for normal-barred transition spiral galaxies. See Figure 9 caption for details.

Figure 12.--Near-infrared central surface brightness as a function of the K_s integrated flux for barred spiral galaxies. See Figure 9 caption for details.
Figure 13.--Near-infrared central surface brightness as a function of the K_s integrated flux for compact, dwarf, irregular and AGN type galaxies. See Figure 9 caption for details.

Figure 14.--Residual central surface brightness for lenticular type galaxies after subtracting the surface brightness to integrated flux relation (Eq. 1). The J-band residual is denoted with a solid line and the K_s with a dashed line. The distribution dispersion is ~0.12 mag arcsec^-2.

Figure 15.--Residual central surface brightness for normal spiral galaxies after subtracting the lenticular surface brightness to integrated flux relation (Eq. 1).

Figure 16--Comparison between optical and infrared imaging. Five galaxies are illustrated: spheroidal E2 galaxy (ESO 507-G046), early-type Sa spiral (NGC 4305), barred SBb spiral (NGC 3485), late-type Sd spiral (NGC 5892) and Seyfert-type active galaxy (MRK 463E). The Ks-band integrated flux for each galaxy is ~10.0 mag (~75 mJy). The left column show gray-scale Digitized Sky Survey images with contours overlaid. The right column shows 2MASS K_s-band images (gray-scale halftone), overlaid with J-band contours. The gray-scale stretch and contour levels were chosen to highlight similar features between the DSS and 2MASS data. The average surface brightness of the lowest contour for the optical images is ~24.6, 24.0, 23.4, 23.0, and 24.0 mag arcsec^-2, reading from top to bottom (early to late-types).

Figure 17-- Comparison between the optical and near-infrared surface brightness for different Hubble types. The B-band surface brightness, µ_e, is computed from the RC3 "effective" aperture (half-light diameter, A_e). The K-band surface brightness, µ_Ke, is computed using a matched aperture. The error bars represent the rms scatter in the distribution.

Figure 18--Comparison between the optical and near-infrared surface brightness for elliptical (filled circles) and late-type barred spirals (crosses). The B-band surface brightness, µ_e, is computed from the RC3 "effective" aperture (half-light diameter, A_e). The K-band surface brightness, µ_Ke, is computed using a matched aperture.
 
 

Table 1. Near-Infrared Morphology Atlas Plates

 
Table notes: D₂₀ is the K-band elliptical isophotal diameter corresponding to 20 mag arcsec^-2 surface brightness. D₂₅ is the RC3 optical radius (25 B mag arcsec^-2). A_e is the RC3 optical apparent diameter of the "effective" aperture (i.e., the half-light radius). B_T is the RC3 total B mag (derived from D₂₅) and Ks is the 2MASS isophotal mag (at 20 mag arcsec^-2). The mean surface brightness (mag arcsec^-2) corresponding to the effective aperture (the half-light radius) is µ_e from the RC3, and µ_Ke from 2MASS using the same RC3 aperture. The K-band 2MASS minor-to-major axis ratio is K_b/a and the RC3 major-to-minor isophotal axis is R₂₅. Each bin has between 50 and 100 sources. The RMS uncertainties are given in the (brackets).
 

Near-Infrared "Tuning Fork", Plate 0

Spheroid/Ellipticals

Lenticulars

Normal Spirals

Transition Spirals

Barred Spirals

Compact

 Dwarf

 Irregular/Peculiar

 AGN/Seyfert