1: \documentclass{emulateapj}
2:
3: \bibliographystyle{apj}
4:
5:
6: \slugcomment{Accepted for publication in ApJ.}
7: \shorttitle{Galaxy Classification}
8: \shortauthors{Bailin and Harris}
9:
10: % citation short forms
11: \defcitealias{shao-etal07}{S07}
12:
13:
14: % short commands
15: \newcommand{\gr}{{\ensuremath{(g-r)}}}
16: \newcommand{\grfo}{{\ensuremath{(g-r)^{\mathrm{F}}}}}
17: \newcommand{\absmag}{{\ensuremath{M_r}}}
18: \newcommand{\absmagfo}{{\ensuremath{M_r^{\mathrm{F}}}}}
19: \newcommand{\cpetro}{{\ensuremath{C_{\mathrm{Petro}}}}}
20: \newcommand{\cnorm}{{\ensuremath{C_{\mathrm{norm}}}}}
21: \newcommand{\cmd}{{\ensuremath{\mathrm{CMD}}}}
22: \newcommand{\cmdfo}{{\ensuremath{\mathrm{CMD}^{\mathrm{F}}}}}
23:
24: \begin{document}
25:
26: \title{Inclination-Independent Galaxy Classification}
27:
28: \author{Jeremy Bailin\altaffilmark{1} \and William E. Harris}
29: \affil{Department of Physics \& Astronomy, McMaster University,
30: 1280 Main Street West, Hamilton, ON, L8S 4M1, Canada}
31: \altaffiltext{1}{bailinj@mcmaster.ca}
32:
33:
34: \begin{abstract}
35: We present a new method to classify galaxies from large surveys
36: like the Sloan Digital Sky Survey using
37: inclination-corrected concentration, inclination-corrected location
38: on the color-magnitude diagram, and apparent axis ratio.
39: Explicitly accounting for inclination
40: tightens the distribution of each of these parameters and enables
41: simple boundaries to be drawn that delineate three different galaxy
42: populations: Early-type galaxies, which are red, highly concentrated,
43: and round; Late-type galaxies, which are blue, have low concentrations,
44: and are disk dominated; and Intermediate-type galaxies, which are red,
45: have intermediate concentrations, and have disks.
46: We have validated our method by comparing to visual
47: classifications of high-quality imaging data from the Millennium
48: Galaxy Catalogue.
49: The inclination correction is crucial to unveiling the
50: previously unrecognized Intermediate class.
51: Intermediate-type galaxies, roughly corresponding to lenticulars
52: and early spirals,
53: lie on the red sequence. The red sequence is therefore composed
54: of two distinct morphological types,
55: suggesting that there are two distinct mechanisms
56: for transiting to the red sequence. We propose that Intermediate-type
57: galaxies are those that have lost their cold gas via strangulation,
58: while Early-type
59: galaxies are those that have experienced a major merger that either
60: consumed their cold gas, or whose merger progenitors were already
61: devoid of cold gas (the ``dry merger'' scenario).
62: \end{abstract}
63:
64: \keywords{galaxies: fundamental parameters --- galaxies: structure}
65:
66: \section{Introduction}
67: Galaxies do not smoothly occupy the full range of
68: possible structural and photometric parameter space. This fact was
69: first pointed out by \citet{hubble26}, whose ``Tuning Fork''
70: classified galaxies into ellipticals, spirals, and barred
71: spirals. The basic categorization of ``early''-type galaxies,
72: with a smooth quiescent elliptical appearance, and ``late''-type galaxies,
73: characterized by a spiral disk with sites of active star formation,
74: has remained to this day.
75:
76: Many studies have attempted to put this classification on a
77: quantitative footing.
78: Bimodality in the color-magnitude plane has been well established
79: \citep[e.g.][]{strateva-etal01,blanton-etal03-properties,baldry-etal04,
80: balogh-etal04-colour}, and provides the most straightforward
81: classification method.
82: \citet{weinmann-etal06a} classified galaxies into three
83: types based on their colors and spectra: early-types, which
84: are red and passive; late-types, which are blue and active;
85: and intermediate-types, which are red and active.
86: These authors argue that their intermediate class are most
87: likely a combination of early-types whose star formation rates are
88: overestimated and reddened edge-on late-types.
89: \citet{ellis-etal05} statistically studied the multivariate distribution of
90: a large number of observed galaxy properties and found that exactly
91: two classes are required.
92: One particularly useful diagnostic plane for classifying galaxies
93: is color versus global concentration of the light profile,
94: where early-type galaxies appear red and concentrated while
95: late-type galaxies appear blue and have low concentrations
96: \citep[e.g.][]{driver-etal06}.
97:
98: The use of such automated classifications is crucial in large galaxy
99: surveys such as those now available, for which visual examination
100: of hundreds of thousands of galaxies is unfeasible%
101: \footnote{See, however, the Galaxy Zoo project:
102: http://galaxyzoo.org/.}. Moreover, automated methods are important
103: even when visual classifications are available, as they are
104: quantitative and allow finer and unbiased classification in cases
105: where the patterns identified by eye do not map correctly onto the
106: physically-important parameters.
107:
108: Most observational parameters that are used in classification methods
109: are either calculated within a circular aperture or in circular
110: annuli. However, disk galaxies are not usually circular when viewed
111: on the sky, and their appearance varies systematically as a function
112: of inclination. For example, the front and back sides of a disk
113: contribute most of their light at small radii if the disk is highly
114: inclined, but at large radii if the disk is nearly face-on; inclined
115: disks therefore appear to be more concentrated.
116: Observed colors and magnitudes
117: are also affected by the increased dust absorption in highly inclined
118: disks relative to nearly face-on disks. Inclination effects therefore
119: introduce significant scatter in the observed parameters of individual
120: disk galaxies.
121:
122: Fortunately, it is possible to calculate and correct for the effects
123: of inclination on several observed parameters.
124: In \citet{bailinharris08-trimodality} (Paper~I), a
125: simple inclination-independent measure of galaxy concentration was derived
126: based on the observed Petrosian concentration index \cpetro\ and
127: the observed isophotal axis ratio $b/a$
128: \citep[see also][who recalculated the concentration in elliptical
129: apertures directly from the galaxy images]{yamauchi-etal05},
130: Recently, four studies have used large samples of galaxies to
131: determine the effects of inclination on galaxy photometry:
132: \citet{shao-etal07} (hereafter \citetalias{shao-etal07}),
133: \citet{ur08} and \citet{maller-etal08},
134: all using the Sloan Digital Sky Survey (SDSS), and
135: \citet{driver-etal07}, using the Millennium Galaxy Catalogue (MGC),
136: measured how the characteristic turnover in the luminosity function
137: of spiral galaxies varies
138: as a function of inclination, and provide simple parametrizations
139: of the relative extinction in a galaxy as a function of disk inclination.
140: These new methods allow us to correct galaxy properties for
141: inclination effects and revisit the problem of automated galaxy
142: classification.
143:
144: In this paper, we present a new method for classifying galaxies
145: that uses
146: inclination-corrected values for galactic concentration, color,
147: and absolute magnitude, along with the apparent axis ratio.
148: We use the statistical power of SDSS, which
149: allows us to examine the multivariate distribution of galaxy properties
150: with ample signal-to-noise, and the MGC, which provides
151: deep high-resolution images for direct visual classification.
152: We describe the selection of the galaxy samples in section~\ref{sample-section}.
153: The expressions for the inclination-corrected
154: quantities are defined in section~\ref{face-on-values-section}\ and
155: a face-on color-magnitude diagram of SDSS galaxies is presented.
156: In section~\ref{classification-section} we present our method of
157: classifying galaxies and demonstrate its effectiveness, and finally
158: we present our conclusions in section~\ref{conclusions-section}.
159:
160:
161: \section{Galaxy Samples}\label{sample-section}
162: Our primary galaxy sample comes from SDSS,
163: a 5-band optical and near-infrared
164: imaging and spectroscopic survey covering one quarter of the sky
165: \citep{sdss-technical-summary}. The final data release,
166: DR6 \citep{sdss-dr6}, contains imaging and spectroscopy of over half
167: a million galaxies with well defined selection criteria. Our sample
168: consists of the $486305$ galaxies that meet the Main Galaxy
169: Sample targeting criteria \citep{sdss-maingalaxysample}, have
170: spectra that are classified as galaxies with confident redshifts
171: ($z_{\mathrm{conf}}>0.85$), have dereddened Petrosian $r$-band
172: magnitudes $r<17.7$, lie in the redshift range $0.01 < z < 0.2$, and have
173: absolute magnitudes $\absmag < -17.5$. Petrosian magnitudes are
174: used throughout, and $r$-band quantities are used for all photometric
175: parameters. Colors and absolute magnitudes are k-corrected to $z=0$
176: with KCORRECT~v4\_1\_4 \citep{kcorrect}. We adopt
177: $\Omega_0=0.3$, $\Omega_{\Lambda}=0.7$,
178: and $H_0=70~\mathrm{km~s^{-1}~Mpc^{-1}}$.
179:
180: While SDSS provides us with the statistics necessary to measure the
181: multivariate distribution of galactic parameters, the imaging is relatively
182: shallow and often taken in poor seeing conditions. Therefore, in order
183: to obtain confident visual classifications of a subset of galaxies,
184: we have used the deeper and higher resolution imaging
185: data from the MGC \citep{mgc-imaging}.
186: This $37.5~\mathrm{deg}^2$ $B$-band survey is entirely contained within
187: the SDSS footprint, allowing cross-identification of all galaxies.
188: We have randomly selected $400$ SDSS galaxies from the above sample
189: that are also contained in the cleaned bright MGC galaxy catalog%
190: \footnote{Note that the default matched mgc\_sdss catalog contains
191: objects that lie in MGC exclusion regions and are
192: therefore not suitable for scientific
193: analysis. We have obtained a version of the mgc\_sdss catalog
194: cleaned of such objects from J. Liske.}
195: for direct visual classification
196:
197:
198:
199: \section{Correcting for Inclination}\label{face-on-values-section}
200: \subsection{Concentration}\label{face-on-concentration-section}
201: In Paper~I, we presented a method to correct the Petrosian concentration
202: \cpetro\ for inclination effects. We briefly review the method here.
203:
204: As disk galaxies are seen closer to edge-on, the isophotal axis ratio
205: $b/a$ decreases while the global concentration
206: of the light profile increases. As a result, the loci of
207: intrinsically similar galaxies in
208: the \cpetro-$b/a$ plane curve to higher \cpetro\ at smaller
209: $b/a$. By constructing models of a pure exponential disk viewed at
210: a variety of inclinations, we have determined the expected
211: concentration of the disk $C_{\mathrm{Petro}}^{\mathrm{disk}}$
212: as a function of the observed isophotal axis ratio $b/a$.
213: We use this relationship to define the normalized
214: concentration, $\cnorm_{,i}$, for a galaxy with observed isophotal
215: axis ratio $(b/a)_i$ and observed Petrosian concentration $\cpetro_{,i}$ to be:
216: \begin{equation}
217: \cnorm_{,i} \equiv \frac{\cpetro_{,i}}{C_{\mathrm{Petro}}^{\mathrm{disk}}
218: (b/a_i)}.
219: \end{equation}
220: The loci of galaxies in the inclination-corrected \cnorm-$b/a$ plane no
221: longer depend on axis ratio, validating the use of \cnorm\ as
222: an inclination-independent measure of galaxy concentration.
223:
224: Note that although elliptical galaxies do not suffer from the same
225: inclination effect as disks, they all have large apparent axis ratios.
226: At these large axis ratios, $C_{\mathrm{Petro}}^{\mathrm{disk}}$ is
227: almost independent of $b/a$, and therefore the magnitude of the
228: inclination correction is minimal. Therefore, applying the correction
229: to all galaxies does not introduce any bias in the concentrations
230: of elliptical galaxies.
231:
232:
233:
234: \subsection{Magnitudes and colors}\label{face-on-magnitudes-section}
235: \citetalias{shao-etal07} find that the extinction in an inclined
236: disk galaxy relative to a face-on disk galaxy in a given band
237: is well described by:
238: \begin{equation}
239: \label{magcorrect-eq}
240: M - M^{\mathrm{F}} = -\gamma_2 \log(\cos i)
241: \end{equation}
242: for observed absolute magnitude $M$, face-on
243: absolute magnitude $M^{\mathrm{F}}$, inclination $i$
244: (defined to be $0\degr$ for face-on disks and $90\degr$ for
245: edge-on disks), and
246: relative extinction coefficient $\gamma_2$. Values of
247: $\gamma_2$ for each SDSS band are given in \citetalias{shao-etal07}.
248:
249: In \citetalias{shao-etal07}, the inclination $i$ was determined from the
250: apparent axis ratio of the exponential component of the surface brightness
251: decomposition (``expab'' in the SDSS database)
252: by Monte Carlo simulations. We simply use the median mapping between
253: expab and $\cos i$ given in their figure~3. We then calculate the face-on
254: absolute $r$-band magnitude, \absmagfo, and
255: face-on \gr\ color, \grfo, using
256: equation~(\ref{magcorrect-eq}):
257: \begin{equation}
258: \absmagfo = \absmag + \gamma_2^r \log(\cos i)
259: \end{equation}
260: \begin{equation}
261: \grfo = \gr + (\gamma_2^g - \gamma_2^r) \log(\cos i).
262: \end{equation}
263:
264: Correcting for the disk inclination
265: presumes the presence of a disk, while many galaxies are almost
266: entirely spheroidal systems.
267: Unlike the concentration
268: (\S~\ref{face-on-concentration-section}), the photometry is
269: sensitive to the inclination correction even at relatively large
270: axis ratios,
271: and should be omitted for elliptical galaxies.
272: We therefore only apply the correction to galaxies for which the bulge+disk
273: surface brightness decomposition in SDSS contains at least a $20\%$
274: disk, i.e. for which $\mathrm{fracdeV} < 0.8$. We include
275: galaxies where the bulge contributes a large fraction of the flux
276: (e.g. compared to \citetalias{shao-etal07}, who restrict their
277: analysis to $\mathrm{fracdeV} < 0.5$,
278: and \citealt{ur08}, who restrict their
279: analysis to $\mathrm{fracdeV} < 0.1$)
280: because the centrally-concentrated
281: bulge light is severely attenuated by disk dust even in the absence of
282: dust within the bulge itself, and
283: therefore the inclination correction is vitally important even in
284: systems with large bulges \citep{driver-etal07}.
285:
286: \begin{figure*}
287: \plottwo{f1a.eps}{f1b.eps}
288: \caption{\label{cmd}%
289: \textit{(Left)} Observed rest-frame \gr\ color versus \absmag\ color-magnitude
290: diagram for SDSS galaxies. The distribution of \gr\ marginalized over
291: absolute magnitude is shown in the histogram on the right. The solid
292: line denotes the red sequence.
293: \textit{(Right)} As in the left panel, but for face-on color \grfo\ versus
294: face-on absolute magnitude \absmagfo. The vector denotes the inclination
295: correction for a galaxy with $i=60\degr$.%
296: }
297: \end{figure*}
298:
299: We plot the observed color-magnitude diagram (CMD) and the
300: inclination-corrected face-on color-magnitude diagram in Figure~\ref{cmd}.
301: The marginal distributions of \gr\ and \grfo\ are added as histograms.
302: The basic bimodality of the CMD is apparent in both plots, with a
303: red sequence and a blue cloud. In the observed CMD (left panel),
304: the blue cloud appears
305: to run into the red sequence at the luminous end and it is difficult to
306: distinguish red from blue galaxies in this region. In contrast, the
307: populations are much more clearly separated in the face-on CMD (right panel).
308: Similarly, while the marginalized distribution of \gr\ contains a red
309: peak with a blue tail, the marginalized distribution of \grfo\ contains
310: a distinct blue peak
311: (see also \citealp{maller-etal08}, who noted the increased separation
312: between red and blue galaxies; and \citealp{ur08}, who noted the
313: tightening of the blue sequence among galaxies with exponential profiles).
314:
315: The effect of the inclination correction is directly demonstrated by
316: the vector in the right panel, which indicates the inclination correction
317: for a galaxy with $i=60\degr$. All inclination corrections occur parallel
318: to this vector.
319: The inclination correction moves galaxies almost perpendicular to the
320: red sequence, which is why it makes a relatively large difference in
321: the appearance of the CMD.
322:
323: We have fit the red sequence by eye as
324: \begin{equation}
325: (g-r)^{\mathrm{red}} = 0.77 + 0.02 (\absmag + 22).
326: \end{equation}
327: This is indicated by the solid line in both panels of Figure~\ref{cmd}.
328: We define the location of galaxy $k$ on the CMD relative to the red
329: sequence as its \cmd\ location parameter%
330: \footnote{Note that these definitions differ somewhat from the \cmd\ location
331: parameter used in Paper~I.}:
332: \begin{equation}
333: \cmd_k = \gr_k - (g-r)^{\mathrm{red}}(\absmag_{,k})
334: \end{equation}
335: and analogously for the face-on CMD location parameter \cmdfo:
336: \begin{equation}
337: \cmdfo_k = \grfo_k - (g-r)^{\mathrm{red}}(\absmagfo_{,k}).
338: \end{equation}
339:
340:
341:
342: \section{Inclination-Independent Classification}\label{classification-section}
343: Traditionally, galaxies have been classified into two types:
344: ``early-type'' or
345: ``elliptical'' galaxies, which are red, round, smooth, and
346: have centrally-concentrated light profiles well approximated by
347: the \citet{devaucouleurs48} $R^{1/4}$ law or
348: the \citet{sersic-profile} profile with large values of $n$;
349: and ``late-type'' or ``spiral'' galaxies, which are blue,
350: disk-shaped, contain spiral and clumpy structure, and have
351: less concentrated exponential light profiles
352: or \citeauthor{sersic-profile} profiles with low values of $n$.
353: An intermediate
354: ``lenticular'' or ``S0'' classification is also sometimes used, although
355: it is often considered an extension of either the elliptical or
356: spiral class depending on context.
357:
358: In this section, we attempt to put these classifications on a
359: firm quantitative footing by using the inclination corrections
360: developed in section~\ref{face-on-values-section} to define an
361: inclination-independent method of classifying galaxies.
362:
363:
364: \subsection{Visual classification}\label{mgcclass-section}
365: We have randomly selected $400$ galaxies contained in both our
366: SDSS sample and the MGC and inspected the MGC images by eye. The advantages
367: of the MGC images are several: (1) they are deeper than the
368: SDSS images, allowing us to see fainter features in the outer regions
369: of the galaxy that can aid in classification, (2) the typical seeing
370: in the MGC is better than in SDSS, allowing effectively higher
371: resolution, and (3) by using single-band images we explicitly ensure
372: that our visual classification is purely based on the morphology
373: and has no color bias.
374:
375: Each galaxy was given one of 5 classifications: ``elliptical'' (E),
376: ``elliptical/lenticular'' (E/L), ``spiral/lenticular'' (S/L),
377: ``spiral'' (S), or ``unclassifiable'' (X; usually a major merger).
378: Table~\ref{class-table} lists
379: the number of galaxies given each classification.
380:
381:
382: \begin{deluxetable}{lc|ccc}
383: \tablewidth{0pt}
384: \tablecaption{Comparison Between
385: Visual and Automated Classifications\label{class-table}}
386: \tablehead{%
387: \colhead{Visual} & & \multicolumn{3}{c}{Automated Classification}\\
388: \colhead{Classification} & \colhead{Total} &
389: \colhead{Early} & \colhead{Intermediate} & \colhead{Late}}
390: \startdata
391: E & $133$ & $72$ & $49$ & $12$\\
392: E/L & $36$ & $15$ & $15$ & $6$\\
393: S/L & $45$ & $8$ & $17$ & $20$\\
394: S & $174$ & $8$ & $13$ & $153$\\
395: X & $12$ & $6$ & $3$ & $3$\\ \hline
396: Total & $400$ & $109$ & $97$ & $194$\\
397: \enddata
398: \end{deluxetable}
399:
400:
401: \subsection{Automated classification}\label{autoclass-section}
402:
403: \begin{figure*}
404: \plottwo{f2a.eps}{f2b_color.eps}
405: %\plottwo{f2a.eps}{f2b_bw.eps}
406: \caption{\label{cmdcon}%
407: \textit{(Left)} Distribution of Petrosian concentration \cpetro\ versus
408: observed \cmd\ location parameter for SDSS galaxies. The marginalized
409: distributions are shown in the top histogram for \cpetro\ and the
410: right histogram for \cmd.
411: \textit{(Right)} As in the left panel for the inclination-corrected
412: concentration \cnorm\ and the face-on CMD location parameter \cmdfo.
413: The green dashed line in the \cmdfo\ histogram shows the bimodal Gaussian
414: fit to the distribution, while the red and blue dotted lines show the two
415: components of the fit.%
416: }
417: \end{figure*}
418:
419: Our primary classification is performed in the color-concentration plane
420: \citep[e.g.][]{driver-etal06}. We adopt
421: the \cmd\ location parameter as the measure of color, as it provides
422: a better estimate of whether a galaxy lies on or off the red sequence
423: than \gr\ alone.
424: The joint distribution of observed concentrations \cpetro\ and observed
425: \cmd\ locations is shown in the left panel of Figure~\ref{cmdcon},
426: along with the distribution of each parameter independently. The tendency
427: for red galaxies to be more concentrated is apparent, as is an overall
428: bimodality of the galaxy population: there is a strong peak
429: of red galaxies at high concentrations ($2.5 \la \cpetro \la 3.4$)
430: and a more spread-out tail of blue galaxies at low concentrations.
431:
432: The right panel of Figure~\ref{cmdcon} shows the joint and individual
433: distributions of the inclination-corrected concentration \cnorm\ and
434: the face-on color \cmdfo. The inclination correction makes several
435: appreciable improvements to this diagnostic plane. Firstly, as
436: described in Paper~I, the distribution of \cnorm\ is
437: \textit{tri}modal:
438: there is a distinct grouping of galaxies at intermediate
439: concentrations.
440: This trimodality has been previously unrecognized because inclination
441: effects smear together the intermediate- and high-concentration
442: subpopulations.
443: The color distributions, and therefore the luminosity-weighted stellar age,
444: of these two populations are similar, but they have distinct morphologies.
445: Secondly, the distribution of \cmdfo\ is much more bimodal
446: than that of \cmd: the previous blue tail that merged into the red sequence
447: has become a distinct peak.
448: We have employed the histogram fitting package RMIX
449: \citep[see also our more extensive description in Paper~I]{macd07}
450: to experiment with a range of possible multimodal fits that employ
451: relatively simple basis functions (Gaussian, lognormal, gamma
452: functions, etc.). Unlike the marginal distribution for \cnorm,
453: which clearly requires three components (see Paper~I), a bimodal
454: distribution fits the \cmdfo\ histogram successfully with no clear
455: evidence that would call for more than two components
456: \citep[see also][]{ellis-etal05}.
457: In Figure~\ref{cmdcon}b, we show the best-fit RMIX solution
458: that matches the \cmdfo\ distribution as the sum of two
459: Gaussian components. In this solution,
460: $60\%$ of the galaxies are assigned to the blue
461: peak while $40\%$ are assigned to the red peak.
462: Finally, the separation between classes in the joint distribution is much
463: stronger.
464: Rather than simply lying in
465: a spread-out tail, the late-type galaxies are relatively localized
466: in a distinct and well-separated region of parameter space from the
467: intermediate and early-type galaxies, which are themselves distinct.
468:
469: \begin{figure*}
470: \plottwo{f3a_color.eps}{f3b_color.eps}
471: %\plottwo{f3a_bw.eps}{f3b_bw.eps}
472: \caption{\label{cmdconeye}%
473: \textit{(Left)} The grayscale denotes the distribution of Petrosian
474: concentration parameters \cpetro\ versus observed \cmd\ location parameter
475: for SDSS galaxies, as in Figure~\ref{cmdcon}. The locations of the
476: visually classified galaxies are overplotted. The visual classifications
477: are: elliptical (``E'', red diamonds), elliptical/lenticular (``E/L'',
478: purple triangles), spiral/lenticular (``S/L'', green asterisks),
479: spiral (``S'', blue pluses), and unclassifiable (``X'', black crosses).
480: \textit{(Right)} As in the left panel for the inclination-corrected
481: concentration \cnorm\ and the face-on CMD location parameter \cmdfo.
482: The solid line denotes our adopted boundary between late-type galaxies
483: and intermediate/early-type galaxies, while the dashed line denotes
484: a possible boundary between intermediate and early-type galaxies based
485: on the trimodality of the \cnorm\ parameter.%
486: }
487: \end{figure*}
488:
489: In Figure~\ref{cmdconeye} we have overplotted the locations of
490: the visually classified galaxies onto the \cmd-\cpetro\ and
491: \cmdfo-\cnorm\ distributions. The symbol types and colors represent the
492: visual classification. In both cases, the distribution at the
493: low-concentration blue end is dominated by spiral galaxies while the
494: high-concentration red end is dominated by elliptical galaxies.
495: In the uncorrected distribution, the intermediate type galaxies span
496: a wide range of parameter space, and in particular while elliptical
497: galaxies dominate the high \cpetro-high \cmd\ regime, there are also
498: a large number of lenticular galaxies and a non-negligible number of
499: spiral galaxies that also lie in this region of parameter space.
500: In contrast, the separation between the regions is much cleaner
501: in the inclination-corrected distribution. The inclination
502: correction moves galaxies to the left and downward, almost
503: completely evacuating the high \cnorm\ region of spirals and
504: spiral/lenticulars.
505:
506: Based on the locations of visually classified galaxies on this plane
507: and on the shape of the joint distribution, we have drawn
508: a boundary in the \cnorm-\cmdfo\ plane to separate late-type galaxies
509: from intermediate- and early-type galaxies, indicated by the solid
510: line in the right panel of Figure~\ref{cmdconeye}. Within the
511: intermediate- and early-type region of parameter space, there is some
512: overlap in the distributions of the various visual classifications,
513: although there is a clear trend for more later types at smaller \cnorm.
514: We tentatively place a border between intermediate- and early-type
515: galaxies at $\cnorm=1.23$ (indicated by the dashed line), which is
516: the cross-over point between the intermediate and high-\cnorm\ peaks
517: in the \cnorm\ histogram (see Figure~\ref{cmdcon}).
518:
519:
520: \begin{figure}
521: \plotone{f4_color.eps}
522: %\plotone{f4_bw.eps}
523: \caption{\label{bacon}%
524: The grayscale denotes the distribution of inclination-corrected concentration
525: \cnorm\ versus isophotal $b/a$ axis ratio for SDSS galaxies
526: (see Paper~I). The locations of visually classified galaxies that lie
527: in the ``Intermediate'' and ``Early'' regions of Figure~\ref{cmdconeye}
528: are overplotted, with the same symbol types as in Figure~\ref{cmdconeye}.
529: The solid line denotes our adopted separation between intermediate-type
530: and early-type galaxies; this line is identical to the red line
531: defining the boundary of early-type galaxies in figure~2 of Paper~I.
532: }
533: \end{figure}
534:
535: We have further examined the distribution of visually-classified galaxies
536: in the axis ratio-concentration ($b/a$-\cnorm) plane of Paper~I.
537: There are several reasons for doing this.
538: Firstly, the overlap in the distribution of visually-classified
539: galaxies of a given type within the Intermediate and Early regions
540: of Figure~\ref{cmdconeye} may perhaps be disentangled if the
541: populations differ in
542: another morphological parameter.
543: Secondly, we note that in Paper~I, we found that flattened galaxies with high
544: \cnorm\ have more in common with intermediate-concentration galaxies
545: than with ellipticals. While this may suggest a problem with our inclination
546: correction, the cause is simply statistics: the intrinsic axis ratio
547: distribution of ellipticals drops off dramatically at low $b/a$,
548: while the concentration distribution of intermediate-types is relatively
549: wide. Therefore, flattened galaxies with high \cnorm\ are more likely
550: to be unusually concentrated intermediate-types than unusually flattened
551: ellipticals.
552:
553: In Figure~\ref{bacon}, we show the distribution of all SDSS galaxies
554: as the grayscale,
555: along with the visual classification of those galaxies
556: that lie in the ``Intermediate'' and ``Early'' regions of
557: Figure~\ref{cmdconeye}. The visual classification of these galaxies
558: varies systematically as a function of axis ratio even at
559: a constant concentration,
560: and the ellipticals are very concentrated in the high-\cnorm\ high-$b/a$
561: region of parameter space. We therefore use this plane to separate
562: early-type galaxies from intermediate-type galaxies as shown.
563:
564: Our final classification is as follows:
565: \begin{itemize}
566: \item Early-type galaxies lie to the upper-right of the solid
567: boundary in Figure~\ref{cmdconeye} and to the upper-right of the
568: boundary in Figure~\ref{bacon}.
569: \item Intermediate-type galaxies lie to the upper-right of the solid
570: boundary in Figure~\ref{cmdconeye} and to the bottom-left of the
571: boundary in Figure~\ref{bacon}.
572: \item Late-type galaxies lie to the bottom-left of the solid
573: boundary in Figure~\ref{cmdconeye}.
574: \end{itemize}
575:
576: A comparison of our automated and visual classifications for each
577: galaxy is given in Table~\ref{class-table}. It can be seen that
578: we do an excellent job of recovering the visual classification.
579: $80\%$ of our Early-type galaxies are classified by
580: eye as E or E/L, while $90\%$ of Late-type galaxies are
581: classified as S or S/L by eye.
582: This is as good as could be expected given that
583: \citet{ellis-etal05} found that two people independently
584: visually classifying MGC images only agree at the $\sim 80\%$ level.
585:
586: The Intermediate classification,
587: despite occupying a relatively limited region of parameter space,
588: contains galaxies that have a range of visual morphologies.
589: However, an examination of Figure~\ref{bacon} reveals that
590: the visual classification of these galaxies, which have intrinsically
591: very similar structural and photometric parameters, varies systematically
592: with apparent axis ratio, i.e. inclination. While
593: these galaxies contain disks, their passive appearance makes it
594: difficult to identify the disk structure when seen face-on.
595: Without the quantitative information available from the automated
596: classification, it would be impossible to distinguish these galaxies
597: from elliptical Early-type galaxies.
598:
599:
600: \begin{figure}
601: \plotone{f5_color.eps}
602: %\plotone{f5_bw.eps}
603: \caption{\label{lumhist}%
604: Face-on absolute magnitude distribution of all SDSS galaxies (black/solid),
605: and those classified as Early (red/dot-dashed),
606: Intermediate (green/dashed), and Late (blue/dotted).}
607: \end{figure}
608:
609: We evaluate the relative importance of each population
610: as a function of luminosity
611: by examining their absolute magnitude distribution
612: (Figure~\ref{lumhist}).
613: We recover the well-known result that the high-luminosity and low-luminosity
614: ends of the galaxy population are dominated by early- and late-type
615: galaxies respectively \citep[e.g.][]{nakamura-etal03}.
616: As suggested in Paper~I, the intermediate-type
617: galaxies have a narrower luminosity distribution.
618: Although a detailed analysis of the
619: luminosity functions is beyond the scope this paper, it is apparent
620: from Figure~\ref{lumhist} that the intermediate-type galaxies are
621: unimportant at the faintest magnitudes, and therefore the observed
622: intermediate types are representative of the population as a whole
623: and not simply the tip of a large low-luminosity ``iceberg''.
624:
625:
626:
627: \section{Conclusions}\label{conclusions-section}
628: We have presented a new method to automatically classify galaxies
629: from the SDSS or any equivalent survey given four basic photometric
630: parameters: concentration, color, axis ratio, and absolute magnitude.
631: This method explicitly corrects for inclination
632: effects, allowing us to avoid the systematic misclassification
633: of inclined versus face-on systems endemic to most other automated
634: classifications.
635: We prefer this method even over visual classification, usually
636: considered to be the gold standard, which has considerable person-to-person
637: scatter \citep[see][]{ellis-etal05} and
638: is also susceptible to systematic inclination biases, as suggested by
639: the variation of visual classifications as a function of axis
640: ratio along the intermediate \cnorm\ locus in Figure~\ref{bacon}.
641:
642: In Paper~I, we discovered that applying an inclination correction to
643: the observed concentration of SDSS galaxies
644: tightens an apparently bimodal distribution
645: into three distinct peaks. In this work, we have discovered that correcting
646: colors and absolute magnitudes for inclination effects also tightens the
647: ``blue cloud'' of the CMD into a much more distinct peak in parameter space
648: \citep[see also][]{ur08}.
649: Based on the joint distribution of concentration and CMD
650: location, we find that the galaxy population separates well into
651: \textit{three} distinct classes, labeled here as ``Early'',
652: ``Intermediate'', and ``Late''. We have validated our classification
653: scheme by comparing
654: to visual classifications of a subset of galaxies in the
655: high-quality MGC images.
656:
657: A consequence of the results presented here and in Paper~I
658: is that there is an Intermediate galaxy type,
659: distinct from both Early-types (traditional ellipticals) and Late-types
660: (traditional spirals). These galaxies have intermediate concentrations
661: consistent with typical bulge-to-disk ratios of $\sim 3$,
662: a wide range of apparent axis ratios implying a disk morphology,
663: and colors indistinguishable from those of ellipticals of the same
664: luminosity (i.e. the \cmdfo\ distributions are similar, although this
665: corresponds to a bluer median color for the Intermediate types due to their
666: lower median luminosity).
667: Intermediate types span the full range of
668: visual classifications, although this may be more indicative of
669: the fallibility of visual classification than of the intrinsic properties
670: of Intermediate types.
671:
672: It is interesting to contrast our galaxy types to those of
673: \citet{weinmann-etal06a}, who also identified an intermediate
674: class of galaxy. However, their intermediate types do not clump
675: in one particular region of parameter space but rather lie
676: at the intersection of the early- and late-type populations.
677: These authors consider that their intermediate-types may be
678: (a) edge-on reddened late-type galaxies, (b) early-type galaxies
679: whose star formation rates are overestimated, (c) a mixture of
680: (a) and (b) (their preferred explanation), or (d) a distinct population.
681: Our intermediate-types cannot be (a), because we have explicitly
682: corrected for inclination, nor (b) or (c), because the star formation
683: rate does not enter into our classification. Therefore, unlike
684: the intermediate-types identified by \citet{weinmann-etal06a},
685: our Intermediate classification must represent a truly distinct
686: galactic population. The lenticulars and bulgy disks that
687: constitute our Intermediate type are red and passive, and would
688: be classified as ``early'' in \citet{weinmann-etal06a}.
689:
690: The standard explanation for the bimodality of the CMD is that
691: galaxies remain blue and star-forming while
692: they contain cold gas, but at some point in their evolution they
693: rapidly lose their cold gas
694: (perhaps due to ram pressure stripping, strangulation,
695: or galaxy harassment when entering a dense cluster or group environment;
696: removal of gas through a galactic wind or active galactic nucleus;
697: or exhaustion of their
698: fuel in a merger-induced starburst),
699: and quickly transit to the red sequence as the young stars fade.
700: Our discovery that there are two distinct populations on the red
701: sequence, an intermediate-concentration population that contains
702: a disk and a high-concentration population that is entirely spheroidal,
703: suggests that there are
704: two distinct mechanisms for transiting to the red sequence.
705: We propose that Intermediate-type galaxies are those whose gas-removal process
706: did not significantly perturb the global morphology of the galaxy
707: (e.g.~strangulation), while Early-type galaxies are those
708: whose gas was removed in a violent event like a major merger.
709: Alternatively, perhaps there is only one method of transiting
710: to the red sequence, which results in Intermediate-type galaxies,
711: while Early-types are formed from violent ``dry'' mergers of
712: Intermediate and Early-type galaxies.
713: Regardless, the presence of these distinct populations provides a
714: key observational clue to the processes that drive galaxy evolution.
715:
716: It is important to note that our
717: samples consist of galaxies within the local universe,
718: i.e.~which have had a Hubble time for their structure to
719: regularize. Classification methods based on local galaxies might
720: not apply to younger high redshift galaxies.
721: Determining the epoch when these regular patterns emerged
722: would provide an important constraint in their origin.
723:
724:
725: \acknowledgements
726: JB thanks Joe Liske for help with the MGC database and the SDSS
727: help desk for their help with the Catalog Archive Server.
728:
729:
730: Funding for the Sloan Digital Sky Survey (SDSS) and SDSS-II has been provided by the Alfred P. Sloan Foundation, the Participating Institutions, the National Science Foundation, the U.S. Department of Energy, the National Aeronautics and Space Administration, the Japanese Monbukagakusho, and the Max Planck Society, and the Higher Education Funding Council for England. The SDSS Web site is http://www.sdss.org/.
731:
732: The SDSS is managed by the Astrophysical Research Consortium (ARC) for the Participating Institutions. The Participating Institutions are the American Museum of Natural History, Astrophysical Institute Potsdam, University of Basel, University of Cambridge, Case Western Reserve University, The University of Chicago, Drexel University, Fermilab, the Institute for Advanced Study, the Japan Participation Group, The Johns Hopkins University, the Joint Institute for Nuclear Astrophysics, the Kavli Institute for Particle Astrophysics and Cosmology, the Korean Scientist Group, the Chinese Academy of Sciences (LAMOST), Los Alamos National Laboratory, the Max-Planck-Institute for Astronomy (MPIA), the Max-Planck-Institute for Astrophysics (MPA), New Mexico State University, Ohio State University, University of Pittsburgh, University of Portsmouth, Princeton University, the United States Naval Observatory, and the University of Washington.
733:
734: The Millennium Galaxy Catalogue consists of imaging data from the
735: Isaac Newton Telescope and spectroscopic data from the Anglo
736: Australian Telescope, the ANU 2.3m, the ESO New Technology Telescope,
737: the Telescopio Nazionale Galileo and the Gemini North Telescope. The
738: survey has been supported through grants from the Particle Physics and
739: Astronomy Research Council (UK) and the Australian Research Council
740: (AUS). The data and data products are publicly available from
741: http://www.eso.org/$\sim$jliske/mgc/ or on request from J. Liske or
742: S.P. Driver.
743:
744:
745: {\it Facilities:} \facility{Sloan, ING:Newton, AAT, ATT, NTT, TNG, Gemini:Gillett}
746:
747:
748: \bibliography{/home/bailinj/papers/masterref.bib}
749:
750:
751: \end{document}
752:
753: