Publications of the Astronomical Society of the Pacific Vol. 109 1997

Publications of the Astronomical Society of the Pacific

Vol. 109 1997 July No. 737

Publications of the Astronomical Society of the Pacific 109: 745-758, 1997 July

Invited Review Paper

Low-Surface-Brightness Galaxies: Hidden Galaxies Revealed

Greg Bothun Department of Physics, University of Oregon, Eugene, Oregon 97403 Electronic mail: [email protected] Chris Impey Steward Observatory, University of Arizona, Tucson, Arizona 85721 Electronic mail: [email protected] Stacy McGaugh Department of Terrestrial Magnetism, Carnegie Institution of Washington, DC 20005 Electronic mail: [email protected] Received 1997 February 12; accepted 1997 April 29

ABSTRACT. In 20 years, low-surface-brightness (LSB) galaxies have evolved from being an idiosyncratic notion to being one of the major baryonic repositories in the Universe. The story of their discovery and the characterization of their properties is told here. Their recovery from the noise of the night-sky background is a strong testament to the severity of surface-brightness selection effects. LSB galaxies have a number of remarkable properties which distinguish them from the more familiar Hubble sequence of spirals. The two most important are (1) they evolve at a significantly slower rate and may well experience star formation outside of the molecular-cloud environment, (2) they are embedded in dark-matter halos which are of lower density and more extended than the halos around high-surface-brightness (HSB) disk galaxies. Compared to HSB disks, LSB disks are strongly dark-matter dominated at all radii and show a systematic increase in M/L with decreasing central surface brightness. In addition, the recognition that large numbers of LSB galaxies actually exist has changed the form of the galaxy luminosity function and has clearly increased the space density of galaxies at ζ = 0. Recent CCD surveys have uncovered a population of red LSB disks that may be related to the excess of faint blue galaxies detected at moderate redshifts. LSB galaxies offer us a new window into galaxy evolution and formation which is every bit as important as those processes which have produced easy-to-detect galaxies. Indeed, the apparent youth of some LSB galaxies suggest that galaxy formation is a greatly extended process. While the discovery of LSB galaxies has led to new insights, it remains unwise to presume that we now have a representative sample which encompasses all galaxy types and forms.

1. INTRODUCTION ' 'Are there diffuse nebulae that cannot be cataloged because Through the noisy haze of sky photons, astronomers they remain masked by the night sky?" For the case of since the time of Messier have detected and cataloged the galaxy detection, this question is quite relevant in the con- positions and shapes of diffuse, resolved objects known as text of the Cosmological Principle, a corollary of which nebulae. The cataloger well knows the limits on sensitivity asserts that all observers in the universe should construct posed by the observing environment, yet these limits are similar catalogs of galaxies. If this were not the case, then rarely quantified and passed on to the next generation of different observers might have biased views and informa- astronomers. Indeed, if Messier were alive in today's light- tion about (1) the nature of the general galaxy population in polluted world, his catalog would certainly be much sparser the Universe, (2) the three-dimensional distribution of gal- since he could only catalog the nebulae he is able to see. axies, and (3) the amount of baryonic matter that is con- Given this basic constraint, the natural question to ask is tained in galactic potentials. On the largest scales, we ex-

(e.g.. Fish 1964). Like any law, it was apparently made to be broken. This review, 20 years after Disney's original analysis, shows that his basic argument has been vindicated. Selection effects have been severe and as a result no representative sample of nearby galaxies has yet been compiled, cataloged, and investigated. □ Phülipps et al. (1987) Δ Davies (1990) fluí selected The most dramatic confirmation that these selection ef- O Davies (1990) diameter " ■ Schombert et al. (1Θ92) fects are real and significant has been provided by Mc- ▲ Sprayberry (1994) Gaugh et al. (1995) and is reproduced here in Fig. 1. Ten • de Jong (1996) years of hunting for galaxies of low surface brightness 25 24 23 22 (LSB) has revealed a surprising result which subverts the -2 μ0 {Β mag arcsec ) conventional wisdom as embodied by Freeman's Law. Fig- Fig. 1—The space density of galaxies as a function of central surface ure 1 shows that up to 50% of the general population of brightness. LSB objects appear to the left in this diagram. Raw counts galaxies resides in a continuous tail extending towards low from the indicated surveys have been converted to space density through the use of volumetric corrections discussed here and in more detail in μ-ο. Thus, the space density of LSBs is significant. This McGaugh et al. (1995). The solid line shows the surface-brightness distri- conclusion has also been reached by Dalcanton et al. (1997) bution which Freeman's Law suggests. The flat line fit to the data, from from a study of seven LSB galaxies detected in the Palomar McGaugh (1996), has a space density which is 6 orders of magnitude 5-m transit scan data. With measured redshifts they as- higher than predicted from Freeman's Law. signed a tentative space density of O.OSÍ^os^ioo -3 -2 Mpc for galaxies with /¿0 fainter than 23.5 mag arcsec , pect the Universe to exhibit a homogeneous appearance, but where h ^ = í/q/IOO. our only signposts for matter are the galaxies whose light As emphasized by McGaugh et al. (1995), the most we detect with optical telescopes against a noisy back- physically reasonable approach in converting raw counts to ground of finite brightness. Given this condition one can space densities is to assume that scale length and absolute easily conceive of observing environments that would make galaxy magnitude are uncorrelated. As shown explicitly be- galaxy detection difficult. low, this results in smaller volumes being accessible to For example, suppose we lived on a planet that was surveys for LSB galaxies compared to surveys for "nor- located in the inner regions of an elliptical galaxy. The high mal" or high-surface-brightness (HSB) galaxies. It is HSB stellar density would produce a night-sky background that galaxies that define the Hubble Sequence from which Free- would be relatively bright and therefore not conducive to man (1970) derived his sample. The space distribution of the discovery of galaxies. Similarly, if the Solar System in galaxies as a function of μ o after the volume sampling its journey around the galaxy were unlucky enough to be correction has been applied produces the distribution shown located near or in a Giant Molecular Cloud (GMC) at the in Fig. 1. The dark point defined by the Schombert et al. 5 same time that evolutionary processes produced telescopes (1992) survey has a space density which is 10 times higher on the Earth, then our observational horizon would be se- than the extrapolation of Freeman's Law would predict. 5 verely limited by the local dust associated with the GMC. Factors of 10 are significant. The space density derived by As it is, we are fortunate enough to be located at a rela- Dalcanton et al. (1997) is even higher than this, perhaps tively dust free area ~ 2.5 scale lengths from the center of suggesting that galaxies become smaller at lower surface a spiral galaxy. At this distance, the local surface brightness brightness. The opposite trend is seen in other data (i.e., of the projected galactic disk is ~ 24 mag arc sec"2 in the LSB galaxies tend if anything to be larger; de Jong 1996); blue. Thus in the direction of the galactic poles, the galactic this illustrates the enormous uncertainty that remains in our stellar density along the line of site is relatively low which knowledge of the local galaxy population. Nevertheless, the affords us a relatively dark window to peer out of in hopes implication of Fig. 1 is clear—very diffuse galaxies exist of discovering diffuse objects. Does this relatively unob- and they exist in large numbers. Their properties are only scured view guarantee that Earth-bound extragalactic as- now being elucidated. tronomers are able to detect a representative sample of This review deals primarily with the physical properties galaxies? of these newly discovered galaxies and their connection to The idea that the night-sky emission places limits on the galaxy evolution. For distance dependent quantities, we as- -1 -1 kinds of galaxies which can be detected was first com- sume Hq = 100 kms Mpc , and scale by /i100 mented on by Zwicky (1957). The first quantitative analysis = (7/0/100). A companion review (Impey and Bothun of the potential magnitude of this selection effect was pre- 1997) more fully details the selection effects that have pre- sented by Disney (1976). Disney's efforts were largely mo- viously prevented the discovery of LSBs, and how their tivated by the discovery of Freeman (1970) that spiral gal- actual discovery impacts the proper determination of the axies seemed to exhibit a constant central surface brightness galaxy luminosity function and its relation to QSO absorp- (/¿o in mag arcsec-2) in the blue. The formal value found tion fines as well as deep-galaxy surveys that have revealed by Freeman was /¿0 = 21.65±0.35 for a sample of a few an apparent excess of intermediate luminosity galaxies at dozen spirals. This constancy of became known as intermediate redshifts. The overall context of this review is "Freeman's Law" which strictly apphes to only disk gal- the idea that LSB disk galaxies represent a parallel track of axies although an analogous law for ellipticals also exists galaxy evolution that is largely decoupled from the pro-

© Astronomical Society of the Pacific · Provided by the NASA Astrophysics Data System HIDDEN GALAXIES 747 cesses that have determined the Hubble Sequence. While (1981) suggested that Freeman's Law was a conspiracy of this has significant implications for galaxy formation sce- dust obscuration, since if galaxies had appreciable optical narios and galaxy evolution the existence of LSB disks depth then we could only see their front surfaces. This themselves is not a surprise. The interesting issue is epoch could result in the appearance that disk surface brightnesses at which they begin to appear and proliferate. Eventually, were fairly constant. Bothun (1981) discovered from a thor- when astration is complete in disks, the Universe will con- ough survey of disk galaxies in the Pegasus I cluster that tain nothing but LSB objects. there were equal numbers of galaxies per magnitude bin with Β band central surface brightness values in the range 2. SURFACE-BRIGHTNESS SELECTION EFFECTS of 21-23.5 mag arcsec-2. The existence of any disk galaxy -2 2.1 How Surface Brightness is Measured with /¿o fainter than 23.0 mag arcsec would represent a 4σ deviation from Freeman's Law and hence should not In simple terms, surface brightness is very similar to have been found so easily. Unfortunately, Bothun (1981) surface air pressure. The amount of air molecules in a was not sufficiently astute to notice or appreciate the sig- three-dimensional column of air in the atmosphere deter- nificance of this at that time so the effect went unnoticed, mines the total amount of pressure which is exerted at a until recently (e.g., Fig. 1). point on the surface of the Earth. If we imagine a disk In principle, the surface-brightness profiles of galaxies galaxy as an optically thin cylinder, then the surface bright- can be traced to arbitrarily large radii. Recent data by ness is a measure of the space density of stars as projected Zaritsky et al. (1997) show that the dark matter halos of through a cylindrical cross section. The mean luminosity typical spiral galaxies may be extremely large (up to 200 density through this cylinder, which is determined both by h kpc in radius) and hence disk galaxies could extend out the stellar luminosity function and the mean separation be- to ^ 50 scale lengths! Integration of Eq. (1), as a function tween stars, is what observers measure as a projected sur- 3 of scale length, shows that \ai contains 26% of the total face brightness. Since the number of stars per Mpc has a luminosity, while 4a/ and 5«/ contain 90% and 96%, re- strong radial dependence, the projected surface-brightness spectively. In practical terms, a diameter defined by four profile shows a falloff with radius. This falloff is generally scale lengths provides a good measure of the total luminos- exponential in character and can be expressed as ity of the system. For a Freeman disk, this corresponds to an isophotal /¿(r) = /x0+1.086—, ai level of 26.05 mag arcsec-2. The darkest night skies that -1 can be found for terrestrial observing have ~ 23.0 mag where 1.086 = 2.5 log ^ . In this formulation two param- -2 eters completely characterize the light distribution: μ$ is the arcsec . Thus, the 90% luminosity isophote is some 3 central surface light intensity and α ι is the scale length of mags below even the darkest skies or 6% of the sky level. the exponential light fall off. In what follows will refer We have consistently defined a LSB disk as one which has μ0 -2 to the central surface brightness in the blue. If Freeman's /i0 fainter than 23.0 mag arcsec or a 90% luminosity -2 law is correct, the number of parameters relevant to galaxy isophote of 27.4 mag arcsec which is 2% of the night selection reduces to one as variations in size modulate those sky brightness. The Freeman value for /¿0 is about 1 mag in luminosity. Since μ0 is a measure of the characteristic brighter than the surface brightness of the darkest night sky. surface mass density of a disk. Freeman's Law requires that That the number of galaxies with faint central surface all the physical processes of disk galaxy formation and brightnesses appears to decline rapidly as -+ μ^ is evolution conspire to result in this very specific value for all suspicious and if true of the real galaxy population implies galaxies. Either the surface mass density must be the same that our observational viewpoint is privileged in that we are for all galaxies (in itself a peculiar result) with little varia- capable of detecting most of the galaxies that exist, at least tion in the mass to light ratio, or variations in the star- when the Moon is down. This is the essence of the argu- formation history, collapse epoch, and initial angular mo- ment voiced by Disney (1976) in characterizing the Free- mentum content must all conspire to balance at this man Law as a selection effect (see also Davies 1990). arbitrary value. So if surface-brightness selection effects are important, The historical importance of Freeman's Law is that it what is the best way to measure surface brightness? Con- was the first real attempt at quantifying the surface- sider the following two hypothetical galaxies which have brightness distribution of spiral disks. As such, Freeman the same total luminosity {MB = —20): (1970) has been cited over 700 times indicating that it has Galaxy A: /x0 = 21.0 az = 3 kpc. had an enormous impact on the field, particularly on studies of the galaxy-luminosity function, which, until recently has Galaxy Β: μ,0 = 24.0 ai = 19 kpc. assumed to be independent of surface brightness. As a con- There are three conceivable ways of measuring the surface sequence of its popularity and widespread influence in ex- brightness in these disks: (1) central surface brightness tragalactic astronomy there were several attempts to show (μ,ο), (2) average surface brightness within a standard iso- -2 that Freeman's Law was not correct. While Disney (1976) phote ( = 25.0 mag arcsec , /^iS0), or (3) effective surface dismissed Freeman's Law as an artifact of selection, others brightness (measured within the half light radius = 1.7 α ι were not so sure. For instance, Kormendy (1977) asserted = μ & instead of measured at reff). Note that methods 2 that Freeman's Law could be an artifact of improper sub- and 3 do not require that the light profile is well fit by an traction of the bulge component in disk galaxies. Boroson exponential function. The results of applying each of these

© Astronomical Society of the Pacific · Provided by the NASA Astrophysics Data System 748 BOTHUN ET AL. three measures of surface brightness are the following: lar total luminosity (MB ^ —21.1); the fifth galaxy has the same scale length as Galaxy Β but a factor of ten lower Galaxy Α: μ,0 = 21.0; /xiso = 23.2; /¿eff = 22.3. total luminosity. This adheres to the McGaugh et al. (1995) Galaxy B: /x = 24.0; ¿¿ = 25.8; /¿ = 24.8. 0 iso eff assumption that scale length and total luminosity are uncor- The largest difference in surface brightness occurs when related in a representative sample of disk galaxies. μ0 is used as the measure. Hence we have adopted this to -2 define the disk galaxy surface brightness. While we realize Galaxy A: cti = 0.5 kpc, μ0 = 16.0 mag arcsec , -2 that this definition requires that the galaxy be adequately fit Galaxy Β: αι = 5.0 kpc, μ0 = 21.0 mag arcsec , by an exponential profile, most LSB galaxies at all lumi- Galaxy C: = 25.0 kpc, = 24.5 mag arcsec-2. nosities meet this criterion (see McGaugh and Bothun 1994; α ι μ0 -2 O'Neil et al. 1997a; Sprayberry et al. 1995a). This exercise Galaxy D: α ι = 50.0 kpc, μ0 = 26.0 mag arcsec , also makes the trivial point that galaxies with low μ.0 re- -2 Galaxy Ε: α ι = 5.0 kpc, μ0 = 23.5 mag arcsec . quire surface photometry out to very faint isophotal radii in order to determine a total luminosity. Assume that the intrinsic space density of these five galaxies are equal, and suppose that we conduct a survey to 2.2 A Censored View of the Galaxy Population catalog galaxies which have diameters measured at the μ0 = 25.0 mag arcsec-2 level (D25) of greater than 1 arcmin. Much of our knowledge of galaxies has stemmed from Under these conditions, we are interested in determining the detailed studies of objects that populate the Hubble se- maximum distance that each galaxy can be detected: quence. Over 70 years ago, Hubble (1922) warned of rely- Galaxy A: This galaxy is quite compact (ratio of 1/2 ing too much on this venture: light diameter to D25 = 0.33) and would fall below the ''Subdivision of non-galactic nebulae is a much more catalog limit beyond a distance of 60 Mpc. difficult problem. At present and for many years to Galaxy B: This is a typical large spiral (like M31); come, their classification must rest solely upon the simple inspection of photographic images, and will be D25 corresponds to 3.74 a/ which is 18.7 kpc or a diameter confused, by the use of telescopes of widely differing of 37 kpc. This projects to an angular size of 1 arcmin. at scales and resolving powers. Whatever selection of types a distance of 125 Mpc. is made, longer exposures and higher resolving powers Galaxy C: D25 corresponds to 0.45 a/ or 11.5 kpc. This will surely cause a reclassification of many individual projects to an angular diameter of 1 arcmin at a distance of nebulae ..." 76 Mpc. Galaxy D: £>25 not exisi: galaxy would In this quote Hubble estabhshes that galaxy classifica- never be discovered in such a survey. tion, and therefore implicitly galaxy detection, is highly Galaxy E: D25 corresponds to 0.92 or 4.6 kpc. This dependent upon observing equipment and resolution. The projects to an angular diameter of 1 arcmin. at a distance of essential issues are: (1) how severe is the bias in terms of 30 Mpc. the potential component of the galaxy population that has The total survey volume is defined by Galaxy B, as they been missed to date, and (2) how would this affect our current understanding of galaxy formation and evolution? In can be seen to the largest distance. The ratio of sampled hindsight it is somewhat mysterious why this issue of gal- volumes for each of the other Galaxy types is considerably axy detection was not considered more seriously 25 years smaller. For instance, the volume ratio of Galaxy Β to ago. Tinsley's (e.g., Tinsley 1975) elegant and accurate Galaxy Ε is a factor of 70! Hence, a survey like this would modeling of the stellar populations of galaxies in the late take the real space density distribution (which is equal) and, 60's and early 70's certainly indicated that galaxies could through the survey selection effect, produce a catalog which undergo significant luminosity evolution, thereby producing would contain 72% type Β galaxies, 18% type C galaxies, faded and diffuse galaxies ai ζ = 0. Alternatively, there 9% type A galaxies, and 1% type Ε galaxies. Type D might be a population of intrinsically low surface mass galaxies would not be represented at all. This is a severe density systems whose evolution is quite different from bias which would lead us to erroneously conclude that there "normal" galaxies, but which nevertheless are important is predominantly one type of disk galaxy in the Universe, repositories of baryonic matter. Disney was the most reso- and that this naturally leads to Freeman's Law. Presently, in nant voice to suggest that such diffuse systems could exist, the real zoology of LSB galaxies, type C and D galaxies are and therefore that we could be missing an important con- quite rare but type Ε galaxies are common. Hence, their stituent of the general galaxy population. detection locally automatically means the space density is A very simple way of describing the effect of surface brightness selection is offered below. While Disney and relatively large because (1) they are so heavily selected Phillips (1983) and McGaugh et al. (1995) have quantified against and (2) the volume correction factors which are these effects to produce Fig. 1, the essential point is that significant. This was the essence of Disney's original argu- LSB disks, at any luminosity/circular velocity, are detect- ment but there was no real data to support it at that time. able out to a significantly smaller distance than HSB disks. Twenty years of progress enables us to plot Fig. 1, which Consider the five hypothetical galaxies listed below. The has the remarkable property that there is no indication of a first four have pure exponential light distributions and simi- rapid fall of in μ0.

3. SEARCH AND DISCOVERY artifacts of the processing, water spots, or specks of dust. Even the detection of 21 cm emission from one of these The story of the discovery and characterization of LSB "plate flaws" did not assuage one particularly recalcitrant galaxies as important members of the general galaxy popu- referee. With youth, low pay, and foolishness on our side, lation began in 1963 with the publication of the David we persisted in our efforts to verify the reality of the Dunlap Observatory (DDO) Catalog of Galaxies by Sidney smudge galaxies. van den Bergh (van den Bergh 1959). This catalog consists In late 1985 and early 1986 we used the Las Campanas of galaxies which exhibit a diffuse appearance and that have angular sizes larger than 3 arcmin. While the DDO objects 100 in. telescope for CCD imaging the very diffuse galaxies are the first bona fide collection of a sample of LSB galax- found in the Malinization process (all of which turned out ies, they are not at all representative of the phenomenon. to be real). Most of these galaxies were devoid of structure. The galaxies contained in the DDO catalog are exclusively However, one had what appeared to be very faint spiral of low mass (some are members of the Local Group). This structure which was connected to a point-like nuclear re- has fostered an erroneous perception that all LSB galaxies gion. On the Palomar Sky Survey, this nuclear region is are dwarf galaxies. Today, we know that all masses of unresolved with no associated nebulosity apparent. This was galaxies have representation in the LSB class. one of the few Malin objects bright enough for optical The first substantial contribution to our understanding of spectroscopy, and on 1986 May at the Palomar 200-in LSB disk galaxies was made by William Romanishin and telescope, Jeremy Mould and Bothun took a spectrum of its his collaborators Steve and Karen Strom in 1983 (see Ro- nucleus. Astonishingly, the spectrum exhibited emission lines at a redshift oí ζ = 0.083, or a recessional velocity of manishin et al. 1983). They derived their sample from the -1 Uppsala General Catalog of Galaxies (UGC—Nilsson about 25,000 kms . Now, we had pursued the Maliniza- 1973). The UGC is a diameter-selected catalog and is there- tion process on UK Schmidt plates of the Virgo cluster in order to find extremely LSB galaxies in the cluster. Virgo fore less sensitive to surface brightness selection effects -1 compared to galaxy selection based on apparent flux (see has a mean recessional velocity of 1150 kms , so this also McGaugh et al. 1995). As a result, the UGC does nucleated object clearly was far beyond Virgo. Since the total angular size of the object on our CCD frame was contain some LSB disk galaxies but most have μ0 in the range 22-23.0 mag arcsec-2. In early 1984, Allan Sandage approximately 2.5 arcmin, quick scaling then indicated that and his collaborators published some of the first results of if a galaxy like this were indeed in Virgo then its angular the Las Campanas Photographic Survey of the Virgo Clus- size would be a degree. If it were as close as the Androm- ter. Contained in those papers were some dramatic ex- eda Galaxy its angular size would be about 20 degrees and amples of dwarf galaxies in the Virgo cluster which were of course we would look right through it without noticing it. quite diffuse. The existence of such diffuse objects in a This seemed absurd, and there was a good chance that this cluster was very interesting as this environment should strange galaxy was a composite system, consisting of a prove hostile to their formation and survival. In addition, background emission-line galaxy shining through a fore- Ellis et al. (1984) discovered additional examples of ex- ground dwarf. tremely diffuse galaxies in the field. This immediately In 1986 October, 21 cm observations at Arecibo revealed raised the possibility that perhaps galaxies like this were the characteristic signature of a rotating disk galaxy whose common but, for the most part, not yet detected and cata- systemic velocity was equal to that of the emission-line loged. If so, perhaps these faint diffuse galaxies were the object. The accidental discovery of Malin 1 (Bothun et al. source of the enigmatic QSO absorption line systems. These 1987) strongly confirmed Disney's original speculation of considerations forged the Impey/Bothun collaboration as we the existence of "crouching giants." The existence of Ma- collectively wondered if the Sandage survey had missed hn 1 (type D in the previous example) certainly implied a galaxies of even lower surface brightness. To answer this non-negligible space density of these kinds of objects. The question we enlisted the help of David Malin in Australia. properties of Malin 1 are described in detail by Impey and Ultimately, we were trying to improve the determination of Bothun (1989). Recent H I observations of Malin 1 using the galaxy luminosity function by concentrating on those the VLA by Pickering et al. (1997) confirm the presence of galaxies which would be the most difficult to detect, due to a greatly extended gaseous disk around the normal bulge extreme diffuseness. We had no idea if such extreme LSB component of the galaxy. galaxies really existed; to say we knew what we were doing The remaining "smudges" in the Virgo cluster area did would really exploit the advantage of hindsight! not turn out to be as spectacular as Malin 1. These smudges Malin's method of photographic amplification had been were most likely LSB dwarf-galaxy members of the Virgo used to find low surface brightness shells and other tidal cluster. Their discovery and characterization by Impey et al. debris around normal galaxies, and it could be extended to (1988) readily showed that the faint-end slope of the LE in find entire very LSB galaxies. Indeed, Malin already had clusters of galaxies was significantly steeper than previously anecdotal evidence that whenever he Malinized a plate, he measured. In the case of Virgo, this meant there were gal- would find these "faint little buggers" everywhere. So axies in the range MB = —12 to —16 which were below David agreed to Malinize selected one-square-degree areas the isophotal limits of the plate material used by Sandage. of the Virgo cluster from which several small diffuse ob- Most of these have μ0 fainter than 24.5 mag jects emerged. Skeptical colleagues insisted that the peculiar arcsec-2 and α ι larger than 1 kpc. The presence of these collection of faint smudges that could be seen were all diffuse galaxies at modest luminosities increased the faint-

galaxies for detailed follow-up studies are incomplete, in- adequate, and biased. These surveys are now complete and most of the results have been published. They have opened up a new field of inquiry in extragalactic astronomy. De- tailed studies of the properties of individual LSB galaxies, and the class as a whole, has resulted in a number of recent Ph.D. theses: Knezek (1993); McGaugh (1992); Sprayberry (1994); Dalcanton (1995); de Jong (1995); Driver (1995); de Blok (1997); O'Neil (1997); and Pickering (1997). In just over a decade, a whole new population of galaxies has been discovered. Figure 1 provides only a hint on their overall space density. The number of objects with /¿o ^ 24.0 mag arcsec-2 is unknown, and can only be guessed by extrapolation of the trend in Fig. 1. Significant numbers of galaxies with ^ 24.0 mag arcsec-2 have been detected in CCD surveys, suggesting that the trend remains fairly constant (Dalcanton 1995, O'Neil et al. 1997a) or even rises towards fainter μ o (Schwartzenberg et al. 1995). Figure 2 shows an example of one of the more extreme LSB galaxies turned up in the O'Neil et al. survey. Reproducing nearly invisible galaxies on paper is difficult. Interested readers should inspect the digital gallery of LSB Fig. 2—An example from the Texas survey (e.g., O'Neil et al. 1997a) of an extremely diffuse object. The image size is 1.2X1.5 arcmin and the galaxy galaxies available at http://zebu.uoregon.edu/sb2.html. looks to be at least 1 arcmin in "diameter" at a probable redshift of These LSB galaxies are of cosmological significance and 4-6000 km s. A better presentation of these diffuse galaxies can be found have properties which are quite different from those of their at http://zebu.uoregon.edu/sb2.html HSB counterparts which dominate existing galaxy catalogs. end power law slope of the LF to a value of —1.55; sig- 4. FORM, PROPERTIES, AND STELLAR CONTENT nificantly steeper than the value of —1.1 which was thought to hold for clusters and perhaps the field (see Efthasthiou et In the Hubble sequence of spirals, morphological classi- al. 1988; Loveday et al. 1992; Marzke et al. 1994). fication is based on spiral arm texture and definition (see The success of the Malin hunt for very diffuse galaxies Sandage 1961). In this case, classification becomes directly prompted three new surveys. The first relied on the good- linked to the relative star formation rate (SFR) via the will of Jim Schombert, who was a Caltech Postdoc associ- illumination of the spiral pattern by young stars. In an ated with the second Palomar Sky Survey. Jim was the attempt to separate out this effect from the structural portion quality control person and thus had direct access to the of morphological classification, van den Bergh (1976) pro- plates themselves for a limited period of time before they posed a 2D classification system involving spiral arm defi- were secured in the vault. This allowed an opportunity for nition and bulge-to-disk ratio. In that system, he introduced hit-and-run visual inspection in predefined declination strips a class of disk galaxies known as anemic galaxies. These to search for diffuse galaxies with sizes larger than 1 arc- are systems with weak spiral structure but they are gener- min. This produced the catalogs of Schombert and Bothun ally not of low surface brightness (see also Bothun and (1988) and Schombert et al. (1992). A second survey was Sullivan 1980). initiated in the Fornax cluster using the Malinization tech- It is well established that in Hubble sequence spirals, the nique in order to compare the results to Virgo. These ob- bulk of the star formation occurs within GMCs. This jects are cataloged and described in Bothun et al. (1991) method of star formation gives rise to the formation of which built on the earlier work of Caldwell and Bothun massive stars within stellar clusters, which ultimately drives (1987). The third survey was initiated with Mike Irwin at the chemical evolution of galaxies. Galactic disks engage in Cambridge and this involved using the Automatic Plate star-formation activity according to a criterion first estab- Machine (APM) to scan UK Schmidt plates using an algo- lished by Quirk (1972). The more recent expression of this rithm optimized to find galaxies of low contrast. This forms criterion, by Kennicutt (1989), reflects a convolution of the the most extensive catalog of LSB galaxies to date (Impey rotation curve with the surface density distribution of gas. et al. 1996). The result is a critical surface density of Η I, as a function The goals of these new surveys was to discover, using of radius, for the formation of molecular clouds and subse- various techniques, the extent and nature of this new popu- quence massive star formation. All data to date (e.g., van lation of galaxies which had low contrast with respect to the der Hülst et al. 1993; de Blok et al. 1996; Pickering et al. sky background and hence have remained undetected and 1997) strongly suggests that LSB disks have an Η I distri- uncataloged. The importance of discovering this new popu- bution which is largely below this threshold. It is therefore lation cannot be overstated. The existence of LSB galaxies likely that LSB disks represent the evolutionary track taken is a clear signal that the samples from which we select by flattened gas-rich systems that are simply unable to form

HI line widths -i—ι ι ι Μ 11 -ι—ι ι ι ι ι ι \ O \ C\2

C\2 CV2 :: ••mß * ·· o •. 4·.· V

C\2 ''.Λ \ ·· ^ • \ £ Ζ CO C\2 \ \ ·

10 100 600 h (kpc)

Wsq (km s ) Fig. 4—Plot of μ0 vs al for samples of LSB and HSB galaxies. The Fig. 3—Fractional distribution of 21 cm velocity widths for samples of absence of HSB galaxies with large values of or/ is real; other than that, the LSB and HSB galaxies. The data are 50% velocity widths. The LSB two variables are uncorrelated. The LSB data come from McGaugh and sample represents 131 galaxies which exhibit double-homed profiles from Bothun (1994) and Sprayberry (1994) and the HSB data come from de the Schombert et al. (1992) catalog and the HSB represents 1500 galaxies Jong (1995). with double-homed profiles whose data have been compiled in the litera- ture. widths which is indistinguishable from any sample of Hubble sequence spirals, as shown in Fig. 3. LSB disks stars in the familiar environment of molecular clouds. If have similar sizes and masses as those high-surface correct, this leads to the following set of expectations: brightness (HSB) that define the Hubble sequence. However, despite this evidence to the contrary, there is a (1) The arm-interarm contrast in LSB disks should be quite widespread notion that LSB galaxies are exclusively low due to a paucity of zero-stars and Η Π regions that low-mass (or "dwarf") galaxies. Indeed, a small percentage generally define spiral arm structure of the LSB population has truly impressive overall sizes (2) The overall star formation efficiency should be reduced with scale lengths that exceed 15 kpc. Figure 4 shows and hence LSB s should evolve at a slower rate than that when one closes in on a representative sample of those disks that form the Hubble sequence. galaxies, the entire disk structural plane defined by α ι and (3) Massive star formation and hence metal production /¿o is populated. should be inhibited in LSB disks. (2) Very few LSB galaxies show evidence for nuclear (4) LSB disks should be relatively low in dust content. activity. This is in marked contrast to HSB disk galaxies (5) If the characteristic spacing between newly formed stars where the percentage which have active galactic nuclei can depends upon the local environment, then star forma- be as high as 50%, depending on the mean luminosity of tion may be occurring in a more diffuse atomic H I the sample. The most probable explanation for this is that medium. This would produce a lower luminosity den- LSB disks generally lack two structural features that sity and a lower than average surface brightness disk. facilitate gas flows and/or the formation of a compact object (6) No stars at all should form in LSB disks, if molecular in the nucleus—bulges and bars. In this context it is clouds are truly required to initiate this process in all interesting to note that about 50% of the large-scale length galactic environments. LSB disk galaxies (e.g., Malin 1 and cousins) show some The observational properties that have been established nuclear activity, and all have a normal bulge component for samples of LSB disk galaxies are largely consistent with with luminosity MB = —18 to —20. These galaxies are these expectations. The most salient features are enumerated truly enigmatic in that "normal" formation processes were below: at work to create the bulge component but no conspicuous (1) LSB disk galaxies are not exclusively of low mass. stellar disk ever formed around this bulge. This is clearly seen in the distribution of 21 cm velocity (3) The amount of neutral hydrogen in LSB disks is very

CV2o • · 02 CV2 * ..· .* * *·· : *· /·· ν

C\2

0.4 0.6 0.8 Radius (kpc) B-V Fig. 5—An example of a LSB disk galaxy rotation curve and surface Η ι Fig. 6—Non-correlation between the integrated disk color and observed distribution plotted against the critical surface density criterion of Kenni- /¿o for the LSB sample of McGaugh and Bothun (1994). There is clearly cutt (1989). The galaxy depicted here is UGC 6614 where the data come no trend of increasing B—V with decreasing as would be expected for from Pickering et al. (1997). The surface density of Η I is everywhere below the critical density. At a radius of 30 kpc it rises to meet the critical a fading scenario. If anything, the mean B — V color gets slightly bluer density and a small amount of star formation, as evidenced by weak Η π with decreasing . regions, is observed there (see data in McGaugh 1994). Most all LSB galaxies have measured surface densities of Hi which are below the paths of chemical evolution exist for disk galaxies, which in critical density (see also van der Hülst et al. 1993; de Blok et al. 1996). turn demands multiple star formation histories. The LSB disks represent some of the most une volved objects in the universe. similar to HSB spirals except for their lower than average (4) Practically all of the LSB galaxies discovered to date surface densities. A typical case is shown in Fig. 5 where it in photographically based surveys (described previously) can be seen that the gas density is generally below the are blue despite the obvious lack of star formation. Mean threshold for molecular cloud formation and hence colors for these disks range from U—B= —0.17±0.05, widespread disk star formation and metal production. LSB B-V = 0.49±0.04, and V-I = 0.89±0.04 (McGaugh spirals are significantly deficient in molecular gas compared and Bothun 1994) to B-V = 0.73 ±0.05 and to HSB spirals of the same mass. Schombert et al. (1990) V—R = 0.50±0.04 for the sample of very large scale failed to detect a single LSB disk in CO (see also Knezek length LSBs of Sprayberry et al. (1995a). In the most ex- 1993). The mean upper limit on is an order of treme cases, these galaxies, despite very low star-formation magnitude less than the MH/ = expectation. In some rates ( ~ 0.1 Mq per year), have disk colors of cases it is nearly 2 orders of magnitude less. A recent B — V — 0.1 and V—I ~ 0.6. By way of comparison with calibration of the H2/CO conversion factor as a function a large sample of galaxies which obey the Freeman Law, de of metallicity (Wilson 1996), when applied to LSB s Jong and van der Kruit (1994) found mean colors of changes these limits by only a factor of 3. Furthermore, B-V = 0.75±0.03 and V-R = 0.53±0.04. One of the virtually none of the LSB galaxies detected by Schombert more curious aspects of the color distributions of LSB et al. (1992) or Impey et al. (1996) are IRAS sources. samples is shown in Fig. 6 which indicates that there is no Although a typical LSB disk contains a handful of bright correlation between μ0 and overall color. This means that Η Π regions, indicating some current star-formation, the the LSB disks cannot be the faded remnants of HSB gal- average star formation rate is at least an order of magnitude axies after star formation has subsided. lower than in similar mass disk galaxies. In our galaxy, star Figure 6 is very difficult to interpret. Galaxies must fade formation is observed to only occur in GMCs (e.g.. Young and redden as they age, but we see no evidence of this in and Scoville 1991). LSB galaxies may well be devoid of our photographically selected sample of LSBs. McGaugh molecular clouds in their interstellar medium but neverthe- (1996) argues that this is a selection effect which, depend- less formed some stars. We might speculate that the ing on the plate material, might be sufficiently severe to pressure-temperature-density manifold in LSB disks physi- prohibit the detection of LSB disks with B-V = 1.0. cally precludes molecular cloud formation. The expectation However, we expect that this selection effect is not that of low metaUicity is confirmed by the observations of Mc- severe as anything with B — V ^ \ and B{0) in the range Gaugh (1992). As derived from the Η π region spectros- 23.0-24.0 mag arcsec-2 should have been detected in copy, the abundance of heavy elements such as oxygen, these surveys. Since we expect that such objects should nitrogen, neon, and sulfur can be an order of magnitude exist, this suggests that the intrinsic local space density of lower than HSB disks of the same mass. On average, LSB LSB galaxies is even larger than suggested by Fig. 1. Per- disk galaxies with L* luminosities have abundances of haps the apparent absence of red LSBs can be reconciled ~ 1/3 solar (McGaugh 1994). The largest LSBs known with a stellar population model and fading scenario which (e.g.. Mahn 2, UGC 6614) may have abundances near solar does not lead to significant reddening. One way to offset the or in excess of solar—but these are rare objects. Thus, LSB tendency for stellar populations to redden as they age is to galaxies strongly violate the mass-metallicity relation de- introduce an extremely blue horizontal branch which is fed fined by Hubble Sequence spirals. This shows that multiple by the evolving stars on the giant and asymptotic giant

© Astronomical Society of the Pacific · Provided by the NASA Astrophysics Data System HIDDEN GALAXIES 753 branch. Such populations have been seen in the bulges of spirals and in ellipticals (e.g., O'Connell et al. 1992; Fer- guson et al. 1991; Brown et al. 1996). A minority of other hot components such as Post-Asymptotic-Giant-Branch (PAGB) stars or Planetary Nebulae (PN) nuclei would also aid in keeping these galaxies bluer as they fade. In fact, we can successfully reproduce the observed colors of LSB disk systems by mixing together three populations of stars: (1) F2 V stars with £-7 = 0.36, 7-/ = 0.55 and Mv = 3.3. (2) KO ΠΙ stars with B-V = 1.00; V-I = 1.30 and My = 0.0. (3) BHB stars (as in M4) with B-V = -0.2; V-I = b-v (total) -0.3 and My = 1.0. Fig. 7—Plot of observed and ß — V for the Texas CCD based sample of O'Neil et al. (1997a). While there is still no correlation between and The colors of LSB disks can be reproduced when B — V, the percentage of objects with B — V 5¾ 0.8 is significantly higher in this sample compared to samples of LSB galaxies which are photo- ^ 50% of the stars are F2 and the ratio of hot evolved graphically selected. Consistent, with those previous samples, however, stars to giants is ^ 3:1. Unfortunately, stellar evolution very blue and very low galaxies are also present in this sample. does not predict such a large ratio of hot evolved stars to giants since the giant branch feeds the BHB stage where the lifetime is somewhat shorter (see Bruzual and Chariot blue objects. The first wide field CCD survey specifically 1993). Hence, this ratio should be unity at most. A ratio of designed to detect LSB galaxies has been initiated by unity, however, produces a too red V—l color. Theoreti- O'Neil et al. (1997a,b). This survey was performed with the cally, it is also possible that if the hydrogen envelope of University of Texas McDonald Observatory 0.8-m telescope giant stars is completely lost during their evolution then the between 1993 and 1996 using an LF1 2048X2048 pixel remnant core will fall on the helium main sequence which CCD camera with a Γ.'32 pixel size. The sensitivity of this is considerably hotter and more luminous, at a given mass, survey is about 1 mag arcsec-2 fainter than previous pho- than the normal main sequence (e.g., a 0.5 Mq helium tographically based surveys. Fields in the Pegasus and Can- main-sequence star has reff ^ 45,000 K). This would result cer clusters (these are Virgo-like structures at redshifts in a substantial UV/blue contribution from evolved stars. 4000-5000 km s_1 which are spiral rich) were selected The above scenario is admittedly quite extreme, and it along with many fields in the low-density environment de- cannot be tested without UV observations of LSB galaxies. fined by the Great Wall. The survey yielded the discovery A large program aimed at determining fluxes of LSB disks of 120 new LSB galaxies including, finally, a red compo- at 2000 Â was approved as a Guest Investigator Program nent to the LSB population. Figure 7 shows the color dis- for the UIT Telescope aboard Astro-2. Unfortunately, there tributions in B — V for this sample. Note again there is no was a complete failure in the first stage image intensifier of correlation between color and μ0. Even though this new the A camera during the Astro-2 mission and none of the survey has detected LSB galaxies with fairly red integrated required data could be taken. Hence, we still do not know if colors, there is no sign of a fading sequence. For this par- optically LSB disk galaxies, in fact, have normal surface ticular sample the average colors are B — V = 0.79 ±0.04, brightnesses in the UV. If they do, there are important U — B = 0.08±0.04, V—I = 0.94±0.06. Approximately implications regarding deep galaxy surveys (see below). In 20% of the sample has B—V ^ 0.9 and U—B ^ 0.4, the fact, a study of a deep Astro-I image of Fornax (see O'Neil colors that might be expected for a faded, relatively metal- et al. 1996) did show that some optically LSB objects could rich Hubble sequence spiral. be detected at 1500 and 2000 Â. (5) Until very recently, almost nothing was known about If LSB galaxies do not have significant UV fluxes the dynamics of rotating LSB disks. Optically, LSB disk (which also contribute to the U and Β passbands) then their galaxies rarely show the symmetry that is usually exhibited blue colors are very difficult to understand. In most cases, by Hubble Sequence spirals (but see Zaritsky and Rix the only physically plausible explanation is a reduced con- 1997). To first order, strongly symmetric optical disks are a tribution to the integrated light by giants. This requires signature of rotation dominated kinematics within relatively these galaxies to have mean ages several Gyr younger than cold disks. The "chaotic" optical appearance of many LSB HSB disks of the same luminosity. This notion of delayed disks might suggest peculiar kinematics, yet the global 21 formation is somewhat consistent with the observed low cm profiles generally show the two-homed signature of a densities of these objects as the dynamical timescale (a rotating disk. Some LSB disks are sufficiently gas rich that characteristic timescale over which an object collapses) can their rotation curves can be derived from aperture-synthesis be an order of magnitude larger than HSB galaxies. Thus data. Two groups (de Blok et al. 1996; Pickering et al. we might expect most LSB galaxies to have late collapse 1997) have used the VLA and WSRT radio arrays to de- times and hence delayed formation of their first stars. termine the 2D distribution of Η I in about a dozen LSB A recent development in the survey and detection of galaxies. The results of these efforts can be summarized as LSB galaxies can counter the previous tendency to find only follows:

_1 -1 (1) LSB galaxies have rotation curves that are shallower at models requires either H0 =¾ 30 km s Mpc , or that small radii compared to HSB galaxies of the same ro- total cluster masses have been systematically underesti- tation velocity and mass. mated. The latter possibility does not appear to be the case (2) The gaseous component of LSB s is dynamically signifi- (Evrard et al. 1996) and hence the measured values of fb in cant at virtually all radii. clusters appears quite inconsistent with fl = 1. In contrast, (3) LSB disks are dark matter dominated at virtually all fb as measured in LSB disks is significantly lower than the radii. Compared to HSB galaxies of the same rotational cluster value and is consistent with Ω = 1. So which is the velocity, LSB disks have globally and locally higher more representative baryonic repository—clusters of galax- values of MIL (see also de Blok and McGaugh 1996). ies or LSB disks? Arguments given in Impey and Bothun (4) In contrast to HSB galaxies, no 4'maximum disk" mass (1997) show that LSB galaxies make a significant contribu- model fits the rotation curve. This means that there is tion to the total baryonic mass density in the Universe, no region in an LSB disk where the luminous (bary- providing up to half of it depending on the faint end slope onic) matter dominates the potential and hence deter- of the luminosity function (see below). mines the form of the rotation curve. (5) Mass models derived from the rotation curves of LSB 5. COSMOLOGICAL RELEVANCE OF LSB and HSB galaxies show that LSB galaxies inhabit less dense and more extended dark matter halos. However, GALAXIES they have dynamical masses comparable to those of The discovery of a substantial population of LSB galax- HSB galaxies. ies at ζ = 0 is relevant to a number of cosmological issues. (6) The most extreme examples give very hard upper limits These issues are enumerated below; those issues that are to the ratio of disk to halo mass. These result in baryon only briefly discussed here are covered in more detail in fractions of ^ 10% at the most, and more likely Impey and Bothun (1997). ^ 3%. QSO absorption lines: Increasing the number density of galaxies is a very good first step towards understanding the This dynamical information can be interpreted in terms origin of QSO absorption lines. In addition, the discovery of of the profound implication that the density profile of the Malin 1, which has 105 kpc2 of gaseous cross section at a dark matter halo ultimately determines the surface density level of N ^ 3X 1019 cm-2 is a helpful addition to the of the galaxy which forms in that potential. Thus, LSB h zoology of objects that might produce damped Lya sys- galaxies may well have fundamentally different dark matter tems. distributions from HSB galaxies. This makes LSB disks physically distinct from HSB disks, even though they may Large-Scale Structure (LSS): To date, most studies of LSS are based on redshift surveys of optically selected have similar global properties. The paradoxical observation samples, which, by definition, do not contain many LSB that LSB disks he on the same Tully-Fisher relation as HSB disks (see Sprayberry et al. 1995b), albeit with sig- galaxies. If LSB galaxies are better tracers of the mass distribution, then the linear biasing factor between mass and nificantly more scatter, can be understood if the ratio of light has a surface brightness dependence. This would be an dark-to-light matter in disk galaxies is independent of the ugly complication to the various sophisticated attempts to dark matter halo density. In this case, galaxies of the similar 0 6 circular velocity but dissimilar surface brightnesses can determine ¿7Í1 · , where b is the (scale-independent) linear bias parameter (see Strauss and Willick 1995). In an at- have the same luminosity because the LSB galaxy is de- tempt to shed light on this, Bothun et al. (1985, 1986) fined by a larger radius which contains the mass that deter- performed a redshift survey of =« 400 LSB disk galaxies, mines the circular velocity. This implies that LSB disks also primarily selected from the UGC. Few of these galaxies have lower surface mass density. These recent observations of LSB s also give rise to an were in the CFA redshift survey at that time. That data, important new cosmological inconsistency. The rotation coupled with a thorough analysis of the clustering proper- curves establish that LSB disks are everywhere dark matter ties of LSB galaxies by Bothun et al. (1993) and Mo et al. (1994) provide the following insights: dominated and reveal a baryonic mass fraction {fb) of ~ 5% or less. Another environment which is everywhere (1) On scales ^ 5Mpc LSB galaxies trace the iden- dark matter dominated is clusters of galaxies. White tical structure as HSB galaxies. et al. (1993) have argued that the ratio CíbICí0 measured (2) LSB galaxies generally avoid virialized regions. for a cluster should not be significantly different than the (3) On scales =¾ 2/1^ Mpc LSB galaxies are significantly universal value. If this is the case, then Ω0 can be deter- less clustered. In fact, there is a significant deficit of mined from the relation: companions within a projected distance of 0.5/^ Mpc indicating that LSB disks are generally isolated. In general, the correlation function has a lower amplitude for LSB s (see Mo et al. 1994) resulting in them being less where is constrained by the observed abundances of clustered on all scales, compared to the HSB galaxies light elements. Current observations indicate that in clusters, contained in the CFA redshift survey. 2 fb ^0.04¾ J . Combined with the nucleosynthesis esti- (4) Whether in groups or clusters, LSB galaxies are usually mate of ~ 0.015/z^ (Walker et al. 1991), this leads to near the edge of the galaxy distribution. The results of íí0 ^ 0.3ΛΓοΓ To reconcile this with Einstein-DeSitter O'Neil et al. (1997a) show that, in the cluster environ-

ment, there is a limiting value of μ o which depends on local galaxy density. If one considers that the highest surface brightness galaxies of all, ellipticals, tend to occur in virialized regions then we have the surface brightness extension of the morphology-density relation t t of Dressier (1980). That is, the local galaxy density + + ? i \ limits the surface mass/luminosity density of galaxies. (5) LSB galaxies are not preferentially found in large scale voids. For instance, there are none in either the CFA bubble (Bothun et al. 1992) or the Bootes Void (Alder- ing et al. 1997). The following scenario can be advanced. Given a Gaus- sian initial spectrum of density perturbations which will eventually form galaxies, there should be many more low- Fig. 8—The luminosity function of the Virgo cluster galaxies showing the density perturbations than high-density ones. Many of these difference between corrections based on apparent magnitude from those based on surface-brightness selection effects. The solid circles/squares give low-density perturbations are subject to disruption or as- the raw counts and error bars as a function of apparent magnitude. The similation into other perturbations and hence will not pro- filled circles are the result of applying an incompleteness correction based duce individual galaxies. However, if a substantial percent- only on apparent magnitude. The open circles are the result of applying a age of them do survive to produce individual galaxies then correction based on surface brightness. These galaxies are missed not because of reduced apparent flux, but because too much of their flux is we would expect them to dominate the galaxy population. below the night-sky background. As a result, the correction for incomplete- We tentatively identify the LSB galaxy population with ness based on surface brightness considerations sets in at a substantially these 1-2σ peaks in the initial Gaussian perturbation spec- brighter magnitude than corrections based only on apparent flux. These are trum. If that is the case, we expect LSB galaxies to (1) have the galaxies in Virgo that were discovered by Impey et al. 1988. formed in isolation, and (2) be fair tracers of the mass distribution on large scales. This isolation on small scales illustrates the basic point for the Virgo cluster. When these must clearly affect their evolution since, compared to HSB ''missing galaxies" are properly accounted for, there is a galaxies, LSB galaxies have experienced fewer tidal en- significant increase in the faint end slope of the luminosity counters with nearby galaxies over a Hubble time. Tidal function relative to that which is obtained simply by apply- encounters are effective at clumping gas and driving global ing a correction for incompleteness based on apparent flux. star formation. Without this external hammer, LSB galaxies The correlation between surface brightness and magnitude would continue to evolve slowly and passively. that seems to hold for dwarf elliptical galaxies (e.g., The Faint End Slope of the Galaxy Luminosity Function: Sandage et al. 1985; Binggeli et al. 1985; Caldwell and In recent years, a lot of different values have appeared for Bothun 1987; Bothun et al. 1989) has been shown to break the faint end slope, a, of the Schechter luminosity function. down below μ0 ^ 24.5 (e.g., Bothun et al. 1991) indicating The canonical value of α ^ —1.25 for HSB field galaxies that corrections for apparent flux do not mimic the more can be compared with observations of nearby clusters that proper correction based on surface brightness selection ef- show α of ^ —1.1 (but see Wilson et al. 1997 for signifi- fects (see also Sprayberry et al. 1997; Impey and Bothun cantly steeper values for more distant clusters). However, 1997). Combining the Virgo and Fornax cluster sample of a = —1.9 has been determined for low-mass irregular gal- dwarfs yields a = —1.55 ±0.05 after these proper correc- axies in the CFA redshift survey (Marzke et al. 1994). The tions are made. Such a steepening indicates that LSB gal- safest thing to then conclude is that recent surveys have axies dominate numerically in clusters. If stellar/baryonic now produced a variety of different values of a indicating MIL increases with decreasing surface brightness for this that it is (a) uncertain and (b) may evolve with time. cluster population then they contain a significant fraction of In general, corrections for incompleteness to the galaxy the baryonic material in clusters (which might exacerbate luminosity function have been erroneously applied in the the fb problem in clusters). It is interesting to speculate that past based on apparent magnitude only. This is conceptually cluster LSB galaxies might represent the results of a phase incorrect since galaxies are selected on the basis of a com- of intense baryonic blowout at higher redshifts. These relic bination of surface brightness and luminosity. This is best galaxies then might be the source of the gas seen in clusters illustrated in the Virgo and Fornax clusters, where the dif- as well as the enrichment of that gas. In the case of the fuse and fairly large LSB dwarf were missed in earlier Virgo cluster, there are approximately twice as many dwarfs surveys but detected using the Malin method. The inte- as brighter galaxies. grated luminosity of these galaxies is brighter then many of The ARM survey of Impey et al. (1996) allows the con- the Binggeli et al. (1984) dwarfs due to their large scale tribution of LSB galaxies to the field galaxy luminosity length. The surveys in Virgo (Impey et al. 1988) and For- function to be estimated. This has been done in detail by nax (Bothun et al. 1991) revealed LSB galaxies that were Sprayberry et al. (1997) who find a faint end slope of up to 3 magnitudes brighter than the nominal magnitude a = -1.42 compared to a = -1.09 for all disk types in limit of the Binggeli et al. (1984) survey but which were the CFA redshift survey. It is too soon to tell if there is a missed by that survey. significant difference between the cluster and field faint end Figure 8 (adapted from Fig. 10 in Impey et al. 1988) slopes but in both cases, the discovery of LSB galaxies in

© Astronomical Society of the Pacific · Provided by the NASA Astrophysics Data System 756 BOTHUN ET AL. both environments yields the same result: the faint end initial-mass function that allows only a limited window of slope steepens. visibility before the galaxies fade to extremely low-surface The Faint Blue Galaxy Connection: One immediate re- brightness levels by ζ = 0. As an example, an IMF which sult from various deep surveys of galaxies (e.g., Lilly et al. forms no stars less than 1 M© produces a galaxy that is 1991; Lin et al. 1996; Odewahn et al. 1996; Driver et al. destined to have a very bright phase and then fade 1996; Glazebrook et al. 1995) was the discovery of an into mostly white dwarf remnants at ζ = 0. A collection of apparent excess, relative to local observations, of faint gal- 1011 white dwarfs distributed in a disk with scale length axies with blue colors. These enigmatic faint blue galaxies of 2 kpc would have a central surface brightness of (FBGs) are the subject of much research. Since the FBG = 27.5 which would be below the current detection population apparently does not exist at ζ = 0, reasons for threshold of LSB galaxies. This situation illustrates the kind their absence from galaxy catalogs of the local universe of extreme fading which would be necessary to hide this must be found. Two possible approaches involve either a component from being detected locally. Interestingly, the very large merger rate since ζ ^ 0.7 or rapid fading of this latest MACHO results (see Sutherland et al. 1996) strongly FBG population (Efstathiou et al. 1991; Colless et al. 1990). suggest that halo white dwarfs are the main cause of the 8 This faded population is potentially an important constituent observed lensing events towards the LMC. In that sense, the of the total baryonic content of the universe (see Cowie Galactic halo might be regarded as a LSB galaxy. 1991; Bouwens and Silk 1996). However, arguments pre- The FBGs are an artifact of uncertainties in the determi- sented above suggest that these faded remnants have not yet nation of the local galaxy luminosity function. In particular, been detected in large numbers (see also Dalcanton 1993). the faint end slope has been seriously underestimated from A number of explanations for the FBG "problem" have nearby samples. Alternatively, the local normalization, a natural linkage with LSB galaxies at ζ = 0. Some of the Φ(0), of the galaxy luminosity function could be too low. possible explanations are the following: This would result if, for instance, deep surveys were more If optically selected LSB galaxies do have normal UV efficient at selecting galaxies than nearby surveys. Driver et fluxes, then this population can become conspicuous in faint al. (1994a,b) have explored this possibility and concluded galaxy surveys when this light is redshifted into the ground that the local normalization may well be seriously underes- based filter set. This would occur, in general, over the timated. However, the effect of increasing the space density redshift range 0.5-1. It is possible that some of the FBGs at ζ = 0 can only partially offset the excess FBG counts. are in fact LSB galaxies! For example, the median luminos- A much larger lever arm is provided by steepening a. ity of the FBGs in Lilly et al. (1991) is identical to the LSB Sprayberry et al. (1997) have considered the effects of sample of McGaugh and Bottum (1994) and both samples increasing a (due to LSB inclusion) and increasing Φ(0) on are dominated by late type galaxies which are weakly clus- the ability for LSB galaxies to explain the FBG excess. tered. They conclude that the CFA survey has missed about 1/3 of The FBGs are a population of star bursting dwarf galax- the total galaxy population at ζ = 0. While this helps to ies located at modest redshift. This suggestion takes advan- resolve the apparent difference between the large numbers tage of the fact that in any GLF with α ^ -1, low mass of FBGs and the local galaxy population, this effect by dwarf galaxies dominate the space density. To produce the itself is not large enough to completely resolve the issue FBGs, however, these dwarf galaxies have to be at least an (see also Ferguson and McGaugh 1995). For additional as- order of magnitude brighter at modest redshifts which re- sistance we turn to the Lilly et al. (1995) redshift survey of quires a fairly significant star formation rate. Subsequent ^ 500 faint galaxies. Their sample has excellent quality heating of the ISM by massive stars and supernova should control and is fairly free from selection effects and is pri- be sufficient to heat it beyond the escape velocity of these marily aimed at determining the galaxy luminosity function low-mass systems (see Silk et al. 1987; Wyse and Gilmore up to a redshift ^ 1. 1992) and hence such galaxies have a significant phase of Lilly et al. detect a change in the luminosity function of baryonic blowout after which they fade to very low absolute blue galaxies by approximately one magnitude between luminosities and are hard to detect at ζ = 0. This mecha- ζ ^ 0.38 and ζ ^ 0.62, and another magnitude between nism gives the universe a channel for making baryons ' 'dis- ζ ^ 0.62 and ζ ^ 0.85. Moreover, many of these galaxies appear" with time. This hypothesis has been investigated in have been observed with HST in order to measure charac- detail by Phillips and Driver (1995) and Babul and Fergu- teristic surface brightnesses. Schade et al. (1995) find that son (1996). the disks of these blue galaxies are ^ 1 magnitude higher The number density of galaxies is not conserved and the in surface brightness at ζ = 0.8 than ζ = 0.3. These studies FBGs merge with other galaxies. It is difficult to support provide rather strong evidence for luminosity evolution in this hypothesis because (1) the FGBs are already weakly the FBGs. In a 15 Gyr old Universe, there are approxi- clustered, and (2) the required merging rate is significantly mately 3.3 billion years between ζ = 0.85 and ζ = 0.38. higher than the inferred rate at modest redshift by Patton et The data indicate that a typical FBG would decline in lu- al. (1997). The merger idea works best if the FBGs are minosity by a factor of 6 (e.g., 2 mags) over this time predominately at higher redshift, where the merger rate is period. This modest decline is quite consistent with standard higher owing to the smaller volume of the universe. population synthesis models involving a normal IMF which The FGBs represent an entirely new population of fades after a significant burst of star formation has occurred galaxies—one defined by a star-formation history or an to make these objects visible at higher redshift. The decline

© Astronomical Society of the Pacific · Provided by the NASA Astrophysics Data System HIDDEN GALAXIES 757 in luminosity is primarily a reflection of the disappearance significantly slower rate and may well experience star for- of the upper main sequence. By ζ = 0, these galaxies will mation outside of the molecular-cloud environment. certainly not have faded to levels that preclude their detec- (2) Surface-brightness selection effects have been severe. tion, and indeed many of them are likely to be represented A proper accounting of them has increased the local number by the ζ = 0 red LSB population which has now been density of galaxies and steepened the faint end slope of the discovered. galaxy luminosity function. Despite this progress, these selection effects still exist and thus we do not yet have a representative, volume-limited sample of nearby galaxies. 6. SUMMARY (3) LSB galaxies span the entire galactic mass range. We conclude this review by again referring to Fig. 1. They are not exclusively low-mass galaxies but include the These results on the space density of galaxies as a function most massive disk galaxies discovered to date (e.g.. Malin of central surface-brightness require a basic adjustment in 1). LSB disks likely are the manifestation of 1-2σ isolated the way we think about galaxies. Much of the current peaks in the initial density fluctuation spectrum. These thought is implicitly one dimensional, with one parameter lower-density peaks have longer collapse times and trace (like luminosity or morphology) dominating the way prob- the mass distribution in a relatively unbiased way. lems are approached. This is no longer sufficient. Surface (4) LSB galaxies are embedded in dark-matter halos brightness selection effects have been severe. Our results which are of lower density and more extended than HSB now indicate that up to 50% of the general galaxy popula- galaxy halos. In this sense, disk galaxy surface mass density tion is in the form of galaxies with central surface bright- and subsequent evolution may be predetermined by the ness below 22.0 mag arcsec-2. Moreover, the space density form of the dark-matter halo. Surface mass density appears remains flat out to the limits of the data, and the space to be the single biggest driver of disk-galaxy evolution. density of the lowest surface brightness disks (/¿ο ^ 25.0 -2 A number of people have helped support this project mag arc sec , the limits of current data) is vastly higher over the last decade. We gratefully acknowledge Mike Dis- than would have been anticipated based on Freeman's Law. ney for making us think, David Mahn for his wizardry and LSB galaxies offer a new window onto galaxy evolution. patience with us, Jim Schombert for being there, Steve Because of this, the quest to find LSB galaxies continues. Strom for pointing the way, Jay Gallagher for telling one of Over the next few years we hope to extend our sensitivity us (G.D.B.) to work on something ''hard," Mark Cornell by using wide field CCD surveys of the sky at dark sites. for assistance with the Texas observations, and Allan These and new surveys by other groups should extend the Sandage for originally showcasing the smudge galaxies. We current data by two magnitudes down to μ0 = 27.0 mag 2 gratefully acknowledge support from the NSF under Grant arcsec" . The major goals of these new surveys are to AST-9005115 and AST-9003158 without which this project determine if the space density of galaxies as a function of could never have been sustained. μ,0 continues to remain flat over a factor of 100 in μ0, and to detect what we so far have failed to detect in large numbers—the red LSB population that must result from the REFERENCES faded remnants of galaxies that no longer can form stars. In Aldering, G., Bothun, G., Kirshner, R., and Marzke, R. 1997, ApJ, this regard, the Sloan Digital Sky Survey will be extremely submitted helpful if its automatic image recognition system can reli- Babul, Α., and Ferguson, H. 1996, ApJ, 458, 100 ably detect LSB galaxies. At levels well below the sky Binggeli, B., Sandage, Α., and Tarenghi, M. 1984, AJ, 89, 64 brightness, however, LSB galaxies are often defined by dis- Binggeli, B., Sandage, Α., and Tammann, G. Α. 1985, AJ, 90, connected regions of pixel intensity. While the human eye 1681 is rather good at finding these, its not clear if that particular Boroson, T. 1981, ApJ, 46, 177 algorithm can be reproduced in a machine. Bothun, G., and Sullivan, W. 1980, ApJ, 242, 903 Despite the power of CCD surveys and automated detec- Bothun, G. 1981, Ph.D. thesis. University of Washington tion algorithms, we are inevitably brought back to the start- Bothun, G., Beers, T., Mould, J., and Huebra, J. 1985, AJ, 90, ing point of this review. Messier cataloged galaxies long 2487 Bothun, G., Beers, T., Mould, J., and Huchra, J. 1986, ApJ, 308, before their extragalactic nature was understood, and it is 510 ironic that his catalog was a reject list of stationary objects Bothun, G., Caldwell, N., and Schombert, J. 1989, AJ, 98, 1542 for comet hunters to avoid. We now know a lot more about Bothun, G., Geller, M., Kurtz, M., Huchra, J., and Schild, R. 1992, galaxies, but the selection biases that have operated for 200 ApJ, 395, 349 years have not been fully overcome. It would be unwise to Bothun, G., Schombert, J., Impey, C, Sprayberry, D., and Mc- presume that we have yet revealed the true population of Gaugh, S. 1993, AJ, 106, 530 galaxies. Bothun, G. D., Impey, G. D., Malin, D. F., and Mould, J. R. 1987, We leave the reader with a small list of things that we AJ, 94, 23 hoped they have learned from this review article: Bothun, G. D., Impey, C. D., and Malin, D. F. 1991, ApJ, 376, 404 (1) LSB galaxies exist. Lurking beneath the brightness of Bouwens, R., and Silk, J. 1996, ApJ, 471, L19 the night sky are real galaxies with evolutionary histories Brown, T., Ferguson, H, and Davidsen, A. 1996, ApJ, 472, 327 substantially different from the processes that produced the Bruzual, G, and Chariot, S. 1993, ApJ, 405, 538 Hubble sequence of spirals. In particular, LSBs evolve at a Caldwell, N., and Bothun, G. D. 1987, AJ, 94, 1126

Colless, M., Ellis, R., Taylor, Κ., and Hook, R. 1990, MNRAS, O'Connell, R. et al. 1992, ApJ, 395, L45 244, 408 Odewahn, S., Windhorst, R., Driver, S., and Keel, W. 1996, ApJ, Cowie, L. 1991, Phys. Scr., 36, 102 472, 13 Dalcanton, J. J. 1993, ApJ, 415, L87 O'Neil, K., Bothun, G., Smith, E. P., and Stecher, T. 1996, AJ, Dalcanton, J. J. 1995, Ph.D. thesis, Princeton University 112, 431 Dalcanton, J. J., Spergel, D. N., Gunn, J. E., Schmidt, M., and O'Neil, K. 1997, Ph.D. thesis, University of Oregon Schneider, D. P. 1997, AJ, in press O'Neil, K., Bothun, G., and Cornell, M. 1997a, AJ, in press Davies, J. 1990, MNRAS, 244, 8 O'Neil, K., Bothun, G., Schombert, J., Cornell, M., and Impey, C. de Blok, E. 1997, Ph.D. thesis. University of Groningen 1997b, AJ, submitted de Blok, Ε., and McGaugh, S. 1996, ApJ, 469, L89 Patton, D., Pritchet, C., Yee, H., Ellingson, E., and Carlberg, R. de Blok, E., McGaugh, S., and van der Hülst, T. 1996, MNRAS, 1997, ApJ, 475, 29 283, 18 Phillips, S., Disney, M., Kibblewhite, E., and Cawson, M. 1987, de Jong, R. S. 1995, Ph.D. thesis. University of Groningen MNRAS, 229, 505 de Jong, R. S. 1996, A&A, 313, 45 Phillips, S., and Driver, S. P. 1995 MNRAS, 274, 83 de Jong, R. S., and van der Kruit, P. C. 1994, A&AS, 106, 451 Pickering, T. 1997, Ph.D. thesis. University of Arizona Disney, M. J. 1976, Nat., 263, 573 Pickering, T., van Gorkom, J., Impey, C., and Bothun, G. 1997, Disney, M. J., and Phillips, S. 1983, MNRAS, 205, 1253 AJ, submitted Dressier, A. 1980, ApJ, 236, 351 Quirk, W. J. 1972, ApJ, 176, L9 Driver, S. P., Phillips, S., Davies, J. I., Morgan, I., and Disney, M. Romanishin, W., Strom, K. M., and Strom, S. E. 1983, ApJ, 263, J. 1994a, MNRAS, 268, 39 94 Driver, S. P., Phillips, S., Davies, J. I., Morgan, I., and Disney, M. Sandage, A. 1961, The Hubble Atlas of Galaxies, Carnegie Insti- J. 1994b, MNRAS, 266, 155 tute of Washington, Pub No. 618 Driver, S. P. 1995, Ph.D. thesis. University of Cardiff Sandage, Α., Binggeli, B., and Tammann, G. Α. 1985, AJ, 90, Driver, S., Couch, W., Phillips, S., and Windhorst, R. 1996, ApJ, 1759 466, L5 Schade, D., Lilly, S., Crampton, D., Hammer, F., LeFevre, O., and Efstathiou, G., Ellis, R. S., and Peterson, B. A. 1988, MNRAS, Tresse, L. 1995, ApJ, 451, LI 232, 431 Schombert, J. M., and Bothun, G. D. 1988, AJ, 95, 1389 Efstathiou, G., Bernstein, G., Tyson, J. Α., Katz, N., and Gu- Schombert, J. M., Bothun, G. D., Impey, C. D., and Mundy, L. hathakurta, P. 1991, ApJ, 380, L47 1990, AJ, 100, 1523 Ellis, G., Grayson, E., and Bond, Η. E. 1984, PASP, 96, 283 Schombert, J. M., Bothun, G. D., Schneider, S. E., and McGaugh, Evrard, Α., Metzler, C., and Navarro, J. 1996, ApJ, 469, 494 S. S. 1992, AJ, 103, 1107 Ferguson, H. et al. 1991, ApJ, 382, 69 Schwartzenberg, J. M., Phillips, S., Smith, R. M., Couch, W. J., Ferguson, H. C., and McGaugh, S. S. 1995, ApJ, 440, 470 and Boy le, B. J. 1995, MNRAS, 275, 171 Fish, R. A. 1964, ApJ, 139, 284 Silk, J., Wyse, R., and Shields, G. 1987, ApJ, 322, L59 Freeman, Κ. C. 1970, ApJ, 160, 811 Sprayberry, D. 1994, Ph.D. thesis. University of Arizona Glazebrook, K., Ellis, R., Colless, M., Braodhurst, T., Allington- Sprayberry, D., Impey, C. D., Bothun, G. D., and Irwin, M. J. Smith, J., and Tanvir, N. 1995, MNRAS, 273, 157 1995a, AJ, 109, 558 Hubble, E. 1922, ApJ, 56, 162 Sprayberry, D., Bernstein, G. M., Impey, C. D., and Bothun, G. D. Impey, C. D., and Bothun, G. D. 1989, ApJ, 341, 89 1995b, ApJ, 438, 72 Impey, C. D., Bothun, G. D., and Mahn, D. F. 1988 ApJ, 330, 634 Sprayberry, D., Impey, C. D., Irwin, M. J., and Bothun, G. D. Impey, C. D., Sprayberry, D., Irwin, M. J., and Bothun, G. D. 1997, ApJ, 482, 104 1996, AJ, 105, 209 Strauss, M., and Willick, J. 1995, Phys. Rep., 261, 271 Impey, C D., and Bothun, G D. 1997, ARRA, in press Sutherland, W., et al. 1996, to appear in Proceedings of Workshop Kennicutt, R. C. 1989, ApJ, 344, 685 on Identification of Dark Matter, Sheffield, Sept. 1996 Knezek, P. M. 1993, Ph.D. thesis, University of Massachusetts Tinsley, B. M. 1975, Lectures on Stellar Populations, Scuola Su- Kormendy, J. 1977, ApJ, 217, 406 periore di Astronomía, Erice, Italy, May 20-31, 1974, Societa Lilly, S. J., Cowie, L. L., and Gardner, J. P. 1991, ApJ, 369, 79 Astronómica Italiana, Memorie, vol. 46, Feb. 1975, p. 3-68. Lilly, S. J., Tresse, L., Hammer, F., Crampton, D., and Lefevre, O. van den Bergh, S. 1959, Publ. David Dunlap Obs., 2, 147 1995, ApJ, 455, 108 van den Bergh, S. 1976, ApJ, 206, 883 Lin, H., Kirshner, R., Schectman, S., Landy, S., Oemler, Α., van der Hülst, J. M., Skillman, Ε. D., Smith, T. R., Bothun, G. D., Tucker, D., and Schechter, P. 1996, ApJ, 464, 6 McGaugh, S. S., and de Blok, W. J. G. 1993, AJ, 106, 548 Loveday, J., Peterson, Β. Α., Efstathiou, G., and Maddox, S. J. Walker, T. P., Steigman, G., Schramm, D. N., Olive, Κ. Α., and 1992, ApJ, 390, 338 Kang, H.-S. 1991, ApJ, 376, 51 Marzke, R. 0., Geller, M. J., Huchra, J. P., and Corwin, Jr., H. G. White, S. D. M., Navarro, J. F., Evrard, A. E., and Frenk, C. S. 1994, AJ, 108, 437 1993, Nat., 366, 429 McGaugh, S.S. 1992, Ph.D. thesis. University of Michigan Wilson, C. 1996, preprint McGaugh, S. S. 1994, ApJ, 426, 135 Wilson, G., Smail, L, Ellis, R., and Couch, W. 1997, MNRAS, McGaugh, S. S., and Bothun, G. D. 1994, AJ, 107, 530 284, 915 McGaugh, S. S., Bothun, G. D., and Schombert, J. M. 1995, AJ, Wyse, R., and Gilmore, G. 1992, AJ, 104, 144 110, 573 Young, J., and Scoville, N. 1991, ARAA, 29, 581 McGaugh, S. S. 1996, MNRAS, 280, 337 Zaritsky, D., Smith, R., Frenck, C., and White, S. 1997, ApJ, 478, Mo, H. J., McGaugh, S. S., and Bothun, G. D. 1994, MNRAS, 39 267, 129 Zaritsky, D., and Rix, H. 1997, ApJ, 477, 118 Nilsson, P. N, 1973, Uppsala General Catalog of Galaxies (Upp- Zwicky, F. 1957, in Morphological Astronomy (New York, sala, Uppsala Astronomical Observatory) Springer-V erlag)