Back to the article index

The descriptions of galaxies in the NGC, which go back to the Herschels, refer essentially to the sizes and give rough indications of brightness. More could hardly be expected, because these early observations were all visual, and everybody who has looked at galaxies like the Andromeda spiral knows that the visual pictures are lacking in detail. Actually it is only a hundred years since Lord Rosse, with his big reflector, found spiral structure in a number of galaxies.

The situation changed when visual observations were replaced by photography at the turn of the century, particularly when the large reflectors came into operation - first the Crossley reflector at Lick Observatory, later on the Mount Wilson reflectors. Pictures of the nearer galaxies showed such a wealth of detail that it was necessary to bring some order into the observed forms and structures.

Some early classification schemes, such as that of Max Wolf, were based on photographs made with astrographs of intermediate size, but they did not find much acceptance. In the early 1920's, two classification schemes were proposed almost simultaneously, one by Lundmark, the other by Hubble. Both were based on the large collections of plates that had been made at the Lick and Mount Wilson Observatories. There was not much difference between the two systems. Lundmark tried to classify many details in addition to the main features. It so happened that in the long run the simpler classification of Hubble won out, and it has been in general use since that time.

Let me remind you of the essential features of Hubble's system. He started with a class that he called the E nebulae - the spheroidal galaxies. They run from round forms (E0) to elliptical forms (E5). We can leave out E7, which Hubble classified in an entirely different way, and S0, about which I shall speak later. The letter E stands for elliptical, and the numeral for the ellipticity, simply defined by the ratio (a - b) / a, where a is the major axis and b the minor axis in the isophotes. I talked with Hubble about the whole thing a few years ago, just before he died, and it seemed that he was convinced that the limiting case was about E6, with the smaller axis about 40 percent of the larger axis.

The E galaxies are distinguished by a remarkably smooth intensity distribution, as we know from the early work of Hubble and the later work of de Vaucouleurs and others. Many of these galaxies have a very remarkable feature - a very bright semistellar nucleus at the center. The companion of the Andromeda spiral, M32, furnishes one of the brightest cases in which this nucleus can easily be seen.

We shall see later that a central nucleus is also a feature of the spirals; though sometimes exceedingly weak or even absent, it is present in many cases. If it is present in spirals, which are flattened systems, it shows flattening. For instance, the central nucleus of the Andromeda spiral has dimensions about 2".5 x 1".5, and its axis is directed beautifully in the direction of the axis of the Andromeda spiral itself. So this nucleus, this high condensation at the center, is a feature of all these galaxies.

Here we have a very remarkable thing. We are dealing essentially with two systems - a central spherical system, surrounded by a disk that exhibits spiral structure. In a spiral galaxy, the central system (especially evident in NGC 5494, for example) can shrink until it is finally reduced to a semistellar point. A very short exposure would be needed to show the semistellar point in M101; what is visible in the photograph is a much larger bright area.

In classifying the spirals, Hubble distinguished the groups Sa, Sb and Sc, the distinguishing criterion being essentially the behavior of the spiral arms. For instance, in his description of Sa the spiral arms emerge at the edge of the central system; in the early spirals both they and the central lens are still unresolved, and the arms are densely coiled. As we proceed along the series, the central nuclear area shrinks at the expense of the growth of the arms, which by and by uncoil, until finally the central area has shrunk to a semistellar point, and all the mass seems to be in the spiral arms. This is Hubble's original description, but he agreed completely that it would be simpler today to classify the spirals simply by the size of the central lens. Those with very large central lenses can be called Sa; those with intermediate central lenses, Sb; and those where the lens has shrunk to a semistellar point (actually a huge cluster of stars), Sc. In what follows I shall adopt this very simplified system, based on the size of the central lens.

Parallel to this series of ordinary spirals is the series of so-called barred spirals, which typically have a central system - a bar within a ring area - and spiral arms springing from this ring. In Hubble's classification, SBa is a type that is crossed by a bar, and the rest of the material is in the form of a ring around it; this ring breaks up at two points, and begins to show spiral structure. In the final stage the bar shrinks at the expense of the finer spiral arms, and we then find the latest type of barred spiral, SBc.

Whereas all the systems discussed so far show obvious rotational symmetry, there were left about 2 to 3 percent of the galaxies in which rotational symmetry was not obvious. They were irregular in shape, and Hubble classified them as irregular galaxies. Today we know that they are exclusively of the Magellanic Cloud type.

The final form in which Hubble presented his sequence is shown in Fig. 1. He started with E0; E6 is a limiting case for elliptical galaxies, and beyond that there was a bifurcation into two parallel series, the normal spirals and the barred spirals, with some intermediate cases in between. At the junction he postulated a hypothetical class S0, which marks the transition from the E galaxies to the spirals.

FIG. 1. The sequence of nebular types.

In his first paper Hubble admitted that, although he could perhaps find transitional cases between the E and SB series, he had not yet found a single case of spiral systems without spiral arms. Later on, such systems were actually discovered - systems where spiral arms would be expected but were not found; Hubble distinguished such systems as S0. Such galaxies have a central lens - a disk - but no spiral structure. We know today that S0 systems are the prevalent type in dense clusters of galaxies, and, besides showing no spiral structure, they show no trace of gas or anything of the sort. The explanation of these systems is entirely different from what Hubble thought it was. We know today that the reason is that the dust has been wiped out by collisions between the cluster galaxies. We shall see later on that this is not true for the field galaxies of type S0. But it has nothing to do with evolution.

In his latest classification (which is given in the Hubble Atlas, edited by Sandage) Hubble changed the position of S0 and, as we shall see later, put in its place a new series of stripped spirals, the S0's: S0a, S0b, S0c, parallel to Sa, Sb, Sc. This S0 series has been stripped of gas, or (as in some cases that we shall see later) the gas has been exhausted. This series is devoid of dust and gas, and shows no spiral structure. You would get exactly such systems if you were to strip the spiral structure from any of the great spirals.

The irregular systems show at first sight very little indication of rotational symmetry. But for a number of reasons the appearance is deceptive. There are clear indications that these systems are very flat and in a state of rotation. When we come to IC 1613 we shall see that it is a very irregular galaxy with a most ragged outline, but still, when the Population II is traced out with the radio telescope, it is found to have a perfectly beautiful elliptical outline. The ragged outline shown by the photographs simply indicates the area in which I think star formation is at present going on; the mass distribution is absolutely regular.

Another argument relates to the Magellanic Cloud systems, which we see in all orientations in the sky. When we see them sideways they are always highly flattened. But when we see them face on, or very nearly face on, like IC 1613, we observe no concentration toward the center, or only a very slight one; if they had any other form we should observe it. Therefore they are highly flattened systems. Finally I may add that a regular rotation of the Magellanic Clouds has been found from both optical and radio observations.

Magellanic systems very often show a central bar. These systems may be closely related to the barred spirals. We practically always find one bar, and we must be cautious in calling these systems irregular. The outline looks irregular, but the chaotic outline is that of the present area of star formation. I should like to set the matter of the irregular galaxies right at this point: we have every reason to believe that they are highly flattened systems, though not as flattened as Sb or Sc spirals, and in consequence are probably in a state of rotation.

The irregular systems cannot be very young, since they contain Population II stars and cluster-type variables. The irregular features consist of young stars, supergiants and so forth, which make a terrific splash and contribute lots of light. But their total mass can be very small compared to the total mass of the system.

Now it is obvious from the scheme as Hubble described it that he had an impression or a belief, although he never quite admitted it, that it represented a continuous sequence. But I believe, on the contrary, that Lindblad put his finger on the essence of Hubble's classification when he suggested that it is a series of increasing flattening, of increasing angular momentum.

It is a very remarkable thing that in the E series we have very simple systems from E0 to E6, and among the spirals we have in general two systems, one of high angular momentum (the disk), and one of low angular momentum (the lens), for a system of the same mass.

Even if we leave out considerations of angular momentum, we have a series of increasing flattening. This certainly presents a problem, a very important problem. Aside from the E galaxies, why do the spirals always show the combination of a disk and a central spheroidal system? It must reflect the original density distribution of the gas. It is hardly possible to think anything else, because otherwise we simply could not understand the great irregularities of the density distribution of subsystems, which are well known; they could not even have established themselves.

This is really a very serious problem for the theoreticians to look into. It almost seems as if, in the spirals, most of the original angular momentum has gone into the disk. Can we imagine that, at some era in the past, the central spheroidal system of low rotation and the disk with very fast rotation actually resembled the equilibrium figure of the gas? One should really look into these things.

To return to the Hubble classification: of course his catalogue referred to the brightest galaxies; there was a selection according to luminosity. His classification of the E's has to be extended because of the discovery in recent years of E galaxies of very small size and low luminosity, of the type of the Sculptor and Fornax systems. When they were discovered at Harvard, Shapley regarded them as a new type of stellar system. We know today that they are simply the continuation of the ordinary series of E galaxies, which starts from giants like M87 in the Virgo cluster or M32; they represent the smallest E galaxies that we know today. Examples are the Leo II system, a spherical galaxy about the same size as the Sculptor system, containing over 300 cluster type variables; the Leo I system, an elliptical galaxy near Regulus; and the elliptical Draco system. Such galaxies must in future be covered by the classification system. There is a whole series of luminosities, and as the luminosity decreases, the pictures look very different from the beautiful standard E galaxies in Hubble's original paper.

Let me add a few words about the merits of Hubble's system. It is a very simple one, but we shall see later, when we try to understand the make-up of galaxies, that there is really not much sense in making a system that covers all the little details of spiral structure. About the merits of Hubble's system I can speak from experience; I have used it for 30 years, and, although I have searched obstinately for systems that do not fit it, the number of such systems that I finally found - systems that really present difficulties to Hubble's classification - is so small that I can count it on the fingers of my hand.

Actually, Hubble's system is even better than one would expect from his own classifications. In his work on the Shapley-Ames catalogue he often made the remark ``peculiar'', as if his system was not a good one. The reason for this ``peculiar'' was that Hubble had a blind spot where two galaxies were involved instead of one, a case that happens quite frequently. I remember the difficulty that I had in convincing him that a certain galaxy is double. When we finally showed that the two systems had different radial velocities, he still called it a hypothesis. If you eliminate the double systems, I am sure that the number of exceptions is unbelievably small, so efficient is the system.

A good system should be applicable over the whole observed range. Every time I took long exposures with the 100-inch or 200-inch (90 minutes or so in the blue, about 4 hours in the red) under excellent conditions, I always examined the numerous small and distant galaxies distributed over the field to see how they behaved and whether they followed the classification scheme, even when the plates were taken for another purpose. Even systems with diameters of 5" could easily be classified, especially with the red and blue plates put together in the blink microscope. The Sa, Sb, and E galaxies were easy. The more difficult cases were the Sc's, where the central nucleus has shrunk to a semistellar point; on the blue plates one very often saw only a little elliptical patch. But when one switched to the red image, the central nucleus was visible, and one was absolutely certain that one was dealing with an Sc.

I think that the Hubble classification, which just deals with the basic features, conveys all that we want to know. It will become clear later that all the details of the spiral structure are what might be called accidental - variations on the same theme. I believe that the more recent systems, like those of de Vaucouleurs, are a retrogression to the time of Lundmark, and even beyond. I can see no sense in covering these details; even if you put them all in you still do not get the full picture. If you want to study the variations on the theme Sc, you simply have to take the plates and examine them - only then do you get the full story. No code system can replace this. The code system finally becomes so complicated that only direct inspection of the plates helps.

I think that the search for exceptions to Hubble's classification is very significant. One of the examples that I said I could count on the fingers of one hand is IC 51. I found it on Schmidt plates and could not make head or tail of it. Then I took a photograph with the 200-inch. The system has a very regular disk, but there is a remarkable thing: smoke rings come out of the nuclear region. If it were not for this feature, it would be an Sb-Sc. It is not a serious exception. In another case, although the system shows central symmetry, the dust pattern violates the symmetry completely, shooting out to large distances, although there are no external systems in the neighborhood. If we were dealing with two systems that had totally collided, on a simple expansion picture with the velocity of separation we should find the culprit within a few degrees, even for the brightest galaxies. In this case nothing was to be found; it is a real exception.

The irregulars were the wastebasket in Hubble's system, because he dumped his doubles into this class. But he still stated that the greater part of the irregulars were of the Magellanic Cloud type. We know now that they are exclusively of the Magellanic Cloud type. We shall see later that there is a group among the irregulars that do not show any Population I stars (giants and so forth); M82 is one of these. It is remarkable, with color index about +0.9 mag and spectral type A5. Holmberg has distinguished Irregular I (where Population I dominates) and Irregular II (where Population II dominates) on the basis of color index. Hubble's classification of irregulars by forms is obviously a good thing; and with the forms alone we get a very close correlation with content.

The distribution of galaxies in space is very chaotic, and far from uniform; single systems are rather rare. There is a story here: Hubble and I had a long-standing bet of $20 for the one who could first convince the other that a system which he had found was single. We never could decide the bet; neither of us could pull out some distant fellow - in some cases there really was a companion and in other cases there could be. So single galaxies may be rare. Doubles and multiples are frequent; from these we go to groups, from a dozen to a few hundred; from there to clusters. In the Coma cluster and others we even see a strong concentration toward the center. Finally we come to clouds of galaxies, which are usually not mentioned. At the present time we can draw no sharp distinction between these different groupings.

If you stagger all these clouds through space, only a few of the nearer ones will stick out. Earlier, Shapley at Harvard found some. One of the most amazing of these clouds is the one found by Shane in his Lick Survey at about 15^h 20^m, +5°; the globular cluster M5 lies at the edge of this cloud. The Lick Survey gives some 200 galaxies per square degree. At Shane's suggestion I had plates taken a number of years ago in the blue and the red with the 48-inch Schmidt, which is much more powerful than the 20-inch. It is really most fantastic; the area is literally covered with galaxies. And the interesting thing is that on this plate you find clusters of galaxies dispersed everywhere in this large cloud, which means that a large cloud contains a large number of clusters of galaxies. Another interesting point is that the galaxies of the cloud are preferentially Sb and Sc, but in the clusters themselves we find S0's, although all of them are obviously connected with the cloud.

We do not know the borderlines of the subdivisions. Holmberg's beautiful paper in the Lund Publications refers only to the doubles among the brighter galaxies of the NGC; it is a model of reasoning. Among the groups we have Stefan's Quintet, and the one that Seyfert found a few years ago; you find these groups everywhere. If you look at the Palomar Sky Survey plates they are unavoidable. You see them together in relation to their types; you know it is a unit; you can check on the radial velocities. But you can only divide according to numbers after they are better understood.

If we restrict the term ``clusters'' to the very rich associations of galaxies which at the same time show a strong concentration toward the center, like the Coma and Corona Borealis clusters, then we should have to include the Virgo ``cluster'' among the ``groups''. So the ``groups'' at the present time are simply a catch-all for intermediate concentrations, from multiples on; they would contain anything from half a dozen, eight, ten galaxies to several hundred, as in the Virgo cluster. It remains to be investigated whether the larger groups, such as that in Virgo, should be regarded as clusters.

It is very simple to decide whether a star is a member of a star cluster; because the total mass is very small, one can always fall back on proper motion and radial velocity and can always decide about membership. In the case of clusters and large groups of galaxies, the dispersion in velocity is very large, so the red shift does not mean anything. It is very hard to assign individual members even to a cluster of galaxies, and in the case of the Virgo group the problem is truly difficult. It is quite easy to derive the brighter part of the luminosity function for clusters like the Virgo and Coma clusters. But it is practically impossible to get the fainter end, because we are counting against a rising background; as we come to the faint end we are simply lost statistically, and have to use physical arguments. We shall find that such physical arguments can eventually enable us to disentangle these problems, but it has not yet been done.

The problem of the stability of the groups is analogous to that for associations of stars in our Galaxy. The groups with positive energy will long ago have dispersed, and we may call semistable the groups that are left now. Occasionally, owing to a close encounter, a member gains sufficient energy to be ejected, but we are certain that the chance for a capture is negligible in the present state of the expansion of the universe, and we can forget about it. And it is an important corollary that the members of the groups that we observe today are of common origin. This will be especially important later on in connection with some conclusions that we shall draw regarding our own group of galaxies: they are all of common origin, and therefore we have good reason to believe that as galaxies they are of the same age.

Our own Local Group may consist of 17 or 18 members. Among the next brightest galaxies, NGC 2403 is the main member of a group; M81 is the main member of a brighter group, probably just as extensive as our own, because we know we have not found its faintest members like our Sculptor system. Third comes M101, where again we are in an extensive group, all belonging together. And there are dwarf systems marching in between, such as the Sextans and Wolf-Lundmark-Melotte system. So in our immediate neighborhood we have the typical arrangement - all the systems in dense groups, and these groups part of a larger one.

We are fortunate that our Galaxy is a member of one of these groups, the so-called Local Group. In Table 1 I have arranged them according to luminosity. The system of highest luminosity (about -20 mag photographic, actually a little brighter) is the Andromeda Nebula, M31. Next comes our own galaxy, though of course we do not know the luminosity. The Large Magellanic Cloud is probably more luminous than M33; NGC 6822 and IC 1613 are dwarf galaxies of the Magellanic Cloud type. A probable new member is IC 5152, a typical dwarf galaxy in the southern hemisphere recently photographed by Evans; from the reports and the picture I should guess that it is an actual member of our Local Group. Then come all the E galaxies. The series of spirals and irregulars seems to stop around -13 mag. The ellipticals include the near companions of M31, then NGC 185 and NGC 147, the more distant companions of M31, then Fornax, absolute magnitude unknown; the last four in the table have about the same luminosity. The figures given are all on the correct new photometric scale, and will not be changed any more.

Leaving out the new member, our Local Group consists of 17 systems, of which 10 (59 percent) are E galaxies, of intermediate to faint luminosity. This picture is very different from what we get over the sky at large, where we give preference by selection to the brighter systems. The luminosity range of galaxies is from about -20 mag to about -9.5 mag or -10 mag; this is a range of 10 magnitudes, a ratio of 10,000:1.

The largest system in the Local Group, and one of the largest in the Universe, is the Andromeda spiral. If we transfer it to the Virgo cluster, whose distance is known from the red-shift - velocity relation, we see that there may still be a handful of systems in the Virgo cluster that are still brighter, but it is really near the top. Its absolute magnitude is -20 mag, and this is conservative it has a linear diameter of about 50 kpc out to the distance of the large HII regions on either side. As an example of the smallest diameter (all these small ones are about the same) I give that of the Draco system, 1.6 kpc, determined from the most outlying cluster-type variables. So the ratio in diameters is 1:30. Thus there is a huge variety among galaxies in luminosities and in diameters. I need hardly say that the main mass of the Local Group is contained essentially in M31 and our own Galaxy - or if you wish you can include the systems up to the Small Magellanic Cloud. If you add up the masses you come to something very close to 5 x 10¹¹ solar masses.

A similar variety in size and luminosity is shown by M31, M32, IC 1613, NGC 185, and NGC 147, which are all practically at the same distance, so a comparison of their apparent dimensions gives a true picture.

It has been suggested that NGC 6946 and IC 10 may be members of the Local Group, but NGC 6946 is definitely not a member; the absorption is well determined and the redshift is too high. Its distance modulus on the new scale is 26.8 or 26.9 mag. But IC 10 could be a member; it is one of the incredible cases that one would not believe possible. A photograph shows a beautiful piece of a spiral arm, with three H II regions of which we have taken the spectra. And that is the only thing you can see; you look everywhere around, and no extragalactic nebulae come through - there must be just a hole through which we see this piece of spiral arm. The radial velocity is not yet decisive, because the red shift depends very much on the assumptions one makes about the component of galactic rotation in that direction. To get the absorption an enthusiast, an optimist, would have to observe some of the blue stars in this system; but it could be done.