A galaxy that goes red before it changes shape

There are two old ways of sorting galaxies. One goes by shape. Edwin Hubble’s sequence runs from smooth ellipticals through lenticulars to spirals with open arms, and it describes how a galaxy’s stars move. The other goes by colour, and its histogram has two peaks with almost nothing between. Red galaxies have stopped forming stars and blue ones have not, so it describes what a galaxy’s gas is doing now.

These two sorted lists agree remarkably often. Red galaxies tend to be ellipticals and blue galaxies tend to be spirals, and for decades one was taken as a proxy for the other. Morphology and colour both correlate with environment, with mass, and with each other. Yet the two are measuring different things, so it is not obvious they should agree at all. A galaxy’s colour is a question about its last few hundred million years. Its shape is a question about the orbits of its stars, which take many rotations and a violent event to change.

So the agreement implies that the things that happen to a galaxy’s orbits and those that happen to its gas are connected. The disagreement is more informative, and it comes to about a fifth of the population. Since 2008 the volunteers of the Galaxy Zoo have made it possible to count those exceptions by eye, in large samples. They have found the red spirals, which have stopped forming stars while keeping their discs, and blue ellipticals, which are the opposite. Each of those is a galaxy caught between two states, and the numbers and the order in which they happen together tell a story that a crossing time measured from colour alone cannot.

How quickly a disc goes red #

Begin with a disc that has formed stars steadily for ten billion years, and turn off its star formation at a stroke. Nothing dynamical has to happen for its colour to change. The blue light was coming from massive stars that live for a few tens or hundreds of millions of years, and those are the first to die. The older stars have been sitting there all along, and they emerge from behind the young ones.

A serviceable account of this needs only that a single-age population fades as a power of its age, a little faster in the blue than in the red. Integrate that over the disc’s ten billion years of history, and the colour and brightness are determined as functions of the time since star formation stopped.

A billion years is short for a disc. It is about four rotations of the part of the disc where the Sun lives. So a galaxy is very likely to show an optical colour typical of the red sequence while its stars are still on the same nearly circular orbits they had when it was blue. As an earlier argument pointed out, a gradual shutdown lengthens this, and each colour index has its own memory. The conclusion holds regardless of those differences: reddening is fast compared with anything that would rearrange the stars.

A disc that dims and a bulge that does not #

The fading is the part usually left out of accounts of the colour, and it changes what morphology means.

The asymmetry between the two parts is simple and large. A bulge is an old population, and an old population has no young stars to lose. Its light fades slowly as the turn-off moves to lower masses. A disc has been supported by recent star formation, so turning that off loses it a good share of its brightness at a stroke. After two billion years it has dimmed by almost a magnitude, while the bulge has dimmed by about a sixth of one.

This is the point at which a stellar-population calculation enters into a dynamical description. Whether a galaxy looks like an early type has long been assessed from how centrally concentrated its light is, or by the fraction of the light in its central component. That ratio, the bulge-to-total, is commonly described as a property of its structure. But it is a ratio of light. Hold every star’s orbit fixed and fade one component more than the other, and the measured ratio changes.

So some part of the movement towards an early-type shape is not a change in shape at all. It is the removal of a bright thin overlay from a structure that was already there. The mass in the bulge has not grown, nor has its share of the mass; a measurement based on stellar mass maps, or one using the near-infrared where the young stars matter less, would show a much smaller change. This was understood in outline long ago for the lenticular galaxies. Lyman Spitzer and Walter Baade proposed in 1951 that spirals in crowded clusters lose their gas in collisions and become lenticulars, and later work on stripped discs supplied the fading that goes with it. What the figure adds is how much of the transition can be done this way, and how soon.

The answer is that it depends strongly on the starting point. A galaxy with a substantial bulge can be reclassified without anything else happening. A galaxy with almost no bulge, however, fades indefinitely and never looks like an early type at all. There is a limit, and it marks the point beyond which the argument needs something to actually change the orbits.

The plane of colour and shape #

Put the two quantities on one figure and the possibilities separate.

The first track gives a natural home to the red spirals. A disc reddens first, almost vertically on this figure, and arrives on the red side with its bulge share scarcely changed. It then spends billions of years moving slowly to the right. A galaxy found at any moment during that interval is a red disc, a passive spiral. In their study of those galaxies, Karen Masters and collaborators selected face-on discs specifically to exclude spirals that are merely reddened by dust, and they found truly passive ones, with older populations and less recent star formation than normal spirals. They are rare at low mass and a substantial share of the most massive spirals. So massive galaxies are red regardless of their shape, which is a strong hint that whatever stops star formation at high mass does so without destroying the disc.

The red spirals carry another feature nobody expected. A very high fraction of them have bars: about seventy per cent against twenty-seven per cent for blue spirals. That might mean bars stop star formation, by driving gas inward to where the orbits turn and using it up, and it might mean that bars form more easily in a disc without gas. A correlation cannot say which came first, and in this case both are physically reasonable.

The second track points to the other minority. A blue elliptical has been made to look like a spheroid by a merger, and it has not yet faded out of the burst the merger triggered. Its colour moves up the diagram with its shape already fixed, and it reaches the red sequence in a billion years or so. These are less common at low redshift, a few per cent of early types. The two minorities, then, take opposite orders. In one, the gas goes first and the orbits later if at all. In the other, the orbits go first and the gas goes in a burst immediately afterwards. A merger can do both, but it is not the only way of doing the first.

The delay as a clock #

The separation between going red and crossing a shape cut has a length. This is the clock proposed by the earlier treatments of this valley: a measurable lag on a process nobody has watched.

The result is less simple than the slogan that colour leads and shape follows, and the sign of the delay is worth attention. For a galaxy with a bulge share already near the cut, the shape can cross first. Its colour is still in the valley while its bulge share has already passed 0.3. The ordering of colour and shape depends on where the galaxy began. Only a true disc goes red long before it looks round, and for the most disc-dominated galaxies the delay exceeds the age of the universe unless some process moves the stars.

The bulge share is one measurement of morphology, though, and it is not what a volunteer in Galaxy Zoo is recording. A human classifier looks for arms and a disc. Spiral arms are short-lived wave patterns, and they depend on the disc staying cold enough to respond. The work the arms do heats the stars. Without new gas to form cold new stars, the disc warms up and the arms weaken over a few rotations. This gives a second clock, independent of fading, and it should begin at the shutdown and run on the timescale of the disc’s rotation: one to a few billion years. So a red spiral still with visible arms has either stopped forming stars only recently, or something still feeds its arms. A red disc with no arms is an S0, whose stars are on disc orbits and which looks smooth only because it has lost its pattern. So morphology by eye is itself a clock, or several.

Reading the delay from a census #

A delay can be measured without following any galaxy at all, because a population in a steady flow holds a record of how long each stage lasts. The argument is the same as the one that turned an empty valley into a crossing time: the count at a stage is proportional to the time spent in it.

So the abundance of red spirals is a clock, with the mean time on the red side as its unit. Studies of the census give a share of about a quarter or so of red galaxies with a disc morphology at intermediate masses, though that number depends strongly on mass and on how one counts disc-like S0s. If it holds, then the delay is around one or two billion years, and that fits with the fading and arm-decay arguments. It does not require any transformation of orbits to be going on at that same time.

The clock has its limits. Mass-dependent quenching, environmental quenching and mergers have different delays, and the census averages over them. It also relies on a steady flow, while the rate at which galaxies have been arriving on the red side has fallen a good deal since redshift 1. But the same reasoning has a virtue: unlike the colour track, it relies on no population-synthesis model, and it uses only the counting of objects in two categories.

A sequence once read in the wrong direction #

The labels on the sequence carry a historical fossil that this argument helps to place. Astronomers still call ellipticals early types and spirals late types, and Hubble was careful to say that those words described complexity of form and not age. Many readers of his diagram nonetheless took it as a track, with smooth galaxies developing arms as they evolved. The stellar populations ended that reading. Ellipticals hold the oldest stars, and the star-forming galaxies at the far end of the sequence are the ones still building. If there is any flow along the diagram, it runs from right to left.

A flow of that kind needs a mechanism, and the first candidates were all violent. Mergers could certainly make an elliptical from two spirals, and simulations from the 1970s onward showed tails and bridges that matched real pairs. But a merger is a poor account of most red galaxies with discs, since a major merger destroys the disc it would have to leave behind. The growth of massive ellipticals by small additions is a late stage acting on objects that were already red and already round.

Fading gives the flow a much quieter first step. It moves a galaxy leftward along the sequence by the reckoning of light, without any encounter at all, and the amount it can do is fixed by a stellar population and not by a collision rate. That makes it the null hypothesis for any claim that a process has transformed a galaxy. Before a change of shape is attributed to harassment in a cluster, or to a bar, or to a merger, the first thing to subtract is what the stars would have done on their own by growing old. Comparisons of cluster and field samples have not always done this, partly because a measurement of shape in one band cannot separate the two effects. The band matters as well: an image in the blue will show a galaxy near the end of its transition as more bulge-dominated than an image in the near-infrared does, and two surveys in different filters will then disagree about how many transformed galaxies there are. That disagreement is another form of what this essay has been measuring.

The fifth galaxy, and the two processes #

There is a broad connection hidden in this, and it answers a question often treated as one of cause and effect. Does a galaxy change shape because it stopped forming stars, or stop forming stars because it changed shape? Each has a mechanism. A growing central bulge can stabilise a disc against fragmentation, so its gas stops forming stars even if it stays, a process called morphological quenching. Conversely, a galaxy that has lost its gas will have its arms fade and its disc look smoother, so that shape follows colour. The fading adds a third relation: a galaxy that has stopped forming stars gets a higher bulge share without any cause at all, simply because of arithmetic.

So a measured correlation between bulge fraction and being red is partly caused by fading, and it has often been presented as evidence for the role of bulges in quenching. How much is due to each has to be separated by using mass rather than light, or by measuring the motions of stars rather than their light. Since a spheroid is supported by random motion and a disc by rotation, the velocity maps from integral-field surveys record what the orbits are doing, and they are unaffected by how bright each component happens to be. Measured this way, the fraction of the population made of slow rotators is small, and a large share of the galaxies classed as early-type turn out to be rotating discs. That is consistent with this account: a good many of the galaxies that moved from blue to red did not change their orbits much.

The surprising connection is therefore between stellar evolution and galaxy classification. The law that says massive stars burn out first has a consequence for a label assigned to a whole galaxy by a human looking at a picture. A statement about nuclear fuel in a single star, integrated over a disc, moves a galaxy along the Hubble sequence.

What the picture cannot show #

The stellar populations here are power laws fitted to the behaviour of real models, and the normalisation of the colours has been chosen for convenience. The cuts themselves, at a colour of 0.65 and a bulge share of 0.3, are round numbers for illustration, and any actual classification depends on the survey, the band and the fitting method. No dust is included, even though a dusty disc is the main impostor among red spirals, and metallicity is taken as constant.

The shutdown is instantaneous, and the disc’s star-formation rate has been constant before it. Real histories rise and fall, and a slow decline makes the disc redden less abruptly and over a longer time. Nor does the model include any change in orbits at all, which is the point of it but also its main limit. In a real galaxy that has lost its gas, stars are heated by interactions and by whatever spiral structure remains, the disc thickens, and any mergers will add to the bulge. The figures isolate what fading alone does. They give an upper bound on the fading explanation and no bound on the others.

Similarly, the figure in which bulge share rises as discs fade is drawn for one measurement, the red band. In the near-infrared the effect is much weaker, and in the ultraviolet much stronger, so the numbers quoted are not a universal measure of the effect.

Still open: whether shape ever moves first in a disc #

The ordering of colour and shape has been studied statistically and not in single objects. A galaxy’s stars keep a record, however: the ages of disc stars and bulge stars can in principle be separated in integral-field spectroscopy of nearby galaxies, and they give the time at which each component last formed stars. If bulges grow before the discs quench, the youngest stars of the bulge should predate the last stars of the disc in most galaxies, and if the reverse holds, they should not. Early results point in different directions depending on mass, and the sample sizes for clean decompositions are only now becoming large enough for this to be attempted. Whether there are two sequences with opposite ordering, as the two tracks of the plane suggest, or one with a broad spread, is what those data should determine, and the answer bears on where the gas of a massive disc went when it finally stopped being used.