The Role of Mergers in Galaxy Evolution

Disk Galaxy: NGC 3370
Credit: NASA/ESA

When we look around us in what we astronomers call "the nearby Universe", most of the galaxies that we see can be divided into two basic groups. There are the "disk" galaxies, sometimes called "spiral" or "late type" galaxies, which are flat like a saucer. We of course live in a disk galaxy, and our nearby companion, the Andromeda Galaxy (M31), is one also.

Giant Elliptical galaxy M87
Credit: NASA/ESA

Then there are the "elliptical" or "early type" galaxies. These look more like round balls of stars from any angle, though they can be slightly flattened. Have a look at this previous post for more information and more  pictures of disk galaxies and elliptical galaxies. Also see this recent post for a discussion of how we measure and quantify galaxy type or morphology. Although it is not apparent just from looking at the images of these galaxies, disk and elliptical galaxies are different in several other ways besides their morphology. Disk galaxies also contain cold gas, which provides fuel for new stars, while elliptical galaxies don't have much gas and contain very few young stars. The motions, or orbits, of the stars within these galaxies are also very different. In disk galaxies, the stars and gas move around the galaxy on regular, nearly circular orbits, with smaller up and down motions, like animals on a merry-go-round. In elliptical galaxies, the stars move around with more random motions like a swarm of bees.

Ever since astronomers first noticed that galaxies came in these different types (which goes all the way back to Edwin Hubble), they have been wondering why. Are these galaxies different because they had different properties from birth? Or could something happen to galaxies to make them one way or the other -- were they shaped by their environment or even perhaps by a traumatic event? This is sometimes called the "Nature or Nurture" debate.

An important clue came from galaxies that don't fall neatly into either of these categories, like the ones shown above. They aren't very common, but we see them often enough to know that they could be telling us something important. These strange-looking galaxies tell us that sometimes, galaxies can interact and even collide. See this previous post introducing galaxy mergers.

Computer simulation of a merger of two disk galaxies
Image Credit: Cox et al. 2008, MNRAS, 384, 386

This inspired theorists to try work out in more detail just what would happen to galaxies if indeed they did interact with one another. We set up "particles" that represent stars and gas in two disk galaxies, for example similar to the Milky Way and M31. We also include the "halos" (extended spherical envelopes) of dark matter that we now believe surround all galaxies, and make up most of their mass (why we believe that is another topic for another day). Fortunately, we think that dark matter interacts with itself and with normal matter according to the usual laws of gravity, and doesn't feel any other forces, so it is actually relatively easy to program a computer to predict what it will do (even though we don't know what it is). Then we set the galaxies on a collision course and use a supercomputer to compute what would happen to the stars, gas, and dark matter as the galaxies move towards one another and eventually begin to interact. There are several nice animations of these kinds of simulations in previous blog posts -- here and here.

The picture above shows a time sequence of snapshots from such a computer simulation of a merger of two nearly equal-mass galaxies. The color scale shows the density of the stars, and the little number in the top left of each panel is the time that has elapsed since the beginning of the merger, in billions of years (Gigayears). The dotted line shows the trajectory of the orbit. The first thing you probably notice is the long streams of stars that are drawn out on both sides. These are called "tidal tails" and are caused by the same kind of tidal forces that the Moon exerts on the Earth (only of course much, much stronger). You might also notice that the centers of the galaxies seem to get denser and more compact. By the end of the simulation, 6 billion years later (remember we think the Universe is about 13.5 billion years old), the two galaxies have merged into one and the remaining galaxy no longer looks like a nice thin disk of stars -- it's a much rounder structure, more like the elliptical galaxies that we saw above.

What is actually happening here? There is a lot of space between stars in galaxies relative to the size of the stars, so the stars themselves do not collide with one another. However, gravity can perturb those nice circular orbits that the disk stars were on. Basically some of the energy from the galaxies' motions relative to one other gets transferred to the stars, scrambling the orbits and making the stars move around more randomly.

Rate of new stars born as function time during a galaxy merger. Image credit: Patrik Jonsson

The gas that was in those two disk galaxies is also dramatically affected by the interaction. The gas gets driven into the nuclei of the galaxies, and when gas gets dense, it can form new stars more efficiently. So the rate of new starbirth goes way up as the galaxies interact and merge. I've shown a little graph here showing the rate at which new stars are being born as a function of time, along with pictures of the merging system along the way. As you can see, the rate of new stars being born spikes up as the galaxies start to interact, and peaks when the galaxies coalesce. It then dies off again as the gas gets used up. In addition, some of the massive stars start to explode as supernovae, which deposits a large amount of energy in the gas. This can heat the gas up and blow it away, removing the fuel for further star formation. Observational studies have shown that galaxies in close pairs do seem to be forming stars more efficiently than isolated galaxies, which seems to support this picture. CANDELS will allow us to further study the connection between mergers and star formation, which will provide important tests for theories of galaxy formation.

Artist's depiction of an accretion disk around a black hole
Image Credit: A. Hobart (CXC)

There is another very intriguing possible consequence of galaxy interactions. If there are massive Black Holes lurking in the centers of the progenitor galaxies, the strong torques during the merger could funnel the gas so close to them that it would begin to be accreted onto the Black Holes. As gas approaches very close to the Black Hole, it forms a hot dense structure called an "accretion disk", which can glow very brightly (I've shown an artist's rendition here). These accreting Black Holes are called Active Galactic Nuclei (AGN), or Quasars, and have been the subject of other posts. Computer simulations of galaxy mergers suggest that these events could cause the Black Holes to grow very rapidly. Theorists have further suggested that the energy radiated by the accreting Black Hole could heat up the remaining gas and even blow most of it out of the galaxy! There is a beautiful animation of a simulation that tries to model that process here.

To sum up, we think that mergers can change disk galaxies from flat to round, scramble their stars from regular circular orbits to random orbits, and maybe can activate black holes that blow away their gas and shut off their star formation. So perhaps all galaxies were born as disks and some get transformed into ellipticals by mergers. It's a nice story, but a number of open questions remain. Do we see enough mergers? Could there be other processes that are important? If mergers cause black holes to shine, why don't we see AGN preferentially in disturbed-looking galaxies? CANDELS is helping us to answer these questions.

Summary of the General Assembly of the International Astronomical Union in Bejing

The International Astronomical Union XXVIII general assembly took place in Beijing, China during the last two weeks of August. The IAU General Assembly is perhaps one the largest astronomical meetings, gathering hundreds to thousands of astronomers working in many different topics: from stars and planets to galaxies, the interstellar medium, or cosmology. It only happens once every three years (can't wait for Hawaii 2015) and it holds roughly 12 different simultaneous sessions every day ranging from small joint discussions and special sessions to the large symposia. Here I'll try to present a brief (and extremely biased) summary of my experience there. Other astronomers with different perspectives are encouraged to complement my side of the story!

Banner for the IAU General Assembly Meeting in Beijing, China

The most interesting meeting for me was the "The intriguing life of massive galaxies" which included both an observational and theoretical perspective on the formation and evolution of the most massive galaxies. This symposium took place only during the second week, but there were other very interesting discussions during the first week such as special sessions on UV-emission in galaxies, cosmic evolution in galaxy clusters, and secular evolution. The latter featured very interesting discussions on the processes driving morphological transformations of disk galaxies and also quenching mechanisms related with these transformations. I specially enjoyed John Kormendy's summary on bulges (classical ellipticals), pseudo-bulges (bulges within spiral galaxies) and the processes involved in their formation, namely dissipational processes for the bulges and secular processes (bars, rings) for the latter, and their kinematic properties: pseudo-bulges are fast-rotators whereas bulges are mainly dispersion dominated (check his last paper and the 2004 review for more details). There were also observational results from the SAURON and ATLAS3D teams showing the motion of these structures for galaxies in the local universe. A quick note from the UV-emission special session, G. Bruzual presented some results from the recently updated version of their stellar population synthesis models (commonly used for SED fitting). Among the upgrades they will include improved libraries in the UV region of the spectra and a revised treatment of the thermally pulsating AGB phase to reduce its contribution to the NIR emission, bringing the values closer to their 2003 models.

Let's go back to the symposium on massive galaxies, where I spent most of my time. The session started with some talks about recent observational results on z~7-10 galaxies. Bouwens and Stark talked about the weak evolution of the specific star formation rate (sSFR, the star formation rate divided by the stellar mass) with redshift, and their efforts to improve their results including new data from deepest IRAC surveys. We certainly missed our CANDELS experts here because that was all the mention of very high redshift galaxies for the remainder of the week. On the cosmic-noon side, S. Wuyts, I. Trujillo and E. Daddi presented summaries on our current knowledge of the structural, kinematic, and star formation properties of galaxies at z~2 and the evolution of these down to z~0. Let's see if I can summarize some of the highlights mentioned in their talks (and many others) without making a horribly long post!

The most popular figure of the meeting was, without discussion, the stellar mass-size relation. Although it was intended to depict the sizes of all massive galaxies, it was almost exclusively used to describe the remarkable size evolution of non-star forming galaxies. Since the notion that the small sizes of these galaxies are robustly measured is now widely accepted, most of the discussion revolved around the possible growing mechanism, with minor merging being the most frequently advocated. Nonetheless, several speakers acknowledged that current measurements of the merger ratio seem to be insufficient to explain the observed size growth from z~2 to 1. It was also highlighted in the closing summary that the size evolution of the massive non-star forming population does not involve a single population, but a mixture of older and recently quenched (star formation has recently been halted) galaxies that are being continuously added, therefore making the interpretation even more complicated.

The second most popular figure (or referred term) was the so-called main-sequence that refers to the correlation between stellar mass and SFR. This relation has been observed up to z~3 with little change in the slope and an evolving normalization towards higher SFRs at early times. The importance of this relation was highlighted by both observers and theorists. The first showed increasing observational evidence of a secular, steady growth, phase for galaxies in the main-sequence co-existing with a starburst phase for galaxies with enhanced star-formation efficiencies (forming more stars per volume of gas), probably indicative of differences in the interstellar medium. Several talks highlighted the importance of the slope of the main-sequence and the relative fraction of galaxies in the starburst phase to characterize the evolution of the stellar mass function at both the most and least massive ends. A slope of unity implies that the galaxy mass function (the number of galaxies with a given stellar mass in a given volume of the Universe) simply shifts to higher masses without changing its shape, typically characterized by a Schechter function. Recent CANDELS results from Wuyts, Targett, and Kartaltepe were mentioned in the context of studying the evolutionary mechanisms that move galaxies from the main-sequence to the quiescent red-sequence after perhaps a brief transit through the starburst phase.  

Regarding talks on galaxy structure and kinematics, we saw really impressive results from SINS and the more recent MASSIV surveys using the NIR/IFU SINFONI to make velocity maps of star-forming massive galaxies up to z~2.  These surveys provide evidence of rotating disk structures in most of these galaxies, although the increased velocity dispersions suggest a more turbulent star-formation (along with short lived clumps) different from the more ordered disks observed at z<1. On the quiescent side, CANDELS results from Van der Wel, Bruce, and Mozena seem to identify a population of passive disks co-existing with the compact spheroids. This could place strong constraints in the quenching mechanisms and transformation processes, as we know that the majority of passive galaxies will become large ellipticals. Different velocity dispersion measurements for small samples of quiescent galaxies at z > 1 confirm that in fact these are 2 to 5 times denser than their local analogs as suggested by their small sizes. Van de Sande presented some new results from their XSHOOTER survey including a galaxy with a velocity dispersion above 400km/s! 

Before making this post too long, I'd like to finish this brief overview with a quick note on the theory side and the models for galaxy formation. Talks from Oser, Naab, and Johansson discussed what they call a two-phase formation scenario, which consists of the in-situ formation of a massive compact galaxy at z>~2 followed by a continuous growth in mass and size as a result of mergers and satellite accretion. Gabor and also Peng and Lilly focused on a more general process controlling galaxy growth and quenching through the evolution of the dark matter halo (its mass and temperature) and the galaxy environment. In this regard, and to conclude on the observational side, Duc showed some extremely deep amazing images from the Virgo NGVS survey reaching down to 29 mag/arcsec. With this kind of data it is possible to study galaxy-galaxy interactions with a high level of detail: enormous tidal features, minor merging, and even satellites of satellites around massive galaxies!

All in all, the IAU general assembly provides a wonderful venue to catch up on the most recent updates for your favourite topic, and also a fantastic opportunity to discover interesting results in other related topics. The large number of attendants ensures that you'll meet a lot of different people and make new connections, so I strongly recommend the experience. Don't miss the next one in Hawai'i 2015!

How Do We Measure Galaxy Morphology?

We've discussed in a previous post how much of what we know about galaxies comes from studying their morphology (in other words, what they look like). This gives us a sense of what a galaxy's internal structure is and, at least in the local universe, galaxy morphology correlates with various other galaxy properties, such as color, star formation rate, and environment. Data from the Hubble Space Telescope is ideal for galaxy morphology studies, especially at high redshift, due to its exquisite high resolution and HST's location above the Earth's blurring atmosphere. With HST data, we are able to see detailed structure in distant galaxies that would be very difficult or impossible from the ground. So how do we actually go about measuring a galaxy's morphology?

A sampling of galaxies with various morphological types and colors.
Credit: Zsolt Frei, Institute of Physics, Eötvös University, Budapest

One of the typically used methods for studying galaxy morphology is to visually look at individual galaxies and classify them into a previously established system. This is what Edwin Hubble did when he first put forth the Hubble sequence. Hubble's classification scheme is what most modern schemes are based on (dividing galaxies roughly into spirals, ellipticals, and irregulars) but there are many other structural features one may wish to include, such as whether or not a bar or tidal features are present. Once a system has been decided upon, the next step is doing the work and looking at a lot of galaxies! Visually classifying galaxies can be very time consuming and can possibly be biased by several different factors (Are you more likely than others to identify certain features? Were you tired when looking at the last set of objects?). However, the human eye is particularly adept at this type of pattern recognition and is unmatched when it comes to identifying faint features in galaxies.

Since classifying galaxies takes a long time, it can be difficult to build up large samples of objects with classifications. It was this problem that lead a group of astronomers to create the Galaxy Zoo project. These astronomers involved the general public in galaxy classifications. These citizen scientists (including hundreds of thousands of volunteers) classified galaxies in the nearby Universe from the Sloan Digital Sky Survey (originally; today Galaxy Zoo includes images from Hubble data as well) via the Galaxy Zoo website. Over time, these classifications have built up to an incredible database of classifications and have lead to public data releases and a number of publications.

Despite the benefits of visual classifications, astronomers have searched for ways to automate galaxy morphology methods so that thousands to millions of objects can be investigated without the need to look at each one individually. Plus, a computer program that can classify galaxies would be able to do the exact same thing for each object, without bias. Several methods have been developed by different groups and they each have their own advantages and disadvantages.

One common method is to use a computer program to quantify how a galaxy's light distribution changes from the center of a galaxy to its edge. Is it very concentrated in the center like an elliptical galaxy or is it extended like a disk galaxy? Perhaps it's somewhere in between? This method is very successful at separating galaxies into spiral and elliptical categories. Another common technique is to quantify how asymmetric a galaxy is. This is done by rotating an image of a galaxy by 180 degrees and subtracting it from the original. Whatever is left tells you how asymmetric it is (if it was perfectly symmetric, like a circle, no signal would remain after subtracting). This method is commonly used to identify galaxy mergers since they are typically messy irregular systems. Another method, which was used in the paper presented in this previous post, combines two different parameters, one called 'Gini' - which is a measure of a galaxy's light distribution, and the other called 'M20' - the second order moment of a galaxy's brightest pixels. By combining these two, galaxies can be split into bulge dominated and disk dominated objects. This method has also been used to identify galaxy mergers.

By combining some of these structural measures with visual classifications, we can calibrate each method (what are the parameters of galaxies with known types?) and study how well each works as a function of a galaxy's brightness and its distance from us. The more distant a galaxy is, the fainter its features are, and the more difficult its morphology can be to measure. In future posts, we will discuss some CANDELS science results using all of these different methods and applying them to galaxies in the distant universe.

Did CANDELS find the first stars in the Universe?

Did that title grab your attention? Spoiler alert: the answer is no (probably), but the journey to that answer was interesting! Read on!

In today's blog post, I'm reporting on a paper I wrote with the help of many other members of the CANDELS team which has just been published in the Astrophysical Journal. The title of the paper is "The Evolution of Galaxy Rest-Frame Ultraviolet Colors from z = 8 to 4" (and you can find it here). There's a lot in the title, so let's break it down: the word "evolution" at the beginning means that we're studying how galaxies change (i.e. evolve) with time. In this particular paper, we're focusing on a time period in the early universe, from around 0.5 to 1.5 billion years after the Big Bang. The universe is now 13.8 billion years old, so this is early on indeed, but as I'll show, things change quickly even in this short time period. The "from z = 8 to 4" denotes this time period; astronomers use a term called "redshift" to denote distance (see these earlier blog posts, #1 and #2, for more details), so a redshift (or "z") of 4 corresponds to 1.5 billion years after the Big Bang, and a redshift of 8 is about 0.5 billion years (or 500 million years) after the Big Bang.

The meat of the title is in the middle: "rest-frame ultraviolet colors"; this means we are looking at the colors in the ultraviolet, which is just bluer than the eye can see. However, due to the expansion of the universe, at these high redshifts this ultraviolet light is shifted to the near-infrared. Thus, we add the qualifier "rest-frame". What this means is that even though the light we observe is in the near-infrared, we have a good idea of the redshift of a given galaxy, so we can figure out what the intrinsic ultraviolet (or UV) color is (you may have seen this referred to as "k-correcting", which is a similar concept).

So, why is this interesting? The UV light tells us a lot about ongoing star-formation. This is because the UV light from galaxies is dominated by massive stars (many times more massive than the Sun). Massive stars are very hot, so they are very bright in the ultraviolet, and they look blue (by comparison, the Sun appears yellowish, and lower mass stars look red). But, massive stars are very short-lived, some exploding as supernovae after only 10 million years. So, if you look at a galaxy and it is bright in the UV and also blue in color, you immediately know that massive stars are still around, so it must be actively forming a lot of stars!  Cool, right?

The story doesn't end here. A number of studies predict that when we look at galaxies very early on, they should appear even bluer than galaxies observed at later times. This is because in the Big Bang the only elements to form were hydrogen and helium (with trace amounts of lithium). If you make a star with just those elements, it will be very hot; this is because gas needs to cool to very low temperatures to condense and form "normal" stars like those we see in today's universe. However, to cool down, gas typically relies on atoms heavier than helium (which we astronomers call "metals") to radiate away energy. If you don't have those elements, then stars do form, but we think that they end up being much more massive (maybe even more than 100 times the mass of the Sun); these massive stars are extremely hot, so they are very blue, much bluer than any star with a normal allotment of metals.

Figure 1: The ultraviolet colors of galaxies at different redshifts (the top vertical axis shows the amount of time in billions of years since a redshift of 20; add 0.2 to change it to since the Big Bang). The red circles show the average of all galaxies, while the blue circle shows the average colors of faint galaxies at a redshift of 7.

So, we were motivated to look through the CANDELS data for distant galaxies, and measure their colors to see how blue they looked. Sounds simple, right?  I'll skip over a lot of the details (though you can learn how we found the galaxies here), but our main results are shown in Figure 1. This plot shows the colors of galaxies at each redshift we study.  The red points show the average colors of all galaxies, but we'll focus on the blue point first, since it represents the average color of faint galaxies only (which tend to be bluer). The gray line and arrow shows the color we would expect if these galaxies had stars with very little metals; if our galaxies had very little metals, we would expect them to be below this line. What we see is that even when we look at the faintest galaxies, while they do look very blue, they are not so blue that we think they host stars with very low metallicities. In fact, taking into account the error bar on this point, these galaxies have colors consistent with the local galaxy NGC 1705 (shown as the cyan bar); while NGC 1705 is blue, it has been studied very intensely, and it doesn't have any metal-free stars.

What this tells us is that even though we're looking less than one billion years after the Big Bang, galaxies have already been enriched by metals a significant amount. This means that star-formation must have already been occurring for a while, so that when we have the capability to look to yet higher redshifts (which we will later this decade with the James Webb Space Telescope), there should be plenty of galaxies burning bright for us to see!

There is one last point to make, this time looking at the red points, which are the average colors of all galaxies at a given redshift.  We see that they start out fairly blue at z=7, and get progressively redder. We think that this is due to the increased presence of cosmic dust grains (composite particles made up primarily of carbon and silicon) in these galaxies. Dust has the funny effect of making the colors of a galaxy look redder, since they absorb and scatter blue light more efficiently than red light, hence the phrase "dust reddening". What we're witnessing here is the build-up of vast reservoirs of dust in these galaxies, in only one billion years of cosmic time! Although dust may sound boring, it has a dramatic effect on the colors of galaxies, as you can see. Also, all of this UV light absorbed by the dust has to go somewhere, and its energy actually gets re-emitted in the far-infrared (much redder than your eye can see). In fact, when we look at galaxies closer to home, many of them are brighter in the far-infrared than in the UV, implying that much of their UV light has been absorbed by dust. What our results here tell us is that while this is likely not an issue at a redshift of 7, it becomes progressively stronger as you move to lower redshift. So, if you want to study galaxies at high redshift, much like the Alamo you must "remember the dust!"

How I Ended up Studying Distant Galaxies

As Janine Pforr mentioned, we've decided that from time to time we will be using blog posts to introduce ourselves and say something about how we got into astronomy. I'm writing this post on a long flight back from Arizona to Baltimore, so apologies in advance if I ramble a bit.

I am Henry C. Ferguson, known to my colleagues as Harry. I am co-Principal investigator of CANDELS along with Sandy Faber. I am an astronomer at the Space Telescope Science Institute (STScI), in Baltimore Maryland. STScI is the science operations center for the Hubble telescope, and will be operating the James Webb Space Telescope (JWST) when it launches in 2018. All the scientists at the institute divide their time between their own research and providing support for the observatories. My research interests broadly center around galaxies and cosmology, and I've been spending most of my time over the past couple decades on surveys of the distant universe.  For my support work, I have most recently been running the Webb Instrument Team as we try to work out the details of how to operate and calibrate the cameras and spectrographs on the new telescope. I've done similar work in the past for Hubble. In my spare time, I enjoy playing Irish traditional music on hammered dulcimer and accordion, most recently as a member of the Tanzania Ceili band.

I've been interested in astronomy since my early teens. I went to high school at St. Paul's, a boarding school in Concord, New Hampshire. The school had a small observatory, and a remarkable science teacher, Walter Hawley. There were no formal astronomy classes, but there was an astronomy club. Skies were reasonably dark, and a friend and I spent many nights observing and sketching the Messier objects. Walter Hawley arranged trips for the astronomy club to the Harvard College observatory, where we got to do some astrophotography with a much larger telescope (a 16-inch reflecting telescope). He had also made the connection with some folks that my parents knew, who had a ten-inch refracting telescope that was equipped with a camera and spectrograph. The telescope had been purchased in a previous generation by a vice-president of General Electric, who was also an amateur astronomer. When it was built, I suspect it was probably the most advanced telescope in the country not owned by a university. The telescope together with the mount probably weighed a couple of tons, and it was inside a big copper-clad observatory with a rotating dome. The camera was a professional device that used glass plates, mounted on a moving stage so that you could track a nearby guide star to keep the images sharp during a long time exposure. I basically had the telescope to myself for a couple of summers, and taught myself by trial and error how to take deep sky images, process them in the dark room, and make nice prints. Back in high-school, the astronomy club would have open nights for elementary school students, scout troops, and the like, and club members would give a little astronomy lecture or slide show before the telescope viewing. So I learned something about galaxies, stars and planets in putting together those shows.

Walter N. Hawley, science teacher
at St. Paul's School. In 1998 he was
awarded the Thomas Brennan award
for excellence in astronomy teaching
by the Astronomical Society of the
Pacific. He also has a main-belt asteroid
named after him. 

Before going on, I'd like to add a few more notes about Walter Hawley. In addition to having a real passion for astronomy, he has to be one of the most intrepid, energetic, and organized people I ever met. When I was a student at the school, we had two telescopes - the 5-inch Alvan Clarke refractor in the observatory, and portable Questar telescope that we could take out into a field. By the time he retired, the school had an Astronomy Center with four domes, a chart house, and a 27-inch robotically-operated reflecting telescope equipped with a CCD camera. About a quarter of the school now takes one of the astronomy courses sometime before they graduate. Walter Hawley's vision and persistence didn't stop at the edge of the school grounds. He was instrumental in building the Christa McAuliffe planetarium in Concord, in honor of the local science teacher who was to have been the first teacher in space when she perished in the space-shuttle Challenger explosion in 1986. His other passion was sailing. He was not a sailor for sport, but I think approached it as a combination of art and engineering. He supervised the building of a wooden yawl at a shipyard run by two 80-year-old boat builders in New Bedford Mass. It was launched while I was still in high school. When I was in college, he sailed it back and forth across the Atlantic, and spent several months in the Carribean on his boat, with his dog, sailing by day and observing variable stars by night. I count myself extremely fortunate to have encountered such a remarkable person during my formative years.

I majored in astronomy at Harvard. The Center for Astrophysics there has one of the largest concentrations of astronomers in the world. At the time, in the late 70's and early 80's, it was the operations center for the Einstein X-ray observatory. I worked with Rick Harnden as a research assistant. Astronomers working on Einstein at the time were not only making great advances in our knowledge of the X-ray sky, but were also inventing new ways to calibrate, analyze and interpret data. Working with Rick, I learned that solving the mystery of why an instrument behaves the way it does is often a subtle and interesting challenge in and of itself. That's part of what makes working as a support astronomer at an observatory appealing, at least to me.

After college I taught science for a year at Maret School in Washington DC, and then went to the Johns Hopkins University for graduate school. When I entered graduate school, the launch of Hubble was imminent. It was still imminent when I finished grad school eight years later. I worked for a few years on the Hopkins Ultraviolet Telescope with Arthur Davidsen and a close-knit team astronomers, students and postdocs. The telescope was scheduled to be launched in the space shuttle in 1986, coinciding with the passage of Halley's comet. On January 28 1986, we watched in horror as the Challenger exploded. Our mission had been next in line. The telescope finally flew four years later as part of the Astro-1 mission, and again in 1995 as part of the Astro-2 mission. But in the meantime, I was left searching for a new PhD thesis topic.

Allan Sandage in front of the 100-inch Hooker telescope
on mount Wilson (Image from Carnegie Observatories/
Carnegie Institution of Washington)

Fortunately for me, another remarkable person arrived as at Johns Hopkins as a visiting professor at just about the same time. Allan Sandage had been Edwin Hubble's assistant when he was right out of graduate school, and was probably one of the most famous astronomers alive. As a graduate student, he had been instrumental in figuring out how to connect the observations of the colors and brightnesses of stars in globular clusters to the emerging knowledge of stellar structure and nuclear fusion, and was able to use that knowledge to estimate the ages of the oldest stars in our galaxy. As Hubble's assistant and for a good portion of his career thereafter, he built much of the foundation of modern observational cosmology. He is well-known for his measurements of the Hubble constant, the expansion rate of the universe, and for his battles with other astronomers over the various systematic uncertainties in these measurements. But he also laid the foundations for studies of very distant galaxies, in a classic paper in 1961 outlining what the great Mount Palomar Telescope could do, and later in a review article on observational tests of world models that grew out of a course he gave while I was at Hopkins. I believe he has told some people that I was the only student able to do the problem sets. That's not true. I was a senior graduate student and mostly was auditing the course. I don't recall even doing the problem sets. I do recall several amazing 3-4 hour sessions in his office, talking about my thesis, but then free-associating on many questions about galaxies and cosmology and how to use galaxies as measuring rods or chronometers to help explore the evolution of the universe. I also recall that when I handed him a draft of a chapter of my thesis, it would come back to me practically the next day with voluminous comments on all aspects of the work from organization to writing style to the technical minutiae of the analysis. I have never met anyone before or since who paid so much attention to the words on a page.

Sandage had arrived at Hopkins with a stack of enormous glass photographic plates with deep images of the Fornax Cluster of galaxies and a few other nearby groups of galaxies. I spent a couple years inspecting, measuring and classifying galaxies on these plates, and then trying to make sense of the correlations that we found in these measurements. Most of the galaxies in nearby clusters are dwarf-elliptical galaxies. I became very good at identifying these amidst thousands of more distant galaxies. It turns out that these galaxies follow a pretty tight correlation between size and brightness. I helped quantify that with my thesis measurements, but to this day we don't really understand what causes this correlation.

Deep-sky image obtained by Tom Shanks
with the William Herschel Telescope.
The color-stretch on this image shows the faint-blue
quite prominently. If you look closely you will notice that
most of the space between the bright reddish and greenish
galaxies is filled with very faint blue galaxies.

After grad school, I spent three years as a postdoc at the Institute of Astronomy in Cambridge. One of the most talked-about topics in astronomy at the time was the problem of "faint blue galaxies:" there were too many of them. A factor of ten more than expected from the favored cosmological model. Most cosmologists were unwilling to accept that the problem might be in the geometry of the universe itself, rather than in our understanding of galaxies. I spent several years working on the problem from the galaxy perspective, thinking that whatever these faint, blue galaxies were, they might be related to the dwarf galaxies I had been studying in nearby clusters. Together with another postdoc, Arif Babul, I explored a somewhat radical model, which postulated that gas in dwarf galaxies collected there early in the universe, but didn't didn't become cool enought to form stars for billions of years. When the gas finally cooled, it formed stars in a rapid burst, making the galaxies briefly shine brightly enough to be seen at great distance. The theory was at least plausible, and matched a lot of the observations.

A small portion of the Hubble Deep Field.

This was just one of several competing models that were being published when I arrived back in Baltimore as a Hubble Fellow at STScI. One of the ways to try to figure out what was going on would be to take a really long time exposure with Hubble, counting up the number of galaxies to even fainter limits. Ideas like this were floating around when the first images from the refurbished Hubble telescope came back to earth in mid 1993, and showed that it would now be capable of obtaining the sharp images that had been promised. Bob Williams, the STScI director at the time, decided to pour most of his "director's discretionary time" into a very long time exposure -- the Hubble Deep Field (HDF). I became deeply involved in planning and executing these observations, calibrating the data, and trying to interpret the results. The observations immediately ruled out our theory, and many of the others. In fact, within a couple years, observations of distant supernovae, along with other lines of evidence, made it clear that the cosmologists had been wrong: the faint-blue-galaxy problem was due primarily to an incorrect assumption about the geometry of the universe.

The Hubble Deep Field became one of the most studied patches of sky and demonstrated the value of bringing combining observations not just from Hubble but from many different observatories. It also demonstrated the value of making the data immediately public to the astronomical community, without the usual proprietary period to a small group of investigators.  It paved the way for the CANDELS survey and many others.

It has been a long couple of flights, with a delay on the connecting flight. So I fear I have gone on too long. I'm returning to Baltimore from a planning meeting for the Large Synoptic Survey Telescope (LSST), a telescope to be built in Chile to survey the entire southern sky, roughly once every four nights. Albeit without quite the clarity of the Hubble images, this telescope will reach the depths of the original Hubble Deep Field over an area 25 million times larger. As a project, it has much of the attraction of the HDF and CANDELS -- it is aimed at gathering a representative sample of the universe (in this case a very large representative sample!), to enable a huge variety of scientific investigations. And all the data will be public. With that and JWST operating at the same time, there is a lot to look forward to.

The Trials and Tribulations of Finding the Most Distant Galaxy Clusters

OK, now for something a little different. In this post I'm not going to talk about stuff we already know or get right, but the stuff we think we may be getting wrong. That may seem a little negative, but is in fact what astronomers spend much of their time doing. We constantly check, question, argue, ponder … it's only when we've run out of ways to challenge an idea or result that we then tend to think we may have just learnt something.

We've heard before about galaxy environment, and more specifically, the relative importance of galactic nature vs. nurture. Measuring environment (or nurture) has become increasingly popular in recent years, mainly because it's become increasingly possible to do.

In particular, large space-bound telescopes, like Hubble, and the galaxy surveys they produce, like CANDELS, have been especially important in the search for, and classification of, rare yet impressive cosmic structures called galaxy clusters. This has sparked a kind of friendly competition within the community to find increasingly distant clusters in the early Universe. For astronomers, it's a bit like searching for the earliest fossilised pre-humans: a trophy when you can beat the previous record, and (more importantly) juicy new clues to early galaxy life within the cosmos.

An early forming galaxy cluster, but is this a real over-density or
artificial? Credit: Papovich et al. 2010

A perfect example is the 2008 discovery of a early forming cluster at redshift 1.6 (about 9 billion light years distant) by CANDELS team member Casey Papovich and collaborators. This cluster was first identified using the Spitzer Space Telescope in the infrared, and the galaxies within were recently studied using CANDELS data to measure their structural properties, types, masses and colours. However, to date, only 13 galaxies in the cluster have been confirmed to be at the distance of the cluster centre to any precision.

This highlights a problem: when clusters become distant and hard to find, how do we know if we've actually found one? Let me propose that you have to answer three challenges before you get to pop the champagne cork on a new cluster distance record:

1. Demonstrate that the galaxies making up the cluster aren't just a chance alignment of unassociated galaxies along the line-of-sight. When you only have very rough distances, which is often the case (called photometric redshifts in astronomy-speak), such alignments can be a very common occurrence, and tricky to recognise.
2. Show that the "reality" of the cluster doesn't depend on the method used to find it. There are many environment "metrics" employed by astronomers, typically based on counting galaxies projected on the sky within a fixed radius, or by an association of nearest neighbours. But the measurement techniques are varied, the uncertainties large, and the agreement between them approximate at best.
3. Even when you've found a genuine cluster, how do you actually know what it will evolve in to? Simulations have shown that some structures grow fast and some slow, and there's no way to know in advance. It may be that the future for this particular object is rather boring and not the "Australopithecus" you were expecting.

At Swinburne, my graduate student Genevieve Shattow and I have been looking into these issues using CANDELS-based simulations of galaxy and cluster formation. These simulations are useful because we have "perfect" information for every modelled galaxy, unlike the real observations. By perfect I mean that every property of each dark matter halo and galaxy are precisely described (because we made them!), and the history of each is fully identified. We can use this information to our advantage.

With this simulation we looked at the most massive cluster at redshift 2 (i.e. when the Universe was 1/3 its present age, and where the current distance record is) and asked:

• How accurately can we identify the cluster if we degrade our information to be comparable to that of real observations?
• When we apply different environment measurement techniques, do they paint a consistent picture? Or do they return different answers?
• What will this object evolve into? Is it actually interesting? Is there a way to separate the interesting objects in the distant Universe from the more mundane ones without knowing in advance?

We're still working on these problems, but let me show you a few early, yet slightly disturbing results.

The above figure shows how galaxies in the larger environment around this simulated cluster appear on the sky if observed at four different times, ranging from the present day (far left) to redshift 2 (far right). We've coloured the galaxies: purple shows the actual gravitationally bound cluster members, while black are the non-cluster galaxies. The circle represents the typical radius astronomers would count within to look for over-densities, signalling the presence of a cluster.

As you can see, finding a cluster in the present-day Universe is fairly easy as there are many galaxies around to characterise it. In the early Universe though it's not so clear. Even worse, the ratio of cluster to non-cluster galaxies tips in an uncomfortable direction, making it hard to distinguish the actual cluster members from the background of field galaxies.

Different viewing angles of the same
cluster return different cluster
membership due to projection effects

We can play an even more illuminating game with the simulation. Let's again look at this single cluster and its surroundings at redshift 2, again projected onto the sky, but now from many different angles. Since galaxies in three dimensions will project differently on to the two dimensional sky depending on the viewing angle, the measured environmental over-density within the circle will vary. But by how much?

The distribution of counted galaxies based solely on changing the viewing angle is shown in the figure on the right. As can be seen, the environment measure may find 60 cluster members, or it may find 80 cluster members. That's a big difference for what is in essence exactly the same object! Given that we can't correct for such projection effects in real observations, this highlights a substantial uncertainty in any such measurement.

This is just the beginning of our work, but important to quantify if we're to have an honest accounting of the issues around measuring galaxy environment in real data. In doing this we'll be looking for ways to correct such biases, hopefully making the job of cluster identification easier and more accurate. This will enhance the value of data from galaxy surveys such as CANDELS and the science that follows from them.

How Far Away is this Galaxy?

In previous posts you have read a lot about redshift and the distance of galaxies from us. You have read how to find and select very distant galaxies and how to distinguish between different types of galaxies from their spectral energy distributions. What you haven’t read in more detail is how astronomers determine a galaxy’s distance or redshift. I’ll tell you a bit more about this now.

The electromagnetic spectrum: wavelengths get smaller from the radio to gamma rays.
A wavelength is here the distance between two peaks of the red wave. Credit: NASA

First of all, let’s recap again what redshift means. I guess all of you have heard an ambulance with its siren switched on passing by. You might have noticed that the pitch of the siren changes as the ambulance moves. In particular it sounds higher when it approaches you and deeper when it drives away from you. (The same is true for the sound of passing by race cars which make the popular "brrrmmmmm" sound kids love so much.) This effect is called the Doppler Effect. While the ambulance moves, the sound waves coming from the siren are squeezed when it drives in our direction and stretched when it moves away from us, resulting in a different sound.

The same happens to the light of distant galaxies. The Universe is expanding and so distant galaxies are moving away from us. The light emitted by these galaxies is stretched out on its way to us. If the galaxies were moving towards us, their light would be squished. Because light can be described as a wave just like sound, this stretch/squish means that its wavelength is changing. Blue light has a shorter wavelength than red light, so light that is blue when it is emitted turns redder on its way to us when the galaxy is moving away, hence we say it is redshifted. The light from distant galaxies takes a long time to reach us. The further away the galaxy is the longer it takes, because light travels with a particular speed and nothing is quicker than light. So if we know the redshift of a galaxy, we know its distance to us and with this we know how long it took the light we see today to get to us. Hence, we see distant galaxies not as they are right now (this light is just emitted and hasn't reached us yet), but as they were a long time ago. With a particular redshift astronomers can assign a so-called look-back time which is the time the light of a galaxy at this redshift took to reach us. For example, the light of the sun takes 8 minutes to reach Earth, but it takes about 5 hours to reach Pluto. The light of distant galaxies needs billions of years to reach us which means they existed when the Universe was very young.

   redshift                  look-back time                   age of the Universe

   0                                         0                                      13.7 billion years

   0.0043                      59 million years                         13.6 billion years

   0.01                        140 million years                         13.5 billion years

   0.1                           1.3 billion years                          12.4 billion years

   0.5                              5 billion years                            8.6 billion years

   1.0                           7.7 billion years                            5.9 billion years

   2.0                         10.3 billion years                            3.3 billion years

   3.0                          11.5 billion years                           2.2 billion years

   4.0                          12.1 billion years                           1.6 billion years

   5.0                          12.5 billion years                           1.2 billion years

   7.0                          12.9 billion years                          800 million years

   10                           13.2 billion years                          480 million years

I've listed the look-back times and ages of the Universe for some redshifts above (for the very cosmologically interested: using a Hubble constant of 71 km/s/Mpc and Omega Matter, describing the density of matter in the Universe, of 0.27 assuming a flat Universe, but you can play yourself with these values over here). Redshift 0.0043 is the redshift of the closest large galaxy cluster to us. Up until redshift 0.5 the galaxy population is pretty well mapped out and studied by surveys such as SDSS. The range between redshift 1 and 3 is interesting to researchers to study how galaxies assemble their mass as we told you in this previous post. Redshift 7 is close to the redshift of the most distant galaxy known to date. With CANDELS we hope to detect many galaxies at this redshift and even further away. We hope to discover galaxies at redshift 10 someday, which were perhaps the first galaxies to form in the early Universe.

Redshift Determination

Example of emission lines in a galaxy spectrum. Credit: Jeyhan Kartaltepe

There are two ways to determine the redshift or distance of a galaxy, either from the spectrum of that galaxy or from its spectral energy distribution (SED) and colours. If astronomers have spectra available, they look for particular features in each spectrum, such as emission lines (those spikes in the figure to the right) or absorption lines. Emission and absorption lines are caused by elements in the galaxy that emit or absorb the light of that galaxy’s stars at particular wavelengths. From lab experiments we know at which wavelength the lines of various elements should be when not moving, we call this the rest-frame wavelength. We also know what their separation in a spectrum is. In order to determine the redshift of a distant galaxy one can look for the spectral lines caused by elements in that galaxy. Since the elements are located within that distant galaxy, the wavelengths of the spectral lines are redshifted by the same amount as all other light coming from the galaxy. We say these observed spectral lines are in observed-frame. The difference between the rest-frame and observed-frame wavelength of a particular spectral line lets you calculate the redshift of the galaxy. The difficulty with getting redshifts from spectra lies in the identification of the spectral lines, you need to be sure that the line you are seeing comes from the element that you think it is. It is easier if spectra have several lines that can be used to determine the redshift because then you can identify the pattern of lines, and know without a doubt what the observed lines are. A redshift determined from a spectrum is called 'spectroscopic redshift'.

SEDs of typical galaxies at different redshifts. The solid curves show the SED in
the rest-frame or at z=0, meaning how it looks when the galaxy is very close to us. The
dotted and dashed lines show the same SED at redshifts z=1 and z=2, meaning the
light from such galaxies traveled about 8 billion and about 10 billion years to reach us!!
The bottom panel shows various filter bands from U (blue curve, visibly blue light) to
infrared bands (black curves), the filter bands are the same ones we showed you in
the SED post. Galaxy templates are taken from Polletta et al.

Taking spectra of many galaxies is very time-consuming and requires a lot of telescope time, which can be difficult to get. Instead, with imaging, you can observe thousands of galaxies at once and construct an SED by using multiple filters. In this case you can still determine the redshift of a galaxy by using SED-fitting. Astronomers take advantage of the fact that the SEDs of galaxies have particular shapes and features at specific wavelengths. For example the Lyman-break that is used to find distant galaxies and about which Russell told you in his post is always located at the same rest-frame wavelength and shifts to redder wavelengths the further away the galaxy is. Another important feature is the so-called 4000 Angstrom break. The 4000 Angstrom break (at 0.4 micron in the figure to the left) occurs because many heavy elements in the atmospheres of old stars in the galaxy absorb some of the starlight around this particular wavelength. The brightness of a galaxy changes noticeably from the left to the right of this break, so in one filter band the galaxy will be brighter and in the other one it will be fainter. A comparison with model galaxies and the location of these jumps in brightness allows us to estimate the redshift of the galaxy. The better the shape of the SED and features such as the mentioned spectral breaks are sampled with the images, the better we can determine the redshift. A redshift determined from galaxy images where no spectrum is available is called a 'photometric redshift'.

For large galaxy surveys, the latter method is used to measure redshifts for many objects at the same time. Although the uncertainties for photometric redshifts are larger than those for spectroscopic redshifts and photometric redshifts can be wrong for single objects, researchers gain better statistics with large sample sizes. Furthermore, the redshift distribution of that sample will be correct on average. While CANDELS itself is focused on imaging data, many spectroscopic surveys have been undertaken in these deep fields and new surveys are always ongoing to obtain more spectra. A lot of CANDELS science you will read about here will make use of photometric redshifts.