Cosmic Clue: What's Killing the Massive Galaxies?

When I was a kid, my favorite board game was Clue. I loved being handed random, scattered pieces of evidence, and having to deduce through sheer logic and process of elimination, who, where, and how Mr. Boddy was killed. So it's not surprising that years later, as a galaxy formation theorist on the CANDELS team, I am drawn to a murder mystery on a cosmic scale, a mystery that is responsible for the most fundamental dichotomy in the galaxy population yet has puzzled astronomers since galaxies were first discovered: What kills the biggest galaxies in the Universe?

The existence of dead galaxies has been known since the time of Edwin Hubble. Dead galaxies are easy to spot as they are among the largest in the Universe, containing up to a hundred times more stars than the Milky Way. They are also very red, since they have not formed any new stars for a long time. This happens because stars in the sky are much like stars in Hollywood: The biggest ones are the hottest and shine the brightest, but they also die out the quickest (and most spectacularly). Since young, hot stars are blue (a consequence of their black body temperature), after they die out, the galaxy is left with nothing but plebian longer-lived redder stars like our Sun. The galaxy then becomes red and dead.

But why? What serial killer is responsible for systematically killing these erstwhile happily star-forming systems?

Red and dead galaxies, being large and bright, are well studied in the nearby Universe. They tend to be elliptical in shape, devoid of any cold gas, and reside in the densest regions of the cosmos such as groups and clusters of galaxies. These are valuable clues, but today's red and dead galaxies died long ago. In a "cold case", it's always tough to sort out the cause from the effects. Does losing a galaxy's gas cause it to become an elliptical? Or does becoming an elliptical cause it to lose its gas? And why does all this happen preferentially to the most massive galaxies, living in the densest regions? There are many clues, but no clear answer.

Elliptical galaxies often have strong X-ray emission associated with gas at many millions of
degrees. These images show a sample of elliptical galaxies, with the optical image showing
the stars on the right, and the X-ray image showing the hot gas on the left. The X-ray gas
often shows a lot of structure, indicating that it has been disturbed putatively by jets
from the galaxies' supermassive black holes. From the Chandra image archive.

For a long time, the answer seemed fairly obvious: Galaxies begin with a reservoir of cold gas, and once they use that up, they can't form any new stars, and they die. But upon closer scrutiny, this explanation doesn't hold water (or in this case, gas). Firstly, non star-forming galaxies form a tight red sequence in color-magnitude space, distinct from star-forming galaxies in the blue cloud, and the region in between (known as the green valley) is conspicuously devoid of galaxies. This means that whatever turns galaxies red happens quickly; if it was gradual, there would be a continuous distribution in color towards the red sequence, with no green valley gap. Hence galaxies don't die of "natural causes" by just gradually running out of gas. They are actively being murdered.

Even more damning for the "running out of gas" explanation is that red and dead galaxies actually are surrounded by a halo of hot gas, typically at a few million degrees. We can see this gas via its X-ray emission with telescopes such as the Chandra X-ray Observatory. And here's the rub: This gas should be cooling! By any reasonable estimate, the energy loss rate of this gas implies that red and dead galaxies should be acquiring tens to hundreds of solar masses per year of fresh cold gas to fuel star formation. But this is clearly not happening -- we see no cold gas in these galaxies, and no star formation. Something is keeping this gas hot, and not allowing a dead galaxy to revive itself.

So it appears that we may need two killers: One to quickly quench a galaxy's star formation by removing the cold gas, and another to prevent any new cold gas from falling on the galaxy. The plot thickens!

Do astronomers have any suspects? Well, one simple rule in astronomy, analogous to "follow the money" in a terrestrial investigation, is to "follow the energy". Removing cold gas from a galaxy takes a lot of energy. So does keeping an entire halo of hot gas hot. What agent has the ability to provide such enormous amounts of energy?

When viewed this way, the list of suspects narrows dramatically, basically down to one single agent. It is an improbable agent, one already shrouded in mystery as the ultimate harbinger of doom in the Universe:  The Supermassive Black Hole.

It was only around 15 years ago that it was realized that most sizable galaxies contain a supermassive black hole in their center, with masses that can exceed a billion Suns. But the radius of a black hole is miniscule when compared to that of the galaxy. And the vast majority of black holes seem to be fairly inert, like the one in our Milky Way, with only a small fraction emitting any significant amount of energy in a so-called active galactic nucleus, or AGN, phase. So how can these black holes, as dark and malevolent as they may seem, be responsible for killing an entire galaxy whose mass is a thousand times larger? The poor, persecuted black hole pleads innocence!

Not so fast, says the prosecution. Black holes act like a cosmic drain, drawing in any unsuspecting mass near the galaxy's center to be devoured into its space-time singularity. But it turns out that black holes are sloppy eaters. A significant portion of the mass that approaches the black hole, by being accelerated close to the speed of light, is converted into pure energy, and is released back out into the galaxy instead of being swallowed. This follows Einstein's famous relation, the energy released is the mass times the speed of light squared. Since the speed of light is a large number, a little bit of mass going in can mean a lot of energy coming out, perhaps enough to kill a galaxy!

This movie shows a numerical simulation of how a galaxy merger might trigger an AGN that removes the gas from a galaxy. The gas distribution is shown for two spiral galaxies, color-coded by temperature. As the two galaxies collide owing to their mutual gravitational attraction, a black hole is fed, which injects energy into the surrounding gas and evaporates it away. Mergers can also transform the galaxy's morphology from spiral to elliptical, which can explain why dead galaxies are usually ellipticals. Movie by T. Di Matteo, V. Springel, and L. Hernquist from Nature, 433, 604 (2005).

Energy released from black holes is called AGN feedback, and it's been observed. It is seen to be strong during galaxy mergers, when the disorder of the collision results in a sort of feeding frenzy for the central black hole, which can eject large amounts of gas. It is also seen to happen intermittently in galaxy clusters, where the cluster gas shows bubbles in X-ray gas that astronomers suspect have been inflated by powerful jets from the central black hole that are only active about 10% of the time. Crude estimates suggest that the amount of energy released in these events could plausibly provide enough energy to kill a galaxy. Heuristic models that include these effects such as the semi-analytic models of CANDELS theorist Rachel Somerville are able to predict a population of red and dead galaxies mostly as observed, which is encouraging. So it seems to be a plausible scenario.

But plausibility wouldn't yield a conviction in a court of law, and it doesn't hold up in a court of science, either. The energetics are a necessary but not sufficient condition for black holes to kill galaxies. Closer examination reveals many puzzling aspects in this story, such as: How does the energy released by the black hole get distributed on such large scales? How does that energy know to go into exactly the gas needed to kill a galaxy, and not other gas? Why does the black hole start putting out all this murderous energy only once the galaxy is massive and living in a dense environment? Theorists have struggled to come up with well-justified answers to these questions, so the case against black holes remains full of, well, holes.

This is where CANDELS comes in. CANDELS will provide us many more clues than we had before.  For the first time, we will be able to probe back to when the first red and dead galaxies began to appear -- when the murders were fresh -- about 2-3 billion years after the Big Bang. Since red and dead galaxies are relatively rare, one needs a wide survey area to be able to find them, and since the first red and deads appeared long ago, one needs very deep imaging to see so far away. This combination of wide and deep is exactly what CANDELS is designed to provide. At the same time, CANDELS can also be used to track and identify black holes over much of cosmic time, with a little help from its survey friends. Using this data, we hope to directly associate the killing of galaxies with black holes, in essence try to catch the culprit red-handed.

But early results from CANDELS have only deepened the mystery. Work by Dale Kocevski and collaborators has shown that, at higher redshifts, the correlation between galaxy mergers and AGN is not nearly so clear as it is today. David Rosario led a study that showed that galaxies with AGN are not obviously distinguished from galaxies that don't, a puzzling result if AGN are supposed to be a harbinger of turning galaxies red. Jen Donley found that AGN at earlier epochs are increasingly surrounded by lots of obscuring gas and dust, which is odd if AGN are supposed to be removing all the gas. CANDELS has seen the first red and dead galaxies appear long ago, but they are strangely compact, unlike anything we see nearby, adding a new puzzling twist to the saga.

So far, we have yet to find any smoking gun evidence pinning the murder of galaxies on supermassive black holes, all while the black holes sit smugly smirking at our hard detective work.  Or perhaps we are in fact falsely accusing the poor black hole, and there is some other murderous agent responsible.  The end of this game of Clue still appears to be far away, which means that for CANDELS astronomers, the fun is just beginning!

How Do Galaxies Grow?

The beauty of CANDELS is that it allows us to study how galaxies grow over an unprecedented dynamic range in mass and time. A particular focus of CANDELS is studying Cosmic Noon, the epoch from redshifts around 4 to 1, when the universe was about 2 to 6 billion years old. During Cosmic Noon, the universe was forming stars at more than ten times its current rate, making it the most active period of galaxy growth in cosmic history. Clearly, if we want to know how galaxies grow, studying this epoch is pretty important.

One of most influential new observations that has emerged from this epoch is a remarkably simple and tight relation between a galaxy's star formation rate and it's stellar mass: The two are linearly related!  In other words, larger galaxies form stars faster. This relation, now referred to as the "galaxy main sequence", might not sound terribly surprising, but it gives us a crucially important clue: It implies that galaxy growth is a smooth process. This is far from obvious when you look at a CANDELS image, because the universe looks like it is filled with a bunch lumps that we call galaxies! Moreover, we see that galaxy mergers are much more common during Cosmic Noon than today. Mergers are known to temporarily boost star formation by large factors (called a starburst) without increasing the stellar mass (as) much.  If this was the dominant process for galaxy growth, the galaxy main sequence would not be tight, it would show a huge scatter! The fact that we see a tight relation can be used to show that merger-induced starbursts likely contribute only about 10% of star formation during Cosmic Noon.

Simulations of galaxy formation can help us interpret what we see in the real Universe. This movie from a simulation shows only the stars (although the simulation itself also includes gas and dark matter) -- bluer stars are younger, redder are older. The volume depicted is roughly 50 million light years on a side, with face-on (left) and edge-on (right) views. We see how galaxies start out small and young arranged along the filamentary Cosmic Web, and merge together into larger, older galaxies -- this is called hierarchical structure formation. If stars were all that were visible (like in CANDELS images), one might infer that galaxies grow mostly by merging. But simulations suggest that this is only the easily-visible tip of the iceberg, and that relatively smooth gas inflows are actually the driver of galaxy growth (see the movie below).

This tight galaxy main sequence has a surprising corollary: If we only see the "lumps", while galaxies grow smoothly, this must mean that we don't see most of the matter that drives galaxy growth!  So where is this vast, unseen reservoir of fuel? Most believe that it is gas in the intergalactic medium, the vast expanses of space surrounding galaxies. This gas is very difficult to detect (which is why the space between galaxies looks so "empty"), but there is lots of observational evidence from the so-called Lyman alpha forest that the vast majority of atoms in the Universe live in the intergalactic medium, not in galaxies.  

Despite theoretical expectations, nobody has conclusively observed this intergalactic gas "in the act" of falling into galaxies, so at present this fueling mode remains a bit of theorist's fancy. Indeed, this is a major focus for next-generation telescopes such as JWST, GMT, and TMT, whose increased sensitivity may be able to directly detect inflow. Nonetheless, CANDELS is already providing indirect constraints by providing the best measurements to-date of the galaxy main sequence.

This simulation movie shows the gas rather than the stars. The large left panel shows the velocity field, in which one can see gas streaming down three converging filaments into a galaxy. The volume shown is roughly 7 million light years on a side. The "Density" panel shows the gas density -- little bright knots are where galaxies would form, and one would see stars, but the remainder of the gas is very difficult to detect. Other panels show gas temperature, metallicity, and metal-line absorption. The Cosmic Web is full of gas, providing relatively smooth fuel for galaxy growth. At around redshift z=3, one can even see an outflow that ejects gas from the galaxy (which is at the center of the panel) perpendicular to the inflow filaments.

A second remarkable new addition to the landscape of galaxy formation is that matter doesn't just flow in to galaxies, it also flows out. In fact, a lot of it flows out -- observations during Cosmic Noon suggest that the rate of mass flowing out from galaxies likely exceeds the rate forming into stars! Expelling this much gas at hundreds of km/s as observed requires enormous amounts of power, comparable to the power from all the supernovae going off in the galaxy. This seems rather surprising because galaxies appear to be rather serene, isolated object, as opposed to roiling pots of explosive energy that drive huge galactic outflows. But again, that may just be because we can't see the gas!

These new observational ingredients seen in galaxies during Cosmic Noon have caused CANDELS theorists to rethink some fundamental ideas about galaxy formation. How do we put all these insights together into a full story for how galaxies grow? One emergent paradigm is that galaxy growth proceeds via a balance between inflows and outflows. Material is brought in by gravity from the intergalactic medium, some of it gets turned into stars, and the rest is expelled in an outflow. This process is continuous and ongoing, resulting in a fairly smooth rate of growth. This is now colloquially referred to as the baryon cycle.

A helpful analogy for this is a factory. What does a factory do? It takes in raw materials, produces a product, and emits some waste. It does so fairly continuously, modulo small fluctuations in supply and demand. Analogously, a galaxy takes in gas accreted from the intergalactic medium, turns them into stars, and ejects outflows. This fluctuates as the inflow varies (after all, the universe does have some lumps!), but overall is a fairly steady. Hence in the baryon cycle view, galaxies are like gas processing factories, continually churning out stars from gas, at a rate that is mostly smooth with mild variations.

The outflows then represent the factory's waste. In fact, astronomers even use the word "pollution" to describe this ejecta. But in this case, astronomers mean something very specific -- pollution here refers to the fact that outflows carry out the by-products of star formation, which are the heavy elements ("metals") synthesized in the cores of stars. You  have probably heard that we are all made of stardust, that carbon, oxygen, and indeed all elements other than hydrogen and helium are produced by nuclear fusion inside stars. When stars explode in supernovae that drive these outflows, they pollute the outflowing material with these metals. The metals therefore provide an excellent tracer for where outflows have reached. Remarkably, quasar absorption line observations indicate that they have reached millions of light years away from galaxies!

The baryon cycle scenario can straightforwardly explain why the cosmic star formation rate is higher by a factor of 10 during Cosmic Noon compared to today -- it's because the universe was smaller and denser, the hence the expected gravitational inflow rate is higher by exactly that amount back then! It straightforwardly explains the tight relation between star formation rate and stellar mass, with the scatter around the mean trend reflecting the "lumpiness" of accretion. It is attractive in its simplicity and is well-situated within hierarchical structure formation models.

But none of that necessarily makes the baryon cycle scenario correct. Indeed, it is far from the only way to explain these data. The increased merger rate during Cosmic Noon suggests that the lumpiness plays a key role, and could drive the increased star formation rates. Moreover, galaxies around Cosmic Noon tend to look more disturbed, suggesting that even if the inflow is smooth over long timescales, on short timescales it may be quite chaotic and stochastic. Finally, the baryon cycle idea centrally invokes two phenomena -- inflows and outflows -- that have at best only recently been detected, and have yet to be well-characterized.  Hence while from a theoretical standpoint it is a simple and attractive model, from an observational standpoint there is much work to be done to test the baryon cycle scenario.

The unparalleled dynamic range of CANDELS, probing massive galaxies down to dwarfs at Cosmic Noon, will measure galaxy scaling relations like the main sequence with unprecedented accuracy, and relate it to many other interesting galaxy properties such as morphologies, colors, environment, and structural parameters. CANDELS theorists are working hard to take these observations and interpret them to see if they are consistent with the baryon cycle scenario. As scientists always love to subvert the dominant paradigm, observers and theorists alike are looking for ways in which the baryon cycle fails, or at least is an incomplete description of what's going on (which it almost certainly is). Through such investigations, we hope to gain a more detailed understanding how these gas factories process their raw material into stars during the most active phase of universe, Cosmic Noon.

The Dawn of Galaxies

One of the most exciting areas of astrophysics today is understanding how the very first stars and galaxies lit up the Universe.  This happened during the Epoch of Reionization (EoR), which was highlighted in the latest Astronomy Decadal Survey report New Worlds, New Horizons as the area of astronomy with the greatest discover potential in the next decade.  It is a central goal of CANDELS to probe galaxies in the EoR.

The Universe began in a hot Big Bang.  Early on, it was too hot for the protons and electrons to combine into atoms, because the high temperature made particles smash into each other too often.  So the Universe after about three minutes consisted of a fully ionized plasma -- that is, an admixture of positively-charged hydrogen and helium nuclei, and negatively-charged electrons.  Finally, after about 380,000 years, the Universe became cool enough that protons and electrons could bind together into neutral hydrogen atoms.  Thus began the Cosmic Dark Ages, so called because no sources of light were present, and all of the cosmos was enshrouded in a fog of neutral hydrogen and helium gas.

The Dark Ages lasted until a few hundred million years after the Big Bang.  It was then that the very first sources of light appeared, providing energy that ate away at the neutral hydrogen fog.  And so the Universe became ionized again, slowly and inhomogeneously, with electrons and protons being separated by energetic photons emitted by the earliest stars and galaxies.  We call this process cosmic re-ionization.  The Epoch of Reionization lasted until about one billion years after the Big Bang, and left the bulk of the Universe fully ionized and transparent as we see it today.  The EoR is the last major phase transition that the Universe undergoes, and it is the frontier of galaxy evolution studies today.

CANDELS has already detect a few EoR galaxies, which is pretty exciting in of itself.  But that's only part of the job.  What we really want is to understand what these galaxies look like, how they got there, and what they imply for the process of reionization. This is a job for the CANDELS theory crew.

So what do we want to figure out?

The biggest questions here are among the most basic:
a) What are the sources responsible for reionization? and
b) What is the topology of reionization?

While we know what galaxies look like today, there are good reasons to think that the first galaxies responsible for reionization might have looked quite different.  For one thing, like in a new house, there hasn't been much time for dust to accumulate. This is critical, because it turns out that star formation as we know it today requires dust as a catalyst. So how can stars form in the first galaxies with little or no dust?

The answer is: very slowly. But the stars that do form can be incredibly massive -- perhaps hundreds of times heavier than the Sun! These first stars, known as Population III stars, are copious emitters of ionizing radiation that can eat away at the cold fog of neutral hydrogen. On the flip side, because they form slowly, they are rare, so it's unknown whether there will be enough of them to power reionization.  Current thinking says probably not, but don't bet your first-born on it.

  This animation of a simulation from John Wise shows heavy elements (yellow) surrounding Population III stars after they have exploded.  

Moreover, these massive Population III stars have a short life, exploding in spectacular hypernovae after just a few million years. The details of these explosions are crucial:  Heavy elements like carbon, oxygen, and silicon are catalyzed in copious amounts during their short lives, and the hypernova could disperse them widely to form dust that quickly transitions star formation to the more familiar Population II (dust-catalyzed) mode. On the other hand, if the star collapses directly to a massive black hole, it would suck most of these heavy elements into oblivion, and the Population III era would continue for longer. Since Population III stars are no longer around today, it is difficult to see how they work in detail, and insights from models are often all we have to go on.

Even after the Population III epoch ends, it remains unclear whether there are enough stars to power re-ionization. CANDELS, as impressive as it is, only allows us to view the brightest of reionization-epoch galaxies -- Hubble cannot directly detect the fainter galaxies (this is what JWST will do). If there are not enough galaxies to provide the reionizing photons needed, it may indicate that there are more exotic contributors such as early black holes or an unexpected preponderance of Population III stars. Yet many models indicate that these faint galaxies are so numerous, that they actually dominate the radiation output! Clearly, tallying the total photon budget from CANDELS galaxy counts remains a poorly constrained yet critical aspect for understanding the sources of re-ionization.

This movie of a simulation from Tiziana di Matteo shows an evolving cube of the cosmos as sources begin to ionize the surrounding gas, eventually leaving a transparent Universe after about 1 billion years. Note the complex topology of filaments and sheets that houses early galaxy formation; this is known as the Cosmic Web.

As if those uncertainties aren't enough, there is the issue of topology. Topology refers to the spatial distribution, in this case of the protons and electrons.  While radiation from early galaxies can ionize hydrogen, the Universe is still sufficiently dense that the dissociated protons and electrons can quickly re-join back into hydrogen. This is a process known as recombination.

While recombinations are (cosmically) rare today, nature has perversely arranged the timescales for reionization and recombination to be annoyingly comparable during the EoR. This means one has to understand the spatial clustering or topology of protons and electrons, in order to know how often after being so cruelly separated, they will bump into each other again and re-discover their lost electrochemical bonds of love. This can be quantified by the clumping factor of protons and electrons.  If the clumping factor is high, it requires many photons to ionize a single atom, since protons and electrons remain close enough to recombine again after being ionized. If clumping is low, a single photon might be enough to keep an atom ionized. Hence we not only have to count how many photons are being emitted, but we also have to understand matter clumping in order to know how effective each photon is at reionizing the cosmic fog.

With all this uncertain physics flying around, it's not surprising that the EoR represents one of the most difficult modeling problems in astronomy today. Our most sophisticated simulations include all the complex processes we use to model galaxy formation at later epochs, plus the dispersal of heavy elements via outflows along with radiative transfer -- the emission and propagation of photons from cosmic sources. This last aspect is particularly challenging, requiring massive supercomputers to move not only mass but light around the simulated cosmos.

CANDELS theorists have developed a remarkable simulation code, called MARCH, capable of handling all these physical effects with essentially no simplifying approximations. Using MARCH, we have been able to show that the clumping factor is around 3, in contrast to earlier estimates of 10-30, and that CANDELS is directly detecting the sources that provide about one-quarter of the photons needed for reionization. While these results are encouraging, there remain many uncertainties in such calculations, particularly the escape fraction, i.e. the number of ionizing photons that escape from within galaxies. There is a long way to go before we can confidently model all the processes going on during the EoR.

Nonetheless, the EoR remains one of the most vibrant and revolutionary areas of study in the CANDELS team. CANDELS data provides the boundary conditions for theorists' models, while the models inform the interpretation of the observations. The recent CANDELS team meeting in Santa Cruz enabled the High-Redshift Working Group to assess where we stand now in both observations and theory, and how to best proceed in concert. Together, we are shining a new light on the Cosmic Dark Ages by peering boldly into the dawn of galaxies.

Making Sense of the Census

Walk into a mall or a sports stadium, and you'll likely see a diverse snapshot of the human population.  There's an enormous range of shapes, sizes, colors, and ages in such a census.  This diversity arises from some combination of birth properties, i.e. nature, and environmental factors, i.e. nurture.  Even though you are seeing only a fleeting moment in each of each person's life, you know that each human evolves through various stages of life from birth to death.  How that life unfolds depends on both nature and nurture, and the relative influence of each continues to be hotly debated.

CANDELS provides an analogous snapshot of the galaxy population. Like people, galaxies go through distinct stages of birth, growth, and death. And like people, the way in which galaxies progress through these stages depends on both nature (its intrinsic traits) and nurture (its environment). CANDELS is remarkable in that it is the first survey capable of directly observing galaxies in all phases of the life cycle, in a wide variety of environments, from the early Universe until today. This impressive data set is designed to provide us a much more complete census of the galaxy population, so we can answer fundamental questions such as the role of nature vs. nurture in establishing why galaxies look the way they do.

As impressive as CANDELS is, there is one intrinsic difficulty with studying the life of galaxies in contrast to the life of humans: One can watch humans change on human time scales, but galaxies evolve over far longer time scales. So CANDELS only provides that single census in the mall or the stadium; it cannot show what happened before or what will happen next to any individual galaxy.

So how are we to piece together the life story of galaxies from a single snapshot? How do we know which young galaxy will turn into which old galaxy? How can we figure out what causes galaxies to transition from one stage of life into another? How do we "connect the dots" into a full life story of galaxies? For humans, we know the answers because we've watched it happen, from birth to growth to life to death. But we have no such blueprint for the life of galaxies.

 

Video showing stars forming in a cosmological simulation from an early epoch until today, where the stars are color-coded by age (blue=young, red=old).  This is analogous to what might be seen in a CANDELS field at a given redshift.  Stars form in smaller galaxies and then merge together to former larger, older systems.  Note that there are two side-by-side projections of the same volume, face-on (left) and edge-on (right).  Movie by Ben Oppenheimer.

This is where the crack team of CANDELS galaxy formation theorists comes in.  Our job is to fit the entirety of the vast CANDELS data set into a single coherent story about why galaxies look the way they do, all within the context of our now well-established concordance cosmology framework.  Simply put, our job is to figure out a blueprint for the life of galaxies.  If that sounds just a tad bit ambitious, well, it is.  Not surprisingly, we haven't succeeded yet -- otherwise there wouldn't be much point in doing CANDELS!  Still, by carefully examining CANDELS data in the context of our current understanding of the Universe, we can make increasingly more educated guesses as to what makes galaxies tick.

How do theorists do this?  To begin with, we need a model for how galaxies form and evolve.  Within such a model, we can directly watch galaxies move through their lives, and try to understand what physical processes are responsible for setting a galaxy's properties at any given life phase.  If our model is successful at matching available observations from CANDELS and other data sets, we can hope (though not guarantee) that the insights we get are applicable to real galaxies, not just model ones.

 

Evolving fly-through of a hydrodynamic simulation showing the gas density, color-coded by temperature.  Note the Cosmic Web of filaments and sheets forming due to gravitational instability.  The nodes of the cosmic web are where galaxies that would be seen by CANDELS form.  The intersections of the filaments cause gas to shock-heat to high temperatures; by the end (redshift z=0), the video is centered on a galaxy group with gas heated to 10,000,000 K by gravity alone.  Movie by Ben Oppenheimer.

Step one is building a model for the entire galaxy population.  The technique I use is called a cosmological hydrodynamic simulation.  Big words, so let's break it down:  "Cosmological" means that I am trying to model a representative portion of the entire Universe.  "Hydrodynamic" means that I am interested in modeling the gas directly, including process such as shocks and radiative cooling (discussed below).  And "simulation" means that it's run on giant supercomputers, often taking months for a single run.  Such simulations are fast becoming a key cog in connecting the physics of galaxy formation to the galaxy observables seen in CANDELS data, because the evolving interplay between cosmology, gas physics, and galaxy formation processes are too complex to solve using pencil and paper.

These simulations begin with the (well-established) conditions shortly after the Big Bang -- a relatively smooth, hot Universe with dark matter and gas.  We typically model a cubical portion of the Universe by representing it with particles, each one representing a portion of cosmic mass.  Dark matter particles dominate by mass but interact only via gravity, while less massive gas particles additionally interact hydrodynamically.  The gas particles are initially comprised of hydrogen and helium, but as stars form they can additionally hold heavy elements (or "metals", in yet another misnomer of astro-lingo) such as carbon and oxygen.  Gas, unlike dark matter, has the important property that it can lose energy by emitting radiation, and thereby can release its gravitational potential energy and sink to the center of the dark matter halo.  This process, known as radiative cooling, enables gas to condense out of the Cosmic Web of dark matter into dense knots that form galaxies.  Once gas condenses into galaxies, things get tricky, since many poorly-understood processes govern the conversion of gas into luminous matter (i.e. stars).  We try our best to include all these processes in our models, but since we don't understand them in detail, we typically end up running lots of models with a range of parameters to see which one matches data most closely.

A surprising complexity is that matter doesn't just flow in to galaxies, it also flows out.  Galactic winds are now seen ubiquitously in rapidly star-forming galaxies, but what powers the expulsion of gas against the enormous gravitational pull of an entire galaxy remains a mystery.  Candidates include the collective effects of thousands of supernovae, or radiation pressure from hot young stars, or energy released by the monstrous black hole that lurks at the center of most every galaxy.  Once ejected, some material can fall back in, in a process we call wind recycling.  The net effect is a cycle of mass, metals, and energy flowing into an out of galaxies.  This means that galaxies and surrounding intergalactic gas form a sort of cosmic ecosystem, and simulations tell us that the way in which matter moves around in this cosmic ecosystem governs the way that galaxies evolve.

 

Close-up of a forming disk galaxy from early epochs to redshift 2 (i.e. "cosmic high noon").  Left panel shows the density, right shows gas color-coded by the temperature.  If you look closely in the right panel late in the movie, you'll see wind recycling happen: a hot plume of gas is ejected from the galaxy, some of which rains back down onto the thin disk.  Movie by Daniel Angles-Alcazar.

Does this mean "nurture" (i.e. surrounding environment) is more important than "nature" for galaxies?  Well, certainly the cosmic ecosystem is important, so a simple interpretation might lead one to say "yes".  But hang on -- it turns out that the properties of this baryon cycle seem to be surprisingly tightly correlated with the galaxy's stellar mass, which is an intrinsic quantity!  This would lean more towards the "nature" interpretation.  As with human behavioralists, the nature versus nurture debate rages on in the galaxy formation community.

The close connection between theory and data within the team is one of the most exciting aspects of the CANDELS survey.  CANDELS is clearly a huge step forward observationally, and it is an ever-increasing challenge for us theorists to keep pace.  In upcoming blog posts, I'll describe what we're learning about the birth, growth, and death phases of galaxies by combining state-of-the-art simulations with CANDELS data.  As we'll see, there are still far more questions than answers.  Yet it's clear that the galaxy population displays a beautiful and complex diversity, much like the human population.  For me, this makes galaxy formation an infinitely fascinating subject of study.