Astronomers have known for some 10 years that nearly every large galaxy contains at its core an immense black hole—an object having such intense gravity that even light cannot escape. The death of stars can produce small black holes—with masses ranging from about three to 100 times the mass of the sun—but such stellar-mass black holes are tiny compared with the behemoths at the centers of galaxies, measuring millions to billions of solar masses.

These supermassive black holes pose major puzzles: Why are they so common in galaxies? Which came first—the galaxy or the hole? And how did they form in the first place?

The mystery is intensified because supermassive black holes were already in place when the universe was very young. Last June, for instance, astronomers reported the earliest example detected so far—a hole of about two billion solar masses that existed 13 billion years ago, a mere 770 million years after the big bang. How could black holes get so big so quickly?

Such rapid formation is perplexing because although black holes have a reputation as mighty vacuum cleaners, they can also act like immense leaf blowers. Gas falling toward a black hole ends up swirling around the hole in a huge disk, the so-called accretion disk. The material heats up and emits radiation, particularly as it approaches the point of no return at the inner margins of the disk. The radiation pushes away other infalling material, limiting how fast the hole can ordinarily grow by accretion. Physicists calculate that a black hole sucking in surrounding matter continuously at its maximal rate would double its mass every 50 million years. That is too slow for a “seed” black hole of stellar mass to grow into a billion-sun monster in less than a billion years.

Astrophysicists have proposed two general ways for seed black holes to form. The first, considered for many years, assumes that the earliest huge black holes were indeed the remnants of stars. The first stars ever to have taken shape in the universe were likely to have been extremely massive compared with those that came later, such as our sun, because the primordial gas clouds were free of elements that help the gas to cool and form smaller clumps. These big stars would have burned out fast and produced black holes of perhaps 100 times the mass of the sun. Some process must then bulk up these holes faster than ordinary accretion. For example, if a big hole formed in a dense cluster of stars, it would end up near the cluster center, with other massive stars and black holes. It could then quickly grow to 10,000 solar masses by swallowing other holes and thereby beating the usual feeding limits. Further growth to supermassive scale could be by more ordinary accretion, perhaps with other largish holes on the menu as well.

Once astronomers knew that large supermassive black holes existed very early on, though, they began to wonder if stellar-mass holes could become supermassive quickly enough, even beginning life with this kind of accelerated growth. People started to look for alternative ways to produce seed black holes, routes that would generate bigger holes than those that could form in the death throes of stars.

Researchers proposed models that made bigger seeds by skipping the middlemen (stars). Rather a large cloud of gas would collapse to create a black hole directly, one larger than the product of a dead star. By making seeds with masses of 10,000 to 100,000 suns, this process somewhat alleviates the time crunch to form supermassive black holes at early times. This kind of direct collapse does not happen in our universe today, but conditions were different when the universe was young.

Unfortunately, it is hard to figure out which of these two scenarios took place—whether seed black holes started off small, as the products of dying stars, or on the contrary, started off larger, as the products of gas implosions. Although astronomers can peer far back in time by looking out to vast distances with telescopes, they cannot yet hope to detect seed holes in the act of forming; even the biggest seeds would be too small to be seen so far away. (The James Webb Space Telescope could reveal them, but it is not due for launch until 2018 and must survive political battles over its funding.) So my colleagues and I have been pursuing another strategy: looking for leftover seeds that have, for whatever reason, survived to the present day without growing supermassive.

If seed black holes started off as stars, we would expect to find many leftover seeds, in both the centers and outskirts of galaxies, because primordial stars could have died anywhere in the galaxy. We would also expect to find a continuous range of masses from 100 to 100,000 suns because their growth could get interrupted for lack of food at any stage along the way to supermassive status. In contrast, if seeds formed mainly by direct gas collapse, then the leftovers should be pretty rare; the direct collapse process, if it happened at all, would have occurred less often than ordinary star death. Instead of a wide range of masses, we would find most leftover seed black holes to be heavier than 100,000 suns (the theoretical models indicate that is likely the typical mass of the seeds that would form by direct gas collapse).

Other astronomers and I have therefore been scouring the skies for a new type of black hole, neither stellar in mass nor supermassive but somewhere in between: the so-called intermediate-mass, or middleweight, black hole. Our aim is to see if their prevalence and range of sizes are more consistent with the star collapse or gas collapse models. When we began this effort about a decade ago, it did not look promising. Astronomers knew of only one middleweight hole and considered it to be a fluke. Since then, though, we have found hundreds.

What counts as a “middleweight”? Here I will take it to mean a black hole with an estimated mass between 1,000 and two million suns. That upper limit is somewhat arbitrary, but it excludes the smallest well-known supermassives, such as the Milky Way’s four-million-sun hole. In any case, the boundary is inherently fuzzy. In practice, measurements of a hole’s mass often start out very uncertain—for instance, the masses of our first batch of middleweights all shifted up by about a factor of two a few years ago when we improved our measurement technique. The precise boundary does not matter as long as we study the entire population of holes extending down from the low supermassive range. What we have learned so far has already provided us a new view of the interactions between black holes and the galaxies that they live in.

Elusive Middleweights
Black holes can reveal themselves in a number of ways. For instance, stars whipping around orbits at the very center of a galaxy are a telltale sign of a lurking supermassive black hole. Middleweight holes, however, are too puny to give away their presence by their gravity in this way. Instead we focus on “active” black holes—ones that happen to be eating stuff—because the hot infalling material emits a tremendous amount of light.

Over decades of studies astronomers found that active black holes usually live in big galaxies of a certain kind. Galaxies, particularly massive ones, come in two general types. Some, like our own, have a large, rotating disk of stars. These disk galaxies look like dinner plates when seen edge-on. The other kind, elliptical galaxies, are basically balls of stars. Some disk galaxies actually have small elliptical galaxies at their centers, known as bulges. Active black holes are most commonly found in large elliptical galaxies and disk galaxies with healthy bulges. Nearly every bulge astronomers look at that is close enough to tell turns out to harbor a black hole of millions to billions of solar masses. Furthermore, bigger bulges have bigger holes—the black hole’s mass is usually about 1,000th of the bulge’s mass. This surprising correlation is a mystery in its own right, implying that galaxies and supermassive black holes evolve together in ways that astrophysicists have not yet fathomed. More prosaically, this pattern suggested where to look for middleweight holes: in the smallest galaxies. But which ones?

A very puzzling little galaxy offered one idea. My thesis adviser, Luis C. Ho of the Carnegie Observatories, studied about 500 of the nearest bright galaxies for his own thesis in 1995. He found that while most of the galaxies with big bulges contain active black holes, galaxies without bulges do not—with one interesting exception. NGC 4395 is a disk galaxy with an active black hole and no bulge at all. Ho’s own thesis advisers had noted this oddity as long ago as 1989, but most researchers considered it an anomaly. Except for NGC 4395, Ho’s survey confirmed the broader rule: black holes are not found in bulgeless galaxies.

Accurately estimating the mass of the hole in NGC 4395 is a challenge. The most direct mass measurements in astronomy involve measuring orbital motion. For instance, the speed of a planet and the size of its orbit around the sun let us calculate the sun’s mass. Similarly, the orbits of stars in a galaxy can reveal a black hole’s mass but only if it is large enough for the effects of its gravity to be discernible in astronomers’ observations of the star motions. The hole in NGC 4395 is too small.

Astronomers must therefore rely on less direct clues. For instance, x-rays coming from active black holes change in intensity over time, and the larger the black hole, the more slowly these variations occur. In 2003 David C. Shih and his colleagues, then at the University of Cambridge, found that the intensity of x-rays coming from NGC 4395 varies so quickly that it must be relatively small—most likely 10,000 to 100,000 solar masses. Ho arrived at the same rough mass range based on other evidence, also in 2003.

A slightly more direct measurement of the mass came in 2005, from Bradley M. Peterson of Ohio State University and his co-workers. They used the Hubble Space Telescope and a technique called reverberation mapping, which relies on gas clouds orbiting the hole similar to using planets orbiting the sun. The timing of echoes of light from the clouds provides the size of the orbits. Peterson and company concluded that the black hole is about 360,000 solar masses. Even with this technique, however, the mass has a large uncertainty—as much as a factor of three—because of assumptions that feed into the number crunching.

The bulgeless galaxy NGC 4395 appears to host just the kind of intermediate-mass black hole we were looking for. Yet of the 500 galaxies examined by Ho, it was the only bulgeless one with clear evidence for an active black hole. The second was found in 2002. Aaron J. Barth, then at the California Institute of Technology, used the Keck II telescope in Hawaii to take a spectrum of a peculiar but little-studied galaxy called POX 52. Like NGC 4395, this galaxy had shown some signs of an active hole even though it is not one of the usual suspects for harboring a supermassive black hole (it is a rare type known as a spheroidal, which is distinct from the bulged disk and elliptical galaxies).

Barth sent the new POX 52 spectrum to Ho, who took one look at it and immediately asked Barth, “Where did you find such a beautiful spectrum of NGC 4395?” The two objects’ spectra looked so similar that Ho could not tell them apart. (Features in the spectrum are what indicate the presence of a black hole.)

Because POX 52 is 300 million light-years distant (20 times farther away than NGC 4395), astronomers’ mass estimates for its black hole are considerably less direct. Still, a variety of evidence all indicates that the galaxy harbors a black hole of around 100,000 suns. Middleweight black holes in bulgeless galaxies now formed a class of two.

Of course, to solve the bigger problem of how the seeds of supermassive black holes formed, we needed more middleweight specimens to answer a lot of basic questions: How common are middleweight black holes? Does every bulgeless galaxy contain one, or are most such galaxies holeless? Do these middleweight holes occur anywhere else? And are there specimens even smaller than these first two waiting to be found? Only by answering these questions could we learn about how seed black holes formed and what role they played in the early universe.

Combing for Holes
Unfortunately, astronomers’ standard techniques are biased against finding active middleweight black holes. The larger the black hole, the more it can eat and the brighter it can shine. Smallish black holes are faint and therefore harder to find. But it gets worse. The elliptical galaxies where large black holes tend to occur are extremely well behaved. These galaxies do not have much gas and are not making new stars, leaving a clean and unobstructed view of the galaxy center. In contrast, disk-dominated galaxies (like where we suspected middleweight black holes might commonly lurk) are often forming stars, and the young starlight and associated gas and dust can hide the active hole.

To overcome these obstacles, in 2004 Ho and I turned to an invaluable library of data designed for finding needles in the cosmic haystack—the Sloan Digital Sky Survey. Since 2000 this project’s dedicated telescope in New Mexico has snapped images across more than a quarter of the sky and has recorded the spectra of millions of individual stars and galaxies.

We combed through 200,000 galaxy spectra and found 19 new candidates similar to NGC 4395—small galaxies that contained active black holes with masses we estimated at less than a million suns. Similar searches over the past few years, using more recent Sloan survey data, have expanded the total to about three dozen holes with masses under a million suns and more than 100 just over the million-sun threshold.

The method used to estimate these masses is relatively indirect. The Sloan light spectra tell us the speed of hot gas orbiting a hole. That is only half the information needed to compute the hole’s mass directly (the other half is orbit-size). Still, astronomers know from observing active holes in the million- to billion-sun range how the gas speed usually trends with the hole’s mass (smaller hole means slower gas). Extrapolating to somewhat lower-mass holes lets us pick out our little guys from the Sloan data.

These searches confirmed what we expected based on NGC 4395 and POX 52: a wider population of intermediate-mass black holes exists. Also in line with expectations, they are found preferentially in galaxies without bulges. Yet these holes still seem to be very rare. Only one in every 2,000 of the galaxies bright enough to study in the Sloan survey shows evidence for an active intermediate-mass black hole.

The Sloan searches, however, could be missing many black holes. They rely exclusively on optical light (the range of wavelengths our eyes can see), and dust clouds could well be hiding many black holes from sight. To get around this, astronomers are using wavelengths of light that can pass right through dust, such as x-ray, radio and midinfrared. Shobita Satyapal of George Mason University and her collaborators have been using mid­infrared light to look for signs of hidden active black holes in bulgeless galaxies. Extreme ultraviolet light coming from material plunging into an active hole would wreak havoc in the surrounding gas, creating unusual species, such as excited states of highly ionized neon. Emissions from these ions would leave characteristic fingerprints in the midinfrared spectra. Relatively few galaxies are amenable to this kind of search, and Satyapal’s team has found only a couple of new active middleweight black holes. Astronomers have also seen signs of possible middleweight or smallish supermassive black holes at x-ray and radio wavelengths, and follow-up observations to confirm these candidates continue.

These results indicate that the optical searches are indeed overlooking numerous bulgeless galaxies that hide their middleweight holes behind dust—but not enough to make mid­dleweight holes common. The verdict is still out, but perhaps only 5 to 25 percent of bulgeless galaxies harbor a middleweight hole big enough to detect.

Growing Galaxies and Holes
The observations of middleweight holes in bulgeless galaxies may help explain the connection between larger holes and big bulges. As I mentioned earlier, supermassive holes in massive bulged galaxies tend to be about 1,000th of the mass of the bulge. The growth of a supermassive black hole appears to be intimately linked with the growth of the surrounding bulge. If the correlations between black holes and galaxies are established during the formation of the bulge, then there should be no correlations between the properties of bulgeless galaxies and their middleweight black holes.

A leading theory to explain how this tight correlation comes about in bulged galaxies goes like this: Elliptical galaxies and large bulges form when disk galaxies merge. During the merger, gravitational forces stir up the disks, so the stars no longer orbit in a disk but move around randomly in a ball (the new elliptical or bulge shape). Gas clouds collide during the merger and are funneled toward the center of the bulge, triggering a major burst of star formation, which increases the total mass of stars in the bulge. At the same time, the black holes from each galaxy merge together and eat some of the new gas in the galaxy center. In this way, large bulges and supermassive black holes can grow and evolve together through these large-scale processes that occur in galaxy mergers. By the time the hole reaches about  1,000th of the bulge mass, its leaf-blower aspect comes to the fore, pushing the remaining gas out of the galaxy center and ending the growth spurt.

Middleweights in bulgeless galaxies, such as NGC 4395, would never have had the benefit of these organized feasts. Instead they would be leftover seeds that have grown only by more happenstance meals of gas at the galaxy center—snacks that are not connected with events shaping the overall evolution of the galaxy. Some galaxies without bulges may not grow a black hole at all. That is the case for the pure disk galaxy M33 (a galaxy much like NGC 4395 in physical appearance), which very clearly contains no black hole more massive than 1,500 suns. Evidence is mounting for this picture, linking black hole growth with bulge formation, but many details remain to be worked out, and the case is not completely settled.

On the question of how black hole seeds formed in the first place, the rarity of middleweight holes lends weight to the theory of direct collapse of gas clouds in the early universe. If star collapse accounted for the earliest seeds, we would expect almost all those galaxies to contain a black hole of at least 10,000 suns at their center. It seems, however, that most small bulgeless galaxies do not contain such a hole at their center.

Other evidence also points toward the direct collapse scenario. In particular, the weak correlation of the middleweights’ masses with their host galaxy masses more closely resembles that scenario’s predictions. And it is much easier to make a billion-sun hole in a few hundred million years if the seeds start out heavy.

Of course, as more data come in, the conclusions drawn so far could change. For instance, if astronomers were to look at galaxies slightly fainter than those with spectra in the Sloan survey, the fraction of galaxies with middleweight holes might rise or fall. And it is possible that some galaxies contain middleweight black holes outside of galactic centers. Indeed, the search for middleweight holes is continuing on many fronts, as is described in detail at

For now, many critical questions about middleweight black holes remain open. Are middleweight holes more common in specific types of small galaxies? (Such correlations might suggest new ways that holes and their host galaxies interact even before the merges that generate bulges and supermassive holes.) Do most bulgeless galaxies completely lack a middleweight hole, or do they have holes just slightly too small to be detected so far—perhaps in the range of 1,000 solar masses? (Such holes would surely have grown from remnants of dead stars and not formed by direct gas collapse.) Or do all bulgeless galaxies have hefty 10,000- to 100,000-sun holes, although most of them do not happen to be eating and spewing out x-rays and light? (That would change the conclusion that middleweights are rare.) The answers could push astrophysicists’ theories of how galaxies and black hole seeds first formed in radically different directions.