Q: Your PhD dealt with the ionization of hydrogen by O and B stars. Who was your adviser and what were your main results?

A: I did my PhD thesis with Lyman Spitzer, Jr. in 1975 at Princeton University, studying the pervasive interstellar ionization that was recently discovered by Ron Reynolds and the University of Wisconsin group. I was an undergraduate at UW and knew the people well. I learned about Strömgren spheres, diffuse clouds, and all the usual ISM topics, while hardly noticing the revolution that was just beginning with the discovery of CO at nearby Bell Labs. It turned out that applications of ionization theory to molecular clouds would prove useful when Charlie Lada walked into my office several months later during my Junior Fellowship at Harvard University.

Q: Shortly after came your 1977 paper with Charlie Lada on sequential star formation. What was the genesis of that paper?

A: Charlie had new CO observations of a bright-rimmed molecular cloud, which might be called a “pillar of creation” now, and wondered if there could be star formation inside. This was before infrared observations, and only a few groups, including the Harvard-Smithsonian Center for Astrophysics, had access to CO detectors.

Charlie had the novel idea for his PhD thesis a year earlier to point the CO telescope to the side of an HII region, unlike others who were searching for CO inside HII regions, which worked well for Orion. This led him to discover the CO core that was located to the side of M17. So we talked about what an HII region might do to a cloud on the side. This differed from the Strömgren sphere situation where an ionization source is in the center of a gas region, and it had not been considered much aside from Spitzer and Oort’s rocket effect paper in 1955. With Charlie, I thought about the gas that gets swept up into a dense neutral layer between the HII region and the cloud. I was surprised to see that the delay time before this swept-up layer collapsed gravitationally was about the same as the collapse time of the untouched part of the cloud ahead of the compression. This seemed important to us as the result was independent of everything. It seemed to say that compressed regions should be forming stars all the time. Also it gave the right morphology for star formation at cloud edges, which was a surprise too and yet consistent with a growing number of observations — including the Orion region, which only happens to be pointing to us so we miss the side view. In all the cases we looked at, dense HII regions were not Strömgren spheres in the usual sense.

That was an exciting time for us. We went on to map the rest of the M17 cloud at the McDonald Observatory of the University of Texas and found it to be far more massive than anyone had reason to expect (1976). It was the first truly “giant” molecular cloud.

Very few people took our triggering model seriously though. The paper’s referee said it would set back the field of star formation by 10 years. It actually took much longer for pervasive triggering to be recognized in the form of giant gas shells and star formation sequences of age, velocity, and cloud dispersal. At the time, though, and for Charlie and me, it was the beginning of the realization that the interstellar medium is a dynamic place with supersonic motions and rapid changes. Our ISM looked very different from the standard model where diffuse clouds were in pressure balance with a pervasive intercloud medium, and where cloud formation was slowly driven by a subsonic thermal instability. Our model fit better with the contemporary proposal by Chris McKee and Jerry Ostriker that ISM structures and motions are controlled by supernovae.

Q: You began working on galaxies with your wife Debra in the late 1970’s. How did that collaboration start?

A: Debra was an undergraduate at Princeton University when I was a graduate student there. She started graduate school at the University of California in Santa Cruz when I went to Harvard. We got married a year later and she transferred to Harvard. Debra learned from Bill Liller and John Huchra how to use Kodak IV-N plates to do near-infrared photography of galaxies. For her thesis, Debra took broadband images with these and other glass plates using telescopes at Mt. Palomar Observatory and Cerro Tololo Inter-American Observatory. After I went to Columbia University for an assistant professorship, she continued to work on her thesis in New York, modeling dust structures with radiative transfer solutions to determine their densities and masses. We thought we might see giant molecular clouds like the M17 cloud, which has $10^5\;M_\odot$ of molecules, but were amazed to find instead that the cloudy-looking objects along spiral arms contained more like $10^7\;M_\odot$ of gas. These dark clouds were also much bigger than GMCs. This led us to the idea that GMCs are the core regions of much larger and more primary clouds. We first called them superclouds at a brief talk for the 1979 IAU Symposium 87 in Mont Tremblant Canada. Also in this year I wrote a paper analogous to my work with Charlie suggesting that accumulated gas in the shock fronts of spiral density wave arms became gravitationally unstable and condensed into clouds with the local Jean mass, which is also $10^7\;M_\odot$.

A few years later when I was at IBM, Debra and I mapped HI clouds of this size in the first galactic quadrant using the Weaver-Williams HI survey and tabulated the star-forming regions and GMCs associated with them (1987). I recall working at home with a desktop computer and a telephone modem linked to a tape drive with the survey at IBM (the latest technology is always amazing). We wrote several other papers at this time on giant star-forming regions in the spiral arms of other galaxies, and I wrote several papers on how these regions might form. But I think, like the triggering model with Charlie Lada, few people took these ideas seriously. Everyone seemed interested only in GMCs, which were still a pretty new concept, and not their larger structures seen mostly in other galaxies, often in atomic form. Today, large-scale gas clouds associated with GMCs and clusters of GMCs are well studied in the Milky Way and other galaxies.

Q: You also did a lot of work on spiral structure with Debra. How did that begin?

Spiral structure in galaxies was an active topic in stellar dynamics theory in the 1960s and 1970s, and with Debra’s unique near-infrared images of several dozen spiral galaxies, which showed the underlying stellar distributions for the first time, she began to classify them based on similarities to density wave predictions. We discovered a large class of galaxies with prominent star formation in short spiral arms but very weak underlying spiral structure in the older stars. These “flocculent” galaxies had not received much attention before as the main theoretical interest was in 2-arm “grand-design” spirals that contained density waves. We considered flocculent arms to be the result of locally sheared gravitational instabilities in the gas, leading to star formation. A decade later, in 1996, I wrote a paper with Yuri Efremov suggesting that flocculent arms are the largest scale in a hierarchy of structures connected with interstellar turbulence, gaseous self-gravity, and star formation.

Q: As you just mentioned, in the 1990’s you began working with the Russian astronomer Yuri Efremov. How did that begin and what were the results?

A: I had read Yuri’s papers in the translated Russian journals and realized he was seeing coherent star formation that he called “star complexes” on the same large scales as Debra and I were seeing giant cloud complexes. I wrote to him several times and we finally met at a 1994 conference in Elba. Then, in 1996, Yuri visited our house and over dinner at a nearby restaurant, we decided to look at the hierarchy on even larger scales to see if we could tie it to flocculent spiral arms. The next day he recalled a strange pattern in the distribution of LMC Cepheid variables where the stars closer to each other had similar periods, which meant similar ages. Period is related to luminosity of course, which is related to the stellar mass and therefore the age of the star at the time when it passes through the Cepheid instability phase. We combined these ideas into the paper on flocculent structure mentioned above. We returned several years later to look for an analogous relationship among star clusters in the LMC, i.e., between their spatial separation and age difference. As for the Cepheids, clusters showed the same power law relation. This relation is what Richard Larson would have derived for his 1981 GMC correlations had he plotted cloud crossing time versus cloud size as a supplement to his famous size-linewidth relation. The power laws we were all finding suggested turbulence and self-gravity played a role in regulating star formation over a wide range of spatial and temporal scales.

Yuri and I also realized at this time (1997) that interstellar turbulence was the most likely cause for the mass distribution of all star clusters, including globular clusters, which have a different mass function today. And with Edith Falgarone in 1996, I proposed the same turbulent origin for the mass distribution of interstellar clouds. This all followed from the hierarchy of star-formation structures; i.e., big clouds contain little clouds in a self-similar pattern, and old regions on large scales contain sub-regions on small scales that come and go more quickly. In all cases, the timescale for star formation appeared to be just one or two crossing times on the relevant scale. This was much faster than previously realized and struck at the paradigm of stable clouds forming stars while the magnetic field slowly diffused away.

Q: You and Debra began looking at high-redshift galaxies in the mid-2000’s. What started that and how did it fit into your previous work?

A: When the Advanced Camera for Surveys was put on the Hubble Space Telescope, and a public image of the Tadpole interacting galaxies was presented (2002), Debra noticed in the background of these two galaxies a broad field of distant and peculiar galaxies. She decided to classify these peculiar galaxies according to physics-based structural properties. It had been known for several years already that more distant galaxies are more irregular, but some of these iregularities looked physically impossible, such as the chain galaxies discovered by Len Cowie and his collaborators at the University of Hawaii. We discovered in 2004 with a large enough sample from the Tadpole background field that the precise irregularity was a clumpiness to star formation, and that chain galaxies were only the edge-on versions of generally clumpy disks. It was then a small leap from our previous work to say that the clumps were normal star-forming regions, made larger and more massive than normal by higher turbulent speeds (2004). These large turbulent speeds were soon discovered by Reinhard Genzel, Natascha Förster-Schreiber and others at the Max-Planck-Institut für Extraterrestrische Physik in Munich, and also by other groups, while Genzel’s group and others were starting to notice the giant clumps too. Within a few years, and aided by new deep surveys and novel simulations by Frederic Bournaud, Debra and I found that star-forming regions at high redshift were larger, brighter, and more dynamically important than star complexes and even superclouds in local spiral galaxies. The high-redshift clumps were apparently massive enough to produce strong torques that could drive the evolution of structure in the whole galaxy. Debra and I also measured photometry of edge-on young galaxies and found them to be as thick as the old thick disks of today’s galaxies. This observation fit the model well because of the high turbulent speeds needed to make the big clumps through gaseous self-gravity.

Q: Another of your key collaborators was Deidre Hunter, who studies dwarf galaxies at Lowell Observatory. How did that begin and what were your contributions?

A: Deidre was an early observer of faint local dwarf irregular galaxies and did important papers with Jay Gallagher in the 1980s and 1990s. These galaxies have prominent star formation for their size, relatively big HII regions, and high gas fractions. We began collaborating in the late 1990’s, and she eventually started a big program of Jansky Very Large Array observations of HI to supplement her extensive ground-based observations. This was the LITTLE THINGS survey (2012). After we joined with Monica Rubio from the University of Chile in Santiago, we observed CO in a low-metallicity dwarf (2013, 2015) and could begin to see star formation and turbulence in dwarfs as a useful contrast to what others were studying in spirals. For example, the CO in dwarfs is highly confined to the inner regions of giant cloud complexes.

With Deidre’s continued observations to deeper and deeper brightness levels, we found star formation at the most extreme conditions, where the average gas surface density is $\sim1\;M_\odot$ pc$^{-2}$, or 10 times lower than the conventional threshold to make GMCs, and at rates down to $10^{-5}\;M_\odot$ pc$^{-2}$ Myr$^{-1}$, which is only 1 Orion-like cluster per square kiloparsec in 500 million years. Our most recent work attempts to tie this in with star formation in spiral galaxies, as we find that the star-forming clouds and processes themselves are about the same in the two cases, with primarily the lack of CO emission from low metallicity affecting how the clouds look in dwarf Irregulars.

Q: You have discussed your long-time interest in galactic-scale star formation, which contrasts with the small-scale observations and theory covered by most of this newsletter. Given the enormous difference in resolution between local regions and extragalactic star formation, how do you think the two fields can be tied together?

A: Galactic scale processes are needed to understand the beginnings of star formation and how dense clouds form. Galactic diversity also shows differences related to environment. It turned out over these years that star formation begins on much larger scales than GMCs and OB associations, and with processes that are strong and perhaps even dominant only on large scales, such as interstellar gravity. These processes also seem to have some degree of self-regulation so that the average galaxy star formation rate remains within a finite range for the galaxy’s mass. Also, star formation is not determined entirely by local processes such as turbulence and GMC gravity, or even disk-scale processes such as spiral waves or ISM gravity, but also by the galactic environment, including gas accretion and interactions. Most likely, there are two or more regulatory processes, including small-scale feedback from massive stars that break apart GMCs and trigger a little more star formation on the way, and large-scale feedback from stellar dynamics, which helps regulate the overall stability of the disk. This is the proverbial forest we have to see through the trees.

Q: You have had an unusual career path as a lone astronomer at IBM Research. How did that happen and what did you do there?

A: IBM’s main research lab in Yorktown Heights, New York, was a great place to work. I was there for almost 40 years after Columbia University. I went there at the beginning of the personal computer era when all technologies were growing exponentially, and also at the beginning of the internet era when I could collaborate easily with other astronomers from around the world. I had the privilege to watch all that growth as an insider and participate in some ways. I worked on a variety of topics, always alongside some expert, including magnetic and phase change materials, piezo-electronics, traffic flow simulations, AI for weather applications, and quantum computing for optimization. Thinking about and publishing research on these topics was very much like thinking about and publishing astrophysics, but some of it led to patents and the satisfaction that they might end up commercially useful.