Your PhD from 1972 dealt with radio observations of massive molecular clouds. How did you choose this subject and who was your advisor?
My graduate advisor at Columbia in 1969 was Phil Solomon. He and Pat Thaddeus had become intrigued by the recent discovery in 1969 that the 6 cm H2CO (formaldehyde) transition was seen in absorption of the cosmic background radiation in nearby dark clouds. This implied that the excitation temperature of the levels was less than the background temperature of 2.7 K and the CMB temperature would be expected to be a lower limit to the excitation temperature. They were granted time on the Greenbank 140 ft telescope to follow this up.
As an aside, I should point out that the 140 ft telescope is the largest fully steerable EQUATORIAL mount telescope -- clearly an awkward structural configuration. We discovered that the reason for this was that when the telescope was built in ~1960 the engineers didn't trust that computers would be fast enough to do the equatorial to horizon-based coordinate transformation so they decided on an equatorial mount -- we've come a long way !!
Phil became my thesis advisor since I was most interested in the astrophysics of the molecular clouds rather than the chemistry, although Pat became a life-long friend. My thesis involved three parts: a survey of the H2CO absorption structures and CO emission in the central 3 deg of the galaxy, and modeling the distributions. Before our work I think the prevailing view had been that the molecular gas there was distributed as a continuous medium in extended galactic structures. We advanced the now accepted view that the molecular gas was mostly contained in self-gravitating clouds (such as those associated with Sgr A, SgrB2 and W51 (although these clouds resided in the larger scale galactic arms seen in HI).
One of the characteristics which intrigued me was the very large line widths (3 - 30 km/s ) of the molecular emission lines, corresponding to 10 - 100 times the thermal sound speed. In addition we found the line profiles of different molecules were quite similar, despite the fact that the different molecules had very different abundances and hence different optical depths. These facts led to our proposing the so-called LVG radiative transfer and excitation model in 1975. For the LVG treatment, I made use of a formalism developed by Leon Lucy who was one of my professors at Columbia. This was an example where personal interactions in science research fortuitously pays off.
In 1975 you and Phil Solomon published a large scale survey in the J=1-0 transition of CO of a large part of the Galactic plane, including the Galactic center. What were the key conclusions?
The first Galactic survey of CO which we conducted about the same time occurred in the simplest possible way -- during an observing run at the NRAO 12 m telescope on Kitt Peak. We started taking CO observations in the Galactic center and then realized that the emission didn't just stop there so we decided to keep going. The Galactic survey consisted of just 90 observations, one per degree of Galactic longitude out to l = 90 deg. I then measured (with a ruler !) off the hard copy spectra and translated the 'digitized survey' to the plot one can see in the published paper. The major conclusion from that paper with just 90 observations was that the overall distribution of molecular gas peaked the Galactic center, and then exhibited another peak or ring at ~5 kpc radius and fell off exponentially out to the solar circle. Based on this and follow-on work, it appeared that the mass of molecular gas dominated the atomic gas in the inner galaxy out to the solar circle; then in the outer galaxy HI dominates.
Of course there were many subsequent surveys by Thaddeus's group including Tom Dame in both hemispheres and higher resolution northern hemisphere data from the UMass-FCRAO 14 m telescope (Sanders, Clemens, Phil and myself). The basic qualitative results from the first survey held up although as is well-known there were often heated but enjoyable debates on the overall mass estimates. The basic understanding that the molecular gas resides in self-gravitating clouds has been well-supported by samples of hundreds of clouds. However, since the self-gravity is only marginally binding and there are important feedback and magnetic forces, the clouds typically have elongated filamentary morphologies.
Shortly after, you and John Kwan developed a model for luminous infrared sources that had recently been discovered in large molecular clouds.
I have always enjoyed moving between different areas of astrophysics rather than staying in one sub-discipline in either theory or observation. The collaboration between John Kwan and myself arose quite naturally since he was a graduate student with my postdoc mentor Peter Goldreich at Caltech. The infrared analysis in the Goldreich and Kwan '76 paper was entirely analytic and John and I undertook the infrared modeling in order to understand the predicted angular or radial distributions which would be seen by IR observers. This research was also partially motivated by interactions with the Neugebauer et al. infrared group at Caltech.
In 1983 you and your collaborators obtained near-infrared Fourier Transform spectra towards the Becklin-Neugebauer object. What did you learn?
This was a most enjoyable collaboration (and expansion of both my theoretical and observational techniques) with Don Hall, Steve Ridgway and Susan Kleinman. The high resolution spectra were obtained in the 2, 4 and 4.6 micron bands which included multiple CO vibrational bands, the H2 vibrational transitions and various HII recombination lines all done with kinematically resolved spectra. The diversity of environments probed was a really new discovery space -- including high temperature and very high density (109 cm-3) from the CO bandhead emission, accretion and outflows probed in the individual CO vibrational lines at lower J and the HII recombination emission. From this analysis we were able to estimate inflow and outflow rates and determine that the central star was a very young B-type star.
A few years later you, with Dave Sanders and Dan Clemens, explored the role of cloud-cloud collisions in forming OB stars.
The motivations of this model for forming star clusters were several: 1) The observed Mach 10-100 motions within the GMCs implied that it would be very hard to initiate large scale collapse of the clouds without a strong compressive force of similar magnitude pressure (100 - 1000 times the H2 gas thermal pressure) and 2) the appearance of almost all OB star clusters only in the spiral arms despite the fact that there was plentiful molecular gas in the interarm regions. The latter is deduced from the fact that in the interior of local spiral galaxies the H2 gas content dominates the diffuse atomic gas -- this can not be the case if the molecular gas is confined solely to the arms since most of the orbital period gas is spent in the interarm regions. Within each radial annulus, gas mass must be conserved -- aside from the small fraction per orbit which goes in to star formation. In a density wave flow, the dominant gas component can not be confined to a small range of azimuth in the spiral arms. An implication of this insight is that the molecules (not necessarily the cloud structures) probably last more than 100 Myr -- not the 1 to 10 Myr, commonly assumed by many astronomers and theoreticians, even today.
Moreover, within the density-wave spiral arms, it is well-known that there will be significant orbit crowding in the spiral arms and higher cloud-cloud velocity dispersions -- both of these are likely to lead to a higher rate of cloud-cloud collisions and since the clouds will be colliding at 5 - 20 km/s the ram pressure at the interface could be sufficient to overcome the turbulence and compress the gas.
In 1987 you and your collaborators used data from the Massachusetts-Stony Brook Galactic CO survey to analyze the properties of the molecular cloud population in the first Galactic quadrant, including some associated with hot cloud cores or radio HII regions. What were your conclusions?
These samples of Galactic GMCs were used to measure the distributions of cloud properties, the frequency of OB star formation, the correlation of cloud line-widths and their diameters, their temperature distributions, and IR luminosities. From the UMASS survey there were samples of 500 - 1000 clouds -- all with resolved spectral line-widths and brightness temperature distributions. The Galactic rotation curve was also determined with unprecedented precision by Clemens et al.
The integrated CO luminosities were found to be well correlated with the dynamical masses of the clouds -- this was a most important finding since it then enabled using the CO 1-0 luminosities of whole external galaxies to estimate their molecular gas masses. (Unfortunately, many investigators now using ALMA in higher J CO transitions have extensively used similar relations to estimate the gas masses.)
Over the years you have increasingly focused on extragalactic star formation. In 2007 you and your team presented an overview of the COSMOS project, which has imaged the largest contiguous area with HST with the aim to understand the evolution of high-redshift galaxies, star formation, AGNs and dark matter and their correlation with large-scale structure. Several hundred papers have already resulted from this initiative, including many follow-up studies with ground- and space-based facilities. Please describe the gestation of COSMOS and some of the key insights gleaned so far.
I believe it was in 2004 that the HST was scheduled for a number of large projects PI'ed by science staff at STSCI. There was a protest in the academic community that STSCI staff had an unfair advantage. (Personally,I believe this belief was wrong.) But in any case Steve Beckwith (the STSCI director) decided to schedule a community workshop and he asked me to chair the working panel on cosmic evolution and cosmology (even though I hadn't really worked in that research area).
Looking at what had been done so far and at theoretical simulations for the evolution, I was struck by how the expected angular scales of the predicted dark matter Large Scale Structure (LSS) were much larger than any existing survey areas at the relevant redshifts. I also realized that one could cover the expected scales using the new ACS camera with a mosaic of single exposures. Most of the members of my panel were very enthusiastic about this general scheme. Also, having come from the star formation community I realized that such a survey must be done in a field at an accessible declination for both the VLA and more importantly ALMA which was just coming on line. Not doing that would be 'almost criminal'. The project with 38 original co-I's was granted 500 orbits for a single filter exposure at each pointing. The equatorial siting of the field and the size of the HST allocation also meant that virtually all large facilities would want to survey in the field where there was rich ancillary data, rather than far North and South fields.
In the beginning of the project, we decided to have a very open team -- that is, any other project contributing imaging or spectroscopic data in the field could then have access to all the other COSMOS datasets after one year of their own proprietary time. We have had over 500 people actively working on COSMOS at different times. This worked out extremely well since those teams contributing one dataset could then count on a huge set of ancillary data, including other bands, catalogs and surveys at radio, submm, and X-ray wavelengths. The photometric redshift catalog is now made from deep imaging in over 30 separate optical, IR and UV bands and has precision redshifts with ≲ 1% accuracy for approximately 800,000 galaxies.
As mentioned above, the driving science motivation was to investigate the correlation of the cosmological dark matter LSS with the evolutionary properties of the baryonic galaxies. A few external people had said that the LSS couldn't be mapped with the accuracy of photometric redshifts but it was immediately obvious to us that they were wrong and given the large number of galaxies in the field, one could not expect to get that many spectroscopic redshifts. At the time the proposal was written, we subjected the cosmological simulation for Lambda CDM to the same ‘noise’ in redshift as that of the expected photometric redshifts and still easily saw the simulation LSS. So, within 2 years of the start of COSMOS, we published 2 deg wide maps of the LSS with the galaxy locations superposed. These LSS structure maps were then correlated with the dark matter distribution derived from weak lensing by Massey et al. It was immediately found that the major LSS / galaxy type correlations don't really set in until z ~ 1.5. The survey has also provided an enormous galaxy sample to look for different types of galaxies at early epochs (back to when the cosmos was only about 1 Gyr after the Big Bang).
Most recently, I have used submm continuum fluxes from the ALMA archive for 708 galaxies (within the archive pointings and having Herschel satellite IR fluxes) to measure the gas masses in the galaxies using the observed Rayleigh-Jeans continuum to estimate the dust and gas masses out to z = 4. These masses are much more reliable than from the CO higher J lines, and the continuum images are there in the archive for any ALMA pointing. I believe this should become the preferred future technique to measure galaxy gas masses at least out to z = 5. (This is the main subject described in more detail in my Russell lecture paper published in ApJ in early 2023.) The technique is simply an extension of what Roger Hildebrand used 40 years ago for measuring masses of Galactic dark cloud cores.
In this project, we obtained gas mass estimates for 708 star-forming galaxies from z = 0.3 to 4, along with estimates of the total star formation rates (both the unobscured SF seen in the opt/uv continuum and the obscured SFR from Herschel IR luminosities). We find that the interstellar gas peaks at z = 2 and dominates the stellar mass in star forming galaxies down to z = 1.2. Roughly 70% rise in SF out to z = 2 is found to be due to the rising gas masses at early epochs and 30% is due to increase in star formation per unit gas mass (i.e. the star formation efficiency). Thus, the galactic gas contents are likely the driving determinant for both the rise in SF and AGN activity from z = 5 to their peak at z = 2 and subsequent fall to lower z.
One of the very special aspects of COSMOS has been the open and trusting atmosphere of the team. The project has now been going 20 yrs and the yearly team meetings (all over the world) typically still have 50 - 100 participants. The project has only a loose set of rules and guidelines -- not the rigid multi-page documents characteristic of most other survey projects.
In closing, I'd just like to say how amazing I find all the profound understanding from where we humans were 50 to 100 years ago. Sometime one forgets how recent the Great Curtis-Shapley debate was -- then astronomers weren't even sure there were other galaxies outside the Milky Way. And 50 years ago when I started in Astronomy, the nature of the dark clouds which form star and planetary systems was hidden behind dust. We can now see into these regions but find many things we still don't understand such as the nature of the turbulence; we can now observe star formation back in time to within a Gyr of the birth of the universe. This is truly amazing ! -- but there remains much more to be explored. The great progress has also come about both from instrumentation opening up the EM spectrum, but also from the physical insights of humans -- astronomers and astrophysicists.