An interview with Pat Hartigan
Interview by Bo Reipurth, SFN #354 - June 2022
How did you end up deciding to study star formation?
I've always loved astronomy. I was an amateur astronomer from age 8, and pursued it as an undergraduate even though the employment prospects seemed poor. Around the time I was graduating, Voyager 2 was just about to visit Saturn, and in that excitement, I was drawn to planetary sciences and ended up in that department as a graduate student at Arizona. I found that while I liked geology, atmospheric chemistry, meteorites, and so on, it was a bit too detailed for me. But the other extreme, studying a faint fuzzy galaxy, seemed too vague. Star formation was the perfect mix of being able to combine good spatial and spectral resolution, without too much detail, and I particularly liked being able to do quantitative emission line analysis. Each of those lines was trying to tell us something.
Your PhD in 1987 dealt with radiative bow shock models of Herbig-Haro objects. How did you choose this subject, and what was the context of HH studies back in the 1980s?
Star formation was not a research area in the Department of Planetary Sciences at Arizona but it was in the Astronomy Department where Charlie Lada headed a group studying bipolar outflows. Charlie was already funding several students, but fortunately he was able to connect me with John Raymond and Lee Hartmann for a summer, where I learned about radiative shock waves. They had done a paper in 1984 where they tried to explain the large widths in the emission lines of HH 1 and HH 2 with a bow shock model, but they didn't have a comprehensive grid that could be used for more general lines and bow shapes. So I ran a large grid (for the time) of models, and found an interesting result that the line width depended only on the shock velocity, and not upon parameters like the density, shape of the bow, orientation, preionization, etc. Most of the spectra we had were from the Multiple Mirror Telescope, and they had very good spectral resolution and signal-to-noise, but were taken through an aperture so had poor spatial information. At about the same time, Alex Raga was analyzing long-slit spectra taken by Karl-Heinz Böhm and Josef Solf, which was a complementary way of going about it.
In 1976, Dick Schwartz had shown that the spectra of HH objects were similar to those of supernovae remnants, and Gary Schmidt's 1979 observation of lack of polarization cemented radiative shock waves as being the correct physics interpretation over reflection nebulae, the other main competing idea at the time. You could either get a bow shock from a stationary cloudlet embedded in a wind, or from a “bullet” ejected somehow from the source. Then jets were discovered in the early 1980s, and there was a lot of comparison to extragalactic jet models, crossing shocks in beams and the like. But those models did not explain line widths or proper motions well. By the late 1980s the idea of a pulsed jet (I first heard it from you at a meeting in Portugal!) came to the fore. Having low-velocity variations embedded in a high-velocity flow meant you could have a highly collimated flow, and nested bow shocks with high proper motions, while still retaining relatively low-velocity shocks within the beam. The jet was running into itself in this picture, it didn't need a surrounding medium to form shocks. Now we just had to figure out how to drive jets and learn why they were so time-variable.
A few years later you published an analysis of the visibility of the Mach disk and bow shock of a Herbig-Haro jet.
A bow shock driven by a collimated jet ought to have two major shock waves, the bow itself which accelerates material in front of it, and a Mach disk (sometimes called the jet shock) which decelerates the jet material. It struck me that both should be observable in most cases, and would have different excitations. That turned out to be true, although you can have the Mach disk disappear for a time if the jet is highly variable. These are well-documented now from HST observations, Fabry-Perots, and image slicers in HH 34, HH 47, and HH 32A, among others. In the paper I pointed out that images of HH 34 had a blobby structure and suggested one of those would be the Mach disk, though the one that was my favorite turned out not to be the right one. Extended HH bow shocks can have many knots, so to identify a Mach disk with confidence requires spectra to measure velocity profiles and line ratios.
Later again you, with Jon Morse and John Raymond, calculated emission-line ratios of planar radiative shocks and compared to spectra of collimated HH jets to derive mass-loss rates and ionization fractions of these flows.
The point was that the ionization fractions can be as low as a few percent in the emitting regions, depending on the object. It is important to think of such regions not as having a single density or temperature, but rather as a cooling zone. The interpretation can be a bit tricky if you want to calculate an actual mass loss rate. One feature of a pulsed jet is that the velocity variations in the jet determine the shock velocities, and these can be much lower than the average jet velocity. In these cases the density can easily vary by an order of magnitude between the preshock and postshock regions without that much change in velocity. So one would measure markedly different mass fluxes in the two adjacent regions by simply multiplying the density and the velocity. Measurements of mass loss rates based on emission line fluxes must average over these regions somehow. The spectrum is always biased towards the densest regions because fluxes scale like n$^2$ when the densities are low. Generally I feel a lot more secure looking at correlations between, say, mass accretion and outflow rates than I do with the actual values of the rates because of systematic effects such as these.
In another early study you led a study of how to unveil a T Tauri star. What was the essence of your method?
It was known for a long time that absorption lines in T Tauri stars were not as deep (i.e. they were “veiled”) when compared with normal stars of that spectral type. This could happen if the centers of the lines filled in with chromospheric emission, but except for a few really strong lines, the high-resolution spectra generally didn't look like chromospheric emission – they looked like a constant flux had been added. So I simply developed a model that solved for this constant over narrow wavelength intervals chosen where there were good lines. With the ratio of the extra emission to photospheric flux in hand, the known dereddened photosphere spectral shape then provides the dereddened spectral energy distribution of the excess.
The procedure has some interesting mathematics. As one would expect, the method works best when there are several well-defined spectral lines in the interval. If the photospheric template doesn't match or if there is a lot of noise, the method will apply too much veiling because it basically smoothes over places where the fit is poor. Templates must have their lines artificially broadened to match the rotational v*sin(i) of the T Tauri star. Our spectra were of high quality so none of these things were problems. The impressive thing to me was you can pull out weak lines like Li 6103 that were in the T Tauri star but not in the template, and it gave really excellent photospheric removals around forbidden lines, so you now could see high-velocity and low-velocity components clearly.
This was followed by a study in which you and your team applied your models to a sample of 22 T Tauri stars. What did you find?
The excesses were flat or slowly rising towards the blue. Scott Kenyon and I worked on a simple `slab of hydrogen' model and estimated temperature, optical depths, and filling factors. The Paschen continuum has a red slope in the optically thin limit, and goes over to the Planck function when optically thick, which will be blue for a hot slab. So I suppose it isn't surprising that you need τ~1 to get an intermediate slope. A more realistic density model than a slab would still have most emission coming from τ~1 anyway just from the radiative transfer equation. But it was clear that the emission came from hot gas over a small area, consistent with the notion that accretion drove veiling. Gibor Basri and collaborators were doing something similar with different instrumentation at about this time.
With the benefit of hindsight, our mass accretion estimates were too high. There was a factor of ~ 1.6 from using a boundary layer model rather than a magnetic accretion model, and we did not have great blue data or good photospheric colors so the reddening estimates were too high, and both effects tend to increase mass accretion estimates. There is still discussion about what fraction of the accretion energy gets absorbed by the stellar photosphere and is not easily observable. Both laboratory experiments and 3-D MHD models show that the geometry of the accretion shocks and cooling zones can be pretty complicated, so there are always going to be significant systematic uncertainties in mass accretion rates.
In 1995 you, with Suzan Edwards and Louma Ghandour, published a highly cited paper in which you analyzed forbidden line profiles of 42 T Tauri stars and were able to correlate the disk accretion and the mass loss of the stars.
If jets are driven by accretion then regardless of any systematics inherent in both the mass accretion and outflow rates, these quantities should correlate. With the de-veiling technique working well to isolate the weak high-velocity emission we could now at least get some estimate of the mass loss rates, and sure enough, they did correlate with mass accretion, though there was always some annoying scatter. For me, one of the really interesting aspects of what we found was to be able to invert the forbidden line profile, assuming Keplerian motion, to infer the surface brightness of the forbidden line as a function of radius. This sort of thing is commonly done with ALMA data of CO and other lines now.
Around the same time you, with Stephen and Karen Strom, observed 39 wide pre-main sequence binaries and posed the question whether such systems are coeval when placed in an HR-diagram with evolutionary tracks.
This was an interesting project with the goal either to help test pre-main-sequence evolutionary tracks or to find out whether or not binaries are coeval, depending on your point of view. Indeed, 2/3 of the pairs ended up looking coeval. The remaining 1/3 were puzzling. Some of them later turned out to be very close unresolved pairs, and others may not have been physical binaries. These were pretty wide pairs, so at some point one can ask if they `count' as a binary pair. I suppose the best data set would be a plot of inferred ages with separation across all spatial scales. It's a fun problem and there are a lot of better statistical methods available now to tackle such data.
A decade later you and Scott Kenyon used the Hubble Space Telescope to compare the components in 20 close T Tauri binaries. What similarities and differences did you find between the components of these binaries?
These pairs are a lot closer than the binaries were in our earlier study, and as one might expect the reddenings are similar between the components. It's interesting to me that the accretion rates of the primaries and secondaries correlated really well. That's a comforting result based on how we usually think of these systems. There are a few mixed pairs, with one of the components a classical T Tauri star and one a weak-lined T Tauri star, though sometimes infrared excesses, [O I] emission and Hα emission disagree as to what should be called a classical T Tauri star. Most of the pairs were coeval, but several systems simply misbehaved, most often with the less massive one a little younger. My takeaway from this is that while the overall picture of what we expect from young binaries hangs together, there is some additional scatter that is poorly characterized. Sometimes there are still unresolved very close binaries for one of the components, but even with that you can't force all pairs to be exactly coeval. At some level, individual stars have different mass-averaged ages because their accretion histories will not be identical, and there may be some long-term variability we don't fully appreciate because we observe these systems at one point in time.
Over the years you have published numerous papers on young stellar jets, both theoretical and observational. Most recently you carried out a detailed study with HST of the famous HH 7-11 flow. What did you learn?
Jets are dynamical systems so being able to combine emission line maps that have the spatial resolution of HST at different epochs removes a lot of the guesswork for interpreting what you see. Even with multiple emission line images it is easy to come up with a 3-D interpretation of images taken at a single time that is just wrong, and you know it immediately once you see a second or third epoch. In the case of HH 7-11 the multi-epoch images clearly showed what looks like a `smoke ring' in HH 8 and HH 10, regions where the jet has punched through what looks like a sheet of material. We can also see what is going on with multiple bow shocks in HH 7. In other objects like HH 32A, high-resolution spectra from the new Keck image slicer revealed a localized narrow Fe II component in the working surface, which is direct evidence of dust destruction. It is interesting to consider how dust managed to end up in a jet that moves at several hundred km/s. Measurements of magnetosonic Mach numbers are another exciting development, where in 2015 Anna Wright and I measured the cooling distance from the offset of the Balmer emission at the shock and forbidden lines in the cooling zone. That number is just what you need to infer Alfvénic Mach numbers, which in stellar jets are on the order of a few. This result is consistent with expectations from pulsed magnetized flows.
Another thing I've found that helps to interpret jets is to get involved with laboratory astrophysics experiments with lasers and pulsed power facilities. I've spent some effort in the past decade trying to understand the observational consequences of intersecting shock waves. It is a humbling dose of reality to see the complexities that arise in real shock fronts. I find it useful to be able to observe shocks from different perspectives in the laboratory, see how they actually evolve in time, and assess the degree to which the experiments match simulations.
What do you expect that JWST will be able to contribute to our understanding of jets from young stars?
JWST is going to be able to peer into the cloud cores around highly-embedded Class 0 objects, which ought to help define what goes on at those very early stages. Being able to trace fluoresced H2 at high spatial and spectral resolution will allow us to see all the molecular gas, not just the heated shocked component. JWST has much better spectral instrumentation than HST does, so it will be possible to pin down the dynamics of the outflowing gas much better than we have been able to do previously. These capabilities are exactly what are needed to test models of C-shocks and PDRs in molecular gas, and will help to connect the low-velocity gas situated along cloud/outflow boundaries with high-velocity outflows and jets.
Thanks for asking me to interview. It has been fun to think about these old papers again. Many of the conclusions and techniques seem obvious now, though they weren't at the time. It is easy to be led astray in thought by some particular paradigm or geometry. At the same time, the basic physics does lead you to what is right. A lot of the discoveries, or missteps, seem to be driven by the instrumentation that is available at the time.