An interview with Eric Feigelson

Interview by Bo Reipurth, SFN #336 - December 2020

Your thesis was about X-ray imaging of radio galaxies. What led to your interest in X-ray astronomy and in particular young stars?

This was all accidental. Bill DeCampli, an older graduate student at Harvard who I greatly admired, was pursuing a theoretical work involving T Tauri winds which were then thought to be responsible for the broad strong Halpha emission lines (today we know this is a signature of accretion). He, and Bisnovatyi-Kogan earlier, argued that the T Tauri wind might be powered by an X-ray emitting corona at its base. NASA's HEAO-2 Einstein Observatory was successfully operating, so we requested observation of some nearby star forming regions. This seemed crazy at the time (clouds are cold, X-rays are hot) but the decision was made by Riccardo Giacconi alone, as he controlled the satellite scientific program.

We detected 7 out of 19 classical T Tauri stars in brief 1/2-hour exposures at levels above 10³⁰ ergs. We did not think it arose from the wind; instead our discussion was dominated by the remarkable and “unexpected finding” of rapid X-ray variability from DG Tau: most of its photons arrived in the last 4 minutes of observation. “The DG Tau X-ray "flare" released ~ 1 x 10³⁴ ergs in ~ 2 x 10² s”. Note the quotation marks: we really were uncertain.

In 1981 you published a second highly influential X-ray study with Kriss. What was the new insight here?

I was then graduating and moved to MIT working in Claude Canizares' X-ray group. My emphasis was still on radio galaxies, but I met there a graduate student Gerard Kriss. He was on his way to a 1.3-meter telescope that MIT shared on Kitt Peak for his research, and I asked him to squeeze in spectra of three stars in the Einstein fields that were not in the Herbig-Rao catalog of classical T Tauri stars. Two of the stars were on the edge of the famous L1551 cloud in Taurus where HL Tau recently formed. These proved to be unobscured K7 stars that, at V ≅ 12, were clearly far above the main sequence if they were associated with the cloud.

Here we were braver in interpretation saying that the pre-main sequence level of flaring “implies that magnetic dynamos, active regions, or other conditions required to produce coronal X-ray emission ... to an exaggerated degree compared to main-sequence stars.” This all became completely clear in the 1983 Einstein study of the ⍴ Ophiuchi cloud by Thierry Montmerle and colleagues where many X-ray flares were seen, a blinking “X-ray Christmas Tree”.

I also like that these three stars were the first X-ray discovered post-T Tauri stars that George Herbig had predicted should exist in abundance. We now see thousands of them with the Chandra X-ray Observatory.

Montmerle also appears in a 1985 Letter on an extremely variable radio star in the ⍴ Ophiuchi cloud.

Yes, this has a story behind it. Thierry and I met working in the Einstein computer room at Harvard-Smithsonian in the early 1980s. I knew French since childhood, we became friends, and he invited me to Saclay to discuss T Tauri flaring. He had the idea that, since solar flares are seen in both radio and X-ray bands, we should go to the NRAO VLA (that I had used in my radio galaxy work) to look for T Tauri radio flares. We worked out on an envelope that, based on solar flares, the radio event would be orders-of-magnitude too faint for the VLA. He brushed away this problem, we wrote a successful proposal, and we joined forces in New Mexico to observe.

Working late at night at VLA's powerful VAX computer, we were confused: some of the data showed a pattern of constant extragalactic sources, but some of the data showed a bright additional source. We turned to Ron Ekers, VLA Director, and he couldn't find any error we made in the complex data analysis. So we accepted it, found it was associated with a poorly studied post-T Tauri star named Dolidze-Arakelyan 21 that was an Einstein X-ray source, ROX-8, in Thierry's 1983 study. While we later recognized that the excess radio emission was similar to flares in RS CVn binary systems, I always smile at the memory that Thierry had thrown away my rational arguments against the proposal... and we made a nice discovery.

You have studied the implications that magnetic flaring had on the early solar system.

Years later in 1999, Thierry and I wrote together a review of the pre-Chandra and XMM observations of pre-main sequence flaring in Annual Reviews of Astronomy & Astrophysics. My favorite section was the speculative links between the stellar flaring, an active early Sun, and isotopic anomalies in primitive meteorites from the solar nebula. I had discussed this years earlier with meteoriticist Guy Consolmagno when we were both at MIT. Since the discovery of the spallation-generated 10Be isotopic anomaly, this association is now widely accepted by meteoriticists like Matthieu Gounelle and Marc Chaussidon.

In 1996 you wrote a much discussed paper on dispersed T Tauri stars detected in X-rays. What were the issues then and now?

Ralph Neuhäuser and others were finding small samples of post-T Tauri stars far from active star forming regions in the X-ray ROSAT All-Sky Survey. But I felt that this literature had missed a point: millimeter astronomers were showing that molecular clouds exhibit supersonic turbulent motions, reaching several km/s on large scales for the giant cloud complexes. If clumps of star form in turbulent eddies, they should inherit these motions and be widely dispersed around the cloud complex. The 1996 paper focused on the isolated Chamaeleon cloud where this halo of post-T Tauri stars should be easily found. Kevin Luhman has searched and did not find a large population, indicating that this cloud has not been actively forming stars for very long. But I still think that the more massive, long-lived star forming complexes like Carina and M17 should have tens of thousands of slowly dispersing stars around them.

In 2003 you and your team observed the Orion Nebula Cluster with Chandra continuously for 13 days, leading to the first comprehensive study of X-ray variability in YSOs. What was the genesis of this project?

I was brought into the Penn State faculty in 1983 to support Gordon Garmire who had won the contract to build the ACIS CCD imager for the Chandra X-ray Observatory. The Orion Nebula Cluster, which had been known in X-rays since the 1970s non-imaging UHURU satellite, was meant to be a poster child for Chandra's sub-arcsecond mirrors with thousands of pre-main sequence stars in a few arcminutes. It succeeded beyond our hopes. Using Guaranteed Time for the ACIS Team together with competitive Guest Observing time, we stared at the ONC for nearly two weeks. This Chandra Orion Ultradeep Project (COUP) was the first big science project I led producing a special issue of the journal with a dozen papers involving over 40 scientists. Since the membership of the ONC had already been established from ground-based studies by Lynne Hillenbrand, COUP provided a template for all of our following Chandra studies of less well-studied and more distant star forming regions.

It seems that X-ray flares are not simultaneously appearing as optical flares. Are these two separate phenomena?

It's even worse when radio flares are considered! After a frustrating 1992 ground- and space-based campaign on HD 283447 in Taurus, and finding that only 5% of the COUP stars showed time-correlated optical and X-ray variability in an effort led by Keivan Stassun, I have given up with simultaneous multiwavelength campaigns. While I don't believe these are entirely separate phenomena, simultaneous flaring is difficult to detect. X-ray flaring is most successfully seen. Optical flares can only be seen if they are comparable to the photospheric emission, and radio flares are only at detectable levels if they considerably exceed the Güdel-Benz relation.

You and your team recently completed the MYStIX program. What did you learn?

MYStIX, supplemented by SFiNCs and MOCX, are a 15-year-long effort by our Penn State group to broaden our knowledge of pre-main sequence populations beyond the Orion Cloud. Kosta Getman, Leisa Townsley, and I initiated the projects along with Pat Broos, Michael Kuhn, Matt Povich, Tim Naylor, Alex Richert and others. We have a census of ≅ 40,000 pre-main sequence stars in ~ 40 star forming regions identified from Chandra, Spitzer and JHK surveys with a big emphasis on removing contaminants. We have lots of results: HII regions are filled with 10⁷ K gas (not 10⁴ K) from shocked OB winds; most clusters are expanding on timescales of 10⁶ years; the X-ray luminosity function can give the total population of a cluster; and a new stellar chronometer is found using X-ray and IR photometry. This last result confirms the reality of cluster age spreads and led to the discovery that most clusters exhibit a spatial age gradient. This last finding is an important constraint on cluster formation theories, supporting models where infalling filaments feed star formation in cluster cores for several million years. This work has been lots of fun and new insights are still emerging.

In parallel with your X-ray studies you have taken great interest in statistical methods in astronomy and have published books on the subject. Where did that interest come from?

It came from X-ray astronomy! As an assistant professor, I walked over to the Statistical Consulting Center to ask how to do regression with upper limits in the sample. I was told an entire field of modern statistics is devoted to this (“survival analysis”). This was totally amazing to me, and ever since then I've been learning, educating and promulgating modern statistics to advance astronomical research. Its importance is growing in this era when enormous surveys - SDSS, Gaia, and soon Rubin/LSST - play such a big role in astronomical research.