My Favorite Star

By Sean Carroll | April 7, 2010 7:32 pm

For a long time now, my day job has been “theoretical physicist,” as a quick glance at my papers will confirm. But it was not always thus! Very few people are actually born as theoretical physicists. When I was an undergraduate astronomy major at Villanova, I wasn’t thinking about quantum field theory or differential geometry; I was working on photometric studies of variable stars. My personal favorite star was Epsilon Aurigae, a mysterious eclipsing binary. One of the very few stars out there that has both a Facebook page and a Twitter feed. And now Epsilon is in the news again!

Among this star’s claims to fame is that it has the longest period of any known eclipsing binary: over 27 years. But it’s not just about facile record-holding; this system is truly puzzling, especially the nature of the secondary (the thing that eclipses the primary star). The basic problem is that the eclipse has a fairly flat bottom, as seen in this light curve from the previous eclipse in 1982-84.

Epsilon Aurigae 82-84

A flat-bottomed light curve is usually associated with a total eclipse; the secondary completely blocks the light from the primary for a while. But in this case, the spectrum of the system seemed to remain unchanged, indicating that most of the light was still coming from the primary star, even in the middle of the eclipse. This led Huang in 1965 to propose a clever model, in which the secondary is actually a disk seen edge-on; the eclipse is therefore not total, but the disk blocks out part of the light without emitting much of its own. And indeed, with modern infrared telescopes we can discern the light from the secondary — it does look like a relatively cold disk, about four astronomical units in radius, with a hot central star.

Epsilon Aurigae art

The 1982-84 eclipse raised a problem with Huang’s model, however. If you look closely at that light curve above, you’ll notice that it gets brighter right near the middle. (The gap in data is from when the star was behind the Sun and unobservable.) Your first guess is that this is probably just a fluctuation in the in brightness of the primary star; but it turns out that this can’t be right. The primary is indeed variable, but its color changes in lockstep with its brightness, an effect that can be measured by observing with different filters. And the mid-eclipse brightening shows no variation in color. It’s not due to variability in the primary; somehow the disk is letting more light past, right during mid-eclipse.

This is where I come in, as an undergrad doing research under Ed Guinan. For my undergraduate thesis, we tried to put together the most sensible picture we could manage of Epsilon Aurigae; our picture has now become the consensus model. (One difference is that we thought the primary was a supergiant, but now it appears that it’s likely to be a smaller star.) As far as the disk is concerned, we built on a variation of Huang’s disk model, due to Wilson in 1971. Wilson suggested that the disk should be thin and tilted, rather than thick and edge-on, with a semi-transparent inner region. In that case, we could imagine that the central star (presumably holding the disk together) could clear out a region near the center, and light passing through that hole could account for the brightening. I wrote a simple computer program (QuickBASIC on an IBM PC!) to calculate the light curve in this model, and we were able to get an extremely good fit to the data. Here’s the killer plot from our paper, with three different models: edge-on thick disk, opaque tilted disk with a central hole, and tilted disk with a central hole surrounded by a larger semi-transparent region.

Epsilon Aurigae light curve fit

And now, it’s eclipse time again! (I’m getting old.) Epsilon started going into eclipse in August 2009, right on schedule, and the eclipse is predicted to last until May 2011, so we’re just a bit before mid-eclipse right now. Here’s the current light curve. But technology has advanced quite a bit since my student years. Nowadays, we don’t need to puzzle out the meaning of a light curve and come up with an elaborate story involving tilted disks; we can just take pictures of the thing.

And that’s just what we did. Under the leadership of Brian Kloppenborg and Robert Stencel at the University of Denver, we put together a proposal to observe the eclipse using CHARA, the Center for High Angular Resolution Astronomy. CHARA is an array of optical telescopes on Mt. Wilson that act as an interferometer, enabling extremely high-resolution imaging of astronomical objects. And by “we,” I do mean that I was included in the proposing team — a little return to my roots. My contributions to the final results were not zero, but they were small; the lion’s share of the credit certainly goes to Brian and Bob and the rest of the team.

And our results just appeared in the form of a paper in Nature. Unfortunately behind a paywall, but there is an extensive NSF press release. But who cares about the words? The stunning things are these pictures — you can actually see the disk begin to move across the surface of the star. There’s even a video where you can compare a model to the actual data. Click to embiggen.

Epsilon Aurigae eclipse

I have to admit that, while these images are unambiguously amazing, the result is somewhat bittersweet for me. John Monnier, another one of the team leaders, admits that he was skeptical about all this tilted-disk business; it just seemed like a house of cards. But I wasn’t skeptical for a moment, having gone through the work of trying to fit the data with various different possibilities. The extraordinary thing to me about observational astronomy was always how you could put together an apparently baroque model of some complicated system, just on the basis of a precious few data points, and yet have some degree of confidence that you were on the right track. Reality is very constraining. So in some perverse sense, it almost seems like cheating to actually take pictures of the thing. Where’s the fun in that? (Of course it’s a great deal of fun.)

And now we need to see what happens at mid-eclipse! I predict we’ll be able to image that small central hole we posited many years ago. Not at all certain about that — I suspect the star at the center of the disk is pretty active itself, and there are probably significant variations in the opacity of the central disk region. That’s what data are for. I will always love being a crazy-eyed theoretical physicist, but there’s something uniquely rewarding about digging into the data and coming to an understanding of far-flung pieces of our universe.

  • Julianne

    Sean, those observations are insane. Love. It.

  • Tod R. Lauer

    Wow! I’m impressed!

    Sean, you may want to try your hand with some of the goodies that Kepler is coming up with…

  • Christian Ready

    Wow, this brings back memories for me as well! As an undergraduate at Villanova a couple of years later, we took data on the same system (albeit well after the eclipse.) I remember reading your paper and writing my share of QuickBASIC programs on those same IBMs!

  • Aaron

    Brian Kloppenborg and “Dr. Bob” are part of a NSF-funded citizen science project called Citizen Sky, dedicated to understanding this star. If anyone is interested in participating in this project, just visit the web site. Amateur astronomers, armchair astronomers, programmers, statisticians, writers, artists and anyone interested in participating in a science project are needed!

  • Ijon Tichy

    Your favourite star. My favourite blog post. Great work.

  • Jennifer Hoffman

    Hooray, congratulations to my colleagues Brian and Bob! Brian has more press info online at .

  • CoffeeCupContrails

    “previous eclipse in 1982-84″…
    “The 1984-86 eclipse raised a problem with Huang’s model”

    Different eclipse?

    • Sean

      Oops, typo fixed.

  • Kuas

    Isn’t that more satisfying than, say, predicting that there are four spacetime dimensions, or that time moves forward?

  • Pingback: A weird kind of eclipse « Meng Bomin()

  • Pingback: Sean Carroll’s (the biologist) double-duty day | Gene Expression | Discover Magazine()

  • I.P. Freeley

    Sounds like the companion is a degenerate star with a disk around it. Shouldn’t there be some sign of a supernova remnant or planetary nebula?

  • Tristram Brelstaff

    I always thought it significant that two of the brightest eclipsing binaries in the sky, Epsilon Aurigae and Beta Lyrae, were so strange.

  • db

    But.. but… that’s a usefully detailed picture of another star. 2000 light years away! I had no idea that was possible yet. Amazing.

  • Lab Lemming

    That is very cool.

  • Telek

    Ah! the memories. One long 11-hour day to write a basic program on a Commodore 64 sometime at the end of 1984 in Fayetteville, Arkansas, working under Dr. Lacy, a starry-eyed undergraduate, a single set of data, a single curve that came out so nicely, and after that, all my puny knowledge of physics eclipsed forever… ah, the memories… beautiful Epsilon Aurigae, great job, great pictures…

  • ObsessiveMathsFreak

    Top shelf stuff here.

    I’ve never heard of stars with such dusty disc coronas before, so this post has been very educational. It’s also nice to see the evolution of a physical model of data.

    I’m curious about a few things though. Firstly, why is there a gap in the original data at around June 1983? Cloudy skies? Secondly, in relation to this, how would observations in this region have influenced your original modelling? Would you, say, have tried a disk with a hole and central star?

    Finally, considering that your disc only model (if I understand correctly) fit so well, is it possible that there is not a star, but instead only, say, a Jupiter sized body at the center of the disc?

  • Nik Reeves-McLaren

    More on the BBC website here in the UK about this work – nice job!

  • Rohan Mehra – Chocolate Films

    Has the world gone mad?
    That is clearly and unambiguously an image of a gigantic Ian M. Banks-esque megaship!

  • Sili

    But it was not always thus! Very few people are actually born as theoretical physicists.

    Boo. Now I love you a little less.

  • TomC

    Flat-bottomed curves, you make the rockin’ world go ’round!

  • Roman

    “But.. but… that’s a usefully detailed picture of another star. 2000 light years away! I had no idea that was possible yet. Amazing.”
    Exactly! I thought stars were, are and will be for some time just point light sources. Or is some kind of the trick involved. Some explanation please.

  • Brendon J. Brewer

    The process for turning interferometry data into pretty pictures is pretty cool too – it’s not as though they literally just took an image. :-)

  • Dirk F.

    Yeah Villanova Astronomy! I still look for AW UMa in the sky, my favorite eclipsing contact binary. Light curves don’t lie, but they can be very ambiguous — just try spot modeling! Nice work.

  • Roman

    I understand that this is not a direct image. But it is not just an artist impression like the first image in the post. So do we see (see in a broad sense) the star as a sphere here or not?
    Can I take the scale bar from the pic and measure the star diameter?

  • uncle sam

    If I remember correctly from reading and e.g. the wonderful (now out of date) “Music of the Spheres” by Guy Murchie, Epsilon Aurigae was then (1960) thought to be one of the largest, lowest-temperature and lowest density stars. It was a “red hot vacuum” and indeed, likely near transparent per the history in Wikipedia. In a science fiction story maybe by Andre Nortion, there was an infrared star so visitors to that system needed goggles just to see. I think only very small dwarfs (barely cooking, so to speak) would be like that – not giant stars.

  • R33

    Sean: “Wilson suggested that the disk should be thin and tilted, rather than thick and edge-on, with a semi-transparent inner region.”

    It’s nice but from what you say the credit clearly goes to Wilson. Not that it was hard to figure anyway.

  • TheRadicalModerate

    Shouldn’t a star + disk like this have very high angular momentum? And shouldn’t the fact that it’s tilted cause the primary to put some kind of torque on the whole system? And shouldn’t that cause all sorts of interesting (and extremely transient, on an astronomical scale) dynamics? Is the silly thing going to precess, or something? Shouldn’t this generate some testable predictions about the evolution of the light curve over the next few cycles?

  • dreamer

    Cool results!

    Talking about pay wall, why don’t you guys put the paper on arXiv?

  • Hiranya

    Best. Post. Ever!

  • JohnJohn

    Very cool. When I first glanced at those CHARA images I thought it was another model! Unambiguously amazing, indeed!

    This system reminds me a bit of my favorite variable star, KH 15D. In the late 90’s it was discovered by Kearns & Herbst to undergo 20-day eclipses every 48 days, with a central rebrightening at mid-eclipse:….116..261K

    In 2004 we discovered that it was a spectroscopic binary

    Using archival plates from a tiny observatory in Italy, Josh Winn and I found that the eclipses were very different in the 1960s (shallower, brighter out of eclipse, and phase-shifted):

    Finally, a model was proposed, involving a 48-day binary pair that periodically peeks around a warped circumbinary disk

    The disk has precessed to the point that both stars are now blocked from view and may remain blocked for the next century.

  • David Copson

    The answer to the problem is obvious to me. The disk is doughnut shaped. At peak of the eclipse more light from the star peeks through the hole.

  • Don Davidson

    Just out of curiosity, is there any reason to think the disc could be a product of astro-engineering? It’s flat as a pancake and fairly symmetrical. Perhaps a Dyson Ring, or some variant thereof? Astrobiologists have recently suggested looking for evidence of large engineering products in the search for extraterrestrial intelligence.

  • Invader Xan

    I was just wondering — You mentioned above that at minimum light, the spectrum of ε Aur remains unchanged… but are you referring to the continuum here?

    My immediate thought (as an astrochemist) would be that at minimum light, it would be an ideal time to look for absorption lines in the spectrum and attempt to gain some insight on the actual composition of that disk. To your knowledge, has anyone attempted this?

  • Eric

    Hi Sean,

    Just wanted to thank you for your talk yesterday, so I apologize that this post is a bit off-topic. I thought you explained most of your points very clearly, and your delivery was excellent. I was a little bit confused about one point, though:

    My understanding was that baby universes can have either forward- or backward-propagating time because the “empty space” state from which they fluctuate does not have a unidirectional arrow of time. But you also said that these baby universes could themselves have children once they reached the empty space state, and (I think) that time in these offspring would not necessarily point in the same direction as it had in the parent.

    Considering our (sub-)universe now, we can say that (assuming exponential expansion) it will eventually find itself in essentially the empty space state. But for all times in our universe, the current state has a well-defined history describing its evolution from the big bang. An observer looking at a late-time state would not be able to deduce that history, but it is (seems to be?) real nonetheless.

    As such, the empty space state in our universe seems like it would still have a well-defined arrow of time pointing away from the big bang. If a baby universe were to fluctuate into existence with its arrow of time pointing in the opposite direction, its evolution would conflict with the pre-existing history. This seems to imply that any baby universes that fluctuate from our eventual empty space state would have a preset arrow of time.

    I’m sure I’m missing some subtlety, but I’d really appreciate it if you could clarify this for me. Thanks again for the talk, it was very enjoyable.

    PS – it was pretty funny to see the kid who spends 15 minutes every class asking poorly-thought-out questions put on a show after you finished!

  • Sean

    Hi Eric– Baby universes aren’t born with an arrow of time pointing in some arbitrary direction; it points in the direction away from which the universe didn’t exist. So yes, in our universe any babies would have arrows lining up with ours, pointing away from the Big Bang. That’s because our universe is not (in this scenario) an eternal “parent” universe, but is a baby itself.

  • Brent

    Actually, for the disk, the secondary star within the disk might have gravitationally captured a significant aspect of the inner debris inside the disk turning it into more of a torus or thick ring around the secondary and when the midpoint of the secondary’s disc passes in front of the star, the primary’s light is able to penetrate the thinner part of the tilted disc and the light curve swerves upward, accordingly. This might explain the changes in the light curve.

  • Pharlophus

    To be able to see the reality from the few points of data and then years later to see the pictures of that reality as it slowly unfolds, you sir are allowed to sit back and just smile without explanation for the rest of your life.

  • Jeff

    For those wishing to follow the eclipse, the Epsilon Aurigae International Campaign can be seen at

    We have nearly 60 observers from 20 countries around the world contributing thousands of high quality observations, both photometric and spectroscopic. We have produced 16 Newsletters available as free pfds from the web site and #17 will be out n a few days. We are well into totality and the light curves and spectroscopic data are showing some surprises. The mysterious mid-eclipse brightening is predicted to start in May and last through fall.


Discover's Newsletter

Sign up to get the latest science news delivered weekly right to your inbox!

Cosmic Variance

Random samplings from a universe of ideas.

About Sean Carroll

Sean Carroll is a Senior Research Associate in the Department of Physics at the California Institute of Technology. His research interests include theoretical aspects of cosmology, field theory, and gravitation. His most recent book is The Particle at the End of the Universe, about the Large Hadron Collider and the search for the Higgs boson. Here are some of his favorite blog posts, home page, and email: carroll [at] .


See More

Collapse bottom bar