Dec 162015

By Amaya Moro-Martin, AURA Astronomer at STScI

The solar system is densely packed with planets and also contains an asteroid and a Kuiper belts, remnants from the planet-formation epoch.  Are planetary systems with high-mass planets any different in terms of remnant planetesimal belts from those with low-mass planets or those with no known planets? What does this tell us in terms of planetary system formation and evolution?

how-meteors-could-have-brought-life-to-earth-slow-speed_59714_600x450Image credit: Lynette Cook

Planetesimals are the building blocks of planets, and mid and far-infrared observations with Spitzer and Herschel indicate that at least 10–25% of mature stars (10 Myr to 10 Gyr) harbor planetesimal disks with disk sizes of tens to hundreds AU (this frequency is a lower limit because the surveys are limited by sensitivity). The evidence for planetesimals comes from the presence of circumstellar dust: because the lifetime of the dust grains (<1 Myr) is much shorter than the age of the star ( >10 Myr), it is inferred that the dust cannot be primordial but must be the result of steady or stochastic dust production generated by the collision, disruption, and/or sublimation of planetesimals, like the asteroids, comets and Kuiper belt objects in our solar system. The presence of these debris disks in both single- and multiple-star systems, and around A- to M-type stars (also around the progenitors of white dwarfs), spanning several orders of magnitude difference in stellar luminosities, imply that planetesimal formation, a critical step in planet formation, is a robust process that can take place under a wide range of conditions. It is therefore not surprising that in some cases planets and debris disks coexist. But are dust-producing planetesimal disks more or less common around stars with planets? Using the evolution of the solar system as a model, in its early history, a star with planetary companions could be expected to be surrounded by a massive debris disk produced by the planetesimal swarm that formed the planets, the latter exciting planetesimal collisions and dust-production while undergoing orbital migration. On the other hand, at a later stage, the star could harbor a sparse dust disk after the dynamical rearrangement of the planets is complete and the planetesimal swarm has undergone significant dynamical clearing. Do observations support these trends?

Because the study of the planet-debris disk correlation could shed light on the formation and evolution of planetary systems and may help “predict” the presence of planets around stars with certain disk characteristics, we have carried out a statistical study of an unbiased sub-sample of the Herschel DEBRIS and DUNES debris disk surveys, to assess whether the frequency and properties of debris disks around a control sample of solar-type stars are statistically different from those around stars with planets. Out of the 466 and 133 stars in the DEBRIS and DUNES samples, respectively, we have selected a subsample of 204 FGK stars located at distances <20 pc (to maximize survey completeness), with ages >100 Myr (to avoid introducing a bias due to disk evolution), and with no binary companions at <100 AU (to avoid introducing a bias due to the observed differences in both disk frequency and planet frequency between singles and multiples). The debris-disk frequency within this unbiased sample is 0.14 +0.3/-0.2 .

In this clean sample, we don’t find any evidence that debris disks are more common or more dusty around stars harboring high-mass planets (> 30 MEarth) compared to the average population. Overall, this lack of correlation can be understood within the context that the conditions to form debris disks are more easily met than the conditions to form high-mass planets, in which case one would not expect a correlation based on formation conditions; this is also consistent with the studies that show that there is a correlation between stellar metallicity and the presence of massive planets, but there is no correlation between stellar metallicity and the presence of debris disks. Another factor contributing to the lack of a well-defined correlation might be that the dynamical histories likely vary from system to system, and stochastic effects need also to be taken into account, e.g., those produced by dynamical instabilities of multiple-planet systems clearing the outer planetesimal belt or the planetesimal belt itself triggering planet migration and instabilities.

Regarding low-mass planets (< 30 MEarth), one would expect that if the planets formed in the outer region and migrated inward, low-mass planets would have been inefficient at accreting or ejecting planetesimals, leaving them on dynamically stable orbits over longer timescales. On the other hand, high-mass planets would have been more efficient at ejecting planetesimals, leaving behind a depleted population of dust-producing parent bodies. Alternatively, if the planets formed in situ, the timescale for the planet to eject the planetesimals would have been shorter in systems with high-mass planets than with low-mass planets. Under both scenarios, from an evolution point of view, one would expect to find a positive correlation between low-mass planets and the presence of a remnant dust-producing planetesimal disk and, in fact, preliminary analyses of the Herschel surveys have found tentative evidence of such correlation. However, our clean sample does not confirm the presence of this correlation. Why? It could be because the true migration histories of the systems studied may be significantly more complicated than the two scenarios described above; for example, in our own solar system, it is now well established that the ice giants, Uranus and Neptune, migrated outward over a significant distance to reach their current locations, sculpting the trans-Neptunian population as they did so. Another explanation could be because the planets detected by radial velocity surveys and the dust observed at 100 μm occupy well-separated regions of space, limiting the influence of the observed closer-in planets on the dust production rate of the outer planetesimal belt. But it could also be that our sample is too small to detect such a correlation because having a clean sample that avoids the biases mentioned above comes at a price: in our sample, a positive detection of a correlation could have been detected only if the disk frequency around low-mass planet stars were to be about four times higher than the control sample.

Another aspect that we have explored is the role of planet multiplicity. Dynamical simulations of multiple-planet systems with outer planetesimal belts indicate that there might be a correlation between the presence of multiple planets and debris. This is because the presence of the former indicates a dynamically stable environment where dust producing planetesimals may have survived for extended periods of time (as opposed to single-planet systems that in the past may have experienced gravitational scattering events that resulted in the ejection of other planets and dust-producing planetesimals). However, our sample does not show evidence that debris disks are more or less common, or more or less dusty, around stars harboring multiple-planet systems compared to single-planet systems.

And how do the observed debris disks compared to our solar system? Because our sample does not show any evidence of disk evolution in Gyr timescales, we can look at the distribution of disk fractional luminosities (Ldust/Lstar; a distance-independent variable). We find that a Gaussian distribution of fractional luminosities in logarithmic scale centered on the solar system value (taken as 10-6.5) fits the data well, whereas one centered at 10 times the solar system’s debris disks can be rejected. This is of interest in the context of future prospects for terrestrial planet detection. Even though the Herschel observations presented in this study trace cold dust located at tens of AU from the star, for systems with dust at the solar system level, the dust dynamics is dominated by Poynting–Robertson drag. This force makes the dust in the outer system drift into the terrestrial-planet region. This warm dust can impede the future detection of terrestrial planets due to the contaminant exozodiacal emission. Ruling out a distribution of fractional luminosities centered at 10 times the solar system level implies that there are a large number of debris disk systems with dust levels in the KB region low enough not to become a significant source of contaminant exozodiacal emission. Comets and asteroids located closer to the star are other sources of dust that can contribute to the exozodiacal emission (and for those, Herschel observations do not provide constraints), but planetary systems with low KB dust-type of emission likely imply low-populated outer belts leading to low cometary activity. These results, therefore, indicate that there are good prospects for finding a large number of debris disk systems (i.e., systems with evidence of harboring planetesimals) with exozodiacal emission low enough to be appropriate targets for terrestrial planet searches.

Larger samples are needed to improve the statistics of the studies mentioned above, but, as we have done here, care must be taken to avoid biases. But increasing the sample size is not enough. There are two additional aspects that need to be improved upon and, with the data at hand, cannot be addressed at the moment: our ability to detect fainter debris disks (as we may only have detections for the top 20% of the dust distribution), and to detect or rule out the presence of lower-mass planets to greater distances. For the later, of critical importance is that the planet search teams make the non-detections publicly available so we can identify systems for which the presence of planets of a given mass can be excluded out to a certain distance.

For more info see Moro-Martín et al. 2015, ApJ, 801, 143 and references therein.


Nov 152015

By Rachel Osten, AURA Astronomer at STScI

A stellar flare is the release of energy that occurs during a magnetic reconnection event in the upper atmosphere of the star.  Flares involve the release of substantial amounts of ionizing radiation, and are the most dramatic forms of energy release that cool stars will undergo during their time on the main sequence. Stellar flares appear to be an extension of the same phenomena observed in great detail on our own star, the Sun, despite the large difference in the energies involved between the two: the largest solar flares have energies of about 1032 erg, while stellar flares can be a thousand to a million times more energetic. From the solar perspective, flares are part of a triad: the flare, a coronal mass ejection, and highly energetic particles. Even though they all contribute to space weather, the latter two are the most important for determining how damaging a solar eruptive event might be. These are also the two for which we have the least some constraints on occurrence in a stellar context.


Figure: Artist’s conception of a large flare observed on a nearby M dwarf flare star.  The flare involves all layers of the star’s atmosphere, from the photosphere to the chromosphere to the tenuous, hot corona.  These large stellar flares may be accompanied by coronal mass ejections, which would affect the stellar environment around active stars.

While stellar flares have been studied for decades — non-solar stellar flares were first described by Ejnar Hertzsprung in a 1924 article entitled “Note on a peculiar variable star or nova of short duration” — it has been the discovery of planets around other stars which has provided much recent astrophysical motivation for studying stellar flares.  Several recent papers (Segura et al. 2010, Khodachenko et al. 2007) have speculated about the possible impacts on close-in exoplanets of flares and associated events like accelerated particles and coronal mass ejections.  The results suggested that the flares themselves were not the main worrisome aspect of magnetic activity that could affect habitability; rather, it was the other “messengers”, namely the coronal mass ejection and the energetic particles, which had drastic implications for habitability. In the Segura paper, the energetic particles accompanying the flare were responsible for widespread destruction of the ozone layer in the atmosphere of a  terrestrial exoplanet in the habitable zone, which took several years to recover to the pre-flare levels. In the Khodachenko paper, the frequent coronal mass ejections acted like an enhanced stellar wind, compressing the planetary magnetosphere and exposing the exoplanet’s atmosphere to enhanced levels of ionizing radiation, which can cause atmospheric loss.

Stellar flares are a multi-wavelength phenomenon. They involve all layers of the star’s atmosphere, from the photosphere to the magnetically heated chromosphere and corona, and involve a variety of physical processes, including plasma heating and particle acceleration.  Stellar flares are typically observed in a piecemeal fashion, usually covering only one wavelength region, and for short periods of time.  Imagine how little we would know about the Sun’s flares if we were limited to observing it for only a few days every year or so. And that’s if your proposal got accepted!  Even missions like Kepler, which stared at one patch of sky for multiple years and observed many thousands of stellar flares, observed in only one wavelength region; having a way to relate the energy released in one bandpass to the energy released by a flare across all wavelength regions, the so-called bolometric flare energy, is important for an intercomparison of heterogeneously observed stellar flares.

Recent work on solar flares has taken a global view of the energetics of large solar eruptive events (Kretzschmar 2011, Emslie et al. 2012). These papers established how radiated energy is partitioned in the dominant mechanisms, namely chromospheric and transition region line emission, continuum emission from a hot blackbody presumed to form in the photosphere, and coronal emission from plasma heated to temperatures in excess of a million degrees.  This solar work has also demonstrated a near equipartition between the total amount of radiated energy (the bolometric flare energy), and the kinetic energy in the associated coronal mass ejection (Emslie et al. 2012, Drake et al. 2013).

My recent paper (Osten & Wolk 2015) unites these two concepts with application to stellar flares.  When considering the potential impact of stellar eruptive events to exoplanets, one needs to “follow the photons”; current astronomical limitations means that we have few options for direct detection of coronal mass ejections, and none for the existence of these very energetic particles expected to be produced in stellar flares as they are in solar flares.  So if we want to gain an initial grasp of the potential influence of stellar eruptive events, the easiest way to do so is to look at the flares from a holistic standpoint, and relate them to coronal mass ejections using a physically motivated way of connecting them.

In order to do the latter — that is, connect the total radiated flare energy with the coronal mass ejection’s kinetic energy — we need a way to “correct” for the flare energy in a given bandpass to the bolometric amount.  Using previously published papers of a few well-observed stellar flares, I established that these flares appear to have similar fractions of energy as aggregates of solar flares (see the Table).  So far, so good.  This is another confirmation of the basic approach, in treating solar and stellar flares as originating from the same basic physical process, despite the orders of magnitude difference in energy release. Being able to relate flares observed in one wavelength range to the total amount of radiated energy released enables a more global intercomparison of those flares.


Table: Fraction of total radiated energy in Solar and stellar flares released in particular bandpasses.  The numbers for solar flares and those for active stars are remarkably similar, and confirm the expectation that the same physical process is occurring. This energy partition allows for a better intercomparison of flares observed in different wavelength regions.

We observe that stellar flares of differing energies have an occurrence that is a power-law in frequency as a function of size.   Relating the kinetic energy of coronal mass ejections to the bolometric energy of an accompanying flare (assuming equipartition, based on solar flare studies) means that we have a way to estimate the cumulative impact of the transient mass loss that is occurring in these flare-related coronal mass ejections. To make a long story short (if you want the long story, read the paper!), we find that this cumulative effect can be large.  We examined published flare frequency distributions from a variety of different types of flaring stars: young solar-mass stars; a 70 MY young solar analog; recently observed superflaring Sun-like stars; young low mass stars; nearby hyperactive M dwarfs with flare frequency distributions measured at both optical and X-ray wavelengths; and inactive early- and mid-M dwarfs.  At the high end, the flare-related transient mass loss may be as much as three orders of magnitude higher than the present-day total solar mass loss rate of 2×10-14 Msun/yr.

So what are the implications of these findings? The main impact is on the stellar environment, not the star itself. The cumulative mass loss rate, though potentially enhanced compared to the Sun’s current total rate of mass loss, is not enough to change the course of the star’s evolution, as the total mass lost is not an appreciable fraction of the total stellar mass. As mentioned before, a high rate of coronal mass ejections means that any close-in exoplanet would be subjected to enhanced compression of its magnetosphere and resulting exposure of its atmosphere to ionizing flare radiation.  The interaction between an enhanced stellar wind and an exoplanet’s magnetic field can generate a magnetospheric field that would have a negative feedback on the planet’s internal dynamo, according to one calculation. Frequent strong CMEs could quench dynamo growth in the planet, leading to weak planetary magnetic fields and a reduced ability to protect the atmosphere from exposure to ionizing radiation. Other impacts on the stellar environment include seeding the planet-forming disk with processed stellar material, or removing material from the debris disk at a later stage in the stellar system’s evolution. Results from young solar analogs suggest that the Sun’s youth was likely marked by such frequent strong CMEs. While these results are suggestive, direct constraints on the existence and occurrence of stellar coronal mass ejections are a vital next step to a fuller understanding of how stars affect their surroundings.



Drake, J. et al. 2013 ApJ 764, 170

Emslie, A.G. et al. 2012 ApJ 759, 71

Hertzsprung, E. 1924 BAN, 2, 87

Khodachenko, M. et al. 2007 Astrobioloy 7, 167

Kretzschmar, M. 2011 A&A 530, A84

Osten, R. A. & Wolk, S. J. 2015

Segura, A. et al. 2010 Astrobiology 10, 751

Oct 152015

By Massimo Stiavelli, Astronomer and JWST Mission Head at STScI.

Google Earth and its related products Google Sky, Google Moon, etc. are incredible tools for real work as well as hours of endless amusement. Unfortunately, the shear amount of imaging data that went into them made it impossible for a human being to check that the combined images were actually correct in all cases. My main field in astronomy is imaging and for every imaging project I have done, from the Hubble Ultra Deep Field to the study of the cores of nearby galaxies, a painstaking pixel-by-pixel inspection of the images has been essential. When you look for rare objects at very high redshift there is little room for error and an image artifact can easily ruin your day and perhaps your reputation. Despite this careful checking of the results of computer processing, I have had my share of objects that have disappeared once better data were obtained. One can never be too careful.

The number of errors in Google Earth is amazingly small but not zero. Any automated imaging combination of heterogeneous data is bound to occasionally fail and produce some non-sense that is immediately recognizable as such by a trained eye. Years ago we had the report of a mysterious island in the Pacific discovered on Google Earth. I checked it out and it didn’t take more than a second to see that it was an image artifact, probably the result of two separate images that had not been appropriately matched or perhaps due to insufficient coverage in that area. The give away was the very dark color and the unphysically sharp boundary. Every imaging system has a finite resolution and if you see something that appears much sharper than other objects in its neighborhood then it is very likely an artifact. So it was for the mysterious island that – as the media reported some time later – “disappeared” in a later version of Google Earth data.

Recently, I came across a report and a video of “alien bases” found on the Moon using Google Moon. I was virtually sure that there were no alien bases there but still decided to check it out. In fact, as soon as the video starts you can see that in the pole-on view of the Moon there is an area that looks like a stripe: it is where imaging data of different quality are mixed and matched (Figure 1). The edges between different data are very unreliable. In this type of situation in a science project one would simply disregard those areas and their immediately surrounding portions of the image. Unfortunately, UFO aficionados are not so easily deterred. Zooming in on the edge between the two types of data they found lots of “alien bases”: most are elongated, very sharp, structures, much sharper in fact than the crater contours visible next to them. One “base” is in fact close to half a crater, a crater that is visible only for half of its circumference, with the other half crater being found in a lower quality lunar image. Clearly all those suspicious features are artifacts and any other interpretation is non-sense.

People are spending time looking at science data. It is in everybody’s interest to make their efforts useful for the progress of science.  By combining carefully chosen problems, training sets, and redundancy, a citizen science experiment can reduce the margin for errors and lead to very interesting results. For example, Planet Hunters (, a recent citizen science project based on Kepler Space Telescope data, was able to identify a very peculiar star that seems to be surrounded by a lot of large debris, perhaps comets or remnants of a planetary collision. These results have led to a scientific publication (Boyajian et al. 2015, submitted to MNRAS and, also mentioned on the Atlantic in an article by Ross Andersen). Another very successful citizen science project has been Galaxy Zoo ( that has led to 48 science papers. Circumstellar disks are the target of the NASA-funded citizen project is Disk Detective (, with the goal of identifying in the WISE data stars that may host or may be forming extrasolar planets. Citizen science has the potential to engage many more people than it does today and tap on a large reservoir of talent available in the public. If we succeed in doing so we will have better science and perhaps fewer non-sensical news reports.


Figure 1. When loading the Moon in Google Earth the stripe with data-matching issues is immediately visible. Along the edges of that stripe finding artifacts is very easy (Image credit: Google Earth).

Sep 162015

By Andrea Banzatti, Postodctoral Fellow at STScI

Antoine de Saint-Exupéry, the world-famous writer of The Little Prince, served as an aviator for the French Aéropostale over the Paris-Dakar route, crossing the African desert for many years in the 1930s. In his book Wind, Sand and Stars he reports of having once perceived the forthcoming conflagration of a sand storm before taking off from an intermediate station in the Sahara, by observing the peculiar behavior of two dragonflies. “What filled me with joy was that I had understood a murmured monosyllable of this secret language, that I had been able to read the anger of the desert in the beating wings of a dragonfly.” This joy is not unknown to the astrophysicist. Here is a story.

Over the last year, I happened to be working on a survey of infrared emission spectra of carbon monoxide (CO) observed in young “protoplanetary” disks, the birthplaces of the plethora of exoplanets detected so far.  The CO molecule is generally abundant in planet-forming regions, at disk radii comprised to within approximately 10 Astronomical Units (AU) from the central star. CO had been observed in disks for over thirty years [1], and recent instrumental developments had made possible to perform a survey of unprecedented sensitivity, spectral resolution, and sample size in the years 2007-2010 [2]. While studying the peculiar flickering behavior of CO and water emission from the disk of a variable star, I noticed that the CO spectra looked like the superposition of two emission line components, one being distinctly broader than the other [3]. I attempted a spectral decomposition analysis, encouraged by the exquisite quality of the data, and found that while many protoplanetary disks showed both CO components, some had only the narrow one [4]. By measuring the temperature (from the line flux ratios) and the disk radius of CO emission (from the line widths) in each disk of the survey, I composed the diagram shown below. When I and Klaus Pontoppidan, my collaborator and mentor, looked at it, we were astonished by the appearing of a sequence.


Figure (click to enlarge): The temperature-radius (T-R) diagram of rovibrational CO emission in disks [4]. The red and blue data points are individual disks from high quality, high spectral resolution surveys done with CRIRES at the VLT (resolving power of ~100,000) [2,5]. The sample spans a range in stellar masses of 0.5-3 solar masses (indicated by the symbol size). The location of each disk in the diagram indicates the vibrational temperature of the innermost CO gas present in the disk. At the bottom of the figure, for comparison, are shown the Solar System planets, together with the distribution of semi-major axes of observed exoplanets with Msini > 0.5 Jupiter masses [6].

Given its high dissociation temperature, CO traces the innermost disk radius where molecular gas can survive in any disk. Therefore, the location of each disk in the diagram indicates the temperature of the innermost molecular gas present in its planet-forming region. The red disks in the diagram are those found to have two CO components and are identified as “primordial”, where the inner radius is set by the stellar magnetospheric accretion or by dust sublimation (truncating the disk out to ~0.1 AU at most for the whole sample). Blue disks lack the broad CO component, and have something else going on preventing CO gas from extending all the way to the smallest distance allowed by the stellar properties…

As the CO emission analyzed here is rovibrational, the measured line ratios give a vibrational temperature, which is a sensitive thermometer of the local radiation field. The temperature-radius (T-R) diagram, taken as a whole, reveals a sequence composed of two regimes. In the inner 0.03-2 AU the temperature decreases as a power-law profile, as expected for the dust temperature in models of inner disks irradiated by the central star. This regime is identified as due to infrared pumping of CO by the local warm dust, and provides an empirical temperature profile for inner disks around solar-mass stars. The second regime takes over beyond ~2 AU, and shows an inversion in the temperature. This temperature inversion strongly points at another excitation mechanism that is known to effectively populate CO lines in low-density and cold environments: ultraviolet (UV) fluorescence [7]. In order for UV radiation to be effective at such large distances from the central star (2-20 AU), the innermost region of these disks must be largely depleted in both dust grains and molecular gas. These disks must host large inner gaps in their radial structure. Overall, CO emission suggests that all blue disks are developing or have developed large inner gaps, and some of them (filled symbols in the figure) have already been identified as “transitional” by dust emission modeling or by direct imaging. The T-R diagram has the power to provide prime targets for direct imaging campaigns, pushing inward the detection of inner gaps to radii that will become accessible to future infrared imagers (e.g. by E-ELT-METIS [8]).

But the best is yet to come. This research provides an empirical framework to investigate gap-opening processes in disks, including planet formation and migration. Comparison of the CO temperature sequence to the distribution of giant exoplanets detected so far reveals two interesting facts. The so-called “hot”-Jupiters are found at the innermost radial location of CO gas in disks, ensuring that abundant gas is present to allow gas-supported planet migration as proposed by models [9]. The distribution of exo-Jupiters, instead, rises at the break point between the two regimes in the CO diagram, supporting the existence of a link between exo-Jupiters formation and the opening of gaps in the natal disks [10], which eventually leads to their dispersion through the “debris disk” phase [11]. The journey of an exoplanet from its birth is long and can be full of surprises, ending up in the large diversity suggested by the foremost research in planetary architectures and compositions [e.g. 12]. And for us, at the horizon, stands the possibility of finding something similar to what we know here on Earth, a journey that is breathtaking for our entire world. It may still be far ahead in time, but every word we catch of this secret story of nature is welcomed with joy by those who spend their lives aspiring to hear it in full. Sometimes, these words are found in the most unexpected data, or in a diagram composed almost by chance. Sometimes, we can understand a murmured monosyllable of this secret language simply by “following the wings of a dragonfly”.



  1. Najita et al., 2000
  2. Pontoppidan et al., 2011
  3. Banzatti et al. 2015
  4. Banzatti & Pontoppidan, 2015
  5. van der Plas et al., 2015
  7. Brittain et al. 2007
  8. Bradl et al. 2010
  9. Kley & Nelson, 2012
  10. Zhaohuan et al. 2011
  11. Wyatt et al. 2015
  12. Howard, 2013
Aug 012015

By Camilla Pacifici, Postdoctoral Fellow at STScI

A main caveat in current statistical studies of galaxies at z ~ 1 is that the way in which the physical properties of galaxies, such as stellar mass (M∗) and star formation rate (SFR), are generally derived from multi-wavelength datasets does not reflect recent advances in the modeling of galaxy spectral energy distributions (SEDs). For example, spectral analyses often rely on oversimplified modeling of the stellar spectral continuum using simple star formation histories (SFHs), such as exponentially declining τ-models. Some studies have shown that more sophisticated SFH parametrizations provide better agreement with the data (e.g. Lee et al. 2010, 2014; Pacifici et al. 2013; Behroozi et al. 2013). The inclusion of nebular emission is also important to interpret observed SEDs of galaxies. Elaborate prescriptions have been proposed, based on combinations of stellar population synthesis and photoionization codes. We have investigated, in a systematic way, how different SED modeling approaches affect the constraints derived on the physical parameters of high-redshift galaxies.

We used version 4.1 of the 3D-HST Survey photometric catalogue for the GOODS-South field covering an area of 171 arcmin2 (Skelton et al. , 2014). We compiled a sample of 1048 galaxies at redshifts 0.7 < z < 2.8 (H < 23) with accurate photometry at rest-frame UV to near-IR wavelengths (U, ACS-F435W, ACS-F606W, ACS-F775W, ACS-F850lp, WFC3-F125W, WFC3-F140W, WFC3-F160W and IRAC 3.6μm). Grism (low-resolution) observations provided us with reliable spectroscopic redshifts (Brammer et al. 2012) for all galaxies and good optical emission-line equivalent widths (EW) for a subsample of 364 galaxies.

We considered three modeling approaches relying on different assumptions: the explored (prior) ranges of star formation and chemical enrichment histories; attenuation by dust; and nebular emission. We built:

  • A classical spectral library (CLSC) adopting exponentially declining SFHs, fixed stellar metallicities, a two-component dust model with fixed parameters, and no nebular emission;
  • A sophisticated spectral library (P12nEL) adopting star formation and chemical enrichment histories from cosmological simulations, a two-component dust model with variable parameters, and no nebular emission;
  • The same P12nEL spectral library, only including also the nebular component (P12, Pacifici et al. 2012).

In Figure 1, we compare the observer-frame colors of the galaxies in the sample (grey symbols) with the predictions of the three model spectral libraries (colored contours). This figure shows that the CLSC spectral library leaves few observed galaxies with no model counterpart. Thus, SED fits for these galaxies will be biased towards the models that lie at the edge of the spectral library. The P12nEL spectral library can cover reasonably well the bulk of the observations, showing the importance of accounting for more realistic ranges of SFHs and dust properties than included in the CLSC spectral library. Few observed galaxies fall outside the contours of the P12nEL model spectral library, presumably because of the contamination of the WFC3-F160W flux by strong Hα emission. The P12 spectral library allows us to cover reasonably well the entire observed color-color space.

Figure 1: Optical-NIR color-color diagrams comparing the 3D-HST sample (grey symbols; open circles mark objects for which error bars are larger than 0.2 mag) with the three model libraries as labeled on top (contours; the three lines mark 50, 16 and 2 per cent of the maximum density).

We compared the constraints on M∗ and SFR derived for all 1048 galaxies in the sample using the CLSC and P12nEL model spectral libraries to those obtained using the more comprehensive P12 library. We summarize the results in Table 1. The use of simple exponentially declining SFHs (CLSC spectral library) can cause strong biases on both the M∗ (~ 0.1 dex) and the SFR (~ −0.6 dex). Not including the emission lines in the broad-band fluxes (P12nEL) does not strongly affect the estimates of M∗, but can induce an overestimation of the SFR (~ 0.1 dex).

Table 1: 16, 50 and 84 percentiles of the distributions of the differences between best estimates of M∗ and SFR when comparing the constraints obtained with different libraries.



To further quantify how emission lines contaminate observed broad-band fluxes, we recorded, for a subsample of the 3D-HST galaxies (364), the contribution by nebular emission to the WFC3-F140W magnitude of the best-fitting P12 model (as derived when including the constraints from both 9-band photometry and EW measurements from 3D-HST grism data). This is shown in Figure 2 as a function of stellar mass. For galaxies at redshifts 0.8 < z < 1.4, both Hα and [S II] fall in the WFC3-F140W filter. Crosses represent the combined contamination caused by these lines. In the same way, we plotted the combined contamination by Hβ and [OIII] for galaxies in the range 1.5 < z < 2.2 (empty circles). Each galaxy is color-coded according to star formation activity, from low (red) to high (black) specific SFR. For galaxies at 0.8 < z < 1.4, the contamination decreases from ~ 0.1 to ~ 0.02 mag as the stellar mass increases from ~ 109.5 to ~ 1011 solar masses. At higher redshift, where [O III] and Hβ are sampled in the band, the contamination is slightly larger because the SFR is on average larger at higher than at lower redshifts (Noeske et al. 2007).

Figure 2: Contribution by nebular emission to the WFC3-F140W magnitude of the best-fitting (P12) model for a subsample of 3D-HST galaxies (for which good grism observations are available) as a function of stellar mass. Crosses represent galaxies with measured Hα and [S II] (0.8 < z < 1.4), while circles represent galaxies with measured Hβ and [O III] (1.5 < z < 2.2). The points are color-coded according to star formation activity from low (red) to high (black) specific SFR. The contamination of the emission lines in the broad-band WFC3-F140W filter increases from ~ 0.01 to ~ 0.5 mag as the stellar mass decreases.

The results obtained in this work revealed the importance of choosing appropriate spectral models to interpret deep galaxy observations. In particular, the biases introduced by the use of classical spectral libraries to derive estimates of M∗ and SFR can significantly affect the interpretation of standard diagnostic diagrams of galaxy evolution, such as the galaxy stellar-mass function and the main sequence of star-forming galaxies. In this context, the spectral library developed by Pacifici et al. (2012) offers the possibility to interpret these and other fundamental diagnostics on the basis of more realistic, and at the same time more versatile models. This is all the more valuable in that the approach can be straightforwardly tailored to the analysis of any combination of photometric and spectroscopic observations of galaxies at any redshift.


Behroozi, P. S., Wechsler, R. H. & Conroy, C., 2013, ApJ, 770, 57
Brammer G. B. et al., 2012, ApJS, 200, 13
Lee S.-K., Ferguson H. C., Somerville R. S., et al., 2010, ApJ, 725, 1644
Lee S.-K., Ferguson H. C., Somerville R. S., et al., 2014, ApJ, 783, 81
Pacifici, C., Charlot, S., Blaizot, J., & Brinchmann, J., 2012, MNRAS, 421, 2002
Pacifici, C., Kassin, S. A., Weiner, B. et al., 2013, ApJ, 762, L15
Skelton, R. E., Whitaker, K. E., Momcheva, I. G., et al., 2014, ApJS, 214, 24


Jul 012015

By Marco Chiaberge, ESA Astronomer at the Space Telescope Science Institute

One of the most important problems in modern astrophysics is to understand the co-evolution of galaxies and their central supermassive black holes (SMBH) (see e.g. Heckman & Best 2014 for a recent review). Since the matter that ultimately accretes onto the central black hole needs to lose almost all (~99.9%) of its angular momentum, studies of mergers, tidal interactions, stellar bars and disk instabilities are central for understanding the details of such a process. However, observational efforts to assess the importance of mergers in Active Galactic Nuclei (AGN) so far have led to conflicting results. A major issue is related to the so-called radio-loud/radio-quiet dichotomy of active nuclei. Radio-loud AGNs have powerful relativistic plasma jets that are launched from a region very close to the central SMBH. The most popular scenario among those proposed so far assumes that energy may be extracted from the black hole via the innermost region of a magnetized accretion disk around a rapidly spinning black hole Blandford & Znajek (1977). In such a framework, the radio-quiet/radio-loud dichotomy can be explained in terms of a corresponding low/high black hole spin separation (Blandford et al. 1990). It is also important to stress that radio-loud AGN are invariably associated with central black holes of masses larger than ~108 solar masses (e.g. Chiaberge & Marconi 2011). Therefore, the black hole mass must play a role.

With the aim of determining the importance of mergers in triggering different types of AGN activity, my collaborators and I selected 6 samples of both radio-loud (RLAGN) and radio-quiet (RQAGN) AGN, and of non-active galaxies matched to the AGN samples in magnitude (or stellar mass). We focused in particular on redshifts between z=1 and z=2.5. All objects were observed with HST/WFC3-IR at 1.4 or 1.6mm, in order to ensure appropriate sensitivity at rest-frame optical wavelengths, and to allow us to detect faint signatures of a merger event. Most of the objects were taken from large surveys performed with Hubble (CANDELS, 3D-HST). The images of the high-luminosity radio galaxies were taken from a “snapshot” program we performed as part of our 3CR-HST survey of radio-loud AGN.



Figure 1 HST/WFC3-IR images of 3 radio-loud AGNs (top, credits: NASA, ESA, M. Chiaberge) and 4 radio-quiet AGNs (from CANDELS, Koekemoer et al. 2011). The 3 radio-loud AGNs are all classified as mergers. Only the 2 radio-quiet AGN in the right-most panels are mergers.





The fun part of the work was to visually inspect the WFC3-IR image of each of the 168 objects (Fig. 1) and determine whether each object was or was not showing signatures of a merger, according to a pre-defined classification scheme. It was at that point that we found something really interesting. Without knowing what type of object we were looking at, we classified almost all (95%) of the RLAGNs as “mergers”. On the other hand, the RQAGN samples and the non-active galaxies had merger fractions between 20% and 37%.  We performed a careful statistical analysis of the results, and we concluded that the merger fraction in RLAGN is significantly higher than that in RQAGNs and non-active galaxies (Fig. 2). It is possible that all RLAGNs are associated with mergers. This result was also confirmed for lower redshift samples of radio galaxies (one at z~0.5 and one at z<0.3), and for objects of both low and high power. On the other hand, the merger fraction of RQAGNs is statistically not different from that of non-active galaxies.

blog_fig2Figure 2 Merger fraction vs. average radio loudness parameter Rx (ratio of the radio to X-ray luminosity) for the different AGN samples. Radio-quiet AGNs are on the left of the dashed line, radio-loud AGN are on the right. The filled symbols are the radio-loud samples and the empty symbols are radio-quiet. The dashed line represents the radio loudness threshold for PG QSOs. The solid line marks the 60% merger fraction that appears to roughly separate radio-loud and radio-quiet samples.





This result has very important implications. Firstly, it shows a clear association between mergers and AGN with relativistic jets (the RLAGN subclass), with no dependence on either redshift or luminosity. Secondly, we firmly determined that not all AGNs are triggered by mergers. The question now is how do mergers trigger AGN with jets? A possible scenario we envisage is that when a galaxy merger happens, the central supermassive black holes merge as well. In general, the resulting spin of the BH after coalescence is lower than the original spin values. But for particular spin alignments and for BHs of similar masses, the spin can be significantly higher (see Schnittman 2013, for a recent review). In that case, if the mass of the BH is at least ~108 solar masses, the energy extracted through the Blandford-Znejek mechanism may be large enough to power the jet. This is not a completely new idea, since it was already proposed in a slightly different form by Wilson & Colbert (1995).  In the near future, we will focus on confirming the strong connection between RLAGNs and mergers with a larger dataset of HST observations, ALMA observations, and integral-field spectroscopy.

These results have been published in an ApJ paper, in an ESA press release, and in a Nature “News and Comments” article.



Blandford R. D., Netzer H., Woltjer L., Courvoisier T. J.-L. and Mayor M. 1990 Active Galactic Nuclei, Vol. 280 (Berlin, Heidelberg, New York: Springer)

Blandford R. D. and Znajek R. L. 1977 MNRAS 179 433

Chiaberge M. and Marconi A. 2011 MNRAS 416 917

Heckman T. M. and Best P. N. 2014 ARA&A 52 589

Koekemoer A. M., Faber S. M., Ferguson H. C. et al 2011 ApJS 197 36

Schnittman J. D. 2013 CQGra 30 244007

Wilson A. S. and Colbert E. J. M. 1995 ApJ 438 62

Jun 012015

By Jacqueline Radigan, Giacconi Fellow at STScI

Formed like stars, but not massive enough to fuse hydrogen, brown dwarfs straddle the boundary between low mass stars and giant planets.  Without a reservoir of nuclear fuel they spend their lives cooling.  Consequently, the atmospheres of old and cold field brown dwarfs are higher surface gravity analogs to those of extrasolar planets.  However, observations of free-floating brown dwarfs carry a key advantage over those of planets:  they can be directly imaged without being washed out by the light of a parent star.  As a result, the spectra obtained of nearby field brown dwarfs in the solar neighborhood are more numerous and of higher quality than their exoplanet counterparts.  In fact the coolest brown dwarf yet discovered (Luhman 2014) has a temperature of only ~250 K (below the freezing point for water!), making it the coolest directly imaged object outside of our own Solar System.

At temperatures below ~2500 K (L spectral types), refractory species in brown dwarf atmospheres begin to condense to form exotic “dust’’ clouds (Burrows & Sharp 1999; Lodders 1999; Burrows et al. 2006).  These first condensates are made of metal-oxides, silicate grains, and liquid iron.  Further cooling leads to grain growth and the eventual settling of cloud layers below the visible photosphere. Thus at temperatures below ~1200 K, dust clouds have largely disappeared from view (T spectral types).



Figure 1.  Reproduced from Radigan 2014.  Color magnitude diagram of M, L and T dwarfs with known parallaxes (small circles) from the database of T. J. Dupuy.  The points are divided by color into the spectral type bins used by Radigan et al. 2014:  L-dwarfs (≤L8.5) in red, T-dwarfs (≥T4) in purple, and L/T transition dwarfs (L9–T3.5) in dark and light blue. Within the L/T transition bin T0–T2.5 dwarfs are shown in dark blue, while objects with L9–L9.5 and T3–T3.5 spectral types (the endpoints) are plotted in a lighter color to illustrate the degree of overlap with the L-dwarf and T-dwarf bins. Two L/T transition objects which are over-luminous for their spectral types (L9 and T0) are noted as possible binaries.


The brown dwarf spectral sequence, especially the transition from cloudy L- to clear T- spectral types is challenging to understand owing to the presence of clouds, complex chemistry, and weather.  These complexities are highlighted by perplexing behavior observed at the L/T transition: at wavelengths around ~1 micron (near the peak of the spectral energy distribution) brown dwarfs briefly become brighter as they cool (Figure 1;  Dahn et al. 2002; Tinney et al. 2003; Vrba et al. 2004; Dupuy & Liu 2012; Faherty et al. 2012).  This phenomenon, which has proved notoriously difficult for modelers to reproduce (e.g. Marley et al. 2002, Allard et al. 2003, Burrows et al. 2006), is likely related to the mechanism of cloud dispersal at the L/T transition, and illustrates the limiting role clouds play in our understanding of cool atmospheres.

Dust cloud “weather” has long provided the most promising explanation for the counter-intuitive L/T transition brightening (Ackerman & Marley 2001, Burgasser et al. 2002) .  The idea is that as dust clouds settle deeper in the atmosphere they begin to intersect with the dynamic troposphere, leading to the development of “holes”, or lower opacity sight-lines into the deep photosphere. Consequently, flux from warmer atmospheric layers is able to escape, and would explain the abrupt brightening seen at 1 micron wavelengths, where clouds are a dominant opacity source. This hypothesis makes a straightforward prediction: L/T transition brown dwarfs will have patchy atmospheres and should therefore exhibit variability on rotational timescales.

We tested this hypothesis by observing 56 brown dwarfs with mid-L to T spectral types over 60 nights using the Wide Field Infrared Camera (WIRC) on the Du Pont 2.5 m telescope at the Las Campanas Observatory (Radigan et al. 2014).  Each target was observed continuously in the J-band (wavelengths of ~1.2 microns) for ~3-5 hours (a large fraction of a rotation period), along with several reference stars that happen to fall in the camera’s field of view.  Differential photometry for the target and reference stars were computed to produce light curves. Using data for nearly 800 (presumably non-variable) reference stars we determined an empirical criterion for significant variability (with an expected 1% rate of false positives). Using our empirically calibrated criterion we found 9 of our 56 targets to be significantly variable (figure 2).

lcs_sigFigure 2. Reproduced from Radigan et al. 2014.  Light curves of significantly variable (p > 99%) targets in our sample. Light curves are shown in the upper panels, and the light curve of a similar-brightness comparison star observed simultaneously is shown in the lower panels. All data shown were obtained using WIRC on the Du Pont 2.5 m telescope except for those of 2M0758+32, which were obtained using WIRCam on the Canada-France-Hawaii Telescope. All light curves have been binned down from their original cadence by factors of 3–7.

The spectral types, colors, and amplitudes of our variable targets are shown in figure 3.  The high-amplitude variables are clustered directly at the L/T transition, providing the first direct evidence in support of cloud break-up. However, since not all L/T transition objects are highly variable, there may be more to the story. While large contrast between clouds and clearings produce large variability for some L/T transition brown dwarfs, how do we explain 1-micron brightening for the larger non-variable population? Some objects with patchy clouds may not be variable due to azimuthal symmetry, small scale features, or long rotation periods.  Other supposed L/T transition objects may in fact be unresolved binaries with components outside the transition. Are these factors sufficient to explain all the non-variables?  In the published paper we concluded: maybe.  So while it is clear that large variability is more frequent at the L/T transition, more data are required to pin down a causal relationship between cloud break-up and 1 micron brightening.


Figure 3. Reproduced from Radigan et al. 2014.  NIR spectral type vs. 2MASS J − Ks color for all targets observed in our program. Gray points show the population of known field L and T dwarfs with J <16.5. Purple circles show detections, with the linear symbol size is proportional to the peak-to-peak amplitude of variability detected (ranging from 0.9%–9% for the smallest to largest symbols). A gray dashed line encircles the objects considered part of the L/T transition sample.  Our average sensitivity or completeness to sinusoidal signals of a given peak-to-peak amplitude is approximately equal in all spectral type bins.

Variable brown dwarfs, and notably the new population of high-amplitude variables at the substellar L/T transition, provide novel opportunities to constrain cloud properties and dynamics in cool atmospheres. Surface variations in cloud thickness produce a chromatic variability signature, providing an unprecedented opportunity to probe cloud structure via multi-wavelength monitoring as in e.g. Radigan et al. (2012) , Buenzli et al. (2012), Apai et al. (2013).  Finally, mapping the evolution of features over multiple rotations will provide a way to study atmospheric dynamics in the high-gravity, non-irradiated regime.

Allard, F., Guillot, T., & Ludwig, H. et al. 2003, Brown Dwarfs (IAU Symp. 211), ed. E. Martín (Cambridge: Cambridge Univ. Press), 325
Ackerman, A. S. & Marley, M. S. 2001, ApJ, 556, 872
Apai, D., Radigan, J., & Buenzli, E. et al. 2013, ApJ, 768, 121
Buenzli, E., Apai, D., & Morley, C. V. et al. 2012, ApJL, 760, L31
Burgasser, A. J., Marley, M. S., & Ackerman, A. S. et al. 2002b, ApJL, 571, L151
Burrows, A. & Sharp, C. M. 1999, ApJ, 512, 843
Burrows, A., Sudarsky, D., & Hubeny, I. 2006, ApJ, 640, 1063
Dahn, C. C., Harris, H. C., & Vrba, F. J. et al. 2002, AJ, 124, 1170
Dupuy, T. J. & Liu, M. C. 2012, ApJS, 201, 19
Faherty, J. K., Burgasser, A. J., & Walter, F. M. et al. 2012, ApJ, 752, 56
Lodders, K. 1999, ApJ, 519, 793
Luhman, K. L. 2014, ApJL, 786, L18
Marley, M. S., Seager, S., & Saumon, D. et al. 2002, ApJ, 568, 335
Radigan, J., Jayawardhana, R., & Lafrenière, D. et al. 2012, ApJ, 750, 105
Radigan, J., Lafrenière, D., Jayawardhana, R., & Artigau, E. 2014, ApJ, 793, 75
Radigan, J. 2014, ApJ, 797, 120
Tinney, C. G., Burgasser, A. J., & Kirkpatrick, J. D. 2003, AJ, 126, 975
Vrba, F. J., Henden, A. A., & Luginbuhl, C. B. et al. 2004, AJ, 127, 2948

Apr 012015

By John Debes, ESA/AURA astronomer at STScI

It is quite the April Fool’s prank to hide in plain sight, and that’s just what the rather pedestrian-appearing WD 1242-105 did for several years, masquerading as a single white dwarf.  The secret it held, and the story of how my team of researchers unraveled it, makes for a useful lesson in how science sometimes progresses—not always by careful predictions, but sometimes by serendipity.  Furthermore, our discovery also demonstrates how astronomy often requires close collaboration.

WD 1242-105 (See Figure 1), resides near the constellation Virgo, and was first discovered as part of the large Palomar-Green Survey of UV-excess sources (Green et al., 1986).  In that survey, it was promptly misclassified as a subdwarf star.  Sixteen years later, Salim & Gould (2002) recognized that this might be a white dwarf candidate based on its apparent motion on the sky—it was large compared to its faint apparent magnitude.  This is often how new white dwarfs are discovered, since astrophysical objects closer to the Earth have larger apparent motions, and white dwarfs are intrinsically faint.


Figure 1: A false color image of WD 1242-105 (center), which is a rather inconspicuous star.  The hint of its high motion on the sky is given by the slight blue/red color–this is due to the fact that the binary had moved between when the two photographs of this star were taken.

Surprisingly, low-resolution spectra were taken and did not detect anything unusual about the white dwarf (Kawka et al. 2004; Kawka & Vennes 2006).  Based on this spectrum, it appeared to be a relatively close, single white dwarf that was about 75 lightyears away.  By a quirk of fate, there was a large high resolution spectroscopic survey of white dwarfs looking for binary objects and using the Very Large Telescope, but this particular white dwarf was not included.

This is where I came in.  As a postdoctoral researcher within the Department of Terrestrial Magnetism (DTM) at the time, I had access to the premiere high resolution spectrograph at Las Campanas Observatory’s Magellan Telescope, called MIKE.  I was using it to survey as many southern hemisphere white dwarfs as I possibly could for traces of metals in their atmospheres.  Since no one had yet published a high resolution spectrum of WD 1242-105, and it appeared to reside within the solar neighborhood, it was a perfect target.  I consequently observed the star three times, each exposure separated by ten minutes.

When I finally got around to analyzing the spectrum of WD 1242-105 a few months later, I was in for a shock.  Instead of the usual single spectral line of Hydrogen (See Figure 2) one sees in white dwarf spectra, there were two separate lines—indicative of a binary system consisting of two hydrogen white dwarfs.  By a sheer stroke of luck, I had caught the binary when both stars were at their maximal relative velocities to our line of sight—the lines were separated by almost 6.5 Angstroms, or 300 km/s.  They were moving so fast, that I could detect changes in their radial velocities on the timescale of each exposure, or ten minutes!


Figure 2: (left) A comparison between synthetic model white dwarf spectra (red lines) with the MIKE observations of WD 1242-105.  The two Hydrogen line components are most visible around the Hα spectral line. (right) A comparison between our model photometry and the observed photometry of this star from optical to mid-IR wavelengths.

Over the next year, I prepared to take more spectra, but by this time I needed to call in some favors.  This binary could have been a progenitor of the famous Type Ia supernovae, and so far not many convincing progenitors have been found.  Many of my colleagues at DTM also used MIKE, and I asked for help in taking additional spectra.  Others were experienced at taking high precision time series photometry of stars, and so I asked them to monitor WD 1242-105.  Finally, another friend of mine was working on a project to measure the parallaxes of nearby stars for evidence of planetary systems—I enlisted that team’s help to measure the distance to this interesting binary.

In the end we had a complete picture for this system’s binary parameters, which enabled us to estimate the individual white dwarf masses (See Figure 3).  We also were able to estimate the masses of the two components using the hydrogen lines themselves as tracers of the white dwarfs’ temperatures and gravities.  Since this was a binary, the distance obtained from our parallax measurement was actually about 60% larger than originally assumed, or 40 pc  (130 light years).  The binary has an orbital period just shy of 3 hours, and a total mass of 0.9 M.  While its mass makes it too light to be a progenitor of a Type Ia supernova, the two components of WD 1242-105 will merge within the next 800 million years in what will no doubt be a remarkable show.  Binaries like WD 1242-105 are believed to be the progenitors of  objects such as R Corona Borealis stars, which are believed to arise from the merger of two white dwarfs.


Figure 3: Radial Velocity curve of the WD 1242-105 double degenerate binary system.  Red points are derived from the less massive component, while the blue points represent the more massive white dwarf.  The second panel shows the residuals after subtracting off the orbital fit to the data.

Since it is so close to Earth, it is one of the largest known sources of expected gravitational wave radiation at frequencies within the mHz range.  Unfortunately, this is too low a frequency for detection with some proposed space-based future gravitational wave observatories such as eLisa, but nevertheless, it will “shine” brightly in the gravitational wave sky, despite its rather bland optical appearance.

Sometimes, the most interesting things are hiding in plain sight, and they just require the right approach or instrument to discover them.



The paper related to WD 1242-105 can be found at this link.

Other references:

Green, R. F., Schmidt, M., & Liebert, J. 1986, ApJS, 61, 305

Kawka, A. & Vennes, S. 2006, ApJ, 643, 402

Kawka, A., Vennes, S., & Thorstensen, J. R. 2004, AJ, 127, 1702

Salim, S. & Gould, A. 2002 ApJ, 575, L83


You can follow John Debes on twitter, @JohnDebes

Mar 022015

By Olivia Jones, Postdoctoral Fellow at STScI

The evolution of dust in galaxies is intrinsically linked with that of the stars, with the formation of new grain material coinciding with stellar deaths. Ultimately this material is returned (via gentle stellar outflows or explosively in the case of supernovae) to the interstellar medium, where it resides for approximately 10^8 years before it is gradually consumed by subsequent generations of star and planet formation. This perpetual recycling and associated enrichment of the gas and dust gradually alters the chemical composition of a galaxy.

Asymptotic Giant Branch (AGB) stars are important contributors to enrichment of the interstellar medium. AGB stars are cool, luminous giants that lose mass at high rates via pulsation-enhanced, dust-driven winds. During this phase of evolution the effective stellar temperature is low enough for dust grains and molecules, notably CO, to condense in their winds, forming substantial circumstellar envelopes detectable in the infrared and millimeter domains.

AGB stars are classified into two major spectral types: carbon-rich stars and oxygen-rich stars. The classification depends on the C/O ratio, which is decisive in the future chemical evolution. In oxygen-rich environments (where the C/O ratio < 1), the dust tends to be composed of amorphous or crystalline forms of silicates and metal-oxides such as amorphous aluminum oxide. Conversely, carbonaceous dust species such as amorphous carbon, SiC and MgS dominate the dust production in carbon-stars (C/O >1). Observational estimates suggest that two thirds of the dust we detect in the Milky Way has been produced by oxygen-rich AGB stars (Gehrz 1989). However, at sub-solar metallicities their total dust contribution relative to supernovae, the efficiency of dust condensation and the chemical composition of the dust remains uncertain.
Fig1_Orich_spec Fig1_crich_spec











Figure 1 (click on plots to enlarge): Example Spitzer-IRS spectra of oxygen-rich and carbon-rich evolved stars with a dust excess in the Small Magellanic Cloud (Ruffel et al. submitted). The key spectral features due to silicates and carbonaceous dust are marked in the top panel.

Figure2a Figure2b
Figure 2 (click on plots to enlarge): Variation in the chemical composition of the crystalline silicate dust alters the peak positions and shapes of the narrow spectral features produced by resonances in the crystalline silicate lattice. Comparing the relative strengths of spectral features (at 23- and 28- microns) which are attributed to different crystalline silicate species in the Large Magellanic Cloud, Small Magellanic Cloud and Milky Way, show a change in the crystalline silicate dust mineralogy with metallicity (Jones et al. 2012). The solid symbols indicate that both features are present in the spectra.

With Spitzer we have observed evolved stars across a range of metal content; from present day solar abundances to metallicities that resemble star-forming galaxies at high redshift (e.g Meixner et al. 2006). These observations show that metallicity has a significant influence on the production and chemical composition of the oxygen-rich dust. In oxygen-rich stars silicates become oxygen poor at lower metallicity, with forsterite becoming less common than enstatite in the Magellanic Clouds, compared to the Galaxy (Jones et al. 2012). This is also seen in low-mass globular cluster stars, where conventional silicate features are seen to disappear in stars below [Fe/H] = -1; instead a featureless mid-IR excess is seen which is possibly caused by metallic iron dust (McDonald et al. 2010). Amorphous alumina oxide is also reduced compared to the Milky Way (Jones et al. 2014). This is because the molecules available for dust production in oxygen-rich stars are limited by the abundances of heavy elements in the photosphere, which reflect the initial abundances of the molecular cloud from which the star formed; this is not the case for carbon stars which produce their own carbon, so the resulting dust condensation sequence in carbon-rich AGB stars shows only minor changes with metallicity (Sloan et al. 2014). Furthermore, at lower metallicities there is less initial oxygen and the numerical ratio of carbon to oxygen atom abundances increases. Consequently, the timescales before dredge-up causes the star to become carbon rich are shorter. Observations of the Magellanic Clouds suggest that the mass return from AGB stars is dominated by carbon stars (Riebel et. al 2012), and cumulatively AGB stars may account for over 50% of the dust production in these galaxies (Boyer et al. 2012, Matsuura et al. 2013).

Recent observational results indicate that large amounts of dust can form in very metal-poor AGB stars (Fe/H] < -2 ; Boyer et al. 2015), and that they contribute significantly to the total dust budgets of metal-poor, high-redshift galaxies. However, it unclear how AGB stars in metal-poor environments are able to produce substantial quantities of dust. Understanding how the conditions in the early universe affect the production and chemical composition of the dust grains will require the superior sensitivity and mid-infrared spectroscopy available with the JWST.


Boyer et al. 2012 (ApJ, 748, 40)
Boyer et al. 2015 (ApJ, 800, 51)
Gehrz 1989 (IAUS, 135, 445)
Jones et al. 2012 (MNRAS 427, 3209)
Jones et al. 2014 (MNRAS 440, 631)
McDonald et al. 2010 (ApJ, 717, 92)
Matsuura et al. 2013 (MNRAS, 429, 2527)
Meixner et al. 2006 (AJ, 132, 2268)
Ruffle et al. (MNRAS, submitted)
Sloan et al. 2014 (ApJ, 791, 28)

Feb 032015

By Annalisa Calamida, Postdoctoral Fellow at STScI

The Milky Way bulge is the closest galaxy bulge that can be observed and studied in detail. The bulge could include as much as a quarter of the stellar mass of the Milky Way (Sofue et al. 2009), and the characterization of its properties can provide crucial information for the understanding of the formation of the Galaxy and similar, more distant galaxies. We observed the Sagittarius low-reddening window, a region of relative low extinction in the bulge, E(B-V)  ≤ 0.6 mag (Oosterhoff & Ponsen 1968), with the Wide-Field Channel of the Advanced Camera for Surveys (ACS) and the Wide Field Camera 3 (WFC3) on board the Hubble Space Telescope (HST). The time-series images were collected in the F606W and F814W filters and cover three seasons of observations from October 2011 to October 2014. The main goal of the project led by Dr. Kailash Sahu is to detect isolated stellar mass black holes and neutron stars in the Galactic bulge and disk through gravitational microlensing.

However, these data, covering 8 WFC3 and 4 ACS fields and including a sample of about 2 million stars, are a gold mine for other Galactic bulge stellar population studies. I started a project aimed at characterizing the bulge stellar populations through the study of different evolutionary phases. The work is based on the same data set used for the microlensing project and on observations taken with ACS in 2004. By combining the observations of one of the ACS fields with those taken in 2004, we measured very precise proper motions, better than ~ 0.5 mas/yr (~ 20 Km/s) at F606W ~ 28 mag, in both axes. Proper motions allowed us to separate disk and bulge stars and to obtain the deepest clean color-magnitude diagram of the bulge. As a consequence we identified for the first time a clearly defined white dwarf cooling sequence in the Galactic bulge (Calamida et al. 2014).

The characterization of the white dwarf population is an effective method to understand the formation history of the bulge and the Milky Way itself, since most stars end their life as white dwarfs. The white dwarf population of the bulge contains important information on the early star formation history of the Milky Way. Our knowledge is most extensive for the white dwarf population of the Galactic disk, in which numerous white dwarfs were discovered through imaging surveys, such as the Sloan Digital Sky Survey (Eisenstein et al. 2006, Kepler et al. 2007, Kleinman et al. 2013). White dwarfs have been identified and studied in a few close Galactic globular clusters too, thanks to deep imaging with HST (Richer et al. 2004; Hansen et al. 2007; Kalirai et al. 2007; Calamida et al. 2008; Bedin et al. 2009). Many of the Milky Way bulge stars are metal rich – several of them even reaching super solar metallicity – which means that they are similar to the Galactic disk population, and the bulge stellar space density is closer to that of the disk than that of globular clusters. However, many of the bulge stars are old, like globular cluster stars, which means that they show common properties with both the disk and the cluster populations. Understanding whether the white dwarfs in the Galactic bulge more closely resemble those found in either the disk or the clusters is therefore an important part of developing our understanding of how the Milky Way was formed, as well as the bulge formation, and, indeed, the nature of the Galactic bulge itself.

The color-magnitude-diagram in the F606W and F814W filters of selected bulge stars is shown in Fig.1, where confirmed white dwarfs are marked with larger filled dots. We used theoretical cooling tracks from the BaSTI database (Pietrinferni et al. 2004; 2006) and models from Althaus et al. (2009) to fit the bulge white dwarf cooling sequence, and a distance modulus DM0 = 14.45 mag and reddening E(B-V) = 0.5 mag (Sahu et al. 2006). These models assume different core and atmospheric compositions: CO-core white dwarfs with an hydrogen-rich envelope (DA), CO-core white dwarfs with an helium-rich envelope (DB) and He-core white dwarfs. Assuming an age of about 11 Gyr and an average solar chemical composition for bulge stars, isochrones predict a turn-off mass of ~ 0.95 M, and through the initial-to-final mass relationship, a white dwarf mass of ~ 0.53 – 0.55 M (Weiss & Ferguson 2009, Salaris et al. 2010).  The figure below shows cooling tracks for DA (dashed blue line) and DB (dashed green) CO-core white dwarfs with mass M = 0.54 M and He-core white dwarfs (dashed red) with mass M = 0.23 M plotted on the Galactic bulge cooling sequence. The DA and DB CO-core cooling tracks are unable to reproduce the entire color range of the observed white dwarf cooling sequence. An increase in the mean mass of the white dwarfs would move these models towards bluer colors, further increasing the discrepancy. The lower mass He-core cooling track for M = 0.23 M fits the red side of the bulge white dwarf sequence (note that empirical evidence shows that the lower mass limit for white dwarfs is ~ 0.2 M, Kepler et al. 2007). These results support the presence of a significant fraction (~ 30%) of low-mass (M ≤ 0.45 M) He-core white dwarfs in the Galactic bulge. According to stellar evolution models, in a Hubble time, these low-mass white dwarfs can only be produced by stars experiencing extreme mass loss events, such as in compact binaries. Among the brighter very red white dwarfs we found indeed one ellipsoidal variable (marked with a blue dot in Fig.1), probably composed of a white dwarf accreting from a main-sequence companion, and two dwarf novae (magenta dots). The fainter counterparts of these binaries could populate the region where the reddest white dwarfs are observed in the color-magnitude diagram. These systems could be composed of a white dwarf and a low-mass (M <  0. 3M) main-sequence companion. This hypothesis is further supported by the finding of five cataclysmic variable candidates in the same field (green dots).


Figure 1: proper-motion cleaned bulge color-magnitude diagram in the F606W and F814W filters. White dwarfs are marked with larger filled dots. Dashed lines display cooling tracks for CO- and He-core white dwarfs. The ellipsoidal variable and the dwarf novae are marked with blue and magenta dots, respectively, while green dots mark cataclysmic variable candidates. Error bars are also labeled.

Our observational campaign ended in November 2014 and now, by adding the third season of observations and including all the twelve ACS and WFC3 fields, we will be able to increase our sample of stars by one order of magnitude. The increased statistics will allow us to better constrain the nature of the white dwarf population in the bulge, for instance, through comparing white dwarf and main-sequence star counts.

In the future, this study will greatly benefit from the advent of the James Webb Space Telescope (JWST), which will have an improved sensitivity and spatial resolution compared to HST. Moreover, JWST will observe in the near-infrared, where the extinction is a factor of ten lower compared to the optical. The new telescope will allow us to observe the bulge white dwarfs in the near-infrared, something that is now barely feasible with HST. The white dwarf cooling sequence will then be used to estimate the age of stellar populations in the bulge, to be compared to estimates obtained by adopting other diagnostics, such as the main-sequence turn-off. This study will be fundamental for constraining the presence of an age spread in the Galactic bulge, which is now a hotly debated topic.


For more details see Calamida et al. 2014, ApJ, 790, 164



  • Sofue et al. 2009, PASJ, 61, 227
  • Oosterhoff & Ponsen 1968, BANS, 3, 79
  • Calamida et al. 2014, ApJ, 790, 164
  • Eisenstein et al. 2006, ApJS, 167, 40
  • Kepler et al. 2007, MNRAS, 375, 1315
  • Kleinman et al. 2013, ApJS, 204, 5
  • Richer et al. 2004, AJ, 127, 2904
  • Hansen et al. 2007, ApJ, 671, 380
  • Kalirai et al. 2007, ApJ, 671, 748
  • Calamida et al. 2008, ApJ, 673, L29
  • Bedin et al. 2009, ApJ, 697, 965
  • Pietrinferni et al. 2004, ApJ, 612, 168
  • Pietrinferni et al. 2006, ApJ, 642, 797
  • Althaus et al. 2009, A&A, 502, 207
  • Sahu et al. 2006, Nature, 443, 534
  • Weiss & Ferguson 2009, A&A, 508, 1343
  • Salaris et al. 2010, ApJ, 716, 1241