Astronomical Science

An Unbiased Near-infrared Interferometric Survey for Hot Exozodiacal Dust

Steve Ertel1,2,3 composition and distribution of the dust vide clues pointing towards the presence Jean-Charles Augereau2,3 produced in debris discs through colli- of a planet, but clumps in the dust distri- Olivier Absil4 sions and outgassing of these larger bod- bution may also be misinterpreted as Denis Defrère5 ies is a powerful tool that can help to actual planets due to the limited resolu- Jean-Baptiste Le Bouquin2,3 constrain the architecture, dynamics and tion and sensitivity of current instruments. Lindsay Marion4 evolution of extrasolar planetary systems. Thus, detecting and characterising exo- Amy Bonsor6 However, debris discs that are relatively zodiacal dust systems provides critical Jérémy Lebreton7,8 easy to detect are located several astro- input for the design of such space mis- nomical units (au) to a few hundreds of au sions. However, detecting the dust itself from their host stars, and thus only trace with present instruments is complicated 1 ESO the outer regions of planetary systems. by the fact that the warm and hot emis- 2 Université Grenoble Alpes, France In order to study the inner regions close sion peaks at near- to mid-infrared wave- 3 CNRS, Institut de Planétologie et to the habitable zones, one has to study lengths where the dust emission is out- d’Astrophysique de Grenoble, France warm and hot dust closer to the star. This shone by the host star. Thus, only a few 4 Départment d’Astrophysique, Géo­ dust is called exozodiacal dust, or exo- very bright — and perhaps unrepresenta- physique et Océanographie, Université zodi for short, by analogue with the Zodi- tive — systems can be detected photo- de Liège, Belgium acal dust in the Solar System. metrically. 5 Department of Astronomy, Steward Observatory, University of Arizona, USA The Zodiacal light can be observed on 6 School of Physics, H. H. Wills Physics dark nights directly after dusk and before Infrared interferometric detection of Laboratory, University of Bristol, United dawn as a cone of faint light stretching ­e x o z o d i s Kingdom from the horizon in the west (after dusk) 7 NASA Exoplanet Science Institute, or in the east (before dawn). It is caused Due to the small extent of exozodiacal ­California Institute of Technology, by sunlight scattered off small dust parti- systems — one au at a typical distance ­Pasadena, USA cles close to the orbit of the Earth. More of 10 pc for nearby stars corresponds to 8 Infrared Processing and Analysis Center, generally, Zodiacal dust is distributed in an angular size of 100 milliarcseconds — California Institute of Technology, a disc inside the asteroid belt, extending only interferometry is currently able to ­Pasadena, USA all the way down to the sublimation dis- spatially resolve them. When used at tance of the dust from the Sun, which baselines of a few tens of metres, near- corresponds to a few Solar radii. The infrared interferometry is able to fully Exozodiacal dust is warm or hot dust dust temperatures range from about resolve the extended emission of the dust found in the inner regions of planetary 100 K to about 2000 K, depending on disc while the star still remains largely systems orbiting main sequence stars, the distance from the Sun. In the inner- unresolved. The result is a small deficit in in or around their habitable zones. The most regions it forms the Fraunhofer the measured squared visibilities (the main dust can be the most luminous compo- corona (F-corona) of the Sun, a region of observable of infrared interferometry) nent of extrasolar planetary systems, the corona where the prominent absorp- compared to the values expected from but predominantly emits in the near- to tion lines in the Solar spectrum are visible the star alone (see Figure 1 for an expla- mid-infrared where it is outshone by because the light seen there is nearly nation). Using this technique the disc can the host star. Interferometry provides a unaltered sunlight scattered by the dust be spatially disentangled from the star, unique method of separating this dusty particles. It is noteworthy that the Zodia- allowing the disc to be detected and the emission from the stellar emission. cal light is the most luminous component ratio between the disc and stellar emis- The visitor instrument PIONIER at the of the Solar System after the Sun itself. sion measured. This method has so far Very Large Telescope Interferometer been the most powerful and efficient tool (VLTI) has been used to search for hot Not unlike the Zodiacal dust, exozodiacal in the search for faint exozodiacal dust. exozodiacal dust around a large sample dust is located in the inner regions of of nearby main sequence stars. The extrasolar planetary systems, within a few However, the dust detected by this results of this survey are summarised: au of main sequence stars. This region method in the near-infrared (NIR) is ex­­ 9 out of 85 stars show excess exo- often encompasses their habitable zone. pected to be very hot, close to its subli- zo­diacal emission over the stellar Historically, this circumstance has brought mation temperature, and its relation to photo­spheric emission. it a lot of attention, because the presence slightly cooler dust in the habitable zone of exozodis is expected to complicate is unclear. This habitable zone dust is the direct-imaging detection and charac- brighter in the mid-infrared (MIR) where it Planetesimals and comets are a major terisation of Earth-like planets in the hab- can be detected by the VLTI MID-infrared component of the Solar System (in the itable zones around other stars by future interferometric instrument (MIDI) for a Kuiper Belt and the asteroid belt), as well space missions. The faint light of these few bright systems (e.g., Smith et al., as of extrasolar planetary systems, where potential planets can be hidden in the 2009) or more efficiently by nulling inter- they occur in debris discs. Besides the extended emission of the dust disc. The ferometry. With this latter method the planets, they are the main outcome of the structures created in the dust distribution stellar light from two telescopes is brought planet formation process. Studying the due to planet–disc interaction may pro- to destructive interference, while light

24 The Messenger 159 – March 2015   Figure 1. (Left) Illustration of the detection strategy for exozodis. The dashed curve shows a realistic

SX   9NNL case assuming a uniform disc for both the star and   the flux distribution from the exozodiacal dust and   HRHAHKH a disc-to-star flux ratio off = 1%. For the simplified   CU case, the solid curve shows the same assumptions,   but with the approximation following the equation. Diameters of the star and (face-on) disc have been

SX     2PT@QD chosen to be 2.5 milliarcseconds (about an A-type   star at 10 pc) and 500 milliarcseconds (5 au at HRHAHKH   10 pc), but exact numbers are not relevant for this

CU #HRBNMKX !@RDKHMDL illustration. 2S@Q CHRBQD@K   2S@Q CHRBRHLOKHjDC 2PT@QD 2S@QŸNMCH@LDSDQ Figure 2. (Below) A diagram (not to scale) to illustrate #HRBRS@QkTWQ@SHN the scattering of planetesimals by an outer planet,   that leads to an exchange of angular momentum and ) /Aeh £ the outward migration of that planet (Bonsor et al., 5£¶lE  /Aeh 2014). Some of the scattered particles are ejected, whilst some are scattered into the inner planetary   system, where they interact with the inner planets.  This scattering leads to a flux of material into the !@RDKHMDL exozodi region.

2S@Q

$WNYNCH(MMDQOK@MDSR .TSDQOK@MDSR .TSDQOK@MDSDRHL@KADKS

­off-centre from the star is transmitted, required to prepare future exoEarth imag- inwards to regions where it is detected. which improves the dynamic range of ing missions. The Large Binocular Tele- Furthermore, a higher dust mass would the observations. Both methods have scope Interferometer (LBTI) is designed require a larger number of planetesimals been used in parallel in the past to search to achieve the required sensitivity and will which are colliding more often to pro- for exozodiacal dust, mostly with the soon start a MIR survey of 50 to 60 care- duce the dust present; this process Fiber Linked Unit for Optical Recombi­ fully chosen nearby main sequence stars would in turn destroy the planetesimals nation (FLUOR) instrument at the Center (Weinberger et al., 2015). faster, so that at the ages of the systems for High Angular Resolution Astronomy observed, few planetesimals would be (CHARA) array (in the NIR; e.g., Absil et left (Wyatt et al., 2007). The production of al., 2006) on Mt. Wilson, California and Potential origins of exozodiacal dust the observed amounts of dust through the Keck Interferometer Nuller (in the comet evaporation would require a large MIR; Mennesson et al., 2014) on Mauna The fact that many exozodiacal systems number of comets, approximately one Kea, Hawaii. have already been found, given the lim- thousand events per year similar to Hale– ited sensitivity of present instruments, is Bopp reaching its perihelion. Due to the recent development of the vis- surprising (Absil et al., 2013). Such high itor instrument PIONIER (Precision Inte- levels of dust are difficult to sustain. A potential scenario explaining such a grated Optics Near Infrared ExpeRiment; The Zodiacal dust has two main origins: large number of comets would be a con- LeBouquin et al., 2011) at the VLTI, which collisions of asteroids in the asteroid nection to an outer debris disc where operates in the H-band, the search for belt result in dust that is dragged inwards planetesimals are scattered inwards exozodis in the NIR has become more by the interaction with stellar radiation through gravitational interaction with an efficient and of similar accuracy to the (Poynting–Robertson drag) and the evap- existing planetary system (see Figure 2). previous surveys, allowing for a signifi- oration of comets heated when they Considering the realistic interaction of cant increase in the number of surveyed get close to the Sun supplies dust. While these planets with the planetesimals stars. The sensitivity of the available these scenarios might work for a few (planetesimal-driven planetary migration), instruments is, however, only capable of MIR-detected systems, they generally do this scenario is able to produce detec­ detecting exozodis a couple of hundred not work well for more massive systems, table exozodis even if the outer debris times brighter (more massive) than the as have been detected in the NIR around disc were too faint to be detected with Zodiacal dust. This state-of-the-art exo- other stars. If the dust density in the disc present instruments (Bonsor et al., 2014). zodi sensitivity is approximately one is too high, the dust will further collide However, such a mechanism puts strong order of magnitude larger than that and be destroyed before being dragged constraints on the architecture of the

The Messenger 159 – March 2015 25 Astronomical Science Ertel S. et al., Near-infrared Interferometric Survey for Hot Exozodiacal Dust

%KTWQ@SHNCHRSQHATSHNM4MBDQS@HMSXCHRSQHATSHNM  2HFL@NESGDCHRSQHATSHNM   ,DCH@MTMBDQS@HMSX — l   

    -TLADQ -TLADQ        l l               l $WBDRRRHFMHjB@MBDE"2$mE ARNKTSDTMBDQS@HMHSXNEkTWQ@SHNmE 

Figure 3. Excess distribution (left) and distribution more massive discs are present around for the statistical analysis. The targets for of uncertainties (right) on the disc-to-star flux ratio more massive stars. A strong depend- both surveys were selected carefully in are shown. The Gaussian overplotted on the excess distribution has a width of σ = 1 and is used to guide ence on the system’s age (most exozodis order to avoid any relevant selection bias the eye, illustrating that the data are consistent with are detected around very young stars) that could affect the statistics. this ideal behaviour. Vertical dashed lines are plotted would suggest that the origin is a collision- at f = −3σ and +3σ for the excess distribution and ally depleting planetesimal belt. Alterna- In addition to the higher efficiency, at the median uncertainty (2.6 × 10−3) for the uncer- tainty distribution. tively a strong correlation with the pres- ­PIONIER provides a few advantages over ence of an outer disc would suggest that FLUOR. The simultaneous use of four the hot material is supplied through some ­telescopes allows the closure phase of underlying planetary system. While this mechanism from this outer reservoir. the detected systems to be measured, would be a very interesting scenario, which can only be achieved by combining it is questionable whether all systems Our team carried out the first large, NIR the light from combinations of three tele- observed are likely to meet these con- interferometric survey for hot exozodiacal scopes. This quantity measures the devi- straints. A combination of planetesimal- dust in order to statistically address the ation of the brightness distribution of driven migration and dust trapping could questions on the origin and evolution of an observed target from point symmetry, explain the presence of the detected these enigmatic systems (Absil et al., allowing a star surrounded by a dust dust. Potential dust-trapping mechanisms 2013; Ertel et al., 2014). We started our disc and a star orbited by a faint, close are suggested from the ­realistic treat- search using CHARA/FLUOR, the first companion, which otherwise have a very ment of the dust sublimation (Lebreton instrument to routinely provide the ­similar observed visibility deficit, to be et al., 2013) or the interaction of dust required accuracy of the measurements. distinguished. Another advantage is the grains with the stellar magnetic field However, FLUOR can only use two tele- fact that PIONIER data can be obtained (Czechowski & Mann, 2010). scopes at a time, which results in a lim- with a small spectral resolution of three ited observational efficiency. Thus, in or seven spectral channels across about seven years, only around 40 stars the H-band. This information allows one An unbiased near-infrared interferometric could be observed with sufficient accu- to constrain the spectral slope of the survey racy, but this still allowed the first statisti- detected excess and thus the emission cal conclusions to be drawn (Absil et al., mechanism and — in case of thermal Strong constraints on the potential origins 2013). emission — the temperature of the dust. of exozodiacal dust, described above, can be placed by a statistical analysis of The development of the PIONIER instru- the detection rate and excess levels of ment for the VLTI, which can use four Survey results exozodiacal dust with respect to other ­telescopes simultaneously, allowed a sig- properties of the systems, such as stellar nificant increase in our observing effi- From our PIONIER survey we have found spectral type, age of the system, or the ciency, so that in only 12 nights during that nine out of the 85 targets which presence of a detectable cool debris disc 2012 a total of 92 stars could be ob­­ proved to be suitable for our analysis in the outer regions of the systems. A served. A cumulative median 1σ accuracy (~11%) show a significant excess, typi- dependence on the stellar spectral type per target of 0.26 % on the disc-to-star cally around 0.5 to 1%, above the stellar (i.e., stellar mass), similar to the one that flux ratio was reached (Figure 3). To photosphere. Five targets were found has been found for debris discs and increase our sample and thereby to to have a faint, stellar companion and younger protoplanetary discs, would improve our statistics we combined the needed to be rejected as exozodis. These ­suggest a similar or even common origin results from the PIONIER survey with latter detections were analysed in a sepa- of the material detected in these different those from FLUOR, resulting in a total of rate study and interesting implications dust populations. For these systems, about 130 stars observed and available for the formation of binary stars were

26 The Messenger 159 – March 2015   Figure 4. Single (left) %+4.1* A@MC and combined (right)

X "NLAHMDC%+4.1/(.-($1MD@Q HMEQ@QDCR@LOKD statistics from the /(.-($1' A@MC   FLUOR (red) and PTDMBX

DPTDMB ­PIONIER (blue) sample. QD

RE The top row shows the R EQ  DR DR excess fraction with  respect to the stellar CDWB CDWB  spectral type, the mid-

Q@QD dle row shows the A@M same, but, in addition, HME

Q * 

N  separated for stars with ' -D@Q and without a debris disc detected, and the   SXOD % SXOD &* SXOD SNS@K SXOD % SXOD &* SXOD SNS@K bottom row the excess fraction for different spectral types and   ­separated for stars %+4.1 VHSGCDAQHRCHRB "NLAHMDC VHSGCDAQHRCHRB X younger and older than %+4.1 VHSGNTSCDAQHRCHRB   "NLAHMDC VHSGNTSCDAQHRCHRB the median age in each /(.-($1 VHSGCDAQHRCHRB

PTDMBX spectral type bin. DPTDMB /(.-($1 VHSGNTSCDAQHRCHRB QD  RE R EQ  DR BDR  CDW CDWB 

Q@QD  A@M HME Q *  N  ' -D@Q

  SXOD % SXOD&* SXODSNS@K SXOD % SXOD &* SXOD SNS@K

  %+4.1 XNTMF "NLAHMDC XNTMF %+4.1 NKC   "NLAHMDC NKC /(.-($1 XNTMF PTDMBX PTDMBX QD QD /(.-($1 NKC  RE RE  DR DR  CDWB CDWB 

Q@QD  A@M HME Q *  N  ' -D@Q

  SXOD % SXOD&* SXODSNS@K SXOD % SXOD &* SXOD SNS@K found (Marion et al., 2014). One more star  Figure 5. Excess levels was rejected from the analysis because it for the stars in our sam- %&* SXOD RHFMHjB@MSDWBDRR ple shown against stellar shows hints of post-main sequence evo-  %&* SXOD MNDWBDRR age. Filled symbols are lution which could cause other effects, for stars with significant such as mass loss or strong stellar winds  SXOD RHFMHjB@MSDWBDRR excess, while empty N  SXOD MNDWBDRR symbols are for stars

that might mimic an exozodi in our data. @SH without a significant WQ

T detection. The sample is A detailed statistical analysis of the separated into A-type

Ck  ­combined PIONIER and FLUOR sample stars and stars of later (Figure 4) shows that the distribution of spectral types, in order to account for the differ- HMEQ@QD  the exozodiacal dust detection rate, with ent main sequence life- respect to the stellar spectral type, is times of these stars. The -D@Q similar to that of debris discs. This sug- l star with a 3% flux ratio gests that both kinds of dust discs are is α Aql (see Absil et al. [2013] for details). produced by circumstellar material origi- l nating in the protoplanetary disc during    the planet formation process. However, FD&XQ there is no correlation between the pres- ence of a debris disc and an exozodi

The Messenger 159 – March 2015 27 Astronomical Science Ertel S. et al., Near-infrared Interferometric Survey for Hot Exozodiacal Dust

  Figure 6. Spectral slopes of the excesses detected     with PIONIER (first three rows) and some ­potential '# '# '# excesses only identified by their spectral slopes,   considered tentative (fourth row). The horizontal,   black line illustrates the case of purely (grey) scat-   tered stellar light, while the coloured curves show     the shapes expected from blackbody thermal emis- sion at different temperatures. Error bars are 1σ.      

 '# '# '#       very hot dust emitting significantly in the     H-band needs to be very close to the   star, in the case of scattered light the dust     could be slightly more distant, closer to       the habitable zone. At that location, the A@MCkTWQ@SHN

' dust has a stronger impact on the detect- '#   '#'  #   ability of habitable Earth-like planets.           Future prospects       Our sample of near-infrared bright exo­                   zodis provides an excellent basis for fur- 6@UDKDMFSGƅL ther studies. At the same time, our obser-    vations are very timely for the new VLTI   instruments, GRAVITY and MATISSE,

 '# '#'#  

N   expected to arrive in 2015 and 2016,   @SH   respectively. GRAVITY will enable us to WQ     observe the discs in the K-band, at

%KT l  slightly longer wavelengths than PIONIER, l  l                    thereby better constraining the slope of 6@UDKDMFSGƅL the excess and thus the emission mecha- nism. From the increased emission in K-band compared to H-band, GRAVITY ­visible in our data. This might suggest sion; however, given the high detection will also allow the excesses to be meas- that there is indeed no physical correla- rate of exozodis, the short lifetime of the ured at very high significance. This will tion between the two phenomena. On dust and the expected scarcity of such allow detection of small variations in the the other hand, it is important to note that events, this is not a likely explanation for excess levels, indicative of variation in we are only able to detect the brightest, the detected systems. Thus, the inde- the mass and location of the dust, ena- most extreme exozodis with current pendence of the detection rate and bling strong constraints to be put on its instrumentation, and the debris discs excess levels from the ages of the sys- evolution. MATISSE will cover the range detectable so far are also at least a few tems surveyed remains enigmatic. of even longer wavelengths up to the times more massive than the Kuiper Belt N-band in the MIR, constraining very well in the Solar System. Thus, a significant An important result from PIONIER derives the most relevant region in which the fraction of both kinds of discs remain from the spectrally dispersed data, en­­ dust emits (Figure 7). Combining all VLTI undetected and a correlation might be abling us to constrain, for the first time, instruments, we will be able to study in hidden by that fact. the spectral slopes of a large number of detail the dust distribution and composi- excess detections (Figure 6). The thermal tion in these systems. The strength of A very surprising result is that the detec- emission of even the hottest dust is combined modelling of different interfero- tion rate (Figure 4) and excess levels (Fig- increasing towards wavelengths longer metric data has — to some extent — ure 5) of exozodis do not decrease with than the H-band. A consequence of this already been shown by our team (Defrère the ages of the systems. This would be thermal emission would be an increasing et al., 2011; Lebreton et al., 2013). expected for any phenomenon that disc-to-star contrast with wavelength. evolves over time, such as a belt of collid- However, we find for most of our detec- One of the main questions to be ing planetesimals. In contrast, the exozo- tions that the contrast is rather constant answered concerns the nature of any dis seem to be caused by a stochastic with wavelength, indicating that the emis- connections between the hot dust process or a process that can be trig- sion is dominated by starlight scattered detected in the NIR, the warm dust in the gered stochastically at any time during off the dust grains. This unexpected habitable zone and even the colder dust the evolution of the planetary system. result has important implications for the further out. So far there are indications One example could be a planetary colli- dust distribution in these systems. While that there might be an anti-correlation

28 The Messenger 159 – March 2015 between the presence of NIR- and MIR- Figure 7. Illustration of * /(.-($1 , 3(22$ the spectral coverage

detected dust. This might suggest the p *

% of PIONIER and the presence of planets in the systems with  * &1 5(38 SX future VLTI instruments  MIR bright dust emission that prevents  GRAVITY and MATISSE. the dust from the outer regions migrating DMRH * The shape of the stellar C emission is shown in further in, where it would be detected in TW * grey as the blackbody Kk the NIR. Instead the dust would pile up emission of two differ- near the orbits of those planets. ent, typical stellar tem-

ODBSQ@ peratures. Dust black-  l 2S@Q In the context of future observational CR body curves are shown as coloured lines. The ­perspectives, two instruments in the spectral coverage of the C C C

northern hemisphere will soon provide C VLTI instruments is illus-

critical complementary information on -NQL@KHRD trated by the coloured, A@M A@M A@M A@M vertical bars. exozodis: the CHARA/FLUOR instrument ' * + , - A@MC in the NIR and the LBTI in the MIR. A  l   new mode at CHARA/FLUOR will provide   6@UDKDMFSGƅL spectrally dispersed observations of northern objects inaccessible from ­Paranal, while the LBTI will observe with dust and will give limited constraints on References unprecedented sensitivity a sample of the dust properties. Only a complete Absil, O. et al. 2006, A&A, 452, 237 50 to 60 nearby main sequence stars to exploitation with both NIR and MIR Absil, O. et al. 2013, A&A, 555, 104 characterise the faint end of the exozodi observations, and in particular those of Bonsor, A. et al. 2014, MNRAS, 441, 2380 luminosity function (see first LBTI results the near-future instruments at the VLTI, Czechowski, A. & Mann, I. 2010, ApJ, 714, 89 in Defrère et al. [2015]). However, the will allow us to precisely characterise Defrère, D. et al. 2011, A&A, 534, 5 Defrère, D. et al. 2015, ApJ, 799, 1 ­efficiency of both instruments is limited exozodiacal dust and study its origin and Ertel, S. et al. 2014, A&A, 570, 128 due to the use of only one interferometric evolution, even if this is only possible for Le Bouquin, J.-B. et al. 2011, A&A, 535, 67 baseline at a time and the need for a systems bright enough to be detected by Lebreton, J. et al. 2013, A&A, 555, 146 large number of observations to reach our survey. Marion, L. et al. 2014, A&A, 570, 127 Mennesson, B. et al. 2014, ApJ, 797, 119 the nominal sensitivity. Furthermore, the Smith, R. et al. 2009, A&A, 503, 265 LBTI exozodi survey is designed for Weinberger, A. J. et al. 2015, ApJS, in press broadband detection of faint levels of Wyatt, M. C. et al. 2007, ApJ, 658, 569 ESO/VVV consortium/D. Minniti

VISTA VVV (the Variables in the Via Lactea Public Survey) near-infrared colour image (JHKs) of the centre of the Galactic H II region M20 (NGC 6514). At near-infrared wave- lengths the line emission is weaker than in the optical and the dust extinc- tion lower, so background stars are easily located. Two Cepheid variables on the other side of the Galactic disc at 11 kpc were detected in this field; see Release eso1504 for details.

The Messenger 159 – March 2015 29