An interferometric study of the Fomalhaut inner debris disk. I. Near-infrared detection of hot dust with VLTI/VINCI
Olivier Absil, Bertrand Mennesson, Jean-Baptiste Le Bouquin, Emmanuel Di Folco, Pierre Kervella, Jean-Charles Augereau
aa r X i v : . [ a s t r o - ph . S R ] A ug To appear in the Astrophysical Journal
Preprint typeset using L A TEX style emulateapj v. 03/07/07
AN INTERFEROMETRIC STUDY OF THE FOMALHAUT INNER DEBRIS DISKI. NEAR-INFRARED DETECTION OF HOT DUST WITH VLTI/VINCI ∗ Olivier Absil IAGL, Universit´e de Li`ege, 17 All´ee du Six Aoˆut, B-4000 Sart Tilman, Belgium andLAOG–UMR 5571, CNRS and Universit´e Joseph Fourier, BP 53, F-38041 Grenoble, France
Bertrand Mennesson
Jet Propulsion Laboratory, California Institute of Technology, 4800 Oak Grove Drive, Pasadena, CA 91109, USA
Jean-Baptiste Le Bouquin
European Southern Observatory, Casilla 19001, Santiago 19, Chile
Emmanuel Di Folco and Pierre Kervella
LESIA–UMR 8109, CNRS and Observatoire de Paris-Meudon, 5 place J. Janssen, F-92195 Meudon, France andJean-Charles Augereau
LAOG–UMR 5571, CNRS and Universit´e Joseph Fourier, BP 53, F-38041 Grenoble, France
Received March 30, 2009; Accepted August 20, 2009.
ABSTRACTThe innermost parts of dusty debris disks around main sequence stars are currently poorly knowndue to the high contrast and small angular separation with their parent stars. Using near-infraredinterferometry, we aim to detect the signature of hot dust around the nearby A4 V star Fomalhaut,which has already been suggested to harbor a warm dust population in addition to a cold dust ringlocated at about 140 AU. Archival data obtained with the VINCI instrument at the VLTI are usedto study the fringe visibility of the Fomalhaut system at projected baseline lengths ranging from 4 mto 140 m in the K band. A significant visibility deficit is observed at short baselines with respectto the expected visibility of the sole stellar photosphere. This is interpreted as the signature ofresolved circumstellar emission, producing a relative flux of 0 . ± .
12% with respect to the stellarphotosphere. While our interferometric data cannot directly constrain the morphology of the excessemission source, complementary data from the literature allow us to discard an off-axis point-likeobject as the source of circumstellar emission. We argue that the thermal emission from hot dustygrains located within 6 AU from Fomalhaut is the most plausible explanation for the detected excess.Our study also provides a revised limb-darkened diameter for Fomalhaut ( θ LD = 2 . ± .
022 mas),taking into account the effect of the resolved circumstellar emission.
Subject headings:
Circumstellar matter — techniques: interferometric — stars: individual (Fomalhaut) INTRODUCTION
The young ( ∼
200 Myr, Di Folco et al. 2004) andnearby (7.7 pc) A4 main sequence star Fomalhaut( α PsA, HD 216956) has been the focus of much at-tention during the last decade in the context of plane-tary system studies. Although the discovery of a colddebris disk around this bright star dates back to earlyobservations with the InfraRed Astronomical Satellite(Aumann 1985), the first resolved observations of itsdisk have been obtained only in the late 90’s in the sub-millimeter regime (Holland et al. 1998), and a few yearslater at optical wavelengths (Kalas et al. 2005). Fomal-haut has nowadays one of the best studied debris disk,with most of its (cold) dust arranged in a narrow ringat 140 AU from its host star. This ring shows interestingfeatures, such as a sharp inner edge that has been in- ∗ Based on observations made with ESO Telescopes at theParanal Observatory (public VINCI commissioning data). FNRS Postdoctoral Researcher terpreted as the result of the gravitational influence of amassive planet located just inside the dust ring (Quillen2006). The existence of this predicted planetary com-panion was recently confirmed by direct observations ofFomalhaut in the optical regime (Kalas et al. 2008), fur-ther boosting the general interest of this system. TheFomalhaut debris disk has additionally been suggested tocontain a population of warm dust thanks to partially re-solved Spitzer/MIPS observations and Spitzer/IRS spec-troscopy (Stapelfeldt et al. 2004).Although the Fomalhaut debris disk has been stud-ied in many details during the last few years, its innerdust content remains rather elusive, with only weak con-straints on a warm dust population within 20 AU pro-vided by Spitzer observations. The main challenges forcharacterizing this warm dust population are the smallangular separation and the high contrast between thestar and the inner disk. Infrared interferometry is anappropriate tool to tackle these challenges: by providingan angular resolution as good as a few milli-arcseconds Absil et al.(mas), it can potentially resolve dust populations downto a fraction of an AU from Fomalhaut. This paper is thefirst of a series that aims at using infrared interferometryto characterize the warm dust content in the inner fewAUs around Fomalhaut, and thereby provide a betterview on the global architecture of its planetary system.The present paper focuses on the search for hot cir-cumstellar emission in the near-infrared regime with theVINCI instrument of the VLT Interferometer. The prin-ciple for warm dust detection with interferometry isbased on the fact that the stellar photosphere and itssurrounding dust disk have different spatial scales. Foran A-type star at 7.7 pc, the angular diameter of the pho-tosphere is about 2 mas, while the circumstellar disk ex-tends beyond the sublimation radius of dust grains, typi-cally located around 0.15 AU (i.e., 20 mas) for black bodygrains sublimating at T sub ≃ τ Cet (Di Folco et al. 2007), ζ Aql(Absil et al. 2008) and β Leo (Akeson et al. 2009). Inthis paper, we revisit archival VLTI/VINCI observationsof Fomalhaut and apply the same method to obtain firstevidence for hot dust within 6 AU from Fomalhaut. OBSERVATIONS AND DATA REDUCTION
The bright star Fomalhaut has been observed on sev-eral occasions between 2001 and 2004 with VINCI, theVLT Interferometer Commissioning Instrument, whichcoherently combines the infrared light coming from twotelescopes in the infrared H and K bands (for a detaileddescription, see Kervella et al. 2003a). Some of these ob-servations have already been reported in two papers byDi Folco et al. (2004) and Le Bouquin et al. (2006). Inthis section, we describe these two data sets as well asan unpublished data set that we have extracted from theESO archives. Most of the observations described herehave been obtained with the 40-cm test siderostats of theVLTI. Their field-of-view (FOV) is limited by the use ofsingle-mode fibers inside the VINCI instrument, and canbe described by a 2D Gaussian function with a full widthat half maximum (FWHM) of 1 . ′′ K band understandard atmospheric conditions at Cerro Paranal. Thistranslates into a linear FOV radius at half maximum ofabout 6 AU at the distance of Fomalhaut. Long-baseline data from Di Folco et al. (2004)
Observations of Fomalhaut have been obtained in the K band by Di Folco et al. (2004) between November2001 and October 2002 with three baselines of vari-ous lengths and orientations: E0–G1 (66 m) and B3–M0(140 m) with the 40 cm test siderostats, and UT1–UT3(102 m) with the 8 m Unit Telescopes (see Fig. 1). Afew observations have also been obtained in the H bandon the E0–G1 baseline, but because H -band data are not available on other baselines and therefore cannot beused to assess the presence of circumstellar emission inthis particular waveband, these observations will not befurther discussed here.In the present study, we will use the calibrated squaredvisibilities published by Di Folco et al. (2004). Data re-duction was performed by using the wavelet analysis ofthe fringe power spectrum described by Kervella et al.(2004a) and implemented in the VNDRS data reduc-tion software . The resulting coherence factors (or rawsquared visibilities) have been converted into calibratedvisibilities by estimating the instrumental transfer func-tion with interleaved observations of calibrator stars withknown diameters. The uncertainties on the resultingsquared visibilities have been separated into statisticalerror (related to the dispersion of the raw visibilities)and systematic error (induced by the uncertainty on thecalibrators’ angular diameters).This data set has allowed Di Folco et al. (2004) to de-rive an accurate estimation of the uniform-disk angulardiameter ( θ UD ) of Fomalhaut: 2 . ± . ± . θ LD )following Hanbury Brown et al. (1974): θ LD ( λ ) θ UD ( λ ) = s − u λ / − u λ / , (1)with u λ the linear limb-darkening coefficient at a givenwavelength λ . Using the tabulated limb-darkening coef-ficients of Claret (2000), the resulting LD diameter reads2 . ± . ± .
020 mas. Due to the lack of short base-lines, Di Folco et al. (2004) could not thoroughly studythe circumstellar environment of Fomalhaut, and wereonly able to derive an upper limit of about 3% on the fluxratio between the circumstellar and photospheric emis-sions.
Medium-baseline data from Le Bouquin et al.(2006)
Observations of Fomalhaut have been obtained withVINCI in the K band using two 40 cm siderostats on the64 m D0–H0 baseline (see Fig. 1). The observations wereperformed on five consecutive nights from October 8thto October 12th, 2004. A large part of this data set wasreduced by Le Bouquin et al. (2006) using a custom datareduction procedure, and was subsequently used to val-idate the new integrated optics beam-combiner IONIC-2TK, which was installed in replacement of the origi-nal MONA fibered beam-combiner in August 2004. Thestudy performed by Le Bouquin et al. (2006) shows avery nice consistency between the MONA and IONIC-2TK versions of the VINCI instrument, including verysimilar spectral transmissions. The study also demon-strates the validity of a Fourier-based estimator to reducethe IONIC-2TK data.These observations were not intended to study the cir-cumstellar environment of Fomalhaut, so that the ob-serving strategy was not optimal in that respect. In Publicly available on the ESO VLTI web site (see ). n Interferometric Study of the Fomalhaut Inner Debris Disk. I. 3 Fig. 1.—
Schematic view of the VLTI baselines used in this study. Data at long baselines (UT1–UT3, B3–M0, E0–G1; in red) andintermediate baselines (D0–H0, in blue) are taken from previous studies respectively by Di Folco et al. (2004) and Le Bouquin et al. (2006),while the data obtained at the shortest baselines (E0–G0, in green) are unpublished data from the ESO archives. particular, only one calibrator star (88 Aqr) was usedthroughout the five nights. This is expected to producesignificant systematic errors on the calibrated visibilities,all the more that, with its estimated angular diameter of3 . ± .
057 mas (Richichi et al. 2009), this calibratorstar is partly resolved on the D0–H0 baseline. However,because some of the observations have been obtained forprojected baseline lengths as short as 20 m, this data setis still very useful to assess the presence of circumstellaremission. At such baselines, the stellar photosphere isalmost unresolved ( V ≃ . . ± .
06 mas for88 Aqr, which we will use in this study, as well as anindependent estimation of the UD angular diameter ofFomalhaut (2 . ± .
07 mas), in good agreement with theestimation of Di Folco et al. (2004).For the present study, we have retrieved the full origi-nal data set (5 nights) and reduced it with VNDRS 3.1,the latest version of the VINCI data reduction software.A total of 86 individual observations are available onFomalhaut for the five nights. The output of the soft-ware consists in wavelet-based Fourier domain estima-tions of the squared coherence factors µ for the targetstar (Fomalhaut) and its calibrator (88 Aqr). These “un-calibrated visibilities” are illustrated for a representativenight in Fig. 2. Coherence factors must then be converted into squared visibilities ( V ) for the scientific object bythe relation: V = µ T (2)where T is the interferometric transfer function (TF),i.e., the response of the system to a point source. Theestimation of the TF is based on the interleaved obser-vations of the calibrator star, whose V can be com-puted from the a priori knowledge of its angular diam-eter ( θ UD = 3 . ± .
06 mas). This V must be com-puted for the wide spectral bandwidth on which theVINCI observations are performed, taking into accountthe actual spectrum of the star (88 Aqr, a K1 III gi-ant) and the spectral transmission of the VINCI instru-ment. The spectral transmission of VINCI is taken fromKervella et al. (2003b), and we have used for 88 Aqrthe tabulated K -band spectrum for a K1 III star in thePickles (1998) stellar spectral flux library. The interfer-ometric TF is then estimated as T = µ /V , and isillustrated for one representative night in Fig. 2. Thestatistical and systematic error bars on this estimationare computed following Appendix C of Kervella et al.(2004a).Once the interferometric TF has been estimated at thetime of the calibrator observations, the value of the TFand its error bars must be estimated at the time of theactual scientific observations. We have used two differ-ent techniques to perform the interpolation of the TFbetween the calibrator data points: a linear interpola-tion between the two calibrator observations bracketing a Absil et al. Fig. 2.—
Squared coherence factors ( µ ) and interferometrictransfer function estimations ( T ) for one representative observ-ing night on the D0–H0 baseline with VINCI-IONIC-2TK. Filledsymbols correspond to µ data points for the scientific target (Fo-malhaut, red diamonds) and the calibrator target (88 Aqr, bluesquares). The interferometric transfer function (TF) is representedby the empty blue squares, which are derived from the calibra-tor measurements taking into account its known diameter. Thesolid line is an interpolation of the TF, obtained by a third de-gree polynomial fit to its estimated values. The 1- σ statistical andsystematic error bars on the estimation of the TF are respectivelyrepresented by dotted and dashed lines. given scientific observations (as in Kervella et al. 2004a),and a global polynomial fit on the whole night (as inLe Bouquin et al. 2006). We have checked that the twomethods give identical results within error bars, and de-cided to use the second one for the rest of the study.A third order polynomial was found appropriate to fol-low the (small) variations of the interferometric TF dur-ing the whole night, as the target and calibrator movedacross the sky. With this method, the statistical errorbar on the interferometric TF is computed locally at anygiven time by using the covariance matrix on the param-eters of the polynomial fit, which is available on output ofthe standard IDL routine poly fit.pro . The systematicerror bar, on the other hand, is computed globally witha weighted sum of the systematic error bars on all T estimations, taking into account the correlation betweencalibrators. In the present case, since only one calibra-tor was used throughout the night, the global systematicerror is equal to the average of the systematic error barson all T estimations. Short-baseline data from the ESO archives
Even though the projected baseline lengths coveredby the Le Bouquin et al. (2006) data set are sufficientlyshort to study the circumstellar environment of Fomal-haut, we have searched the ESO archives for calibratedVINCI observations on even shorter baselines. We haveidentified four interesting nights in 2003 (September27th, September 30th, November 2nd and November3rd), where observations of Fomalhaut interleaved with acalibrator star (88 Aqr) have been obtained on the 16 mE0–G0 baseline, using the 40 cm test siderostats and theMONA beam combiner. We have used VNDRS 3.1 toreduce this data set. One night (November 2nd) wasdiscarded due to poor atmospheric conditions, and theremaining three nights produced 44 individual observa-tions of Fomalhaut of sufficient quality.Following the method described in Section 2.2, we have
Fig. 3.—
Same as Fig. 2 for one representative observing nighton the E0–G0 baseline with VINCI-MONA. A second degree poly-nomial was used to fit the estimated values of the interferometricTF.
Fig. 4.—
Sampling of the Fourier ( u , v ) plane for our completedata set, using the same color code as in Fig. 1. used the interleaved observations of the calibrator star88 Aqr to estimate the interferometric TF. In the presentcase, a second-degree polynomial was found convenientto model the TF, as illustrated for a representative nightin Fig. 3. The scatter in the interferometric TF esti-mations in Fig. 3 may seem larger than in Fig. 2, butthis is actually mostly due to the different scales for thetwo plots and to the smaller systematic error bars in thepresent data set. The reduced size of the systematic er-ror bars is related to the fact that the calibrator star ismostly unresolved on the very short E0–G0 baseline, sothat the uncertainty on its angular diameter translatesinto a much smaller systematic error on the TF estima-tion.Our final VINCI data set on Fomalhaut, spanning alarge range of baseline lengths and azimuths, comprisesa total of 171 calibrated squared visibilities. The spatialfrequencies sampled by our observations are well spreadin the Fourier ( u, v ) plane, as illustrated in Fig. 4. EVALUATING THE AMOUNT OF CIRCUMSTELLAREMISSION n Interferometric Study of the Fomalhaut Inner Debris Disk. I. 5
Fig. 5.—
Result of the fit of an oblate limb-darkened stellar photosphere to our full data set. For the sake of clarity, the solid linecorresponds to a circular photosphere with a diameter equal to the geometric mean of the diameter of the best-fit oblate stellar model, whilethe residuals are computed with the full 2D photospheric model. The dashed and dotted lines in the bottom panel represent respectivelya 1 σ and a 3 σ deviation with respect to the best-fit model. Let us first assess whether a realistic stellar photo-spheric model for Fomalhaut could reproduce the wholedata set. The description of our oblate limb-darkenedphotospheric model is given in Appendix A, showing thatonly one parameter needs to be fitted ( θ LD , the geomet-ric mean of the limb-darkened diameter). Fitting ourmodel to the whole data set gives θ LD = 2 . ± . ± .
050 mas (Fig. 5), where the first error bar accounts forthe statistical dispersion of the data while the seconderror bar is related to the systematic error in the eval-uation of the interferometric TF. This result is within2 σ of the previous estimation by Di Folco et al. (2004),but the residuals of the fit in Fig. 5 and the large re-duced chi square ( χ r = 2 .
43) clearly demonstrate thatan oblate limb-darkened photosphere cannot reproducethe data satisfactorily. It must be noted that our datasample various baseline azimuths, which makes it diffi-cult to represent the model together with the data on asingle plot. The solid line of Fig. 5 is actually computedfor a perfectly circular photosphere of diameter θ LD andis only an approximation of the actual model used tocompute the fit residuals, which take into account thevarious baseline azimuths and the stellar oblateness. Accepting the photospheric position angle recentlymeasured by Le Bouquin et al. (2009), possible improve-ments of the fit could be obtained by changing eitherthe limb-darkening coefficient ( u K ) or the photosphericoblateness ( ρ ). This is investigated in Table 1, whichgives the goodness-of-fit for various values of these twoparameters. The fit is slightly better for extreme limb-darkening coefficients ( u K . χ r are so small that we can safely statethat our interferometric data do not constrain the shapeof the photosphere in a significant way. We will thereforeuse the most plausible values (oblateness ρ = 1 .
021 andposition angle PA = 156 ◦ , see Appendix A) in the restof this study. Fitting a star-disk model to the complete data set
The failure of a realistic stellar photospheric model toreproduce the interferometric data urges us to increasethe complexity of our model. The apparent decrementof visibility observed at short baselines and the slope inthe residuals of the fit suggest that another source ofemission, fully resolved by the interferometer, must be Absil et al.
TABLE 1Goodness-of-fit ( χ r ) for the photospheric model as afunction of oblateness and limb-darkening u K = 0 . u K = 0 . a u K = 0 . u K = 1 . ρ = 1 .
00 2.46 2.46 2.45 2.43 ρ = 1 . b ρ = 1 .
05 2.42 2.42 2.41 2.40 ρ = 1 .
10 2.46 2.46 2.45 2.44 a Expected value of the limb-darkening coefficient u K in the K band (Claret 2000). b Expected value of the photospheric oblateness ρ based onour oblate photosphere model (see Appendix A). present within the interferometric field-of-view. As afirst step, we assume that this resolved emission is as-sociated with the circumstellar debris disk, and to repre-sent its contribution we use a simple model of a diffusesource uniformly distributed across the whole field-of-view. As discussed by Absil et al. (2006), such a modelis a good approximation provided that the circumstellardisk is fully resolved at all baselines. It must be notedthat in the present case, due to the very short baselines,this condition might not be completely fulfilled. The uni-form emission model should nevertheless provide a goodestimation of the flux ratio between the integrated cir-cumstellar emission and the stellar photosphere. Morerealistic circumstellar emission models will be discussedin Section 5.In Fig. 6, we have fitted the whole VINCI data set witha limb-darkened oblate stellar photosphere surroundedby a uniform circumstellar emission. The quality of thefit is very satisfactory, with χ r = 0 .
95 and no obvi-ous trend in the residuals, which are now nicely spreadaround 0. The best-fit mean limb-darkened diameteris θ LD = 2 . ± . ± .
042 mas, while the best-fitflux ratio between the circumstellar disk and the star is0 . ± . ± . σ level. Another repre-sentation of our detection is given in Fig. 7 and discussedin Appendix B.Systematic errors dominate the noise budget in ourdetection. This is not surprising as our short- andintermediate-baseline data are calibrated with a singlecalibrator star (88 Aqr). The uncertainty on this cal-ibrator’s diameter is the main contributor to the sys-tematic error budget. Using a few different calibratorswould have been preferable to reduce systematic errors,but the archival data were not originally meant to beused for this particular purpose. We are nonetheless con-fident in the robustness of our result for three main rea-sons. First, 88 Aqr is a K giant with a slow rotationalvelocity (3.6 km s − , Hekker & Mel´endez 2007) that hasalready been used as an infrared photometric standard(Bouchet et al. 1991) and as an interferometric calibra-tor by various authors without any hint of unexpectedlyhigh or low visibilities. This includes a recent studydedicated to interferometric calibrators by Richichi et al.(2009), where a UD diameter of 3 . ± .
057 mas isderived. Second, we have independently estimated itsangular diameter using surface-brightness relationships(Kervella et al. 2004b), giving a UD diameter of 3 . ± .
08 mas in good agreement with the value used in this study (3 . ± .
06 mas). Finally, we note that an im-perfect model for 88 Aqr cannot be at the origin of theobserved visibility deficit for the following reasons. Ar-tificially decreasing its angular diameter would increasethe calibrated visibilities of Fomalhaut. This would how-ever not reconcile the Fomalhaut calibrated data set witha simple photospheric model, as it would have a muchlarger influence on the data collected at mid-range base-lines (D0–H0) than on the short baseline data (E0–G0).A change in the limb-darkening coefficient would not helpeither. The presence of a circumstellar environment (diskor companion) around the calibrator could only reducethe measured calibrator visibilities with respect to theirexpected values, which would result in an increase of thecalibrated visibilities for the scientific target and therebyreduce the measured disk/star flux ratio with respect toits actual value. Therefore, the probability that our de-tection is based on a systematic effect related to the cal-ibrator is considered to be extremely low.Our revised estimation of the stellar limb-darkened di-ameter of Fomalhaut ( θ LD = 2 . ± . ± .
042 mas)is within the error bar of the previous estimation byDi Folco et al. (2004), as it is mainly based on long-baseline data which are hardly affected by the presence ofthe circumstellar emission. The large systematic error isdue to the fact that most data (at short- and mid-baselinelengths) have been calibrated with the same referencestar. A better accuracy on the stellar diameter can actu-ally be obtained by fixing the disk/star contrast at 0.88%and by fitting only the long-baseline data, which havebeen obtained with five different calibrators. The finalresult is then θ LD = 2 . ± . ± .
021 mas ( χ r = 0 . . ± .
022 mas and 2 . ± .
022 mas(apparent major and minor axes), taking into accountthe apparent oblateness of 1.021.
Fitting a binary star model to the complete data set
Besides a circumstellar disk, another potential sourceof visibility deficit at short baselines would be a faintpoint-like object within the interferometric FOV aroundthe target star. To reproduce the observed visibilitydrop, the off-axis object should have a flux ratio of about0.88% with respect to Fomalhaut as seen through theinterferometer in the K band, which is equivalent to amagnitude K = 6 . K = 6 . . ′′
8) of the in-strumental FOV. Following the discussion in Absil et al.(2006), we estimate that the presence of such a brightbackground object within the instrumental FOV is veryunlikely (probability ∼ − ). Therefore we only con-sider the case of a bound low-mass companion.To fit our interferometric data set, we construct amodel of a binary star, assuming that the companionorbits within the plane defined by the outer dust ringimaged at visible and sub-millimetric wavelengths. Wetherefore set the inclination and position angle of theprojected orbit to 65 . ◦ ◦ , respectively. In prac-tice, coplanarity with the outer dust ring is probably notn Interferometric Study of the Fomalhaut Inner Debris Disk. I. 7 Fig. 6.—
Result of the fit of a star-disk model to our full data set. The solid line represents the best fit star-disk model, while the dottedline represents the best-fit result with a single star for comparison, using the same conventions as in Fig. 5. mandatory to ensure its stability, but the purpose hereis only to check whether binary star models could actu-ally fit the data rather than to explore all the possibleorbital solutions. For the sake of simplicity, we will fur-ther assume that the orbit is circular, so that we are leftwith three parameters: the semi-major axis, the orbitalphase at a given time t and the binary flux ratio (or contrast ). Using a large number of potential values forthese three parameters, we compute the position of thecompanion at our observing dates and the associated vis-ibility of the binary system. We deduce the chi squarebetween the observed visibilities and the computed ones,and for each couple semi-major axis / contrast, we searchfor the minimum χ as a function of the orbital phase at t . Thereby, we produce a χ map, which is representedin Fig. 8 as a function of semi-major axis and binarycontrast.The χ map shows that a whole range of semi-majoraxes and contrasts fit the interferometric data in a sat-isfactory way. The effect of the Gaussian beam pro-file is evident in this figure, as the binary contrast re-quired to fit the data increases with the binary sepa-ration. We have restricted the plot range to contrastssmaller than 5%, as brighter companions would mostprobably have been detected by simple near-infrared spectro-photometric measurements. Because the instru-mental transmission becomes extremely small at linearradii larger than 30 AU (i.e., angular separations largerthan 4 ′′ ), binaries with larger semi-major axes cannotreproduce our interferometric data set. Conversely, forany semi-major axis smaller than about 25 AU, one canfind suitable combinations of orbital parameters and fluxratio to fit the data satisfactorily (reduced χ ∼ . ′′
8, i.e., about6 AU), the best-fit binary flux ratio is about 0.88% asexpected.In conclusion, even though our interferometric mea-surements span a large range of time, baseline lengthsand azimuths, they are not sufficient to discriminate be-tween a circumstellar disk and a point-like companion asthe source of visibility deficit. This is mostly due to the(lack of) observing strategy, which was not optimized forsuch a goal. NATURE OF THE NEAR-INFRARED EXCESS SOURCE
The previous section has shown that our interferomet-ric data set is not sufficient to discriminate between apoint-like object and an extended source as the origin ofthe near-infrared excess emission. Therefore, we need to Absil et al.
Fig. 7.—
Result of the fit of a star-disk model to our full data set, showing the data for two particular baselines as a function of the hourangle of the observation ( left : the small E0–G0 baseline, right : the mid-range D0–H0 baseline). The blue curve represents the visibilityof the star alone for our best-fit star-disk model, and its thickness corresponds to the 3 σ uncertainty related to the final error bar on itsdiameter. The black solid line represents the best-fit star-disk model, with its 3 σ confidence interval shown as dotted lines. Fig. 8.—
Reduced chi square map for the fit of a binary star model, plotted as a function of the orbital semi-major axis of the companionand of the binary flux ratio. The orbital dynamics are taken into account in this model, assuming a circular orbit, and the initial orbitalphase is optimized to minimize the χ r . We have limited the scale range to χ r ≤ χ r between 1 and 3.5, with a step of 0.5. use complementary data to further constrain the natureof the near-infrared emission around Fomalhaut. In thissection, we focus on the possible presence of a low-masscompanion, as well as on potential non-thermal sourcesof extended emission. The thermal emission from a cir-cumstellar dust disk will be briefly discussed in Sections 5and 6. Further constraints on a low-mass companion
Since a low-mass companion cannot be ruled out bythe interferometric data, we have searched the literature for possible additional constraints on its existence. Threemain type of observations can be used in this context: ra-dial velocities, astrometric measurements, and high dy-namic range single-pupil images. Before discussing theexisting data in the literature, it is useful to convertthe companion flux ratio into an estimated mass, usingthe evolutionary models of Baraffe et al. (1998) for low-mass stars. Assuming an age of 200 Myr for Fomalhaut(Di Folco et al. 2004), and taking into account its dis-tance of 7.7 pc, a K -band magnitude of 6.1 for a close-inlow-mass companion translates into a mass of 0 . M ⊙ .n Interferometric Study of the Fomalhaut Inner Debris Disk. I. 9Wider companions would have larger masses to accountfor the reduced off-axis transmission of VLTI/VINCI,e.g., 0 . M ⊙ at 0 . ′′ . M Jup at 3 days, 3 . M Jup at 10 days, and 10 M Jup at100 days, the latter corresponding to a semi-major axisof 0.5 AU. Since such planetary companions would pro-duce flux ratios largely below 0.88% in the K band, thisbasically rules out any companion closer than 0.5 AU asthe source of the detected near-infrared excess.An anomalous proper acceleration of 6.6 mas/yr hasbeen observed in Hipparcos astrometry towards Fo-malhaut (Chiang et al. 2009). This acceleration ismarginally significant ( ∼ σ ), the Fomalhaut astrometrybeing otherwise stable. Such quasi-steady accelerationcould be caused by a companion whose orbital period islonger than ∼ a ≥ . M Jup at 2.6 AU, and wouldincrease to 50 M Jup at 7 AU, or 300 M Jup (i.e., 0 . M ⊙ )at 15 AU. The latter would produce a K -band flux ratioof about 0.6% with respect to Fomalhaut. Companions oflarger masses at the same orbital distances would be in-compatible with the Hipparcos data, while companionswith orbital periods smaller than 3 yr should have massessmaller than 9 M Jup to be consistent with
Hipparcos ob-servations. Because a companion with a 0.6% flux ratioat an orbital semi-major axis of 15 AU is mostly incon-sistent with the VINCI data ( χ r ∼ .
8, see Fig. 8), theseastrometric measurements rule out the low-mass binaryscenario as an explanation for the VINCI near-infraredexcess up to orbital semi-major axes of about 15 AU. Atlarger distances, the mass of a bound companion com-patible with the
Hipparcos acceleration would becomemarginally consistent with the VINCI near infrared ex-cess.High contrast imaging with large diffraction-limitedtelescopes can be used put complementary constraintson low-mass companions at larger angular distances. Inparticular, the HST/ACS observations of Fomalhaut byKalas et al. (2005, 2008) show no low-mass companionsat the sensitivity limit of the instrument, except for theplanetary-mass companion at 119 AU (well outside theVINCI field). According to Kalas et al. (2002), the ACSinstrument can detect magnitude differences up to 11 at0 . ′′ I = 12 . I band, which roughlycorresponds to a 0 . M ⊙ object at 200 Myr and 7.7 pcfollowing the Baraffe et al. (1998) models. This objectwould be much too faint to explain the K -band excessderived from VINCI observations, so that the effect ofa companion located at an angular distance larger than0 . ′′ . ◦ M band observations with the AO-assisted Clio imager on the 6.5-m MMT have furtherconstrained the range of allowed companion masses inthe 5–40 AU region around Fomalhaut (Kenworthy et al.2009). In particular, objects with masses larger than13 M Jup can be ruled out at distances between 8 and40 AU. Taking into account all these constraints, thepresence of a bound low-mass companion around Foma-lhaut, whatever its angular separation, cannot be at theorigin of the near-infrared excess detected with VINCI.And since the probability to find a K ∼ Possible sources of extended emission
Among the possible sources of extended emission in theclose vicinity of Fomalhaut (mostly within 0 . ′′ K -band excess. Gaseousdisks, on the other hand, are generally occurring in classi-cal Be or Ae stars, a category of hot rapidly rotating starsclose to the main sequence that exhibit prominent hydro-gen emission lines. Although Fomalhaut does not rotateas fast as classical Ae/Be stars, and has a spectral typelater than most emission-line stars, the possible presenceof a gaseous disk cannot be ruled out a priori . Sinceinfrared excesses in such stars are generally correlatedto the equivalent width of their H α emission line (seee.g. van Kerkwijk et al. 1995, and references therein), wehave searched the literature for possible evidence of H α emission around Fomalhaut. The H α photospheric ab-sorption line profile of Fomalhaut was inspected in detailby Gardiner et al. (1999) and Smalley et al. (2002). Inparticular, the observed H α line profile was fitted with amodel Balmer profile to derive the effective stellar tem-perature, which is found by Smalley et al. (2002) to be0 Absil et al.in agreement with the fundamental effective tempera-ture derived by other methods. No peculiar feature isreported in the H α absorption line, including no obviousemission line inside it. Furthermore, the photometric in-dex of the H α line was monitored between 1968 and 1974by Dachs & Schmidt-Kaler (1975), showing no variabil-ity and no evidence of H α emission for Fomalhaut in 19observations.Finally, let us note that Fomalhaut is located wellinside the Local Interstellar Bubble (LIB), so thatthe heating of interstellar material proposed to explainthe presence of infrared excesses around shell stars or λ Bo¨otis stars located about 100 pc away (Abt 2004;Mart´ınez-Galarza et al. 2009) is unlikely to be at the ori-gin of the observed near-infrared excess. Consequently,we consider the presence of hot dust in the inner plan-etary system of Fomalhaut as the only plausible expla-nation to the observed near-infrared excess, and furtherdiscuss this scenario in the next sections. CONSTRAINING THE INNER DISK MORPHOLOGY
We have previously derived in Section 3.1 the inte-grated flux ratio between the circumstellar disk and thestellar photosphere using a very simple brightness distri-bution for the disk: a uniform emission across the instru-mental FOV. Accepting that an inner dust disk is indeedat the origin of the detected near-infrared excess, we dis-cuss here the limitations to this preliminary analysis andinvestigate whether the morphology of the disk could beconstrained based solely on our interferometric data set.As a first step, we have tried to use three very differ-ent morphologies for the circumstellar disk to reproducethe data set: a uniform emission extending across thewhole field-of-view as in Section 3.1, a narrow ring ofdust located at the sublimation radius ( r sub ≃ . T sub = 1700 K), and thezodiacal disk model of Kelsall et al. (1998), which is im-plemented in the Zodipic package . Fig. 9 illustratesthe influence of the disk morphology on the squared vis-ibility at short baselines. The small angular extent ofthe ring structure leads to significant oscillation in thesquared visibility, whose frequency depends on the az-imuth of the baseline since the disk is inclined. Smalleroscillations are visible at a similar frequency for the zo-diacal disk model, because the inner rim dominates the K -band emission of the disk. Fitting these three modelsto our complete data set leads to similar results in termsof best-fit flux ratio and reduced χ (see Table 2), exceptfor the ring structure which gives a larger best-fit con-trast of ∼ .
3% with a slightly increased χ r . The largerbest-fit contrast is mainly due to the fact that the ringis not fully resolved at the shortest baselines ( .
10 m),so that a larger disk emission is required to produce thesame squared visibility deficit at those baselines. Notehowever that the best-fit contrast is only about 3 σ aboveour original estimation, which indicates that, even withthis extreme morphology, the final disk/star flux ratiodoes not heavily depend on the distribution of the cir- The ring width is forced to be much smaller than the interfer-ometer’s angular resolution so that it does not impact the visibilityestimations. Zodipic is an IDL program developed by M. Kuchner et al. forsynthesizing images of exozodiacal clouds. It can be downloadedat http://asd.gsfc.nasa.gov/Marc.Kuchner/home.html . Fig. 9.—
Comparison of our data set with various disk models atshort baselines. The data have been separated into three ranges ofbaseline azimuths to allow the comparison, since the model visibil-ities depend on both the baseline length and azimuth. In our mod-els, the narrow ring is located at the sublimation radius of blackbody dust grains, assuming a sublimation temperature of 1700 K,while the zodiacal disk follows the Kelsall et al. (1998) model.
TABLE 2Best-fit flux ratio and goodness of fit for threedifferent disc models
Uniform emission Zodiacal disc Narrow ring a Flux ratio 0.88% 0.85% 1.26% χ r a The ring is located at the dust sublimation radius ( ∼ . cumstellar emission.As a second step, we have generated images of a geo-metrically and optically thin debris disk toy model basedon pure black body assumption (grain temperature pro-portional to r − . , with T = 1700 K at 0.1 AU), using arange of values for the two most important geometric pa-rameters: the inner disk radius r in and the exponent α ofthe power-law describing the density decrease as a func-tion of distance ( n ( r ) ∝ r − α ). The synthetic images onlyinclude thermal emission, which is expected to be largelydominant for the dust temperatures explored here. Foreach couple of parameters ( r in , α ), we compute the vis-ibility of the star-disk system at the relevant observingdates and adjust the disk/star flux ratio in order to min-imize the χ distance between the modeled and observedvisibilities. By exploring a range of possible values forthe two parameters, we produce a χ map, which is dis-played in Fig. 10 together with the associated best-fitflux ratios. In these toy models, we have assumed a dustsublimation temperature of 1700 K. All models with in-n Interferometric Study of the Fomalhaut Inner Debris Disk. I. 11 Fig. 10.—
Left : Reduced chi square map for the fit of our simple geometrical disk toy model (see text for description), as a function oftwo free parameters: inner disk radius and density power-law exponent. The flux ratio between the disk and the star has been optimizedto minimize the χ r values. Right : Optimum disk/star flux ratio associated to each disk model displayed on the χ map. put inner radii smaller than the sublimation radius havetheir inner radius forced to r in = 0 . χ observed in the map ofFig. 10 suggests that all models fit almost equally wellour data set, so that the morphology of the circumstel-lar disk can not be meaningfully constrained. It muststill be noted that models with very steep density pro-files ( α >
6) starting at the sublimation radius have aslightly larger χ r . These particular models roughly cor-respond to narrow rings of dust located at the sublima-tion radius, which do not reproduce very well the dataat the shortest baselines (see Fig. 9). For dust rings withlarger inner radii, the oscillation frequency in the visi-bility curve is increased (the disk is fully resolved at ashorter baseline) and several local minima are then ob-tained in the χ map, e.g. at r in = 0 . r in = 0 . α <
2) give equal fit qualities whatever the innerradius, because their spatially extended brightness dis-tribution produces almost no oscillation in the visibilitycurve (see green dash-dotted curve in Fig. 9). It mustalso be noted that, whatever the chosen morphology ofthe dust disk, the best-fit disk/star flux ratio remainssimilar to the value derived with the uniform emissionmodel. The largest discrepancy is found for the modelswith steep density power-laws starting at the sublima-tion radius, which give a best-fit contrast of 1.16% (i.e.,about 2 σ above our original estimation). This observa-tion reinforces our statement that the disk morphologyhas a weak influence on the best-fit disk/star contrast. DISCUSSION
The discovery of hot dust grains in the inner plane-tary system of Fomalhaut cannot be regarded as a to-tally unexpected result. First, the data collected byStapelfeldt et al. (2004) with the MIPS and IRS instru-ments onboard Spitzer have already shown the presenceof a warm excess emission at wavelengths between 17and 34 µ m. MIPS imaging suggests that most of the24 µ m excess is unresolved and originates from a 10 ′′ re-gion around Fomalhaut, i.e., well within the main dust ring located at 140 AU. Second, other A-type stars havealready been shown to harbor similar near-infrared ex-cesses that have been interpreted as the signature of hotdust (Absil et al. 2006, 2008; Akeson et al. 2009). In allprevious cases, the absence of significant excess emissionat wavelengths around 10 µ m has suggested that the dustsurface density distribution is very compact and doesnot follow a classical r − . power-law. Modeling theseinterferometric observations has still shown that realis-tic debris disk models can be found to fit the observednear-infrared excesses of a few percent around A-typestars without producing too large amounts of excess atlonger wavelengths, that would have been noticed in mid-infrared spectro-photometric measurements.It is not our purpose here to thoroughly study the prop-erties of the Fomalhaut inner disk, as this will be done ina companion paper, where we include additional interfer-ometric data collected at 10 µ m with the Keck Interfer-ometer Nuller. It is nevertheless useful to check whetherstandard disk models could both reproduce the measured K -band excess and be consistent with photometric dataat longer wavelengths. For that, we have used our debrisdisk toy model based on pure black body grains, assum-ing an r − . surface density power law from the dustsublimation radius up to 140 AU. In order to reproducethe measured K -band excess, this disk must have a bolo-metric luminosity ratio L disk /L ∗ ∼ × − with respectto the stellar photosphere. The excesses produced at 17.5and 24 µ m would then be of 0.38 and 0.22 Jy respectively,which would correspond to the measured IRS “on-star”excess of 0 . ± . µ m, while being fully withinthe measured MIPS “on-star” excess of 0 . ± . µ m (Stapelfeldt et al. 2004). The predicted flux inthe “extended” disk located outside 10 ′′ , amounting to0.06 Jy, would also be within the MIPS PSF-subtractedestimation of 0 . ± .
02 Jy for the outer disk regions.Our model is also consistent with published photomet-ric measurements in the mid-infrared (including an IRAS12 µ m observation), which are generally of poor accuracy.We will however show in a companion paper that thissimple model cannot reproduce interferometric data ob-tained in the mid-infrared with the Keck interferometer,as well as unpublished IRS data in the 10–18 µ m region.This suggests that the disk cannot be continuous under2 Absil et al.the r − . power law assumption, and therefore cannotbe directly connected with the outer belt using classicalmodels (e.g., based on Poynting-Robertson drag). CONCLUSION
In this paper, we have re-analyzed published and un-published archival data from the VLTI/VINCI instru-ment in search for near-infrared circumstellar emissionaround the young A4V star Fomalhaut. The large num-ber of visibility measurements at short baselines (146 in-dividual measurements with projected baselines smallerthan 60 m), and the good accuracy of individual datapoints (median relative error bar of 2%), have allowedus to directly detect the presence of resolved circumstel-lar emission at a level of 0 . ± .
12% with respectto the stellar photosphere in the K band. If this ex-cess emission was produced by a dust disk similar to thesolar zodiacal disk, its surface density would be about5000 times as large as in the solar system. Our analysisof the whole VINCI data set also provides an updatedmean limb-darkened diameter θ LD = 2 . ± .
022 masfor the stellar photosphere, which is predicted to havean apparent oblateness of 1.021. This measurement doesnot significantly differ from the previous estimation byDi Folco et al. (2004), which did not take into accountthe presence of circumstellar emission.We attempted to constrain the morphology of the cir-cumstellar emission source with our interferometric dataset, but failed to discriminate between a point-like sourceand an extended source, mostly due to the inappropriate observing strategy of this heterogeneous data set. Addi-tional constraints obtained with complementary observ-ing techniques have thus been used to determine the na-ture of the excess emission, showing that the presenceof a point-like source within the interferometric field-of-view (either a bound companion or a background object)is very unlikely to reproduce the observed K -band ex-cess. Non-thermal extended sources are also shown to bemostly inconsistent with other observations. We there-fore consider that a hot dust population located mostlywithin the first 6 AU around Fomalhaut is the most likelyexplanation for the observed K -band excess.In a companion paper, we will introduce further con-straints on the inner dust disk thanks to high dynamicrange mid-infrared observations with the Keck Interfer-ometer Nuller, and complement these new constraintswith various spectro-photometric measurements in orderto investigate the nature and distribution of the dustgrains.O.A. acknowledges the financial support from theEuropean Commission’s Sixth Framework Program asa Marie Curie Intra-European Fellow (EIF) while atLAOG, and from a F.R.S.-FNRS Postdoctoral Fellow-ship while at IAGL. This research was partly fundedby the International Space Science Institute (“Exozodi-acal Dust Disks and Darwin” working group) and by anEGIDE/PHC Procope programme ( Facilities:
VLTI(VINCI)
APPENDIX
OBLATE LIMB-DARKENED PHOTOSPHERIC MODEL
To properly fit our interferometric data, which are taken at various baseline azimuths, we must take into account theinfluence of stellar oblateness due to the rapid rotation of Fomalhaut ( v sin i = 93 km s − , Royer et al. 2007). Followingthe discussion of Absil et al. (2008), we compute the apparent oblateness ρ of the Fomalhaut photosphere, defined asthe ratio of the major and minor apparent radii ( ρ = R a /R b ), using the following equation: ρ ≃ ( v sin i ) R m GM ∗ + s (cid:18) ( v sin i ) R m GM ∗ (cid:19) , (A1)where R m = ( R a R b ) / is the geometric mean between the major and minor apparent radii. Taking for R m the stellarradius derived by Di Folco et al. (2004) based on long-baseline interferometric measurements ( R m = 1 . R ⊙ ), andwith M ∗ = 2 . M ⊙ (Di Folco et al. 2004), the apparent oblateness of the photosphere is ρ = 1 . star = 156 ◦ ± ◦ ) and that of the debris ring (PA disk = 156 . ◦ ± . ◦
3, Kalas et al. 2005). We will furtherassume that the photospheric limb darkening can be approximated by a linear law and use the K -band limb-darkeningcoefficient tabulated by Claret (2000) for an A4 V star ( u K = 0 . θ LD . IMPORTANCE OF SHORT BASELINES FOR DEBRIS DISK DETECTION