A&A 387, 285-293 (2002)
R. J. Laureijs1, - M. Jourdain de Muizon2,3, - K. Leech1, - R. Siebenmorgen4 - C. Dominik5 - H. J. Habing6 - N. Trams7 - M. F. Kessler1,
1 - ISO Data Centre, ESA Astrophysics Division, Villafranca del Castillo, PO Box 50727, 28080 Madrid, Spain
2 - LAEFF-INTA, ESA VILSPA, PO Box 50727, 28080 Madrid, Spain
3 - DESPA, Observatoire de Paris, 92190 Meudon, France
4 - ESO, K. Schwarzschildstr. 2, 85748 Garching bei München, Germany
5 - Astr. Inst. Anton Pannekoek, Univ. Amsterdam, Kruislaan 403, 1098 SJ, Amsterdam, The Netherlands
6 - Leiden Observatory, PO Box 9513, 2300 RA Leiden, The Netherlands
7 - Integral Science Operations, ESA ESTEC, Keplerlaan 1, 2201 AZ Noordwijk, The Netherlands
Received 10 August 2001 / Accepted 8 March 2002
We present an ISO 25 m photometric survey of a sample of 81 nearby main-sequence stars in order to determine the incidence of "warm'' dust disks. All stars were detected by ISO. We used an empirical relation to estimate the photospheric flux of the stars at 25 m. We find 5 stars (6%) with excess above the photospheric flux which we attribute to a Vega-like disk. These stars show disk temperatures not warmer than 120 K. Our study indicates that warm disks are relatively rare. Not a single star in our sample older than 400 Myr has a warm disk. We find an upper limit of for the mass of the disks which we did not detect.
Key words: stars: planetary systems - stars: general - infrared: stars
After the initial discoveries by IRAS, the search and analysis of Vega-like disks in the infrared has received a substantial boost with the availability of data from the Infrared Space Observatory (ISO, Kessler et al. 1996). ISO has improved on IRAS in several important ways. Firstly, the number of bands has been increased and the infrared wavelength coverage has been extended to 200 m; secondly, the detection limits have been lowered; and, thirdly, imaging and spectroscopy have been made possible on arcsec scales. One major ISO finding, based on a statistical study, is that the detection of a debris dust disk depends strongly on the age of a star: the probability of detecting a Vega-like disk comes close to unity for main-sequence dwarfs of less than 400 Myr (Habing et al. 1999; Habing et al. 2001, hereafter Paper I). Disks around older main-sequence stars are much less frequent, but they still exist. The precise mechanism that prevents these older disks from dissipating is still an open question (Jourdain de Muizon et al. 2001).
The search for new Vega-like stars has been based either on the analysis of the infrared colours using the IRAS database (Fajardo-Acosta et al. 2000; Mannings & Barlow 1998, and references therein for previous IRAS surveys), or by comparing the far-infrared flux with a prediction based on a photospheric model or extrapolation from optical photometry. In most cases the surveys rely on the measurements at 60 m, because the excess emission is high compared to the photospheric flux and the background confusion is low compared to observations at longer wavelengths.
A significant excess at shorter wavelengths is a signature of warm debris material, presumably closer to the star than the particles emitting at the longer wavelengths. Detecting such an excess generally requires a high photometric accuracy due to the relatively large contribution of the photospheric emission.
In this paper we analyse photometric data at 25 m of a sample of 81 main-sequence stars in order to determine the fraction of the stars which have significant infrared excess at 25 m.
In Sect. 2 we describe the sample, the different data sets and the processing of the ISOPHOT data. In order to achieve the highest possible photometric accuracy we merged the ISO and IRAS data taking into account the systematic differences in calibration between the two data sets. The merging of the data sets and the extraction of the stars with 25 m excess are described in Sect. 3. The results are analysed in Sect. 4. In Sect. 5 we discuss the properties of the excess stars. The conclusions are stated in Sect. 6.
For a full description and discussion of the sample we refer to Paper I; here we briefly describe the properties of the 25 m sample. The stars were selected on the following selection criteria: (1) all stars are within 25 pc of the Sun; (2) a spectral type later than B9 but earlier than M, and only dwarfs of type IV-V or V; (3) a predicted photospheric flux at 60 m of at least 30 mJy; (4) no (optical) binaries within the aperture; (5) no known variables.
The third selection criterion implies that the minimum photospheric flux at 25 m is of the order of 90 mJy. During the mission, some stars were added but these have not biased the statistics (cf. Paper I, Sect. 5.4).
In Table 1 we have listed the properties of the stars of the sample observed at 25 m. Columns 1 to 3 describe the catalogue names of the stars, including the ISO observation identifier. Columns 4 to 6 list the optical stellar parameters from the Hipparcos Catalogue (Perryman et al. 1997), the spectral types in Col. 6 come from the machine-readable version of the Hipparcos Catalogue. Column 7 gives the age of the star from Lachaume et al. (1999). Columns 8 and 9 give the predicted flux density from Eq. (1) (below) and IRAS flux density, respectively. Columns 10 and 11 list flux density plus uncertainty obtained from the ISO data. Columns 12 and 13 give the adopted flux density plus uncertainty as described in Sect. 3.1. Finally, Col. 14 lists possible flags - "E'' indicating a significant excess, and "N'' indicating that the star was not included in the 60 m list of Paper I due to instrumental reasons and visibility constraints during the ISO mission.
Notes: HD106591 was observed twice, the ISO flux density is the weighted average; the second ISO_id is 33700128.
HD110833 is not in the list of Lachaume et al. (1999), the age is estimated by us according to Lachaume et al. (1999)
The IRAS data were obtained from the IRAS faint source catalog (IFSC, Moshir et al. 1989). When no IFSC data are available we have taken data from the IRAS point source catalog.
In order to decide whether a star has any excess emission at 25 m
we have used the relation derived by Plets (1997):
Equation (1) is tied to the IRAS calibration. We adopted a flux density of 6.73 Jy for =0 mag, this value is consistent with the IRAS photometric calibration (IRAS Explanatory Supplement).
For the analysis of likely excess stars we needed to estimate the photospheric flux at 12 m. We used the relationship similar to Eq. (1) derived by Waters et al. (1987). The photospheric flux densities in the far-infrared ( m) were estimated by assuming that for a given star the magnitude longward of 60 m is identical to the 60 m magnitude as given in Paper I.
The values of V and B-V were taken from the Hipparcos Catalogue (Perryman et al. 1997). At the low flux end with 300 mJy, an uncertainty of mV= 0.01 corresponds to about 2 mJy and B-V= 0.01 corresponds to about 5 mJy. Consequently, the statistical uncertainty in the "predicted'' flux density according to Eq. (1) is estimated to be of the order of 2-3%.
The ISOPHOT data at 25 m (Lemke et al. 1996) were collected throughout the ISO mission (Leech & Pollock 2000) with observation template AOT PHT03 in triangular chopped mode (Klaas et al. 1994). The chopper throw was 60'' and the aperture used was 52''. The on-target exposure time was 128 s and an equal amount of time was spent on the two background positions.
The data were processed using ISOPHOT interactive analysis PIA Version 8.1 (Gabriel et al. 1997). All standard signal corrections were applied. A generic chopper pattern of two plateaux of 4 "source plus background'' and 4 "background'' signals were derived. For the signal difference we have taken the average of the last two signals of each plateau. The signals were converted to flux densities under the assumption that the responsivity of the detector has the same value at the beginning of each ISO revolution, and changes with orbital phase due to ionising radiation according to an empirical function tabulated in the "Cal G'' table PPRESP (Laureijs et al. 2001).
To check the ISO results we have correlated the ISO fluxes with the predicted fluxes and found a tight correlation. However, the correlation is not along the line of unit slope but along a power law where the high fluxes ( Jy) are systematically underestimated and the lower fluxes ( mJy) are systematically overestimated with respect to the model predictions, see Fig. 1.
|Figure 1: Comparison between the photometric data of ISO and IRAS. Left panel: the correlation between ISO and IRAS. For the ISOPHOT data at 25 m we assumed a default detector responsivity and standard calibrations. Right panel: the same correlation after correction of the ISO fluxes to obtain an overall match with the predicted fluxes. The solid lines represent the lines of unit slope.|
|Open with DEXTER|
Assuming that the majority of the stars in our sample have no significant excess emission at 25 m we recalibrated the ISO fluxes to the predicted fluxes so that the correlation between subsamples of the two data sets scatters around the line of unit slope. The method is illustrated in Fig. 1 where the correlations are given between the ISO and IRAS fluxes before and after the correction. A detailed description of the recalibration is given elsewhere (Laureijs & Jourdain de Muizon 2000). It should be stressed that the recalibration systematically changes the fluxes in the ISO sample as a whole and does not affect the relative scatter amoung the individual observations. Also, the Plets (1997) predictions are based on the IRAS calibration, whereas the ISOPHOT calibration is based on a different photometric system. The recalibration ensures that the fluxes of the ISO observations are consistent with the IRAS calibration.
The predicted, IRAS, and recalibrated ISO fluxes are listed in
Table 1. The IRAS and ISO fluxes have been colour corrected
assuming a stellar photosphere, the colour correction factors are 1.40
and 1.28 for IRAS and ISO, respectively. The table includes the adopted flux
at 25 m, determined in the following way:
|Figure 2: V- versus B-V colour for all stars in the sample. The solid line is the photospheric emission as predicted by Eq. (1). The dashed line is the relationship for K and M dwarfs with B-V > 0.8 derived by Mathioudakis & Doyle (1993).|
|Open with DEXTER|
The visual-infrared colour-colour diagram for the stars in the sample is presented in Fig. 2. The predicted photospheric flux follows closely the distribution of points in the sample indicating that Eq. (1) is applicable. The good match also indicates that most of the 25 m flux densities are predominantly photospheric. The sample contains 5 stars with B-V >1.0 (Table 1), these stars are all K dwarfs and have V- above the prediction. On the other hand, the relation derived by Mathioudakis & Doyle (1993) for K and M dwarfs predicts values of V- which are too high for the three stars with highest V-.
To assess the photometric quality of the sample we present a histogram of the difference in Fig. 3. The distribution is strongly peaked and suggests a normal distribution close to zero for the majority of stars in the sample.
|Figure 3: Distribution of deviations from the predicted fluxes. The dashed bins give the number of stars below and above the given flux limits. The solid line is a normal distribution based on 35 mJy, mean = 8 mJy.|
|Open with DEXTER|
The parameters of the normal distribution have been derived as follows. Initially, an intermediate mean and standard deviation was derived of the stars in the interval mJy. Judging from the histogram we decided that the stars falling outside this interval must be outliers. Subsequently, all stars were rejected which are more than 2.6 standard deviations away from the mean (i.e. 1% probability of occurrence). This yields an interval of mJy. The mean of the remaining stars is 8 mJy with a dispersion of 35 mJy. The normal distribution has been included in Fig. 3. Application of a Kolmogorov-Smirnov test showed that the distribution in the given interval is normal with a significance level of 5%.
This analysis indicates that for the majority of the stars in the sample, the 25 m fluxes are consistent with the expected photospheric fluxes. The overall scatter between the observations and expectations is 35 mJy. Judging from the individual uncertainties of the 25 m fluxes, we conclude that most of the scatter must come from the infrared measurements and that the predicted fluxes which are only based on optical data are very accurate, within 8 mJy for the sample average. In addition, there is no indication that the distribution is non-normal, suggesting that there is no statistical evidence for a surplus of positive excesses in the distribution for mJy.
|Figure 4: The properties of the stars more than 3 away from the peak of the distribution presented in Fig. 3. The error bars present . The star HD 39060 ( Pic) is not included due to its off-scale positive excess of a lot.|
|Open with DEXTER|
A total of 11 (=14%) targets fall outside the interval, and 2 out of these 11 targets are below the expected flux. For these outlying stars we have plotted in Fig. 4 the ratio . In order to indicate the significance of the deviation we have included error bars, highlighting the uncertainties in the individual measurements.
Figure 4 indicates that 7 stars (6 stars plus Pic) with have fluxes which are more than above the predicted values. These stars have been flagged in Table 1.
Two of the 7 excess stars are classified as K dwarfs (Table 1). It is likely that Eq. (1) does not apply for these type of stars but rather the relation derived by Mathioudakis & Doyle (1993), see also Fig. 2. Indeed, for HD 88230 Mathioudakis & Doyle (1993) predict a photospheric flux of 513 mJy which is above mJy observed by us. For HD 191408 the value for B-V is low, in the regime where the difference between Eq. (1) and the relation by Mathioudakis & Doyle is small. The measured 25 m flux of 462 mJy for HD 191408 is still more than above the flux of 383 mJy expected for K dwarfs. In conclusion, we reject the detection of an excess in HD 88230.
We have listed resulting the excess stars in Table 2. The excess emission at 25 m depends on the shape of the spectral energy distribution and the response of the filterband. We only determined the in-band excess emission for the ISO observations.
If the excess emission is not point-like but comes from a region which is a significant fraction of the beam profile, then the averaging of the two flux measurements is not valid. In all positive excess cases except for HD 191408, the IRAS excess at 25 m is larger than the ISO excess. This might indicate that the excess emission is extended and has partly been resolved by ISO (cf. Fig. 5).
Using IRAS and published ISO observations obtained at other wavelengths we have determined the far-infrared spectral energy distributions of the excess stars. The spectral energy distributions after subtraction of the photospheric emission component have been plotted in Fig. 5. When the observed emission in the IRAS 12 m band is within 5% of the estimated photospheric emission, an upper limit of 5% photospheric emission is presented.
The 12/25 flux ratio for HD 191408 yields a colour temperature of about 600 K assuming a dust emissivity. In combination with the ISO upper limit at 60 m we consider it more likely that the excess emission is due to coronal free-free emission. Also at 12-25 m, a typical power law for free-free emission ( ). We therefore exclude this star in the subsequent analysis.
|Figure 5: The far-infrared spectral energy distributions of the 25 m excess stars (cf. Table 2) after subtraction of the photospheric emission; open squares: the excess emission derived from IRAS data, filled squares: ISO data. The ISO data at 60 and 170 m were taken from Paper I, for Leo we added 60 and 90 m data from the ISO archive. The solid lines are modified blackbody curves fitted to the 25/60 colour of the ISO excess emission. Except for the upper limits and HD 191408, all data points were colour corrected according to the corresponding modified blackbody.|
|Open with DEXTER|
Of the five remaining stars we have derived the 25/60 colour temperatures to analyse the temperature of the dust causing the 25 m excess emission. Assuming a dust emissivity, we find temperatures between 49 and 122 K. The inferred temperatures are included in Table 2. HD 38678 has the highest temperature (122 K) and is the only star for which the 25 m excess flux density is higher than that at 60 m.
The fraction of Vega-like stars in our sample with a significant excess at 25 m is 5 out of 81 or 6%. Did we overlook genuine 25 m excess stars? There are two stars for which IRAS shows a significant excess which cannot be confirmed by ISO. The IRAS measurement of g Lup (HD 139664) at 25 m (493 mJy) would indicate a strong excess above the photosphere. We find, however, an ISO flux ( mJy) which is close to the predicted photospheric flux. This star shows one of the largest discrepancies between ISO and IRAS. HD74576 (see Table 1) is the other star where IRAS would indicate an excess higher than 120 mJy but is rejected because of an inconsistent ISO measurement. Based on these two cases we conclude that from our 25 m sample, the uncertainty in the number of excess stars is at most two, giving a most probable fraction of Vega-like stars with 25 m excess of 6% and a maximum possible fraction of 9%. This is smaller than the fraction found at 60 m (18%) in Paper I.
The non-detection of significant 25 m excess emission for all other stars in the sample shows that the Vega-like disks are generally cool: the largest fraction of the dust in the disk must be colder than 120 K. The median 25 m flux in our sample is mJy (=3.3 mag). To be detectable in our sample the typical contrast C25 between emission from a presumed disk and the stellar photosphere must be greater than , i.e. C25> 0.3. Assuming a disk temperature of 120 K, the minimum detectable dust mass of the disk is estimated to be for an A0 dwarf ( 9600 K) and for a G0 dwarf ( 6000 K). See Appendix A for a description of the calculation. These masses increase for lower dust temperatures. For comparison, the minimum detectable mass in the survey at 60 m is (Paper I). Our Vega-like candidates are all included in the list of Paper I. Since all Vega candidates in Paper I have inferred masses larger than we conclude that we have detected essentially the warmest disks at 25 m.
Three stars in our sample ( Leo, Lyr, and Pic) show significantly more far-infrared emission at m than the modified black body energy distributions would predict, see Fig. 5. This could be an indication of the presence of colder dust material in the disk, presumably at larger radii from the stars.
The minimum detectable mass of for a G0 dwarf assumes an arbitrarily chosen fixed distance between the disk and the star. The detection of only A stars suggests that only stars of this stellar type are sufficiently bright to heat the dust at a minimum distance of the star. For example, a 1 m size silicate particle must be at 35 AU from an A0 star to be at a temperature of 120 K. At this distance, the temperature of a similar dust particle around a G0 star would be 86 K, yielding a minimum mass of . It is therefore more likely that the minimum detectable mass in our sample is .
The low fraction of 6% of main-sequence dwarfs exhibiting Vega-like excess emission at 25 m gives force to the result by Aumann & Probst (1991) who carried out a similar survey at 12 m. They found only 2 statistically significant excess candidates out of 548 nearby stars. These two stars ( Pic and Lep) are also found in our sample of excess candidates. Apparently, warm debris disks are rare.
A similar study by Fajardo-Acosta et al. (2000), where 2MASS data were combined with IRAS data for a sample of 296 main-sequence stars, yielded 8 systems which have a significant excess at 12 m. None of their 8 stars is in our initial sample. This low fraction (<3%) is not inconsistent with our result at 25 m. Only one of these 12 m excess stars is detected at longer wavelengths, and the spectral energy distributions of the 8 stars indicate dust temperatures in excess of 200 K. The temperatures suggest that the systems detected by Fajardo-Acosta et al. (2000) are distinct from the systems we have detected at 25 m which all have been detected at 60 m (see Fig. 5).
From a survey of 38 main-sequence stars using IRAS and ISOPHOT data Fajardo-Acosta et al. (1999) found no star with a significant excess at 12 m, and a fraction of 14% excess stars at 20 m. It is difficult to interpret this fraction since the ISOPHOT data used in their study were inconclusive, and the 20 m detections needed confirmation. In any case, the absence of 12 m detections indicates that these disks are not warmer than 200 K.
The temperatures and the inferred upper limits for the dust emission at 25 m put strong requirements to possible ground based photometric surveys of debris disks at 20 m. In order to be able to detect disks below our detection limit of , the contrast between disk emission and photospheric emission is <0.3 (equivalent to larger than 1.3 mag). On the other hand, the accuracy of predicting the infrared photospheric flux is generally not better than 5% which limits the maximum contrast to 3.3 mag. Significant improvement can only be made by imaging the disk.
All five Vega-like candidates in our sample are young, less than 400 Myr (cf. Table 1) with spectral type A0-A3, confirming the finding by Habing et al. (1999) that debris disks are mostly found around stars that just entered the main-sequence. In fact, of the 8 stars in our sample younger than 400 Myr, 5 have a detectable dust disk at 25 m, whereas none of the older stars show a significant excess.
The lower limits on the mass have been derived assuming that the size of the disk particles is much smaller than the wavelength. At 25 m this corresponds to m, where a is the radius of a grain. Larger grain sizes yield relatively lower absorption cross sections which increase our minimum mass estimate. Detailed modelling by Krügel & Siebenmorgen (1994) and Dent et al. (2000) which includes the (observed) spatial distribution in the disk, suggests much larger grain sizes of the order of a few tens of m. Such sizes could increase our lower limit of the disk mass by one order of magnitude or more.
The dust model calculations by Li & Greenberg (1998) for Pic assuming that the particles are made out of cometary material show that for a given temperature, the grains can span a whole range of distances from the star depending on the composition and mass. For K they find D=20 AU for the biggest porous silicate aggregates (of 10-4 g) to D=200 AU for the smallest ones (of 10-14 g).
A 25 m survey of 81 late type main-sequence dwarfs using ISO and IRAS data showed that 5 (or 6%) of all stars in the sample exhibit significant infrared excess which can be attributed to a Vega-like dust disk. The low fraction and the fact that the disks have already been identified at 60 m indicates that the bulk emission from Vega-like disks is from cool dust ( K).
From the detection limit of Vega-like disks we estimate a lower mass limit of for the disks not detected by us. The survey confirms that there seems to be an absence of detectable amounts of dust at close distances 20 AU from the stars.
The ISOPHOT data presented in this paper were reduced using PIA, which is a joint development by the ESA Astrophysics Division and the ISOPHOT Consortium. We thank the referee, Dr. R. Liseau, for helpful comments leading to improvement of the manuscript.
Following the treatment of Paper I, we define ,
between the dust emission and the photospheric emission in the infrared: