A probable close brown dwarf companion to GJ 1046 (M2.5V)
AAstronomy & Astrophysics manuscript no. gj1046˙final4 c (cid:13)
ESO 2018November 1, 2018
A probable close brown dwarf companion to GJ 1046 (M2.5V) (cid:63)
M. K¨urster, M. Endl and S. Re ff ert Max-Planck-Institut f¨ur Astronomie, K¨onigstuhl 17, D-69117 Heidelberge-mail: [email protected] McDonald Observatory, University of Texas, Austin, TX 78712e-mail: [email protected] Zentrum f¨ur Astronomie Heidelberg, Landessternwarte, K¨onigstuhl 12, D-69117 Heidelberge-mail: [email protected]
Received 18 January 2008 / Accepted 17 March 2008
ABSTRACT
Context.
Brown dwarf companions to stars at separations of a few AU or less are rare objects, and none have been found so far aroundearly-type M dwarfs (M0V-M5V). With GJ 1046 (M2.5V), a strong candidate for such a system with a separation of 0 .
42 AU ispresented.
Aims.
We aim at constraining the mass of the companion in order to decide whether it is a brown dwarf or a low-mass star.
Methods.
We employed precision RV measurements to determine the orbital parameters and the minimum companion mass. We thenderived an upper limit to the companion mass from the lack of disturbances of the RV measurements by a secondary spectrum. Aneven tighter upper limit is subsequently established by combining the RV-derived orbital parameters with the recent new version ofthe Hipparcos Intermediate Astrometric Data.
Results.
For the mass of the companion, we derive m ≥ . Jup from the RV data. Based on the RV data alone, the probability thatthe companion exceeds the stellar mass threshold is just 6 . ff ects from the secondary spectrum lets us constrainthe companion mass to m ≤
229 M
Jup . The combination of RV and Hipparcos data yields a 3 σ upper mass limit to the companionmass of 112 M Jup with a formal optimum value at m = . Jup . From the combination of RV and astrometric data, the chanceprobability that the companion is a star is 2 . Conclusions.
We have found a low-mass, close companion to an early-type M dwarf. While the most likely interpretation of thisobject is that it is a brown dwarf, a low-mass stellar companion is not fully excluded.
Key words.
Stars: low-mass, brown dwarfs – Binaries: spectroscopic – Stars: individual: GJ1046 – Astrometry
1. Introduction
The paucity of brown dwarf companions to solar-like stars atseparations of a few AU or less (a canonical value of ≤ M ≤
13 M
Jup ) andstellar ( M ≥ .
08 M (cid:12) ) companions with relatively little overlapbetween the two. At wide separations no “brown dwarf desert” isobserved. While the frequency of brown dwarf companions sep-arated from their host star by < . > K of the primary (the companion is usuallynot visible in the spectrum) is given by K = (2 π G / P ) / (1 − e ) / m sin i ( M + m ) / , (1) Send o ff print requests to : M. K¨urster (cid:63) Based on observations collected at the European SouthernObservatory, Paranal, Chile, programmes 173.C-0606 and 078.C-0829 where M and m are, respectively, the mass of the star and com-panion, and P , e and i are the orbital period, eccentricity andinclination.As can be seen from equation (1) the chances to detect acompanion object via RVs increase with shorter period (andshorter separation) and higher companion mass. For example,the RV semi-amplitude of the stellar reflex motion caused by a20 M Jup brown dwarf at 1 AU from a solar-mass star is 565 ms − ,if the orbit is circular and seen edge-on ( i = ◦ ). This is two or-ders of magnitude greater than the current state-of-the-art RVmeasurement precision of a few ms − . The detectability also in-creases for lower-mass stars; the same 20 M Jup at 1 AU from an0 . (cid:12) star would produce an RV semi-amplitude of 1009 ms − .And in fact there is observational evidence of the existence ofbrown dwarf companions to low-mass stars such as the proto-type brown dwarf companion GJ 229B that orbits an M1V starat a wide projected separation of 44 AU (Nakajima et al. 1995).Few brown dwarf companion candidates in the separationregime up to a few AU are known. The first such candidate wasHD 114762 (Latham et al. 1989), and a more recent exampleis HD 137510 (Endl et al. 2004). The masses of these candi-date objects have often not been well-determined since the RVmethod just yields a minimum mass and the astrometric preci-sion of the available Hipparcos data (ESA 1997) is mostly notsu ffi cient to confirm brown dwarfs and exclude stellar compan-ions (e.g. Pourbaix 2001; Pourbaix & Arenou 2001). Among the a r X i v : . [ a s t r o - ph ] A p r M. K¨urster et al.: A probable close brown dwarf companion to GJ 1046 (M2.5V)
Fig. 1.
RV time series of our UVES RV data for GJ 1046. The solid line corresponds to the Keplerian solution with a period of169 d and with χ = .
7, 8 degrees of freedom (DoF), and p ( χ ) = . .
56 ms − which compares well with the average measurement error of 3 .
63 ms − which is much smaller than the plotsymbols. For comparison the second best solution with a period of 338 d is shown as a dashed line. While being the second best,the latter solution is clearly excluded due to its extremely large χ value of 10 121.best established brown dwarfs are the companions to the G4IVstar HD 38529 and to the G6IV star HD 168443 with companionmasses of 37 M Jup and 34 M
Jup , orbital periods of 2174 . .
68 AU and 2 .
87 AU, respectively.These masses were determined by Re ff ert & Quirrenbach (2006)who derived new astrometric solutions from the Hipparcos mea-surements given the precisely known RV-derived orbital parame-ters, i.e. period, time of periastron passage, eccentricity, and lon-gitude of periastron. Another example determined in a similarfashion by Zucker & Mazeh (2000) is the G5IV star HD 10697with a companion in a 1078 d orbit at a separation of 2 .
12 AUand a mass of 40 M
Jup . An example for an object with a mini-mum mass of 9 . Jup that turned out to be a star with a mass of142 M
Jup is the companion to HD 33636 (Bean et al. (2007).The present paper presents a new candidate for a brown-dwarf companion which we have found in our precision RVsurvey carried out with the UVES spectrograph at the ESOVLT in search for planetary and substellar companions to Mdwarfs (see K¨urster et al. 2003). The host star, GJ 1046 (M2.5V; V = .
62 mag), has no entry in the Double and MultipleSystems Annex of the Hipparcos data base. Comparing its Vband and J,H,K band colours (from the 2MASS catalogue;Skrutskie et al. 2006) with the mass-luminosity relationshipsby Delfosse et al. (2000) there is no indication of near-infraredemission in excess of the scatter found in these relations.If the companion to GJ 1046 turns out to be a brown dwarf,then the system would be unique in that it would contain the firstclose-in ( ≤ ≥ M6V) are rela-tively frequent (e.g. Montagnier et al. 2006). But even in thesesystems separations ≤ Table 1. Di ff erential RV time series measurements of GJ 1046. Date a) BJD − DRV RV-error2 450 000 [ms − ] [ms − ]2004-10-03 3281.83304 -1132.0 3.22004-11-11 3320.62631 411.5 3.52005-07-27 3578.91097 -1670.6 4.82005-08-26 3608.77350 -1524.6 3.72005-09-11 3624.71807 -928.7 4.32005-09-19 3632.72200 -615.2 3.52005-10-17 3660.68277 492.3 3.52006-01-15 3750.60256 -1787.7 3.52006-09-15 3993.84856 326.4 3.32006-09-28 4006.77107 823.4 3.92006-10-05 4013.80030 1085.8 3.62006-10-06 4014.61077 1108.1 3.72006-11-02 4041.65276 1737.5 3.12006-11-08 4047.65399 1673.7 3.3Note: a) Barycentrically corrected Julian Date ries with low-mass stellar secondaries. Counterexamples are theM8 star LHS 2397a with an L7.5 companion at a separation of2 . α
8, orbited by a brown dwarf at aseparation of 1AU (Joergens & M¨uller 2007).
2. Observations
GJ 1046 was observed with the VLT-UT2 + UVES as one of thetargets of our precision RV survey of M dwarfs in search forextrasolar planets (see K¨urster et al. 2003, 2006). To attain high-precision RV measurements UVES was self-calibrated with its . K¨urster et al.: A probable close brown dwarf companion to GJ 1046 (M2.5V) 3
Table 2.
System parameters of GJ 1046
RV-derived parameters
Orbital period P . ± .
030 [d]Time of periastron T p BJD − . ± . K . ± . − ]Orbital eccentricity e . ± . ω . ± .
50 [ ◦ ]Mass function f ( m ) 9 . ± .
024 [10 − M (cid:12) ] χ (DoF =
8) 12.7 p ( χ ) 0.123Scatter rms 3.56 [ms − ]Mean error ∆ RV − ] Inferred parameters
Stellar mass M . ± .
007 [M (cid:12) ]Minimum companionmass m min . ± .
30 [M
Jup ]Semi-major axis ofcompanion orbit a . ± .
010 [AU]Critical inclination(for m = . (cid:12) ) i crit . ◦ ] a) Probability of i crit , p i crit .
3% (20 . ◦ > i ≥ ◦ ) Parameters derived from absence of companion spectrum
Maximum companionmass m spmax
229 [M
Jup ]Minimum inclination i spmin . ◦ ] a) Probability of i spmin , p i spmin .
2% (8 . ◦ > i ≥ ◦ ) Astrometry-derived parameters
Ascending node Ω . formal optimum [ ◦ ]Inclination i . formal optimum [ ◦ ]Companion mass m . formal optimum [M Jup ] χ (DoF = . i asmin . σ limit [ ◦ ] a) Probability of i asmin , p i asmin .
7% (15 . ◦ > i ≥ ◦ )Maximum companionmass m asmax
112 3 σ limit [M Jup ] b) Probability of a stellarcompanion p ∗ . A priori probability based on the RV derived minimummass and assuming random orientation of the orbit.b) Probability derived from the astrometric model. iodine gas absorption cell operated at a temperature of 70 ◦ C.Image slicer . (cid:48)(cid:48) slit were chosen yielding a resolv-ing power of R =
100 000 −
120 000. The central wavelength of600 nm was selected such that the useful spectral range contain-ing iodine (I ) absorption lines (500 −
600 nm) falls entirely onthe better quality CCD of the mosaic of two 4 K × ff erential ra-dial velocities ( DRV ) we refer the reader to Endl et al. (2000).A concise summary can also be found in Sect. 4 of K¨ursteret al. (2003).A total of 14 spectra of GJ 1046 observed through the iodinecell were obtained in 14 nights between 03 October 2004 and11 November 2006. See Table 1 for the journal of observations.
Fig. 2.
Histogram of the mean internal RV measurement errorsfor our sample of stars. Applying a κ - σ clipping procedure withan iterative rejection of values exceeding the mean plus 2 . σ (for a Gaussian distribution the chance probability of exceed-ing this value in one iteration is 0 . . − . Of the total sample of 41stars only 38 are displayed, because three stars with large errors(79, 81, and 92 ms − , respectively) lie far outside the displayederror range. They are newly discovered double-lined spectro-scopic binaries with such strong contributions from the secon-daries that the employed single-lined spectrum model fails. Sixstars have abnormally high mean errors between 7 and 16 ms − which, in one case, is the result of very low signal-to-noise ratio,but in the other cases could also be due to spectral contamina-tion from a companion. The remaining bulk of the distribution(mean error < . − ) has a mean of 3 .
91 ms − and a width of σ = .
02 ms − .Individual exposure time was 900 s yielding an average S / N perpixel between 39 and 58 for the various spectra (the median andmean being 51 . .
1, respectively). On average our errorof the individual RV measurements is 3 .
63 ms − for this V = .
62 mag object. All RV data were corrected to the solar systembarycenter using the JPL ephemeris DE200 (Standish 1990) forthe flux-weighted temporal midpoint of the exposure as given bythe UVES exposuremeter. For each epoch of observation propermotion corrected stellar coordinates were used. On 19 November2004 we also obtained a triplet of exposures without the iodinecell (exposure time 3 ×
705 s) required as a template spectrum inthe data modelling process (cf. Endl et al. 2000).
3. Results
Our RV time series for GJ 1046 is listed in Table 1 and alsoshown in Fig. 1 along with the best-fit Keplerian orbit yieldingan orbital period of P =
169 d, an eccentricity of e = . K = − , and a mass function f ( m ) = ( m sin i ) / ( M + m ) = . × − M (cid:12) (Table 2).Given the unknown inclination i only a minimum to thecompanion mass can be determined corresponding to the case i = ◦ . For this we need an estimate of the stellar mass ofGJ 1046 which we obtain from the K-band mass-luminosityrelationship by Delfosse et al. (2000). Taking the apparent K-magnitude from the 2MASS catalogue ( K = .
03 mag) and com-bining it with the Hipparcos parallax (71 .
11 mas) we find an ab-solute K-magnitude of 6 .
29 mag. The K-band mass-luminosityrelationship then yields a stellar mass of 0 . ± .
007 M (cid:12) . M. K¨urster et al.: A probable close brown dwarf companion to GJ 1046 (M2.5V)
Table 3.
Simulations for the determination of an upper limit to the companion mass.
Mass V-band Model Mean error Excess error Chance probability from[M (cid:12) ] [M
Jup ] flux ratio spectrum [ms − ] [ms − ] sample comparison excess errora) b) criterion0.00 0 — none 3.63 0.00 0.56 0.61 10.16 167 18.30 GJ 699 3.79 1.07 0.50 0.55 0.700.17 177 14.85 GJ 699 3.92 1.49 0.47 0.50 0.590.18 188 12.26 GJ 699 4.04 1.77 0.47 0.45 0.510.19 198 10.27 GJ 699 4.22 2.16 0.34 0.38 0.400.20 209 8.69 GJ 699 4.47 2.61 0.25 0.29 0.280.20 209 8.69 GJ 682 4.41 2.51 0.31 0.31 0.310.21 219 7.43 GJ 682 4.79 3.12 0.17 0.20 0.140.22 230 6.41 GJ 682 5.16 3.67 0.13 0.11 00.23 240 5.57 GJ 682 5.61 4.28 0.094 0.048 00.24 250 4.87 GJ 682 6.10 4.90 0.031 0.016 00.25 261 4.29 GJ 682 6.70 5.63 0 0.0032 00.26 271 3.79 GJ 682 7.40 6.45 0 0.00032 0Notes: a) Fraction of stars in the distribution with larger mean errorsb) Areal fraction of Gaussian fitted to the distribution exceeding the mean error We then infer a minimum companion mass of m min = . Jup and, from equation (1), a semi-major axis of the com-panion orbit of a = .
42 AU. In order for the true companionmass to exceed the stellar threshold of 0 .
08 M (cid:12) the orbital in-clination i would have to be < . ◦ . For a chance orientationof the orbit the probability that i is smaller than some angle θ isgiven by p ( θ > i ≥ ◦ ) = − cos θ ; (2)hence the chance probability to have an inclination < . ◦ isjust 6.3% making it not very likely that the companion is a star(see also Table 2).
4. Spectroscopic companion mass upper limit
An upper limit to the mass of the companion can be determinedfrom the spectroscopic data by exploiting the notion that withincreasing mass the companion would at some point become sobright that it would noticeably a ff ect the RV measurements. Inour data modelling approach a suitable indicator for the pres-ence of an additional perturbing signal is the magnitude of theRV measurement error, because all RV data are obtained assum-ing that the modelled spectrum is that of a single-lined binaryin which the light from the companion can be neglected. If thisis not the case, then the light contribution from the companionmanifests itself as unusually large errors of the determined RVvalues. For details of the data modelling approach and the esti-mation of the RV errors we refer the reader to Endl et al. (2000).Briefly, the spectra are subdivided into ≈
500 chunks of ≈ ff ected in a non-homogeneous fash-ion, depending on the detailed spectral line patterns within thechunk, with the e ff ect that the derived chunk RV values exhibit astronger scatter and hence combine to an increased value of theinternal RV measurement error. By way of simulations addingfaint companion spectra to the spectra of GJ 1046 we investigatethe following two possibilities of obtaining information on thecompanion mass upper limit from the internal RV error. 1. A comparison of the mean internal RV measurement errorof GJ 1046 for di ff erent added companion spectra with thedistribution of mean observed errors for the sample of mon-itored stars.2. A comparison of the mean internal RV measurement errorof GJ 1046 for di ff erent added companion spectra with itsoriginal value.In the case of considerable contamination the average RVerror for the spectra of the star in question stands out from thedistribution of RV errors for the sample of monitored stars whichis shown in Fig. 2. If only the bulk of the distribution is consid-ered, i.e. stars with a mean RV error < − , this distributionis nearly Gaussian with a mean value of 3 .
91 ms − and a widthof σ = .
02 ms − .The average RV measurement error for GJ 1046 is 3 .
63 ms − with the errors of the 14 individual GJ 1046 spectra ranging from3 .
07 to 4 .
84 ms − . These values are typical of the range of signal-to-noise values of our spectra (between 39 and 58 for GJ 1046)and well inside the distribution of errors in our sample of stars(Fig. 3). We also note that the RV errors increase with decreasingsignal-to-noise ratio of the GJ 1046 spectra which would not bethe case, if the errors were dominated by a perturbing signal.In the performed simulations we added spectra of faint low-mass stars to the original GJ 1046 spectra. As summarised inTable 3 these simulations explored the companion mass regime0 . − .
26 M (cid:12) (first two table columns) which, according tothe V-band mass-luminosity relation by Delfosse et al. (2000),corresponds to companions that are factors of 18 − / N per pixel: 300) and GJ 682 (M3.5V; mean S / N perpixel: 230) for masses ≤ .
20 M (cid:12) and ≥ .
20 M (cid:12) , respectively(fourth column). For each simulated spectrum the companionspectrum was shifted to the appropriate companion RV for theprobed mass value as well as scaled in flux corresponding to thebrightness predicted by the V-band mass-luminosity relation.We find that a companion with a mass ≥ .
254 M (cid:12) or265 M
Jup and a factor of 4 . ≥ − which would make this star stick out from the bulkof the sample (Fig. 2) indicating a contamination of the spec- . K¨urster et al.: A probable close brown dwarf companion to GJ 1046 (M2.5V) 5 trum. The 7th and 8th columns of Table 3 list the chance prob-abilities of the obtained mean errors (fifth column) from a com-parison with the total sample of observed stars.For the second possibility of determining the mass upperlimit of the companion we assume (conservatively) that the meanoriginal RV error is entirely caused by contributions from thecompanion spectrum and not attributable to photon noise or toe ff ects of instrumental nature or intrinsic to the star. We thensearch for the companion spectrum (as a function of companionmass and brightness) whose addition to the observed spectra in-troduces an additional RV error (6th column in Table 3) of thesame magnitude, i.e. it doubles the square of the errors. With anoriginal value of 3 .
63 ms − we search in the simulated data forthe companion mass and brightness that leads to a mean intrinsicerror a factor of √ .
13 ms − . (We note in passingthat this value is in the 88 .
4% percentile of the distribution of thestellar sample truncated at 7 ms − ; see Fig. 2.) The 9th columnof Table 3 lists the chance probability of obtaining the excesserror value listed in the 6th column.For this increased error value we find a companion mass of ≥ .
219 M (cid:12) or 229 M
Jup and a primary-to-secondary V-band fluxratio of 6 . Jup as the spectroscopic upper limit to the mass ofthe companion to GJ 1046. This value corresponds to an orbitalinclination of 8 . ◦ . The probability for an inclination as smallas (or smaller than) this value is 1 .
5. Companion mass upper limit from a combinationof the RV data with Hipparcos measurements
Even if the astrometric signature of the companion is not seenin the Hipparcos data, Hipparcos astrometry can yield stringentupper mass limits on companions detected via the radial velocitymethod.Using the Hipparcos parallax (71 .
11 mas) together with theorbital parameters derived from the RV measurements we canpredict the minimum astrometric signal of the stellar reflex mo-tion to be 3 . ff ect could be consider-ably higher. For the limiting inclination of 20 . ◦ the full minoraxis of the stellar orbit would extend 10 . ff ert & Quirrenbach (2006) by keeping those ofthe orbital parameters that are known from the analysis of theRVs fixed and varying only the inclination and the ascendingnode while fitting an astrometric orbit to the abscissa residu-als. Additional free parameters in the fit were a correction tothe mean position, mean proper motion and parallax of the star.The result is shown in Fig. 3 (left panel).The formally best fit to the Hipparcos data is achieved withan inclination i = . ◦ ( i − ◦ = . ◦ ) corresponding to a truecompanion mass of 47 . Jup pointing at a brown dwarf com- panion (Table 2). However, an F-test measuring the varianceimprovement yields a probability of 17% for the detection of theastrometric orbit implying that is has not been detected with sig-nificance. This can also be seen in Fig. 3 (left panel), where theascending node is completely undetermined since the 2 and 3 σ confidence contour levels span the entire parameter range.In the right panel of Fig. 3, the χ value is shown as a func-tion of inclination only, together with the 1 σ and 3 σ confidenceregions for the inclination. The 3 σ (99 .
73% confidence) lowerlimit to the inclination is i = . ◦ implying a 3 σ upper masslimit for the companion of 112 M Jup .Therefore, a stellar companion cannot be fully excluded,even though it is unlikely. From the astrometric solution thechance probability for the companion to have a stellar mass, orequivalently, for its inclination to be either i < . ◦ or > . ◦ is 2 .
2% and 0 . .
9% (see also Table 2).
6. Conclusions
We have presented the discovery of a probable brown dwarfcompanion to an M dwarf with an orbital period of just under1 / .
42 AU. Our RVmeasurements provide a lower limit to the true companion massof 26 . Jup and a chance probability of just 6 .
2% that the com-panion is actually a star. From the absence of any indications ofa secondary spectrum in our data we can place an upper limit tothe companion mass of m =
229 M
Jup .Combining our RV measurements with the HipparcosIntermediate Astrometric Data from the recent new reductionby van Leeuwen (2007a,b) we find a formal best-fit compan-ion mass value of 47 . Jup , but pertinent to a model that is notsignificant. However, the same data allows us to place a muchtighter companion mass upper limit of 112 M
Jup at 99 .
73% con-fidence. This mass upper limit still allows a stellar companion,but with a low probability. From the astrometric analysis thechance probability that the companion mass exceeds the stellarmass threshold is 2 . Acknowledgements.
We thank the ESO OPC for generous allocation of observ-ing time and the Science Operations Team of Paranal Observatory for carry-ing out the service mode observations for this programme. ME acknowledgessupport by the National Aeronautics and Space Administration under GrantsNNG05G107G issued through the Terrestrial Planet Finder Foundation Scienceprogram and Grant NNX07AL70G issued through the Origins of Solar SystemsProgram.
References
Bean, J.L., McArthur, B.E., Benedict, G.F., et al., 2007, AJ 134, 749Campbell, B., Walker, G.A.H, Yang, S., 1988, ApJ 331, 902Delfosse, X., Forveille, T., S´egransan, et al., 2000, A&A 364, 217Endl, M., K¨urster, M., Els, S., 2000, A&A 362, 585Endl, M., Hatzes, A.P., Cochran, W.D., et al., 2004, ApJ, 611, 1121ESA, 1997, ESA SP-1200Freed, M., Close, L.M., Siegler, N., 2003, ApJ 584, 453Gizis, J.E., Kirkpatrick, J.D., Burgasser, A., et al., 2001, ApJ 551, L163 Varying the RV derived parameters within their errors leads tominute changes in the formal best-fit solution indicating that the uncer-tainties of the latter are absolutely dominated by the astrometric data. The combined chance probability is given by one minus the productof the confidences: 1 − (1 − . − . = . Fig. 3. Left: χ contours for fitting a substellar companion with fixed spectroscopic parameters to the Hipparcos IntermediateAstrometric Data of GJ 1046. The inclination i and the ascending node Ω were free parameters of the fit, as were corrections tothe standard five astrometric parameters in the Hipparcos Catalogue. The contours represent two-parameter joint confidence levelswith probabilities of 68 .
3% (1 σ ), 95 .
4% (2 σ ), and 99 .
7% (3 σ ). The best fit solution is indicated by a cross. Right:
The χ of theastrometric orbit as a function of inclination only. In this case the 1 σ and 3 σ confidence levels indicated by the horizontal dashedlines correspond only to the single parameter i treating Ω as an uninteresting parameter. Again the best fit solution is indicated by across. Grether, D., Lineweaver, C.H., 2006, ApJ 640, 1051Joergens, V., M¨uller, A., 2007, ApJ 666, L113K¨urster, M., Endl, M., Rouesnel, F., et al., 2003, A&A, 403, 1077K¨urster, M., Endl, M., Rodler, F., 2006, The Messenger, 123, p. 21Latham, D.W., Mazeh, T., Stefanik, R.P., Mayor, M., Burki, G., 1989, Nat.,339,38Montagnier, G., S´egransan, D., Beuzit, J.-L., et al., 2006, A&A, 460, L19Nakajima T., Oppenheimer B.R., Kulkarni S.R., et al., 1996, Nat., 378, 463Neuh¨auser R., & Guenther E.W. 2004, A&A 420, 647Pourbaix D., 2001, A&A 369. L22Pourbaix D. & Arenou F., 2001, A&A 372, 935Re ffff