Herschel discovery of a new class of cold, faint debris discs
C. Eiroa, J. P. Marshall, A. Mora, A. V. Krivov, B. Montesinos, O. Absil, D. Ardila, M. Arevalo, J.-Ch.Augereau, A. Bayo, W. Danchi, C. del Burgo, S. Ertel, M. Fridlund, B.M. Gonzalez-Garcıa, A. M. Heras, J. Lebreton, R. Liseau, J. Maldonado, G. Meeus, D. Montes, G.L. Pilbratt, A. Roberge, J. Sanz-Forcada, K. Stapelfeldt, P. Thebault, G. J. White, S. Wolf
AAstronomy & Astrophysics manuscript no. astroph˙DUNES˙colddiscs c (cid:13)
ESO 2018October 13, 2018 L etter to the E ditor Herschel discovery of a new class of cold, faint debris discs (cid:63)
C. Eiroa , J. P. Marshall , A. Mora , A. V. Krivov , B. Montesinos , O. Absil , D. Ardila , M. Ar´evalo , J.-Ch.Augereau , A. Bayo , W. Danchi , C. del Burgo , S. Ertel , M. Fridlund , B.M. Gonz´alez-Garc´ıa , A. M. Heras ,J. Lebreton , R. Liseau , J. Maldonado , G. Meeus , D. Montes , G.L. Pilbratt , A. Roberge , J. Sanz-Forcada , K.Stapelfeldt , P. Th´ebault , G. J. White , , and S. Wolf (A ffi liations can be found after the references) Preprint online version: October 13, 2018
ABSTRACT
We present
Herschel
PACS 100 and 160 µ m observations of the solar-type stars α Men, HD 88230 and HD 210277, which form part of the FGKstars sample of the Herschel Open Time Key Programme (OTKP)
DUNES ( DU st around NE arby S tars). Our observations show small infraredexcesses at 160 µ m for all three stars. HD 210277 also shows a small excess at 100 µ m, while the 100 µ m fluxes of α Men and HD 88230 agreewith the stellar photospheric predictions. We attribute these infrared excesses to a new class of cold, faint debris discs. α Men and HD 88230are spatially resolved in the PACS 160 µ m images, while HD 210277 is point-like at that wavelength. The projected linear sizes of the extendedemission lie in the range from ∼
115 to ≤
250 AU. The estimated black body temperatures from the 100 and 160 µ m fluxes are (cid:46)
22 K, whilethe fractional luminosity of the cold dust is L dust / L (cid:63) ∼ − , close to the luminosity of the Solar-System’s Kuiper belt. These debris discs are thecoldest and faintest discs discovered so far around mature stars and cannot easily be explained by invoking ”classical” debris disc models. Key words. - Stars: planetary systems: planetary discs -Stars: planetary systems: formation - Stars: individual: α Men (HIP 29271) - Stars:individual: HD 88230 (HIP 49908) - Stars: individual: HD 210277 (HIP 109378)
1. Introduction
Debris discs are tenuous structures associated with main se-quence stars formed by second generation dust, which has re-sulted from the collisions of solid bodies continuously supplyingthe circumstellar environment with small dust particles. This for-mation sequence is inferred from the lifetime of the dust grainsagainst destructive collisions, Poynting-Robertson drag and ra-diation pressure, which is much shorter than the ages of the hoststars. These discs are visible in reflected light at optical wave-lengths and in thermal radiation at mid- / far-IR and submillimeterwavelengths (Aumann et al., 1984; Backman & Paresce, 1993).General disc characteristics are grain black body temperatures of ∼ −
100 K, fractional luminosities f > L dust / L (cid:63) ∼ − , andradii from less than 10 AU to several times 100 (e.g. Absil et al.2006, Su et al. 2005, Trilling et al. 2008). Debris discs are con-sidered analogues of the Solar System asteroid and Kuiper belts,although their luminosities are usually more than 100 times theKuiper belt level of L dust / L S un ∼ − − − (Stern 1996,Vitense et al. 2010). The sensitivity of the 3.5 m Herschel far-infrared space telescope (Pilbratt et al. 2010) with its instrumentPACS (Poglitsch et al. 2010) o ff ers the possibility of characteris-ing colder ( ∼
30 K) and fainter ( L dust / L (cid:63) few times 10 − ) debrisdiscs with spatial resolution ∼
60 AU (FWHM) at 10 pc, i.e.,true extra-solar Kuiper belts.DUNES is a
Herschel
OTKP designed to detect and char-acterise extra-solar analogues to the Kuiper belt around mainsequence FGK nearby stars (Eiroa et al. 2010). In this letterwe present the results for three stars from the DUNES sam-ple: α Men, HD 88230, and HD 210277 as clear examples of (cid:63)
Herschel is an ESA space observatory with science instruments pro-vided by European-led Principal Investigator consortia and with impor-tant participation from NASA.
Table 1.
Stellar properties.
Star α Men HD 88230 HD 210277Sp. Type G5V-G7V K6V-M0V G0V, G7V-G9VDistance (pc) 10.2 4.9 21.6 L ∗ ( L (cid:12) ) 0.85 0.15 1.10T ef f (K) 5590 3850 5540Age (Gyrs) 5.5 6.6 6.9 the advantages o ff ered by Herschel observations: they trace anew class of cold, T (cid:46)
22 K, spatially resolved debris discs withvery low fractional luminosities. These discs have remained un-observed by previous far-IR and submillimeter studies. Table 1gives some properties of the stars. Ages are based on the logR (cid:48) HK activity index and have an uncertainty of 60% (Mamajek& Hillenbrand 2008). HD 210277 hosts a Jupiter-like planet(Marcy et al. 1999). This star and α Men have faint stellar com-panions, but neither the measurements nor the photospheric pre-dictions are a ff ected by them. Eiroa et al. in prep will present afull discussion of the stars and the general results of the DUNESsurvey.
2. Observations and data reduction α Men, HD 88230 and HD 210277 were observed with PACS100 / µ m in scan map mode. For each star, two scans at po-sition angles 70 ◦ and 110 ◦ were carried out, each scan consist-ing of 10 legs with separation of 4 (cid:48)(cid:48) , length of 3 (cid:48) and mediumspeed of 20 (cid:48)(cid:48) / s. Table 2 gives the scan identification numbers(Obs. ID) and the total duration of the observations (OT). α Men was observed twice in order to increase the S / N ratio. Data a r X i v : . [ a s t r o - ph . S R ] O c t C. Eiroa et al.:
Herschel discovery of a new class of cold, faint debris discs
Table 2.
Log of the PACS 100 µ m and 160 µ m observations. Star HIP Obs. ID OT (sec) α Men 29271 1342203297 / α Men 1342216043 / / / reduction was made using the Herschel Interactive ProcessingEnvironment (HIPE) version 7.2. The individual scans were mo-saiced to produce the final image at each band. To check the con-sistency of the reduction and analysis (particularly regarding thee ff ect of correlated noise), mosaics were produced at both the na-tive 3 . (cid:48)(cid:48) µ m (green) and 6 . (cid:48)(cid:48) µ m (red), as well assuper-sampled pixel scales 1.0 (cid:48)(cid:48) (green) and 2.0 (cid:48)(cid:48) (red), the latterbeing the default pixel size in HIPE for this type of data. Theimages with the native pixel size scales avoid, at least partly, thecorrelated noise in the PACS images (see also Fruchter & Hook,2002). A high-pass filter was used to remove large scale back-ground emission from the images, with filter widths of 15 (cid:48)(cid:48) and25 (cid:48)(cid:48) in the green and red channels. To prevent the removal of anyextended structure near the stars, regions where the sky bright-ness exceeded a threshold value in the image defined by the stan-dard deviation of all the positive pixels ( S > − Jy / pixel )were masked from that process. Absolute flux calibration un-certainties are ∼
3% and ∼
5% for the green and red bands (seetechnical note below).
3. Results
Table 3 gives the J2000.0 optical equatorial coordinates of thestars as well as their 100 µ m peak positions corrected for theproper motions of the stars (van Leeuwen, 2007). O ff sets be-tween the optical and the PACS positions (column 4 of Table 3)are within ∼ σ the Herschel pointing accuracy of 2 . (cid:48)(cid:48) µ m images and isocontour plotsof the stars. α Men and HD 88230 are point-like at 100 µ m,FWHM ∼ . (cid:48)(cid:48) × . (cid:48)(cid:48)
5, while both stars are resolved at 160 µ m,with angular sizes / position angles ∼ (cid:48)(cid:48) / ◦ and ∼ . (cid:48)(cid:48) / ◦ , re-spectively. The extended emission appears clearly asymmetric inHD 88230, with the star located at the North-Eastern side. Noneof the objects are resolved in the direction orthogonal to the ex-tended emission. HD 210277 is unresolved at both wavelengths.PACS fluxes (Table 3) have been estimated using circu-lar and rectangular aperture photometry, taking special care tochoose the reference background region due to the presence offield objects. Specifically, measurements of HD 210277 take intoaccount the presence of the bright object located at ∼ (cid:48)(cid:48) North-East from the star; in the case of this star we have, in addition,carried out PSF photometry using the DAOPHOT software pack-age. The PSF photometry fluxes are F (100) = ± F (160) = ± ff erent positions in the nearbyfields. Sky sizes for the errors estimates are equal to the area ofthe aperture used for the photometry for all three stars and bothbands; in particular the sizes of the extended emission around α Men and HD 88230.PACS fluxes have been compared to predicted stellar photo-spheric fluxes (Table 3) using Gaia / PHOENIX models (Brott & Technical Note PICC-ME-TN-037 in http: // herschel.esac.esa.int Hauschildt, 2005), with the stellar parameters as given in Eiroaet al. (in prep). Fig. 1 shows the spectral energy distributions(SEDs) of the stars, where PACS fluxes are plotted together withoptical, near-IR, IRAS, AKARI, and
Spitzer / MIPS and IRS data;in addition, the best χ -square photospheric fit is shown. To assessthe presence of an excess at 100 and / or at 160 µ m we require thatthe observed fluxes, F PACS , exceed by at least 3 σ the predictedphotospheric fluxes, F ∗ ( χ λ = ( F PACS − F ∗ ) /σ λ ). No excesses aredetected at 100 µ m for α Men and HD 88230, while it is seen inHD 210277. All three stars do show 160 µ m excesses. We alsonote that the SED slopes from 100 to 160 µ m are α = - 0.5 ± α Men), α = - 0.7 ± α = ± ff er from the expected Rayleigh-Jeans behaviour( α = - 2.0) of a stellar photosphere in this wavelength regime. All three fields show ∼ ∼ (cid:48)(cid:48) towards the SW from HD 88230 and a bright one at ∼ (cid:48)(cid:48) towards the NE from HD 210277. We have consulted theNASA / IPAC extragalactic database to search for counterpartswithout finding any association. Although we cannot firmly ex-clude a coincidental alignment or contamination of a backgroundsource(s) in the line of sight of our stars, we think that it is un-likely for the
Herschel sources presented here due to the closecorrelation between the optical and
Herschel positions, and thephotospheric predictions and the estimated
Herschel µ mfluxes. In fact, following in a first approach the source counts byBerta et al. (2010), the average density of extragalactic sourceswith fluxes ∼ ∼
12 mJy - i.e., the measured excessesat 160 µ m from α Men / HD 88230 and HD 210277, respectively(Table 3) - are 2 / arcmin and 0.7 / arcmin . Thus, given the opti-cal / µ m o ff sets (Fig. 1) the a priori probability of an acciden-tal alignment is very small, clearly smaller than 5%. We refer toa future paper (del Burgo et al., in prep. ) for a detailed study onsource contamination within the DUNES fields.
4. Analysis
The small PACS excesses above the photospheric fluxes are in-terpreted as due to cold debris discs around the stars. Black bodydust temperatures, T dust , can be estimated from the 100 and 160 µ m excess fluxes for HD 210277; in the case of α Men and HD88230, an upper limit for T dust can be calculated taking the 100 µ m flux as 3 σ statistical noise. T dust for the three stars is (cid:46)
22K (Table 4); the corresponding inner radii of discs with blackbody grains at this T dust , considering the luminosity of the stars,are also given in Table 4. Fig. 1 shows the excellent agreementbetween the combined SED of the stellar photospheric fits plus22 K black bodies and the observed SEDs.Deconvolution of the images can be used to estimate the truesize of the resolved discs in α Men and HD 88230. Our methodfirst removes the photospheric contributions from each imageby subtracting a PSF with a peak scaled to the predicted photo-spheric flux level. The PSF model uses α Bo¨otis images, rotatedto match the roll angle of the telescope during the observations ofthe DUNES stars. After star subtraction, the images are decon-volved using both modified Wiener and Richardson-Lucy algo-rithms. The noise model takes into account that the main contrib-utor is the telescope thermal emission. Both algorithms produceconsistent results although with di ff erent noise patterns. Fig. 2shows the 160 µ m star-subtracted and the Wiener-deconvolved . Eiroa et al.: Herschel discovery of a new class of cold, faint debris discs 3
Table 3.
Optical positions of the stars and of their PACS 100 identified counterparts. Observed PACS fluxes with 1 σ statisticalerrors ( F PACS ), and predicted photospheric fluxes ( F (cid:63) ). Flux units are mJy. Star Optical position PACS 100 µ m position O ff set PACS 100 µ m PACS 160 µ m(J2000.0) (J2000.0) (cid:48)(cid:48) F PACS F (cid:63) χ F PACS F (cid:63) χ α Men 06 10 14.47 -74 45 11.0 06 10 14.53 -74 45 11.1 0.3 17.8 ± ± +
49 27 15.3 10 11 21.88 +
49 27 17.2 3.2 22.5 ± ± ± ± Fig. 1.
PACS images, isocontours and SEDs of α Men / HIP 29271 (left), HD 88230 / HIP 49908 (middle) and HD 210277 / HIP 109378(right). Images: 100 µ m (left), 160 µ m (right). North is up and East to the left. Isocontours: 100 µ m contours are in green colourwhile the 160 µ m ones are in red. α Men: 100 µ m contours (10, 20, 40, 80 and 90 % of the flux peak); 160 µ m contours (20,40,60,80,90 % of the flux peak) HD 882308: 100 µ m contours ( 10, 20,30,40,60, 80, 90 % of the flux peak); 160 µ m contours (20, 30, 40,50, 60, 80, 90 % of the flux peak) HD 210277: 100 µ m contours ( (30, 40, 50, 60, 70, 80, 90 % of the flux peak); 160 µ m contours(50, 60, 70, 80, 90 % of the flux peak). The lowest contour in all cases is ≈ σ . The optical position of the stars are indicated by thesymbol “*” in the isocontour plots; a segment indicates the projected linear sizes at the distance of each star. SED plots: black lineis the photospheric fit while the blue line is the photosphere plus a 22 K black body.images of α Men and HD 88230; estimated sizes of the decon-volved sources are 16 (cid:48)(cid:48) and 21 (cid:48)(cid:48) , respectively. Table 4 gives thelinear sizes of the semi-major axes ( ∼ σ contours) from both theoriginal and deconvolved PACS 160 µ m images. In the case ofHD 210277, the value in Table 4 corresponds to an upper limitof 12 (cid:48)(cid:48) , i.e., approximately the 160 µ m beam size. The compar-ison of both the directly observed and deconvolved semi-majoraxes to the estimated radii from T dust clearly indicates that theobserved discs (assuming as a first approach symmetric discs)are smaller than the expected sizes for black body discs.Dust fractional luminosities, f (Table 4), can be obtained bytaking the 160 µ m fluxes and assuming the star temperaturesgiven in Table 1 and T dust =
22 K (Beichman et al., 2006). The estimated f values are of the order of 10 − , close to the Kuiperbelt level. Essentially similar f values are obtained taking themaximum wavelength corresponding to 22 K, i.e., 230 µ m, andits expected flux extrapolating from the one measured at 160 µ m.
5. Discussion
The debris discs in this work are a new class of discs charac-terised by an excess at 160 µ m, little to no excess at 100 µ m,and no excess emission at shorter wavelengths. These discs arethe coldest and least luminous ones known to date; they are sig-nificantly colder and fainter than other observed DUNES discs C. Eiroa et al.:
Herschel discovery of a new class of cold, faint debris discs
Fig. 2. µ m star-subtracted (left) and deconvolved (right) im-ages. Up: α Men / HIP 29271. Contours: 10%, 20%,40%, 80% ofthe peak. Bottom: HD 88230 / HIP 49908. Contours: 20%,40%,80% of the peak. North is up and East to the left. Field size is60 (cid:48)(cid:48) × (cid:48)(cid:48) . Table 4.
Black body dust temperatures from the far-IR excesses, T dust and the corresponding estimated radii. Observed 3 σ linearsizes (semi-major axes) of the extended emission at 160 µ m inthe original and deconvolved images. f is the dust fractional lu-minosity. Star T dust Size Size Size f Est. Orig. Deconv.(K) (AU) (AU) (AU) α Men ≤ ≥
147 92 81 9.7 × − HD 88230 ≤ ≥
62 56 51 1.6 × − HD 210277 22 160 ≤
130 5.4 × − (e.g. Liseau et al. 2010, Marshall et al. 2011). Their interpreta-tion poses significant challenges.The shapes of the SEDs suggest that the dust is located ina ring with a larger inner void. We have seen that the observeddisc radii are smaller than those implied from T dust (Table 4). Inother words, T dust is smaller than the black body temperature, T bb ≈ T dust , the question is how to make thedust cold, while allowing a size distribution, but still keeping thedisc radii within the observed ones.The low temperatures of grains require them to be large andhighly reflective. The required albedo can be derived from T dust and T bb . Taking T dust =
22 K as representative for the three ob-served discs, and T bb = (cid:38) ∼
10 K below the black body value (Gurwell et al.2010). It is natural to expect that dust released from the surfacesof such objects would have similar properties. In addition, one has to explain why small grains in the colddiscs are depleted. One possibility is to assume a very low dy-namical excitation of dust-producing planetesimals, so that discswould be devoid of small particles (Th´ebault & Wu, 2008). Thereason is that low collision velocities between large grains, un-a ff ected by radiation pressure, create an imbalance between therates at which small grains are produced (low) and destroyed(high). Low collision velocities are compatible with low orbitalvelocities at the large radii of the cold discs. Besides, since thesurface density of the solids far from the central stars is alsolow, planetesimal accretion scenarios predict very long accre-tion timescales (Kenyon & Bromley, 2008). Thus, large plan-etesimals that would excite the discs may have failed to grow. Infact, low dynamical excitation has been inferred for other largedebris discs, e.g. HD 207129 (L¨ohne et al 2011).The tenet that the cold discs are probably in a low dynam-ical excitation state would be di ffi cult to reconcile with the ex-istence of large planets in the discs, since they would stir thediscs too strongly. The emission around α Men and HD 88230 isasymmetric, which might be suggesting the presence of a giantplanet as in Fomalhaut (Kalas et al.2008). The problem can bemitigated if the planets are in nearly-circular orbits or the plan-etesimals have low eccentricities, as suggested for Fomalhaut(Chiang et al. 2009). HD 210277 hosts a planet at 1.1 AU with e p = .
47; however, based on the formulae by Mustill & Wyatt(2009), a stirring front from such a planet should not be able toreach the ∼
130 AU-sized disc on Gyr timescales.Another point is the origin of the large inner voids in the colddiscs. These could either be due to clearing by planets or mayreflect the accretional and collisional history of primordial discs.In the latter case, the observed radii are those at which solidscould reach “right” sizes, “right” degree of dynamical excitation,and / or were able to survive over the stellar age, to produce theobserved emission. A detailed analysis of the nature of the colddiscs is beyond the scope of this paper, and we defer to Krivov etal. ( in prep. ) where possible scenarios will be discussed in detail.
6. Conclusions
We have presented
Herschel
PACS observations of three stars ofthe OTKP DUNES sample. The observations reveal a new classof debris discs with fractional luminosities close to the Solar-System Kuiper’s belt, but are colder and larger. These discs area challenge to current models explaining debris discs aroundmature solar-type stars such as either the usual collisional-dominated disc scenario or low dynamical excitation discs.
References
Absil, O., di Folco, E., M´erand, A. et al. 2006, A&A 452, 237 2010 A&A 518,L17Aumann H.H., Beichman, C.A., Gillet, F. C., et al. 1984 ApJ 278, L23Backman, D.E., & Paresce, F. 1993, in Protostars and Planets III, ed. E.H. Levy& J.I. Lunine (Tucson: Univ. Arizona Press), 1253Barucci, M. A., Alvarez-Candal, A., Merlin, F. et al 2011, Icarus 214, 297Beichman, C.A., Bryden, G., Stapelfeldt, K.R., et al. 2006, ApJ 652, 1674Berta, S., Magnelli, B. Lutz, D., et al. 2010 A&A 518, L30Brott, I. & Hauschildt, P.H. 2005, ESA SP-576, C. Turon, K.S. O’Flaherty,M.A.C. Perryman (eds), p.565Chiang, E., Kite, E., Kalas, P., Graham, J.R., Clampin, M. 2009, ApJ, 693, 734Eiroa, C., Fedele, D., Maldonado, J., et al. 2010, A&A 518, L131Fruchter, A. S.& Hook, R.R. 2002, PASP 114, 144Gurwell, M. A., Butler, B. J., Moullet, A. 2010, BAAS 42, 1014Kalas, P., et al. 2008, Science 322, 1345Kenyon, S. J. & Bromley, B.C. 2008, ApJS, 179, 451Liseau, R., Eiroa, C., Fedele, d., et al. 2010, A&A 518, L132L¨ohne, T. et al. 2011, A&A, submitted . Eiroa et al.:
Herschel discovery of a new class of cold, faint debris discs 5
Mamajek,E. E., Hillenbrandt, L.A. 2008 ApJ 687, 1264Marcy, G.W., Butler, R.P. Vogt S.S., Fischer, D., Liu, M.C. 1999 ApJ 520, 239Mustill, A.J. & Wyatt, M.C. 2009, MNRAS, 399, 1403Pilbratt, G., Riedinger, J. R., Passvogel, T., et al. 2010, A&A518, L1Poglitsch, A., Waelkens, C., Geis, N., et al. 2010, A&A518, L2Stern, S.A. 1996, A&A 310, 999Su, K.Y.L., Rieke, G.H., Misselt, K.A., et al., 2005, ApJ 628, 487 653, 675Th´ebault, P., Wu, Y 2008, A&A 481, 713Trilling, D.F., Bryden, G., Beichman, C.A. et al. 2008, ApJ 674, 1086van Leeuwen, F. 2007, A&A 474, 653Vitense, Ch., Krivov, A. V., L¨ohne, T. 2010, A&A 520, 32 Dpt. F´ısica Te´orica, Facultad de Ciencias, UniversidadAut´onoma de Madrid, Cantoblanco, 28049 Madrid, Spaine-mail: [email protected] ESA-ESAC Gaia SOC. P.O. Box 78, E28691 Villanueva de laCa˜nada, Madrid, Spain Astrophysikalisches Institut und Universit¨atssternwarte, Friedrich-Schiller-Universit¨at, Schillerg¨aßchen 2-3, 07745 Jena, Germany Dpt. de Astrof´ısica, Centro de Astrobiolog´ıa (INTA-CSIC), ESACCampus, P.O.Box 78, E-28691 Villanueva de la Ca˜nada, Madrid,Spain Institut d’Astrophysique et de G´eophysique, Universit´e de Liege, 17All´ee du Six Aoˆut, B-4000 Sart Tilman, Belgium NASA Herschel Science Center, California Institute of Technology,1200 E. California Blvd., Pasadena, CA 91125, USA Universit´e Joseph Fourier / CNRS, Laboratoire d’Astrophysique deGrenoble, UMR 5571, Grenoble, France European Space Observatory, Alonso de Cordova 3107, Vitacura,Casilla 19001, Santiago 19, Chile NASA Goddard Space Flight Center, Exoplanets and StellarAstrophysics, Code 667, Greenbelt, MD 20771.USA UNINOVA-CA3, Campus da Caparica, Quinta da Torre, Monte deCaparica, 2825-149 Caparica, Portugal Christian-Albrechts-Universit¨at zu Kiel, Institut f¨ur TheoretischePhysik und Astrophysik, Leibnizstr. 15, 24098 Kiel, Germany ESA Astrophysics & Fundamental Physics Missions Division,ESTEC / SRE-SA, Keplerlaan 1, NL-2201 AZ Noordwijk, TheNetherlands INSA at ESAC, E-28691 Villanueva de la Ca˜nada, Madrid, Spain Onsala Space Observatory, Chalmers University of Technology, Se-439 92 Onsala, Sweden Universidad Complutense de Madrid, Facultad de Ciencias F´ısicas,Dpt. Astrof´ısica, Av. Complutense s / n, 28040 Madrid, Spain Jet Propulsion Laboratory, California Institute of Technology, MS183-900, 4800 Oak Grove Drive, Pasadena, CA 91109, USA LESIA, Observatoire de Paris, 92195 Meudon France Department of Physics and Astrophysics, Open University, WaltonHall, Milton Keynes MK7 6AA, UK19