Deriving extinction laws with O stars: from the IR to the UV
Highlights of Spanish Astrophysics VIII, Proceedings of the XI Scientific Meeting of the Spanish Astronomical Society held on September 8 – 12, 2014, in Teruel, Spain. A. J. Cenarro, F. Figueras, C. Hernández-‐Monteagudo, J. Trujillo, and L. Valdivielso (eds.)
Deriving extinction laws with O stars: from the IRto the UV
J. Ma´ız Apell´aniz Centro de Astrobiolog´ıa, INTA-CSIC, Spain
Abstract
We have recently derived a family of extinction laws for 30 Doradus that provides betterfits to the optical photometry of obscured stars in the Galaxy and the LMC. Simultane-ously, we are extending our Galactic O-Star Spectroscopic Survey (GOSSS) to fainter, moreextinguished stars to obtain accurate spectral types for massive stars with more than 6magnitudes of V -band extinction. I have combined both lines of research with 2MASS,WISE, and Spitzer photometry to obtain the 1-10 micron extinction law for O stars in thesolar neighborhood. I present these results and compare them with the extinction laws inthe same wavelength range derived from late-type stars and H ii regions. I also discuss plansto extend the newly derived optical-IR extinction laws to the UV. Walter Baade was once asked if he would become an astronomer if he were born again.He replied that he would only do it if the extinction law were the same everywhere. Inthe mid-twentieth century astronomers worried about the variability of the extinction lawin the optical but the preoccupation now extends to other ranges, such as the IR and theUV. Cardelli et al. (1989, CCM) produced the first single-parameter family of extinction lawsthat extended from the IR to the UV. In Ma´ız Apell´aniz (2013a) we described some of theproblems with the CCM family laws and in Ma´ız Apell´aniz et al. (2014a) we proposed analternative family of extinction laws for the optical range that used the CCM results for thetwo other ranges. In this contribution we proceed from the IR to the UV by [a] extendingour analysis to longer wavelengths using Galactic O stars, [b] summarizing the new laws ofMa´ız Apell´aniz et al. (2014a) in the optical, and [c] presenting our plans to produce a newfamily of extinction laws that also reaches the UV. The new analysis relies heavily on theGalactic O-Star Spectroscopic Survey (GOSSS), described in Ma´ız Apell´aniz et al. (2011).To date, we have published the blue-violet spectra and the spectral types of 448 Galactic Ostars (Sota et al., 2011, 2014) and we have already observed over 2000 stars (of O type andothers). See the contribution by Ma´ız Apell´aniz et al. in these proceedings for additional a r X i v : . [ a s t r o - ph . S R ] O c t Deriving extinction laws with O stars: from the IR to the UV
GOSSS results on a new spectral library, a spectral classification tool, and a first list of starswrongly classified as O in the literature.Before proceeding, it is worth reiterating something we have mentioned elsewhere: R V ,the band-integrated ratio of total to selective extinction depends not only on the type of dustbut also on its amount and on the type of stellar SED (Ma´ız Apell´aniz, 2013a). Therefore, R V should not be used to characterize an extinction law . Instead, one should use monochromaticquantities such as R . In the IR, the CCM laws use a power law A ( λ ) /A (1 µ m) = x α ( x ≡ /λ , with λ in µ m)with α = 1 .
61 fixed, which was adopted from Rieke & Lebofsky (1985). More recent works inthe IR have found different values of the exponent, possible variability from one sightline toanother, and complex behaviors as a function of wavelength (e.g. Moore et al., 2005; Rom´an-Z´u˜niga et al., 2007; Nishiyama et al., 2009; Gao et al., 2009; Sch¨odel et al., 2010). Thosepapers, however, use as reference targets cool stars and/or H ii regions, whose intrinsic SEDsare subject to larger uncertainties than those of OB stars. Furthermore, if one uses differentobjects and/or environments to derive extinction laws, one cannot be certain that possibledifferences are intrinsic to the extinction law since they may be instead caused by the sampleheterogeneity. Therefore, if one derives extinction laws in different wavelength regimes (e.g.UV, optical, and IR), one should use at least a subsample that spans all of them. This iswhat Fitzpatrick & Massa (2009) attempted with OB stars but their sample was small (14objects) and only included targets with A V up to ∼
4, for which extinction in the IR is small,leading to a measurement of the extinction law with relatively large uncertainties.I have started reanalyzing the IR extinction laws with GOSSS, taking advantage of: • Its large sample size of Galactic targets with spectral types: we are getting close to1000 O and 1000 B stars observed. • The accuracy and detail of its spectral types, which allows for the detection and elim-ination of peculiar objects that may have non-standard intrinsic SEDs in the IR (e.g.strong winds, circumstellar disks, obscured companions). • Its large range of extinctions (up to A V ∼
9) that includes newly obtained high-extinction objects observed with large telescopes. • The recent availability of WISE and Spitzer photometry for many of the GOSSS targets.I have proceeded in the following way:1. I collected the 2MASS J + H + K s , WISE W W W
3, and Spitzer IRAC 3 . . . . . See http://wise2.ipac.caltech.edu/docs/release/allsky/expsup/sec6_3c.html . a´ız Apell´aniz E ( J − H )−0.10.00.10.20.30.40.50.60.70.8 un c o rr e c t ed E ( H − K ) AllOe + O IafpeOf?pEarly Of/WNOther discarded E ( J − H )−0.10.00.10.20.30.40.50.60.70.8 un c o rr e c t ed E ( H − K ) V not discarded I not discarded E ( J − H )−0.10.00.10.20.30.40.50.60.70.8 c o rr e c t ed E ( H − K ) Not discarded α = 2.39 α = 1.61 Figure 1: (top) Uncorrected E ( H − K s ) vs. E ( J − H ) plot for the GOSSS O-star sample.(bottom left) Same plot (without error bars) for spectral luminosity classes I and V afterdiscarding stars with likely IR excesses. The linear fits to the two samples used to determinethe wind correction to E ( H − K ) are also shown. (bottom right) Final plot after applyingthe luminosity-class-dependent corrections. The expected extinction effect for two power lawswith different exponents is also shown. Deriving extinction laws with O stars: from the IR to the UV
2. I assigned TLUSTY intrinsic SEDs (Lanz & Hubeny, 2003) to each star according tothe spectral type- T eff calibration of Martins et al. (2005) adapted to the new spectraltype subdivision (Sota et al., 2014) using the grid of Ma´ız Apell´aniz (2013b). Note thatTLUSTY does not include wind effects, so small corrections are required in the IR, andthat those corrections are expected to be larger for higher luminosity stars.3. I calculated the color excesses E ( J − H ) and E ( H − X ) (where X is any of the 2MASS+ WISE + Spitzer IRAC filters above except J or H ) for each star, discarded the starswith likely IR excesses (Figure 1, top), divided the sample by spectral luminosity classes(I to V), and fitted straight lines to the E ( H − X ) vs. E ( J − H ) for each subsample(Figure 1, bottom left). The intercepts were then used to produce luminosity-class-dependent corrections for E ( H − X ) and corrected plots (Figure 1, bottom right).Only the case where X = K s is shown in Figure 1 but the process was repeated for allfilters. For K s the correction increases monotonically from 0.031 mag (luminosity classV) to 0.070 magnitudes (luminosity class I) and the behavior is simlar for other filters.4. The next effect was to select a extinction law model. In the NIR a power law iscommonly used but beyond 2.5 µ m different structures are expected (Fritz et al., 2011).I chose a simplified three-parameter ( α + β + γ ) model for A λ /A ( A being the extinctionat 1 µ m) consisting of: • A power law with exponent α between 1 µ m and 2.5 µ m. • Another power law with exponent β beyond 2.5 µ m. The two power laws are joined“a la Kroupa” at 2.5 µ m but in order to preserve the derivability a weighted sumof the two power laws is applied between 2.4 µ m and 2.6 µ m. • A gaussian component centered at 3.3 µ m with a σ of 0.2 µ m and a peak intensityof γ . The purpose of this component is to model the combined effect of H Oand aliphates at these wavelengths, which are the lines with the stronger expectedequivalent widths in the 1-8 µ m range (Fritz et al., 2011).5. I fitted the model to the most extinguished stars in the GOSSS sample using a χ -minimizing code with four free parameters: α , β , γ , and A . Note that different starsmay have different photometric bands available: we only used those cases with at leastfive existent magnitudes. Also note that WISE W3 was not used in the fitting for threereasons: [a] low S/N and confusion with background, [b] possible presence of silicateabsorption around 10 µ m (Fritz et al., 2011), and [c] possible presence of excesses dueto circumstellar cool material. Nevertheless, the W3 band photometry is shown in theplots in Figure 2 for reference.6. An average extinction law was built from a selection of the fitted stars. The averagevalues found were: α = 2 . β = 1 . γ = 0 . µ m, in agreementwith other recent results. The intensity of the H O + aliphates peak around 3.3 µ m isroughly consistent with the extinction law of Fritz et al. (2011). a´ız Apell´aniz µ m)−1.0−0.50.00.51.01.5 A λ − A H ( m ag ) A = 2.569, α = 2.083, β = 1.765, γ = 0.0585Average: A = 2.745, α = 2.223, β = 1.439, γ = 0.0540Fitted photometry µ m)−1.0−0.50.00.51.01.52.0 A λ − A H ( m ag ) ALS 21 079 A = 3.035, α = 1.874, β = 1.360, γ = 0.0266Average: A = 3.322, α = 2.223, β = 1.439, γ = 0.0540Fitted photometry µ m)−1.0−0.50.00.51.0 A λ − A H ( m ag ) ALS 15 128 A = 2.167, α = 1.959, β = 0.854, γ = 0.0707Average: A = 2.162, α = 2.223, β = 1.439, γ = 0.0540Fitted photometry µ m)−1.5−1.0−0.50.00.51.01.52.0 A λ − A H ( m ag ) SS 215 A = 3.001, α = 1.067, β = 0.753, γ = 0.0468Average: A = 3.468, α = 2.223, β = 1.439, γ = 0.0540Fitted photometry Figure 2: IR extinction-law plots for four GOSSS stars. 2MASS J20344410+4051584 is acase with only 2MASS + Spitzer IRAC photometry. ALS 21 079 shows the possible effect ofthe 10 µ m silicate band in W3. ALS 15 128 apparently has a 10 µ m excess. SS 215 is a casewhere strong winds invalidate our simplified approach to calculate the IR extinction law. Deriving extinction laws with O stars: from the IR to the UV
7. The four examples in Figure 2 can be used to give an idea of the diversity of thedata. SS 215 is an O2 If*/WN5 star (Sota et al., 2014) that shows the largest apparentdeviation from the average extinction law. However, such stars have strong winds whoseeffect in the IR SED is likely not to be corrected by the simplified procedure describedhere. Therefore, the low value of α measured is likely to be an artifact and not a realextinction effect.The results presented here are preliminary. We are currently working on expandingour high-extinction sample before publishing our final rsults on the IR extinction law usingGOSSS stars. Ma´ız Apell´aniz et al. (2014a) have combined the spectral types from the VLT-FLAMESTarantula Survey (VFTS Evans et al., 2011; Walborn et al., 2014) with photometry from theHST/WFC3 Early Release Science (GO 11 360) to produce a new family of extinction lawsfor the optical region. The new laws: • Maintain the overall behavior and good characteristics of CCM: they are a single-parameter ( R ) continuous and differentiable family. • Eliminate the U band excesses detected in CCM for all values of R . • Alleviate the wiggles introduced by the use of the seventh degree polynomial in 1 /λ usedby CCM and make the extinction laws more Whitford-like (Whitford, 1958; Ardeberg& Virdefors, 1982).The new family of extinction laws produce significant better fits for the optical-NIRphotometry of both Galactic and 30 Doradus targets . However, they should not be the finalword for four reasons: • Ma´ız Apell´aniz et al. (2014a) assumed the Rieke & Lebofsky (1985, the same as CCM)exponent in the IR, which is too shallow, as we have seen here. Therefore, the opticallaws should be “stitched” to a more correct IR law (or laws, if it is finally shown thatthere is real variation in the IR). • The more correct way of deriving extinction laws is with spectrophotometry, not withphotometry, thus avoiding interpolation in wavelength. This is easier to do in theoptical than in the IR and we have already obtained data with this purpose. • The “stitching” with the UV should also be revisited (see next section). The differences between MW, LMC, and SMC extinction laws studied in the past refer to the UV. a´ız Apell´aniz • Finally, some discrete absorption features in the ISM have been extensively studied inthe optical and some of them are highly correlated with extinction. Therefore, it shouldbe possible to include them in a high-spectral-resolution version of the extinction law,a goal towards which we are also working in (Ma´ız Apell´aniz et al., 2014b).In summary, once a new IR extinction law is obtained, it should be possible to improveupon the optical results of Ma´ız Apell´aniz et al. (2014a).
The non-specialist astronomer may think that the UV extinction laws were long agosettled with IUE. However, that is far for the truth, as the following points show: • The CCM laws used a single parameter ( R V or, more properly, R ) to describe thewhole UV-to-IR wavelength range. However, Fitzpatrick & Massa (2007) claim thatwith the exception of a few curves with large values of R , the UV and IR portionsof Galactic extinction curves are not correlated with each other, which is in directcontradiction with CCM. • Fitzpatrick & Massa (2007) also find that “the central position of the 2175 ˚A extinc-tion bump is mildly variable, its width is highly variable, and the two variations areunrelated.” • It is usually expressed that for the SMC there is no 2175 ˚A bump in the extinctionlaw. Actually, Gordon et al. (2003) find four SMC sightlines without a bump andone with a weak bump. More recently, Ma´ız Apell´aniz & Rubio (2012) studied fouradditional SMC sightlines and found one with a significant bump, two with a weakbump, and one without it. Clearly, more sightlines are needed (an ubiquitous issuewith UV extinction). • A bump-less extinction curve may not be exclusive to low metallicity environments suchas the SMC: Valencic et al. (2003) found an example in the Milky Way. • In some objects geometry may be at work. As shown in another contribution to theseproceedings by Ma´ız Apell´aniz et al., a large fraction of the UV light coming fromHerschel 36 and its surroundings is actually scattered radiation from the H ii region, aneffect that is also known to be important in the Orion Nebula. For Herschel 36, IUEwas capable of resolving the star from the nebulosity but in more distant objects wemay be considering the joint flux instead of just that received directly from the star. • Speaking of geometry, when the UV radiation originates from multiple sources or scat-tered radiation is included, one should use the term “attenuation law” instead of “ex-tinction law”. A popular example is the Calzetti law (Calzetti, 2001).
Deriving extinction laws with O stars: from the IR to the UV
In summary, there is a relatively large degree of confusion with UV extinction. Weknow it is highly variable but there is contradictory information in the literature regardinghow that variation takes place. This hampers the solution of the ultimate questions of theorigin of the extinction law and its dependence on metallicity and environment. Fortunately,there are two lines of work that may help us clear the waters.On the one hand, IUE has an excellent archive that can be combined with new data andtechniques. Modern surveys provide optical-IR photometry of better quality than what wasavailable until recently. Spectroscopic surveys such as GOSSS and new spectral libraries andmodelling can also constrain the intrinsic SEDs of the sources better and reduce systematicerrors. A reanalysis of this combination of old and new data is a necessary step that we planto attack in the near future.On the other hand, new UV spectroscopy is clearly required. The number of IUE spec-tra of Galactic sightlines with large R is very small. The number of studied extragalacticsightlines (even for the MCs) is also small and, in many cases, with very low extinctions(which amplifies the effect of systematic errors). Along this line, it would be very useful toobtain UV spectroscopy of some of the sources in Ma´ız Apell´aniz et al. (2014a), since wewould kill those two birds (large R and the LMC) with one stone. Acknowledgments
I acknowledge support from [a] the Spanish Government Ministerio de Educaci´on y Cienciathrough grants AYA2010-15081 and AYA2010-17631 and [b] the Consejer´ıa de Educaci´on of the Juntade Andaluc´ıa through grant P08-TIC-4075.
References
Ardeberg, A., & Virdefors, B. 1982, A&A, 115, 347Calzetti, D. 2001, PASP, 113, 1449Cardelli, J. A., Clayton, G. C., & Mathis, J. S. 1989, ApJ, 345, 245Evans, C. J., et al. 2011, A&A, 530, A108Fitzpatrick, E. L., & Massa, D. 2007, ApJ, 663, 320Fitzpatrick, E. L., & Massa, D. 2009, ApJ, 699, 1209Fritz, T. K., et al. 2011, ApJ, 737, 73Gao, J., Jiang, B. W., & Li, A. 2009, ApJ, 707, 89Gordon, K. D., Clayton, G. C., Misselt, K. A., Landolt, A. U., & Wolff, M. J. 2003, ApJ, 594, 279Lanz, T., & Hubeny, I. 2003, ApJS, 146, 417Ma´ız Apell´aniz, J. 2013a, in Highlights of Spanish Astrophysics VII, 583 a´ız Apell´aniz9