Probing the Inner Regions of Protoplanetary Disks with CO Absorption Line Spectroscopy
Matthew McJunkin, Kevin France, Eric B. Burgh, Gregory J. Herczeg, Eric Schindhelm, Joanna Brown, Alexander Brown
aa r X i v : . [ a s t r o - ph . S R ] F e b Draft version September 15, 2018
Preprint typeset using L A TEX style emulateapj v. 11/10/09
PROBING THE INNER REGIONS OF PROTOPLANETARY DISKS WITH CO ABSORPTION LINESPECTROSCOPY
Matthew McJunkin , Kevin France , Eric B. Burgh , Gregory J. Herczeg , Eric Schindhelm , Joanna M.Brown , Alexander Brown Draft version September 15, 2018
ABSTRACTCarbon monoxide (CO) is the most commonly used tracer of molecular gas in the inner regions ofprotoplanetary disks. CO can be used to constrain the excitation and structure of the circumstellarenvironment. Absorption line spectroscopy provides an accurate assessment of a single line-of-sightthrough the protoplanetary disk system, giving more straightforward estimates of column densitiesand temperatures than CO and molecular hydrogen (H ) emission line studies. We analyze newobservations of ultraviolet CO absorption from the Hubble Space Telescope along the sightlines tosix classical T Tauri stars. Gas velocities consistent with the stellar velocities, combined with themoderate-to-high disk inclinations, argue against the absorbing CO gas originating in a fast-movingdisk wind. We conclude that the far-ultraviolet observations provide a direct measure of the diskatmosphere or possibly a slow disk wind. The CO absorption lines are reproduced by model spectrawith column densities in the range N ( CO) ∼ − cm − and N ( CO) ∼ − cm − ,rotational temperatures T rot ( CO ) ∼
300 – 700 K, and Doppler b -values, b ∼ − . Weuse these results to constrain the line-of-sight density of the warm molecular gas ( n CO ∼ − − ) and put these observations in context with protoplanetary disk models. Subject headings: protoplanetary disks —– stars: individual (AA Tau, HN Tau, DE Tau, RECX-15,RW Aur, SU Aur) INTRODUCTION
Characterizing the gas properties of protoplanetarydisks at planet-forming radii ( a ≤
10 AU) is criti-cal to understanding the way planets form and evolve.The lifetimes of protoplanetary disks are comparableto the time during which giant planet cores form andaccrete gaseous envelopes (10 − yr; Fedele et al.2010; Hern´andez et al. 2007; Hubickyj et al. 2005). Thesurface density of the disk in which the planet formsplaces limits on the amount of migration that is possi-ble (Trilling et al. 2002; Armitage et al. 2003). The life-time of the protoplanetary disk determines the amountof time that giant planets are able to accrete, thereby in-fluencing their final masses (Ida & Lin 2004). Observa-tions of carbon monoxide (CO) emission and absorptionlines from ultraviolet (UV) to infrared (IR) wavelengthshave been shown to be useful probes of molecular gasin the inner disk (Najita et al. 2003; Salyk et al. 2011;Schindhelm et al. 2012).Emission and absorption from the CO Fourth Posi-tive band system ( A Π − X Σ + ) has been widely usedto study the interstellar medium (Federman et al. 1980;Burgh et al. 2007), debris disks (Vidal-Madjar et al.1994), comets (Feldman & Brune 1976; McPhate et al.1997), planets (Durrance 1981; Feldman et al. 2000), andthe atmospheres of cool stars (Carpenter et al. 1994). Center for Astrophysics and Space Astronomy, Univer-sity of Colorado, 389 UCB, Boulder, CO 80309, USA;[email protected] Kavli Institute for Astronomy and Astrophysics, Peking Uni-versity, Beijing 100871, China Southwest Research Institute, 1050 Walnut Street, Suite 300,Boulder, CO 80302, USA Harvard-Smithsonian Center for Astrophysics, 60 GardenStreet, Cambridge, MA 02138, USA
The first detections of CO far-UV emission and absorp-tion spectral features from the inner regions of proto-planetary disks were presented in France et al. (2011a).The detected CO absorption lines have rotation temper-atures T rot ( CO ) ≈ ±
200 K, pointing to an originin the warm inner disk gas. Molecular hydrogen (H )and CO absorption were measured simultaneously forthe first time by France et al. (2012a) in the disk of AATauri. The H was seen in absorption against the Ly α emission line, as in Yang et al. (2011), and the CO A − X absorption bands were observed against the far-UV con-tinuum (France et al. 2011b). The value of CO/H ∼ − . Better constraints on the CO/H ratioare important for determining the total amount of gas inprotoplanetary disks.Many classical T Tauri star (CTTS) disks have beenstudied with near-IR fundamental (4.7 µ m) and over-tone (2.3 µ m) bands of CO, in both emission (e.g.Carr et al. 1993; Najita et al. 2003; Salyk et al. 2011),and absorption (e.g. Rettig et al. 2006; Horne et al.2012). Near-IR CO emission has been used as a tracerof the dominant molecular gas component, H , becausethe homonuclear nature of H makes rovibrational tran-sitions dipole-forbidden and thus difficult to observeat near- and mid-IR wavelengths (Pascucci et al. 2006;Bitner et al. 2008; Carmona et al. 2008; although seeBary et al. 2003; Ramsay Howat & Greaves 2007; andBary et al. 2008). The CO lines at 4.7 µ m appear tooriginate from the inner disk region with T ≈
500 -1500 K and Keplerian velocities consistent with the dustsublimation radius. However, some sources have lineprofiles with excess low velocity emission (Bast et al. McJunkin et al.2011), which Pontoppidan et al. (2011) propose arisesfrom a slow molecular disk wind. In general, the IRCO temperatures from LTE slab models (275 - 1675 K,average ∼ ± ∼
100 K, likely placingthe absorbing gas farther out in the disk at r >
10 AU.The discrepancies between the IR and UV CO worksuggests that the observations are probing multiplemolecular gas populations, and multiple populations maybe responsible for the anomalously high observed CO/H ratio in AA Tau. With deep UV absorption spectroscopy,we have a new observational tool for studies of the innerdisk gas. Absorption lines give the most direct measureof the column density and temperature of the disk gasalong the line-of-sight as the line fitting is largely inde-pendent of the geometry or the photo-exciting source. Inorder to provide observational constraints on inner diskgas, we present new analyses of far-UV spectroscopic ob-servations of six (0.6 – 6 Myr) CTTSs. Our observationsprobe the warm disk atmosphere, which is important forconstraining the three-dimensional structure of disks inwhich planets form. We describe the targets and the ob-servations in §
2. The analysis performed and the modelfit parameters are presented in §
3. The disk geometryand correlations in the data are discussed in §
4. Finally, § TARGETS AND OBSERVATIONS
We analyze far-ultraviolet spectra of six targets thatare a subset of the 34 T Tauri stars presented inFrance et al. (2012b): AA Tau, HN Tau A, DE Tau,RECX-15 (ET Cha), RW Aur A, and SU Aur. HN TauA and RW Aur A (hereafter HN Tau and RW Aur) arebinary stars, but with separations large enough, 1.4” forRW Aur AB and 3.1” for HN Tau AB (White & Ghez2001), such that only the primary is within the aper-ture and dominates the emission analyzed here. AA Tauand HN Tau have been studied previously by France etal. (2011a, 2012a), but we refit their spectra with thesame procedure as the other targets for consistency. Fivetargets are in the Taurus-Auriga star-forming region ata distance of 140 pc (Bertout et al. 1999; Loinard et al.2007). RECX-15 is in the η Chamaeleontis cluster ata distance of 97 pc (Mamajek et al. 1999). Parametersincluding the age, luminosity, and spectral type of thesetargets are listed in Table 1.All targets were observed using the
Hubble SpaceTelescope -Cosmic Origins Spectrograph (
HST -COS;Green et al. 2012) under program ID 11616. The COSFUV M-mode wavelength solution is accurate to ∆v ∼
15 km s − and depends on the object centering . Targetacquisition was through the MIRRORB near-UV imag-ing mode for AA Tau and DE Tau. The rest of the tar-gets were acquired using the MIRRORA mode. Far-UVspectra were obtained using three central wavelengthsfor G160M, two central wavelengths for G130M, andmultiple focal-plane positions to cover the 1133 ≤ λ ≤ Table 2 lists the dates and exposure times for each ob-ject used in this paper. The data were processed throughCALCOS, the COS calibration pipeline, and aligned andco-added with the procedure described in Danforth et al.(2010).The six targets were the only objects in the T Tauristar sample to show unambiguous CO absorption. Therest of the larger sample either had low continuum signal-to-noise (S/N) making modeling difficult, or were com-plete non-detections of CO absorption (most likely fromlow-inclination disks not intercepting the line-of-sight).Our optical depth detection limit was τ = 5, below whichthe absorption depths were comparable to the noise inthe data. The targets that we do not fit have incli-nations ranging from 4 ◦ (TW Hya, Pontoppidan et al.2008) to 85 ◦ (DF Tau, Johns-Krull & Valenti 2001) withan average of 43 ◦ . Thus, the majority of the targetsthat do not show strong CO absorption have lower in-clinations. DF Tau is an unusual target discussed in § .
3. The six targets that we analyze all have medium-to-high inclination (35 ◦ - 77 ◦ , with an average of 61 ◦ )protoplanetary disks. The average does not include HNTau, whose inclination is not well known. The inclina-tions have been determined in a number of different ways,with fairly large uncertainties arising from the techniquesused. The inclination of AA Tau was determined by pe-riodic eclipses of the star by the magnetically warpedaccretion disk combined with a measured line-of-sightprojected rotation velocity in the optical (Donati et al.2010; Bouvier et al. 2003). For DE Tau, it was calcu-lated from stellar radii and rotation periods spectroscop-ically determined in the IR combined with the literaturevalue of v sin i in the red (Johns-Krull & Valenti 2001).Optical spectroscopy of a very strong H α emission lineprofile was modeled to find the inclination of the diskof RECX-15 (Lawson et al. 2004). The interferometricdata of RW Aur was fit with an inclined uniform diskmodel and was sufficient to constrain the inclination ofthe disk (Eisner et al. 2007). This inclination is incon-sistent with the RW Aur inclination ( ∼ ◦ ) found byL´opez-Mart´ın et al. (2003) using the ratio of proper mo-tion to radial velocity toward emission knots in the RWAur jet. The visibility as a function of hour angle in theK-band of SU Aur was fit with a Gaussian brightnessprofile that was inclined on the sky, giving the best fitinclination for the disk (Akeson et al. 2002). The inclina-tion of HN Tau is not well known, and only a lower limitis given to indicate that an inclination higher than 40 ◦ is needed for the line-of-sight to intercept the disk andto explain observed outflows (France et al. 2011a). Thislower limit is increased in § . ∼ ,
000 (10 km s − at6300 ˚A).The CO is observed in absorption against the far-UV continuum of these actively accreting stars. TheO Absorption in Protoplanetary Disks 3 TABLE 1Target Parameters
Object Spectral A V Inclination L ∗ M ∗ ˙M log (Age) Ref. a Type (degrees) (L ⊙ ) (M ⊙ ) (10 − M ⊙ yr − ) (yrs)AA Tau K7 0.5 70 0.71 0.80 0.33 6.38 ± ± >
40 0.19 0.85 0.13 6.27 ± ± ± ± a (1) Akeson et al. (2002); (2) Andrews & Williams (2007); (3) Bertout et al. (1988); (4) Donati et al. (2010);(5) Eisner et al. (2007); (6) France et al. (2011a); (7) Gullbring et al. (1998); (8) Gullbring et al. (2000); (9)Hartigan et al. (1995); (10) Johns-Krull & Valenti (2001); (11) Johns-Krull et al. (2000); (12) Kraus & Hillenbrand(2009); (13) Lawson et al. (2004); (14) Luhman & Steeghs (2004); (15) Ramsay Howat & Greaves (2007); (16)Ricci et al. (2010); (17) White & Ghez (2001). TABLE 2
HST -COS G130M/G160M Observations of Targets
Object R. A. (J2000) Dec. (J2000) Date G130M Exposure G160M Exposure(s) (s)AA Tau 04 34 55.42 +24 28 52.8 2011 Jan 06, 07 5588 4192DE Tau 04 21 55.69 +27 55 06.1 2010 Aug 20 2088 1851HN Tau 04 33 39.37 +17 51 52.1 2010 Feb 10 5724 4528RECX-15 08 43 18.43 -79 05 17.7 2011 Feb 05 3890 4501RW Aur 05 07 49.51 +30 24 04.8 2011 Mar 25 1764 1617SU Aur 04 55 59.39 +30 34 01.2 2011 Mar 25 1788 1759 far-UV continuum arises from stellar photospheric andchromospheric emission as well as an accretion con-tinuum (Calvet & Gullbring 1998; France et al. 2011b;Ingleby et al. 2011; see Table 1 for the accretion rates).The far-UV continuum may be enhanced in someobjects by a molecular dissociation quasi-continuum(Bergin et al. 2004; Herczeg et al. 2004), but the detailsof this process have not been thoroughly characterizedobservationally (see, e.g., France et al. 2011b). The con-tinuum emission is absorbed by circumstellar atoms andmolecules. Depending on the inclination angle, the lightmay pass through multiple layers of the disk and thusprobe multiple temperature and density regimes. At thehigh temperatures and column densities of the observedCO (see § . J >
2. For temperaturestypical of the interstellar medium (ISM), states with J ≥ V . . N (CO) . (see Figure 3 ofBurgh et al. 2007), which is orders of magnitude smallerthan the circumstellar column densities that we observe.Therefore, our CO measurements do not probe the colddiffuse interstellar material with
Data Analysis
We identify nine CO bands, ( v ′ - 0), that span the COSfar-UV bandpass from the (0 - 0) band at λ ∼ λ ∼ and Si IV λ λ emissionlines populate the spectra, complicating the model spec-trum fits as well. All CTTS systems with identifiableCO absorption contain broad bands created by the over-lap of many closely spaced rotational lines. This sug-gests a relatively high rotational temperature, T rot ( CO ),because as higher- J lines are populated, the observedspectral bands become broader. Broad CO absorptionbands were first identified in the spectrum of HN Tau(France et al. 2011a) and later in the spectrum of AATau (France et al. 2012a). We employ a similar tech-nique to the previous work, using spectral synthesis mod-eling to measure T rot ( CO ) and N(CO) in all six sources.The spectrum around each band was normalized by alinear fit to the nearby continuum, avoiding regions withH or other emission lines. The least contaminated ab-sorption bands are the (1 - 0), (2 - 0), (3 - 0), and (4 - 0)bands, which also have the strongest oscillator strengthsin the Fourth Positive system. However, CO and COare not cleanly resolved in the low- v ′ bands. To isolatethese isotopologues, higher- v ′ bands with larger separa-tions between the CO and CO bandheads ( v ′ = 7,8) are used to better constrain the column densities ofboth species (see § . ∼ v ′ = 7, 8 absorption McJunkin et al. Fig. 1.—
The nine far-UV CO rovibrational absorption bandheads identified by orange vertical lines. Going from left to right: ( v ′ - v ′′ = 8 - 0) 1322 . . . . . . . . . O Absorption in Protoplanetary Disks 5 !" ()*+! !- () ./*012 !-3- !-34 !-3$ !-"%5167879:;<*+ Å , ************************************** = > ? @ B7 C * D !- () ()*+F Fig. 2.—
Low- v ′ (above) and high- v ′ (below) model fits forDE Tau showing the larger separation between CO and COfor larger v ′ -values. The separation helps to better constrain the CO/ CO ratio in the model. The data is in black and the best-fit model is in orange. The first absorption dip in each model is the CO bandhead, and the CO bandhead is marked with a bluearrow. (see § J -lines present.DE Tau and SU Aur have the largest depths, but thedata are also noisier, making it harder to determine acontinuum level. The depths will vary, even for similarcolumn densities, because of the error associated withthe continuum fit and population shifts in the lower- J levels due to increases and decreases in temperature (see § § . J values.Here, the normalization will have a bigger effect. Athigh- J values (longer wavelengths within the band) it isdifficult to pinpoint the weaker absorptions amongst theemission lines and the noise in the data. This makesthe data harder to interpret, though RW Aur does havea significant lack of high- J lines compared to the othertargets. HN Tau, however, has fairly prominent high- J lines. All of these factors contribute to the derivedmolecular parameters and errors described in § § Model Description and Fitting Procedure
The CO absorption profiles were modeled for COusing the oscillator strengths and rovibrational linewavelengths from Eidelsberg (private communication).The A Π energy levels of Haridass & Huber (1994)were used to calculate the ground state energies.The oscillator strengths for CO are taken fromEidelsberg et al. (1999), and the wavelengths and oscil-lator strengths for the perturbation states were takenfrom Eidelsberg & Rostas (2003). A common rotationaltemperature was assumed for both CO and CO. All six of the analyzed bands ( v ′ = 1, 2, 3, 4, 7, 8) werecompared to model spectra for a grid of different values ofthe Doppler b -value, rotational temperature, T rot ( CO ),and logarithmic column density of both CO and CO,log ( N ( CO)) and log ( N ( CO)). The model also in-cludes a velocity shift parameter (explored in more de-tail in § § − in steps of 0.1 kms − for the Doppler b -value, 300 – 1000 K with steps of50 K for the rotational temperature, 16.0 – 18.0 in stepsof 0.1 for log ( N ( CO)), and 14.0 – 17.0 in steps of 0.1for log ( N ( CO)). The maximum b -value for our gridsearch comes from our assumption of turbulent velocities . − and CO rotational temperatures . × K. The maximum log ( N ( CO)) value for the searchcomes from the analysis of France et al. (2012a), who ar-gue that any larger column of CO would lead to a col-umn of hydrogen that would extinguish the stellar fluxaround the Lyman alpha line center, which is inconsis-tent with our observations. This limit is confirmed afterthe fit as the majority of the targets (except those withthe highest column, RECX-15 and SU Aur) are incon-sistent with a column of 10 cm − within their errors.Any CO column density lower than our lower limit of10 cm − (which corresponds with our optical depth de-tection limit of τ = 5) would be very difficult to detectwith the S/N, setting the lower bound on our columndensity search. An example of our CO model is shownin Figure 3. The higher rotational states are excited athigher temperatures while the lower rotational states aredepopulated. Thus, the observed absorption band be-comes wider and shallower with increasing temperatureif column density is kept constant. The three simulatedspectra at the top of Figure 3 are at the native modelresolution (∆v ∼ − ) and the bottom spectrumis at the resolution of the HST -COS instrument (∆v ∼
17 km s − ). The CO absorption lines are narrower thanthe instrumental resolution. The unresolved line cores donot go to zero at the observational resolution even whenthe CO optical depths are large.Because the statistical errors in the data were some-times anomalously small , such that a reduced χ fittingroutine with non-uniform errors gave erroneous results,we employed a simple least squares fitting routine thatminimizes the square of the difference between the modeland the data. This procedure weighted all data pointsequally, giving a more reliable fit. Because of the pres-ence of emission lines (mostly H ), we truncated the dataabove a normalized flux level of 1 . The CALCOS pipeline is known to mishandle statistical errorsin the low S/N regime (Froning et al. 2011)
McJunkin et al.
Fig. 3.—
Three different CO models for the (2 - 0) bandof DE Tau with identical Doppler b -value = 1 . − ,log ( N ( CO)) = 17 . ( N ( CO)) = 16 .
0, and velocityshift = 1 . − . (Top 3): Model output with native modelresolution (∆v ∼ − ). (Bottom): Model output after beingconvolved with the HST -COS linespread function (∆v ∼
17 kms − ) to simulate data detected with the instrument. density to fit the higher flux values at these points, in-troducing a bias which contributes to the large errors onthe column density in Table 3.The model uses the five input parameters to create atheoretical spectrum. The Doppler b -value, along withthe column density, determines the opacity of the model,and the temperature determines the band shape as dis-cussed above, with higher temperatures leading to wider,shallower absorption models. The column density inputsdetermine the relative depth of the CO and CO ab-sorption bands. The CO/ CO ratio is primarily de-termined by the differences in the absorption depths atthe relevant bandheads. At the high column densities ofthese disks (see § . J CO lines of the lower- v ′ bands. This means that in-creasing the column density of CO does not lead toa significant increase in the absorbed flux. The columndensity fits rely more on the CO absorption and thehigher- v ′ bands and higher- J lines of CO, which areless saturated. The velocity shift parameter moves theabsorption band of the model to shorter or longer wave-lengths corresponding to the gas moving toward or awayfrom the observer, respectively. The theoretical spec-trum is then convolved with the COS linespread functionand compared to the observations to find the best-fit pa-rameter values.
Model Fits and Errors
The fits to the normalized absorption data are shownin Figure 4 and the parameter values are provided inTable 3. DE Tau and SU Aur, which have the deepestabsorptions, have higher column densities in both COand CO than most of the other targets. The shal-lower low- J lines combined with the prominent high- J lines in HN Tau require a higher temperature and lowercolumn density to fit the data. Most of the targetsshare a common best-fit temperature ( ∼
350 K), withHN Tau being an outlier because of the more prominent
Fig. 4a.—
Low- v ′ model fits (orange) for AA Tau. Thebest-fit values (see Table 3) for the Doppler b -value, T rot ( CO ),log ( N ( CO)) , log ( N ( CO)), and velocity shift are used. Thedata (black) are continuum normalized and the CO bandheadsare marked to clearly show the two different CO species. The emis-sion contaminating the absorption is not included in the model andany flux above a normalized value of 1.1 is truncated during thefit.
Fig. 4b.—
Same as Figure 4A for DE Tau. high- J lines. The column densities are much more var-ied, with large uncertainties because of the high- J linesblending in with the low S/N continuum and contamina-tion by emission lines. Unlike the temperature, there isno common best-fit column density for all of the targetswithin the uncertainties. An intersystem band not takeninto account in the code ( a ′ Σ + − X Σ + (14 - 0)) at λ = 1419 .
50 ˚A (Eidelsberg & Rostas 2003) hinders thefit of the (4 - 0) band of HN Tau, the highest temper-ature target. This absorption is between the CO and CO bandheads, causing the data to be deeper than themodel in this part of the spectrum. A few targets alsoshow this difference in the (2 - 0) band as well, whichcould be another intersystem band.Our fit values for all six targets lie in the rangesof b = 0.5 – 1.4 km s − , T rot ( CO ) = 300 – 700 K,log ( N ( CO)) = 16.0 – 17.8, and log ( N ( CO)) =14.7 – 16.7. The errors were estimated by taking thestandard deviation of the χ distribution with appropri-ate degrees of freedom. Increasing our minimum χ valueby the standard deviation determined the best-fit param-O Absorption in Protoplanetary Disks 7 TABLE 3CO Fit Parameters
Object F cont (1420) log ( N ( CO)) log ( N ( CO)) T rot ( CO ) b v CO v ∗ v ∗ Reference a (10 − erg cm − s − ˚A − ) (K) (km s − ) (km s − ) (km s − )AA Tau 6.1 16 . +0 . − . . +1 . − . +550 − . +0 . − . +24 . ± . . ± .
04 2DE Tau 3.7 17 . +0 . − . . +1 . − . +1000 − . +0 . − . +1 . ± . . ± .
018 2HN Tau 7.9 16 . +0 . − . . +0 . − . +1200 − . +1 . − . +28 . ± . . ± . . +0 . − . . +1 . − . +1850 − . +0 . − . +18 . ± . . +0 . − . . +0 . − . +600 − . +0 . − . +20 ±
20 15 . ± . b . +0 . − . . +1 . − . +1250 − . +0 . − . +16 . ± . . ± .
05 2 a (1) Woitke et al. (2011); (2) Nguyen et al. (2012). b Value for RW Aur B
Fig. 4c.—
Same as Figure 4A for HN Tau.
Fig. 4d.—
Same as Figure 4A for RECX-15. eter range. We define errors as the width of the best-fitparameter space, as defined by the standard deviation ofthe χ distribution. This error procedure is illustratedin Figure 5. The errors on log ( N ( CO)) were moredifficult to define because of the lower column density of CO. For log ( N ( CO)) .
14, the χ value does notincrease substantially as the column density is decreasedbecause we are no longer detecting the absorption. Wetake a lower limit on the column density of CO of 10 cm − for this reason. Within the errors, HN Tau, RWAur, and SU Aur are consistent with a non-detectionof CO (column density of CO < cm − ). The Fig. 4e.—
Same as Figure 4A for RW Aur.
Fig. 4f.—
Same as Figure 4A for SU Aur. temperatures in general tend to have large upper errorbars. Higher temperatures populate the higher J -states,which produce weaker absorption lines due to smallercolumn densities. Due to the low S/N in the data, theseweak lines blend into the continuum making the hightemperature limit difficult to constrain. At temperatures & v ′ - 1) rovibrational bands of CO shouldbe detectable ( τ ≥
5) in the data at our best-fit columndensities. At our best-fit temperature of ∼
350 K, the( v ′ - 1) band would be detected for log ( N ( CO)) & !" ’() *)) *() +)) +(),-.&--/ & :;:< = ’( ’> ’? ’@ ’A;4. ’) BCB ’* D0EE * E*))**)*G)*>)*@)+)) H I J - : .K < - * Fig. 5.—
Illustration of the uncertainty determination. (Above):A plot of the χ distribution for the degrees of freedom of RWAur. The standard deviation of the distribution is shown in red;the mean is defined by the dashed line. (Below): A plot of theunweighted χ value as a function of log ( N ( CO)). We increasedthe minimum χ by the standard deviation of the distribution todefine our error region. The dashed line shows the best-fit columndensity for RW Aur (10 . cm − ), the red line shows the standarddeviation of the above χ distribution, and the orange line definesthe error region where it crosses the unweighted χ line. which suggests that the CO gas is not at densities muchgreater than 10 cm − , or at temperatures much greaterthan 1000 K, though our models of the ( v ′ - 0) bandsalone do not provide strong constraints on the high endof the temperature range. We find similar results withinthe errors for AA Tau to the absorption line analysis ofFrance et al. (2012a). Our b -values are much higher thanthose found with sub-millimeter observations of the diskof TW Hya ( b . − ) by Hughes et al. (2011),however, this could be a consequence of their observa-tions probing colder, more quiescent CO gas at disk radiiof ∼
100 AU.The errors are driven mainly by three sources: con-tinuum determination, low S/N, and saturation in theabsorption lines. Estimates of the continuum fluxes aredifficult due to the large number of emission lines in thesespectra, especially the emission coincident with the ab-sorption bands. The error in the determination of thecontinuum flux is ∼ ∼ .
25 dex. The low S/N in the contin-uum (S/N per resolution element ∼ λ b -values. We includehigher- v ′ bands in our fit as they have smaller oscillatorstrengths and are therefore less saturated.The line optical depths for all the targets (except HNTau) are of order 100 – 600 for the low- v ′ transitions ( v ′ = 0 - 6) up to J -values of 15 to 25, which puts these lines on the flat portion of the COG. HN Tau, however, hasoptical depths of about 20 or less for all v ′ transitionsand even reaches optical depths of order unity in themodels for the (7 - 0) band and therefore was undetectedin the observations. The lower optical depths are causedby the higher temperature and higher b -value of HN Tau.The higher temperature (700 K) of HN Tau lowers theline strength of each J -line compared to if the gas wascooler by distributing the absorption over more transi-tions. Similarly, the higher b -value (1.4 km s − ) low-ers the line center cross-section through larger Dopplerbroadening spreading out the cross-section in wavelengthspace. The product of the lower line strength and lowercross-section in each J -line leads to a significantly loweroptical depth for HN Tau. None of the other targets ap-proach optical depths as small as 1 until the high- J levelsof the v ′ = 7, 8 states. RESULTS AND DISCUSSION
CO Velocity and Isotopic Fraction
The radial velocities of the CO absorption lines listedin Table 3 were obtained with a simple least squares fit.Assuming the best-fit parameters of the CO model, ve-locity shifts from -200 km s − to +200 km s − with 1km s − intervals were explored. Due to low S/N, the (8- 0) band was not used in the velocity calculation. Theaverage and standard deviation of the velocity shifts forthe observed bands were taken as each target’s velocityand velocity error. Shifting the velocity moves the modelhigh- J CO lines relative to the observed spectra, whichcan change the best-fit temperature. However, the tem-perature would change by only ∼
50 K, which is wellwithin our errors. The range of acceptable velocities wassmall enough that the best-fit parameters from the modelwere not affected by the shift.Since RW Aur contains strong, redshifted H emis-sion lines (France et al. 2012b), which sometimes placesthe emission on top of the CO absorption bandhead, wechecked the velocity shift by hand. We required that thebandhead of the model and data match for the (1 - 0),(2 - 0), and (4 - 0) bands, which are the cleanest andstrongest absorption bands. The best-fit velocity shiftis +20 km s − , consistent with the stellar radial veloc-ity (+15 km s − ) in the literature (Nguyen et al. 2012).The fit starts to become noticeably worse at ±
20 km s − from the best-fit value, which we take as our error.The fitted CO absorption line velocities are generallyconsistent with the radial velocities of the stars from theliterature to within the 15 km s − absolute uncertainty inthe COS wavelength scale. However, the CO absorptionin DE Tau appears to be somewhat blueshifted relativeto the stellar velocity (see Table 3), but is still consistentwithin the errors. As noted by Nguyen et al. (2012), thelow stellar radial velocity of HN Tau (4.6 km s − ) devi-ates from the average velocity of the Taurus-Auriga star-forming region ( ∼
15 km s − ; Hartmann et al. 1986).The quoted stellar radial velocity of HN Tau is incon-sistent with our CO velocity, but is only slightly out-side our errors. These small velocity differences betweenthe CO absorption and the star (∆v = 2.04 - 23.4 kms − ) indicate that the CO is approximately at rest in thestellar frame, or at least not in a fast-moving disk wind.This is in agreement with studies presented by Bast et al.O Absorption in Protoplanetary Disks 9(2011) and Najita et al. (2003) who find CO emissiongenerally consistent with stellar velocities, with averagevelocity difference ∆v ∼ − . However, the pos-sibility of a slow-moving disk wind, such as described inPontoppidan et al. (2011), cannot be fully ruled out asan explanation for the location of the absorbing CO gasin our targets.The best-fit CO/ CO ratio in our sample disksranges from ≈ CO/ CO ∼
70 (Sheffer et al. 2007) and theyoung stellar object environment value of ∼
100 foundby Smith et al. (2009), but the S/N and spectral resolu-tion are not optimal for making a precise determinationof the isotopic ratio in the disk.
CO Temperature and Density
We now compare our CO absorption temperatures andcolumn densities to other gas measurements of the innerdisk in the literature. The rotational excitation temper-atures of our disks (300 – 700 K) agree well with theanalyses of UV fluorescent CO emission lines (460 ± ∼ ∼
400 K. The difference is only marginallysignificant, although suggestive, due to the large un-certainties on the upper limit of the UV CO temper-ature distribution (see § . ( N (CO))= 18 .
6, which are both higher than ourparameter values for the same target. Horne et al. (2012)find a CO column density for AA Tau of 1.2 × cm − ,also higher than our value. We compare the Salyk et al.(2011) and Horne et al. (2012) CO parameters for AATau to our fit by using their column density and temper-ature values in our model and plotting them with oursin Figure 6. We assume a CO/ CO ratio of 70, our b -value, and our velocity shift value for the other models.The other models do not appear to be consistent withour models or data. Because emission line studies probea much larger region of the disk than the single line-of-sight of absorption line studies, it is not surprising thatthe Salyk et al. (2011) study appears to probe a differentmolecular gas population. The higher temperature gasmay be located at a disk height larger than the line-of-sight sampled in our absorption line measurements. TheUV spectra may also be preferentially sensitive to lowercolumn density gas due to extinction effects. However,the apparent inconsistency between our model and theHorne et al. (2012) model for AA Tau is surprising be-cause they are both absorption line studies. It may bethat the near-IR continuum emission and the UV con-tinuum emission are produced in different locations, sothat our lines-of-sight to them are different. Alterna-tively, if the CO is located in the inner disk, then thedifference may be related to the warp of the inner disk,which varies with the rotation of the star (Bouvier et al.2007). The temperature fits to our data show that it isinconsistent with a temperature of 2500 K, the H emis- Å )0.00.51.01.5 N o r m a li ze d F l ux DataOur ModelHorne et al. 2012Salyk et al. 2011
Fig. 6.—
We use the temperature and column density valuesof Salyk et al. (2011) and Horne et al. (2012) for AA Tau in ourmodel and assume CO/ CO ∼
70 to compare their CO gasparameters to ours. For the comparison, we use the (4 - 0) band ofAA Tau. The data is in black, our fit is in orange, the Salyk et al.(2011) model is in blue, and the Horne et al. (2012) model is inred. sion/absorption line temperature found by France et al.(2012a,b) in these disks, and the H fluorescence tem-perature for TW Hya, modeled by Herczeg et al. (2004)and Nomura & Millar (2005). Assuming that both theCO and H are in local thermodynamic equilibrium, thehot H emission appears to be spatially separate fromthe CO absorption.Assuming that our CO rotational temperatures arerepresentative of gas kinetic temperatures T gas >
300 K,we estimate the maximum radius where warm CO canbe maintained at this temperature (from Figure C.2 ofWoitke et al. 2011) to be r CO ≈
10 AU. This is consistentwith the UV CO emission studied by Schindhelm et al.(2012) who find that the emission arises from radii & r CO ≥ n CO = N ( CO ) r CO (1)The estimated CO number densities are in the range of n CO ∼ − − , and are shown in Table 4. Withthese densities, we computed the CO/H ratio requiredfor the CO to be thermalized up to J = 25 (the highest J -state that could be reliably identified in our data). Ifwe assume that collisions with H are the leading con-tributor to the CO level populations, and that the H density is sufficiently high, then the kinetic temperaturewill equal the CO rotational temperature, T rot ( CO ). Sig-nificant photoexcitiation would decouple the two temper-atures. We computed the H critical density (the den-sity where the spontaneous emission rate equals the colli-sional de-excitation rate) assuming an H ortho/para ra-tio of 3 and using collision rates calculated by summingover all collisional routes (Yang et al. 2010) downwardout of level J = 25. The H critical density for CO ex-citation in the range 300 – 750 K is ≈ (5.3 – 4.2) × cm − . The CO/H ratios needed to thermalize the ab-sorption lines are listed in Table 4. They are upper limitsand range from 1 . × − − . × − . For the CO den-sities derived with the assumed Woitke maximum radius,0 McJunkin et al. TABLE 4Protoplanetary Disk Warm Gas Parameters
Object n CO CO/H
2a 12
CO/ COcm − AA Tau 5 . +28 . − . × < . × − . +373 . − . DE Tau 1 . +4 . − . × < . × − . +778 . − . HN Tau 6 . +26 . − . × < . × − . +93 . − . RECX-15 4 . +12 . − . × < . × − . +113 . − . RW Aur 5 . +21 . − . × < . × − . +635 . − . SU Aur 1 . +6 . − . × < . × − . +2432 . − . CO/H ratio upper limit. For CO/H ratio less than this value, theabsorption lines are thermalized up to J = 25. typical interstellar translucent and dense cloud CO/H ratios (10 − – 10 − ; Lacy et al. 1994; Burgh et al. 2007)are sufficient to maintain a thermal distribution for theobserved CO absorption lines in five of the targets. Al-though the derived CO densities for most of our targetssuggest that the CO is thermalized, the sample is alsoconsistent with sub-thermal excitation, within the un-certainties. Sub-thermally populated lines would lead tothe CO temperatures being underestimated. If the ab-sorbing CO population is in thermal equilibrium, thenwe conclude that the observed CO/H ratios derived byFrance et al. (2011a, 2012a) are not representative of the local CO/H ratios in the warm molecular disk surface.We compare our derived temperatures and approxi-mated CO densities (which we change to a molecular hy-drogen density with an assumed CO/H ratio of 10 − )to the Woitke et al. (2011) model of RECX-15 as a checkon the vertical disk structure. Continuing our assump-tion that collisions with H are the leading contributor tothe CO level populations, the averaged hydrogen densitywill be dominated by H and we can directly comparethe plots in Figure C.2 of Woitke et al. (2011) with ourtemperatures and densities assuming the CO radius to be r CO ≈
10 AU as above. For each target’s temperatureand density pair, the height of the gas determined fromthe model are consistent with each other. These heightsrange between z/r ∼ . ∼ . Comparison of CO with System Parameters
Using the scale height model from § .
2, we calculatethat only inclinations of > ◦ will intercept the A V = 1surface of the disk, which is larger than the inclinationsin Table 1. At the inclinations of the targets, the sight-lines are not probing into the depth of the disk wherethe visual extinction exceeds unity, which indicates thatour CO is likely well above the A V = 1 disk surface.However, using a dimensionless height of z/r = 0 .
6, thedisk would be intercepted for inclinations & ◦ . Thisincreases our estimate on the unconstrained HN Tau in-clination. The lower limit on inclinations of & ◦ is com-patible with most of our disk inclinations, with the no-table exception of DE Tau. The inclination of DE Tau istoo small for us to be observing the gas in absorption, yet Fig. 7.—
COS data for DF Tau with the model fit of AA Tauoverplotted, showing the absence of CO absorption in the DF Tauspectrum (DF Tau data in black and best-fit model for AA Tauin orange). The CO bandheads are marked to show the twodifferent CO species. we clearly see the absorption bands in the data. DE Taumay be a candidate for a protoplanetary disk with a slow-moving molecular disk wind with a more face-on inclina-tion and blue-shifted lines (Pontoppidan et al. 2011). DFTau, another target in the France et al. (2012b) sampleknown to have CO (Najita et al. 2003; Salyk et al. 2011;Schindhelm et al. 2012) and thought to have a high incli-nation (85 ◦ ; Johns-Krull & Valenti 2001), does not showUV CO absorption, which is surprising. DF Tau has ananomalously large radius for a 10 Myr star (3.37 R ⊙ ,Gullbring et al. 1998), which leads to the large inclina-tion value. The spectrum for DF Tau is shown in Fig-ure 7 with the fit for AA Tau overplotted to illustratethe non-detection. A revised determination of the incli-nation of DE Tau and DF Tau, and possibly the othersources in the sample, would be useful. With notable ex-ceptions, for our mid- to high-inclination disks known tocontain CO, we see absorption if the gas is located at aheight of z/r & .
6, which is compatible with our derivedtemperatures and estimated densities. A cartoon of thisgeometry is shown in Figure 8.In Figure 9, we compare the column density and rota-tional temperature of CO with the mass accretion rateand disk inclination. We do not see any strong correla-tions between these parameters, though there may be aslight decrease in temperature with increasing disk incli-nation. At higher inclinations, we are observing the diskmore edge-on so that the line-of-sight passes through thecolder, denser material lower in the disk atmosphere. CONCLUSIONS
We present model fits to ultraviolet CO absorptionlines in six protoplanetary systems. We find results inAA Tau and HN Tau that are consistent with previ-ous work and extend our analysis to four new targetsobtained with
HST -COS far-UV G130M and G160Mmodes. We find CO rotational temperatures in the range300 - 700 K, which agree well with UV CO fluorescencerotational temperatures. However, our temperatures arecooler than inner disk CO temperatures obtained fromthe analyses of IR CO emission spectra and warmer thanthe Rettig et al. (2006) CO temperatures from IR COabsorption. The IR emission data may be probing a dif-O Absorption in Protoplanetary Disks 11
Fig. 8.—
A schematic representation of the line-of-sight geome-try for the inner region of T Tauri star disks. The red layer is ob-served as low-density, hot (T ∼ emission and absorption(Herczeg et al. 2004; Yang et al. 2011; France et al. 2012a); thegreen material is observed as intermediate density, warm (T ∼ Fig. 9.—
Correlations of the target parameters and best-fit pa-rameters. ferent gas population at smaller disk radii, while the IRabsorption data is likely probing gas at larger radii. Themeasured velocities of the CO absorbing gas rule out anorigin in a fast-moving disk wind. Our derived temper-atures and approximated densities of the gas are con-sistent with models for disk heights of z/r ∼ .
6. ThisCO location constrains the inclination of our disks tobe & ◦ in order to intercept the absorbing gas in themodel.We note that the present analysis is roughly at thelimit of what is feasible with the current generation of HST instrumentation. Higher spectral resolution wouldimprove our molecular parameter determination signifi-cantly. Unfortunately, observations at these flux levelswith
HST -STIS E140M mode are not feasible. A newobservational capability will be required to derive morerobust disk parameters from CO absorption line obser-vations of CTTSs. However, future observations withthe new COS G130M λ absorption in these disks at wave-lengths around 1100 ˚A. These observations would allowus to directly determine the value of the CO/H ratio inmoderate-to-high inclination protoplanetary disks.We thank M. Eidelsberg for providing the CO oscilla-tor strengths and rovibrational line wavelengths used inthis work. This work made use of data from HST guestobserving program 11616 and was supported by NASAgrants NNX08AC146 and NAS5-98043 to the Universityof Colorado at Boulder. REFERENCESAkeson, R. L., Ciardi, D. R., van Belle, G. T., & Creech-Eakman,M. J. 2002, ApJ, 566, 1124Andrews, S. M. & Williams, J. P. 2007, ApJ, 659, 705Armitage, P. J., Clarke, C. J., & Palla, F. 2003, MNRAS, 342,1139Bary, J. S., Weintraub, D. A., & Kastner, J. H. 2003, ApJ, 586,1136Bary, J. S., Weintraub, D. A., Shukla, S. J., Leisenring, J. M., &Kastner, J. H. 2008, ApJ, 678, 1088Bast, J. E., Brown, J. M., Herczeg, G. J., van Dishoeck, E. F., &Pontoppidan, K. M. 2011, A&A, 527, A119Bergin, E., Calvet, N., Sitko, M. L., Abgrall, H., D’Alessio, P.,Herczeg, G. J., Roueff, E., Qi, C., Lynch, D. K., Russell, R. W.,Brafford, S. M., & Perry, R. B. 2004, ApJ, 614, L133Bertout, C., Basri, G., & Bouvier, J. 1988, ApJ, 330, 350Bertout, C., Robichon, N., & Arenou, F. 1999, A&A, 352, 574Bitner, M. A., Richter, M. J., Lacy, J. H., Herczeg, G. J.,Greathouse, T. K., Jaffe, D. T., Salyk, C., Blake, G. A.,Hollenbach, D. J., Doppmann, G. W., Najita, J. R., & Currie,T. 2008, ApJ, 688, 1326Bouvier, J., Alencar, S. H. P., Boutelier, T., Dougados, C., Balog,Z., Grankin, K., Hodgkin, S. T., Ibrahimov, M. A., Kun, M.,Magakian, T. Y., & Pinte, C. 2007, A&A, 463, 1017 Bouvier, J., Grankin, K. N., Alencar, S. H. P., Dougados, C.,Fern´andez, M., Basri, G., Batalha, C., Guenther, E., Ibrahimov,M. A., Magakian, T. Y., Melnikov, S. Y., Petrov, P. P., Rud,M. V., & Zapatero Osorio, M. R. 2003, A&A, 409, 169Burgh, E. B., France, K., & McCandliss, S. R. 2007, ApJ, 658, 446Calvet, N. & Gullbring, E. 1998, ApJ, 509, 802Carmona, A., van den Ancker, M. E., Henning, T.,Pavlyuchenkov, Y., Dullemond, C. P., Goto, M., Thi, W. F.,Bouwman, J., & Waters, L. B. F. M. 2008, A&A, 477, 839Carpenter, K. G., Robinson, R. D., Wahlgren, G. M., Linsky,J. L., & Brown, A. 1994, ApJ, 428, 329Carr, J. S., Tokunaga, A. T., Najita, J., Shu, F. H., & Glassgold,A. E. 1993, ApJ, 411, L37Danforth, C. W., Keeney, B. A., Stocke, J. T., Shull, J. M., &Yao, Y. 2010, ApJ, 720, 976Donati, J.-F., Skelly, M. B., Bouvier, J., Gregory, S. G., Grankin,K. N., Jardine, M. M., Hussain, G. A. J., M´enard, F.,Dougados, C., Unruh, Y., Mohanty, S., Auri`ere, M., Morin, J.,Far`es, R., & MAPP Collaboration. 2010, MNRAS, 409, 1347Durrance, S. T. 1981, J. Geophys. Res., 86, 9115Eidelsberg, M., Jolly, A., Lemaire, J. L., Tchang-Brillet, W.- ¨U.,Breton, J., & Rostas, F. 1999, A&A, 346, 705Eidelsberg, M. & Rostas, F. 2003, ApJS, 145, 89Eisner, J. A., Hillenbrand, L. A., White, R. J., Bloom, J. S.,Akeson, R. L., & Blake, C. H. 2007, ApJ, 669, 10722 McJunkin et al.