A CO-to-H_2 ratio of \approx 10^{-5} towards the Herbig Ae star HK Ori
P. Wilson Cauley, Kevin France, Gregory J. Herzceg, Christopher M. Johns-Krull
DDraft version February 26, 2021
Typeset using L A TEX twocolumn style in AASTeX63
A CO-to-H ratio of ≈ − towards the Herbig Ae star HK Ori P. Wilson Cauley, Kevin France, Gregory J. Herzceg, and Christopher M. Johns-Krull Laboratory for Atmospheric and Space Physics, University of Colorado Boulder, 600 UCB, Boulder, CO 80303 Laboratory for Atmospheric and Space Physics, University of Colorado Boulder, 600 UCB Boulder, CO 80303 Kavli Institute for Astronomy and Astrophysics, Peking University, Yiheyuan Lu 5, Haidian Qu, 100871 Beijing, People’s Republic ofChina Physics & Astronomy Department, Rice University, 6100 Main Street, Houston, TX 77005 (Received 12/18/2020; Accepted 02/24/2021)
Submitted to AAS journalsABSTRACTMeasurements of gas mass in protoplanetary gas disks form the basis for estimating the conditionsof planet formation. Among the most important constraints derived from disk diagnostics are theabundances of gas-phase species critical for understanding disk chemistry. Towards this end, we presentdirect line-of-sight measurements of H and CO, employing UV absorption spectroscopy from HST -COS to characterize disk composition, molecular excitation temperatures, and spatial distributionin the circumstellar material around the Herbig Ae stars HK Ori and T Ori. We observe strongCO (N(CO) = 10 . cm − ; T rot (CO) = 19 K) and H (N(H ) = 10 . cm − ; T rot (H ) = 141K) absorption towards HK Ori with a CO/H ratio ( ≡ N(CO)/N(H )) = 1.3 +1 . − . × − . Thesemeasurements place direct empirical constraints on the CO-to-H conversion factor in the disk arounda Herbig Ae star for the first time, although there is uncertainty concerning the exact viewing geometryof the disk. The spectra of T Ori show CO (N(CO) = 10 . cm − ; T rot (CO) = 124 K) absorption.Interestingly, we do not detect any H absorption towards this star (N(H ) < . cm − ). We discussa potential scenario for the detection of CO without H , which deserves further investigation. The lowabundance ratio measured around HK Ori suggests significant depletion of CO in the circumstellargas, which conforms with the handful of other recent CO abundance measurements in protoplanetarydisks. INTRODUCTIONIn the inner disks ( r <
10 AU) around young stars,molecular gas emission and absorption provide our bestmeans of inferring the physical conditions at the radiiwhere gas giant and rocky planet cores are formingand accreting their nascent atmospheres. The abun-dances of gas-phase species at the location of formationand accretion are expected to affect the chemical abun-dances of exoplanet atmospheres ( ¨Oberg et al. 2011).Recent surveys of molecular emission from Classical TTauri Stars (CTTSs) and Herbig Ae stars have pro-vided new constraints on the radial distribution, temper-ature, and composition of planet-forming disks. Surveysof mid-IR emission from CO (Salyk et al. 2009; Ban-zatti & Pontoppidan 2015), H O and organic molecules(Pontoppidan et al. 2010; Carr & Najita 2011), andspectrally/spatially-resolved near-IR observations (Pon-
Corresponding author: P. Wilson [email protected] toppidan et al. 2011; Brittain et al. 2015) have placedconstraints on the CO and H O inventory at planet-forming radii and the evolution of the inner moleculargas disk radius (Antonellini et al. 2020). All of thesestudies require an accurate knowledge of the absoluteabundances to convert these measurements into localgas masses.An emerging consensus from ALMA observations ofprotoplanetary disks is that on large scales ( ∼ − (e.g., Miotelloet al. 2017; Long et al. 2017) This depletion process maybe fairly rapid, occurring within the first ≈ stellar disk’s lifetime (Zhang et al. 2020). Wheninterpreted with physical-chemical models of disks thatinclude CO freeze-out and photodissociation (Miotelloet al. 2016), the CO line fluxes are much weaker thanexpected from dust emission and the few available de-tections of HD (Bergin et al. 2013; McClure et al. 2016),implying that something other than freeze-out is deplet-ing the CO. The most plausible explanation is that com-plex C-bearing molecules also freeze-out in the disk mid-plane, although chemical reprocessing of CO in the cold a r X i v : . [ a s t r o - ph . S R ] F e b disk mid-plane is also possible for CTTS disks whichare older than ≈ ratio for a pre-main sequence star (France et al. 2014a).UV spectroscopy cannot probe the disk mid-plane, butcurrent models predict that carbon should rapidly getsequestered in solids in the disk mid-plane at all radiiand not return to the gas phase. In the inner disk ( r (cid:46) in the warm molecular layer ac-counts for the vast majority of the carbon and hydrogenrespectively ( ´Ad´amkovics et al. 2014), which suggeststhat CO/H measured through this region of the disk isa good approximation of the true CO/H value.The focus of this work is the circumstellar gas aroundthe Herbig Ae stars T Ori and HK Ori. Both T Oriand HK Ori are well studied pre-main sequence sys-tems. T Ori is a UX Orionis (UXOR) variable star whichhosts a self-shadowed circumstellar disk (Hillenbrandet al. 1992; Hein Bertelsen et al. 2016) and has an esti-mated accretion rate of log( ˙ M ) = − . ± . M (cid:12) yr − (Mendigut´ıa et al. 2011a, however, see Section 4). Theleading explanation for the UXOR phenomenon is theoccultation of the central star by dust in the warped orpuffed up inner disk rim (Dullemond et al. 2003; Kreplinet al. 2016). This requires that the disk be viewed nearlyedge-on ( i (cid:38) v sin i ≈ −
175 km s − (Mora et al. 2001; Alecian et al. 2013; Cauley et al. 2015)implying that the star is viewed equatorially rather thanpole-on. HK Ori is also actively accreting from its circumstellardisk (Hillenbrand et al. 1992; Mendigut´ıa et al. 2011a)with an accretion rate of log( ˙ M ) = –5 . ± . M (cid:12) yr − and has a K4 binary companion at ≈
150 AU, or ≈ (cid:48)(cid:48) . II II equivalent width. Cauley et al. (2015) derived aprojected rotational velocity of v sin i = 20 km s − forHK Ori which is also indicative of a low inclination an-gle for the system. However, the spectrum of HK Ori iscontaminated by a plethora of emission lines and thereare few clean photospheric absorption lines which canbe used to fit spectral templates. Thus the value fromCauley et al. (2015) should be viewed with caution. Inaddition, HK Ori shows significant spectroscopic andphotometric variability (Eiroa et al. 2002; Baines et al.2004; Mendigut´ıa et al. 2011b) similar to the UXORphenomenon although with smaller amplitude. Finally,HK Ori on average shows a double-peaked H α line pro-file with a strong central absorption feature (Reipurthet al. 1996; Mendigut´ıa et al. 2011b; Cauley et al. 2015).This profile shape was shown to be qualitatively repro-duced by accreting HAe stars viewed at high ( (cid:38) ◦ )inclination (Muzerolle et al. 2004). The H α line profilevariability also agrees well with the UXOR obscurationscenario tested by Muzerolle et al. (2004) for UX Ori.Overall, the evidence for a disk with a high inclinationangle (i.e., edge-on disk) is stronger than for a low incli-nation angle.In this paper we present new and archival HubbleSpace Telescope
FUV data for T Ori and HK Ori. Weuse the UV spectra to measure CO and H column den-sities via pencil beam absorption through the circum-stellar material and then compare the derived columndensities to place constraints on the CO/H ratios in thegas. Section 2 describes the UV spectroscopic data setsand Section 3 describes the spectral synthesis modelingused to derive molecular parameters. Section 4 placesthese results in context of current disk and interstellarsight-line studies and a brief summary of our findings ispresented in Section 5. OBSERVATIONS AND DATA REDUCTIONThe data were collected in 2012 and 2019 as part ofthe HST Guest Observing Programs 12996 (P.I. Johns-Krull) and 15070 (P.I. France). Both HK Ori and T Oriwere observed with two different Cosmic Origins Spec-trograph (COS) settings in each program. The COSG130M λ v = 0 Lyman band transitions (Abgrall& Roueff 1989). We note that the G130M λ lines. Thus the H anal-ysis utilizes only the Program 15070 data. The G130Mobservations have a resolving power of R ≈ , v = 19 km s − . HK Ori wasobserved for a total of t = 7758 seconds in this mode,while T Ori was observed for t = 10089 seconds.The COS G160M λ − A-X system transitions from the 0 ≤ v (cid:48) ≤ R ≈ , v = 17 kms − . HK Ori was observed for a total of t = 1376 sec-onds and T Ori was observed for t = 1696 seconds. AllCOS exposures were reduced with the standard calCOS pipeline.Quality G160M spectra are available for T Ori fromboth 2012 and 2019 through both programs. The 2019spectrum taken through Program 15070 has approxi-mately double the amount of flux compared with the2012 spectrum and the CO absorption is weaker by afactor of ≈
2. Interestingly, the flux ratio is roughlyconstant at ≈ (cid:38) ≈ (cid:46) in T Ori’s spectrum (see Sec-tion 3.1) prohibits any useful constraints on CO/H inthe system. Thus the bias introduced does not influenceany of our main conclusions concerning the circumstel-lar material around T Ori. We provide a more detaileddiscussion of T Ori’s FUV flux variability and how itmight be related to the variable CO absorption and lackof H absorption in Section 4.2.Figure 1 show the final co-added spectra surround-ing the H Lyman band transitions of interest (Abgrall& Roueff 1989) and Figure 2 shows the final co-addedspectra bracketing the CO A − X band transitions. Thespectra are smoothed with a 5-pixel boxcar. The UVspectra of many Herbig Ae/Be stars are littered withemission lines (e.g., Grady et al. 1996; Bouret & Catala1998; Cauley et al. 2016), which are most clearly seen inHK Ori’s spectrum. In order to isolate the H and COabsorption, we manually fit a high order spline, com-bined with a first or second degree polynomial, to thecontaminating emission lines to estimate the “contin-uum” emission. The continuum fits in the 1090 − − A − X absorption bands, we compared the observed spectrato PHOENIX stellar photosphere models (Husser et al.2013) to aid in identifying which features are stellar ver-sus circumstellar. Scaled and reddened PHOENIX mod- els are over-plotted in red in Figure 2. Note that HK Oriis a binary system with an A-type primary and G-type TTauri secondary with a separation of ≈
150 AU (Lein-ert et al. 1997; Smith et al. 2005). We only show theA-type photosphere model in Figure 2 due to the muchweaker flux of the G-type photosphere in the FUV. Itis clear that the photosphere of HK Ori is dwarfed bythe emission lines so the photosphere is relatively unim-portant in this case. For T Ori, however, the spectrumfrom 1400 − MODELING THE H AND CO ABSORPTIONCalculating the CO/H ratio in the warm molecu-lar layer of a disk requires a measurement of the co-spatial CO and H densities. To accomplish this wemodel the column densities of H and CO independently.The temperature of each gas is also derived indepen-dently. For the CO models, the rotational temperatureis a model parameter; the H rotational temperature isderived from the best-fit column densities in a separatefitting procedure. All model spectra are convolved withthe COS linespread-function (LSF) (Kriss 2011), whichvaries as a function of wavelength.3.1. H models The v (cid:48)(cid:48) = 0 , J (cid:48)(cid:48) H absorption lines are fit using the h2ools suite of optical depth templates from McCan-dliss (2003), which contain all of the Lyman and Wernertransitions from the ground vibrational level of the elec-tronic ground state (Abgrall & Roueff 1989). Each tem-plate spectrum is calculated for an integer value of theDoppler parameter b , where b = (2 kT rot /m H2 + v ) / ,and assuming a column density of N = 10 cm − . Thefree parameters in the fit are the line-of-sight radial ve-locity of the absorbing gas v rad , the coverage fraction ofthe stellar disk by the gas f cov , the Doppler broadeningparameter b , and the column densities of the rotationaltransitions J (cid:48)(cid:48) = 0 −
8. We note that although the ve-locity resolution of the observations is ≈
17 km s − ,the value of b can be constrained at much higher preci-sion since it has a strong effect on the optical depth ofthe unsaturated H transitions. Because the templatesfrom McCandliss (2003) are optical depth vectors, andthe optical depth is proportional to the column density,the templates can be adjusted for any column densityby multiplying by N/ . The absorption spectrum isthen computed as e − τ i where τ i is the value of the scaledoptical depth template at each wavelength. Rotationaltransitions J (cid:48)(cid:48) > F l u x den s i t y ( − e r g c m − s − ) ( − ) R ( )( − ) R ( ) ( − ) R ( ) ( − ) P ( )( − ) R ( ) ( − ) P ( )( − ) R ( ) ( − ) P ( )( − ) R ( ) ( − ) R ( )( − ) R ( )( − ) P ( )( − ) R ( ) ( − ) R ( ) ( − ) P ( )( − ) R ( ) C I . − . Å C I . − . Å C I . − . Å C I . − . Å T Ori1090 1100 1110 1120 1130Wavelength (Å)02468
Figure 1.
Extracted G130M spectra in the surrounding region of the H Lyman band. The observed spectra are shown in black andthe continuum fits in red. Transitions to v ‘ = 0 are marked in green and transitions to v ‘ = 1 are marked in orange. Transition labels for(1-0)P(1) at 1094.05 ˚A, (1-0)R(6) at 1109.86 ˚A, (0-0)P(1) at 1110.06 ˚A, (0-0)P(2) at 1112.50 ˚A, and (0-0)P(3) at 1115.90 ˚A are omitted forclarity. Clear absorption is visible in the HK Ori spectrum while the T Ori spectrum shows no sign of absorption by H . The absorptionfeatures in the T Ori spectrum are C I lines, most of which are likely accretion-related. We perform the model fits using our custom MCMCroutine based on the algorithm of Goodman & Weare(2010) (see also Foreman-Mackey et al. 2013). Uni-form priors are assumed for all parameters. We run theMCMC for 10,000 steps with 100 walkers, resulting in10 samplings of the posterior distribution. Because thetemplate spectra are calculated for discrete values of b we interpolate between the two templates which bracketthe model value of b . For example, if the MCMC itera-tion requires a spectrum with b = 3 . − we linearlyinterpolate between the b = 3 . − and b = 4 . − templates. Once the interpolation is complete theresulting spectrum is shifted by the radial velocity andscaled to find the best-fit column density. The modelspectrum, which is at the native velocity resolution, isconvolved with the COS LSF for comparison with theobserved spectrum. The median values of the marginal-ized posterior distributions are taken as the best-fit pa-rameter values and the 1 σ confidence intervals are the16 th and 84 th percentiles of the marginalized posteriors. For T Ori there is no visual evidence of H absorp-tion (see Figure 1 and Figure 5) and the spectrum from1090 − fitting procedure to TOri’s spectrum, with some exceptions, to measure theupper limit to T Ori’s H absorption properties. For TOri we fix the coverage fraction at f cov = 1 . v rad = 56 . − (Cauley & Johns-Krull2015). This is necessary because the lack of absorptionlines in the spectrum allows a total degeneracy between f cov and the H column densities. For the same reason v rad is entirely unconstrained.The marginalized posterior distributions from theMCMC are shown for HK Ori in Figure 3 and for TOri in Figure 4. Darker blue colors indicate regions ofhigher posterior density. The fixed parameters for T Ori f cov and v rad are excluded from Figure 4. We also ex-clude the posteriors for the rotational states J (cid:48)(cid:48) = 7 , DataPhotosphere model F l u x den s i t y ( − e r g c m − s − ) Figure 2.
Extracted G160 M spectra in the surrounding region of the CO Fourth Positive system. The stellar spectra are shown inblack and scaled photosphere models from PHOENIX are shown in red. The photospheric models are used as a guide to which featuresare stellar versus circumstellar. The wavelength segments containing the CO A − X bands are shaded in gray. for both objects due to severe blending of the J (cid:48)(cid:48) = 7state and no detectable features corresponding to the J (cid:48)(cid:48) = 8 state.The most likely model parameters and their 1 σ con-fidence intervals are given in Table 1 and the best-fitmodels are shown in Figure 5. As expected, we are onlyable to place an upper limit on the H column den-sity for T Ori due to the lack of any discernible H absorption in the spectrum. Also over-plotted for TOri in Figure 5 is the expected H absorption assumingthe canonical interstellar dense and translucent cloudratios (CO/H = 10 − and = 10 − ) and temperature T rot = 125 K where the temperature and CO columndensity are those derived in Section 3.2. A model withsuch a large H column density is strongly ruled out bythe MCMC posteriors. We will return to potential ex-planations for the lack of circumstellar H absorptionaround T Ori in Section 4.The most likely parameter values for HK Ori, onthe other hand, give a total H column density oflog (N[H ])= 20 . +0 . − . cm − . The broadening param-eter value is b = 4 . +0 . − . km s − suggesting a significant Table 1. H model fit parameters Parameter HK Ori T Ori(1) (2) (3) N ( J (cid:48)(cid:48) = 0) † (cm − ) 20.17 +0 . − . +1 . − . N ( J (cid:48)(cid:48) = 1) (cm − ) 19.80 +0 . − . < N ( J (cid:48)(cid:48) = 2) (cm − ) 18.85 +0 . − . < N ( J (cid:48)(cid:48) = 3) (cm − ) 17.44 +0 . − . < N ( J (cid:48)(cid:48) = 4) (cm − ) 16.75 +0 . − . < N ( J (cid:48)(cid:48) = 5) (cm − ) 16.12 +0 . − . < N ( J (cid:48)(cid:48) = 6) (cm − ) 11.71 +1 . − . < N ( J (cid:48)(cid:48) = 7) (cm − ) · · · · · · N ( J (cid:48)(cid:48) = 8) (cm − ) 12.58 +1 . − . < v rad (km s − ) -19.9 +0 . − . b (km s − ) 4.9 +0 . − . +3 . − . f cov +0 . − . T rot (K) 141 +6 − +649 − † All column densities N are log values. − . − . − . v rad (km s −1 )10.0011.5613.13 l og ( N J = ) . . . b (km s −1 ) . . . f cov . . . log(N J=0 ) . . . log(N J=1 ) . . . log(N J=2 ) . . . log(N J=3 ) . . . log(N J=4 ) . . . log(N J=5 ) . . . . log(N J=6 )15.0515.8216.60 l og ( N J = ) l og ( N J = ) l og ( N J = ) l og ( N J = ) l og ( N J = ) l og ( N J = ) f c o v b ( k m s − ) Figure 3.
Corner plot of the marginalized posterior distributions for the H MCMC of HK Ori. Note that all column densities N are inunits of cm − . contribution from turbulence or unresolved componentsto the line widths. The model also indicates that theH line-of-sight absorption covers the entire stellar diskgiven the value of f cov so close to 1.0.The temperature of the H is not a free parameter inthe spectrum fitting. Instead, we use the derived columndensities to find the most-likely rotational temperature.The low- J (cid:48)(cid:48) H level populations in the disk are expectedto be determined by collisions so we can calculate thetemperature of the gas assuming a Maxwell-Boltzmanndistribution of the rotational states (see France et al.2014a). We accomplish this by using the same MCMCroutine to fit a line to the H densities as a functionof the excitation temperature of the various rotationalstates. The rotational, or kinetic, temperature T rot ofthe gas is then the slope of the best-fit line.The most-likely linear fit is shown with a red line inFigure 6 along with one hundred models from a random sampling of the accepted posterior distributions (graylines). We find that the H gas around HK Ori has arotational temperature of T rot = 142 +6 − K. The small fituncertainties on the J (cid:48)(cid:48) = 0 , , absorption around T Ori, thecorresponding temperature fit is mostly unconstrained:the same procedure yields T rot = 452 +649 − K for T Ori.3.2.
CO models
The CO fitting procedure is outlined in detail inMcJunkin et al. (2013) (see also France et al. 2014a) andwe briefly summarize the salient points here. To derivethe circumstellar CO column densities we fit the COFourth Positive band system shaded in gray in Figure 2.We include the (4 − , (3 − , and(2 −
0) bands in the fitsfor both objects but exclude the (1 −
0) band for the HKOri fit due to contamination by what is most likely an . . . b (km s −1 )10.0011.8913.78 l og ( N J = ) . . . log(N J=0 ) . . . log(N J=1 ) . . . log(N J=2 ) . . . log(N J=3 ) . . . log(N J=4 ) . . . log(N J=5 ) . . . . log(N J=6 )10.0012.2214.45 l og ( N J = ) l og ( N J = ) l og ( N J = ) l og ( N J = ) l og ( N J = ) l og ( N J = ) Figure 4.
Corner plot of the marginalized posterior distributions for the H MCMC of T Ori. The parameters are almost entirelyunconstrained due to the lack of absorption in the observed spectrum. emission line from atomic iron. The synthetic CO spec-tra are generated using literature oscillator strengthsand ground-state energy levels (Haridass & Huber 1994;Eidelsberg et al. 1999; Eidelsberg & Rostas 2003). Thefree parameters in the MCMC fitting procedure are thecolumn densities of CO and CO, the radial velocityof the absorbing gas v rad , the rotational temperature ofthe gas T rot , and the Doppler-broadening parameter b .Similar to the case for H , b has a strong affect on theoptical depth of the CO transitions and is somewhat de-generate with the column density. This makes the valueof b sensitive to not only the line widths but the linedepths as well.We again adopt uniform, non-restrictive priors formost of the parameters and run the 100 walkers for10,000 steps each. The one exception is the Dopplerbroadening parameter b : we restrict the prior range to0 . < b ≤ . − . For each model iteration we generate the CO spectrum using the input T rot , b , andcolumn density values. The spectrum is then shifted invelocity space according to v rad and then convolved withthe COS LSF to produce the output model.The marginalized posterior distributions for the COMCMC fits are shown in Figure 7 and Figure 8. Thebest-fit models (red) and fifty random samplings of theaccepted posteriors (green) are shown in Figure 9. Themost noticeable difference in the CO spectra is the widthof the absorption bands, where HK Ori’s are fairly nar-row while T Ori’s are broad. This is manifested inthe large difference in derived T rot values: for HK Ori T rot = 19 +11 − K and for T Ori T rot = 124 +53 − K. Note thatT Ori’s b -value is just the median of the prior range, i.e.,it is unconstrained. This is a result of the high tempera-ture: as the temperature increases and the optical depthin the higher rotational states increases, the individualline broadening becomes less important and the instru- N o r m a li z ed f l u x HK Ori
Best−fit model1090 1095 1100 1105 1110 1115 1120Wavelength (Å)0.00.51.01.52.0 N o r m a li z ed f l u x T Ori
CO/H =10 −4 , T rot = 125 KCO/H =10 −5 , T rot = 125 K Figure 5.
Normalized
HST
COS spectra of HK Ori (top) and T Ori (bottom) from 1090 – 1120 ˚A. The best-fit H absorption modelsare over-plotted in red. No significant H absorption is detected in the spectrum of T Ori. The magenta model shows the expected H absorption assuming a conservative CO/H ratio of 10 − and the same temperature as that derived for the CO gas (see Section 3.2). Theblue model shows an assumed ratio of 10 − . mental profile makes differences in b negligible. This isnot the case for the low temperature fits for HK Ori. Atlow temperatures larger values of b increase the opticaldepth over a narrow range of rotational states which sig-nificantly increases the absorption depth. Thus there issome degeneracy between N and b at low T and b is con-strained even though it is much smaller than the instru-mental resolution. Neither spectrum shows any strongabsorption due to CO, in contrast to some CTTS disks(e.g. France et al. 2012).We note that the CO/ CO ratios indicated by ouranalysis are much larger than is typically found for theISM or the circumstellar environments of young stellarobjects (YSOs) (Smith et al. 2015). However, the COcolumn densities are highly uncertain and it’s likely thatthe CO absorption is almost entirely lost in the noisefor both objects. Thus we caution against applying thederived CO/ CO ratios to further work on these sys-tems. 3.3. H I Absorption
Table 2.
CO model fit parameters
Parameter HK Ori T Ori(1) (2) (3) N ( CO) † (cm − ) 15.5 +0 . − . +0 . − . N ( CO) (cm − ) 11.0 +2 . − . +3 . − . v rad (km s − ) − +4 . − . +6 . − . b (km s − ) 1.1 +0 . − . +1 . − . T rot (K) 19 +11 − +53 − † Column densities N are log values. Absorption line analyses of interstellar sight-lines of-ten rely on a knowledge of the total neutral hydrogencolumn density, N(H I ), to place molecular absorptionproperties in context. Toward this end, we use the 2012 HST -COS G130M observations of both stars to mea-sure N(H I ) by fitting the Ly α absorption profile. Weshow the observed profiles and reconstructed spectrafrom the model fits in Figure 10. Both HK Ori and T HK Ori exc (K)101214161820 l og ( N / g j ) T rot =141 +6−6 K HK Ori exc (K)101214161820 l og ( N / g j ) Figure 6.
Fit to the H excitation temperature versus columndensity for HK Ori. One hundred random samplings of the pos-terior distributions are shown in gray. The best-fit value for therotational excitation temperature is T rot = 141 +6 − K, indicative ofmaterial in a warm disk layer. .
76 1 .
10 1 . log(T) (K)5.008.1111.22 l og ( N [ C O ] ) ( c m − ) .
54 1 .
17 1 . b (km s −1 ) .
68 10 .
10 19 . v rad (km s −1 ) .
61 15 .
22 15 . log(N[ CO]) (cm −2 ) .
00 8 .
11 11 .
22 14 . log(N[ CO]) (cm −2 )14.6115.2215.83 l og ( N [ C O ] ) ( c m − ) v r ad ( k m s − ) b ( k m s − ) Figure 7.
Corner plot of the marginalized posterior distributionsfor the CO MCMC of HK Ori.
Ori display broad Ly α emission features consistent withaccretion-generated H I emission observed on lower massT Tauri stars (e.g., France et al. 2014b). In analogy tothe methodology employed for H and CO, we assumethe observed profile is the superposition of an under-lying stellar emission spectrum and the Ly α absorber.For HK Ori, we employed a broad Gaussian as the in-trinsic Ly α profile shape, and implemented an iterativefitting routine to find the best fit to the observed COSG130M Ly α spectra (France et al. 2012; McJunkin et al. .
55 1 .
91 2 . log(T) (K)5.008.0711.15 l og ( N [ C O ] ) ( c m − ) .
47 1 .
98 3 . b (km s −1 ) − . − . − . v rad (km s −1 ) .
47 14 .
82 15 . log(N[ CO]) (cm −2 ) .
00 8 .
07 11 .
15 14 . log(N[ CO]) (cm −2 )14.4714.8215.16 l og ( N [ C O ] ) ( c m − ) −64.76−51.27−37.77 v r ad ( k m s − ) b ( k m s − ) Figure 8.
Corner plot of the marginalized posterior distributionsfor the CO MCMC of T Ori. α emission peak is less pronounced in TOri, and here we fit a spline function to the continuumand emission line as a baseline for fitting the H I absorp-tion. The Doppler b -value, b HI , is assumed to be 10 kms − for both stars. For more details on reconstructingintrinsic Ly α profiles see McJunkin et al. (2014).For HK Ori, our fits find a baseline continuum of8 × − erg cm − s − ˚A − , Gaussian peak amplitudeof 6 × − erg cm − s − ˚A − , a Gaussian FWHM =1100 km s − , and log N(H I ) = 20.75 ± N(H I ) = 20.25 ± DISCUSSION4.1.
Comparing H , CO, and H I with the TranslucentISM We compared the absorption properties of the twotargets against typical properties of interstellar sight-lines to help solidify assignment of these componentsto a circumstellar disk origin. The HK Ori CO/H ra-tio and molecular fraction are consistent with interstel-lar clouds at the diffuse-to-translucent boundary (e.g.,Burgh et al. 2007). However, the CO and molecularfraction are considerably higher than expected for a starwith HK Ori’s optical reddening E ( B − V ) (= 0 .
12; as-suming the N(H I ) to reddening conversion from Diplas& Savage 1994). Furthermore, the H column density toHK Ori is orders of magnitude higher than what is seento typical hot star targets near the Orion star-forming The molecular fraction, f H is defined as f H = 2N(H )/(2N(H ) + N(H I )). CO(4−0) N o r m a li z ed f l u x HK Ori N o r m a li z ed f l u x T Ori
CO(3−0)
CO(2−0)
CO(1−0)
Figure 9.
Normalized spectra surrounding the CO (
A-X ) absorption bands for HK Ori (top row) and T Ori (bottom row). The best-fitCO absorption models are over-plotted in red and one hundred random draws from the accepted posterior chains are shown in green. Notethat the CO(1-0) band is excluded from the model determination for HK Ori due to the contamination by the strong emission line near1511 ˚A.
T Ori F l u x ( − e r g c m − s − Å − ) Observed Ly− α profile1205 1210 1215 1220 1225Wavelength (Å)0.00.20.40.60.81.01.21.4 N o r m a li z ed f l u x log N(HI)=20.25
Reconstructed Ly− α profileModeled Ly− α profile HK Ori F l u x ( − e r g c m − s − Å − ) N(HI)=20.75 0.00.51.01.52.02.53.0 F l u x ( − e r g c m − s − Å − ) Figure 10.
Observed (top panels) and reconstructed (bottompanels) Ly α profiles for T Ori and HK Ori. The best-fit neutralhydrogen ISM column densities are labeled in the bottom panels. region (Savage et al. 1977). This points to a circumstel-lar origin of the molecular gas in HK Ori.Further support for a circumstellar origin of the molec-ular gas around HK Ori comes from an analysis of theexcitation temperatures. While T(H ) = 141 K is con- sistent with an interstellar origin, T(CO) = 19 K is muchhigher than what is seen for translucent clouds (e.g., seeFigure 6 of Burgh et al. 2007). We note that the tem-perature discrepancy could be due to the density of CObeing below the critical density (see Section 4.4 for amore detailed explanation). Therefore, numerous linesof evidence suggest a circumstellar, as opposed to inter-stellar, origin for the molecular gas observed toward HKOri.T Ori’s hydrogen sightline is characteristic of hot starsnear the Orion star forming regions (N(H ) < cm − ; N(H I ) = 10 . − . cm − ; Savage et al. 1977).However, the diffuse sightlines in Orion are typically as-sociated with very low CO upper limits (N(CO) < cm − ; Federman et al. 1980). Similarly, the very lowmolecular fraction is incompatible with the large CO col-umn density, arguing that the CO we observe towardsT Ori is almost certainly of a circumstellar origin. Thisis corroborated by the high T(CO), which is a factorof ∼
10 – 20 higher than seen on translucent interstel-lar sight-lines (Burgh et al. 2007; Sheffer et al. 2008).The circumstellar nature of T Ori’s CO absorption isalso bolstered by the fact that it shows a variable col-umn density likely related to inner disk material (seeSection 4.2).4.2.
Where’s the H around T Ori? The lack of H absorption towards T Ori is surprisinggiven the system’s young age and the high likelihoodthat we observe the disk closer to an edge-on orientation.1 Table 3.
Total column densities and N(CO)/N(H ) ratios Object N(H ) N(CO) N(H I ) N(CO)/N(H ) f H (1) (2) (3) (4) (5) (6)HK Ori 20.34 +0 . − . +0 . − . ± +1 . − . × − < +1 . − . +0 . − . ± · · · < × − The properties of the CO gas we derive confirm that thegas is circumstellar ( T ≈
125 K) and not interstellar,where the gas typically has temperatures
T <
10 K.Furthermore, the radial velocity of the CO absorption, ≈
45 km s − , is consistent with, although mildly blue-shifted relative to, the system velocity of ≈
56 km s − (noting that HST -COS has a wavelength accuracy of 15km s − ).We can look to the variability between the 2012 and2019 observations for clues about the lack of H absorp-tion towards the star. In Figure 11 we show a compari-son between the HST
COS spectra from 2012 (red) and2019 (purple). The bottom panel shows the raw spectrafor wavelengths common to both epochs and the middlepanel shows the ratio of the fluxes. The top panel dis-plays zoomed-in normalized comparisons between strongspectral features. A few things are evident from Fig-ure 11: 1. The flux at wavelengths longer than ≈ ≈ × greater in 2019 than in 2012; 2. The fluxat wavelengths shorter than ≈ V -mag ASAS-SNlight curves (Shappee et al. 2014; Kochanek et al. 2017)of T Ori and interpolated the observed magnitudes ontoour HST
COS observation dates: for the 2012 observa-tions V ≈ . V ≈ . ≈ . V -band measurements confirm that the optical flux variesat a similar level to the FUV flux.We speculate that T Ori’s nature as a rapid rotatorand UXOR variable can qualitatively explain the phe-nomena seen in Figure 11. First, rapidly rotating starsare subject to non-negligible gravity darkening wherethe oblateness of the star causes the poles to be hot-ter than the equator (von Zeipel 1924; Espinosa Lara& Rieutord 2011). If the UXOR screen preferentiallyblocks equatorial stellar latitudes, or is at least moreoptically thick across these regions, the star will be dim-mer but the stellar spectrum will be weighted towardsthe hotter polar regions. This can account for the ob-served shallower absorption lines in the 2012 faint phase.Second, the increased CO absorption during the faintphase could be a result of the increased gas column den-sity produced by the occulting screen. Finally, the lack of H absorption in the Lyman transitions may be theresult of the sub-1300 ˚A flux, which is related to materialaccreting onto the star, being scattered into the line-of-sight over a large volume in the inner disk region. If theflux is the result of scattering over a large volume, weare not probing the same pencil beam through the diskthat produces the CO absorption since the flux from1400 - 1500 ˚A is largely photospheric in origin. Thisalso explains the relatively stable sub-1300 ˚A flux sincethe UXOR screen does not occult the entire scatteringsurface.We stress that the scenario outlined above is specula-tive; there are likely other plausible explanations for thevarious phenomena (e.g., an occulting accretion columnand related veiling of the stellar photosphere). In partic-ular, the proposed viewing geometry must be fine-tunedin order to preferentially occult equatorial latitudes onthe star, leaving the hotter poles visible. AdditionalFUV time-series spectra will be useful in determining amore precise view of the system geometry and we defera more quantitative characterization of the variability tofuture work.4.3. Large Doppler broadening parameters
The H and CO analysis for HK Ori suggest large( ≈ − − ) non-thermal broadening componentsare necessary to reproduce the observed H and COabsorption. This would imply significant contributionsfrom turbulence or spatially unresolved gas motions re-sponsible for the line broadening. The turbulence inter-pretation seems unlikely given that CO radio observa-tions suggest that turbulence in protoplanetary disks isweak at distances of ≈ s of AU (Hughes et al. 2011;Flaherty et al. 2018, 2020). On the other hand, the pen-cil beam H observations, if passing through the disk,likely probe higher surface layers than the thermal COemission.We caution against an over-interpretation of the largevalue for the Doppler parameter for two reasons. First,absorption features with widths of a few kilometers persecond are not resolved in our spectra. The large b -values are mainly a result of the models fitting the ab-sorption line depths, which are determined by both b and the column density N . Without resolving the linewidths of individual H and CO transitions it is difficultto say with confidence what is driving the large valuesof b . Second, there is considerable uncertainty around2 F l u x ( − e r g c m − s − Å − ) eff = 8800 K photosphere F l u x ( − e r g c m − s − Å − ) F / F F / F N o r m a li z ed f l u x N o r m a li z ed f l u x Figure 11.
Comparison of the 2012 (red) and 2019 (purple) FUV T Ori spectra. The bottom panel displays the flux spectra from bothepochs, as well as a photospheric template scaled to match the 2019 flux. The Lyman- α emission lines are interpolated over for clarity.Only wavelengths common to both sets of observations are shown. The middle panel shows the ratio of the 2012 spectrum to the 2019spectrum and the top panels show zoomed-in normalized spectrum comparisons between strong spectral features from each epoch. The CObandheads are marked with gray shaded regions and the wavelength ranges shown in the top panels are marked with light blue shading inthe middle panel. the viewing geometry for the disk around HK Ori. Moreprecise constraints on our viewing angle through thesecircumstellar environments could help clarify the inter-pretation of the large value for the Doppler parameter.4.4. Spatial Location and Distribution of MolecularAbsorption Components around HK Ori
We have measured significant columns of H and COtowards HK Ori. The presence of the K4 companionat 0 (cid:48)(cid:48) . ≈
30 AU. Given thelack of circumstellar material around the secondary, wecontinue under the assumption that the absorption iscreated by material occulting the primary star.The most critical assumption necessary for deriving aCO/H ratio in a disk via pencil beam absorption is thatthe CO and H populations are co-spatial, i.e., we aresampling the same parcels of the disk in both species.At first glance, the differences in the derived rotationaltemperatures between the H and CO populations sug-gest we might be sampling different gas components.However, while the CO rotational levels are populatedby collisions, radiative processes can alter the observed3distributions of J -levels. If the local number density islower than the CO critical density, these levels can ra-diatively depopulate fast enough that the populationsare not representative of the local kinetic temperature.Referencing the density and temperature structures ofHerbig Ae disks presented in Ag´undez et al. (2018), weestimate the local H number density in the region probedby our HK Ori H absorption spectra is n H ∼ − cm − . Assuming an H ortho–para ratio of 3 and thatthe T H ≈ T kin = 150 K, the total ortho- + para-H2collision rate, summed over all possible lower levels, isΓ T OT = 2.6 × − cm s − at J = 10 (Yang et al.2010). The CO critical density for J = 10 at 150 K is n H ≈ × cm − . Therefore, it is possible thatthe high-J CO populations that drive the determinationof T(CO) are significantly sub-thermal, which would ex-plain the temperature difference between the observedH and CO populations. To explore this possibility, wecomputed RADEX models (van der Tak et al. 2007)using the online interface . For our measured values ofT(H ), N(CO), b CO , and n H = 10 cm , we find that COrotational levels 3 ≤ J ≤
30 are characterized by temper-atures less than the H rotational temperature. Finally,sub-thermal CO rotational temperatures are almost al-ways observed in co-spatial CO and H populations inthe ISM (Burgh et al. 2007; Sheffer et al. 2008). Weconclude that while there is uncertainty about the exactspatial distribution of the gas and the viewing geome-try, it is reasonable to infer that the H and CO are co-spatial around HK Ori with CO/H ratio ≈ . × − .The system radial velocity for HK Ori is ≈ − − (Reipurth et al. 1996; Baines et al. 2004) whichmeans the best-fit radial velocities for both the H andCO absorption are blue-shifted from the system veloc-ity by ≈ −
40 km s − since we fit the spectrum inthe Earth’s rest frame. Again, it is worth noting thatthe absolute velocity precision of COS is ≈
15 km s − .Even taking into account the velocity precision of COS,the measured blue-shifts for the H and CO absorptionare large and inconsistent with a pencil beam observedthrough a Keplerian disk since material in a circulardisk along any line-of-sight should have velocities cen-tered around ≈ − . However, it’s possible that theabsorption arises at the base of a molecular disk windlaunched magnetocentrifugally from the disk’s upper at-mosphere. Evidence of such a wind has been detectedin CO emission from the Herbig Ae star HD 163296 byKlaassen et al. (2013) with characteristic velocities of ≈
18 km s − . Thus depending on the wind launchingangle and the disk inclination angle, blue-shifted veloc-ities of ≈
40 km s − are not unreasonable if the absorp-tion arises in a disk wind.Given the evidence for a disk or disk wind origin ofthe gas absorption towards HK Ori, our derived CO/H http://var.sron.nl/radex/radex.php ratio of ≈ . × − suggests significant depletion ofCO in the circumstellar material relative to the canon-ical molecular cloud value of 10 − (Bergin & Williams2017). Lower values of CO/H are expected for coolerdisks with mid-plane temperatures (cid:46) −
40 K whichis mainly driven by dust grain surface chemistry (Re-boussin et al. 2015). However, the depletion factor ofCO is highly dependent on age and location in the disk(e.g., Krijt et al. 2020) which makes it difficult to sayanything definitive about HK Ori’s disk based on ourderived ratio given the uncertainties around the line-of-sight probed by our observations. If our observationstrace the warm molecular layer of HK Ori’s disk, theCO/H ratio is line with the depleted values reportedfor other protoplanetary disks (e.g., Favre et al. 2013;McClure et al. 2016; Schwarz et al. 2016). CONCLUSIONWe presented new FUV observations from the
HubbleSpace Telescope of H and CO absorption towards theHerbig Ae stars HK Ori and T Ori. We find a signif-icant column of CO absorption and a notable lack ofH absorption towards T Ori, which we posit is relatedto variable occultation of the star by inner disk warp-ing. HK Ori shows blue-shifted absorption in both H and CO with derived temperatures and column densi-ties which place the gas firmly in the circumstellar en-vironment around HK Ori. Although there is consider-able uncertainty as to the exact line-of-sight traced byour observations, the derived CO/H value around HKOri is ≈ . × − suggesting a depleted reservoir ofCO. Spatially-resolved molecular emission maps of bothsystems will help place our pencil-beam observations incontext and provide a more precise understanding of theCO/H value derived for the gas around HK Ori. Acknowledgments:
We thank the referee for theirsuggestions, which helped improve the quality of themanuscript. P. W. C. acknowledges useful conversationswith Nicole Arulanantham about the molecular compo-nents of inner disks. The data in this paper were ob-tained through
HST
Guest Observing programs 12996and 15070. This work has made use of NASA’s Astro-physics Data System and the SIMBAD database, whichis operated at CDS, Strasbourg, France.4 REFERENCES
Abgrall, H., & Roueff, E. 1989, A&AS, 79, 313´Ad´amkovics, M., Glassgold, A. E., & Najita, J. R. 2014,ApJ, 786, 135, doi: 10.1088/0004-637X/786/2/135Ag´undez, M., Roueff, E., Le Petit, F., & Le Bourlot, J.2018, A&A, 616, A19, doi: 10.1051/0004-6361/201732518Alecian, E., Wade, G. A., Catala, C., et al. 2013, MNRAS,429, 1001, doi: 10.1093/mnras/sts383Antonellini, S., Banzatti, A., Kamp, I., Thi, W. F., &Woitke, P. 2020, A&A, 637, A29,doi: 10.1051/0004-6361/201834077Baines, D., Oudmaijer, R. D., Mora, A., et al. 2004,MNRAS, 353, 697, doi: 10.1111/j.1365-2966.2004.08104.xBanzatti, A., & Pontoppidan, K. M. 2015, ApJ, 809, 167,doi: 10.1088/0004-637X/809/2/167Bergin, E. A., & Williams, J. P. 2017, The Determinationof Protoplanetary Disk Masses, ed. M. Pessah &O. Gressel, Vol. 445, 1, doi: 10.1007/978-3-319-60609-5 1Bergin, E. A., Cleeves, L. I., Gorti, U., et al. 2013, Nature,493, 644, doi: 10.1038/nature11805Blondel, P. F. C., & Djie, H. R. E. T. A. 2006, A&A, 456,1045, doi: 10.1051/0004-6361:20040269Bosman, A. D., Walsh, C., & van Dishoeck, E. F. 2018,A&A, 618, A182, doi: 10.1051/0004-6361/201833497Bouret, J. C., & Catala, C. 1998, A&A, 340, 163Brittain, S. D., Najita, J. R., & Carr, J. S. 2015, Ap&SS,357, 54, doi: 10.1007/s10509-015-2260-4Burgh, E. B., France, K., & McCandliss, S. R. 2007, ApJ,658, 446, doi: 10.1086/511259Carr, J. S., & Najita, J. R. 2011, ApJ, 733, 102,doi: 10.1088/0004-637X/733/2/102Cauley, P. W., & Johns-Krull, C. M. 2015, ApJ, 810, 5,doi: 10.1088/0004-637X/810/1/5Cauley, P. W., Redfield, S., Jensen, A. G., & Barman, T.2016, AJ, 152, 20, doi: 10.3847/0004-6256/152/1/20Cauley, P. W., Redfield, S., Jensen, A. G., et al. 2015, ApJ,810, 13, doi: 10.1088/0004-637X/810/1/13Diplas, A., & Savage, B. D. 1994, ApJS, 93, 211,doi: 10.1086/192052Dullemond, C. P., van den Ancker, M. E., Acke, B., & vanBoekel, R. 2003, ApJL, 594, L47, doi: 10.1086/378400Eidelsberg, M., Jolly, A., Lemaire, J. L., et al. 1999, A&A,346, 705Eidelsberg, M., & Rostas, F. 2003, ApJS, 145, 89,doi: 10.1086/345596Eiroa, C., Oudmaijer, R. D., Davies, J. K., et al. 2002,A&A, 384, 1038, doi: 10.1051/0004-6361:20020096Espinosa Lara, F., & Rieutord, M. 2011, A&A, 533, A43,doi: 10.1051/0004-6361/201117252 Favre, C., Cleeves, L. I., Bergin, E. A., Qi, C., & Blake,G. A. 2013, ApJL, 776, L38,doi: 10.1088/2041-8205/776/2/L38Federman, S. R., Glassgold, A. E., Jenkins, E. B., & Shaya,E. J. 1980, ApJ, 242, 545, doi: 10.1086/158489Flaherty, K., Hughes, A. M., Simon, J. B., et al. 2020, ApJ,895, 109, doi: 10.3847/1538-4357/ab8cc5Flaherty, K. M., Hughes, A. M., Teague, R., et al. 2018,ApJ, 856, 117, doi: 10.3847/1538-4357/aab615Foreman-Mackey, D., Hogg, D. W., Lang, D., & Goodman,J. 2013, PASP, 125, 306, doi: 10.1086/670067France, K., Herczeg, G. J., McJunkin, M., & Penton, S. V.2014a, ApJ, 794, 160, doi: 10.1088/0004-637X/794/2/160France, K., Schindhelm, E., Bergin, E. A., Roueff, E., &Abgrall, H. 2014b, ApJ, 784, 127,doi: 10.1088/0004-637X/784/2/127France, K., Burgh, E. B., Herczeg, G. J., et al. 2012, ApJ,744, 22, doi: 10.1088/0004-637X/744/1/22Goodman, J., & Weare, J. 2010, Communications inApplied Mathematics and Computational Science, 5, 65,doi: 10.2140/camcos.2010.5.65Grady, C. A., Perez, M. R., Talavera, A., et al. 1996,A&AS, 120, 157Haridass, C., & Huber, K. P. 1994, ApJ, 420, 433,doi: 10.1086/173573Hein Bertelsen, R. P., Kamp, I., van der Plas, G., et al.2016, A&A, 590, A98, doi: 10.1051/0004-6361/201527652Hillenbrand, L. A., Strom, S. E., Vrba, F. J., & Keene, J.1992, ApJ, 397, 613, doi: 10.1086/171819Hughes, A. M., Wilner, D. J., Andrews, S. M., Qi, C., &Hogerheijde, M. R. 2011, ApJ, 727, 85,doi: 10.1088/0004-637X/727/2/85Husser, T. O., Wende-von Berg, S., Dreizler, S., et al. 2013,A&A, 553, A6, doi: 10.1051/0004-6361/201219058Kama, M., Bruderer, S., van Dishoeck, E. F., et al. 2016,A&A, 592, A83, doi: 10.1051/0004-6361/201526991Klaassen, P. D., Juhasz, A., Mathews, G. S., et al. 2013,A&A, 555, A73, doi: 10.1051/0004-6361/201321129Kochanek, C. S., Shappee, B. J., Stanek, K. Z., et al. 2017,PASP, 129, 104502, doi: 10.1088/1538-3873/aa80d9Kreplin, A., Madlener, D., Chen, L., et al. 2016, A&A, 590,A96, doi: 10.1051/0004-6361/201628281Krijt, S., Bosman, A. D., Zhang, K., et al. 2020, ApJ, 899,134, doi: 10.3847/1538-4357/aba75dKriss, G. A. 2011, Improved Medium Resolution LineSpread Functions for COS FUV Spectra, COSInstrument Science Report 2011-01(v1Leinert, C., Richichi, A., & Haas, M. 1997, A&A, 318, 4725