Probing UV-Sensitive Pathways for CN and HCN Formation in Protoplanetary Disks with the Hubble Space Telescope
Nicole Arulanantham, K. France, P. Cazzoletti, A. Miotello, C. F. Manara, P. C. Schneider, K. Hoadley, E. F. van Dishoeck, H. M. Günther
DDraft version February 24, 2020
Preprint typeset using L A TEX style emulateapj v. 12/16/11
PROBING UV-SENSITIVE PATHWAYS FOR CN & HCN FORMATION IN PROTOPLANETARY DISKS WITHTHE
HUBBLE SPACE TELESCOPE
Nicole Arulanantham
Laboratory for Atmospheric and Space Physics, University of Colorado, 392 UCB, Boulder, CO 80303, USA
K. France
Laboratory for Atmospheric and Space Physics, University of Colorado, 392 UCB, Boulder, CO 80303, USA
P. Cazzoletti
Max-Planck-Institute for Extraterrestrial Physics (MPE), Giessenbachstr. 1, 85748, Garching, Germany
A. Miotello
European Southern Observatory, Karl-Schwarzschild-Str. 2, D-85748 Garching bei M¨unchen, Germany
C. F. Manara
European Southern Observatory, Karl-Schwarzschild-Str. 2, D-85748 Garching bei M¨unchen, Germany
P. C. Schneider
Hamburger Sternwarte, Gojenbergsweg 112, 21029 Hamburg, Germany
K. Hoadley
Department of Astronomy, California Institute of Technology, 1200 East California Blvd., Pasadena, CA 91125, USA
E. F. van Dishoeck
Leiden Observatory, Leiden University, PO Box 9513, 2300 RA Leiden, The Netherlands andMax-Planck-Institute for Extraterrestrial Physics (MPE), Giessenbachstr. 1, 85748, Garching, Germany
H. M. G¨unther
MIT, Kavli Institute for Astrophysics and Space Research, 77 Massachusetts Ave., Cambridge, MA 02139, USA
Draft version February 24, 2020
ABSTRACTThe UV radiation field is a critical regulator of gas-phase chemistry in surface layers of disks aroundyoung stars. In an effort to understand the relationship between photocatalyzing UV radiation fieldsand gas emission observed at infrared and sub-mm wavelengths, we present an analysis of new andarchival
HST , Spitzer , ALMA, IRAM, and SMA data for five targets in the Lupus cloud complex and14 systems in Taurus-Auriga. The
HST spectra were used to measure Ly α and FUV continuum fluxesreaching the disk surface, which are responsible for dissociating relevant molecular species (e.g. HCN,N ). Semi-forbidden C II] λ emission were also measured to constraininner disk populations of C + and vibrationally excited H . We find a significant positive correlationbetween 14 µ m HCN emission and fluxes from the FUV continuum and C II] λ photodissociation and carbon ionization to trigger the main CN/HCNformation pathways. We also report significant negative correlations between sub-mm CN emissionand both C II] and FUV continuum fluxes, implying that CN is also more readily dissociated indisks with stronger FUV irradiation. No clear relationships are detected between either CN or HCNand Ly α or UV-H emission. This is attributed to the spatial stratification of the various molecularspecies, which span several vertical layers and radii across the inner and outer disk. We expect thatfuture observations with JWST will build on this work by enabling more sensitive IR surveys thanwere possible with
Spitzer . Subject headings: stars: pre-main sequence, protoplanetary disks, molecules INTRODUCTION
Multi-wavelength observations of gas- and dust-rich disks around young stars have allowed us to develop rudimentarymaps of the composition and structure of planet-forming material. Infrared surveys with
Spitzer ( ¨Oberg et al. 2008; a r X i v : . [ a s t r o - ph . S R ] F e b Arulanantham et al.Pontoppidan et al. 2010; Bottinelli et al. 2010; Salyk et al. 2011b; Pascucci et al. 2013) and
Herschel (Dent et al. 2013)provided important constraints on warm molecular gas in surface layers of the inner disks ( r <
10 au), contributingcolumn densities and temperatures of critical molecular gas species (e.g. H O, CO ). Sub-mm observations of star-forming regions with ALMA have revealed the structure of cold gas in the outer disks with unprecedented sensitivityand angular resolution (Ansdell et al. 2016, 2017; Barenfeld et al. 2016; Pascucci et al. 2016; Miotello et al. 2017;Tazzari et al. 2017; Long et al. 2018, 2019; van Terwisga et al. 2019; Cazzoletti et al. 2019; Williams et al. 2019),showing statistically significant trends in mass and radial extent as a function of cluster age and initial conditions ofthe parent cloud (e.g. angular momentum, temperature; Cazzoletti et al. 2019). Both observational and theoreticalwork demonstrate that the chemical evolution of molecular gas is strongly dependent on the ultraviolet radiation fieldreaching the surface of the disk (Aikawa et al. 2002; Chapillon et al. 2012; Walsh et al. 2012, 2015; Cazzoletti et al.2018; Cleeves et al. 2018; Ag´undez et al. 2018; Bergner et al. 2019; Miotello et al. 2019). However, the effects of diskgeometry and varying UV flux are degenerate in physical-chemical models of the gas distributions (see e.g. Cazzolettiet al. 2018), making it difficult to trace the precise locations of critical species within the disk in the absence of highangular resolution datasets.Observational constraints on the UV flux reaching the disk surface are available from HST surveys of young starswith circumstellar disks (see e.g. Yang et al. 2012; France et al. 2012, 2014). The wavelength range available with theCosmic Origins Spectrograph (
HST -COS) includes Ly α emission (cid:0) λ = 1215 .
67 ˚A (cid:1) and a portion of the FUV continuum (cid:0) λ > (cid:1) , providing estimates of key photochemical ingredients in the molecular gas disk (Bergin et al. 2003; Liet al. 2013). In addition to these direct tracers of the UV radiation field, emission lines from electronic transitions ofH are also detected in HST -COS and
HST -STIS spectra (Herczeg et al. 2002; France et al. 2012). These featuresoriginate in surface layers close to the star (Hoadley et al. 2015), providing an independent way to estimate the UVflux reaching the innermost regions of the disk (Herczeg et al. 2004; Schindhelm et al. 2012b).In this work, we present new and archival
HST -COS and
HST -STIS observations of a sample of five young systemsin the ∼ Spitzer , including disks in Taurus-Aurigafor comparison. We focus on these two molecules because of the strong dependence of their physical distributionson the UV radation field (Ag´undez et al. 2008, 2018; Cazzoletti et al. 2018; Bergner et al. 2019; Greenwood et al.2019). Our dataset therefore allows us to observationally examine the theoretical relationships between moleculargas emission and UV radiation fields, in turn constraining physical-chemical models that map abundances of volatileelements (C, N, and H). This information may be particularly useful in disk regions where gas-phase oxygen is depletedand emission from more abundant species (e.g. CO) is fainter than expected (Miotello et al. 2017; Schwarz et al. 2018).The radial distributions of these molecules can then inform us about the composition of material available for in situprotoplanetary accretion, setting important initial conditions for atmospheric chemistry (see e.g. Madhusudhan et al.2011). To this end, we discuss the relationships between spectral tracers of the UV radiation field and integratedfluxes from CN and HCN, with particular consideration given to the impact of disk geometry and optical depth of themolecular gas. TARGETS & OBSERVATIONS
A Sample of Young Disks in the Lupus Complex
Our sample consists of five young stars with circumstellar disks in the nearby ( d ∼
160 pc; Bailer-Jones et al. 2018)Lupus cloud complex: RY Lupi, RU Lupi, MY Lupi, Sz 68, and J1608-3070. Table 1 lists the properties of each targetfrom Alcal´a et al. (2017), including stellar mass, disk inclination, and visual extinction ( A V ). Interstellar A V is lowalong the line of sight to the Lupus clouds (Alcal´a et al. 2017), making the region well-suited for UV observations.This group of young systems shows a broad range of outer disk morphologies in ALMA observations of their gas anddust distributions. At the time of the new HST observations, two targets (Sz 68, RU Lupi) were categorized as full,primordial disks and three (MY Lupi, RY Lupi, J1608-3070) were identified as transition disks from sub-mm images( r cav = 25 , ,
75 au; van der Marel et al. 2018) under the traditional classification scheme (see e.g. Strom et al.1989; Skrutskie et al. 1990). However, RY Lupi differs from MY Lupi and J1608-3070 in that it has strong 10 µ msilicate emission (Kessler-Silacci et al. 2006) from warm grains close to the central star and undergoes periodic opticalvariability attributed to occultations by a warped inner disk (Manset et al. 2009). These signatures are not typicalof depleted transition disks, indicating that the clearing of material seen inside 50 au is a gap, rather than a cavity(Arulanantham et al. 2018; van der Marel et al. 2018). RU Lupi and MY Lupi were observed at high resolution ( ∼ HST and ALMA observations presented here include emission from both binary components.However, the secondary star (component B) and its disk are much smaller and fainter than the circumprimary disk( F star B /F star A = 0 .
17, Ghez et al. 1997; I peak,diskB /I peak,diskA = 0 .
23, Kurtovic et al. 2018), indicating that the bulkof the UV emission comes from the primary component.These systems were selected for follow-up with
HST after a large ALMA survey of the Lupus clouds identifiedthem as hosts to some of the most massive dust disks in the region (Ansdell et al. 2016, 2018). However, physical-chemical models of the CO and C O emission demonstrate that the total gas masses are unexpectedly low (Miotelloet al. 2016, 2017), which can be attributed to either shorter timescales than predicted for removing gas from the diskolecular Gas Distributions in Lupus Disks 3
TABLE 1Stellar & Disk Properties
Target Distance M ∗ A V i r cav,dust r cav,gas References[pc] [ M (cid:12) ] [mag] ◦ [au] [au]RU Lupi *
159 0.8 0.07 ∼ ◦
14, 17, 21, 24, 29.1, 34, 42, 50 · · · a b c
RY Lupi 158 1.47 0.1 68 50 50 a d
MY Lupi ∗
156 1.02 0.04 73 8, 20, 30, 40 25 a b d
Sz 68 154 2.13 0.15 34 · · · · · · a e
J1608-3070 155 1.81 0.055 74 75 60 a d*
High-resolution ALMA images of RU Lupi have revealed a series of rings inside ∼
50 au (Huang et al. 2018).The two rings with constrained values have inclinations of 20 ◦ and 17 ◦ , so we use an average of the two. Forboth RU Lupi and MY Lupi, the r cav,dust radii are locations of the dust rings resolved by (Huang et al. 2018). a Bailer-Jones et al. 2018 b Huang et al. 2018 c van der Marel et al. 2018 d Ansdell et al. 2016
TABLE 2Observations of Young Systems in Lupus
RU Lupi * RY Lupi MY Lupi Sz 68 J1608-3070
Program ID
Exposure Time [s]Observation Date
HST -COS G140L · · · λ R ∼ · · · HST -COS G130M λ R ∼ HST -COS G160M λ R ∼ HST -STIS G230L · · · λ R ∼ · · · HST -STIS G430L
120 60 60 60 60( λ R ∼ * The G160M λ λ (e.g. via external photoevaporation) or chemical pathways that trap carbon in larger molecules with higher freeze-outtemperatures. Since UV photons are critical regulators of chemical processes in disk environments (see e.g. Aikawaet al. 2002; Bergin et al. 2003; Bethell & Bergin 2011; Walsh et al. 2012, 2015; Cazzoletti et al. 2018; Visser et al.2018), the HST data we present here provide currently missing observational constraints on the levels of irradiationat the surface of the gas disk.
Observations
All five systems were observed with the
Hubble Space Telescope (HST) , using both the Cosmic Origins Spectrograph(COS; Green et al. 2012) and the Space Telescope Imaging Spectrograph (STIS; Woodgate et al. 1997a,b). Table 2lists exposure times and program IDs for the observations. The FUV spectra for three of our targets were previouslypresented in France et al. 2012 (RU Lupi), Arulanantham et al. 2018 (RY Lupi), and Alcala’ et al. 2019 (MY Lupi).Five different modes of
HST -COS and
HST -STIS were used to observe four of our systems, to cover wavelengthsfrom 1100-5000 ˚A. NUV coverage is not included for the fifth star, RU Lupi, which was observed as part of a differentprogram. These data were used to extrapolate the FUV continuum down to the hydrogen ionization edge at 912 ˚A,which is a critical region for photodissociation of abundant molecular species (e.g. H and CO) and gas-phase chemistrybut not readily accessible with available UV facilities (France et al. 2014). All five spectral modes were then stitchedtogether to produce a SED for each system using the methods outlined in France et al. (2014) and Arulanantham et al.(2018). Figure 1 shows an overview of the SEDs, highlighting the strong contribution from the accretion-dominatedNUV continuum (cid:0) λ ∼ − (cid:1) . Accretion processes enhance the FUV continuum as well, but the total flux at λ < α , C IV, C II). Our resulting library of radiation Arulanantham et al. RY Lupi MY Lupi
Wavelength [Å] F l u x [ e r g s c m Å ] Sz 68 J1608-3070
Fig. 1.—
A panchromatic spectrum was produced for four disks in Lupus by stitching together new data from five different observingmodes of
HST -COS (blue) and
HST -STIS (gray). We include reconstructed model Ly α profiles (orange; see e.g. Schindhelm et al. 2012b)in place of the observed features, which are contaminated by interstellar absorption and geocoronal emission. Interpolated fluxes spanningthe overlap between gratings are shown in magenta. The spectra from RU Lupi and the disks in Taurus-Auriga were previously presentedin Herczeg et al. (2005) and France et al. (2014). fields, which encompasses far-ultraviolet to optical wavelengths, is available to the community . We anticipate thatthe data will be used in gas-phase chemical modeling efforts that require an understanding of stellar irradiation at thedisk surface.Sub-mm CN fluxes for this work were taken from the literature, providing measurements of gas in the cold, outerregions of the disk. The data were acquired with the SMA ( ¨Oberg et al. 2010, 2011), the IRAM 30-m telescope(Guilloteau et al. 2013), and ALMA (van Terwisga et al. 2019), causing the full sample to span several differentobserving configurations and spectral features. While the Lupus ALMA survey measured fluxes from the strongest N = 3 − , J = 7 / − / N = 2 − , J = 5 / − / N = 2 − , J = 5 / − / N = 3 − , J = 7 / − / . N = 2 − , J = 5 / − / N = 2 − , J = 5 / − / http://cos.colorado.edu/ ∼ kevinf/ctts fuvfield.html olecular Gas Distributions in Lupus Disks 5 TABLE 3
Spitzer /IRS and sub-mm CN PIDs and Fluxes
Target Program ID PI Observation Date HCN Fluxes CN Reference[mm/dd/yyyy] [10 − erg s − cm − ]AA Tau 20363 J. Carr 2005 Oct 15 4 . ± . ± ± ± ± ± ± . . ± . · · · LkCa 15 40338 J. Najita 2008 Nov 5 2 . ± . . ± . ± ± ± ± ±
80 Guilloteau et al. 2013J1608-3070 · · · · · · · · · · · · van Terwisga et al. 2019V4046 Sgr 3580 M. Honda 2005 Apr 19 4 . ± . et al. (2010, 2011); however, these measurements are likely slightly underestimated, since they do not include theweaker emission lines at 226.887 and 226.892 GHz, and we depict them as lower limits in the figures presented here.We also note that different methods were used to derive the CN fluxes, with van Terwisga et al. (2019) using aperturephotometry and Guilloteau et al. (2013) and ¨Oberg et al. (2010, 2011) using the integrated spectral lines. Guilloteauet al. (2013) account for beam dilution in their flux measurements by incorporating a “beam filling factor” in theirCN line fitting procedure, while ¨Oberg et al. (2010, 2011) estimate synthesized beam sizes. We emphasize that thedifferences in observing methodologies outlined here introduce systematic uncertainties into our analysis that maycontribute ∼
30% to the Lupus CN fluxes and ∼
50% to the ¨Oberg et al. (2010, 2011) measurements that are scaledand presented here.Infrared HCN features, originating in the warmer inner disk, were measured directly from observations with theInfraRed Spectrograph (IRS) onboard the
Spitzer Space Telescope (Houck et al. 2004). All targets except J1608-3070were observed in the high-resolution mode ( R ∼ Spitzer /IRS observations. Fluxes from the 14 µ m HCN v band were measuredover the wavelength range defined for the feature in Najita et al. (2013). That work used slab models of the moleculargas disk to identify line-free spectral windows for each target, which we used as a reference for continuum subtraction.Although the measurement errors from the Spitzer spectra are ∼ ∼ Archival
HST
Data from YSOs in Taurus-Auriga
To increase the sample size of this study and compare properties between different star-forming regions, we include allTaurus-Auriga sources from the literature with both sub-mm CN fluxes and UV spectra from
HST -STIS and/or
HST -COS. Disks in Taurus are roughly equivalent in age to the Lupus systems ( ∼ In Protoplanetary Systems; PI: K.France; PID: 12876).
HST -COS observations were acquired with both the G130M and G160M gratings for all systems,providing wavelength coverage over the same range of FUV wavelengths ( ∼ Uncertainty in Literature Measurements of A V Observations at FUV wavelengths are highly affected by the amount of dust and gas along the line of sight to a disk( A V ), making accurate reddening corrections critical in interpreting the spectral features. France et al. (2017) report Arulanantham et al. L FUV (traditional extinction) L F U V ( N ( H I ) e x t i n c t i o n ) L Ly (traditional extinction) L L y ( N ( H I ) e x t i n c t i o n ) Fig. 2.—
Comparison of luminosities from the FUV continuum (left) and Ly α (right), calculated using different values of A V to dereddenthe spectrum. The method described in McJunkin et al. (2014), which uses the Ly α wings to estimate N ( HI ) along the line of sight,typically yields smaller A V and lower luminosities than A V derived from broadband color excesses or comparisons to stellar photospherictemplates (1-to-1 relationship traced by black, dashed lines). The major outliers in the right panel are disks with A V values that changedby ∆ A V > .
8. Since the N ( HI )-based measurements alleviate the correlation between A V and total flux from UV-H (France et al.2017), we adopt those extinctions for the analysis presented here. a statistically significant positive correlation between A V (derived from e.g. broadband color excesses, deviations fromstellar photospheric templates; Kenyon & Hartmann 1995; Hartigan & Kenyon 2003) and the total luminosity from UV-fluorescent H . The trend implies that more molecular gas emission is seen from disks with more intervening material;however, there is no physical or chemical process that would produce such a relationship between circumstellar andinterstellar material, making systematic overestimates in the A V measurements a more likely driver. This effect mustbe removed in order to accurately assess relationships between the UV spectral features.To provide an estimate of the line-of-sight interstellar extinction that is less sensitive to the circumstellar dustproperties, McJunkin et al. (2014) used observed Ly α profiles to directly measure the amount of neutral hydrogen(H I) along the line of sight. A V values were then calculated as A V /R V = N (H I) / (cid:0) . × atoms cm − mag − (cid:1) (Bohlin et al. 1978). France et al. (2017) find that the correlation between A V and L (H ) becomes statisticallyinsignificant when the McJunkin et al. (2014) reddening values are adopted, so we adopt the H I-derived A V valuesfor the analysis presented here.Extinctions derived using the McJunkin et al. (2014) method are typically significantly lower than reported inthe literature, since the measurements isolate interstellar N(H I) from circumstellar material that may significantlyincrease A V estimates (e.g. DR Tau; ∆ A V ∼ . N (H I)-based A V values to the measurements using more traditional methods to estimate A V , using targetsfrom McJunkin et al. (2014) and France et al. (2017). We find that the two luminosities are roughly consistent formost disks, demonstrating that the trends presented later in this work are independent of the choice of A V . Figure 2also presents the same comparison for Ly α luminosities measured using the two different A V values, showing just fivemajor outliers that have the largest changes in A V as well (∆ A V > . A V < .
8. We therefore conclude that the trendsin Ly α emission presented here are also not driven by the choice of A V . Normalized UV Luminosities
Models of UV-sensitive molecular gas distributions have included a broad range of stellar blackbody temperatures,UV excesses due to accretion, and total UV luminosities, allowing the authors to isolate the impact of UV irradiationfrom disk geometric properties such as disk mass, flaring angle, and pressure scale height (see e.g. van Zadelhoff et al.2003; Walsh et al. 2015; Cazzoletti et al. 2018; Bergner et al. 2019). Other observational work has split the targets byspectral type, allowing the authors to roughly correct for stellar mass and temperature (Pascucci et al. 2009). However,the sample we present here is too small to meaningfully bin the data by both disk and stellar properties. Instead, wedivide the individual UV components by the total UV luminosity ( L UV,total ), defined as the sum of fluxes from theFUV continuum, Ly α , UV-H , C IV, and C II].Since both L UV,total and fluxes from the individual UV components are impacted by target properties in the sameway (e.g. i d , M ∗ ), this correction allows us to evaluate the entire sample without binning the data based on diskor stellar parameters. The normalization also reduces the impact of systematic uncertainties in measuring A V , asdescribed in Section 2.4, although we note that resonant scattering effects may preferentially enhance Ly α fluxes indeeper layers of the disk (see e.g. Bethell & Bergin 2011), relative to emission at other FUV and NUV wavelengths.When the data were analyzed before accounting for the diversity of disk and stellar parameters, no correlations wereobserved between any of the UV tracers and the 14 µ m HCN or sub-mm CN fluxes. We therefore conclude thatolecular Gas Distributions in Lupus Disks 7normalizing the UV luminosities allows us to provide more physically meaningful information about UV-dependentgas-phase chemistry than using the raw fluxes alone. All plots also differentiate between targets with resolved dustsub-structure (open markers) and full, primordial disks (filled markers), demonstrating that the results are roughlyindependent of evolutionary phase. RESULTS
The
HST -COS and
HST -STIS spectra described above provide direct measurements of the UV radiation field,which we use to estimate the UV flux reaching the surface of the gas disk. Here we focus on the role of UV photons inproducing CN, which has been detected in ALMA Band 7 observations of a large sample of disks in the Lupus clouds(van Terwisga et al. 2019), including the five studied here. We present the following results under the assumption thatthe dominant reaction pathway for CN production in disks isN + H ∗ −→ NH + HC + + NH −→ CN + + HCN + + H −→ CN + H + , (2)where H ∗ is gas that has been pumped into excited vibrational states by FUV photons (Walsh et al. 2015; Heays et al.2017; Cazzoletti et al. 2018). HCN can then be formed via reactions with H and CH CN + H −→ HCN + HCN + CH −→ HCN + CH (3)(Baulch et al. 1994; Walsh et al. 2015; Visser et al. 2018). Destruction of HCN by UV photons also significantlyinfluences the total abundance of CN in the disk (Walsh et al. 2015; Cazzoletti et al. 2018; Pontoppidan et al. 2019),with photodissociation occurring at a rate of 1 . × − s − under a typical interstellar radiation field and a rate of5 . × − s − under a standard T Tauri Ly α profile (van Dishoeck et al. 2006; Heays et al. 2017). Although TheUMIST Database for Astrochemistry (McElroy et al. 2013) lists many other reactions for CN formation, physical-chemical models demonstrate that the H ∗ pathway is the most important route (see e.g. Visser et al. 2018). This maybe attributed to the high abundance of H in disks, relative to molecular species like C H and OH that are requiredfor alternate pathways (Visser et al. 2018).
Photodissociation of N by the FUV Continuum Gas-phase chemical models of protoplanetary disks find that column densities of nitrogen-dependent molecularspecies (e.g. CN, HCN) vary with the total FUV flux at the disk surface (Pascucci et al. 2009; Walsh et al. 2012,2015). The trend is attributed to photoabsorption, since five excited electronic transitions of N fall between 912-1100˚A. Absorption at these wavelengths produces pre-dissociated N via coupling to the continuum (Li et al. 2013; Visseret al. 2018). As shown in Eq. 1, the atomic nitrogen products react with H ∗ to produce NH, therefore catalyzingformation of molecules like CN and HCN (Walsh et al. 2015; Cazzoletti et al. 2018).The predicted relationship between molecular abundances and UV irradiation is corroborated by observationalstudies at IR wavelengths, which find that disks around cool M stars show less emission from nitrogen bearing moleculesthan disks around hotter solar-type stars (Pascucci et al. 2009, 2013; Najita et al. 2013). The difference is attributed tovarying UV photon production rates, which are expected to be lower in cooler stars (van Zadelhoff et al. 2003; Walshet al. 2012, 2015). To further investigate this relationship from an observational perspective, we compare CN fluxesmeasured from ALMA observations (van Terwisga et al. 2019) to FUV fluxes from the HST -COS and
HST -STISspectra presented here.Although our
HST data are truncated below 1000 ˚A, the method from France et al. (2014) was used to estimate theFUV continuum at shorter wavelengths. The extrapolation was performed on binned fluxes, which were calculated in210 line-free regions from the longer wavelength data. The continua were fitted with a second order polynomial andextended down to 912 ˚A. For uniformity with the larger sample presented in France et al. (2014), we report the totalintegrated flux from wavelengths < ρ = 0 . p = 0 . L HCN = m × L F UV /L UV,total + b shows that the DR Tau and AA Tau (2013) spectra are highlyinfluential data points. When these outliers are removed from the correlation, the Spearman rank coefficient increasesto ρ = 0 .
66, although p remains at 0.01 because of the reduction in sample size. Both targets are included in Figure 3,but we note that the AA Tau (2013) spectrum in particular is impacted by an inner disk warp (Schneider et al. 2015;Hoadley et al. 2015; Loomis et al. 2017) that is attenuating the FUV flux. In contrast with both the HCN results andmodeling work, we find a significant negative correlation between the FUV and sub-mm CN fluxes ( ρ = − . p = 0 . L FUV / L UV , total L C N [ e r g s ] B P T a u D E T a u D M T a u D R T a u G M A u r J - L k C a M Y L u p i R U L u p i R Y L u p i S U A u r S z V S g r Spearman = -0.6p-value = 0.0165 10 L FUV / L UV , total L H C N [ e r g s ] A A T a u ( ) A A T a u ( ) B P T a u D E T a u D M T a u D R T a u G M A u r H N T a u A L k C a M Y L u p i R U L u p i R Y L u p i S U A u r S z V S g r Spearman = 0.64p-value = 0.0099
Fig. 3.—
Sub-mm CN (left) and 14 µ m HCN (right) luminosities versus fractional flux from the FUV continuum (integrated from912-1650 ˚A). The five Lupus systems are shown as red squares and the subset of disks from France et al. (2017) as black circles, with openmarkers representing disks with resolved dust substructure. N molecules are readily predissociated by photons between 912-1000 ˚A (Liet al. 2013), and the atomic nitrogen then reacts with H ∗ as a first step in CN formation (Walsh et al. 2015; Cazzoletti et al. 2018). Thetwo relationships demonstrate that while N photodissociation may proceed more efficiently in disks that are more strongly irradiated byFUV photons relative to the rest of the UV spectrum, CN photodissociation may increase as well. Measured FUV luminosities in this plotare accurate to within a factor of ∼ A V dominating the uncertainties. emission from warmer disks. The negative correlation between sub-mm CN emission and the FUV continuum impliesthat this trend at least is dominated by photodissociation, since increased UV irradiation could lead to a warmer diskand therefore an opposite relationship. However, the trend of increased IR HCN emission in systems with stronger FUVfluxes could be attributed to either scenario, with both a larger number of HCN molecules and higher temperaturesleading to enhanced populations in the upper state of the 14 µ m transition. This result will become easier to interpretas more measurements of sub-mm HCN emission (see e.g. ¨Oberg et al. 2010, 2011; Bergner et al. 2019) are acquiredwith ALMA in coming years. Despite the uncertainty in the driving force behind the HCN trend, the negative andpositive correlations between the FUV continuum and CN and HCN, respectively, demonstrate that the amount ofFUV flux in a given disk helps control the balance of formation and destruction pathways that determine abundancesof UV-sensitive species. Ly α as a Regulator of Disk Chemistry Ly α emission is by far the strongest component of the UV radiation field, comprising roughly 75-95% of the totalflux from ∼ − (Herczeg et al. 2002, 2004; Hoadley et al. 2017) and CO (France et al. 2011a;Schindhelm et al. 2012a) and photodissociation energies of HCN, C H (Bergin et al. 2003; Walsh et al. 2015; Heayset al. 2017), and H O (France et al. 2017). Destruction via Ly α photons is not necessarily more efficient thandissociative transitions at other wavelengths, but this is mitigated by the large number of Ly α photons relative to otherregions of the UV spectrum. Unfortunately, the observed Ly α feature is always contaminated by geocoronal emissionand ISM absorption along the line-of-sight, making direct measurements difficult. Instead, we use the method fromSchindhelm et al. (2012b) to reconstruct the Ly α profiles for the five Lupus disks from observations of UV-fluorescentH .UV-H fluorescence is activated when a population of hot, vibrationally excited H ( T > (cid:0) X Σ + g (cid:1) into the first andsecond dipole-allowed excited electronic state (cid:0) B Σ + u , C Π u (cid:1) by photons with energies that fall along the Ly α lineprofile (Herczeg et al. 2002, 2004; France et al. 2011b). A cascade of UV emission lines is then observed as themolecular gas population transitions back to the ground electronic state. The features can be divided into groupscalled progressions, where a single progression, denoted [ ν (cid:48) , J (cid:48) ], consists of all transitions out of the same upperelectronic level with vibrational state ν (cid:48) and rotational state J (cid:48) (Herczeg et al. 2002, 2004). We measured fluxes fromthe strongest emission lines in 12 progressions (see Table 2 of France et al. 2012) by integrating over models of aGaussian profile convolved with the HST -COS line-spread function (LSF) and superimposed on a linear continuum.Upper limits on features indistinguishable from the continuum were calculated as the RMS flux within a 3 ˚A rangeacross the expected line center. The total progression fluxes ( F H , i ) were then used as Ly α “data” ( y ), where the x value for each data point is the Ly α pumping wavelength for the electronic transition. We then fit a model Ly α profileto the ( x i , y i ) = ( λ LyA,i , F H ,i ) data points.The Ly α model consists of an initial “intrinsic” Gaussian emission line, an H I outflow between the star/accretionshock and the molecular gas disk, and a population of H that absorbs the Ly α photons. We allow five modelolecular Gas Distributions in Lupus Disks 9 Wavelength [ Å ] L y I n t e n s i t y [ e r g s c m Å ] MY LupRU LupRY LupSz 68TYC 7851-810-1
Fig. 4.—
A comparison of the reconstructed Ly α profiles at the disk surface for all five Lupus targets in our sample, with colored contoursshowing rough uncertainties associated with the modeling procedure. The reconstruction is done using observed UV-H emission lines asdata points (Schindhelm et al. 2012b), since molecules are pumped into these excited electronic states by Ly α photons. The absorptionseen on the blue side of the line profiles is due to an atomic outflow between the star and disk, rather than interstellar H I along theline-of-sight. TABLE 4Best-Fit Parameters * for Ly α Reconstruction
Target I Lyα v out N out T H N H [erg s − cm − ˚A − ] [km s − ] [dex] [K] [dex]RU Lupi (3 . ± . × − − +6 − . +0 . − . ±
900 18 . +0 . − . RY Lupi 2 +2 − × − − +12 − . +0 . − . ± . +1 . − . MY Lupi 3 +50 − × − − +50 − . +0 . − +1500 − . ± +50 − × − − +30 − ± ± . +1 . − J1608-3070 1 . +0 . − . × − − +1 − . +0 . − . +400 − +2 − Model parameters are: I Lyα = amplitude of intrinsic Ly α profile, v out =velocity of intervening outflowing gas, N out = column density of interven-ing outflowing gas, T H = temperature of fluorescent H , N H = columndensity of fluorescent H . The FWHM of the intrinsic profile for each sys-tem was held constant in the model, set to the average (708 km s − ) fromSchindhelm et al. 2012b. Error bounds on the parameters were estimatedby running the model with the FWHM fixed to the minimum (573 kms − ) and maximum (912 km s − ) values from that work. parameters to vary: the amplitude of the intrinsic profile ( I Lyα ), the velocity and column density of the outflowing HI ( v out , N out ), and the temperature and column density of the absorbing H ( T H , N H ). The FWHM of the intrinsicprofile for each system was fixed to the average, maximum, and minimum values from Schindhelm et al. (2012b),resulting in three model profiles for each target. Posterior distributions for the model parameters were constructedusing MCMC sampling (Foreman-Mackey et al. 2013) within the bounds defined by Schindhelm et al. (2012b). However,we find that the model uncertainties are better captured by the variations in the average, maximum, and minimumFWHM profiles. Figure 4 shows the median Ly α profile at the disk surface for all five Lupus systems, with coloredcontours representing the bounds set by the three FWHM values (see Table 4).We compare the total luminosities from our reconstructed Ly α profiles to the the CN and HCN luminosities in Figure5. Since CN molecules can also form as byproducts of HCN or CH CN photodissociation (Walsh et al. 2015) via Ly α photons near 1216 ˚A (Nuth & Glicker 1982; Bergin et al. 2003), we expect that increased Ly α irradiation of the0 Arulanantham et al. L Ly / L UV , total L C N [ e r g s ] A A T a u ( ) B P T a u D E T a u D M T a u D R T a u G M A u r J - L k C a M Y L u p i R U L u p i R Y L u p i S U A u r S z V S g r Spearman = 0.38p-value = 0.1745 0.2 0.4 0.6 0.8 1.0 L Ly / L UV , total L H C N [ e r g s ] A A T a u ( ) B P T a u D E T a u D M T a u D R T a u G M A u r H N T a u A L k C a M Y L u p i R U L u p i R Y L u p i S U A u r S z V S g r Spearman = -0.4p-value = 0.1021
Fig. 5.—
Sub-mm CN (left) and 14 µ m HCN (right) emission versus fractional Ly α luminosity. The Lupus disks are shown as red squaresand systems from France et al. (2017) as black circles, with open markers representing systems with resolved dust substructure. Neitherspecies is significantly correlated with Ly α emission, but the Spearman rank coefficients are tentatively consistent with models predictingincreased CN and decreased HCN abundances with increased Ly α irradiation. The 2013 AA Tau spectrum is omitted from these plots,since the Ly α fluxes were very similar to the spectrum from 2011. disk surface will increase the significance of dissociative pathways in regulating both CN and HCN column densities.However, no statistically significant relationships are detected between Ly α and either HCN ( ρ = − . p = 0 .
10) orCN ( ρ = 0 . p = 0 . ρ ) are consistent with modelpredictions, with increased CN and decreased HCN emission observed from targets with stronger Ly α emission.Since the Ly α profiles are derived from UV-fluorescent H ∗ , the lack of correlation between CN and Ly α emissionmay be attributed to the radial stratification of the UV-H and sub-mm CN. The UV-H emission originates from gasin surface layers of the inner disk ( r <
10 au), while the CN population extends to radii of ∼ α emission, with UV-H emission originating from much closer to the disksurface than the 14 µ m HCN features. A direct comparison between observed Ly α , HCN, and CN luminosities likelyrequires more careful treatment of optical depth effects. Furthermore, a bootstrapping analysis of the data returns ± σ confidence intervals on the Spearman rank coefficients of [-0.71, -0.29] and [0.2, 0.67] for the HCN and CN vs.L (Ly α ) / L (UV , total) correlations. Although the upper limits of the confidence intervals are consistent with robustlinear relationships, it is possible that targets with larger uncertainties on the reconstructed Ly α profiles (e.g. MYLupi, Sz 68) are masking underlying trends in the data.We also note that the HST -COS,
Spitzer -IRS, and sub-mm CN observations were not conducted simultaneously(see Tables 2, 3), implying that the reconstructed Ly α profiles may not be representative of the flux reaching thedisk surface at the time of the molecular gas observations. However, an HST analysis of older K and M dwarfs hasdemonstrated that Ly α line strengths do not increase as much as other atomic features during flares (e.g. Si IV λ α emission also remains steady during typical YSO variability. AlthoughFrance et al. (2011) report a strong correlation between the FUV continuum and the C IV λ Director’s Discretionary program on
HST . Mapping the population of H ∗ Ly α -pumped H As introduced in Section 3.2, we detect a suite of emission lines from hot ( T ∼ − in the HST -COS spectra of all five disks. Although our
HST -COS spectra are not spatially resolved,
HST -STIS spectra of the same features in the disk around TW Hya show that the H emission must be within ∼ ∗ available for CN and HCN formation in the inner disk region. Figure 6 provides a cartoon demonstrating the roughspatial locations of the emitting gas.The UV-H emission lines are spectrally resolved (i.e. broader than the HST -COS resolution), allowing us to extractinformation about the spatial distribution of hot, fluorescent gas. Assuming that the material is in a Keplerian disk,the FWHMs of the emission lines can be mapped to an average radial location as (cid:104) R H (cid:105) = GM ∗ (cid:18) iF W HM (cid:19) (4) olecular Gas Distributions in Lupus Disks 11 Fig. 6.—
Rough spatial locations of emitting gas that produces C II] λ , IR-HCN, and sub-mm-CN emission lines, compared toradius where physical-chemical models predict peak abundances of H ∗ , CN, and HCN (Cazzoletti et al. 2018). The observed gas populationsshould overlap at radii close to the central star, although the sub-mm CN emission is the only component that generally extends acrossthe full disk. TABLE 5Average Emission Radii of Hot, Fluorescent H Target
F W HM [1 , (cid:10) R H (cid:11) [1 , F W HM [1 , (cid:10) R H (cid:11) [1 , F W HM [0 , (cid:10) R H (cid:11) [0 , F W HM [0 , (cid:10) R H (cid:11) [0 , [km/s] [au] [km/s] [au] [km/s] [au] [km/s] [au]RU Lupi 51 ± . ± .
03 49 ± . ± .
03 31 ± . ± .
08 50 ± . ± . * . ± . . ± . ± . ± . ±
10 1 ± ± . ± . ± . ± . ± . ± . ± . ± . ± . ± . ± . ± . ± . ± . ± . ± . ± . ± . . ± . ± . ± . ± ±
10 0 . ± . ±
10 0 . ± . * The FWHMs listed here for RY Lupi are from single-component fits to the emission lines. Arulanantham et al. (2018) presents a more detailedanalysis of the H line shapes, showing that the strongest features in the [1,4] progression are better fit by a two-component model. We also adoptthe disk inclination from van der Marel et al. (2018) (68 ◦ ), instead of the scattered light inclination from Manset et al. (2009) (85 . ◦ ). (Salyk et al. 2011a; France et al. 2012), where M ∗ is the stellar mass and i is the disk inclination. We average theFWHMs of the strongest emission lines from the [ ν (cid:48) , J (cid:48) ] = [1,4], [1,7], [0,1], and [0,2] progressions and calculate a radiusfor each progression (see Table 5), finding that the bulk UV-H emission originates inside ∼ r ∼
75 au; van der Marel et al. 2018). This is consistent with the results of Hoadley et al. (2015),who found that disks with less advanced dust evolution typically have broad UV-H features that are dominated bygas located close to the star. Although resonant scattering should allow pumping photons from the Ly α line wings topenetrate deeper into the disk than those from the line center (Bethell & Bergin 2011), we find no trends between theaverage emitting radius calculated for each progression and the Ly α pumping wavelength responsible for exciting thetransitions.Since H ∗ is required to produce CN (see Eq. 1), the UV-H features are a probe of the uppermost layer of availablereactants located at the average radius of emitting gas ( R H ). We explore this relationship by comparing the CN andHCN luminosities to the total flux from fluorescent UV-H emission lines (see Figure 7), finding that neither speciesis correlated with the UV-H . In the case of HCN, the scatter can be attributed to the vertical stratification effectsdiscussed in Section 3.2, since the H ∗ and HCN abundances are expected to peak at different heights relative to thedisk midplane (Cazzoletti et al. 2018). Ly α photons are only able to pump H ∗ in a thin surface layer, so the UV-H features do not contain information from vibrationally excited gas present deeper in the disk that would be co-located2 Arulanantham et al. L H / L UV , total L C N [ e r g s ] A A T a u ( ) A A T a u ( ) B P T a u D E T a u D M T a u D R T a u G M A u r J - L k C a M Y L u p i R U L u p i R Y L u p i S U A u r S z V S g r Spearman = -0.0p-value = 0.9546 10 L H / L UV , total L H C N [ e r g s ] A A T a u ( ) A A T a u ( ) B P T a u D E T a u D M T a u D R T a u G M A u r H N T a u A L k C a M Y L u p i R U L u p i R Y L u p i S U A u r S z V S g r Spearman = 0.39p-value = 0.1450
Fig. 7.—
CN and HCN luminosities versus fractional luminosity from UV-pumped H , with open markers representing disks with resolveddust substructure (i.e. rings, gaps, or cavities). The Lupus disks are shown as red squares, while the black circles are systems from Franceet al. (2017). Neither sub-mm CN or IR HCN is significantly correlated with UV-H , which is likely due to the disparate radial and verticalstratification of the three populations of emitting gas. TABLE 6Hot Gas & Stellar Luminosities
Target L H L ˚ A L Lyα L CII ] L FUV c [10 erg s − ] [10 erg s − ] [10 erg s − ] [10 erg s − ] [10 erg s − ]RU Lupi 6 . ± . . ± . a ± · · · a RY Lupi 1 . ± .
08 1 . ± . a ±
25 2 . ± .
02 8.3MY Lupi 0 . ± .
04 15 . ± b . ± .
01 3.5Sz 68 0 . ± .
03 3 . ± b . ± .
01 3.2J1608-3070 0 . ± . . ± . b . ± .
01 4.4 a Values from France et al. (2017). b Reconstructed Ly α profiles have large uncertainties, due to noisy UV-H emission lines. c FUV continuum luminosities are accurate to within a factor of ∼ with the HCN. By contrast, the sub-mm emission traces CN molecules in the cold outer disk, where the population ofH ∗ declines due to extinction of pumping photons from the UV continuum (Visser et al. 2018; Cazzoletti et al. 2018).The radial distribution of UV-H may provide estimates of either how far Ly α emission is able to travel in the diskor a rough boundary for the population of H ∗ . We explore this degeneracy further in Section 4 by using 2-D radiativetransfer models to reproduce the UV-H emission lines. O Dissociation
Previous studies of UV continuum emission from young stars with disks have identified an excess “bump” in thespectra around ∼ α -pumped H ∗ , where the population of H ∗ is indirectly produced during H O dissociationin the inner disk ( r < α and bump luminosities are strongly correlated (cid:0) ρ = 0 . , p = 1 . × − ; France et al. 2017 (cid:1) , no relationship is observed between the bump and X-ray luminosities(Espaillat et al. 2019). This implies that Ly α photons may play a more prominent role than the X-ray radiation fieldin regulating the distribution of hot H O and vibrationally excited H at radii close to the central star.We use the method described in France et al. (2017) to measure bump fluxes and calculate luminosities for thethree disks in our sample that were not included in that work. A second order polynomial fit to the FUV continuum,representative of the underlying flux, was integrated from 1490-1690 ˚A and subtracted from the total observed flux inthe same wavelength region. The residual flux can be attributed to the bump alone (see Table 6). Figure 8 comparesthe CN/HCN and bump luminosities, again showing no clear linear trend between the spectral features. Similar tothe Ly α -pumped fluorescent gas, the H ∗ responsible for producing the bump is therefore likely located higher in thedisk surface than the 14 µ m HCN and constrained to closer radii than the sub-mm CN. C II] λ + The semi-forbidden C II] emission lines at λ HST -STIS spectra. Models of the feature in other young systems find that its shape is consistent with an origin atolecular Gas Distributions in Lupus Disks 13 L Å / L UV , total L C N [ e r g s ] A A T a u ( ) A A T a u ( ) B P T a u D E T a u D M T a u D R T a u G M A u r J - L k C a M Y L u p i R U L u p i R Y L u p i S U A u r S z V S g r Spearman = -0.2p-value = 0.3613 10 L Å / L UV , total L H C N [ e r g s ] A A T a u ( ) A A T a u ( ) B P T a u D E T a u D M T a u D R T a u G M A u r H N T a u A M Y L u p i R U L u p i R Y L u p i S U A u r S z V S g r Spearman = 0.03p-value = 0.8989
Fig. 8.—
CN and HCN luminosities versus fractional luminosity from the 1600 ˚A bump (produced by Ly α -pumped H ∗ left behind duringH O photodissociation; France et al. 2017). The Lupus disks are shown as red squares, while the black circles are systems from Franceet al. (2017). The strength of the bump is correlated with time-varying accretion luminosities (Espaillat et al. 2019), implying that themeasurements shown here are snapshots of the population of hot, non-thermal gas and not necessarily reflective of equilibrium chemicalconditions. the base of a warm, inner disk ( r ∼ . − + .Simon et al. (2016) find that the low-velocity component of the [O I] profile is consistent with Keplerian rotation, withemitting gas originating in radially separated components located between 0.05 and 5 au. Only one target (AA Tau)in the sample presented here was observed at high enough spectral resolution and signal-to-noise to obtain kinematicinformation from the C II] emission, with three individual features resolved at λ λ λ .
12 ˚A.We find that the central feature ( λ − ; Banzatti et al. 2019), which likely implies that the two emission lines have similar inner disk origins. TheC II] λ + ) is a key reactant in the main formation pathways of C H (Henning et al. 2010; Walsh et al.2015; Miotello et al. 2019) and CN (Walsh et al. 2015; Visser et al. 2018; Cazzoletti et al. 2018). C + also plays animportant role in CO destruction, with physical-chemical models showing an enhancement in the CO column densityat r ∼
10 au, where gas self-shielding allows the C + /C ratio to drop below unity (Walsh et al. 2012). We note that theFUV C II λ + that is directly involved in molecule formation and destruction than needed for the C II] λ HST -STIS spectra over a wavelength range spanning all threeC II] features and subtracting a linear continuum (see Table 6). Figure 9 compares the CN, HCN, and C II] λ ρ = 0 . p = 0 . ρ = − . p = 0 . L CN = m × L CII ] /L UV,total + b .The negative correlation between fractional C II] and sub-mm CN emission points to a relationship between C II] λ + is similarly driven. Although we find that theC II] emission is not significantly correlated with the mass accretion rate ( ρ = 0 . p = 0 . HST -STIS observations of the C II] λ L CII ] / L UV , total L C N [ e r g s ] A A T a u ( ) A A T a u ( ) B P T a u D E T a u D M T a u D R T a u G M A u r J - L k C a M Y L u p i R Y L u p i S U A u r S z Spearman = -0.5p-value = 0.0322 10 L CII ] / L UV , total L H C N [ e r g s ] A A T a u ( ) A A T a u ( ) B P T a u D E T a u D M T a u D R T a u G M A u r H N T a u A L k C a M Y L u p i R Y L u p i S U A u r S z Spearman = 0.82p-value = 0.0005
Fig. 9.—
CN and HCN luminosities versus fractional C II] λ L CII ] / L UV , total L F U V / L U V , t o t a l B P T a u S z D E T a u D M T a u H N T a u A L k C a G M A u r A A T a u ( ) A A T a u ( ) S U A u r R Y L u p i D R T a u M Y L u p i J - Spearman = 0.62p-value = 0.0165
Fig. 10.—
Fractional FUV continuum emission versus fractional C II] λ µ m HCN emission, implying that systems with stronger FUV and C II] fluxes are better able to produce HCN in the inner disk. Afull kinematic analysis of the C II] line profiles is likely required to determine whether the C II] emission traces the C + population involvedin gas-phase chemistry or accretion processes that enhance the strength of the feature. analysis of the line properties is outside the scope of this work. DISCUSSION
Physical-chemical models of disks have suggested that emission from Ly α and the FUV continuum directly impactmolecular gas abundances, providing photons at the energies required for gas-phase reactions to proceed. Althoughwe find significant correlations between both 14 µ m HCN and sub-mm CN and the FUV continuum, neither speciesappears to be related to Ly α emission. In order to understand this discrepancy, we examine the impact of diskgeometry on the observed spectra and consider whether the optical depth of the inner disk has a significant impact onour results. Extent of UV-H Emitting Region olecular Gas Distributions in Lupus Disks 15 -1 Radius [au] N o r m a li z e d F l u x Sp e c t r a l R e s o l u t i o n L i m i t Sz 68MY Lupi RY LupiJ1608-3070 RU Lupi
Fig. 11.—
Radial distributions of flux from hot, UV-fluorescent H pumped by Ly α in the five Lupus disks presented here, obtained byfitting a 2-D radiative transfer code to individual emission lines. The flux distribution in J1608-3070 extends to much more distant radiithan the other three systems, consistent with a depletion of small dust grains inside a large sub-mm cavity ( r cav ∼
75 au; van der Marelet al. 2018). Although the flux distribution from MY Lupi spans the radii of its first two dust rings ( r ∼ ,
20 au; Huang et al. 2018), wedetect no sign of a break in the population of hot gas. Finally, we note that the flux distribution from Sz 68 is sharply truncated around10 au, implying that the UV-H emission extends to the edge of the system’s circumprimary disk (Kurtovic et al. 2018). UV-H features observed with HST -COS are typically much broader than the instrument resolution (cid:0) ∆ v ∼
17 km s − (cid:1) ,implying that most of the detected flux is emitted at radii ≤
10 au from the central star (see e.g. France et al. 2012).This is supported by
HST -STIS spectra of H in the disk around TW Hya, which show that the UV-H emissionlines are not spatially extended and are therefore confined to the inner disk (Herczeg et al. 2002). However, physical-chemical models suggest that both Ly α and FUV photons reach large swaths of the outer disk (see e.g. Cleeves et al.2016). The FUV photons can pump H into vibrationally excited states at radii where the gas temperature is toolow for thermal populations of H ∗ to survive (see e.g. Cazzoletti et al. 2018). Ly α photons then act as a searchlightilluminating the vibrationally excited population.Since fluorescent emission is not detected from the outer disk (see Figure 11), the radial extent of the UV-H emittingregion may be restricted by either:1. the extent to which Ly α pumping photons can travel into the disk (Ly α -limited), or2. the total abundance of H ∗ , excited thermally ( T > ∗ -limited).The negative correlation between sub-mm CN fluxes and the FUV continuum (measured both directly and via CII] λ ∗ in those regions. With this in mind, we focus on a simple model of the Ly α -limited scenario. A moreadvanced full disk H ∗ distribution will be analyzed in a future paper on the H ∗ -limited scenario. Emission
To investigate whether our UV-H observations are Ly α -limited, we use the 2-D radiative transfer model developedby Hoadley et al. (2015) to reproduce the distributions of fluorescent gas in 4/5 of the Lupus disks presented here.The model propagates Ly α photons into a Keplerian disk with a power-law temperature gradient of coefficient q , fixedtemperature T au at a radial distance of 1 au, and a minimum T = 1000 K for UV-fluorescence, T ( r ) = T au (cid:16) r (cid:17) − q (5)a pressure scale height dependent on the radial temperature distribution T ( r ) and stellar mass M ∗ , H p ( r ) = (cid:115) kT ( r ) µm H r GM ∗ (6)6 Arulanantham et al.and a surface density distribution with characteristic radius r c , power-law coefficient γ , and normalization factorΣ c = M H (2 − γ ) /µ r char (2 π ) / Σ ( r ) = Σ c (cid:18) rr c (cid:19) − γ exp (cid:34) − (cid:18) rr c (cid:19) − γ (cid:35) (7)We calculate the mass density distribution at some height z above the disk midplane ( ρ ( r, z )) for the entire volume ofH gas and the corresponding number density (cid:0) n [ ν,J ] ( r, z ) (cid:1) and optical depth ( τ λ ( r, z )) of molecules in the upper levelof each progression. Once the physical structure of the underlying hot H population has been derived, the distributionof UV-H flux from each transition is calculated as F λ H ( r, z ) = ηF ∗ ,Lyα (cid:18) R ∗ r (cid:19) (cid:32) ( r cos i disk ) s ( r, z ) (cid:33) × B mn τ (cid:48) λ (cid:88) (cid:16) − e − τ (cid:48) λ ( r,z ) (cid:17) , (8)where η represents the geometric fraction of the disk exposed to Ly α photons (held constant at 0.25; Herczeg et al.2004), F ∗ ,Lyα is the Ly α flux reaching the gas, B mn is the branching ratio that describes the likelihood of a giventransition relative to all other transitions from the same upper level, and s ( r, z ) is the sightline from the observerto a gas parcel at position ( r, z ) in the disk. The flux distribution is then summed over the entire disk, producingan emission line profile that we fit directly to the observed UV-H spectra. The resulting model distribution of gasinforms us about where in the disk the H ∗ is exposed to Ly α radiation, providing radial constraints on the uppermostlayer of reactants for producing CN molecules in the inner disk.To fit these models to the observed UV-H features from the Lupus disks, we used the reconstructed Ly α profilesshown in Figure 4 to estimate F ∗ ,Lyα for each progression. The disk inclinations ( i disk ) and stellar masses ( M ∗ ) werefixed to values from the literature (see Table 1). The parameters z , γ , T , q , r char , and M H were allowed to vary,and uncertainties on the best-fit models were estimated using MCMC re-sampling (Foreman-Mackey et al. 2013) withuniform priors spanning the grid space defined by Hoadley et al. (2015). The MCMC algorithm used 3000 walkers toexplore the parameter space, finding no strong degeneracies between the six variables (see Figure A.1 for an examplecorner plot). The final distributions of UV-H flux were most sensitive to the values of T and q used to define theradial temperature structure defined in Equation 5. Radial Distributions of Flux from UV-H Figure 11 shows the radial distributions of UV-H flux that best reproduce the observed emission lines from theLupus disks. The shapes of the gas distributions are generally correlated with the sub-mm dust distributions, inagreement with the results from Hoadley et al. (2015) that showed less UV-H close to the star in disks with dustgaps or cavities. Sz 68, which is a close binary (Ghez et al. 1997), shows a distribution that is sharply truncated at10 au. This is consistent with UV-H emission from the primary component alone (Kurtovic et al. 2018), with Ly α photons reaching the gas surface layers out to the circumprimary disk edge. MY Lupi, which has two shallow gaps at8 and 20 au (Huang et al. 2018), shows no sign of breaks in the gas disk at those radii, although the flux distributiondeclines rapidly from its peak at ∼ emission in J1608-3070 extends tomore distant radii than the other systems, with a flat distribution from ∼ into vibrationallyexcited states. However, we note that the outer radius of the UV-H distribution is limited by the HST -COS spectralresolution (cid:0) ∆ v ∼
17 km s − (cid:1) . Given the stellar mass and disk inclination of J1608-3070, this corresponds to a spatialscale maximum of ∼
20 au.Although the UV-H lines do not originate from the same region as the sub-mm CN emission, the distributions ofUV-H flux provide constraints on either the radial extent of the population of H ∗ in surface layers of the gas disk orthe location where those surface layers become optically thick to Ly α photons. However, the 2-D radiative transfermodels described above only include the thermal population of H ∗ and do not account for FUV-pumped gas locatedat more distant radii. Distinguishing between the Ly α -limited and H ∗ -limited scenarios for UV-H emission, andsubsequently identifying disk regions where H ∗ -driven CN formation pathways can proceed, will require a model thataccounts for both the thermal and non-thermal gas populations. This relationship between the spatial distributionsof H ∗ and CN will therefore be explored in more detail in a forthcoming paper. SUMMARY & CONCLUSIONS
We have analyzed the ultraviolet spectral properties of 19 young stars in the Lupus and Taurus-Auriga associations,using spectra from
HST -COS and
HST -STIS to directly measure fluxes from Ly α , the FUV continuum, semi-forbiddenC II] λ . Each of these is a potential tracer of the photochemical pathways responsible forproducing CN and HCN molecules in disks. To investigate the formation chemistry of these two species, we comparethe UV tracers to sub-mm CN and 14 µ m HCN fluxes. We find that1. HCN fluxes are positively correlated with relative fluxes from the FUV continuum and C II] λ + reactants required inthe first step of the main HCN formation pathway.olecular Gas Distributions in Lupus Disks 172. By contrast, CN fluxes are negatively correlated with relative fluxes from the FUV continuum and C II] λ α emission. However, we report very tentativepositive (CN) and negative (HCN) correlations that are consistent with modeling work that predicts increasedphotodissociation with stronger Ly α irradiation.We attribute the lack of correlations between CN and HCN emission and UV-H fluxes to the spatial distributions of thethree molecular species: the UV-H is concentrated in surface layers of the inner disk, the sub-mm CN emission extendsfrom the inner to the outer disk, and the HCN emission originates in deeper layers of the inner disk. By combiningUV spectra with IR and sub-mm fluxes from UV-dependent molecular gas species, we are able to investigate modelpredictions of molecule formation pathways and observationally confirm that the FUV continuum plays an importantrole in regulating CN and HCN populations in protoplanetary disks. The analysis presented here can be extended toadditional species (e.g. hydrocarbons) in the era of JWST , which will enable higher spectral resolution observationsof warm molecular gas and more accurate physical-chemical models of surface reactions in planet-forming systems. ACKNOWLEDGEMENTS
We are thankful to the referee for their thoughtful comments that helped strengthen the analysis presented here.NA is supported by NASA Earth and Space Science Fellowship grant 80NSSC17K0531 and HST-GO-14604 (PIs:C.F. Manara, P.C. Schneider). H.M.G. was supported by program HST-GO-15204.001, which was provided by NASAthrough a grant from the Space Telescope Science Institute, which is operated by the Associations of Universities forResearch in Astronomy, Incorporated, under NASA contract NAS5-26 555. We are grateful to M.K. McClure and C.Walsh for helpful discussions regarding the analysis presented here. This paper makes use of the following ALMAdata: ADS/JAO.ALMA a community-developed core Python package for Astronomy (Astropy Collaborationet al. 2013; The Astropy Collaboration et al. 2018). APPENDIX
UV-H SPECTRA AND MODELING RESULTS
We present the nine strongest observed UV-H emission lines for RU Lupi, MY Lupi, Sz 68, and J1608-3070,with “best-fit” models from the 2-D radiative transfer approach described in Section 4.2 overlaid on the data. Thecorresponding model radial distributions of UV-H flux for each system, with contours representing + / − − σ bounds,are shown as well. REFERENCES´Ad´amkovics, M., Najita, J. R., & Glassgold, A. E. 2016, ApJ,817, 82Ag´undez, M., Cernicharo, J., & Goicoechea, J. R. 2008, A&A,483, 831Ag´undez, M., Roueff, E., Le Petit, F., & Le Bourlot, J. 2018,A&A, 616, A19Aikawa, Y., van Zadelhoff, G. J., van Dishoeck, E. F., & Herbst,E. 2002, A&A, 386, 622Alcala’, J. M., Manara, C. F., France, K., Schneider, C. P.,Arulanantham, N., Miotello, A., Guenther, H. M., & Brown, A.2019, arXiv e-printsAlcal´a, J. M., et al. 2017, A&A, 600, A20—. 2014, A&A, 561, A2Andrews, S. M., et al. 2018, ApJ, 869, L41Andrews, S. M., Rosenfeld, K. A., Kraus, A. L., & Wilner, D. J.2013, ApJ, 771, 129Ansdell, M., Williams, J. P., Manara, C. F., Miotello, A.,Facchini, S., van der Marel, N., Testi, L., & van Dishoeck, E. F.2017, AJ, 153, 240Ansdell, M., et al. 2018, ApJ, 859, 21—. 2016, ApJ, 828, 46Ardila, D. R., et al. 2013, ApJS, 207, 1Arulanantham, N., et al. 2018, ApJ, 855, 98Astropy Collaboration et al. 2013, A&A, 558, A33 . . . . . g a mm a T q
481 21 62 0 r c h a r z . . . . . M H . . . . . gamma T q rchar . . . . . MH2
Fig. A.1.—
Corner plot showing marginalized probability distributions of the variable parameters, derived from MCMC re-sampling ofthe 2-D radiative transfer models described in Section 4.1. This set of distributions was acquired by calculating log-likelihood values usingthe observed fluxes from the MY Lupi spectrum.Dent, W. R. F., et al. 2013, PASP, 125, 477Espaillat, C. C., Robinson, C., Grant, S., & Reynolds, M. 2019,arXiv e-printsForeman-Mackey, D., Hogg, D. W., Lang, D., & Goodman, J.2013, PASP, 125, 306France, K., Roueff, E., & Abgrall, H. 2017, ApJ, 844, 169France, K., Schindhelm, E., Bergin, E. A., Roueff, E., & Abgrall,H. 2014, ApJ, 784, 127France, K., et al. 2011a, ApJ, 734, 31—. 2012, ApJ, 756, 171France, K., Yang, H., & Linsky, J. L. 2011b, ApJ, 729, 7Ghez, A. M., McCarthy, D. W., Patience, J. L., & Beck, T. L.1997, ApJ, 481, 378G´omez de Castro, A. I., & Ferro-Font´an, C. 2005, MNRAS, 362,569Green, J. C., et al. 2012, ApJ, 744, 60Greenwood, A. J., Kamp, I., Waters, L. B. F. M., Woitke, P., &Thi, W.-F. 2019, A&A, 626, A6Guilloteau, S., Di Folco, E., Dutrey, A., Simon, M., Grosso, N., &Pi´etu, V. 2013, A&A, 549, A92Hartigan, P., & Kenyon, S. J. 2003, ApJ, 583, 334Heays, A. N., Bosman, A. D., & van Dishoeck, E. F. 2017, A&A,602, A105Henning, T., et al. 2010, ApJ, 714, 1511Herczeg, G. J., Linsky, J. L., Valenti, J. A., Johns-Krull, C. M.,& Wood, B. E. 2002, ApJ, 572, 310Herczeg, G. J., et al. 2005, AJ, 129, 2777Herczeg, G. J., Wood, B. E., Linsky, J. L., Valenti, J. A., &Johns-Krull, C. M. 2004, ApJ, 607, 369Hoadley, K., France, K., Alexander, R. D., McJunkin, M., &Schneider, P. C. 2015, ApJ, 812, 41 Hoadley, K., France, K., Arulanantham, N., Loyd, R. O. P., &Kruczek, N. 2017, ApJ, 846, 6Houck, J. R., et al. 2004, ApJS, 154, 18Huang, J., et al. 2018, ApJ, 869, L42Ingleby, L., et al. 2013, ApJ, 767, 112Johns-Krull, C. M., & Herczeg, G. J. 2007, ApJ, 655, 345Kenyon, S. J., & Hartmann, L. 1995, ApJS, 101, 117Kessler-Silacci, J., et al. 2006, ApJ, 639, 275Kurtovic, N. T., et al. 2018, ApJ, 869, L44Lebouteiller, V., Barry, D. J., Goes, C., Sloan, G. C., Spoon,H. W. W., Weedman, D. W., Bernard-Salas, J., & Houck, J. R.2015, The Astrophysical Journal Supplement Series, 218, 21Lebouteiller, V., Barry, D. J., Spoon, H. W. W., Bernard-Salas,J., Sloan, G. C., Houck, J. R., & Weedman, D. W. 2011, ApJS,196, 8Li, X., Heays, A. N., Visser, R., Ubachs, W., Lewis, B. R.,Gibson, S. T., & van Dishoeck, E. F. 2013, A&A, 555, A14Long, F., et al. 2019, ApJ, 882, 49—. 2018, ApJ, 863, 61Loomis, R. A., ¨Oberg, K. I., Andrews, S. M., & MacGregor,M. A. 2017, ApJ, 840, 23Loyd, R. O. P., et al. 2018, ApJ, 867, 71Madhusudhan, N., Mousis, O., Johnson, T. V., & Lunine, J. I.2011, ApJ, 743, 191Manset, N., Bastien, P., M´enard, F., Bertout, C., Le van Suu, A.,& Boivin, L. 2009, A&A, 499, 137McClure, M. K. 2019, arXiv e-printsMcElroy, D., Walsh, C., Markwick, A. J., Cordiner, M. A., Smith,K., & Millar, T. J. 2013, A&A, 550, A36McJunkin, M., France, K., Schneider, P. C., Herczeg, G. J.,Brown, A., Hillenbrand, L., Schindhelm, E., & Edwards, S.2014, ApJ, 780, 150 olecular Gas Distributions in Lupus Disks 19 - - F l u x [ - e r g s - c m - Å - ] [1,7]: 1467.08 Å - - Velocity [km/s] - - Fig. A.2.—
Observed UV-H emission lines (black) and model distributions (teal) for MY Lupi. Dashed lines show the model profileprior to convolution with the HST -COS line spread function (LSF), while the solid lines are used to show the convolved line profile. -1 Radius [au] F H ( r ) [ - e r g c m - s - ] MYLup 1- Uncertainties
Fig. A.3.—
Best-fit model radial distribution of UV-H flux from the disk around MY Lupi, with contours marking + / − σ boundson the median distribution. Uncertainties were estimated using MCMC sampling (Foreman-Mackey et al. 2013) over the parameter spacedefined by Hoadley et al. (2015). - - Velocity [km/s] F l u x [ - e r g s - c m - Å - ] [1,7]: 1467.08 Å - - - - Fig. A.4.—
Observed UV-H emission lines (black) and model distributions (teal) for Sz 68. Dashed lines show the model profile prior to convolution with the HST -COS line spread function (LSF), while the solid lines are used toshow the convolved line profile. -1 Radius [au]0 F H ( r ) [ - e r g c m - s - ] Sz68 1- Uncertainties
Fig. A.5.—
Best-fit model radial distribution of UV-H flux from the disk around Sz 68, with contours marking + / − σ bounds on themedian distribution. Uncertainties were estimated using MCMC sampling (Foreman-Mackey et al. 2013) over the parameter space definedby Hoadley et al. (2015). olecular Gas Distributions in Lupus Disks 21 - - F l u x [ - e r g s - c m - Å - ] [1,7]: 1467.08 Å - - Velocity [km/s] - - Fig. A.6.—
Observed UV-H emission lines (black) and model distributions (teal) for J1608-3070. Dashed lines show the model profileprior to convolution with the HST -COS line spread function (LSF), while the solid lines are used to show the convolved line profile. -1 Radius [au] F H ( r ) [ - e r g c m - s - ] TYC7851 1- Uncertainties
Fig. A.7.—
Best-fit model radial distribution of UV-H flux from the disk around J1608-3070, with contours marking + / − σ bounds onthe median distribution. We note that the MCMC sampling (Foreman-Mackey et al. 2013) used to estimate the uncertainties on the fluxdistribution was carried out over a tighter parameter space than for the other two targets, since the signal-to-noise in the UV-H emissionlines is lower for J1608-3070.2 Arulanantham et al.