Multiple nitrogen reservoirs in a protoplanetary disk at the epoch of comet and giant planet formation

The isotopic ratio of nitrogen measured in primitive Solar System bodies shows a broad range of values, the origin of which remains unknown. One key question is whether these isotopic reservoirs of nitrogen predate the comet formation stage or are posterior to it. Another central question is elucidating the processes that can produce the observed variations in the 14N/15N isotopic ratio. Disks that orbit pre-main-sequence stars provide unique opportunities for observing the chemical content of analogs of the protosolar nebula and therefore for building a comprehensive scenario for the origin of nitrogen in the Solar System and in planet-forming disks. We present spectrally and spatially resolved observations of the hyperfine structure of the 4-3 rotational transition of HCN and its main isotopologs H13CN and HC15N in the disk orbiting the 8 Myr old TTauri star TW Hya. The spatially averaged HCN/H13CN and HCN/HC15N abundance ratios are 86+/-4 and 223+/-21, respectively. The latter value is significantly lower than the CN:C15N ratio 323+/-30 in this disk and thus provides the first evidence that two isotopic reservoirs of nitrogen are present in a disk at the stage of giant planet and comet formation. Furthermore, we find clear evidence for an increase in the ratio of HCN to HC15N with radius. The ratio in the outer disk is 339+/-28, in excellent agreement with direct measurements in the local interstellar medium, and with the bulk nitrogen isotopic ratio predicted from galactic evolution calculations. In the comet formation region at r=20 au, the ratio is a factor ~3 lower, 121+/-11. This radial increase qualitatively agrees with the scenario in which selective photodissociation of N 2 is the dominant fractionation process. However, our isotopic ratios and kinetic temperature of the HCN-emitting layers quantitatively disagree with models of nitrogen chemistry in disks.


Introduction
The isotopic ratio of nitrogen, 14 N/ 15 N, in various bodies of the Solar System (meteorites, comets, planets, etc.) shows the largest variations among the most abundant constituents (carbon and oxygen), with values ranging from ∼50 in submicron grains that are immersed in a chondrite matrix (so-called hotspots) to 441 in the Sun and Jupiter (Bonal et al. 2010;Marty et al. 2011;Fouchet et al. 2000;Hily-Blant et al. 2013). When, where, and how these isotopic reservoirs of nitrogen formed during the evolution from a molecular cloud to planetary systems remains to be determined. Comets provide a special case among primitive cosmomaterials: regardless of the cometary type (short or long period) and of the carrier of nitrogen that is observed in their coma (CN, HCN, NH 2 , NH 3 , N 2 , or NO), the isotopic ratio is consistently found to be ≈140 (Wampfler et al. 2018;Hily-Blant et al. 2017, hereafter Paper I, and references therein). This is lower by a factor three than the ratio of the bulk proto-Sun, which is 441. This well-established observational fact remains unexplained.
Evolved protoplanetary disks at the stage of giant planet and comet formation provide unique opportunities for investigating how and when this 15 N-rich reservoir of nitrogen was built up. The nitrogen isotopic ratio was measured via rare isotopologs 1 of HCN, specifically, H 13 CN and HC 15 N, for a 1 Hereafter, the three isotopologs H 12 C 14 N,H 13 C 14 N,and H 12 C 15 N,are denoted HCN,H 13 CN,and HC 15 N,respectively. small sample of disks with the ALMA interferometer (Guzmán et al. 2017), yielding an HCN/HC 15 N between 83±32 and 156±71, and a mean value of 111±19, under the assumption that HCN/H 13 CN=70. On the other hand, a direct measurement, one that does not involve assuming a 12 C/ 13 C ratio, of CN/C 15 N in the TW Hya system yielded a significantly higher ratio, 323±30 (Paper I). Based on this value and on directly measured values in the dense local interstellar medium (ISM), a value of 14 N/ 15 N=330±30 was proposed for the present-day bulk solar neighborhood. This also agrees excellently well with predictions from the galactic chemical evolution model of Romano et al. (2017) for the local ISM. A new picture thus emerged in which HCN and CN trace two isotopic reservoirs of nitrogen that are present in analogs of the protosolar nebula (PSN) at the time of comet formation, with HCN representing a 15 N-enriched reservoir of nitrogen similar to that recorded in Solar System comets. However, HCN/HC 15 N was only measured indirectly, and moreover, this was not done in TW Hya, therefore clear-cut evidence for this scenario was lacking.
To elucidate how the various nitrogen reservoirs in the Solar System were formed requires identifying the processes that lead to variations of 14 N/ 15 N (Hily-Blant et al. 2013Füri & Marty 2015). The efficiency of chemical mass fractionation, which is the only process expected in cold and shielded gas (Terzieva & Herbst 2000), has recently been called into ques-  HCN,H 13 CN,and HC 15 N (from left to right) in Jy/beam km s −1 . The maps are centered at α=11:01:51.81150, δ=-34:42:17.2636 (J2000.0). The visibilities were projected on the same spatial grid and convolved to a synthesized beam of HPBW 0.26 ×0.21 . The emission is centered on the systemic velocity of 2.83 km s −1 and is integrated over 5 km s −1 for HCN and 2 km s −1 for the less abundant isotopologs. tion both theoretically (Roueff et al. 2015;Wirström & Charnley 2018) and observationally by measuring HCN/HC 15 N in the L1498 pre-stellar core (Magalhães et al. 2018). The obtained value, 338±28, agrees very well with the present-day elemental ratio of ∼330 proposed in Paper I and thus supports the prediction that chemical fractionation is not an efficient process even in cold pre-stellar cores. Further credence was provided by subsequent direct measurements in HC 3 N and HC 5 N toward similar sources Taniguchi & Saito 2017). The concordance between observations and models may be only an apparent one, however, because sharp deviations are recurrently obtained for N 2 H + (Bizzocchi et al. 2013;Redaelli et al. 2018).
The other fractionation process, selective photodissociation of N 2 , which is expected in unshielded gas, has been investigated in the context of protoplanetary disks and molecular clouds (Heays et al. 2014;Visser et al. 2018;Furuya & Aikawa 2018). The opacity of N 15 N is far lower than that of N 2 , therefore more 15 N atoms are released upon dissociation than 14 N. This leads to an atomic isotopic ratio that is lower than the bulk. Gas-phase and/or surface chemistry then propagates this deviation into other species. A correlation (positive or negative) between the UV flux and the 14 N/ 15 N is therefore expected in some species. For HCN in disks, the HCN/HC 15 N is predicted to increase with decreasing UV flux.
In this Letter, we aim at providing the requisite evidence for two distinct N reservoirs within the disk that orbits the nearby TTauri star TW Hya through an analysis of the emission of HCN and isotopologs, based on archival ALMA data (2016.1.00629.S, Cleeves et al). We also explore the possibility of a radial gradient in the HCN/HC 15 N to determine whether selective photodissociation is the driving fractionation process in PSN analogs.

Observations and data reduction
The (4-3) rotational transitions of HCN,H 13 CN,and HC 15 N at 354.5,345.4,and 344.2 GHz, respectively, have been mapped with the ALMA interferometer with a spectral resolution of 61 KHz or 0.053 km s −1 (see Appendix A). All the lines were observed simultaneously, thus mitigating cross-calibration biases. The standard pipeline of CASA 5.4.0 was used to generate the complex visibilities, which were then self-calibrated in Gildas 2 . The final sensitivities are 2.3, 1.9, and 2.0 mJy/beam per 61 kHz channel for HCN, H 13 CN, and HC 15 N respectively. The resulting integrated intensity maps are shown in Fig. 1.
The HCN(4-3) spectra (see Fig. 2) contain two weak hyperfine (hf) transitions of equal intensity, each carrying 2.1% of the total flux, and shifted by -1.672 and +1.360 km s −1 with respect to our reference frequency 354505.477 MHz. They surround a strong, non-Gaussian, central feature made of three overlapping hf transitions representing 95.8% of the flux. The intrinsically weaker intensity of H 13 CN and HC 15 N compared to HCN leads to one single feature only, which in the case of H 13 CN carries 95.8% of the total flux. For HC 15 N, the splitting is entirely unresolved and the single feature thus carries 100% of the rotational transition flux (see Table A.1).

Results
The conversion of line flux into abundance ratios is a simple procedure provided the lines are optically thin and the excitation temperatures are known. Even for optically thin transitions, however, the overlap in velocity space of lines that are emitted at different locations in the disk, which is a result of the Keplerian rotation of the disk, complicates the conversion from flux into abundance ratios because this mixes different regions of the disk, broadens the lines, and may also weaken the assumption that the emitting species are cospatial. The latter is especially true in the case of species with very different abundances, such as nitrogen isotopologs. Mitigating both the optical depth and Keplerian blurring effects thus offers clear advantages.
The nearly face-on configuration of the TW Hya disk naturally reduces the Keplerian blurring. In addition, the contribution from different locations along the line of sight to the emerging lines is limited to the vertical direction. Moreover, the spectrally resolved hf multiplet of HCN provides two weak lines that are most likely optically thin. Their flux is thus directly proportional to their column density.
Because the emission of all three isotopologs is well detected and spatially resolved, one goal might be measuring the HCN isotopic ratios as a function of radius, provided that line broadening can be minimized. This implies that the spatial area over which spectra are averaged needs to be limited. Another complementary approach consists of deriving a spatially averaged isotopic ratio by measuring line flux ratios over the whole disk. This method provides higher sensitivity, but Keplerian line broadening must still be controlled. We applied both approaches and detail them in the following.

Spatially averaged line fluxes
In order to limit the Keplerian line-broadening, the spectra were averaged over several groups of pixels that were selected for their closely similar projected Keplerian velocity. In practice, a 2D mask was thus determined by selecting the disk area where the 354507.455 MHz hyperfine emission of HCN is higher than 40 mJy/beam in each velocity channel. Although the overlap of different Keplerian velocities is well controlled, it is still noticeable close to the disk center, and we further masked out the spectra located at radii smaller than 0.3 . The resulting mask recovered the usual dipole-like pattern, but with a central hole due to the radial filter (see Fig. B.1). As a result, the spectra averaged within each Keplerian velocity mask have a high signal-to-noise ratio (see Fig. B.2), allowing the two weak hf transitions of HCN to be properly fitted. The line fluxes of the two weak hf lines at 354503.869 and 354507.455 MHz were measured through Gaussian fitting and their averaged value, corrected for their summed relative intensity (4.2%), was then used as an estimator of the HCN flux (see Table B.1).

Radially dependent fluxes
In order to measure the flux ratios as a function of distance from the central TTauri star and still avoid Keplerian line-broadening, a radial mask was superimposed on the previously described Keplerian masks. To calculate disk radial offsets in au, we adopted a distance of 59.5±0.9 pc to TW Hya (Gaia Collaboration 2018). Spectra contained within concentric annuli, centered at radii r = 5 to 55 au in steps of 5 au (one-third of the half-power beam width, HPBW), and of constant thickness δr =5 au, were averaged. The outer radius was imposed by the HC 15 N detection limit (see Fig. 1). The number of spectra in each annulus increased with r, which compensated for the natural decrease of the signal-to-noise ratio with radius. In the process, spectra were weighted according to their flux rms. As in the spatially averaged approach, the line fluxes of HCN, H 13 CN, and HC 15 N were measured through Gaussian fitting.
Radially averaged spectra corresponding to the Keplerian mask at 2.83 km s −1 are shown in Fig. 2. The velocity confusion at small radii is evident, while at radii above 20 au, it is efficiently removed by our masking procedure. Average spectra in all Keplerian channels are displayed in Fig. B.3. While the peak intensity of HCN hardly changes beyond 20 au, the peak intensities of H 13 CN and HC 15 N decrease when moving outward. Also, all lines become narrower as the radius increases (see Table B.3). For HCN, the FWHM is 1.5±0.7 km s −1 at r ≤ 0.2 , decreasing to 0.34±0.02 km s −1 at 0.9 .

Isotopic ratios
Because the three molecules have very similar spectroscopic constants, their rotational flux ratios are equal to their column Table 1. Isotopic ratios derived from our two approaches, the spatially averaged and the radially dependent methods (see also Fig. 3 density ratios to within 1%, provided the lines are optically thin and have the same excitation temperature. Under these assumptions, which we verify below, we obtain the isotopic ratios directly from the fluxes. The results are shown in Fig. 3 and are listed in Tables 1 and B.2. For HCN/HC 15 N, we find ratios between 185 to 252 across the Keplerian channels (typical uncertainty of 5-10%) with an average isotopic ratio 223 ± 21. Here and elsewhere, uncertainties on the absolute flux include a conservative 10% error (statistical) from amplitude calibration. The quoted uncertainty on average values is the largest of the weighted uncertainties and the dispersion of the averaged values. The lowest values (185, 186, and 198) are obtained in the three central Keplerian channels, and the average ratio in the eight remaining channels is 233±12. HCN/H 13 CN is more uniform, with values ranging from 82 to 92, which averages to 86±4. The corresponding average of H 13 CN/HC 15 N is 2.6±0.2, with values ranging from 2.22 to 2.75. The low values of HCN/HC 15 N might arise because the opacity of HCN is substantial even in its weak hf lines. However, the almost constant HCN/H 13 CN suggests that these lines are instead optically thin. Still, assuming a constant isotopic ratio of 233 and adopting the usual expression τ/(1 − exp(−τ)) for the opacity correction for the integrated intensity, with τ the opacity (Goldsmith & Langer 1999), an opacity of 0.4 for the weakest hf lines, corresponding to a rotational opacity τ 12 ≈ 20, would be needed to explain the flux ratios in these three channels. Alternatively, assuming a constant H 13 CN/HC 15 N ratio of 2.7 based on the outer channels (see Table B.2), the total H 13 CN opacity would be τ 13 = 0.3 in the three middle channels. It is thus likely that the opacity of the weakest hf line of HCN(4-3) and of H 13 CN(4-3) is ≈ 0.3, which is not entirely negligible, but still fulfills our optically thin assumption. We note that in these Article number, page 3 of 9 A&A proofs: manuscript no. 36750corr Left: HCN/HC 15 N and HCN/H 13 CN isotopic ratios in each Keplerian channel derived from the spatially averaged procedure (see Sect. 3.1). The colored bands indicate the average value ±1σ. The spatially averaged CN/C 15 N ratio from Paper I is shown as a hatched rectangle. Right: Radial dependence (see Sect. 3.2) of the HCN/H 13 CN and HCN/HC 15 N ratios. Each box is 5 au wide (one-third of the HPBW) and ±1σ in height. The full line shows the best fit of a power law to the 15 N ratio, with a slope of 1.3. The gray shaded areas are the present-day 14 N/ 15 N and 12 C/ 13 C elemental ratios in the local ISM (330±30 (Paper I) and 65±10 (Halfen et al. 2017), respectively). three channels, the HCN/H 13 CN would change by less than 4% if these corrections were applied.
The radially dependent approach shows that the HCN-to-HC 15 N isotopic ratio varies with radius while HCN/H 13 CN remains constant within 1σ. More specifically, the former ratio increases from 121 at 20 au to 339 at 45 au, and may decrease at larger distances. The nitrogen isotopic ratio in HCN close to the star is thus a factor 2.7 lower than the spatially averaged CN/C 15 N, 323±30, whereas between 45 to 50 au, the two ratios become consistent within 1σ. The ratios change by less than 4% when the three central Keplerian channels are excluded. The best fit of a power law to the HCN/HC 15 N gives 117±10×(r/20 au) 1.3±0.13 , indicating a shallower dependence with radius than a pure geometrical dilution of the UV flux from the central star (∝ r −2 ).
A key assumption in our analysis is that a single excitation temperature characterizes the weak hf lines of HCN and the overlapping hf lines of H 13 CN and of HC 15 N. However, line overlap may lead to unequal excitation temperatures among the hf manifold of HCN (Magalhães et al. 2018). We have tested our assumption (see Appendix C) by computing the excitation temperatures of HCN, H 13 CN, and HC 15 N lines while taking line overlap in the excitation into account. Our calculations assume uniform physical conditions and cover the density and gas temperature regimes from 10 6 to 10 11 cm −3 and 5 to 30 K, respectively. We found that deviations from a single excitation temperature are only obtained at densities below 3 ×10 7 cm −3 (see Fig. C.1). On the other hand, the full width at half-maximum (FWHM) of the HCN, H 13 CN, and HC 15 N lines at a radius of 0.6 (≈35 au) and beyond is ≈0.35 km s −1 , which places an upper limit on the kinetic temperature of (FWHM/2.35) 2 × µm H /(2k B ) ≈35 K (Teague et al. 2016). Referring to the TW Hya model of van't Hoff et al. (2017), this upper limit indicates that the density is at these radii higher than 10 9 cm −3 , which means that it is in a regime where lines are expected to be thermalized. This validates our basic assumption. Our upper limit on the kinetic temperature is also well below the gas temperature of ∼ 200 K where the HCN abundance is the highest in the generic models of Visser et al. (2018).

Discussion
The spatially averaged directly measured CN/C 15 N and HCN/HC 15 N in the TW Hya circumstellar disk provide clearcut confirmation of at least two isotopic reservoirs of nitrogen, as anticipated by Paper I. In CN, the ratio is 1.4±0.2 times higher than in HCN, and their difference is 90±36. Based on the ∼8 Myr age of TW Hya, these results indicate that multiple isotopic reservoirs are present at the key protoplanetary disk evolutionary stages when comets and giant planets are likely forming.
In the comet formation zone between 20 and 30 au, the HCN/HC 15 N is 121±11, which is a factor of 2.8 compared to the outer disk at 55 au, where the isotopic ratio of HCN matches the revised elemental ratio of 330 well. The 15 N-enrichment in the region inside of 30 au is reminiscent of the threefold 15 Nenrichment of comets compared to the bulk proto-Sun, providing a strong indication that HCN traces the fractionated reservoir that is recorded by comets. Guzmán et al. (2017) also found marginal evidence for an increase of the HCN-based 14 N/ 15 N ratio with radius in the inner ∼50 au of the V4046 Sgr disk. We also note that the inner disk ratio confirms the mean HCN/HC 15 N ratio of 111±19 (Paper I) that was obtained indirectly by Guzmán et al. (2017) assuming HCN/H 13 CN=70±20. Adopting our measured averaged of 86 for the latter would bring the former ratio to 136±23. We also very tentatively observe a decrease of the ratio beyond 55 au, which could indicate chemical processes in a gas not exposed to strong UV radiation and where depletion is limited. Measurements with a high signal-to-noise ratio beyond 55 au are needed to confirm this behavior, however.
The average HCN/H 13 CN, 86±4, is significantly higher than the 12 C/ 13 C ratio of 65 ± 10 in the local ISM (Halfen et al. 2017), but agrees surprisingly well with the value of 89±1 in the PSN (Clayton & Nittler 2004). This indicates that HCN is slightly depleted in 13 C in TW Hya, opposite to what was found in the L1498 pre-stellar core (Magalhães et al. 2018), suggesting that the underlying process may not be the same in these two environments. The depletion of HCN in 13 C is at variance with the models of Visser et al. (2018), however, which instead predict that HCN is enriched in 13 C at radii below ∼100 au. Optically thick H 13 CN emission could artificially increase the HCN/H 13 CN provided that the weak HCN hf transition remains thin. However, the H 13 CN opacity required to decrease the ratio to the elemental value is 0.55, which is higher than the value of 0.3 derived previously for the central Keplerian channels. This also implies that the total opacity of HCN would be 35, or 0.75 for the weakest HCN hf, which is twice higher than the value of 0.3 that is required to bring the spatially averaged HCN/HC 15 N to a constant value (see Sect. 4). We note that high 12 C/ 13 C ratios have already been measured in carbon monoxide from ices toward young stellar objects (Smith et al. 2015).
The increasing HCN/HC 15 N with radius may indicate that selective photodissociation represents the dominant fractiona-tion process in disks (Visser et al. 2018). In these models, the HCN-to-HC 15 N column density ratio first decreases with radius before it increases back to the elemental ratio as a consequence of UV flux attenuation that is due to a combination of geometrical dilution and dust extinction. We note that the minimum ratio in the grid of models of Visser et al. (2018) is ≈240-280 for disks around TTauri stars, which would translate into 180-210 when 330 (rather than 441) is assumed for the elemental ratio. This is still higher than the lowest ratio reported here and might indicate that the UV flux is underestimated in these models. However, dedicated models of TW Hya are required because the precise radial dependence of the isotopic ratio strongly depends on the physical parameters of the model (disk mass, grain size, and dust-to-gas mass ratio).
Overall, the increasing HCN/HC 15 N from 121 up to the bulk value in the outer disk supports a scenario in which a single reservoir of interstellar, not fractionated, nitrogen dominates at large radii in PSN analogs. This nitrogen reservoir is enriched in 15 N by the selective photodissociation of N 2 at smaller distance from the central star.  H 2 density ( 10 6 to 10 11 cm −3 ) and kinetic temperature (from 5 to 30 K) relevant to the TW Hya orbiting disk. The results of these calculations in terms of T ex are shown in Fig. C.1. There are significant departures at densities between 10 6 cm −3 and 10 8 cm −3 . At higher density, all lines are thermalized (models at n H 2 higher than 10 9 cm −3 are not shown). From the generic chemical models of Visser et al. (2018), it appears that HCN is concentrated at radii r = 5 − 50 au and scale heights z/r = 0.2 to 0.3, where the gas kinetic temperature is ≈ 200K and the density in the range 10 6 to 10 8 cm −3 , which would indicate that the single excitation temperature assumption fails.
On the other hand, the FWHM of the HCN, H 13 CN, and HC 15 N lines at a radius of 0.6 (≈35 au) and beyond is ≈ 0.35 km s −1 (see Table B.3), placing a conservative upper limit on the kinetic temperature of (Teague et al. 2016) (FWHM/2.35) 2 × µm H /(2k B ) ≈ 35 K. (C.1) Referring to the tailored TW Hya model of van't Hoff et al. (2017), this upper limit, applied to the relevant range of radii, indicates that the density is higher than 10 9 cm −3 , in a regime where lines should be thermalized. This validates our single excitation temperature assumption. Our upper limit on the kinetic temperature is also well below the gas temperature of ∼ 200 K where the HCN abundance is the highest in the models of Visser et al. (2018), although we note that these models are generic and are not tailored to TW Hya.     (12) 0.16 870.9(16.9) 3.010(1) 9.6(2) 2.988(4) 3.6(2) 2.990(10) 0.21 881.2(16.1) 3.067(1) 9.6(2) 3.035(3) 3.5(2) 3.048( 9) 0.27 870.2(15.5) 3.117(1) 9.6(2) 3.080(4) 3.5(2) 3.113(10) 0.32 859.7(14.2) 3.166(1) 9.8(2) 3.138(3) 3.7(2) 3.165 (11) NOTE.-Statistical uncertainties (at the 1σ level) are written in parentheses in units of the last digit. § Keplerian minus systemic velocity ( km s −1 ), W is the integrated flux in mJy/beam km s −1 , and v the central velocity in km s −1 . a For HCN, the integrated intensity is computed as the total flux of the two weak lines divided by the sum of their relative intensities, 4.2%, while the center velocity is that of the main central component. b,c The hf lines are not separated, and a single Gaussian was fit to the the main line.