XUE: Molecular Inventory in the Inner Region of an Extremely Irradiated Protoplanetary Disk

XUE 1 is part of a triple system (see Appendix A) where the third star (B) is at a distance of 15 (∼2600 au) from a binary system (A1+A2) whose components are ≈02 (∼300 au) apart. Based on the MIRI imaging, we conclude that A1 is the star with the PPD giving rise to the observed MIRI spectrum (see Appendix A).

We derive the masses of XUE 1 A1 and XUE 1 A2 from their extinction-corrected z-band magnitudes (z_A1 ≃ 17.9 mag and z_A2 ≃ 18.9 mag). We assume that both components are subject to the same extinction A_V ∼ 8.7 mag as derived for the unresolved object A using UKIRT photometry (Broos et al. 2013; King et al. 2013). The z band is not much affected by the presence of accretion/disks, and assuming a reddening law of A_z /A_V = 0.505 (Rieke & Lebofsky 1985), the comparison with the PARSEC 1.2S models (Chen et al. 2014; Tang et al. 2014; Chen et al. 2015) for the 0.7 Myr isochrone gives the masses M_A1 ≈ 1.1 M_⊙ and M_A2 ≈ 0.6 M_⊙ for a distance of 1690 pc.

XUE 1 was observed on 2022 August 3 as part of the XUE project in Cycle 1 (GO-1759; Ramirez-Tannus et al. 2021) with the MIRI MRS (Rieke et al. 2015; Wells et al. 2015; Wright et al. 2015). All three wavelength settings (SHORT, MEDIUM, and LONG) are used for the observations. A four-point dither optimized for a point source was performed in the negative direction. The observations are obtained in the FASTR1 readout mode with 40 groups per integration and 2 integrations per dither position. This results in a total exposure time of 900 s per wavelength setting and 2700 s in total. No dedicated background observations were taken for this observation. The background in the observation is dominated by the emission of the surrounding H ii region and photodissociation region and spatially highly variable.

We reduced the data using the JWST Pipeline version 1.9.4, with context 1046 of the Calibration Reference Data System (CRDS; Bushouse et al. 2023). The Detector1 part of the pipeline was run without modifications. In Spec2 we implemented a custom background subtraction. After the WCS assignment, we performed a background subtraction by removing the two nodding positions from each other. This minimizes the effect of the variable background with respect to taking a dedicated background observation. We then run the standard steps in Spec2, including the residual fringe correction. In Spec3, we skipped the outlier rejection step, as it causes artifacts in the extracted spectrum owing to the undersampling of the point-spread function. The final cubes are combined using the drizzle algorithm.

Both components A (A1+A2) and B are visible in the MRS data cubes, but the two components in source A are not resolved. To identify the disk-bearing source, we used archival Hubble Space Telescope (HST)/Advanced Camera for Surveys (ACS) imaging from the Hubble Legacy Archive and the simultaneous MIRI F560W imaging corresponding to our program (see Appendix A). We measured the position of source A in the datacube and ran the extract1d step with the position of the source as input. We extracted the source with a modified aperture 2 times smaller than the default value, linearly increasing from 03 at 5 μm to 128 at 22 μm. We use the aperture corrections corresponding to this extraction aperture. The full 4.9–22 μm spectrum of XUE 1 is shown in Figure 1.

Zoom In Zoom Out Reset image size

We detect the P-branch CO fundamental emission between 4.9 and 5.35 μm with upper J levels between 24 and 46 for the v = 1 − 0 transitions and between 19 and 41 for the v = 2 − 1 transitions. Using the rotational diagram method (Goldsmith & Langer 1999), we estimate the excitation temperature and column density of the CO molecule. We fit the unblended lines with Gaussian profiles to find their flux (see Table 2 in Appendix. C.1.1 for the measured fluxes). We then calculate their specific intensity by assuming that the emission originates from a disk annulus with a radius of 0.5 au (see below) and a distance of 1690 pc. The rotation diagrams in Figure 2 are consistent with straight lines, suggesting LTE and optically thin conditions. We derive excitation temperatures of T_ex = 2883 ± 276 K for the v = 1−0 bands and of T_ex = 2606 ± 568 K for the v = 2−1 bands. A different choice of emitting area and/or distance would only affect the absolute values for the column densities but does not change the temperature estimate or the relative column density of v = 1−0 with respect to v = 2−1.

Zoom In Zoom Out Reset image size

The fact that we can perform linear fits for each vibrational level indicates that the rotational level populations within each vibrational level are in LTE. However, the measured column densities of the v = 2−1 bands are lower than those of v = 1−0, indicating that collision partner densities are not high enough to significantly populate the v = 2−1 level, requiring non-LTE level populations (see also Appendix C).

We fit radex non-LTE slab models (van der Tak et al. 2007) to the v = 1 − 0 and v = 2 − 1 bands simultaneously, including blended lines, following the methodology described in Tabone et al. (2023) (see Appendix C.1.2 for more details). The best-fitting non-LTE model for CO is shown in the top panel of Figure 3, and the best-fitting parameters are listed in Table 1. It has a temperature of 2300 K, which is consistent with the results from the rotation diagrams. The best-fitting column density is N = 3 × 10¹⁷ cm⁻², and the emitting radius is R = 0.44 au. The density of collisional partners (see Appendix C) of the best-fitting model is n_<H> = 6 × 10¹² cm⁻³, which is close to the critical density for v = 1 − 0 (Thi et al. 2013). This indicates that collisions populate the rotational levels within the v = 0 according to LTE, but the densities are too low to bring the vibrational temperature to the gas temperature. The temperature is well constrained to values between 2000 and 2800 K, and N is well constrained to values between 10¹⁶ and 10¹⁸ cm⁻² (see χ² map, in Appendix D). However, it has been shown that including the low-energy lines, not present in the MIRI spectrum, could lead to higher column densities and lower temperatures (e.g., Banzatti et al. 2022). Near-IR (∼4.2–5 μm) observations are needed to better constrain the CO fundamental's physical parameters. In order to fit the CO lines to the slab models, we need a lower resolution than that of the MRS at 5 μm (Argyriou et al. 2023). The CO lines are spectrally resolved with a deconvolved FWHM of 51 km s⁻¹. Assuming that the emission originates in a disk annulus and Keplerian rotation, we can calculate the characteristic radius. We adopt a mass of 1 M_⊙ and, as the inclination is not known, assume a value of 60° for XUE 1. This results in a characteristic radius of 0.25 ± 0.09 au, which is in agreement with the best-fitting slab model. Assuming an edge-on disk would lead to a radius of 0.34 au, while an inclination of 30° would result in a 0.09 au radius.

Zoom In Zoom Out Reset image size

Table 1. Best-fit Model Parameters for the Studied Molecules

Species	N	T	R	${{ \mathcal N }}_{\mathrm{tot}}$	〈n〉H	N0.5 au	T0.5 au
	(cm−2)	(K)	(au)	(mol.)	(cm−3)	(K)
H2O [7 μm]	2.2 × 1018	975	0.13	2.6 × 1043	⋯	6.8 × 1016	1150
H2O [15 μm]	6.8 × 1019	550	0.46	1.0 × 1046	⋯	2.2 × 1019	575
HCN	2.2 × 1017	575	0.57	5.0 × 1043	⋯	4.6 × 1016	675
C2H2	2.2 × 1018	475	0.23	8.2 × 1043	⋯	3.2 × 1017	550
CO2	2.2 × 1014	450	5.30	4.3 × 1042	⋯	3.2 × 1016	450
CO	3.0 × 1017	2300	0.44	4.1 ×1043	6.0 × 1012

Note. Columns (2), (3), (4), (5), and (6) show the disk properties obtained when leaving the radius as a free parameter. Columns (7), and (8) show the N, and T for a radius of 0.5 au.

Download table as:  ASCIITypeset image

We fit the spectra of water around 7 μm and in the 13–17 μm region with LTE slab models (Tabone et al. 2023; see Appendix C.2 for details). The bottom panel of Figure 3 shows the best-fit model to the spectral region between 13 and 16.3 μm, and the best-fit model parameters for each molecule are listed in Table 1. The χ² maps resulting from the fitting procedure for each molecule are shown in Appendix D. From these maps it is possible to read the typical uncertainties on the best-fit parameters; the red, orange, and yellow lines show the 1σ, 2σ, and 3σ levels, respectively. In addition to the three free parameters (T, N, and R), we also list the total number of molecules (${{ \mathcal N }}_{\mathrm{tot}}=\pi {{NR}}^{2}$), which is well constrained for optically thin emission.

3.2.1. H₂O

3.2.2.  ¹²CO₂

3.2.3. C₂H₂ and HCN

We compare our measurements with Spitzer and newly obtained JWST observations of PPDs that are not strongly irradiated, i.e., in low-mass star-forming regions. Figure 4 shows the line luminosity of H₂O, measured from the 17 μm lines, against the HCN/H₂O, C₂H₂/H₂O, CO/H₂O, and ¹²CO₂/H₂O of XUE 1 in comparison with a sample of T Tauri stars located in nearby (<200 pc), young (1–3 Myr) star-forming regions from Banzatti et al. (2020, 2022) and the newly obtained JWST observations of isolated PPDs (Grant et al. 2023; Kóspál et al. 2023; Perotti et al. 2023). We find that the H₂O luminosity and the ¹²CO₂/H₂O ratio are at the high end of the distribution. However, The HCN/H₂O, C₂H₂/H₂O, and CO/H₂O ratios are higher than found in the T Tauri sample.

Zoom In Zoom Out Reset image size

High line luminosities of molecular species are a signpost of a high-energy input into the disk, which can be intrinsic, i.e., due to a high stellar luminosity and/or a high gas accretion rate, or extrinsic, i.e., a strong external irradiation field, or a combination of these effects. Since the line luminosities of XUE 1 are in the range of those observed in nonirradiated disks (see Appendix E; but note the high C₂H₂ and HCN line luminosities; see below), disk irradiation is not needed to explain our observations.

Perotti et al. (2023) compare the H₂O luminosity of a sample of disks from Banzatti et al. (2020, 2022) to the continuum spectral index n_13–30, which is a measure of the dust depletion of the inner disk. We follow this approach, but we calculate the n_13–25 from the Spitzer spectra from the Spexodisks database (Pontoppidan et al. 2010; Salyk et al. 2011; Banzatti et al. 2020) to allow direct comparison to our MIRI data. The right panel of Figure 4 shows that XUE 1 does not have a dust-depleted inner disk, which implies that the inner disk dust and gas content of XUE 1 is comparable to that of the average T Tauri disks in nearby low-mass star-forming regions.

In summary, we find no substantial evidence for an externally irradiated disk surface in XUE 1. This is surprising because models of disk spectra for continuous, irradiated disks show much stronger line enhancements than detected for XUE 1 (Antonellini et al. 2015). These strong line enhancements are the result of the entire disk surface being warm enough to contribute to the line emission. In the following we discuss two scenarios that can explain our observations, i.e., the disk is shielded from UV irradiation, or the disk is truncated.

A scenario in which XUE 1 is shielded from the UV radiation from the massive stars could explain why its spectral properties in the MIRI spectral range resemble those of nearby disks. The physical and chemical properties of the innermost parts of the disks traced by MIRI are dominated by the radiation from the central star. Therefore, if XUE 1 is shielded from the radiation, given the mass of the central star, there would be no reason to expect the MIRI MRS spectrum to be different from those of nearby regions. In the top panel of Figure 5 we show the extinction A_K reported by Ramírez-Tannus et al. (2020) for a sample of stars in Pis24 including the extinction measured for XUE 1 in this work. The extinction toward XUE 1 does not seem higher than that of the neighboring stars. This is consistent with the lack of a significant dust cloud between XUE 1 and the O-type stars in the Herschel maps. However, the resolution of Herschel is not high enough to detect small-scale molecular features.

Zoom In Zoom Out Reset image size

The bottom panel of Figure 5 shows the projected distribution of massive stars (T_eff ≥ 24,000 K) in Pis24 with respect to the position of XUE 1. The external FUV radiation felt by XUE 1 assuming a 2D projected distance is represented by the colors and width of the lines. The stars contributing the most to the FUV radiation are all located toward the left side of the plot. Given this asymmetric FUV radiation field, it is not possible to discard the presence of a coherent dust structure, like a filament, shielding XUE 1 from most of the external radiation.

Nevertheless, Getman et al. (2014) show that heavily embedded objects within clouds, such as XUE 1 (A_V ≳ 10 mag), may undergo an environmental transformation due to cloud dispersal. These objects can become less absorbed (A_V ∼ 5 mag) and thus be significantly influenced by the radiation emitted by nearby massive stars, on timescales of less than 0.5 Myr. Therefore, it is still possible that XUE 1 has been exposed to the UV radiation at some point during its evolution. It is also possible that XUE 1 and the massive stars are not exactly at the same distance. If that were the case, the FUV field felt by XUE 1 could be substantially lower than reported in Figure 5.

Disk truncation would result in a change of the SED. Given the wavelength coverage of the MIRI MRS spectrum and the lack of longer-wavelength data, disks as small as a few au would all show a similar spectrum in the MIRI data, i.e., poorly constraining the outer disk radius. The line emission provides much stronger constraints on the disk outer radius. Antonellini et al. (2015) predict the mid-infrared emission of externally irradiated disks to be much stronger than in isolated disks. This is because of an increase of the line-emitting area. If XUE 1's disk were truncated, it would not be possible to form a hot layer of gas extending beyond the inner disk, and therefore the emitting area would be similar to that observed in nearby disks, which is dominated by the radiation from the central star and by accretion heating. Therefore, a truncation of XUE 1's disk could explain the fact that we do not see extraordinarily strong emission lines.

The observed HCN and C₂H₂ line luminosities of XUE 1 are higher than in the Banzatti sample, while this is not the case for ¹²CO₂. We note that this is consistent with disk irradiation because in nonirradiated disks the C₂H₂ and HCN emission is more concentrated in the innermost disk, while ¹²CO₂ has a larger emitting area. In an irradiated disk C₂H₂ and HCN will therefore be relatively more enhanced. However, we note that other effects, such as a higher inner disk C/O ratio (perhaps caused by a lower inward pebble flux), may also result in high C₂H₂ and HCN line luminosities.

Disk truncation may also be traced by the presence or absence of polycyclic aromatic hydrocarbon (PAHs). For PAHs to emit in their infrared vibrational resonances, they need to be exposed to a UV radiation field. T Tauri stars in low-mass star-forming regions only rarely show PAH emission (Geers et al. 2006). This is most likely due to a lack of a strong enough stellar UV radiation field. Indeed, In the hotter Herbig Ae/Be stars PAHs are much more frequently detected, particularly in sources with a large dust disk gap and/or inner disk dust depletion (Acke et al. 2010). The PAH emission in Herbig stars is dominated by the outer disk surface layers (e.g., Lagage et al. 2006; Yoffe et al. 2023) and traces the spatial distribution of gas. If the disk of XUE 1 is exposed to a radiation field G₀ as high as indicated in the color bar of Figure 5, we expect the full gas disk surface to contribute to the PAH emission, because such radiation fields are similar to or exceed those provided by the central stars of disks surrounding Herbig stars. However, the background-subtracted spectrum of XUE 1 (Figure 1) does not show any evidence for PAH emission. The central star is probably too cool to produce sufficient UV photons. Assuming a similar abundance of PAHs in the disk of XUE 1 to that in disks surrounding Herbig stars, this suggests that any gaseous disk surrounding XUE 1 cannot be spatially very extended, unless the disk is shielded from UV radiation.

With the data at hand, it is possible to fully discard neither the extinction nor the disk truncation scenario. Nevertheless, given the evidence presented, we have a slight preference for the truncation scenario. Dedicated ALMA observations are needed to prove or disprove the hypothesis that XUE 1's disk is truncated. Additionally, we are performing a parameter study using the radiation thermochemical disk model ProDiMo (Woitke et al. 2009; Kamp et al. 2017), in which we study the effect of disk truncation and PAH abundance on the spectra of irradiated PPDs. With this study we will be able to better understand the MIRI MRS observations (S. Hernández et al. 2023, in preparation).

With this first observation of water and other molecules in the inner, terrestrial-planet-forming regions of a disk in one of the most extreme environments in our Galaxy, we show that conditions for (terrestrial) planet formation, routinely found in disks in low-mass star-forming regions, can also occur in massive star-forming regions. The fact that dust grain growth happens fastest in the inner disk (e.g., Birnstiel et al. 2012) and that there have been substructures detected in disks that could be caused by planets in systems at 0.5 Myr old (Sheehan & Eisner 2018; Segura-Cox et al. 2020) leaves the possibility that planet formation has already happened or has gone a long way at the age of XUE 1 (∼1 Myr). Therefore, the fact that XUE 1's disk might be truncated does not necessarily prohibit planet formation.

However, it is reasonable to expect different masses and orbital periods for the XUE 1's planets, compared to those that formed in lower UV environments. For example, if planets form by pebble accretion, then the young planets would be fed by a stream of pebbles from the outer disk. If the outer disk is dissipated, the supply of material that helps the planets grow is reduced (see, e.g., Qiao et al. 2023). Indeed, outer disk truncation due to UV radiation is well documented by ALMA observations (Ansdell et al. 2017; van Terwisga et al. 2019, 2020; van Terwisga & Hacar 2023), and our observations could be the first evidence based on MIRI observations that the outer disk radius in such an extreme environment has been truncated.

XUE 1 shows that the conditions for terrestrial planet formation can also happen in extreme environments. Nevertheless, the remaining observations from the XUE program are crucial to establish the frequency with which this occurs.

The MIRI spectrum contains a part of the P-branch of the CO fundamental. Both the v = 1 − 0 and the v = 2 − 1 rovibrational transitions are detected in the spectra. In this section we present a first approach at deriving the excitation temperatures via the population diagram method (Goldsmith & Langer 1999), and we determine the temperature, column density, and emitting area by means of fitting non-LTE slab models to the spectra.

C.1.1. Rotation Diagram

We selected the lines to be fitted by visually inspecting the spectrum and picking only those lines in which the v = 1 − 0 and v = 2 − 1 transitions are resolved and not contaminated by H₂O. We fitted the v = 1 − 0 corresponding to upper J levels between 24 and 33 and J_up = 46 and v = 2 − 1 with upper J levels between 18 and 28 excluding J_up = 23 and including J_up = 41. We determined the flux of each line and the continuum level by simultaneously fitting a zero-degree polynomial and all the selected lines with Gaussian profiles. To do the Gaussian fits, we assumed an error of 10⁻⁴ Jy, which corresponds to the typical error of line-free regions. The errors in the line fluxes (reported in Table 2) were calculated as the square root of the covariance matrix. The specific intensity, I, was determined by integrating over each profile and dividing the observed flux by the solid angle, Ω, assuming an emitting area of 0.5 au and a distance to XUE 1 of 1690 pc. The column density was then calculated using the formula

$$\begin{eqnarray}&&{N}_{j}=\displaystyle \frac{4\pi {\lambda }_{j}{I}_{j}}{{A}_{j}{hc}},\end{eqnarray} \tag{ C1 }$$

where λ is the central wavelength of each transition, A_j is the Einstein A-coefficient of each transition, h is the Planck constant, and c is the speed of light. Assuming that CO is in LTE, the energy levels can be described by a Boltzmann distribution. We can therefore express the relative column densities of any two excitation levels in terms of an excitation temperature T_ex as

$$\begin{eqnarray}&&\displaystyle \frac{{N}_{i}}{{N}_{j}}=\displaystyle \frac{{g}_{i}}{{g}_{j}}\exp \left(-\displaystyle \frac{{E}_{i}-{E}_{j}}{{{kT}}_{\mathrm{ex}}}\right),\end{eqnarray} \tag{ C2 }$$

where g_j is the statistical weight of the rotational level, E_j is the energy of the upper level, and k is the Boltzmann constant. The excitation temperature T_ex corresponds to the inverse slope of the $\mathrm{ln}(N/g)$ versus E_up relation.

Table 2. Measured Fluxes of the CO Fundamental Lines Used for the Rotation Diagram and the CO χ² Modeling (Section 3.1)

Wavelength	vup	Jup	Flux
(μm)			(10−16 erg s−1 cm−2)
5.1887	1	46	1.853 ± 1.135
5.1192	1,2	41,36	3.158 ± 0.889
5.1057	1,2	40,35	3.551 ± 0.933
5.0924	1,2	39,34	3.814 ± 0.945
5.0792	1,2	38,33	3.443 ± 0.91
5.0661	1,2	37,32	3.942 ± 0.96
5.0532	1,2	36,31	3.871 ± 0.919
5.0405	1,2	35,30	3.907 ± 0.989
5.0279	1,2	34,29	2.971 ± 0.919
5.0154	1	33	3.286 ± 0.911
5.0031	1	32	4.318 ± 0.945
4.9908	1	31	3.998 ± 0.907
4.9788	1	30	4.979 ± 0.933
4.9668	1	29	4.602 ± 0.933
4.9550	1	28	4.785 ± 0.925
4.9434	1	27	4.461 ± 0.884
4.9318	1	26	4.205 ± 0.904
4.9204	1	25	4.89 ± 0.901
4.9091	1	24	6.379 ± 1.036
5.1855	2	41	0.518 ± 0.941
5.0184	2	28	0.778 ± 0.932
5.0065	2	27	0.515 ± 0.761
4.9947	2	26	1.015 ± 1.042
4.9831	2	25	0.768 ± 1.054
4.9716	2	24	1.193 ± 1.122
4.9490	2	22	0.918 ± 0.871
4.9379	2	21	0.777 ± 0.892
4.9269	2	20	1.373 ± 1.021
4.9161	2	19	1.211 ± 0.827
4.9054	2	18	1.194 ± 0.912

Note. The blended lines (excluded from the rotation diagram) are indicated by listing both v_up and J_up levels.

Download table as:  ASCIITypeset image

C.1.2. Non-LTE Slab Models for CO

The ratio between the CO v = 1 − 0 and the v = 2 − 1 emission lines in XUE 1 is particularly high and cannot be reproduced by LTE slab models. We therefore opt to fit the observed CO spectra with radex non-LTE slab models (van der Tak et al. 2007) using the molecular data from Song et al. (2015). We follow the methodology from Tabone et al. (2023); we assume the line profiles to be Gaussian with a broadening ΔV = 4.7 km s⁻¹ as in Salyk et al. (2011). In the non-LTE models the density of collision partners n_<H> is a free parameter, allowing us to decrease it in order to match the line ratio between the fluxes of the v = 1−0 and v = 2−1 lines. The collision partners are H₂, He (0.1n_<H>), and electrons (10⁻⁵ n_<H>). Other free parameters are the column density (N_CO), the excitation temperature (T_ex), and the emitting area given by $\pi {R}_{\mathrm{disk}}^{2}$, where R_disk does not necessarily correspond to the disk radius but could also represent a ring with the same area. Further studies are needed to account for the effect of fluorescence, which can also affect the v = 1 − 0/v = 2 − 1 line ratio, but this is beyond the scope of this paper.

We fit the CO v = 1 − 0 upper J levels between 24 and 41 and J_up = 46 and v = 2 − 1 with upper J levels between 18 and 36, excluding J_up = 23, where the lines with J_up between 29 and 36 are blended with the v = 1 − 0 lines with J_up between 34 and 41 and J_up = 41. We use a χ² minimization to find the best model that adjusts to the observations. We compared the fluxes derived in Section C.1.1 with their respective errors to the model fluxes.

In an optically thin regime, N_CO and R_disk would be degenerate. Therefore, we can take advantage of the resolution of JWST in order to get an extra constraint on the emitting radius. We measure the width of the individual transitions and assume that the extra line broadening is due to Keplerian rotation. The projected rotational velocity can be expressed as

$$\begin{eqnarray}&&v\sin i=c\displaystyle \frac{{\mathrm{FWHM}}_{\mathrm{rot}}}{2\lambda },\end{eqnarray} \tag{ C3 }$$

where i is the angle at which we observe the disk and ${\mathrm{FWHM}}_{\mathrm{rot}}=\sqrt{{\mathrm{FWHM}}_{\mathrm{obs}}^{2}-{\mathrm{FWHM}}_{\mathrm{inst}}^{2}}$. Here FWHM_inst is the expected line width due to the resolution of the instrument (R ∼ 4000; Argyriou et al. 2023), and FWHM_obs is the measured line width. The resulting velocity can then be used to provide additional constraints on the emitting surface radius provided that we know the mass of the central star, allowing us to circumvent the degeneracy between the column density and the emitting surface radius.

Table 3. Overview of the Parameter Ranges for the Modeling Procedure

Species	$\mathrm{log}(N)$	T	Fitting Ranges
	(cm−2)	(K)	(μm)
H2O @7 μm	15–22	100–1400	6.404–6.422
			6.445–6.456
			6.467–6.478
			6.630–6.641
			6.678–6.689
			6.781–6.792
			6.858–6.871
			7.039–7.051
			7.202–7.212
H2O @15 μm	15–22	100–1400	13.498–13.507
			14.173–14.182
			14.205–14.215
			14.322–14.332
			14.341–14.350
			14.402–14.412
			14.422–14.432
			14.482–14.493
			14.820–14.830
			14.880–14.899
			15.050–15.060
			15.178–15.187
			15.324–15.332
			15.341–15.353
			15.412–15.423
			15.450–15.460
			15.564–15.574
			15.621–15.631
			15.720–15.763
			15.790–15.808
			15.962–16.015
			16.105–16.120
			16.217–16.227
			16.271–16.274
			16.317–16.327
HCN	14–19	200–1000	13.980–13.997
			14.022–14.057
			14.290–14.308
C2H2	17–20	100–1200	13.562–13.748
12CO2	13–19	100–1400	14.932–14.940
			14.959–14.990
CO	14–20	100–3400	4.900–5.152 a
			5.18–5.19

Note.

^a See Appendix C.1.2 for the list of lines used.

Download table as:  ASCIITypeset image

In order to characterize the remaining molecular emission features present in the spectrum of XUE 1, we model them with the LTE model presented in Tabone et al. (2023), which has also been used in Grant et al. (2023), Perotti et al. (2023), and Gasman et al. (2023). We included emission from H₂O (around 7 and 15 μm), ¹²CO₂, ¹³CO₂, OH, C₂H₂, and HCN, with molecular data from the HITRAN database (Gordon et al. 2022). As with CO, Gaussian intrinsic line profiles with ΔV = 4.7 km s⁻¹ are assumed. For the HCN and ¹²CO₂ Q-branches, we included mutual shielding from adjacent lines as described in Tabone et al. (2023). Under the LTE assumption we are able to fit the data with three free parameters: the gas temperature T, the column density N, and the emitting area as parameterized by an emitting radius R_disk ($\pi {R}_{\mathrm{disk}}^{2}$). As in the case of CO, R_disk is not necessarily the disk radius but could also represent a ring.

In order to fit the MIRI MRS spectrum, we first subtract the continuum locally by selecting line-free regions (see Table 3). We fit a spline function to determine the continuum level and subtract this from the MIRI spectrum. We then proceed to fit individual regions of the spectrum containing different molecules following a similar procedure to that described in Grant et al. (2023).

For each molecule, we ran a grid of models varying N in steps of 0.166 in ${\mathrm{log}}_{10}$-space, T in steps of 25 K in linear space, and R_disk from 0.01 to 10 au in steps of 0.03 au in ${\mathrm{log}}_{10}$-space. The best-fitting models were found by means of a χ² fitting procedure. We measured the typical error in the flux in several line-free regions of the spectrum and adopted an error of 10⁻⁴ Jy for the wavelength range between 13 and 17 μm. We convolved the model spectra with a resolution R = 3000 for the water around 7μm and with R = 2500 between 13 and 17 μm to mimic the limited spectral resolution of MIRI MRS. We then resampled the convolved model to the wavelength grid of the observations. The χ² was calculated at given spectral regions in order to avoid regions of the spectrum that are strongly contaminated by other species while still including lines that are sensitive to temperature and optically thin lines. In the case of H₂O we selected the regions in which to calculate χ² by running an LTE model with T = 800 K, N = 10¹⁷ cm⁻², and R = 1 au to identify the spectral region where strong water lines are expected around 7 and 15 μm. The parameter ranges, as well as the wavelength regions used to calculate the χ², are listed in Table 3.

To further reduce the contamination by other species, we follow the iterative procedure described in Grant et al. (2023). In short, we first fit H₂O, subtract the best-fitting model from the continuum-subtracted data, and then proceed to do the same for HCN, C₂H₂, and ¹²CO₂. We then subtract the composite best model of HCN, C₂H₂, and ¹²CO₂ from the data and repeat the procedure for H₂O. We perform this process once and find no change in the best parameters, nor any improvement on the residuals.