Water-Rich Disks around Late M-stars Unveiled: Exploring the Remarkable Case of Sz114

We present an analysis of the JWST/MIRI-MRS spectrum of Sz 114, an accreting M5 star surrounded by a large dust disk with a gap at ∼ 39 au. The spectrum is molecular-rich: we report the detection of water, CO, CO 2 , HCN, C 2 H 2 , and H 2 . The only identified atomic/ionic transition is from [Ne II ] at 12.81 µ m. A distinct feature of this spectrum is the forest of water lines with the 17.22 µ m emission surpassing that of most late M-star disks by an order of magnitude in flux and aligning instead with disks of earlier-type stars. Moreover, flux ratios of C 2 H 2 /H 2 O and HCN/H 2 O in Sz 114 also resemble those of earlier-type disks, with a slightly elevated CO 2 /H 2 O ratio. While accretional heating can boost all infrared lines, the unusual properties of Sz 114 could be explained by the young age of the source, its formation under unusual initial conditions (a large massive disk), and/or the presence of dust substructures. The latter delay the inward drift of icy pebbles and thus preserve a lower C/O ratio over an extended period. In contrast, late M-star disks—which are typically faint, small in size, and likely lack significant substructures—may have more quickly depleted the outer icy reservoir and already evolved out of a water-rich inner disk phase. Our findings underscore the unexpected diversity within mid-infrared spectra of late


INTRODUCTION
Observations of disks around young stars provide a unique opportunity to witness planet formation in action (e.g., Bae et al. 2022) and constrain the distribution of volatiles while planets are assembling (e.g., Öberg & Bergin 2021).Spitzer/IRS observations at moderate resolution (R ∼ 700) mostly targeted young solar analogs (FGK and early M) and demonstrated that their disk atmospheres are rich in volatiles.Most of the volatile oxygen and carbon are carried by H 2 O and CO, while most nitrogen is likely in the undetected form of gas-phase N 2 (e.g., Pontoppidan et al. 2014, and references therein).For the brightest disks, high-resolution (R ∼ 25, 000 − 95, 000) ground-based telescopes targeted few accessible transitions and spectroscopically confirmed that infrared molecular emission mostly traces gas within few au from the star (e.g., Pontoppidan et al. 2010;Mandell et al. 2012;Najita et al. 2018;Salyk et al. 2019;Banzatti et al. 2023a).This is the gas inside the water iceline (hereafter, snowline) that is accreted by those gaseous planets, including sub-Neptunes, within ∼ 1 au (e.g., Bean et al. 2021;Izidoro et al. 2022).
In contrast to young solar analogs, only a handful of mid-to-late M-star disks (disks around stars whose spectral types are later than ∼M3,) have IRS spectra and only at modest S/N due to their relatively low luminosity.Nevertheless, their infrared spectra are different and characterized by: i) stronger C 2 H 2 than HCN emission, a signature seen even in lower resolution IRS spectra (Pascucci et al. 2009), and ii) weak/undetected H 2 O lines (Pascucci et al. 2013).Retrieval slab models of gas in local thermodynamic equilibrium (LTE) suggest that their inner disk has an enhanced abundance of C 2 H 2 and HCN relative to water (Pascucci et al. 2013), pointing to elevated C/O ratios of ∼ 1 (e.g., Najita et al. 2011).In contrast, when the same models are applied to disks around earlier-type stars, C 2 H 2 and HCN are typically orders of magnitude less abundant than H 2 O (e.g., Carr & Najita 2011;Salyk et al. 2011;Walsh et al. 2015).A similar chemical trend has been also reported with ALMA for the outer disks (≥ 20 au) of five late M-stars whose C 2 H/HCN column density ratios are a factor of a few higher than in disks around earlier-type stars1 (Pegues et al. 2021).
These findings hint at a change in the volatile content of mid-to-late M-star disks, with hydrocarbons and nitriles likely being more abundant than in disks around earlier-type stars.Supporting this inference, the first JWST/MIRI-MRS spectrum (R ∼ 2, 000) of an early M-star (M1.5, M * ∼ 0.5 M ⊙ , Andrews et al. 2018) disk shows solar-like H 2 O, C 2 H 2 , and HCN abundances (GW Lup, Grant et al. 2023) while that of a late M-star (M5, M * ∼ 0.2 M ⊙ ) disk has no water detection but strong emission from C-bearing molecules including from benzene C 6 H 6 (2MASS J16053215-1933159, hereafter J160532, Tabone et al. 2023).However, the selection of these two disks was based on the availability of Spitzer/IRS spectra at R ∼ 700, which provided the foundation for the aforementioned identified trends.As it is difficult to draw any firm conclusion from such a small sample, targeting disks that have not been previously observed with Spitzer/IRS is essential, particularly for the underrepresented late M stars, as it allows for a comprehensive exploration of the potential diversity within this class.
Here, we present the JWST/MIRI-MRS spectrum of the disk of Sz 114, an accreting M5 star (M * ∼ 0.17 M ⊙ , e.g., Hughes et al. 1994;Mortier et al. 2011;Alcalá et al. 2017) located in the Lupus III sub-group (median distance of 158.9 pc and median age of 2.5 Myr, Galli et al. 2020).This disk, which is massive (M dust ∼ 30 M ⊕ , Ansdell et al. 2018) and millimiter-bright, was observed as part of the ALMA Disk Substructures at High Angular Resolution Project (DSHARP, Andrews et al. 2018) and found to have a shallow dust gap at ∼ 39 au (Huang et al. 2018).Followup visibility modeling using the super-resolution code frank suggests additional unresolved substructures between ≈ 7 − 12 au (Jennings et al. 2022).Sz 114 was not observed with Spitzer/IRS at R ∼ 700.We show here that the mid-IR spectrum of this late M-star disk is remarkably water-rich (Sects. 2 and 3).We discuss molecular flux ratios relative to water in the context of other mid-to-late M-star disks as well as disks around earlier-type stars observed with Spitzer/IRS and JWST/MIRI-MRS.We propose that the rich water spectrum of Sz 114 is driven by the star's relatively high accretion luminosity in combination with a young age and a more massive large disk with substructures that might prolong the inward drift of icy pebbles (Sect.4).We conclude in Sect. 5 by summarizing the key findings of our study.Additionally, we emphasize the importance of expanding the sample of mid-to-late M-star disks beyond that observed with Spitzer/IRS.The extension to systems with a range of ages, accretion luminosities, and outer disk substructures is crucial for capturing the diversity of inner M-star disks and thoroughly testing our current understanding of pebble drift and planet formation.

OBSERVATIONS AND DATA REDUCTION
Sz 114 was observed with the Medium Resolution spectrometer (MRS, Wells et al. 2015) of the JWST Mid-Infrared Instrument (MIRI, Rieke et al. 2015) as part of the Cycle 1 Treasury Project "A DSHARP-MIRI Treasury Survey of Chemistry in Planet-forming Regions" (ID: 1584, PI: C. Salyk, Co-PI: K. Pontoppidan, hereafter JDISC).To minimize fringes, the source was first acquired using the MRS target acquisition procedure with a neutral density filter.This procedure identifies the brightest pixel in the image and offsets it to the Integral Field Unit's (IFU) center with an accuracy of 0. ′′ 1. JDISC employed the 4-point dither pattern, which is optimized for point sources, in the negative direction, i.e. the pattern moves from the bottom right to upper left.All three MIRI-MRS grating settings were selected to ensure full spectral coverage (4.9 − 28 µm).To achieve a signal-to-noise (S/N) of ∼ 100 at the longest wavelengths, 75 groups per ramp were selected for an exposure time per module of 832.5 s.
The data were reduced following Pontoppidan et al. in prep.In short, the JWST Calibration Pipeline (Bushouse et al. 2022) version 1.11.0 and Calibration Reference Data System (CRDS) context jwst_1105.pmapwas used to process the data up to stage 2 resulting in datacubes for every exposure, channel, and sub-band.1D-spectra were extracted with an aperture that widens linearly with wavelength within each sub-band and the background was subtracted from the two opposing dither positions.Then spectra were individually divided by a relative spectral response function constructed from the asteroid 515 Athalia (PID: 1549, PI: K. Pontoppidan) before median-clipping of the four dither positions.Finally, we shifted the spectrum by -1 km/s to align the lines with the mean heliocentric radial velocity of the Lupus III subgroup, where Sz 114 is located (Galli et al. 2013).The final reduced spectrum is shown in black in Figure 1.The MIRI-MRS spectrum of Sz 114 shows a broad and flat-topped 10 µm silicate emission feature typical of mid-tolate M-star disks (e.g., Apai et al. 2005;Pascucci et al. 2009).These properties indicate that the typical dust grains in the disk atmosphere are microns in size, i.e. about ten times larger than interstellar medium dust and typical grains in solar-type disk atmospheres (e.g., Natta et al. 2007).In addition, the prominent ro-vibrational bands from C 2 H 2 , HCN, and CO 2 , all of which were previously detected with Spitzer/IRS in mid-to-late M-star disks (Pascucci et al. 2009(Pascucci et al. , 2013)), are also present in the Sz 114 MIRI-MRS spectrum, see Figure 1.At the shortest wavelengths, several CO ν = 1 − 0 P-branch lines are also found, while CO emission in disks has often been found to have non local thermodynamic equilibrium (non-LTE) excitation (e.g., Thi et al. 2013;van der Plas et al. 2015).
The most prominent characteristic of this spectrum is the forest of water lines: from the hot bending ro-vibrational band at ∼ 6.6 µm to the cooler pure rotational lines at longer wavelengths (see e.g.Fig. 1 in Banzatti et al. 2023a).In relation to water emission, Sz 114 is very different from other mid-to-late M-star disks observed with Spitzer/IRS (Pascucci et al. 2013) and the similarly late M5 object recently observed with MIRI-MRS (J160532, Tabone et al. 2023) in which water emission was either very weak or absent.Such a rich water spectrum is more akin to that of disks around earlier-type stars (spectral types earlier than M3, e.g.Carr & Najita 2011;Salyk et al. 2011), as discussed further in Sect. 4. Visual inspection of the MIRI datacubes does not show any extended emission.Therefore, as in previous papers (e.g., Salyk et al. 2011;Pascucci et al. 2013), we assume that the water emission is tracing the disk surface instead of jets or winds.
Properly identifying weak emission lines from other molecules requires subtracting the dust continuum emission and the rich water spectrum.To identify line-free channels for continuum subtraction, we first generated a water spectrum with CLIcK (Liu et al. 2019).CLIcK assumes a plane-parallel slab of gas in LTE and adopts molecular parameters from the HITRAN database(e.g., Gordon et al. 2022).Line overlap is properly accounted for using eq. 1 in Tabone et al. (2023).With the generated water spectrum, we identified by eye continuum regions that avoided water lines and the prominent ro-vibrational bands mentioned above (see light blue points in Figure 1).After experimenting fits with polynomials of different orders, we found that a 5th-order polynomial provides a good fit for the entire continuum emission2 .The continuum-subtracted spectrum of Sz 114 is depicted in grey below the original spectrum in Figure 1.
Next, we work with the continuum-subtracted spectrum and run CLIcK again to find the water models that fit best the water emission between ∼ 10 − 19 µm.The reason for focusing on this restricted wavelength range is that it contains the ro-vibrational bands of interest for placing Sz 114 in context of other disks observed with Spitzer/IRS and, more recently, JWST/MIRI-MRS (see Sect. 4).In spite of the restricted wavelength range, the water spectrum covers a broad range of upper energy levels and Einstein-A coefficients which may not be fitted well by a single-temperature slab of gas in LTE (e.g., Banzatti et al. 2023b).This is indeed the case for Sz 114, where we find that two temperatures are necessary to reproduce the observed spectrum (see the sum of the two slab models as light blue lines in the bottom panel of Figure 2 and in Figure 3).The best fit is found by first fitting the hot water in the region between ∼ 10−12 µm plus the high energy line at 15.18 µm (E up ∼ 8, 000 K) simultanously.Next, the best fit to the hot water is removed and then all water lines are fitted between ∼ 10 − 12 µm and ∼ 16.3 − 19 µm but excluding regions with known molecular lines other than water.We stress that the aim of this modeling is not to perfectly match the complete JWST-MIRI spectrum.Instead, our goal is merely to identify a suitably precise model within the molecular feature-rich range (∼ 13 − 17 µm) to untangle the contribution from individual molecules and measure their line fluxes.
These slab models have three input parameters: the emitting radius (R em , the model fits for an equivalent area of πR 2 em ), the column density (N ), and the gas temperature (T ).To fit the water lines, we generate a grid of 4,000 models using a custom-written random walk python code (see appendix A for more details) with starting values R em =0.2 au, N = 10 18 cm −2 , and T = 1, 000 K for the hot and 300 K for the cool components, based on prior models of solar-type disks (Banzatti et al. 2023a).For each step, random values are chosen between ±0.05 au for ∆R em,i , ±2 for ln(∆N i ), and ±100 K for ∆T i .Using equation 2 in Mulders et al. (2015) with M * and Ṁacc for Sz 114 (Table 1) and assuming typical values for the Rosseland mean opacity (from 20 to 570 cm 2 g −1 ), a gas-to-dust ratio of 100, and two values for the turbulent mixing strength (0.01 and 0.001), we estimate the midplane snowline to be between ∼ 0.1 − 0.5 au.Since the cool component is still at several hundred K, we set an upper limit of 0.5 au for R em , i.e. most of the water emission is coming within the snowline.The best-fit model is the one with the minimum χ 2 and derived temperatures, column densities, and emitting radii are summarized in Table 3 of Appendix A.
After subtracting the water spectrum, we proceed to fit with CLIcK the next major molecular emitters in the 10 − 19 µm region.First, we fit the CO 2 spectrum between 14.6 − 15.4 µm thus covering many P-and R-branches in addition to the strongest Q-branch emission, see Figure 3. Next, we remove the best-fit spectrum extended to the entire wavelength region and fit together the C 2 H 2 and HCN ro-vibrational bands between 13.5 − 14.3 µm.The best-fit parameters for these molecular species are also summarized in Table 3 of Appendix A while the composite model compared to the data is shown in the upper panel of Figure 2 with individual molecular emission models in the lower panel.
Aware of the degeneracies among parameters even at the sensitivity and spectral resolution of MIRI-MRS (e.g.Grant et al. 2023), we use these fits merely to separate the contribution from different molecules and calculate line fluxes.The fluxes are calculated by integrating the data (or model for the blended molecules C 2 H 2 , HCN and CO 2 ) within the wavelength range reported in Table 2. To facilitate the comparison between Sz 114 and literature sources these ranges are chosen to match those adopted in Spitzer studies (e.g., Salyk et al. 2011).Regarding water, we note that the two tentative detections in late M-star disks are at 17.22 µm (in 2MASS J04381486+2611399 and 2MASS J04442713+2512164, hereafter J043814 and J044427, see Pascucci et al. 2013), hence we focus on the flux of this line for Sz 114.The calculated line fluxes can be found in Table 2.Because the disk surface snowline is further out than the midplane snowline, we have also run models that do not restrict R em to be within ∼ 0.5 au.Although the best-fit values for T and N are different from those reported in Table 3, molecular line fluxes, which are the focus of this paper, are within 5% of those reported in Table 2.
Inspection of the continuum-subtracted spectrum with the best-fit water, CO 2 , C 2 H 2 , and HCN synthetic models reveals additional molecular transitions (see Table 2): two from H 2 and a tentative detection of the ro-vibrational band from the rarer isotopologue 13 CO 2 at ∼ 15.41 µm (see Grant et al. 2023 for a stronger detection in GW Lup).We have also searched for C 4 H 2 , C 6 H 6 , and CH 4 , since they were detected toward J160532 with MIRI-MRS (Tabone et al. 2023), but did not identify them in the spectrum of Sz 114.The only detected atomic/ionic line in the spectrum of Sz 114 is that from [Ne II] at 12.81 µm (see Table 2).The H 2 peak centroids are consistent with the stellar radial velocity while the [Ne II] is clearly blueshifted, see Figure 3.A discussion of the H 2 and [Ne II] lines is provided in Appendix B. Here, we focus on the prominent water lines and the CO 2 , C 2 H 2 , and HCN ro-vibrational bands.2. Shaded regions indicate the wavelength ranges over which fluxes are measured.For C2H2 (purple), HCN (red), and CO2 (green) these are the same ranges as those adopted in earlier Spitzer/IRS studies to facilitate comparison with literature sources.For H2O (light blue) the wavelength range is restricted to the complex around 17.22 µm.The best-fit spectra obtained with CLIcK are also shown with solid lines.
sample of late M-type stars is from Pascucci et al. (2013) and includes at most six sources with SpTy M7.25−M4.5 for which line flux ratios (including upper or lower limits) can be calculated.The sample of earlier-type stars (M3−G0) is from Banzatti et al. (2020) and includes at most 49 sources3 .For this latter sample, we note that the water line flux has been re-calculated using the restricted wavelength range around the 17.22 µm transition.
Figure 4 demonstrates that Sz 114 has an unusually strong 17.22 µm water flux for its late SpTy compared to other mid-to-late M-star disks observed before, more similar to that of disks around early M and K-type stars.The only other outlier in this sample is the disk of XX Cha, a slightly earlier SpTy source (M3.5) observed with Spitzer/IRS and more recently with JWST/MIRI but whose MIRI spectrum has not been published yet.Additionally, Sz 114 shares similar HCN/H 2 O and C 2 H 2 /H 2 O flux ratios with XX Cha, both of which fall well within the range of values observed in disks around earlier-type stars.One difference is the slightly higher CO 2 /H 2 O flux ratio exhibited by Sz 114.In contrast to Sz 114, late M-star disks tend to have higher C 2 H 2 /H 2 O flux ratios (Figure 4 lower panel) and higher C 2 H 2 /HCN ratios (see Fig. 13 from Pascucci et al. 2009).These higher ratios have been attributed to a very high C/O ratio (≥ 0.8) in the inner disk of late M-type stars (Pascucci et al. 2013;Tabone et al. 2023), as the extra carbon that does not end up in CO can efficiently form hydrocarbons in the gas phase (e.g., Najita et al. 2011).Sz 114 clearly shows that there is a larger diversity of mid-IR spectra down to late spectral types than previously inferred from the Spitzer/IRS surveys.
To place the Sz 114 system into context, we have also collected the accretion luminosity (L acc ) and the millimeter flux (F mm ) for the earlier spectral type sample (Banzatti et al. 2020) and an expanded late M-star disk sample that now includes systems observed at the lowest Spitzer/IRS resolution (R∼100) from Pascucci et al. (2009), see Figure 5.This latter sample had, on average, a higher C 2 H 2 /HCN flux ratio than disks around earlier-type stars but nothing could be said about the strength of the water emission due to the very low spectral resolution.L acc and F mm from this expanded sample are from Manara et al. (2022) which homogenize measurements from various surveys (e.g., Andrews et al. 2013;Ansdell et al. 2016;Pascucci et al. 2016;Alcalá et al. 2017).The reason for focusing on these two parameters is that multiple Spitzer studies have shown that water emission is positively correlated with accretion luminosity (e.g., Salyk et al. 2011;Banzatti et al. 2020) while the water/HCN ratio is anti-correlated with the disk millimeter flux (Najita et al. 2013).Banzatti et al. (2020) also found the water/HCN ratio, as well as the water luminosity corrected for the L acc trend, is anti-correlated with dust disk radius as millimeter fluxes are correlated with dust disk sizes (e.g., Tripathi et al. 2017;Hendler et al. 2020).
The positive correlation between water and accretion luminosity is likely driven by the additional heating that accretion can provide for the excitation of the lines (e.g., Najita & Ádámkovics 2017;Glassgold et al. 2009).Indeed, the luminosity of several other infrared lines is also positively correlated with L acc (e.g., Salyk et al. 2011 and Fig. 8).The anti-correlation between the water and millimeter flux (and dust disk size) has been interpreted in the context of pebbles' drift: low millimeter fluxes might result from the efficient inward drift of icy pebbles (roughly the mm-sized grains detected by ALMA) which increase the water vapor in the inner disk upon crossing the snowline (Banzatti et al. 2020).This scenario is supported by new MIRI spectra showing excess cold water emission in compact disks, which could originate from enhanced ice sublimation due to a larger flux of inward drifting pebbles (Banzatti et al. 2023b).
As seen in Figure 5, Sz 114 has a relatively large accretion luminosity for its spectral type, further supporting the expectation that extra heating boosts infrared lines, including from H 2 O.However, Sz 114 and J160532 do not follow the anti-correlation between water luminosity and F mm : The infrared spectrum of the millimeter brighter, larger, and structured disk of Sz 114 is significantly more water-rich than that of the millimeter-faint J160532.Note that the spectral type, as well as the stellar mass and accretion luminosity of these two objects are strikingly similar (see Figure 5).One major difference, in addition to the millimeter flux (and potential size and substructures), is the likely age of the two objects: Sz 114 belongs to the ∼ 1 − 3 Myr-old Lupus star-forming region (Galli et al. 2020) and has an isochronal age of ∼ 1 Myr (Alcalá et al. 2017), while J160532 belongs to the ∼ 5 − 10 Myr-old Upper Sco association (e.g.Preibisch & Mamajek 2008;Luhman & Esplin 2020) and has a dynamical age of 2.5±1.6 Myr (Miret-Roig et al. 2022).With the same spectral type but a stellar luminosity that is ∼ 5 times lower than that of Sz 114, J160532 is clearly older.
How might age play a role in explaining the different mid-IR spectra?Recently, Mah et al. (2023) carried out simulations using a coupled disk evolution and planet formation code that includes pebble drift and evaporation at icelines in radially smooth disks (that is a disk with no gaps or pressure bumps).They showed that, for such disks, the C/O ratio inside the snowline initially decreases below the stellar value, due to the inward drift and evaporation of water-ice-rich pebbles, before it increases to superstellar values due to the inward diffusion of carbon-rich vapor, mostly from the evaporation of CH 4 , from the accreting gas surface flow.Importantly, the transition from low to high C/O ratios (water-rich to water-poor) occurs faster around very low-mass stars because of the closer-in icelines (e.g., Mulders et al. 2015), shorter viscous timescales, and faster radial drift (e.g., Pinilla et al. 2013).Specifically, Mah et al. 2023 estimated that the inner disk of a 0.1 M ⊙ star dries out in ∼0.5 Myr if the α viscosity is 10 −3 and in ∼ 2 Myr  (Pascucci et al. 2013) and disks around earlier-type stars (Banzatti et al. 2020).In these comparisons, the water line at 17.22 µm is adopted.The two downward arrows with a question mark next to them are for the tentative Spitzer water detections in the mid-to-late M-star disks of J043814 and J044427.J160532 is a source similar to Sz114 in spectral type, stellar mass and accretion luminosity but is extremely carbon-rich.XX Cha is a slightly earlier-type (M3.5) star with a disk rich in water.Note how Sz 114 (star symbol) has more similar molecular flux ratios to earlier-than later-SpTy except for CO2/H2O which is higher than the average earlier-type disk value.
for α = 10 −4 while that of a 0.7 M ⊙ star dries out in ∼2 Myr and ∼10 Myr, respectively (see their Figure 1).This scenario explains the general trend that the inner disks of late M-stars are relatively more water-poor (C-rich) than similarly aged disks around earlier-type stars as well as the difference in the spectra of Sz 114 and J160532.By the relatively later age of J160532, the icy pebble supply from the outer disk has been likely consumed (as suggested by the low millimeter flux) and the inner disk C/O ratio has been enhanced by the inner diffusion of carbon-rich vapor, which then triggers the formation of hydrocarbons.In contrast, at the early time of Sz 114, the inward drift of icy pebbles and the water vapor released when crossing the snowline, in combination with accretional heating, may boost the water emission.Interestingly, the two tentative detections of water in Spitzer/IRS spectra of similarly late M-star disks are also from young systems in the Taurus star-forming region (Pascucci et al. 2013).

Xie et al.
While these ideas may account for the relative properties of Sz 114 and J160532, it is also interesting to consider why Sz 114 is so different from the majority of the mid-to-late M-star disks of similar age observed with Spitzer/IRS.One possibility is that Sz114 formed with unusual initial conditions, i.e. a large and massive disk at birth.Sz 114 has an exceptionally large millimeter flux and extended dust disk among mid-to-late M-stars.Sz 114 may also be unusual among mid-to-late M-star disks in hosting substructures (gaps).The shallow gap detected by DSHARP at ∼ 39 au, along with potential other substructures between ≈ 7 − 12 au, might maintain some icy pebbles further out, thus maintaining a large disk, and, at the same time, prolong water enrichment in the inner disk (see e.g.Fig. 4 in Kalyaan et al. 2023 for this effect in disks around solar-type stars).In contrast, most mid-to-late M-star disks are characterized by low millimeter fluxes (e.g., Pascucci et al. 2016 and Figure 5 for those targeted with Spitzer/IRS).These faint disks are low in the total amount of icy pebbles: they might have started with an overall smaller and less massive disk than Sz 114 and have not developed any substructure to slow down the inner drift of icy pebbles.As a result, the typical inner disk of late M-stars would follow the evolution described in Mah et al. (2023) and "dry out" fast as discussed above.A sketch that places the temporal evolution of the C/O ratio in context, for both Sz 114 and other late M-star disks, is provided in Figure 6.
These ideas can be tested with MIRI-MRS spectroscopy of larger samples of disks surrounding mid-to-late M stars.JWST GTO and GO Cycle 1 and 2 programs will also cover the low millimeter flux disks around late M-stars (see crosses in Figure 5) for which, similar to J160532, we expect water-poor and C-rich spectra even at young ages.It would be valuable to carry out MIRI-MRS observations of the late M-star disk in Taurus with tentative water detection and moderately high F mm (J044427), as well as of ESO Hα 559, which has a millimeter flux comparable to Sz 114.If our interpretation of the data at hand is correct, we expect these spectra to be water-rich as the spectrum of Sz 114.Fully testing the proposed scenario would require sensitive MIRI-MRS spectra of disks spanning a large range in spectral type, millimeter fluxes, accretion luminosities, and ages across several star-forming regions.

SUMMARY AND OUTLOOK
We have acquired and analyzed the JWST/MIRI-MRS spectrum of Sz 114, an accreting M5 star in the ∼ 1 − 3 Myrold Lupus star-forming region surrounded by a Class II disk.This system has a high millimeter flux and was selected for the ALMA Disk Substructures at High Angular Resolution Project (DSHARP), which discovered a large dust disk with a size of ∼ 58 au and a shallow gap at ∼ 39 au.Additional unresolved substructures may be also present between ≈ 7 − 12 au (Jennings et al. 2022).This disk was not previously observed with Spitzer/IRS at R ∼ 700, hence there are no gas line detections previously reported.We model the MRS spectrum of Sz 114 with a slab of gas in LTE with the main goal of removing line blending and calculating fluxes for the detected gas lines.We also collect Spitzer/IRS fluxes from disks around stars of similar and earlier spectral type and discuss the spectrum of Sz 114 in the context of this larger sample.Our main results can be summarized as follows:

C/O ratio
• The MIRI-MRS spectrum of Sz 114 shows a forest of water emission lines, from the hot water bending rovibration band at ∼ 6.6 µm through the cooler pure rotational lines at ∼ 25 µm.In addition to water, we also detect CO, CO 2 , HCN, and C 2 H 2 , with CO 2 having the strongest ro-vibrational band.Two H 2 lines and the [Ne II] at 12.81 µm are also robustly detected while 13 CO 2 is only marginally detected.
• The water spectrum of Sz 114 between ∼ 10−19 µm requires at least two LTE slabs of gas: hot close-in (∼ 800 K) water emission and an additional cooler (∼ 500 K) component whose emission comes from further out in the disk.
• The 17.22 µm water flux of Sz 114 exceeds that of mid-to-late M-star disks, including that of the carbon-rich disk of J160532 observed with JWST, by an order of magnitude, and falls within the range observed in disks around earlier-type stars by Spitzer.Similarly, the C 2 H 2 /H 2 O and HCN/H 2 O flux ratios of Sz 114 align with those of disks around earlier-type stars, while exhibiting a slightly higher CO 2 /H 2 O ratio.
As for the earlier-type stars, most of the molecular emission from Sz 114 is expected to arise in the disk atmosphere inside the snowline.While accretion luminosity can boost molecular emission, it alone cannot account for the rich water spectrum observed in Sz 114: J160532, for example, exhibits a similar L acc but a water-poor and C-rich spectrum.Several other factors might contribute to the exceptionally rich water spectrum of this late M-star disk.We have highlighted age because the water-rich inner disk phase is much shorter in smooth disks around lower-mass stars: This is because icy pebbles, which are O-rich, are depleted faster and the inner disk C/O ratio is enhanced earlier through the diffusion of C-rich vapor (Mah et al. 2023).Age could explain the difference between the MIRI spectra of Sz 114 and the relatively older J160532.In addition, we propose that unusual initial conditions, a large massive disk at birth, combined with the development of substructures, might be necessary to understand Sz 114 in the context of other mid-to-late M-star disks of similar age observed with Spitzer/IRS.The existence of the ∼ 39 au shallow gap in the disk of Sz 114 might help to retain a large dust size but it is not a strong enough barrier to pebbles drifting inward and could just help prolong the water-rich phase of the inner disk (Kalyaan et al. 2023).The other inferred substructures between ≈ 7 − 12 au might further delay the delivery of water vapor inside the midplane snowline which is expected to be only at ∼ 0.1 − 0.5 au from the star.This delay would preserve a lower C/O ratio for a more extended period compared to other mid-to-late M-star disks, the majority of which are millimeter faint and likely lack significant dust traps.The lower frequency of dust traps, in combination with a typically low disk mass at birth, could have important implications for the type of planets forming in disks around late M-stars.For instance, in the context of planet formation via pebble accretion, Mulders et al. (2021) predict that the frequency of close-in super-Earths should decrease for M * ≤ 0.5M ⊙ , meaning that systems like TRAPPIST-1 would be rare.
Our findings demonstrate that the mid-infrared spectra of mid-to-late M-star disks exhibit greater diversity than previously inferred from the limited subset observed with Spitzer.This underscores the importance of expanding the sample size to comprehensively explore and better understand the potential range of variations.column density (N) and temperature (T) of the two molecules are highly degenerate with each other, thus the error bars of the fitted parameters are very large and not shown here (e.g., table 3).We note that our major goal is to separate the contribution from different molecules to calculate line fluxes, and use these strategies mentioned above to speed up our modelling.See Figure 2, 3 for the comparison between our fitted models and the data.Note-We identify two sets of well-fit parameters for C2H2 and HCN given the intrinsic degeneracies between column density, temperature, and emitting area in LTE slab models of warm gas in protoplanetary disks (e.g., Salyk et al. 2011;Najita et al. 2018).The values reported here correspond to the maximum likelihood in our fitting routine, and the other parameter set is T ∼ 400 K and log10(N ) ∼ 19 cm −2 for both C2H2 and HCN.

Figure 3 .
Figure3.Expanded view of wavelength regions covering the ionic, atomic, and molecular lines reported in Table2.Shaded regions indicate the wavelength ranges over which fluxes are measured.For C2H2 (purple), HCN (red), and CO2 (green) these are the same ranges as those adopted in earlier Spitzer/IRS studies to facilitate comparison with literature sources.For H2O (light blue) the wavelength range is restricted to the complex around 17.22 µm.The best-fit spectra obtained with CLIcK are also shown with solid lines.

Figure 4 .
Figure 4. Molecular flux ratios vs. scaled water flux for mid-to-late M-star disks(Pascucci et al. 2013) and disks around earlier-type stars(Banzatti et al. 2020).In these comparisons, the water line at 17.22 µm is adopted.The two downward arrows with a question mark next to them are for the tentative Spitzer water detections in the mid-to-late M-star disks of J043814 and J044427.J160532 is a source similar to Sz114 in spectral type, stellar mass and accretion luminosity but is extremely carbon-rich.XX Cha is a slightly earlier-type (M3.5) star with a disk rich in water.Note how Sz 114 (star symbol) has more similar molecular flux ratios to earlier-than later-SpTy except for CO2/H2O which is higher than the average earlier-type disk value.

Figure 5 .
Figure 5. Flux at 0.89 mm scaled at 160 pc vs accretion luminosity.Symbols are color-coded by SpTy.The sample of midto-late M-star disks includes those in Pascucci et al. (2009) which was only observed at the lowest Spitzer/IRS resolution.Down arrows indicate millimeter upper limits.Triangles are for the mid-to-late M-star disks present in Figure 4 while crosses indicate JWST/MIRI targets.Sz 114 is indicated with a star symbol and shows a significantly higher millimeter flux than other mid-to-late M-star disks observed with Spitzer/IRS.

Figure 6 .
Figure6.Sketch of the inner disk C/O ratio evolution with time.Substructures, depending on their locations and depths, can slow down the inward drift of icy pebbles.The time at which the C/O ratio starts to increase indicates the depletion time of the inner water-rich phase and depends on several factors including stellar mass, initial disk size, mass, and viscosity, and the presence and depth of substructures.Model predictions for disks with substructures around solar-type stars can be found inKalyaan et al. (2023).Models for disks without substructures around stars of different masses are presented inMah et al. (2023).

Figure 7 .
Figure7.Corner plot of our hot water models as an example of our fitting.The vertical dashed lines indicate the 1σ of each parameters as the middle 68% of the accepted models.We can see the degeneracy between the three parameters from the shape of the distribution.

Table 2 .
SZ 114 IN THE CONTEXT OF OTHER DISKSAs described in Sect.3, the MIRI-MRS spectrum of Sz 114 is dominated by molecular emission, especially from water vapour.As such, its spectrum is very different from that of late M-star disks observed with Spitzer/IRS and Upper panel: composite model (orange shaded region) between 13.4-15.5µm compared to the Sz 114 JWST-MIRI spectrum (black).Lower panel: individual best fits to water, C2H2, HCN, and CO2.Summary of line fluxes Note-Sz 114 has a forest of water emission lines.We first fit the water spectrum and subtract the best-fit water model before computing the flux from other species or fit the emission from other molecular lines.a Emission from these species is blended, hence we use our best-fit model flux here.b This is a tentative detection after subtracting blended water and 12 CO2 lines.
4.more similar to that of disks around earlier-type stars.To better illustrate the difference, we show in Figure4line flux ratios with respect to water emission at 17.22 µm, scaled at 160 pc.We focus on C 2 H 2 , HCN, and CO 2 as these ro-vibrational bands were previously detected in Spitzer/IRS and their fluxes or upper limits are tabulated.The

Table 3 .
Summary of model fitting