MINDS. JWST/MIRI Reveals a Dynamic Gas-rich Inner Disk inside the Cavity of SY Cha

SY Cha was observed with MIRI ¹⁹ in MRS ²⁰ mode on 2022 August 8 as part of the MINDS GTO Program (PID: 1282, PI: T. Henning). The total exposure time was 3696 s using a four-point dither pattern in the positive direction.

The data were processed using version 1.9.4 of the JWST pipeline (Bushouse et al. 2023). Bespoke routines based on the vip package were used for background subtraction, bad pixel correction after stage 2, and spectral extraction (Gomez Gonzalez et al. 2017; Christiaens et al. 2023). At the Spec2 stage the standard flat-fielding and fringe corrections from the JWST pipeline were used. The background was then created using the minimum values among the four dither positions and smoothed with a median filter. Spectra were extracted from the final cubes using apertures twice the FWHM in size centered on the source centroid, identified with a 2D Gaussian fit. After extraction, residual fringe correction was carried out for each wavelength band. The final spectrum is presented in Figure 1.

Zoom In Zoom Out Reset image size

Visual inspection of the data cubes reveals the H₂ and [Ne ii] emission to be extended, while the rest of the line emission and the continuum is spatially unresolved. The above data reduction was not optimized for extended emission. Analysis of the extended H₂ emission, likely indicating the molecular component of a wind, will be presented in a separate work and is not discussed further here.

The overall spectrum shows stronger emission than observed by Spitzer in either low-resolution or high-resolution mode, with the Spitzer high-resolution observations (2004 August 31 and 2008 May 1) sitting between the low-resolution (2009 April 9) and MIRI observations (2022 August 8, see Figure 1). The largest differences are seen between the new MIRI observations and the low-resolution Spitzer observations from 2009, with the MIRI spectrum roughly 10% stronger at 6 μm, 40% at 10 μm, and 30% at 20 μm. The spectrophotometric accuracy is 2%–10% for Spitzer/IRS and 5.6% ± 0.7% for JWST/MIRI-MRS (Furlan et al. 2006; Watson et al. 2009; Argyriou et al. 2023). The increased spatial resolution of JWST lessens the chance of contaminating emission from other objects. Such contamination in the Spitzer data would result in a higher flux relative to JWST, the opposite of what we observe. While the difference in flux at short wavelengths is within the calibration uncertainty, the difference beyond 10 μm is not.

Transition disks are known to be variable in the IR. The level of variability seen between the Spitzer and JWST observations of SY Cha is similar to the variability of transition and pretransition disks observed over multiple epochs with Spitzer by Espaillat et al. (2011). That the variability is greater at longer wavelengths points to changes in the inner disk geometry as the source of the variability in SY Cha. A more nuanced understanding of the cause of the mid-infrared variability in SY Cha requires detailed radiative transfer modeling, which will be presented in subsequent work.

The strongest atomic line features are shown in Figure 2. To determine the total flux from each transition, we fit a Gaussian line profile using the Markov Chain Monte Carlo package emcee (Foreman-Mackey et al. 2013). Because many of the atomic lines overlap with nearby transitions, a two-component Gaussian is used. Only the Gaussian fit to the atomic transition is used when integrating to get the total flux. Each Gaussian fit is integrated over the respective wavelength range shown in Figure 2. For subtraction purposes, the continuum is defined by selecting apparently line-free pixels (see Appendix A) and then interpolating with a spline fit. The spline fit to these points is then subtracted from the spectrum. The rms noise, σ, of the continuum-subtracted spectrum is calculated by taking the standard deviation in a line-free wavelength range. Values for σ are given in Table 2. Best-fit models are identified using a modified reduced χ² fitting:

$$\begin{eqnarray}&&{\chi }^{2}=\displaystyle \frac{1}{N}\sum _{i=1}^{N}\displaystyle \frac{{\left({F}_{{\rm{obs}},i}-{F}_{{\rm{mod}},i}\right)}^{2}}{{\sigma }^{2}},\end{eqnarray} \tag{ 1 }$$

where N is the number of data points used in the fitting, F_obs and F_mod are the observed and model line flux, respectively, and σ is the rms noise calculated in a nearby line-free region.

Zoom In Zoom Out Reset image size

We characterize the observed molecular emission by fitting local thermodynamic equilibrium (LTE) 0D slab models, as presented by Tabone et al. (2023), to the individual subbands. Each model is a slab of gas with a uniform temperature and density and no implicit spatial information. The LTE assumption is valid when all energy levels of a given species are well characterized by a single excitation temperature. Continuum subtraction and noise estimation are carried out in the same way as for the atomic emission (Section 2.3). The points used for continuum subtraction are listed in Appendix A. The wavelength range used to calculate the noise and the resulting values are shown in Figures 3 and 4.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

When generating the slab models, the intrinsic line profile is assumed to be a Gaussian. We include mutual shielding from closely spaced transitions of the same molecule following the method described by Tabone et al. (2023):

$$\begin{eqnarray}&&\tau (\lambda )=\displaystyle \sum _{i}{\tau }_{0,i}{e}^{-{\left(\lambda -{\lambda }_{0,i}\right)}^{2}{/2\sigma }_{\lambda }^{2}},\end{eqnarray} \tag{ 2 }$$

where λ_0,i is the rest wavelength of given line i, σ_λ is the intrinsic broadening of the line in microns, assumed to be equivalent to σ_λ = 2 km s⁻¹ (Salyk et al. 2011a) and ΔV = 4.7 km s⁻¹, and τ_0,i is the optical depth at the line center:

$$\begin{eqnarray}&&{\tau }_{0}=\sqrt{\displaystyle \frac{\mathrm{ln}2}{\pi }}\displaystyle \frac{{A}_{{ul}}N{\lambda }_{0}^{3}}{4\pi {\rm{\Delta }}V}\left({x}_{l}\displaystyle \frac{{g}_{u}}{{g}_{l}}-{x}_{u}\right).\end{eqnarray} \tag{ 3 }$$

Here, x_u and x_l are the level population of the upper and lower states, respectively, g_u and g_l are their respective statistical weights, A_ul is the Einstein A coefficient, giving the rate of spontaneous emission, and N is the total column density of the molecule.

To find the best fit for each molecule a grid of models is generated with three free parameters: column density N (grid spacing ${\rm{\Delta }}\mathrm{log}N=0.2$ dex), excitation temperature T (ΔT = 50 K), and emitting area π R² (${\rm{\Delta }}\mathrm{log}R=0.003$ dex):

$$\begin{eqnarray}&&F(\lambda )=\pi {\left(\displaystyle \frac{R}{d}\right)}^{2}{B}_{\nu }(T)(1-{e}^{-\tau (\lambda )}),\end{eqnarray} \tag{ 4 }$$

where d is the assumed distance to SY Cha of 180.7 pc. Note that R does not correspond to a physical radius in the disk, but rather is the radius of a circle with the same area as the derived emitting area. As such, it R can be considered a lower limit on the physical radius of the emission, though one that is not motivated by any physical model. The resulting model spectra are convolved to the resolution of the corresponding MRS band, using the average values reported by Jones et al. (2023) for each band, then resampled to the same wavelength grid as the observations using the SpectRes package (Carnall 2017).

For a given subband, the spectral lines are fit for one molecule at a time, following the method of Grant et al. (2023). The goodness of fit for a given model is determined by comparing to the observations at each wavelength using a χ² routine. The χ² fitting is restricted to regions where emission from the given molecule is expected and there is no expected strong emission from other species based on initial slab models using typical emitting temperatures for each species. Additionally, pixels where the continuum-subtracted flux is less than –5 mJy are excluded. This excludes both potential absorption features and single pixels with anomalously low values.

After a best-fit model for a given molecule has been identified, this model is subtracted from the spectrum before fitting the next species. Molecules are fit in the following order: H₂O, OH, CO, CO₂, HCN, then C₂H₂. After obtaining an initial fit for all species, we rerun the fit for H₂O with the best-fit models from all other species subtracted in order to account for any small contributions from other species overlapping with the water emission. This second round of fitting results in only small variations, within the reported uncertainties, compared to the first round, demonstrating that the emission from other species has only a minor impact on our fitting routine, at least for species with strong lines. For weaker lines, it is possible that this method leads to an underestimation of the line flux. Molecular fits are only run when the molecule is expected to emit within the subband. The continuum-subtracted spectra and best-fit models for the fitted regions are shown in Figures 3 and 4 and the χ² maps for the model grid are shown in Figures 8–10.

Atomic hydrogen emission at mid-infrared wavelengths arises primarily from the hot disk atmosphere, though jets and outflows can also contribute. We detect four hydrogen recombination lines shortward of 10 μm, with the strongest being the H i (6–5) line at ∼7.5 μm. We do not detect the H i (7–6) line, which is often used as an accretion tracer, (e.g., Rigliaco et al. 2015; Beuther et al. 2023), though it may be masked by the presence of multiple strong H₂O lines in the vicinity.

Figures 3 and 4 show the continuum-subtracted spectra as well as the best-fit slab models. The best-fit model parameters are given in Table 3. After continuum subtraction, the MIRI-MRS spectrum of SY Cha is dominated by rovibrational and pure rotational H₂O vapor emission, which is seen in all subbands (Figures 3 and 4). As with observations of other disks observed with MIRI-MRS, e.g., Kóspál et al. (2023), Banzatti et al. (2023a), Gasman et al. (2023), the lines at longer wavelengths preferentially trace cooler gas (Table 3). In addition to H₂O, we detect CO, C₂H₂, HCN, CO₂, and OH. The best-fit column densities for H₂O are high compared to most other disks observed with MIRI, where typical column densities are of order 10¹⁸ cm⁻², while the column densities for the carbon-bearing species are on the low end (Banzatti et al. 2023a; Gasman et al. 2023; Grant et al. 2023; Perotti et al. 2023; Tabone et al. 2023). Our slab model fits do not provide tight constraints on the temperature of the emitting gas. This is likely due to the limits of fitting multiple transitions with a single slab model. As demonstrated by Banzatti et al. (2023a, 2023b), H₂O lines close together in frequency space can originate from physically distinct locations, with distinct densities and temperatures. In the future more sophisticated modeling, e.g., using nested slab models or full radiative transfer of a model disks, will be needed to provide tighter constraints.

Table 3. Best-fit Model Parameters per Spectral Channel

Molecule	N	T	R
	(cm−2)	(K)	(au)
4.9–5.7 μm	σ5.44−5.452 = 0.59 mJy
H2O	2.0 × 1019	1020	0.026
CO	3.0 × 1016	1570	0.13
5.7–6.6 μm	σ5.998−6.004 = 0.22 mJy
H2O	1.2 × 1020 ${}_{-0.7}^{+6.4}$	${1080}_{-650}^{+580}$	${0.022}_{-0.014}^{+0.42}$
6.6–7.6 μm	σ6.89−6.90 = 0.40 mJy
H2O	2.0 × 1019 ${}_{-1.0}^{+4.7}$	${970}_{-660}^{+850}$	${0.029}_{-0.019}^{+0.51}$
13.5–15.5 μm	σ14.74−14.755 = 0.62 mJy
H2O	8.4 × 1018 ${}_{-0.8}^{+1.0}$	${660}_{-120}^{+130}$	${0.17}_{-0.08}^{+0.35}$
OH	3.3 × 1018	1750 b	0.038
CO2	4.9 × 1017	210	0.27
HCN	2.7 × 1013 ${}_{-0.5}^{+4.4}$	${720}_{-190}^{+265}$	${6.9}_{-6.88}^{+3.07}$
C2H2 a	2.3 × 1017	100	2.7
15.5–17.5 μm	σ16.06−16.1 = 0.97 mJy
H2O	7.0 × 1018 ${}_{-0.7}^{+0.6}$	${510}_{-65}^{+60}$	${0.24}_{-0.08}^{+0.19}$
OH	1.2 × 1018	810 b	0.15
17.9–20.7 μm	σ19.1−19.153 = 0.33 mJy
H2O	3.1 × 1021	360	0.27
OH	1.1 × 1017	3500 b	0.060

Notes. The quoted uncertainties are the confidence intervals per fit parameter, given based on the χ² maps, and are included only for maps with a closed 1σ contour (see Figures 8–10). For N the confidence intervals are given in log-space.

^a Tentative detection. ^b Temperatures for OH reflect collisional or chemical pumping, not a physical gas temperature.

Download table as:  ASCIITypeset image

As previously mentioned, CO rovibrational emission at 4.7 μm was not detected in ground-based observations of SY Cha (Brown et al. 2013). However, rovibrational CO emission is clearly seen in our MIRI-MRS Band 1 data, which begin at 4.9 μm. Given the presence of extended emission seen in H₂, it is possible that the lower-J CO lines are obscured by an absorbing disk wind between the disk and the viewer. Previous high-spectral-resolution ground-based studies have shown that lower-J CO lines are more likely to be seen in absorption than higher-J lines, particularly for disks such as SY Cha's with inclinations greater than 40° (Banzatti et al. 2022). This is interpreted as due to a disk wind component. Alternatively the 4.7 μm emission could be hidden by telluric features in the ground-based observations, as suggested by Brown et al. (2013). Our best-fit slab model for CO gives an excitation temperature of 1570 K and an equivalent emission radius (emitting area) of 0.13 au (0.053 au²). This is consistent with the temperatures and emitting area derived from LTE slab model fits to Keck/NIRSPEC data for full T Tauri disks by Salyk et al. (2011a). However, they found transition disks tended to be best fit by a lower temperature model and smaller emitting area. That the CO emission in SY Cha is consistent with emission from full protoplanetary disks suggests that the inner disk in SY Cha remains gas rich.

¹³CO₂ is not detected. Based on the best-fit temperature and column density for CO₂, the ¹³CO₂ feature at 15.42 μm is expected to be extremely weak, peaking at less than 1 mJy. This feature also overlaps with a bright H₂O line, preventing constraints on the ¹³CO₂/CO₂ ratio.

Most mid-infrared disk spectra show either strong H₂O emission, with H₂O emission stronger than emission from other molecular species, or hydrocarbon emission (Pascucci et al. 2013; Banzatti et al. 2020, 2023a; Gasman et al. 2023; Grant et al. 2023; Tabone et al. 2023). The SY Cha spectrum contains both strong H₂O lines as well as, potentially, a weak C₂H₂ line. Figure 5 shows the best-fit C₂H₂ slab model as well as the residual observed spectrum after subtracting the best-fit H₂O, OH, CO₂, and HCN models. Emission from C₂H₂ explains residual emission that is not well fit by the fundamental bending mode of HCN. This potential C₂H₂ emission feature at 13.69 μm peaks 5σ above the rms noise of 0.6 mJy. However, this is of the same order as the residuals in other regions of the spectrum where we do not expect molecular emission. The column density and temperature of C₂H₂ are also not well constrained by the slab model fits (Figure 10). We therefore classify the C₂H₂ detection as tentative. Previously reported values for C₂H₂ in SY Cha could have been contaminated by emission from other species due to the lower spectral resolution of Spitzer (Salyk et al. 2011b; Banzatti et al. 2020). Alternatively, the C₂H₂ column density in the inner disk has changed in the 12 yr between observations.

Zoom In Zoom Out Reset image size

MIRI-MRS spectra of another transition disk, PDS 70, were recently studied by Perotti et al. (2023). The SY Cha and PDS 70 disks are superficially similar in the millimeter continuum, with a small inner disk, a wide (>40 au) dust cavity, an outer ring of millimeter grains, at roughly 70 au and 45 au, respectively, and moderate inclinations of 511 and 517, respectively (Benisty et al. 2021; Orihara et al. 2023). Both disks have been observed with ALMA at ∼20 mas resolution, though the rms noise in the PDS 70 observations is significantly lower. The inner disk of PDS 70 is marginally resolved by ALMA, extending out to 18 au, while the inner disk in SY Cha is unresolved but constrained to be smaller than 3.5 au. As previously mentioned, PDS 70 is one of the only transition disks known to host young massive planets within its dust cavity. However, the MIRI-MRS spectra of these two systems are vastly different, with the SY Cha continuum a factor of 3–6 stronger and much stronger molecular and atomic emission when PDS 70 is scaled to the distance of SY Cha (Figure 6). The shapes of the continua are also different. The slope of the continuum at short wavelengths is extremely steep in PDS 70 when compared to most T Tauri disks. The crystalline forsterite feature at 11 μm, indicative of a past heating event, is more pronounced in PDS 70, while the slope of the continuum beyond 15 μm is steeper for SY Cha.

Zoom In Zoom Out Reset image size

The best-fit slab models for H₂O from 6.7 to 7.3 μm give similar temperatures and emitting areas for both sources, roughly 600 K and 0.04 au. However, the column density for SY Cha is several orders of magnitude greater, at 5.2 × 10²¹ cm⁻² compared to 1.4 × 10¹⁸ cm⁻² in PDS 70. Emission lines from other atomic and molecular species are likewise much stronger for SY Cha.

There are several potential explanations for the different chemical compositions in the inner disks of these two systems. With a bolometric luminosity of 0.35 L_⊙, PDS 70 is less luminous than SY Cha (Pecaut & Mamajek 2016). Thus, the incident ionizing flux from the central star, a driver of disk evolution, is different. The inner disk in SY Cha could be less evolved, with less gas depletion, than the PDS 70 disk. Although the ages of these systems are highly uncertain, the best estimates suggest SY Cha is roughly 2 Myr younger than PDS 70 (Müller et al. 2018; Galli et al. 2021).

Alternatively, it is possible that the dust cavity in SY Cha was not carved by massive planets, instead being formed by other methods (e.g., Flock et al. 2015) or smaller planets, which are less efficient at preventing the inward motion of gas and small dust grains. In the case of planets, the strength of the pressure trap increases with increasing planet mass, as well as depending on disk temperature and viscosity (e.g., Fung et al. 2014; Kanagawa et al. 2015; Asensio-Torres et al. 2021). There are no published data on SY Cha from optical/near-infrared surveys searching for planets in transition disks. Likewise, the published observations of SY Cha with ALMA are not sensitive enough to detect emission from protoplanets of similar brightness to the ones detected in PDS 70. The ¹²CO gas observed by ALMA does not display a central cavity, while the less abundant ¹³CO and C¹⁸O isotopologues exhibit a ring-like morphology. In PDS 70 all molecular species, apart from the optically thick ¹²CO, observed with ALMA appear depleted inside of 02 and modeling of the disk finds a gas surface density minimum at 20 au (Keppler et al. 2019; Facchini et al. 2021; Portilla-Revelo et al. 2023). The cavity in PDS 70 is clearly far more depleted than that in SY Cha. This is likely the cause of the different chemical compositions in the inner disk.

If the cavity in SY Cha is carved by less massive planets this would point to the composition of the inner disk being strongly linked to the mass of, not just the presence of, planets at larger radii. Confirmation of this idea will require additional JWST observations of transition disks to constrain the inner disk composition as well as deep searches for substellar companions. Overall, the molecular column densities in the inner disk of SY Cha are more similar to those derived for the full protoplanetary disk of Sz 98 despite its different millimeter dust morphology (Gasman et al. 2023). The Sz 98 disk is a full protoplanetary disk with several narrow gaps observed in the submillimeter continuum. Its dust disk extends to R = 278 au, compared to ∼140 au for the SY Cha disk (Tazzari et al. 2017; Orihara et al. 2023).

It is also possible that the strong molecular emission lines in SY Cha are indicative of some level of dust grain growth in the inner disk. As dust grains grow, the dust opacity is reduced, exposing a larger gas column and resulting in stronger molecular line emission without a large change in the SED at IR wavelengths (Antonellini et al. 2023). In this case the unresolved inner disk emission seen with ALMA should be from dust emission, rather than free–free emission or an ionized wind. Additional observations at longer wavelengths would help clarify the origin of the submillimeter emission. It should be noted that the JWST data presented in the current work demonstrate the presence of small dust grains in the inner disk but do not rule out the presence of larger grains.

As discussed in Section 2.2, there is substantial variability observed in the continuum flux between the JWST observations and the three Spitzer epochs. To determine whether the molecular line flux is likewise variable we degrade the MIRI-MRS spectrum to the resolution of the Spitzer high-resolution observations from 2004 and 2008. Continuum subtraction is then carried out as described in Section 2, with the same points used for all epochs. We then integrate over the same wavelength range to get the total line flux. The wavelength ranges used for C₂H₂, HCN, and CO₂ are the same as those used by Banzatti et al. (2020) while the wavelength range for H₂O is taken from Najita et al. (2013). The results are given in Figure 7 and Table 4.

Zoom In Zoom Out Reset image size

The total line flux appears to vary not just between the JWST and Spitzer observations, but also between the two Spitzer epochs. This apparent change in brightness is of the same order as that reported by Banzatti et al. (2012) for EX Lupi for similar wavelength ranges. EX Lupi is a prototype of the highly variable T Tauri sources know as EXors, where changes are seen when comparing observations during quiescent and outbursting epochs. However, the luminosity of the line emission in SY Cha is weaker than that in EX Lupi.

To determine if the variability seen in the molecular line emission for SY Cha is significant, we calculate the uncertainty in the flux ratio between two epochs, where the uncertainty for a single observation is the calculated rms noise added in quadrature to the spectrophotometric accuracy of the instrument. Given the spectrophotometric accuracy of both Spitzer/IRS and JWST/MIRI, the variability in line flux in SY Cha is not statistically significant. The potential variability in line emission seen in SY Cha, which has not undergone any known recent outbursts, suggests that the physical conditions in the inner regions of even quiescent protoplanetary disks could change on observable timescales. Careful study of line emission in a single source at high signal-to-noise ratios over multiple epochs is needed to confirm this idea.