Orion SrcI's Disk Is Salty

In SrcI, the salt lines originate from the surface layers of the disk (the Band 7 lines exhibit NaCl emission peaks at ±0032 = 13 au above the disk), unlike SiO and H₂O that both exhibit high vertical extents consistent with outflow (Ginsburg et al. 2018a). No emission in the salt lines is observed toward the optically thick continuum in the disk midplane (Figure 4). Ginsburg et al. (2018a) modeled the line emission now assigned to these salts as a truncated Keplerian disk, finding that the radial distribution of the emission has an inner cutoff ≈35–40 au and an outer cutoff ≈55–60 au.

The equilibrium temperature in this radius range can be computed assuming the central source has a luminosity $L={10}^{4}\,{L}_{\odot }=4\pi {r}^{2}{\sigma }_{\mathrm{SB}}{T}^{4}$ (T_* = 4000 K, R_* = 210 ${R}_{\odot }$), where we solve for temperature with r at the inner and outer radius, giving T_eq = 500–670 K for r = 35–60 au, assuming the disk intercepts all of the starlight at this radius. More realistically, providing the opposite limiting case, T_eq = 120–180 K for a flat disk (Equation (4) of Chiang & Goldreich 1997). These cooler temperatures are consistent with the observed brightness temperatures in the outer part of the continuum disk (in Figure 4, the Band 6 and 7 NaCl emission begins just beyond the T = 300 K continuum contour), although the temperature in the inner portion of the disk reaches ≳500 K.

We also compare the salt emission locations to the SiO emission regions. Figure 5 shows the SiO v = 0 J = 8–7 line, which is thermally excited (it shows no sign of masing at any velocity) and traces the outflow. The SiO v = 5 J = 8–7 line, with an upper state energy level of 8747 K, is also shown. It clearly traces a smaller radius than the NaCl line, but a similar height in the disk. By contrast, the SiO v = 0 J = 8–7 emission, with E_U = 75 K, starts above the vertical centroid of the NaCl line. These spatial anticorrelations suggest that the difference between the molecules' emission patterns is driven partly by excitation, since the salt lines have upper state energy levels intermediate between the SiO v = 0 and v = 5 states and trace an intermediate region.

Zoom In Zoom Out Reset image size

The detection of vibrationally excited transitions from v = 0 to v = 6 at similar brightness suggests that the vibrational excitation temperature ${T}_{\mathrm{ex},\mathrm{vib}}$ of the molecules is high. However, we see sharply decreasing populations in the higher rotational levels, indicating a much lower ${T}_{\mathrm{ex},\mathrm{rot}}$.

Within each vibrational state for which we observed several transitions, we attempt to measure a rotational temperature. For the KCl v = 0 state, in which we have detected 4 rotational transitions from 30 < E_U < 400 K, the best fit is T_rot ∼ 105 K (Figure 6). At such a low temperature, the populations of the vibrationally excited states should be effectively zero.

Zoom In Zoom Out Reset image size

Similarly, for each rotational transition observed from multiple vibrational levels, we measure a vibrational temperature. We find values that range from 1000 to 5000 K, although these fits are subject to very large uncertainties. There are also statistically significant outliers from the single-temperature model, suggesting that the level populations may not be well-characterized by a single temperature.

Below, we explore several mechanisms for the observed line excitation:

(1) Collisional excitation by ${{\rm{H}}}_{2}$. Quintana-Lacaci et al. (2016) calculated collision rates for NaCl with He and provided a Fortran code to compute those rates. Typical collision rate coefficients for ΔJ = 1 or ΔJ = 2 transitions are 10⁻¹⁰ cm³ ${{\rm{s}}}^{-1}$. Making the usual assumption that the collision rates with ${{\rm{H}}}_{2}$ are similar to those for He, we used RADEX (van der Tak et al. 2007) to examine the excitation.⁹

The NaCl transitions we observe have Einstein A-values in the range A ∼ 10⁻³–10⁻² s⁻¹, while the vibrationally excited states have infrared transitions with rates A ∼ 1 s⁻¹ Barton et al. (2014), Cabezas et al. (2016). The collision rates into the v = 0 states are C ∼ 10⁻¹¹ cm³ s⁻¹ (insensitive to temperature), while collision rates into v > 0 states from v = 0 states are lower, C ∼ 10⁻¹³–10⁻¹² cm³ s⁻¹ at T = 1000 K, but only C < 10⁻¹⁵ cm³ s⁻¹ at T = 100 K Quintana-Lacaci et al. (2016). Thus, extremely high gas temperatures and densities would be required to collisionally excite a significant population of the molecules into vibrational states; collisional excitation alone also cannot explain different vibrational states being observed at a similar brightness level.

(2) Collisional excitation by electrons. Because NaCl and KCl are extremely polar molecules, with dipole moments of 9.1 and 10.3 Debye, respectively (Barton et al. 2014), collisions with electrons could also be important. Using the approximations from Dickinson & Richards (1975), we find that collision rates between both NaCl and KCl and electrons in the J = 5–50 v = 0 states are C ≈ 10⁻⁵ cm³ ${{\rm{s}}}^{-1}$. Thus, electrons are 10⁵ times as efficient at exciting these molecules than ${{\rm{H}}}_{2}$. In typical low-mass disks, ionization fractions X_e ≥ 10⁻⁴ occur in the upper layers of the disk, where some UV radiation is able to ionize H and C (Bergin & Tafalla 2007). Collisions with electrons would dominate the excitation of the salt lines in such regions. However, deeper into the disks, only X-rays and cosmic rays can penetrate, resulting in fractional ionizations X_e < 10⁻⁶, too small for electron collisions to be relevant.

The low rotational temperatures that we infer seem to rule out both ${{\rm{H}}}_{2}$ and electron collisional excitation of excited vibrational states, since a collision rate high enough to excite these vibrational levels would quickly thermalize their rotational ladders at high temperatures.

(3) Infrared excitation. The observed emission from vibrationally excited states suggests the presence of a significant radiation field at 25–45 $\mu {\rm{m}}$, which covers the range of ${\rm{\Delta }}v=1$ rovibrational transitions with v ≤ 6 (for NaCl, the range is 25–35 $\mu {\rm{m}}$, for KCl, it is 35–45 $\mu {\rm{m}}$). Since the selection rules for these transitions require that Δv = 1,¹⁰ the radiation density must be high enough to maintain a large population in the v = 1, 2, 3, 4, 5 states such that some fraction of photons are able to excite molecules to still higher states. Such a strong radiation field would be expected to thermalize the rotational ladders within each vibrational state to the same high temperature, which is not observed.

We experimented with several RADEX models with varying backgrounds. The fiducial system, with a half-sky-filling T_bg = 200 K (representing the optically thick disk) and a dilute 4000 K radiation field ${I}_{\nu }={(100{R}_{\odot })}^{2}/{(30\mathrm{au})}^{2}{B}_{\nu }(4000\,{\rm{K}})$ (representing the star) reasonably predicts the v = 1 and v = 0 intensities, but substantially underpredicts the v > 1 intensities. A stronger radiation field in the 25–45 $\mu {\rm{m}}$ regime is required, but such a field cannot be reasonably produced by combinations of blackbodies. If there were a dust emission feature in that range, such as crystalline silicates forsterite and enstatite (e.g., Molster & Kemper 2005), that dominated the emission spectrum in the far-infrared, it could potentially explain the observed vibrational excitation patterns. Alternatively, bright emission lines from other molecules, such as water, in the mid-infrared might also drive the peculiar excitation.

To demonstrate that radiative excitation in the mid-infrared is a plausible mechanism to explain the rotation diagrams, we performed additional RADEX modeling in which we modified the background radiation field. We adopted a model with a dilute 150 K blackbody representing the disk, a very dilute 4000 K blackbody representing the star, plus a strong Gaussian feature centered at 29 $\mu {\rm{m}}$ with width 1 $\mu {\rm{m}}$. With this radiation field, the vibrational lines are highly excited and exhibit a vibrational temperature significantly higher than the rotational temperature seen within any vibrational state. We show one example model in Figure 7. This toy model illustrates that it is possible, via radiative excitation, to obtain level population ratios similar to those observed, but only with a peculiar radiation field; similar experiments exploring a range of physical parameters with a blackbody radiation field could not reproduce the different vibrational and rotational temperatures.

Zoom In Zoom Out Reset image size

(4) Ultraviolet excitation. The high vibrational and low rotational temperatures of the salt lines resemble the "sawtooth" pattern observed for the rovibrational populations of ${{\rm{H}}}_{2}$ toward the Orion bar photodissociation region. This pattern is explained by ultraviolet excitation into higher electronic states, followed by decay back to excited vibrational levels in the ground electronic state (Kaplan et al. 2017). ${{\rm{H}}}_{2}$ rotational populations in the ground vibrational state are thermalized by collisions; UV excitation transposes this pattern into higher vibrational states.

Unfortunately, there are several problems applying this model to the salt lines. First, we find an abrupt decline in the populations of v > 6 vibrational states, with weak or ambiguous detections of NaCl in the v = 7 and v = 8 levels, and non-detections for v > 9. Unless there is some selection effect in the electronic de-excitation disfavoring higher vibrational states, it is difficult to explain this cutoff. Second, the lowest excited electronic energy levels for both NaCl and KCl are about 5 eV above the ground state, very close to the photodissociation threshold for these molecules (Zeiri & Balint-Kurti 1983; Silver et al. 1986), which may therefore be dissociated rather than excited by UV photons. Third, there may not be any UV radiation in the outer disk where these lines are detected. While UV emission may be produced in shocks in the outflow, and ∼5 eV photons could be produced by the central ∼4000 K photosphere (Testi et al. 2010), it is unclear whether it could penetrate into the disk where the salt emission is observed. Even a light haze of dust in the upper layers of the disk would likely shield the molecules from such illumination.

(5) Dust opacity. Finally, we consider the possibility that rotational temperatures, which are based on measurements of spectral lines over a wide frequency range, are underestimated because of greater dust extinction at high frequencies. Vibrational temperatures should be little affected by extinction because they are derived from lines in the same ALMA band that suffer similar dust attenuation.

Comparison of the observed line and continuum ratios across the bands suggests that dust extinction is not the dominant effect governing the rotational temperature. The KCl v = 0 lines, which are the only vibrational state detected in all three observed bands, have similar peak brightnesses at ∼100 and ∼345 GHz (${T}_{B,\max }(\mathrm{KCl}\,v=0\,13\mbox{--}12)=15.3\pm 0.9$ K and ${T}_{B,\max }(\mathrm{KCl}\,v=0\,45\mbox{--}44)=9.7\pm 1.8$ K). If the rotational excitation temperature T_ex were comparable with the observed vibrational excitation temperatures T_vib ∼ 1000 K, the 45–44 transition would be nine times as bright as the 13–12 transition. If we assume a dust opacity index β = 2, which is conservative in that it provides the greatest difference in extinction between the two frequencies, the dust opacities required to produce the observed line ratio are ${\tau }_{345\mathrm{GHz}}=2.8$ and ${\tau }_{100\mathrm{GHz}}=0.2$. These optical depths would result in a substantially (≈4×) higher dust brightness at 0.85 $\mu {\rm{m}}$ than at 3 mm, which is not observed.

Dust obscuration on sub-beam scales could preferentially block our view of some of the high frequency salt emission region, which would bias our measurements toward cooler rotational temperatures. However, the morphology of the line emission does not change with frequency, suggesting that this effect is negligible. Figure 8 shows the salt line images in each frequency band convolved to the same resolution, confirming that the morphological differences are minimal.

Zoom In Zoom Out Reset image size

We conclude that there is no simple model that explains the observed pattern of low rotational and high vibrational temperature.

Sodium and potassium are rarely observed in the dense molecular interstellar medium. They are usually assumed to be rapidly incorporated into dust grains after being ejected from dying stars (e.g., Milam et al. 2007). Since we observe NaCl and KCl in the atmosphere of the disk, it is clear that there is a zone where either dust has not formed (which is not very likely in a protostellar system) or where dust is destroyed and returned to the gas phase. We argue that the dust must be destroyed almost immediately as it is launched into the disk-driven outflow (Hirota et al. 2017). We evaluate three scenarios for salt production: sputtering off of dust by high-energy particles, gas-phase formation from atomic gas, and thermal desorption.

(1) Sputtering: In the first scenario, NaCl and KCl molecules are sputtered from the grains. The most plausible means to achieve the energies required to sputter grains is with strong shocks (Schilke et al. 1997; Decin et al. 2016). In SrcI, it seems unlikely that there would be very strong shocks in the disk, but there are high-velocity shocks throughout the outflow. Bright SiO emission is observed in the inner part of the outflow (i.e., within <100 au of SrcI), suggesting that grains are efficiently destroyed there. Without knowing the grain structure, however, it is unclear whether high-energy particles capable of destroying the grain cores are present.

Since the salts are detected close to but above the disk midplane, the grains must be sputtered very rapidly after being launched from the disk surface. The lack of salt and SiO emission in the disk midplane may be because these molecules are not in the gas phase at all in the midplane. Alternatively, radiative transfer can explain their non-detection, i.e., if the continuum background is the same intensity as the emission lines.

In this model, both SiO and the salts are in the gas phase above the disk midplane. However, while low-J SiO lines are seen throughout the outflow at very high elevations above the disk, the salt lines are limited to ≲30 au from the midplane. Even the low-J salt lines, which can be excited at the same modest densities required to excite SiO (n ∼ 10⁴–10⁶ ${\mathrm{cm}}^{-3}$), are not observed at high elevations. Excitation therefore cannot explain the morphological difference between SiO and the salts, and instead we infer that they are chemically distinct. Possible explanations are that the salts deplete back onto grains as they are entrained in the outflow, they react to form other molecules, or they dissociate into atomic form while SiO remains molecular.

(2) Gas-phase formation: Gas-phase formation of NaCl and KCl is possible from ionized Na and K. Potassium and sodium have similar first ionization potentials at 4.34 and 5.14 eV, respectively. These atoms would be nearly fully ionized between 1500 and 2000 K at a density of 10⁶ ${\mathrm{cm}}^{-3}$ (aluminum has a higher first ionization potential of 5.99 eV, and so is ionized at a few hundred K higher temperatures than sodium). These ions react with HCl to form NaCl and KCl. However, at present, only the inner <20 au region, which is currently unresolved, clearly reaches such high temperatures.¹¹

For gas-phase formation to be a significant process, then, either material in that hot inner region is being transported outward, which seems unlikely since there is a bulk outflow lifting material off of the disk, or the disk was previously substantially warmer. Under the dynamical formation scenario of the disk, in which it is the remnant of a previously larger disk from the SrcI-BN interaction (Bally et al. 2017; Luhman et al. 2017; Kim et al. 2018), energy was released as the disk re-settled into its current apparently smooth state, which could have resulted in much higher gas temperatures over the last few hundred years. This scenario would imply that the observable state of NaCl and KCl is very short lived, as these molecules are in the process of depleting onto grains. While plausible, there are many processes that need to coincide perfectly in this scenario (disk temperature history, gas density, gas-phase formation rates), so we favor the dust destruction hypotheses.

We note that, if the formation rates of NaCl and KCl are very high, it is possible that the vibrational excitation observed is from the chemical energy of formation. However, because of the short expected lifetimes of the vibrationally excited states, we regard this explanation for the excitation as very unlikely.

(3) Thermal desorption from grain surfaces. Decin et al. (2016) presented a model for the gas-phase abundance of NaCl in which it is thermally desorbed from grains at grain temperatures between 100 and 300 K, depending on the unknown binding energy of NaCl with the grain surface. Since we expect grain temperatures T > 150 K in the NaCl emission region (see Section 3.1), this model hints that simple grain heating is enough to explain the high gas-phase abundances of NaCl observed. Since NaCl is not observed within r < 30 au, though, it must be dissociated at smaller radii, which most likely requires a substantial UV field within that radius in the wavelength range 912 Å < λ < 3000 Å (Silver et al. 1986).

This model is plausible, though it relies on unknown physical parameters. Additionally, while a 4000 K stellar atmosphere (Testi et al. 2010) produces nonnegligible radiation short of 3000 Å, it is unclear whether that radiation can propagate to the part of the disk where the salts are observed, since small dust grains may pervade the system.

While NaCl and KCl are clearly detected here, there are also several transitions of AlCl and AlF in our observational bands that were not detected. Cernicharo & Guelin (1987) detected these transitions at comparable brightness to NaCl and KCl in IRC+10216. The lack of AlCl may be because SrcI's disk is oxygen-rich, and aluminum is locked into AlOH, as suggested in the Cherchneff (2012) model. Indeed, the tentative detection of AlO in our spectra supports this conclusion.

With several transitions of each of the observed isotopologues, we can in principle measure the relative abundance of the various isotopes. However, because of the uncertainties in the excitation above, these measurements should be taken with a grain of salt.

The ³⁷Cl abundance can be measured by comparing the intensity of the NaCl and Na³⁷Cl v = 3 J = 7–6, v = 4 J = 8–7 and J = 7–6, and v = 5 J = 7–6 lines assuming they are optically thin. The ratio ³⁵/³⁷Cl is r = 1.4 ± 0.4. The KCl v = 0 J = 45–44 line gives a ${}^{35}{/}^{37}$Cl ratio r = 0.8. Agúndez et al. (2012) measured this ratio to be r = 2.9 ± 0.3 in IRC+10216, and Highberger et al. (2003) reported a ratio of 2.1 ± 0.8 in CRL 2688. Our measurement is somewhat lower, but should be regarded as consistent at least until we can obtain more accurate column density measurements.

We only have one commonly observed ⁴¹KCl/³⁹KCl line, v = 0 J = 13–12, which has a ratio ${}^{39}{/}^{41}$ r = 2.4.

The detection of these refractory species in the gas phase in the atmosphere of SrcI's disk hints that other refractory molecules may be present, which could enable their first detection in the ISM (e.g., such rare species as FeO and FeS), and may enable direct measurements of the metallicity in star-forming gas.

The observed NaCl and KCl transitions uniquely trace the disk, unlike most other molecules observed in the ISM that are bright and abundant in hot cores like Sgr B2 (Nummelin et al. 1998; Belloche et al. 2013). These lines therefore potentially represent tracers that can uniquely trace high-mass disks, making them extremely valuable probes of early-stage high-mass star formation, at least at distances where the salt-emitting regions can be resolved. However, it remains unclear whether SrcI is a unique source or is representative of the class of high-mass protostars with disks; if the latter, these lines and potentially many others can be used to probe the radiation environment of extremely embedded high-mass protostars. An obvious next step is to search for these lines in other resolved disks around high-mass protostars (e.g., HH80/81; Girart et al. 2017). These species are particularly interesting for future millimeter and centimeter wavelength observatories like the ngVLA and SKA, since they have lower-J transitions into the few centimeter range, wavelengths at which the dust is very likely to be optically thin even in the inner few au of a disk.

The detection of these species with substantial population in vibrationally excited states highlights the need for laboratory work to measure such transitions in known astronomical molecular species. Indeed, prior work on vibrational excitation in modeling and assigned vibrationally excited ethyl cyanide in ALMA observations of Orion-KL revealed that even when catalogs for these states are available, they are often incomplete (Fortman et al. 2012).