Salt, Hot Water, and Silicon Compounds Tracing Massive Twin Disks

We report results of 0.05"-resolution observations toward the O-type proto-binary system IRAS 16547-4247 with the Atacama Large Millimeter/submillimeter Array (ALMA). We present dynamical and chemical structures of the circumbinary disk, circumstellar disks, outflows and jets, illustrated by multi-wavelength continuum and various molecular lines. In particular, we detect sodium chloride, silicon compounds, and vibrationally-excited water lines as probes of the individual protostellar disks at a scale of 100 au. These are complementary to typical hot-core molecules tracing the circumbinary structures on a 1000-au scale. The H2O line tracing inner-disks has an upper-state energy of Eu/k>3000K, indicating a high temperature of the disks. On the other hand, despite the detected transitions of NaCl, SiO, and SiS not necessarily having high upper-state energies, they are enhanced only in the vicinity of the protostars. We interpret that these molecules are the products of dust destruction, which only happens in the inner disks. This is the second detection of alkali metal halide in protostellar systems after the case of the disk of Orion Source I, and also one of few massive protostellar disks associated with high-energy transition water and silicon compounds. These new results suggest these"hot-disk"lines may be common in innermost disks around massive protostars, and have great potential for future research of massive star formation. We also tentatively find that the twin disks are counter-rotating, which might give a hint of the origin of the massive proto-binary system IRAS 16547-4247.


INTRODUCTION
Massive stars are the important sources of UV radiation, turbulent energy, and heavy elements in galaxies. Massive close binaries are the progenitors of merging black holes, which are detected by their gravitational wave emission. It is of prime importance to understand the formation process of massive stars (e.g., Tan et al. 2014). An essential question is whether massive protostars accrete through disks, so as in low-mass star formation. Recent theoretical/numerical studies have supported the disk accretion (e.g., Rosen et al. 2016;Tanaka et al. 2017;Kuiper & Hosokawa 2018). Particularly, the shielding effect by the disk inside 100 au is the key to solve the long-standing radiation pressure problem (e.g., Wolfire & Cassinelli 1987) in the formation of > 40M stars (Kuiper et al. 2010;Tanaka & Nakamoto 2011). Simulations also predict that an accretion disk tends to be gravitationally unstable, which results in accretion bursts Meyer et al. 2017Meyer et al. , 2018 and the formation of companions (Krumholz et al. 2009;Rosen et al. 2016).
Thanks to the recent development of interferometers, especially the Atacama Large Millimeter/Submillimeter Array (ALMA), more and more disk/envelope structures around massive protostars with Keplerian-like rotation have been reported (see Hirota 2018; Beltrán 2020, for recent reviews). However, so far, the number of studies reaching the resolution of 100 au is still limited (Hirota et al. 2017;Ginsburg et al. 2018;Maud et al. 2019;Motogi et al. 2019;Johnston et al. 2020). The hot and dense nature of the surrounding material of massive protostars leads to the detection of rich molecular lines within < 0.1 pc, known as hot cores. One difficulty for the disk hunting is the lack of knowledge of which lines can trace the innermost region and separate the disk from the envelope. Recently, there are some attempts to identify the disk with both kinematics and chemical patterns (e.g., Zhang et al. 2019a). However, there is no agreed-upon set of such molecular lines yet. This work will provide a tip for the disk tracing line selection based on new ALMA observations.
Our target IRAS 16547-4247 (hereafter IRAS 16547) is an O-type protostellar object with a bolometric luminosity of ∼ 10 5 L , embedded in a 10 3 M clump within a radius of 0.2 pc, at the distance of 2.9 kpc (Garay et al. 2003). Radio observations showed jets aligned in a northwest-southeast direction, across a scale of 0.1 pc on the plane of the sky (Rodríguez et al. 2005(Rodríguez et al. , 2008. The presence of jets indicates ongoing accretion in the vicinity of the protostar. Recently, Zapata et al. (2015Zapata et al. ( , 2019 reported a binary system seen as compact dusty objects with an apparent separation of 300 au, surrounded by a circumbinary disk, using ALMA observations. Using vibrationally-excited CH 3 OCHO and CS transitions with upper-state energies of E u /k > 500 K, Zapata et al. (2019) showed the circumbinary disk is rotating with a Keplerian-like profile of an enclosed mass of 25±3M . However, the dynamics at the several ×100 au scale must be controlled by the individual binary protostars, which has not been well studied yet.
We report new multi-band ALMA observations toward IRAS 16547 with resolutions of 0.05 at 1.3 and 3 mm. In this Letter, we mainly present the detection of sodium chloride, silicon compounds, and water lines as probes of the individual circumstellar disks. We propose that these inner-disk tracers may be common around massive protostars at the scale of 100 au, and valuable in understanding the disk properties in massive star formation.

OBSERVATIONS
The 3 mm observations were carried out with ALMA in Band 3 on June 9, 2019 (ALMA project ID: 2018.1.01656.S), with 45 antennas and baselines ranging from 83.1 m to 16.2 km. J1617-5848 was used for bandpass and flux calibration, and J1706-4600 for phase calibration. The 1.3 mm observations were carried out in ALMA Band 6 on July 15, 2019, in the same project, with 42 antennas and baselines ranging from 138.5 m to 8.5 km. J1427-4206 was used for bandpass and flux calibration, and J1706-4600 for phase calibration. We also utilize the ALMA archived Band 7 (0.85 mm) data, which were obtained on August 21 and 22, 2017 in project 2016.1.00992.S (Zapata et al. 2019), with 44 antennas and baselines ranging from 21.0 m to 3.7 km. J1427-4206 was used for bandpass calibration, J1617-5848 for flux calibration, and J1636-4102 for phase calibration. The total integration times were 90, 78, and 100 min in Band 3, 6 and 7, respectively. The data were calibrated using CASA (McMullin et al. 2007) pipeline v5.6.1. After pipeline calibration, we performed phase self-calibration for all the three bands using the continuum data combining line-free channels of all the spectral windows, and applied the self-calibration solutions to the line data. Images are made with CASA task tclean with the robust weighting parameters of 0.5 for Band 3 and 6 data, and −0.5 for Band 7 data. The resultant synthesized beams of the continuum images are 0.048 × 0.046 at 3 mm, 0.055 × 0.038 at 1.3 mm, and 0.056 × 0.046 at 0.85 mm, which corresponds to ∼ 150 au. Figure 1 shows the 1.3 and 3 mm continuum maps. The dust emission dominates the 1.3 mm continuum, highlighting the circumbinary disk and outflow cavities, while the 3 mm continuum reveals the jet structures. The structures seen in the 1.3 mm continuum are very similar to those in 0.85 mm continuum, which was first reported by Zapata et al. (2019) (Appendix Figure 6). Three protostars are prominent at all wavelengths, namely IRAS 16547-Ea and IRAS 16547-Eb (hereafter, sources A and B) forming the proto-binary with an apparent separation of 300 au, and a much weaker third source IRAS 16547-W. Using the 0.85 mm fluxes, we evaluate circumstellar disk masses of 0.19 M and 0.035M around sources A and B within a radius of 0.035 (150 au) assuming a dust temperature of 350 K (Appendix A). The proto-binary is surrounded by a circumbinary disk of 2500 au, outflow cavities are seen on the northern and southern sides of the circumbinary disk (see also Zapata et al. 2019  The 3 mm continuum newly reveals jet knots from source A aligned in a northwest-southeast direction, consistent with the orientation of the central radio source detected by centimeter observations (P.A. = −16 • , Rodríguez et al. 2005Rodríguez et al. , 2008. The resolution of the ALMA observation is an order of magnitude higher than those in the previous radio observations, which allows us to spatially resolve this central radio source into sources A and B and several jet knots, and to determine that the jet originates from source A. The jet orientation is also close to the elongated distribution of water masers (Franco-Hernández et al. 2009). The prominence of the proto-binary and jet knots in 3 mm continuum suggests that they are dominated by free-free emissions, and the jet knots may also contain significant synchrotron con-tributions. We leave the detailed analysis of the multiband continuum to a forthcoming paper.

Lines
Rich molecular lines are detected in IRAS 16547, especially in Bands 6 and 7. Figure 2 shows the integrated intensity maps of representative emission lines, which trace different components in the proto-binary system from the circumbinary disk to the individual circumstellar disks (see Appendix B for the summary of the lines presented in this work). Methyl cyanide CH 3 CN, which is commonly used as a disk tracer toward massive protostars (e.g., Johnston et al. 2015;Beuther et al. 2017;Johnston et al. 2020), associates with the circumbinary disk and the outflow cavity at the 1000-au scale (panels a and b). We detect the CH 3 CN (12 K -11 K ) K-ladder from K = 0 to K = 11 with excitation temperatures from ∼ 60 to ∼ 600 K. Here as representatives, K = 4 and K = 8 lines are shown as they are less contaminated from neighboring lines. The emission of sulfur dioxide SO 2 , another typical hot-core molecule, with E u /k = 403 K, also traces the circumbinary disk and the outflow cavity (panel c). However, peaks of these lower-energy transitions of CH 3 CN and SO 2 with E u /k 1000 K do not coincide with the positions of sources A and B, due to self-absorption and/or absorption against the compact continuum sources in slightly red-shifted velocities, indicating that they trace the outer cooler infalling material. This wide distribu-tion makes it difficult to study the innermost regions of a few hundred au by these lines. On the other hand, the vibrationally-excited transitions of SO 2 , CS, and H 2 O with upper-state energies of E u /k 1000 K trace the innermost region of the circumbinary disk and the individual protostellar disks (panels d-f). In particular, the H 2 O v 2 = 1 emission with E u /k = 3464 K is concentrated at the positions of sources A and B (panel f; the extended emission comes from contamination of other lines; see Figure 3d). Such a high upper-state energy reflects the high temperature of protostellar disks in massive star formation at several hundred au. With lower E u , the SO 2 v 2 = 1 and CS v = 1 lines also trace the rotation of the circumbi-   nary disk on the 1000 au scale (see below). We note that Zapata et al. (2019) first reported the CS v = 1 emission tracing the rotating circumbinary disk, but its connection to the individual circumstellar disks were not known. Furthermore, we newly found that the emissions of NaCl, SiO, and SiS are also concentrated in the vicinity of the protostars (panels g-i; again the extended emissions come from contamination of other lines), in spite of their low upper-state energies of E u /k < 100K. The critical densities of these lines are not so high (∼ 10 6 -10 7 cm −3 ), which can limit their distribution to the vicinity of the protostars. This fact indicates that these refractory molecules do not exist on the 1000-au scale, but enhanced in the innermost regions of several hundred au. It is worth noting that this is the second detection of the alkali metal halide, NaCl, in protostellar systems after the Orion Source I disk (Ginsburg et al. 2019;Wright et al. 2020).
Using these lines, we can illustrate the kinematics from the circumbinary disk to the individual circumstellar disks. Figure 3a presents the moment 1 map of the SO 2 v 2 = 1 line, showing the rotation of the circumbinary disk as reported by Zapata et al. (2015Zapata et al. ( , 2019. The systemic velocity of IRAS 16547 is about −31 km s −1 (Garay et al. 2003). The rotation direction is consistent with the elongation of the circumbinary structure. Following Zapata et al. (2015Zapata et al. ( , 2019, we plot the positionvelocity (PV) diagrams along the major axis of the circumbinary disk (P.A.=50 • ), passing between sources A and B (panel b). The PV diagram of the SO 2 v 2 = 1 line shows a rotational profile with velocity increasing toward the center. However, inside 0.1 , or 300 au, the SO 2 v 2 = 1 emission does not show the high-velocity component which is expected for the Keplerian disk. Instead, the SiO emission nicely traces the central highvelocity components up to ∆v lsr ±30 km s −1 . Figure 3c shows the PV diagrams of the SO 2 v 2 = 1 and SiO emissions along a slit passing through sources A and B (P.A.= −65 • ). This PV diagram is clearly not a simple Keplerian profile inside 0.1 , suggesting the two protostars dominate the dynamics at this scale. Not only the SiO line but also the SO 2 v 2 = 1 line shows the high-velocity components (especially in red-shifted velocities), associating with sources A and B. The same is also seen in the PV diagrams of the H 2 O v 2 = 1 emission with P.A.= −65 • (panel d), where the hot-water emission prominently shows two circumstellar components. These indicate that the rotation around the binary system is smoothly connected to the rotation around the individual protostars. The two circumstellar components are not quite parallel in the PV diagrams, judging from the different direction of velocity gradient. This suggests misalignment of the rotation directions between the twin disks (see below). Note there is contamination by other lines extending to the southeast direction at v lsr −20 and −27 km s −1 . Similar contamination was also reported by Hirota et al. (2012) in the same H 2 O line in Orion Source I.
We note that the blue-shifted emission of SiO in source A is missing, probably due to self-absorption (see Figure 7 in Appendix B), indicating that the SiO emission traces the outflowing material. However, as opposed to the commonly seen extended SiO emissions tracing shocked regions along the outflow, here the compact morphology of SiO and its close association with the two protostars suggest that it traces the material just launched from the disks or the surface layers of the disks, which can show both rotation and outflowing motions (e.g., Hirota et al. 2017;Maud et al. 2019;Zhang et al. 2019a). Figure 4 shows the blue-and red-shifted emissions of selected lines of water (panel a), silicon-compound (panels b-f), and sodium-chloride (panels g-h). To better resolve the kinematics of the individual circumstellar disks, for the lines in Band 6, we further improve the resolution to ∼ 0.035 (∼ 100 au), by emphasizing data with longer baselines using a robust parameter of −0.5 (see Section 3.2). For source A, the H 2 O and NaCl emissions show velocity gradients in the NE-SW direction, which is similar to the rotation direction of the circumbinary disk (P.A.= 50 • ). Therefore, we interpret this velocity gradient as the disk rotation of source A. This orientation is also consistent with the rotating structure traced by water maser (Franco-Hernández et al. 2009) around source A. The disk A rotation is more difficult to be identified in silicon-compounds emissions because they could be blended with the outflowing motion. The velocity gradients of the v = 0 emissions of SiO and 30 SiO are ambiguous due to the strong absorption in the blue-shifted component. For the SiO emission, comparing the red-shifted emission position and the continuum peak position gives a velocity gradient direction in P.A. ∼ 10 • , which could be the outflow direction (or the direction between the outflow and the disk rotation). If this is the case, resembling the typical star-formation picture, the outflow direction would be nearly perpendicular to the disk rotation, and close to parallel with the jet-knot orientation. On the other hand, the velocity gradient of the vibrationally-excited 29 SiO line is consistent with the disk A rotation, as this line is optically thinner than the SiO v = 0 line due to its rarity and high-excitation state. For SiS, red-shifted components of v = 0 and 1 lines roughly follow the same  velocity gradients seen by H 2 O and NaCl, suggesting the existence of SiS in the disk. However, the blue-shifted component is missing in the low excitation (v = 0) map, probably due to the similar reason for SiO. The NaCl lines trace the disk components even for the lower excitations, suggesting NaCl does not exist in the outflow unlike silicon compounds.
In source B, the velocity gradients seen in the emissions of H 2 O, 30 SiO, 29 SiO, and SiS v = 0 are close to parallel to the disk A rotation, but in the opposite direction, suggesting that the circumstellar disk of source B is rotating in the opposite direction to the disk A and the circumbinary disk. The high-velocity component of the SiO v = 0 emission again shows a gradient perpendicular to the disk rotation, which may also trace the outflowing motion, similar to source A.

Salt, Silicon Compounds, and Hot Water as Disk Probes
Based on the new high-resolution ALMA observations, we identify two groups of molecular lines tracing the innermost 100-au scale of the massive binary system IRAS 16547. The first group is the vibrationally-excited "hot" lines with E u /k > 1000K. Especially, the H 2 O line with E u /k > 3000 K nicely traces the individual circumstellar disks. The second group is the refractory molecules, i.e., alkali halides (NaCl) and silicon compounds (SiO and SiS) in the case of IRAS 16547. The lines of refractory species do not necessarily have high excitation of > 1000 K, but they trace only the innermost regions around the circumstellar disks. This fact indicates they are released to the gas phase within the disks on the 100 au scale. The production pathway of these gasphase refractory molecules are probably through the destruction of dust grains, e.g., sputtering under strong shocks or radiation, and thermal desorption from grain surfaces. SiO and SiS are likely released to the gas phase in the disk and then launched to the outflow, as they are associated with both the disks and the outflows. Further multi-line high-resolution observations and their excitation analysis will be crucial to investigate chemical origin of those species. Some previous observations have reported similar molecular lines tracing the innermost regions of massive protostellar sources. Orion Source I has been intensively studied as it is the closest massive protostar candidate at a distance of 415 pc. All the disk-tracing molecules The central twin disks are revealed by high-energy transition H2O lines with Eu/k > 3000 K, as well as NaCl and silicon-compound lines that are produced by the destruction of dust grains. The circumbinary disk, the dusty outflow cavity, and the jet knots are also seen by the new ALMA observations. The blue and red colors indicate the blue-shifted and red-shifted emissions from rotation. The circumstellar disk B is found to be counter-rotating against the disk A and the circumbinary disk. The outflowing material in source B is also traced by the SiO emission. However, the evolutionary stage of Orion Source I is still debatable, and it could be an evolved star rather than a protostar (see Báez-Rubio et al. 2018). Therefore, it has been difficult to establish these molecules as typical disk probes of massive star formation. Recently, in the B-type protostar G339.88-1.26, Zhang et al. (2019a) found that the disk and envelope can be disentangled not only by kinematics but also by chemical signatures, in particular, the disk is traced by the SiO emission, the envelope is traced by complex organic molecules and the transition zone is highlighted by SO 2 emission. Maud et al. (2018Maud et al. ( , 2019 also presented the SiO and vibrationally-excited H 2 O emissions tracing a Keplerian disk around the O-type protostar G17.64+0.16 at 100 au. Consistent with these previous findings, in the case of IRAS 16547, the individual circumstellar disks on 100 au scale are traced by the NaCl, SiS, and vibrationally-excited H 2 O lines. These studies suggest that hot water, silicon compounds, and alkali halides could be commonly present in dynamical and hot massive protostellar sources, and can be used to trace the inner disk and/or the material just launched from the disk. Further systematic observations are needed to confirm the common presence of those molecular lines in massive protostellar sources. Developing from the conventional hot-core chemistry, the "hot-disk" chemistry would be an essential avenue for future research of massive star formation. This work has demonstrated the usefulness of the hot-disk lines for understanding the dynamics down to ∼ 100 au from the massive protostars. We also note that the lower-energy transitions of refractory molecules are excellent targets for future radio observations by the Square Kilometre Array (SKA) and the Next Generation Very Large Array (ngVLA), which will be able to resolve the sublimation fronts of solid materials at a 10-au scale.
An additional importance of the hot-disk chemistry around protostars is its unique link to meteoritics. The oldest materials contained in primitive meteorites, i.e., Ca-Al-rich inclusions (CAIs) and chondrules, have been sublimated or molten once in the proto-solar disk. This fact suggests that at least some materials in the pre-solar nebula must be heated to 1500K, although protoplanetary disks are typically as cool as few hundred Kelvin in planet-forming regions of several au scale (e.g., Bell et al. 2000). Therefore, it is still under debate how and where CAIs and chondrules formed. Further observations of hot-disk chemistry could provide important constraints on the gas-phase conditions of refractory species, and might give unique insights into the formation of hightemperature meteoritic components.

The Massive Proto-binary IRAS 16547
Finally, we discuss the unveiled picture of the massive proto-binary IRAS 16547, and a possible scenario of its origin (see the schematics in Figure 5). The orbital dynamics could be constrained based on the systemicvelocity difference between two sources (Zhang et al. 2019b). We find the velocity difference is as small as ∆v lsr 2 km s −1 based on the available inner-disk tracing lines (Figure 7 in Appendix B). If the two protostars are coplanar with the circumbinary disk, and orbiting the same circular path following the Keplerian profile of the circumbinary disk (the enclosed mass of 25 M ; the inclination of 55 • by Zapata et al. 2019), the expected velocity difference is about 4 km s −1 . The fact that the observed ∆v lsr is smaller than the simple Keplerian velocity suggests that the protostars are gravitationally bound.
The ionized state of surrounding environments provide hints of the evolutionary stage of massive protostars (Tanaka et al. 2016;Rosero et al. 2019;Zhang et al. 2019c). Based on the free-free fluxes at 3 mm, we estimate ionizing photon rates of 9.6 × 10 45 s −1 and 4.3 × 10 45 s −1 for sources A and B, respectively (Appendix A). Note that those are upper limits as we ignore the contribution from dust emission. The estimated ionizing-photon rates are several orders of magnitude lower than that of zero-age main-sequence (ZAMS) stars with > 2 × 10 4 L ( Davies et al. 2011), suggesting the binary stars are at protostar phase with large radii of ∼ 20 R . The evolutionary calculations of protostars proposed that such large radii of massive protostars are the consequence of high accretion rates of 10 −4 M yr −1 (e.g., Hosokawa & Omukai 2009;Haemmerlé et al. 2016).
We do not detect hydrogen recombination lines such as H40α and H42α. The non-detection of recombination lines can be explained by the line broadening with a width of 100 km s −1 (Appendix A), which is consistent with the presence of jets seen in the 3 mm continuum. Additionally, the northern jet knots are located with approximately equal intervals of 0.4 , or 1000 au, which may be evidence of periodic accretion induced by a hidden companion or some disk instability around source A. Assuming a total mass of 20 M and a jet velocity of 100 km s −1 on the sky plane, we estimate this hidden companion should have a period of 50 yr and a semi-major axis of 40 au. The proper motion of the jet knots would be detectable by follow-up observations with similar resolutions, which will provide important clues for testing the companion's presence and for understanding jet launching and precession (Rodríguez et al. 2008).
The two protostars have similar continuum fluxes and line emissions, and look coplanar with the circumbinary disk. Those features superficially link to the disk fragmentation as the origin of the binary system (e.g., Krumholz et al. 2009). A puzzling finding, however, is the tentative detection of the counter-rotating disks ( Figure 4), which are hard to form by disk fragmentation. An alternative mechanism is turbulent fragmentation at the molecular cloud core scale (e.g., Offner et al. 2010;Bate 2012;Kuffmeier et al. 2019). Although their birthplaces may be distant, some pairs of protostars can migrate to as close as 100 au, forming binary systems. The presence of turbulence leads to the random rotation of protostellar disks, which remains even after the migration (Offner et al. 2016). The turbulent fragmentation scenario would go well with the small-cluster nature of IRAS 16547 on the scale of 0.1 pc, seen in the misalignments of several outflows and jets (Higuchi et al. 2015). However, considering the actual origin of binary systems could be much more complicated, e.g., the combination of both fragmentation processes (Rosen & Krumholz 2020) and the dynamical interactions with highly-eccentric orbits (Saiki & Machida 2020), it is difficult to conclude the formation process based on the currently available information. We want to emphasize that the detection of the counter-rotation is still tentative, and follow-up high-resolution observations are required to conclude the disk orientations of IRAS 16547.

SUMMARY
We report the dynamical and chemical structures of the massive proto-binary system IRAS 16547-4247 using 0.05 -resolution ALMA observations at 3, 1.3, and 0.85 mm. We propose that (1) the lines of destructed-dust molecules, such as alkali metal halides (e.g., NaCl) and silicon compounds (e.g., SiO and SiS), and (2) the highexcitation water line with E u /k > 3000 K can act as good tracers for investigating dynamics of the innermost region of massive star formation at a scale of 100 au. Figure 5 shows the schematic view of the proto-binary IRAS 16547 unveiled by this study. In the scale of 1000 au, the rotation of the circumbinary disk is revealed by emission lines of typical hot-core molecules, such as CH 3 CN and SO 2 , with upper-state energies of E u /k 100-1000 K (Zapata et al. 2019). However, these lines cannot well trace the protostellar disks at a 100-au scale. Instead, we found some molecular lines, including vibrationally-excited water, silicon compounds, and sodium chloride, exclusively trace the individual circumstellar disks. The detection of vibrationally-excited lines in H 2 O, SiO, SiS, and NaCl with upper-state energy as high as > 2000-3000 K indicates the very high temperature in the innermost disks. Because sodium chloride and silicon compounds are produced through the destruction of dust grains in the dynamical disks, their emissions are seen only in the vicinity of protostars, even for the lower-energy transitions with E u /k < 100 K. Using these new disk probes, we analyzed the disk kinematics, and tentatively discovered that the twin disks are counter-rotating. The pair of the counterrotating disks might suggest that the binary system is formed via turbulent fragmentation at the cloud-core scale rather than disk fragmentation. However, more observations are needed to confirm the rotation directions of disks.
Notably, this is the second detection of salt in protostellar systems after the case of the disk of Orion Source I (Ginsburg et al. 2019), and also one of few massive protostellar disks associated with high-energy transition water and silicon compounds. These new results suggest these "hot-disk" lines may be common in innermost disks around massive protostars and can be detected in high-resolution observations. Such "hot-disk" chemistry would have great potential for future research of massive star formation. We estimate the properties of the protostars based on results of multi-wavelength ALMA observations. The disk masses can be estimated using the dust flux of S ν,d ,

ACKNOWLEDGMENTS
where D = 2.9 kpc is the distance to IRAS 16547, Ω is the solid angle of the integrated region, κ ν,d is the dust opacity per gas mass, and B ν (T d ) is the Planck function at the dust temperature T d , respectively. We utilize the 0.85 mm fluxes of S ν,d = 84 and 60 mJy within 0.05 for sources A and B, because the dust emission should dominate at this wavelength ( Figure 6). The dust temperatures in the disks are uncertain from the currently available data, but the peak brightness temperature of 340 K at 0.85 mm suggests the high temperature of T d 350 K. Here we assume the dust temperature range of T d = 350-500 K. Considering the physical condition of the disks, we apply the dust opacity of κ ν,d = 0.097 cm 2 g −1 for the coagulated dust model in the high density of 10 8 cm −3 without the ice mantle (Ossenkopf & Henning 1994) (a gas-to-dust ratio of 100 is assumed). We evaluate the disks masses as 0.033-0.19 M and 0.019-0.035 M for sources A and B, respectively, at the dust temperature range of T d = 350-500 K.
Under the assumption of the optically-thin free-free emission, we can evaluate the ionizing-photon rates of the protostars (Schmiedeke et al. 2016), where T e is the electron temperature, which we use the typical values of 8000 K (Keto et al. 2008). We adopt the 3 mm continuum for the free-free fluxes, which is the upper limit because the dust emission would contribute. Based on S ν,ff = 36 and 17 mJy within 0.05 , we evaluate the ionizing-photon rates of 9.7 × 10 45 s −1 and 4.3 × 10 45 s −1 for sources A and B, respectively. The evaluated rates are orders of magnitude lower than ZAMS stars with luminosities > 2 × 10 4 L (Davies et al. 2011) (note the total luminosity of IRAS 16547 is ∼ 10 5 L ), confirming sources A and B are still at the protostellar phase with large radii. Assuming the bolometric luminosities are 5 × 10 4 L , we estimate the stellar radii of 16 R and 17 R for sources A and B, respectively (Tanaka et al. 2016). Such large radii of the massive protostars suggest that both protostars have grown with high accretion rates 10 −4 M yr −1 (e.g., Hosokawa & Omukai 2009;Haemmerlé et al. 2016). We do not identify the hydrogen recombination lines of H26α, H30α, H40α and H42α (353.6227,231.9009,99.0230,and 85.6884 GHz,respectively), which is a suggestive hint of the broadening effect due to the high-velocity jets. Under the assumption of the optically-thin and local thermal equilibrium (LTE) conditions of the recombination lines, we can estimate the the ratio of hydrogen recombination lines to the free-free continuum (Anglada et al. 2018), where Y + is the ratio of the He + and H + column densities (we use the typical value of Y + = 0.08). Again, we adopt the 3 mm continuum flux as the free-free emission. Taking into account of the peak intensity I ν,ff = 7.2 mJy beam −1 and the root-mean-square noise 0.4 mJy beam −1 , the non detection with 5σ level for recombination lines at Band 3 (H40α and H42α) suggests the line width of 100 km s −1 , which is consistent with the presence of the jets. We note that this is a conservative limit because the 3 mm flux also contains the dust emission. Table 1 summarizes the emission lines presented in this study. We particularly discuss the detection of sodium chloride, silicon compounds, and hot water as the disk probes at the 100-au scale. Figure 7 shows the spectra of these emissions at the continuum peaks of sources A and B. The systemic velocity of IRAS 16547 is about −31 km s −1 (Garay et al. 2003). The line width of water, sodium chloride, and SiS (12-11) (v = 1) are about 20 km s −1 . The Color scale / grey contours: 0.85 mm continuum Cyan contours: 3 mm continuum 1000 au Figure 6. Same as Figure 1, but showing the 0.85 mm continuum image in color scale and grey contours. The contour levels are 5σ × 2 n (n = 0, 1, ...), with 1σ = 1.6 K (0.40 mJy beam −1 ). The synthesized beam is 0.056 × 0.046 (P.A. = 85.9 • ) for the 0.85 mm continuum image (shown at the lower-left corner). water line is contaminated by the other lines at around v lsr = −20 and −27 km s −1 (see also Figure 3d). The SiO (5-4) and SiS (12-11) (v = 0) emissions trace not only the disks but also the outflows. Therefore, they are broader than other lines with clear absorption features in the blue-shifted side (see also Figure 3h and i). We do not identify a clear line-of-sight velocity difference between the two protostars, i.e., ∆v lsr 2 km s −1 , which indicates that the binary system is gravitationally bound. Note-Line information of H2O is taken from the Jet Propulsion Laboratory (JPL) line database (Pickett et al. 1998), the information of other lines are taken from the Cologne Database for Molecular Spectroscopy molecular line catalog (CDMS) (Müller et al. 2005). . Line spectra of water, silicon compounds, and sodium chloride at the continuum peak positions of sources A (red) and B (black).