Main

MWC 480 is a Herbig Ae star with an estimated stellar mass of 1.8 solar masses (M⊙)9 in the Taurus star-forming region at a distance of 140 pc. The star is surrounded by a 0.18 ± 0.1 M⊙ protoplanetary disk; this is an order of magnitude more massive than the 0.01 M⊙ minimum-mass solar nebula, the lowest possible mass of the solar nebula that could have produced the Solar System10,11. Compared to disks around solar-type stars, the MWC 480 disk is 2–3 times warmer at a given radius12,13 and is exposed to levels of ultraviolet radiation orders of magnitude higher. Despite these environmental differences, the composition and abundance of volatiles in the MWC 480 disk appear largely similar to those in disks around solar-type stars, except for a lower abundance of cold (temperature T < 20 K) chemistry tracers in the outer disk14,15. The inner disk chemistry of MWC 480 has not been studied, but Herbig Ae and T Tauri disks are observed to have different volatile compositions close to their stars16.

Using the Atacama Large Millimeter/submillimeter Array (ALMA), we detected two rotational emission lines of CH3CN from the MWC 480 protoplanetary disk (the 140–130 line at 5σ, and the 141–131 line at 3.5σ, with each energy level characterized by the quantum numbers J and K, the total angular momentum and projection of the angular momentum along the molecular symmetry axis, respectively). We also detected emission lines from the N-bearing carbon chain HC3N, and from the 13C isotopologue of HCN. We targeted the 13C isotopologue rather than the more abundant H12CN because the latter is optically thick and therefore a poor tracer of the HCN abundance. Figure 1 shows the spatially resolved line detections together with a dust continuum emission map, demonstrating the spatial coincidence between CH3CN, HC3N, HCN and dust emission from the disk. The angular resolution is 0.4″–0.6″, corresponding to 50–70 astronomical units (1 au is the distance from the Earth to the Sun). The emission is also spectrally resolved, which can be used to probe smaller spatial scales. Figure 1 shows the velocity gradient across the disk that arises from Keplerian rotation in all three lines. Figure 2 shows the spectra of the three lines; each displays the double-peaked structure typical of a rotating disk. Table 1 lists the integrated line fluxes.

Figure 1: ALMA detections of simple and complex cyanides in the MWC 480 protoplanetary disk.
figure 1

a, 1.1 mm emission (black contours are 3σ + σ × 2[1,2,…]). b–d, Integrated emission of H13CN (b), HC3N (c) and CH3CN (d) lines (colour: see colour scale on the right). Black contours are [3,4,5,7,10]σ. e–g, As b–d, but for 2 km s−1 velocity bins around the source mean velocity, displaying the disk rotation. Positions are relative to the continuum phase centre (marked with a cross) at right ascension (α) 04 h 58 min 45.94 s and declination (δ) +29° 50′ 38.4″. The synthesized beam is shown in the bottom left corner of each panel.

PowerPoint slide

Full size image
Figure 2: Spectra of detected cyanides in the MWC 480 protoplanetary disk.
figure 2

The observed spectra (contours) of H13CN, HC3N and CH3CN are extracted from ALMA spectral-image data cubes and are shown as functions of the local standard of rest velocity, Vlsr. The synthetic spectra (light blue shaded histograms) are based on the best-fit disk abundance models in Fig. 3. The CH3CN spectrum contains two partially overlapping lines identified with the 140–130 and 141–131 transitions. The spectra were extracted from the spectral image cubes using a Keplerian mask to maximize the signal-to-noise ratio.

PowerPoint slide

Full size image
Table 1 Molecular data
Full size table

We use the spectrally and spatially resolved line emission to constrain the radial profiles of molecular column density based on parametric abundance models defined with respect to the adopted density and temperature structure of the MWC 480 disk (Methods). Figures 2 and 3a–c show the synthetic line spectra and maps from the best-fit models, demonstrating the good match between models and data. Figure 3d–f shows the best-fit radial column density profile, together with all profiles consistent with the data within 3σ, and abundances at 30 and 100 au from the central star are reported in Table 1. The best-fit profiles have different slopes for the different molecules. The H13CN column density decreases with radius, which is consistent with predictions from disk chemistry models17. The increasing column density with radius out to 100 au of HC3N, effectively a ring, is not predicted by models18, indicating that disk chemistry models are incomplete for HC3N. The CH3CN emission is best reproduced with a flat profile, but other profiles cannot be excluded.

Figure 3: Models of cyanide emission and radial distributions in the MWC 480 disk.
figure 3

a–c, Synthetic ALMA observations of the integrated line emission flux without added noise based on the best-fit radial column density profiles of H13CN, HC3N, and CH3CN (colour: see colour scale on the right). Black contours are as in Fig. 1. d–f, Best-fit radial column density profiles (thick dark blue line), shown together with all column density profiles that are consistent with data within 1, 2 and 3σ confidence intervals. Increasing deviations from the best fit are shown with successively lighter shades and thinner lines.

PowerPoint slide

Full size image

The absolute abundances depend on the assumed disk density structure, and we therefore compare the complex cyanides in the MWC 480 disk to cometary and to protostellar compositions using HCN as a reference species, similar to the practice in cometary studies19. We calculate abundances of CH3CN and HC3N with respect to HCN at 30 au, the smallest disk radius accessible by the ALMA observations, and at 100 au, the outer boundary of the cyanide emission maps. Accounting for the higher luminosity of MWC 480 compared to the young Sun, this radial range in the MWC 480 disk corresponds to the comet-forming zone of 10–30 au in the solar nebula19. Assuming a standard HCN/H13CN ratio of 70, the best-fit HC3N and CH3CN abundances with respect to HCN are 0.4 and 0.05 at 30 au, and 5 and 0.2 at 100 au, respectively. The CH3CN/HCN abundance ratios are robust to model assumptions to within factors of a few, while the HC3N/HCN abundance ratio may be overestimated by an order of magnitude when using our simple abundance model (Methods). Conservatively, both the CH3CN and HC3N abundances with respect to HCN are thus ∼5% at 30 au and ∼20% at 100 au. A typical comet contains 10% of CH3CN and HC3N with respect to HCN (ref. 19). The MWC 480 gas-phase cyanide composition at both 30 and 100 au is thus cometary within the observational and model uncertainties.

The relationship between gas-phase abundance ratios and the abundance ratios in ices, the main reservoirs of volatiles in disks17,18,20, depends on both desorption characteristics and chemistry (Methods). HCN, HC3N and CH3CN are characterized by similar freeze-out and desorption kinetics, but different chemical pathways. In particular, the existence of efficient grain surface formation pathways to CH3CN enhances CH3CN with respect to the other cyanides in the ice mantles. The scale of this enhancement factor varies among models, but it is at least one order of magnitude (Methods). This results in an expected minimum CH3CN/HCN ice ratio of 0.5 at 30 au in the MWC 480 disk, considerably higher than that found in the typical Solar System comet.

The CH3CN/HCN and HC3N/HCN ratios in the MWC 480 disk are also high when compared to protostars. The HCN/HC3N/CH3CN ratio is 1/0.01/0.08 towards the solar-type protostellar binary IRAS 16298-242221, and similar abundance ratios are found towards more massive systems. The MWC 480 cyanide composition is thus difficult to explain by inheritance alone, as has been suggested for H2O, for example (ref. 7). Rather, the observed high disk abundances probably reflect an efficient disk chemistry that readily converts a large portion of the carbon originally in CO and other small molecules into more complex organics22,23 during the first million years of the disk life time.

This early, efficient complex chemistry in protoplanetary disks affects the surface conditions of rocky (exo-)planets. In the ‘grand tack’ model of early Solar System dynamics24, the coupled migration of Jupiter and Saturn in the gas-rich disk from which they formed caused scattering of volatile-rich (and thus complex-organic-rich) planetesimals inwards, mixing with the asteroid belt, and outwards, producing the present-day Kuiper belt and comets25. While comets are expected (and observed) to conserve most of their original compositions due to low temperatures and thus very long desorption and chemical timescales19, asteroid-belt bodies are more processed. A combination of evaporation and chemistry probably resulted in both a net loss of volatiles and in a relative enhancement of complex organics with respect to H2O over the subsequent tens of Myr (ref. 26). A second instability 100 Myr later resulted in a heavy bombardment of the Earth by mainly asteroid-belt bodies, including the icy bodies originating in the outer Solar System27. The frequency of these instabilities in exoplanetary systems is unknown, but the large numbers of ‘hot Jupiters’ and ‘super-Earths’ too close to their host stars (within 0.1 au) to be explained by in situ formation reveal that major planetary migrations are common. When these migrations cause bombardment of icy bodies, the icy bodies in question are probably organic-rich.

The high ratio of complex to simple cyanides in the MWC 480 disk implies that the rich organic composition of comets is not unique to our Solar System, and could be commonplace. Laboratory experiments have shown that the same ice chemistry that produces CH3CN, that is, photo-processing of interstellar ice analogues, also produces simple sugars and amino acids28,29. This suggests that the early surface conditions of Earth, set by comet and asteroid bombardment, may be common for young rocky planets, and that conditions favourable to an even richer chemistry may be ubiquitous.

Methods

ALMA observations and data reduction

MWC 480 was observed with the Atacama Large Millimeter/submillimeter Array (ALMA) as a part of the Cycle 2 proposal 2013.1.00226. Observations were carried out on 2014 June 15, with 33 antennas and baselines of 18–650 m (15–555 kλ). The total on-source integration time was 22 min. The nearby quasar J0510+1800 was used for phase and gain, bandpass and absolute flux calibration (J0510+1800 was itself calibrated to an absolute flux of 1.50 ± 0.08 Jy on 2014 June 29). The correlator was configured to observe 14 spectral windows (SPWs) with resolution δν = 61.04 kHz. The H13CN 3–2 line was placed in SPW 1, HC3N 27–26 in SPW 9, and CH3CN 140–130 and 141–131 in SPW 5, each 60–120 MHz wide and centred on 258.983, 245.622 and 257.498 GHz, respectively.

The visibility data were calibrated by ALMA/NAASC staff following standard procedures. Each SPW was further self-calibrated in phase and amplitude in CASA 4.2.2. The self-calibrated phase and gain SPW-specific solutions were applied to create individual SPW line data cubes using the Briggs weighting scheme, with Brigg’s robustness parameter set to 1–2. For the weak CH3CN lines, a taper of 1″ was applied to the visibilities to maximize the signal-to-noise. The data cubes were subsequently continuum subtracted using channels identified as free of spectral line emission. The resulting r.m.s. noise in 1 km s−1 bins was ∼2.3 mJy per beam. A CLEAN deconvolution was performed with a mask produced manually based on the H13CN emission and applied to the HC3N and CH3CN spectral line cubes (after verifying that the HC3N and CH3CN cubes contained no additional emission). For the 257 GHz continuum map, obtained by combining all SPWs in the USB, the synthesized beam is 0.65″ × 0.39″ (position angle PA = −7°) and r.m.s. noise 0.3 mJy per beam. The 257 GHz continuum flux density was determined to be 331 ± 33 mJy (assuming a 10% uncertainty in the absolute calibration), which agrees well with previous Submillimeter Array results30.

Integrated emission maps were produced using the immoments task in CASA, adding emission >1σ from each channel. Lower clipping levels do not change the result at the 10% level. Spectra were extracted using both the empirical CLEAN masks and theoretical masks based on Keplerian rotation to isolate disk emission in each channel. The CLEAN masks resulted in the best signal-to-noise ratio, and these results are used in the study.

Parametric disk density and temperature model

We adopt a previously developed disk density structure of MWC 48010. The total surface density is given by the similarity solution, which self-consistently describes the density structure of an accretion disk31,32:

where r is the radius (that is, the distance from the central star), is a normalization coefficient, Rc is the characteristic radius, and γ is a gradient parameter describing the density fall-off with radius. The two-dimensional density structure in cylindrical coordinates is

where H(r) = H100(r/100 AU)h is the scale height, H100 the height at a radius of 100 AU, and h a parameter that describes the amount of flaring of the disk. The density structure parameters were constrained using high angular resolution observations of the MWC 480 dust continuum emission at 1.3 mm and 3 mm wavelengths10 and are listed in Extended Data Table 1. The resulting mass surface densities (in g cm−2) are 550, 91 and 4.7 at radii r of 1, 10 and 100 AU, respectively.

Based on existing model fits to dust and 13CO observations33,34, the disk midplane temperature is parameterized as , where T0 is set to 23 K at R0 = 100 au, and the power-law index q is 0.4. To account for the presence of a vertical temperature gradient due to heating of the disk surface by the central star34,35, we parameterize the two-dimensional temperature structure in cylindrical coordinates as

where T0 is the midplane temperature at R0, q is a power-law index describing the decrease in midplane temperature with radius, and β is a factor describing the increase in temperature at increasing disk height, which is set to 1.5 (ref. 34). The density and temperature disk structures are shown in Extended Data Fig. 1.

Molecular abundance retrieval

To retrieve molecular abundance and column density profiles, we define parametric abundance models with respect to the adopted MWC 480 disk density and temperature structure. We use a simple power-law prescription x = x100AU × (r/100 au)α for the molecular abundances, where x is the abundance of the molecule with respect to the total hydrogen density, x100AU is the abundance at 100 au, r is the disk radius in au, and α is a power-law index36. To obtain a column density profile, which is more intuitively related to observations, the abundance prescription is multiplied by disk surface density profile, equation (1). We set an outer cut-off radius (Rout) of 100 au, corresponding to the most extended emission observed for any of the cyanides. We also explored Rout of 60 and 80 au, but this did not improve the fit for any of the molecules. For each molecule, we calculate a grid of abundance models that covers x100AU from 10−14 to 10−10 with respect to nH, and α between 0 and 1 for H13CN, 1 and 2 for HC3N, 0 and 2 for CH3CN. These initial ranges of α were selected on the basis of visual inspection of the observed emission, noting its level of central concentration, that is, α = 0 corresponds to a steeply decreasing column density profile, α = 1 to an almost flat profile, and α = 2 to an increasing profile with radius.

The best-fit models are obtained by minimizing χ2, the weighted difference between the real and imaginary part of the complex visibility measured in the (u, v)-plane sampled by the ALMA observations36,37,38. We use the three-dimensional Monte Carlo code LIME39 to calculate the radiative transfer and molecular excitation. In addition to the molecular abundance profile and the disk density and temperature structure parameters, the radiative transfer modelling requires information on disk inclination, turbulence and velocity field. We adopted a disk inclination of 37°, disk turbulence of 0.05 km s−1 and a Keplerian velocity field with a stellar mass of 1.8 solar masses (ref. 10). For HCN and HC3N, the level populations were computed in non-LTE (local thermodynamic equilibrium) with collision rates listed in the BASECOL database40,41. For CH3CN, we assumed LTE, which is a reasonable approximation, since the expected disk emissive layers have higher densities than the critical density. We checked this assumption by running several HCN and HC3N models assuming LTE and found a good agreement (within 20%) between LTE and non-LTE models. In Extended Data Fig. 2 we show the best-fit models for different α. In the case of HC3N, only the models with α = 2 show the observed ring-like structure and are therefore the only ones considered further.

Within the adopted model framework, the molecular abundances and abundance ratios are constrained within a factor of a few. These models implicitly assume a constant abundance profile with disk height, which may be a poor approximation in the outer disk where freeze-out in the midplane results in a vertically layered structure20,42,43. The existing data are, however, insufficient to put any constraints on the vertical distribution. Rather than pursuing a more complex model, we opted to test the sensitivity of the derived abundances and abundance ratios on the details of the vertical structure assumptions. We ran a second grid of models to simulate the effect of freeze-out in the midplane by reducing the molecular abundances by three orders of magnitude at z/R < 0.2, where z is the disk height and R the disk radius. Using the same fitting procedure as before, we find best-fit x100au of 1 × 10−11 (α = 0), 3.5 × 10−10 (α = 2), and 5 × 10−11 (α = 1), for H13CN, HC3N and CH3CN, respectively. These abundances are an order of magnitude higher than those obtained assuming a vertically constant abundance. The CH3CN/H13CN abundance ratio profile is unchanged within the model uncertainties, however. Despite a higher excitation level of CH3CN the two molecules emit from a similar disk layer when assigned the same vertical abundance profile due to a combination of a high critical density of the CH3CN transition and a rapid fall-off in density with disk height, that is, practically no CH3CN emission originates at z/R > 0.5. In contrast, the derived HC3N/HCN ratio does depend on the vertical structure. The HC3N transition has an even higher energy level and a lower critical density than CH3CN, which implies that some of the HC3N emission can originate in more elevated disk layers than H13CN and CH3CN. The reported CH3CN/H13CN results are thus robust to the abundance model assumptions, while the reported HC3N/H13CN results may be overestimated by up to an order magnitude.

Disk chemistry modelling to constrain the ice-to-gas ratios

To constrain the origin of CH3CN and therefore how its ice-to-gas ratio compares with HCN, we first explore how much of the observed CH3CN gas could come from gas-phase chemistry. We ran a grid of pseudo-time-dependent models with a complete gas-phase chemistry44, but without any grain surface reactions (except for adsorption and desorption, and H2 formation on grains), for 1 Myr. The grid covers several orders of magnitude of densities, temperatures, ultraviolet fields and ionization fractions—the four most important regulators of gas-phase chemistry in disks. For gas-phase CH3CN production, the ionization rate is expected to be of special importance because the main gas-phase formation pathway includes (through , which has a expert-validated reaction rate in the KIDA database45). Extended Data Fig. 3 shows that none of these models can produce CH3CN/HCN > 0.01 and that in the vast majority of the parameter space CH3CN/HCN < 0.001. A ratio of 0.01 is approached for an ionization rate of 10−14 s−1, but such high ionization rates are only attained in the disk atmosphere where stellar X-rays regulate the ionization balance. This disk layer contain a very small fraction of the total disk mass and therefore contributes a negligible amount of the total molecular column. Deeper into the disk, the ionization rate is reduced below 10−17 s−1 because of attenuation of both X-rays and cosmic rays46. A ratio of 0.01 is also approached at lower ionization rates at very low temperatures and high densities. Such environments are characteristic for the outer disk midplane, which contains a lot of mass, but freeze-out at these temperatures results in CH3CN abundances below 10−14, which would not contribute to the observed CH3CN emission. The observed high abundance of CH3CN gas with respect to HCN therefore implies that grain surface chemistry contributes significantly to the observed CH3CN abundance.

We ran a number of complete disk chemistry models for 1 Myr, the estimated age of the MWC 480 star+disk system, which include gas and grain surface chemistry, as well as different levels of turbulence, to calculate the range of plausible ice-to-gas conversion factors for HCN (and other gas-phase chemistry products) and CH3CN (ref. 47). The main grain surface CH3CN formation pathways in these models are hydrogenation of C2N and CH3+CN and ice photochemistry. The main desorption pathway is ultraviolet photodesorption because of the high binding energies of HCN, HC3N and CH3CN: 4,170, 4,580 and 4,680 K, respectively48,49,50, corresponding to sublimation temperatures of ∼90 K at a gas density of 106 cm−3 and 110 K at a gas density of 1010 cm−3. The three cyanides should also present very similar ultraviolet photodesorption efficiencies because of similar radiation cross-sections and binding energies51. Without grain surface chemistry, these molecules would thus have had comparable ice-to-gas ratios.

Because CH3CN can form on the grains in the cold midplane, a key feature is the coupling between the ice formation layer and the disk layers where molecules are efficiently desorbed into the gas phase. Without such a coupling, ice chemistry cannot affect the gas-phase volatile composition. In the disk models, we define the vertical diffusion coefficient as , where αz is a free parameter, cs is the local sound velocity, and Ω is the Keplerian orbital frequency. Different levels of turbulence are simulated by different values of αz between 0 and 0.01. In models with turbulence turned off (αz = 0), there is no mixing between the cold, icy midplane and the intermediate layers where gas-phase molecules are abundant (and the observed cyanides probably reside). Extended Data Fig. 4a shows that without turbulence the predicted CH3CN abundances are low everywhere (<10−11 with respect to nH, where nH is the number density of H nuclei) and the vertically averaged CH3CN/HCN abundance ratio is <0.001; see Extended Data Fig. 5a, d.

In models that include vertical mixing (αz = 0.001–0.01), the CH3CN abundance as well as the CH3CN/HCN abundance ratio are enhanced by an order of magnitude because of mixing of midplane icy grains into ultraviolet exposed disk regions where the ices photodesorb (Extended Data Figs 4 and 5). In particular, the CH3CN/HCN ratio approaches 10−2 outside of 30 au, within an order of magnitude of the observed ratios. Model–data comparison thus suggest that the intermediate disk layers are strongly coupled to the icy midplane through mixing. This implies that gas-phase observations of CH3CN and other cyanides at intermediate disk heights can be used to trace the total volatile reservoir, even if the conversion from gas to ice can be complex.

Using the three models, with no, moderate and strong mixing, we calculate the HCN and CH3CN ice-to-gas density and column density ratios as a function of disk radius (Extended Data Figs 6 and 7). For HCN, the ice-to-gas ratio varies between 105 and 103 between 30 and 100 au. For CH3CN, the ice-to-gas ratio varies between 107 and 104. In other disk chemistry models18,52 the contrast between the two ice-to-gas ratios is even higher. Considering the range of values and their sensitive dependence on assumptions about disk turbulence, we conclude that the ice-to-gas ratio for CH3CN is at least 10 times higher than HCN and HC3N. At longer timescales the balance between formation in the midplane and destruction at more exposed disk layers determines whether the CH3CN abundance increases or decreases—in general, the more turbulence the faster the destruction.

It is important to note that we employ generic disk structures53 that have not been fitted to the MWC 480 data. The temperature and density structures are comparable (compare Extended Data Figs 1 and 4). The ultraviolet field is lower than would be expected towards a Herbig Ae star and the gas-phase CH3CN production may therefore be somewhat overestimated in our model, but as shown above, gas-phase chemistry is a minor contributor to the overall CH3CN gas budget. The grain surface production takes place in ultraviolet shielded regions and should thus not be affected by an increased ultraviolet flux. An increased ultraviolet field intensity may change the desorption/adsorption balance, however, since ultraviolet photodesorption is an important desorption mechanism. This increase in photodesorption will be accompanied by an increased level of photodissociation and the main effect will therefore be to push the desorbing layer closer to the midplane rather than increasing its abundance. An A star type radiation field may aid the detection of gas-phase CH3CN, since less turbulence may be required to bring the icy grains up to ultraviolet exposure and into the gas phase. A decrease in the disk height of the cyanide gas layer, compared to our generic disk model, may also reduce the absolute ice-to-gas ratios, but the relative ratios of species with similar adsorption and desorption characteristics should not be significantly affected.

Desorption mechanism

In our disk chemistry model, ultraviolet photodesorption regulates the desorption of HCN, HC3N and CH3CN ice into the gas phase. This result depends on a combination of disk and star properties and on the theoretical yields of different desorption mechanisms, many of which are poorly constrained. While relative ice-to-gas ratios of related species are insensitive to the details of the desorption process, the absolute ice-to-gas ratios found in our models must be considered highly uncertain. Based on theory, different desorption mechanisms should present different radial and vertical abundance distributions, which could be tested observationally. Desorption due to release of chemical formation energy should be most important in the cold disk midplane, while ultraviolet photodesorption is only efficient in warmer, upper disk layers, and thermal desorption in the disk layers that are warm enough. In particular, colder molecular emission layers should be characterized by lower excitation temperatures, which can be measured when multiple lines of the same species are observed. Observational constraints on ice desorption in disks will thus be attainable with ALMA.

Different desorption mechanisms should also exhibit different chemical signatures, although the details are often model-dependent. In the case of thermal desorption, the gas phase composition in a layer should simply depend on volatility, albeit with the added complication of ice entrapment54. Chemistry can also help distinguish between ice and gas origins and thus provide an independent measure of the relative importance of the two pathways. The high ratio of CH3CN/CH3NC in the Horsehead PDR (Photon Dominated Region), for example, was used as evidence for a grain surface origin of the observed CH3CN (ref. 55). Chemical tracers could also provide useful constraints on the incident radiation fields and temperature structure and thus aid in ruling out some of the potentially important desorption mechanisms. For example, CN/HCN is a proposed tracer of ultraviolet flux and HNC/HCN of temperature56,57.

Sample size. No statistical methods were used to predetermine sample size.