DISCOVERY OF SiCSi IN IRC+10216: A MISSING LINK BETWEEN GAS AND DUST CARRIERS OF Si–C BONDS

Interstellar dust grains are synthesized in two main types of sources: the inner winds of asymptotic giant branch (AGB) stars and the ejecta of massive stars, mainly supernovae (SNe). Dust grains are formed from molecular seeds. In AGB stars, molecules such as TiO, VO, ZrO, C₂, CN, and C₃, among others, are known to be present in their photospheres since the beginning of the 20th century. Because of the high stability of the CO molecule, depending on whether the C/O ratio is >1 (C stars) or <1 (M stars), the gas becomes either carbon- or oxygen-rich after CO reaches its equilibrium abundance, leading to a different chemistry and therefore to different dust formation schemes. In SNe, different types of dust can be formed depending on the degree of enrichment in heavy elements. The dust formation efficiency in SNe is still a matter of debate. Recent studies with the Herschel Space Telescope point to much larger dust masses than previously thought (Matsuura et al. 2011, 2015).

Dust formation can be simplified as a two-step process: formation of nucleation seeds followed by grain growth through condensation of refractory molecules at high temperature and other less refractory species at larger distances from the star. There are, however, many mysteries in this picture of events, starting from the fundamental step of the formation of the nucleation seeds, which essentially are refractory species (Gail 2010). The presence of SiC grains in C-rich AGBs was confirmed by the detection of an emission band at ∼11.3 μm (Treffers & Cohen 1974). This band has been found toward a large number of C-rich stars with the IRAS and Infrared Space Observatory satellites. However, the molecular precursors of SiC dust grains are still unknown. SiC molecules have been detected in the external shells ($\geqslant 300$ R_*) of IRC+10216 (Cernicharo & Gottlieb 1989; Patel et al. 2013). In the inner layers, the most abundant molecule harbouring a Si−C bond is SiC₂ (Thaddeus et al. 1984; Cernicharo et al. 2010), whereas the most abundant Si-bearing species is SiS, which, together with SiO and SiC₂, account for a significant fraction of the available silicon (Agúndez et al. 2012).

CW Leo, located ≃130 pc from us, is a Mira variable star with a period of 630–670 days and an amplitude of $\simeq 1$ mag in the K band (Menten et al. 2012). It is one of the brightest infrared sources in the sky. Due to its close proximity, IRC+10216, the circumstellar envelope of CW Leo, has attracted many studies because it is exceptionally rich in molecular species. Half of the known interstellar species are observed in this C-rich envelope. The observed molecules range from CO, the main tracer of the cool molecular gas, and other diatomic and triatomic species (see, e.g., Cernicharo et al. 2000, 2010), to molecules containing refractory elements (Cernicharo & Guélin 1987), and long carbon chain species C_nH and their anions (Cernicharo 1996a; Guélin et al. 1997; McCarthy et al. 2006; Cernicharo et al. 2008; Thaddeus et al. 2008 and references therein). Among the species detected in this source, the silicon-carbon species Si_nC_m could play an important role as seeds of SiC dust grains. The simplest members of such a family, SiC (Cernicharo & Gottlieb 1989), and SiC₂ (Thaddeus et al. 1984), are known to be present in IRC+10216. However, Si₂C (hereafter SiCSi), which is predicted to be very abundant from chemical equilibrium calculations (Tejero & Cernicharo 1991; Yasuda & Kozasa 2012) and could play a key role in the formation of SiC dust grains, has not been found so far.

In this Letter we present the detection of SiCSi in IRC+10216 based on a close coordination between laboratory, astrophysics, and radio astronomy. This synergy has resulted in the detection of 112 lines of this new species using data from the IRAM 30 m radio telescope.¹⁰ Nine of these lines are also detected with the Submillimeter Array (SMA).¹¹

In the course of searches for new molecules we have covered a large fraction of the 3, 2, 1, and 0.8 mm spectrum of IRC+10216 with a high sensitivity using the 30 m IRAM radio telescope. In the 3 mm window, the data acquired during the last 30 years cover the 80–116 GHz domain with very high sensitivity (1–3 mK). Examples of these data can be found in Cernicharo (1996a), Cernicharo et al. (2007, 2008), Agúndez et al. (2007, 2008, 2014), and references therein. Most of the 2 mm data come from the line survey of IRC+10216 carried out by Cernicharo et al. (2000), complemented with additional data obtained during the search for specific molecular species (see, e.g., Guélin and Cernicharo 2004; Fonfría et al. 2006; Agúndez et al. 2008, 2012). The sensitivity of the 2 mm observations varies between 1.5 and 10 mK. The 1 mm and 0.8 mm data come from observations carried out during the searches quoted above and a line survey between 290 and 355 GHz carried out using the new EMIR receivers in 2010 and 2011 with a sensitivity of 2–5 mK, depending on the atmospheric transmission.

For frequencies above 250 GHz the spectrometers were two autocorrelators with 2 MHz of spectral resolution and 4 GHz of bandwidth. For all other observations the spectral resolution was 1 MHz provided by filter banks or autocorrelators. The observing mode, in which we wobbled the secondary mirror by ±90'' at a rate of 0.5 Hz, and the dry weather conditions (sky opacity at 225 GHz was below 0.1 in most observations) ensured flat baselines and low system noise temperatures (${{T}_{{\rm sys}}}$ $\simeq$ 100–400 K depending on the frequency).

The selected observing method, with the off position located at $180^{\prime\prime}$ from the star, provides reference data free from emission for all molecular species but CO (see Cernicharo et al. 2015). The emission of all other molecular species is restricted to a region ≤20–30''(see, e.g., Guélin et al. 1993). The intensity scale, antenna temperature ($T_{{\rm A}}^{*}$), was corrected for atmospheric absorption using the ATM package (Cernicharo 1985; Pardo et al. 2001). The main beam antenna temperature can be obtained by dividing $T_{{\rm A}}^{*}$ by the main beam efficiency of the telescope, which is 0.81, 0.59, and 0.35 at 86, 230, and 340 GHz, respectively. Calibration uncertainties for data covering such a large observing period have been adopted to be 10%, 15%, 20%, and 30% at 3, 2, 1, and 0.8 mm, respectively. Additional uncertainties could arise from the line intensity fluctuation with time induced by the variation of the stellar infrared flux, which has been recently discovered by Cernicharo et al. (2014). All data have been analyzed using the GILDAS package.¹²

The data revealed several hundreds of spectral lines which could not be assigned to any known molecular species collected in the public CDMS (Müller et al. 2005) and JPL (Pickett et al. 1998) spectral databases and in the MADEX code (Cernicharo 2012). Most of these lines show the characteristic U-shaped or flatted profiles with linewidths of 29 km s⁻¹. However, above 250 GHz a significant number of lines are very narrow and come from the dust formation zone of CW Leo (see, e.g., Patel et al. 2011; Cernicharo et al. 2013). Among the unidentified lines observed with the IRAM 30 m telescope, we have been able to assign 112 to the rotational spectrum of SiCSi through an iterative procedure described below and a close synergy between molecular spectroscopy and astrophysics. Some selected lines among those observed in IRC+10216 are shown in Figure 1. Nine lines observed above 300 GHz with the IRAM 30 m telescopes and assigned to SiCSi correspond to unidentified lines observed with the SMA by Patel et al. (2011). The spatial distribution of some of these lines, as derived with the SMA, is shown in Figure 2.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Disilicon carbide (SiCSi) has a C_2v symmetry and a ¹A₁ electronic ground state with a modest permanent dipole moment of ∼1D along the b-inertial axis (Barone et al. 1992; Bolton et al. 1992; Gabriel et al. 1992; Spielfiedel et al. 1996; McCarthy 2015). Because the two equivalent off-axis silicon atoms are bosons, only half of the rotational levels exist (${{K}_{a}}+{{K}_{c}}$ even).

The search for lines of SiCSi in space began with the lines measured in the laboratory by McCarthy (2015), which allowed us to accurately predict frequencies with ${{K}_{a}}=0,1$ in the 3 mm domain. Although the molecule is predicted to be rather floppy, with deviations up to 10–20 MHz for transitions higher in J than those measured in the laboratory ($J\leqslant 10$), we quickly found three lines in the 3 mm spectrum of IRC+10216 that could be analyzed simultaneously with the laboratory lines. From the newly derived spectroscopic constants, which now include additional centrifugal distortion terms, we searched for additional lines until we found all the intense lines predicted in the 3 mm domain with upper level energies below 100 K. Using the spectroscopic constants derived from a combined fit to the laboratory and IRC+10216 data, we continued the search for ${{K}_{a}}=0,1$ lines in the 2 and 1 mm domains, and up to 20 additional lines were easily identified. Differences between the predicted and observed frequencies were again 10–50 MHz, but because the lines are so intense (typically 10–20 mK) and have a distinctive line shape, it was not difficult to make additional assignments (see Figure 1). Nevertheless, on occasion, we had to discard several initial assignments because subsequent predictions failed to predict new lines. With the laboratory lines and the ∼30 astronomical lines involving K_a = 0 and 1 levels, a refined set of constants were derived, and we then proceeded to search for lines with ${{K}_{a}}=2,3$ in the entire 30 m data set of IRC+10216 that was available to us. With these new astronomical measurements in hand, several additional lines with frequencies up to 180 GHz were measured in the laboratory. Most importantly, several of these lines correspond to transitions involving ${{K}_{a}}=1\;{\rm and}\;2$ levels, which enabled us to precisely determine the D_K term, and confirm the astronomical assignments with higher K_a.

Ultimately, we were able to assign 112 lines in IRC+10216 to SiCSi with $J\leqslant 48$ and ${{K}_{a}}\leqslant 5$. At least 30% are free of blending, and the uncertainties of the derived frequencies are less than 1 MHz. Other lines were slightly blended, but from their characteristic line profile it was still possible to measure their frequencies with an accuracy of ∼1 MHz. Above 250 GHz, most observations have a spectral resolution of 2 MHz and the frequencies are accurate to 1–1.5 MHz. From the unblended lines, an average expansion velocity of 14.0 ± 1.0 km s⁻¹ was derived, which is similar to that obtained for most lines in IRC+10216 (Cernicharo et al. 2000). Hence, the discovery of SiCSi in IRC+10216—which is based on 112 observed lines—is one of the most robust identifications ever published for a first detection of a molecule in space. Some 20 additional lines have been also identified, but they are so heavily blended that accurate frequencies and intensities could not be derived.

The rotational constants were derived with a Watson Hamiltonian that included distortion constants up to sixth order in the A and S reductions (Table 1). The fit with the S reduction was done with the SPFIT program Pickett (1991), while that in the A reduction was done with fitting programs associated with MADEX (Cernicharo 2012). We confirmed that both SPFIT and MADEX yielded the same results in the A reduction. The rotational constants in the A and S reductions agree rather well, but the S reduction is preferred because ${{\delta }_{K}}$ and h_JK are three orders of magnitude larger than the corresponding constants d₁ and h₂ and the predicted frequencies for high J or K_a will be less accurate. Because there are only a few observed lines with ${{K}_{a}}\gt 3$, further improvement of the rotational constants was not possible. We estimate that the calculated frequencies with $J\leqslant 60$, ${{K}_{a}}\leqslant 5$ are reliable up to 400 GHz. The list of observed lines in IRC+10216, together with the derived line parameters, energies of the upper levels, and line strengths is provided in Table 2 (online), which also gives the frequencies and uncertainties for the 22 lines observed in the laboratory (McCarthy 2015).

Table 1.  Spectroscopic Constants (MHz) of SiCSi in Comparison to SiC₂

Constant	S-reduction	A-reduction	SiC2 A-reduction
A	64074.3366(44)	64074.33623(37)	52474.1930(677)
B	4395.51772(41)	4395.621072(844)	13158.71537(204)
C	4102.13098(62)	4102.02789(107)	10441.58029(168)
${{{\Delta }}_{J}}$ $\times {{10}^{3}}$	9.66776(224)	9.73150(178)	13.20278(397)
${{{\Delta }}_{JK}}$	−0.856833(73)	−0.8572075(610)	1.5382238(882)
${{{\Delta }}_{K}}$	23.58788(178)	23.58805(148)	−1.2159(174)
${{d}_{1}}/{{\delta }_{J}}\;\times {{10}^{3}}$	−1.52630(34)	1.519832(437)	2.41191(253)
${{d}_{2}}/{{\delta }_{K}}\times {{10}^{2}}$	−0.00318(32)	5.1591(454)	8.70880(415)
${{H}_{J}}\times {{10}^{8}}$	−3.627(116)	−4.1349(949)	−8.312(460)
${{H}_{JK}}\times {{10}^{5}}$	1.9373(67)	1.93298(547)	−5.0158(993)
${{H}_{KJ}}\times {{10}^{3}}$	−1.8882(95)	−1.88755(791)	0.39025(399)
${{H}_{K}}\times {{10}^{2}}$	4.8632(204)	4.8633(168)	0.2487(529)
h1/${{h}_{J}}\times {{10}^{9}}$	−5.304(225)	−5.231(187)	−3.46(271)
h2/${{h}_{JK}}\times {{10}^{9}}$	−2.614(255)	−6586(361)	−35151(191)
h3/${{h}_{K}}\times {{10}^{3}}$	⋯	⋯	1.1162 (286)

Note. 1σ uncertainties (in parentheses) are in the units of the last significant digits.

Download table as:  ASCIITypeset image

Table 2.  Observed Lines of SiCSi in the Laboratory and in IRC+10216

(J	Ka	Kc)u	(J	Ka	Kc)l	${{\nu }_{{\rm obs}}}$(unc)	$\int T_{{\rm A}}^{*}\times dv$	Predicted$_{{\rm Lab}+{\rm Astro}}$(unc)	Eupp	Sul	σ
						(MHz)	(K km s−1)	(MHz)	(K)		(mK)
5	1	5	6	0	6	6718.9927 (0.002)	⋯	6718.994 (0.001)	8.9	2.586	lab
8	0	8	7	1	7	12018.3548 (0.002)	⋯	12018.353 (0.001)	14.7	3.692	lab
20	1	19	19	2	18	19095.4772 (0.002)	⋯	19095.477 (0.002)	89.8	4.991	lab
15	2	14	16	1	15	24506.4805 (0.002)	⋯	24506.481 (0.002)	60.4	3.644	lab
3	1	3	4	0	4	24959.6919 (0.002)	⋯	24959.692 (0.001)	5.3	1.528	lab
10	0	10	9	1	9	31198.7591 (0.002)	⋯	31198.761 (0.002)	22.4	4.861	lab
1	1	1	2	0	2	42663.0910 (0.002)	⋯	42663.088 (0.002)	3.3	0.504	lab
2	1	1	2	0	2	60248.2200 (0.020)	⋯	60248.209 (0.003)	4.1	2.493	lab
4	1	3	4	0	4	61298.7200 (0.020)	⋯	61298.746 (0.003)	7.0	4.450	lab
6	1	5	6	0	6	62974.9200 (0.020)	⋯	62974.923 (0.005)	11.6	6.340	lab
8	1	7	8	0	8	65310.0300 (0.020)	⋯	65310.040 (0.008)	17.8	8.135	lab
1	1	1	0	0	0	68154.6000 (0.020)	⋯	68154.606 (0.004)	3.3	1.000	lab
10	1	9	10	0	10	68348.4600 (0.020)	⋯	68348.474 (0.011)	25.7	9.806	lab
3	1	3	2	0	2	84424.5800 (0.020)	⋯	84424.602 (0.006)	5.3	2.002	lab
5	1	5	4	0	4	100120.6600 (0.040)	⋯	100120.678 (0.009)	8.9	3.022	lab
8	2	6	9	1	9	109709.6000 (0.040)	⋯	109709.645 (0.018)	26.2	1.447	lab
7	1	7	6	0	6	115262.8500 (0.040)	⋯	115262.899 (0.013)	14.1	4.073	lab
11	1	11	10	0	10	144033.4200 (0.040)	⋯	144033.475 (0.023)	29.3	6.327	lab
3	2	2	3	1	3	180036.9000 (0.040)	⋯	180036.916 (0.017)	13.9	1.456	lab
5	2	4	5	1	5	181405.4500 (0.040)	⋯	181405.432 (0.014)	17.6	2.548	lab
7	2	6	7	1	7	183384.5800 (0.040)	⋯	183384.603 (0.012)	22.9	3.561	lab
9	2	8	9	1	9	185976.7800 (0.040)	⋯	185976.772 (0.012)	29.8	4.527	lab
16	1	15	16	0	16	82260.5000 (0.500)	0.109 (22)	82260.891 (0.029)	59.2	13.826	1.7
3	1	3	2	0	2	84424.1900 (0.330)	0.107 (18)	84424.602 (0.005)	5.3	2.002	1.4
9	2	8	10	1	9	86429.8400 (0.500)	0.059 (10)	86429.537 (0.008)	29.8	1.888	1.1
18	1	17	18	0	18	88716.0400 (1.150)	0.091 (14)	88715.413 (0.040)	73.7	14.774	1.7

Note. 1σ uncertainties (in parentheses) are in the units of the last significant digits.

Only a portion of this table is shown here to demonstrate its form and content. A machine-readable version of the full table is available.

Download table as:  DataTypeset image

The SiCSi distortion constants are as large as those of SiC₂, a well-studied molecule also considered to be very floppy. Table 1 provides, for comparison purposes, the rotational constants derived from a fit to the lines of SiC₂ reported by Müller et al. (2012) including the same constants as for SiCSi plus additional higher distortion constants. Both, ${{{\Delta }}_{K}}$ and H_K are one order of magnitude larger for SiCSi than for SiC₂, while ${{{\Delta }}_{JK}}\;{\rm and}\;{{\delta }_{K}}$ are of the same order despite SiCSi being somewhat heavier than SiC₂. Both molecules have a very low-lying vibrational mode, around 100–200 cm⁻¹. Hence, by analogy with SiC₂ for which the antisymmetric ${{\nu }_{3}}$ mode is readily observed in IRC+10216, we might also expect to observe rotational lines of SiCSi in its symmetric mode ${{\nu }_{2}}$. Nevertheless, SiCSi has its dipole moment along the b-axis while in SiC₂ it is aligned along the a-axis which produce a significantly different rotational spectrum.

The observed line profiles of SiCSi go from U-shaped, in the case of low excitation lines lying at high frequencies and thus observed with a smaller beam, to flat-topped, for high-excitation lines lying at low frequencies (see Figure 1). Such behavior indicates that SiCSi is concentrated around the star but with a relatively extended distribution, as occurs in the case of metal-bearing species (Cernicharo & Guélin 1987; Guélin et al. 1993). The SMA maps, which correspond to transitions involving upper level energies between ∼60 and ∼100 K, show that the emission is concentrated in a region of $\simeq 6^{\prime\prime}$ in diameter (see Figure 2). A slightly larger brightness distribution size can be expected for lines involving lower energy levels (see Figure 1). We have checked if SiCSi could be responsible for some of the lines detected with ALMA in IRC+10216 (Cernicharo et al. 2013). Unfortunately, only three strong lines are within the spectral domain covered by these ALMA cycle0 observations and only one seems to be free of blending and is clearly detected in the central pixel. However, these observations were selected to trace the dust formation zone and their sensitivity is not good enough to trace the spatial distribution of weak lines.

To initially estimate the abundance and origin of the SiCSi emission from transitions involving different energies, we show in the top panel of Figure 3 a rotational diagram based on the 112 observed lines. We have adopted a Gaussian source size with a radius $5^{\prime\prime}$. The observed integrated intensities have been corrected for dilution in the beam and for the main beam efficiency of the telescope. We can distinguish three zones with different rotational temperatures. The lines with upper level energies below 60 K (zone I) have been fit with a rotational temperature, T_rot, of 31 ± 5 K and a column density of $(1.4\pm 0.4)\times {{10}^{15}}$ cm⁻². The lines with upper level energies between 60 and 200 K (zone II) have been fit with ${{T}_{{\rm rot}}}=118\pm 12$ K and a column density of $(3\pm 0.7)$ × 10¹⁵ cm⁻². Finally, the lines with upper level energies above 200 K (zone III) have been fit with ${{T}_{{\rm rot}}}=280\pm 50$ K and a column density of $(3.2\pm 0.9)\times {{10}^{15}}$ cm⁻². The rotational temperature derived for zone I could correspond well with the extended molecular ring observed at $r\;\simeq$ 14'' in SiC₂ (Guélin et al. 1993). Its value agrees well with the rotational temperature derived for other species (Cernicharo 1996a; Agúndez et al. 2008). Zone II corresponds to the emission region of metal-bearing species and high energy lines of SiC₂ with $r\leqslant 5^{\prime\prime}$ (Figure 2 and Velilla Prieto 2015). Finally, the high energy lines (zone III) correspond to the region with $r\leqslant 2^{\prime\prime}$ traced by ${\rm high}-J$ lines and vibrationally excited states of SiC₂, SiS, HCN, HNC among other species (Cernicharo et al. 2010, 2011; Velilla Prieto 2015). In this region, Cernicharo et al. (2010) derived for SiC₂ ${{T}_{{\rm rot}}}\;\simeq$ 204 K and a column density of $\sim 8\times {{10}^{13}}$ cm⁻² averaged over a beam of $30^{\prime\prime}$. When corrected for the source radius of $5^{\prime\prime}$ adopted in this work, the column density of SiC₂ becomes 3 × 10¹⁵ cm⁻². Hence, SiCSi is as abundant as SiC₂ in the dust formation zone of IRC+10216. The large difference in the intensities of the observed lines of SiCSi and SiC₂, near a factor of 100, arise from the lower dipole moment and the larger partition function of SiCSi. When corrected for these effects, the observed intensities of the individual lines of SiCSi and SiC₂ indicate a SiC₂/SiCSi abundance ratio of about 5.

Zoom In Zoom Out Reset image size

Early chemical equilibrium calculations by Tejero & Cernicharo (1991) showed that in the inner envelope of C-rich AGB stars the most abundant gas phase species containing a Si−C bond are SiC₂ and SiCSi, while SiC is predicted to be present with a much lower abundance (see also Yasuda & Kozasa 2012 and Figure 3). To constrain the abundance and spatial distribution of SiCSi in IRC+10216 we have carried out radiative transfer calculations to obtain the emergent line profiles and compare them with the observed ones. The physical model of the envelope is taken from Agúndez et al. (2012) with the downward revision of the density of particles in the regions inner to 5 R_* derived by Cernicharo et al. (2013). Since collisional rate coefficients are not available for SiCSi we have assumed local thermodynamic equilibrium, which is a reasonable approximation given the low dipole moment of SiCSi. We find that the intensities and profiles of the low excitation lines observed can be well reproduced by adopting a SiCSi abundance relative to H₂ of 4 × 10⁻⁸, from the inner layers out to the photodissociation region. We have assumed that SiCSi is photodissociated by the ambient interstellar ultraviolet field with a rate of ${{10}^{-9}}\times {\rm exp} (-2{{A}_{{\rm V}}})$ s⁻¹, where A_V is the visual extinction. However, to reproduce the high intensities observed for lines involving upper level energies above 200 K, it is necessary to increase the abundance of SiCSi in the warm inner regions to 2 × 10⁻⁷ (see Figure 3). Finally, a depletion in the very inner regions, which is consistent with the expectations from chemical equilibrium, yields a better agreement with the observed line profiles (see Figure 1). The derived abundance profile for SiCSi is shown in Figure 3.

High angular resolution interferometric observations are needed to better constrain the abundance gradient of SiCSi in the inner layers. In any case, we find that disilicon carbide is as abundant as silicon dicarbide in the inner layers. The depletion in the abundance of SiCSi at 50 R$_{*}$, which leads to a SiC₂/SiCSi abundance ratio of $\simeq 5$ in the outer regions, would be consistent with its incorporation into SiC dust grains.

We thank spanish MINECO for funding under grants AYA2009-07304, AYA2012-32032, CSD2009-00038, and ERC under ERC-2013-SyG, G. A. 610256 NANOCOSMOS. The new laboratory measurements in Cambridge were supported by NASA grant NNX13AE59G.