OBSERVED RELATIONSHIP BETWEEN CO 2–1 AND DUST EMISSION DURING THE POST-ASYMPTOTIC-GIANT-BRANCH PHASE^*

Although the relation between CO line strength and infrared dust emission strength has been studied before (e.g., Bujarrabal et al. 1992), there is still a lack of in-depth investigation. In this work, we do not try to compare the dust and gas mass loss rates. Instead, we directly compare the observed CO line fluxes with IRAS flux densities (R_CO25 ratio—see Equation (1)). The greatest advantage of the direct comparison is that it allows us to exclude the uncertainties introduced by distances and empirical formulas for mass loss rate. The traditional IRAS color–color (C–C) diagram of the two colors,

$$\begin{eqnarray*} {\rm C12}&=&2.5\log (F_{25\,{\rm \mu m}}/F_{12\,{\rm \mu m}}),\\ {\rm C23}&=&2.5\log (F_{60\,{\rm \mu m}}/F_{25\,{\rm \mu m}}), \end{eqnarray*}$$

is also used. Here, F₁₂, F₂₅, and F₆₀ are the IRAS flux densities. The C23 color is a better tracer of the cool dust emission around pAGB stars than C12. Thus, we will concentrate on R_CO25-C23 relationships of our sample stars.

Our new observational results of the high Galactic latitude (h.g.l.) pAGB stars are combined with literature data. The h.g.l. objects (we adopt |b| > 15°) should be mainly local stars or low mass halo stars statistically less massive than the low Galactic latitude (l.g.l.) counterparts. Thus, we plot their R_CO25-C23 and IRAS C–C diagrams separately in Figure 2. Compared with our new observations, there are no CO 2–1 line detections in the literature toward other h.g.l. objects. However, the average literature CO line fluxes show clear discrepancies with our new measurements for Frosty Leo and 89 Her (see the two symbols linked to the same object names in the figure). The lower literature CO line flux of 89 Her could be due to the fact that it has been partially resolved by the small beam of the IRAM telescope which was used to obtain two of the three available literature CO data sets. For Frosty Leo, the literature data was obtained by a 12 m telescope (as in this work) more than 18 yr ago. The reason for the discrepancy is unclear. The 1σ upper limits of non-detection sources are estimated by assuming a typical CSE expansion velocity of 10 km s⁻¹ (≈7.7 MHz). The R_CO25-C23 diagram will be called the CO-IR diagram hereafter.

Zoom In Zoom Out Reset image size

Shown in the upper panels of Figure 2 are h.g.l. pAGB stars. Both the literature data (gray circles) and our new observations (black dots for the detected objects) are shown, as indicated by double arrows in the figure. In the upper left panel, almost all the detected pAGB stars, excluding only the exceptional case of Frosty Leo, show very similar CO-IR flux ratios of

$$\begin{equation} R_{\rm CO25} \approx 1.04 \, (\pm 0.08) \; {\rm [MHz]} \end{equation} \tag{ 2 }$$

as delineated by the horizontal line in the figure. This relation is valid in the color range of −1.6 < C23 < 0.0 for the pAGB stars considered. The slant dashed line will be explained below.

On the IRAS C–C diagram in the upper right panel, Frosty Leo is not shown because of the poor quality of its IRAS 12 μm flux density data. The remaining four objects are distributed along an elongated region below the black body line. Particularly, the C12 colors vary more than the C23 colors. This trend of increasing colors could be the result of fast weakening of the 12 μm flux density when the inner CSE of the pAGB stars is gradually evacuated. Thus, the increasing IR color sequence of the detected h.g.l. pAGB stars could be an evolutionary sequence. Two special objects are commented on in the footnote.⁵$^,$⁶

In the two lower panels of Figure 2, the l.g.l. pAGB stars are shown. The CO-IR diagram in the lower left panel shows that many stars with CO detection (gray circles) are crowded around a similar region as the high latitude stars (with R_CO25 ≈ 1.04 and C23 < 0.0), while the rest objects seem to be located in an elongated region, extending from blue C23 color and small R_CO25 ratio to red C23 color and large R_CO25 ratio. The object names are marked out in the figure for the latter group of stars. We find that a slanted straight line (the dashed line) of

$$\begin{equation} \log (R_{\rm CO25})\approx {\rm C23}. \end{equation} \tag{ 3 }$$

can represent this trend reasonably well. This log-linear trend holds in the color range of −1.2 < C23 < 1. Equation (3) is not a fit to the data, but a simple function recognized by eye. The lower right corner of this CO-IR diagram is devoid of objects, hinting that this log-linear trend could represent a border of pAGB star distribution on the CO-IR diagram.

The l.g.l. pAGB objects, which are distributed along the slanted line and have C23 < 0 are located on or below the BB line on the IRAS C–C diagram. On the other hand, the rest objects that have C23 > 0 (with the exception of IRAS 19475+3119) are located above the BB line (see the lower panels of Figure 2). In addition, it is seen that the l.g.l. pAGB stars, which appear in a similar region as the h.g.l. ones on the CO-IR diagram (compare the two left panels) also occupy a similar region on the IRAS C–C diagrams (compare the two right panels). This indicates that these h.g.l. and l.g.l. objects share a similar combination of dust and CO gas characteristics.

Although some regularity has begun to emerge on the previously discussed R_CO25-C23 diagrams of the pAGB stars, the discussed trends are still murky due to the limited number of objects involved. In this section, we try to verify these trends in a broader context of the AGB-pAGB-PN evolutionary sequence. The CO 2–1 line data collected from literature for some AGB stars and PNe make this comparison possible. Thus, the CO-IR diagrams of these AGB stars and PNe are plotted and compared with the pAGB stars (represented by the two straight lines that delineate the major trends) in Figure 3. The traditional IRAS C–C diagrams are also shown for them. For clarity, the C-rich and O-rich AGB stars are plotted in separate panels.

Zoom In Zoom Out Reset image size

5.2.1. AGB Stars and PNe on the CO-IR Diagrams

5.2.2. The pAGB Trends on the CO-IR Diagram in the Context of AGB-pAGB-PN Evolution

Our sample of pAGB stars aggregate in different regions on the CO-IR diagram, which is not so well seen on the traditional IRAS C–C diagram. As we will see below, the aggregation on the CO-IR diagram just reflects the different nature (e.g., mass, binarity, chemistry, etc.), and the stage of pAGB evolution of these stars through the contrast of dust and CO line emission. Here, we merge the h.g.l. and l.g.l. pAGB stars and divide them into three CO-IR groups, as shown in the upper panel of Figure 4. Only significant, from the point of view of performed source classification, CO non-detections are plotted in Figure 4.

Zoom In Zoom Out Reset image size

Group-I—those pAGB stars with R_CO25 > 1/3 and C23 < 0 (the red filled circles, or gray filled squares in Figure 4). They compose the largest group of pAGB stars, which is distributed in a horizontally elongated region on the CO-IR diagram, with a similar R_CO25 ratio of ∼1 MHz (actually in the range of 0.42–5.2 MHz, with a median of ∼1 MHz). Their distribution can be roughly represented by the horizontal line that was determined for our h.g.l. pAGB stars in Section 5.1.

Group-II—those pAGB stars with C23 > 0 (the green filled circles and one gray filled square in Figure 4). They are the reddest group with usually enhanced CO 2–1 emission relative to their IR dust emission. They occupy the red part of the log-linear trend as delineated by the dashed line in the figure.

Group-III—those pAGB stars with R_CO25 < 1/3 and C23 < 0 (the blue filled circles in Figure 4). They have exceptionally weak CO 2–1 line emission, although their C23 colors are similar to that of group-I stars. They occupy the blue part of the log-linear trend (the dashed line) and the region immediately above it. This group also includes some CO 2–1 non-detections whose 1σ upper limit of R_CO25 ratios are smaller than 1/3 and thus they are also CO deficient pAGB stars. We did not plot the other non-detections, since they could be members of different groups.

All of the three CO-IR groups of pAGB stars are listed in Table 3, with different groups in separate sub-sections of the table. The vicinity of intersection between the solid and dashed lines in the top panel of Figure 4 do not allow us to be sure to which group a given object belongs. Therefore, the classification of the objects near the group borders (here we have assumed somewhat arbitrarily a rectangular border region as R_CO25 > 1/3 and −0.5 < C23 < 0) is tentative, and we labeled them with a leading "*" symbol before their object names in Table 3. The table is organized in a way to enable easy comparison between the CO-IR and IRAS C–C diagrams in Figure 4. Objects in groups-I and II are sorted with increasing C23 colors, while those of group-III are ordered in decreasing C23 colors. Also listed in the table are some properties of the pAGB stars that will be explained and discussed later in Section 6.

In the lower panel of Figure 4, the three groups are displayed also on the traditional dust emission diagnostic tool, the IRAS C–C diagram. Group-II pAGB stars are clearly separated from the other stars by their large C23 colors. However, the groups-I and III pAGB stars are well mixed on the IRAS C–C diagram, which demonstrates that the involvement of gas emission (the CO 2–1 line) in the grouping has brought us new information about the pAGB stars.

Group-I pAGB stars distribute in an elongated region on the IRAS C–C diagram, with their C23 colors varying only by a factor of about 1.5, while their C12 colors varying by a larger factor of about 4.0. The larger variation range of the C12 colors may be the natural consequence of fast weakening of the 12 μm dust emission in the expanding detached CSE. Thus, group-I pAGB stars are possibly still in the early pAGB stage when the detached CSE is still actively developing, which is also supported by the fact that their R_CO25 ratios are similar to that of AGB stars. Their similar R_CO25 ratios hint that the thermal balance in the gas (represented by the CO 2–1 line) and dust (represented by the 25 μm emission) might still be tightly coupled during the early pAGB stage.

The red C23 color and richness of CO 2–1 emission of the group-II stars indicate that they could have massive and cool CSEs. By contrast, the weakness of CO 2–1 emission and the blue C23 colors of the group-III stars suggest that they could be lower-mass pAGB stars with a more transparent spherical component in their CSEs where CO molecules have been partially or totally destroyed by penetrating UV photons. The well known low-mass binary disk system Red Rectangle (Men'shchikov et al. 2002; Witt et al. 2009) which shows a narrow CO line (Jura et al. 1995; Dayal & Bieging 1996) is such a member of group-III. However, it is not clear why the groups II and III share the same log-linear relationship on the CO-IR diagram.

For discussion in this paper, we assume that the chemical type of a pAGB object is C-rich if it has a C-rich central star and/or C-rich dust in its CSE. Also stars with dual dust chemistry in their CSE (simultaneous presence of C- and O-rich dust features and/or molecules) are treated as having a C-rich chemical type. Knowledge of the chemical type is important since it allows us to roughly estimate mass of the progenitor. Single carbon stars are formed only in a limited progenitor mass range. For solar metallicity this happens for ∼1.5 M_☉ < M_ZAMS < ∼ 5 M_☉, and the mass range shrinks and shifts toward somewhat smaller values for lower metallicities (Piovan et al. 2003). For progenitor masses lower and higher than the above mass range a star will remain O-rich. Note, however, that in close binary systems, mass transfer to a companion star may reduce the star's AGB lifetime and the star may remain O-rich even if its progenitor was of intermediate mass.

The distribution of the chemical types from Table 3 is visualized together with central star spectral types on the CO-IR diagram in Figure 5. C-rich and O-rich pAGB stars are plotted in separate rows, while those with late and early spectral types in different columns.

Zoom In Zoom Out Reset image size

In the left two panels of Figure 5, the C- and O-rich pAGB stars with late spectral types (F, G, K, and M) show distinctive distributions. C-rich pAGB stars of late spectral types are mostly group-I sources, with only few objects, like Red Rectangle, Roberts 22, and IRAS 19477+2401,⁷ belonging to group-III. On the other hand, O-rich pAGB stars with similar spectral types belong predominantly to group-III⁸ and to the transition region between different groups. Group-II objects with late spectral types are not numerous, in general, but among them there is only one C-rich star (IRAS 11385−5517) with SiC emission at 11.3 μm seen in its ISO spectrum, but also with OH emission detected from its shell (see Torun catalog and references in Table 3).

Post-AGB stars of earlier spectral types (right panels of Figure 5) are also not numerous in our sample but are located mostly along the log-linear track (dashed line in the figure), with a clear exception being V886 Her (C23 = −2.07), a massive (Ryans et al. 2003) fast-evolving O-rich pAGB star (Arkhipova et al. 1999).

From point of view of the progenitor mass, a feasible interpretation of the discussed trends is that group-I objects, being predominantly C-rich, are intermediate mass pAGB stars; the group-III sources that are CO-deficient objects are the lowest mass pAGB stars; and the group-II sources are intermediate or high mass stars, which already follow the PNe trend on the CO-IR diagram (see bottom left panel in Figure 3). This simplified interpretation is supported by the significantly large percentage of O-rich sources (about 70%; see Table 3) among groups-III and II. On the other hand, the paucity of O-rich pAGB stars in group-I region (see the two bottom panels of Figure 5) seems to suggest that O-rich pAGB stars do not evolve into this region.

From the point of view of single-star evolution, we expect that during the pAGB stage, the C23 color almost continuously increases with time, while F₂₅ flux generally decreases after a short-lasting increase at the transition phase between AGB and pAGB (see, e.g., Szczerba & Marten 1993; Steffen et al. 1998). Such behavior is due to cooling of the circumstellar shell, which is moving away from the central star. However, evolution of the CO 2–1 line flux, another key factor determining the position of a source on the CO-IR diagram, is neither simple nor investigated theoretically to our knowledge. In this respect the CO-IR diagram serves as an observational tool that allows us to put constraints on the CO 2–1 line flux behavior during pAGB phase of stellar evolution.

PNe are well concentrated along the log-linear region (see bottom left panel in Figure 3) to which pAGB stars (with exception of the low-mass ones that could disperse their circumstellar shells before the onset of photoionization) should ultimately evolve. In the frame of single-star evolution, the existence of such a trend can be understood if the CO 2–1 line flux does not change much during the late pAGB (pre-PNe) and PNe phases of evolution, while the F₂₅ μm flux density monotonically decreases. However, in the earlier stages of pAGB evolution, the F₂₅ continuum (see above) and probably also CO 2–1 line fluxes could change non-monotonically, resulting in a relatively complex distribution in Figure 5 (e.g., the lack of O-rich pAGB sources in group-I and the presence of C-rich ones among group-III).

The SEDs of pAGB stars have been classified according to the scheme introduced by van der Veen et al. (1989), with the addition of a type 0 (Szczerba et al. 2012). There are six SED types in total: types 0, I, and II, which show significant near infrared (NIR) excess, and types III, IVa, and IVb showing cold dust emission together with a second peak at shorter wavelengths from central star emission. de Ruyter et al. (2006) proposed that the NIR excess, which is seen in the first three SED types (0, I, and II) is a signature of gravitationally bounded circumbinary disks perhaps formed during strong binary interaction (van Winckel et al. 2009). The other three SED types (III, IVa, and IVb), are signatures of detached shells and/or expanding tori, which are formed due to mass loss on the AGB, or by interaction of an AGB star with its companion (Zijlstra 2007), respectively.

Among our CO-IR group-I sources, three quarters (13 out of 17) have SED types of III, IVa, or IVb, indicating that their dust emission is dominated by a detached shell/expanding torus. On the other hand, two-thirds (12 out of 18) of the CO-IR group-III pAGB stars have SED types of 0, I, or II, showing that they have excess emission from hot dust, a signature of a disk. It is natural to expect that formation of a disk is due to interaction among the stars when the primary star was a giant. It is also interesting to note that most of the pAGB sources (8 out of 11) that are located in the transition region (those marked by "*" in Table 3), are showing emission from cold (detached) CSEs (with SEDs of type III, IVa, and IVb). The situation in the CO-IR group-II is less evident, as about the same fraction of sources show a presence of hot+cold or only cold dust.

Information about binarity is directly obtained from the Torun Catalog of pAGB stars. Although such information is by no means complete in the catalog, some interesting features still can be recognized in the distribution of known binary pAGB stars on the CO-IR diagram, as shown in Figure 6. First, known binaries appear in all three regions of pAGB stars on the CO-IR diagram. However, most of them appear in the CO-deficient region of group-III. In the current sample, about 39% (7 out of 18) group-III pAGB stars are known binaries. Among the six C-rich stars that belong to group-III, two are known binary systems (Red Rectangle and RV Tau). Binaries are also common in the group-II region (about 40%). What is also striking on Figure 6 is dichotomy in C23 color distribution of binaries. One, more numerous group, have blue C23 colors (no cold dust), while the second group have red C23 colors (significant amount of cold dust, with Frosty Leo being the most extreme example).

Zoom In Zoom Out Reset image size

The CSE expansion velocity of all the 42 pAGB stars with detection of CO 2–1 lines show a profound trend: V_exp is smallest among group-III pAGB stars, intermediate among group-I stars, and largest among group-II objects. Excluding a few exceptional objects, which will be discussed below, we obtain average velocities from the CO line widths for each group: 12 ± 2 km s⁻¹ for group-I, 28 ± 12 km s⁻¹ for group-II, and 4 ± 2 km s⁻¹ for group-III. Note that the vast majority of objects located in the transition region (those marked with "*" in Table 3) have velocities in between those characteristic for group-I and group-II, with the clear exception being IRAS 17106−3046, which has a very low V_exp that is more typical of group-III objects. Because the AGB wind velocity is expected to be higher for more massive AGB stars (see, e.g., the discussions of Nyman et al. 1992), the trend we found suggests that group-II objects are statistically more massive than group-I pAGB stars. On the other hand, such low expansion velocities for group-III objects suggest that CO is observed from circumstellar disks (rotating and/or expanding) rather than from outflows (Bujarrabal et al. 2005, 2013). It is very likely that such disks are formed in binary systems (e.g., van Winckel 2003).

There are two objects in group-I (89 Her and AI Sco), which have very low expansion velocities for their CSE derived from CO lines. Both of them are known to be binaries with 0-type SEDs, O-rich chemistry, and a blue C23 color close to that typical of O-rich AGB stars. They share characteristics of group-III pAGB objects, but at the same time have a relatively strong CO 2–1 emission in comparison with their 25 μm continuum flux.

The other two group-I objects (IRAS 08005−2356 and IRAS 23541+7031), on the other hand, have very high expansion velocities, much larger even than those typical for group-II pAGB objects. IRAS 08005−2356 has a broad CO line, which was only tentatively detected by Hu et al. (1994). However, similarly high velocities are also detected in the optical spectrum of this star by Slijkhuis et al. (1991), and Sánchez Contreras et al. (2008), which may be interpreted as a signature of an ongoing fast wind (or jet) from this source. Its type-I SED indicates significant near-infrared emission, which may be interpreted as emission from hot dust in a disk, but its R_CO25 = 5.20 is the largest among group-I pAGB stars. IRAS 23541+7031 has a disk or torus of molecular gas, which seems to be expanding (Castro-Carrizo et al. 2002). It has very rich and complex wind activity and rapidly evolving shocked material (Sánchez Contreras et al. 2010). In this respect, a broad CO 2–1 line detected by Hajian et al. (1996) is not surprising.

Roberts 22 and IRAS 17245−3951 (Walnut Nebula) are two other special objects with normal or high CSE expansion velocities (31 and 15  km s⁻¹, respectively), but this time being members of our group-III objects (R_CO25 = 0.16 and 0.25, respectively). Again, the broad CO lines may be interpreted by the contribution from bipolar outflows, because both show typical bipolar nebulae. We note that Roberts 22 was known to show broadened H_α emission line and was thus suspected to be a Wolf–Rayet star (Roberts 1962).

Finally, we would like to call attention to the fact that, among sources from Table 3, there are two pAGB classes (Szczerba et al. 2007) that are the most abundant. They are "21 μm" sources (12 objects) and RV Tau stars (8 sources). These two groups have quite well defined properties. The "21 μm" sources are a group of C-rich intermediate mass stars with the still unidentified 21 μm dust feature (Kwok et al. 1989), which also show s-process element enhancement (Van Winckel & Reyniers 2000), and have periodic variability well correlated with their effective temperature (Hrivnak et al. 2010). RV Tau stars are luminous variable stars, which show alternating deep and shallow minima with periods between 30 and 150 days, and spectral types F, G, and K (e.g., Preston et al. 1963). Most of them show IR excess, which is interpreted as a remnant of AGB mass loss (Jura 1986). RV Tau stars with near-IR excess (very likely due to radiation from disks) are probably binaries (de Ruyter et al. 2005).

These two groups have well established positions on the CO-IR diagram. Almost all members of the "21 μm" type sources from our sample are in our group-I, with one belonging to the transition region and only one to group-III (IRAS 19477+2401). It is remarkable that 10 objects (out of 17, or 59%) in group-I are "21 μm" sources. Their IR colors are in the range −1.3 < C23 < −0.55. All but one have a G or F spectral type, 9 (among 12) have an SED of IVa or IVb type, none of them is known to be binary, and their outflow velocities are around 10 km s⁻¹ (with an average of 12 ± 3 km s⁻¹).

On the other hand, 7 (out of 8) RV Tau stars from our sample belong to group-III objects, with only one source (AI Sco) being a member of group-I on the CO-IR diagram. This is not so surprising since RV Tau variables are well-known to be CO-deficient stars (Alcolea & Bujarrabal 1991). RV Tau stars from our sample consist of the majority of the blue part of the group-III objects, with IR colors in a narrow range of −1.5 < C23 < −1.1. They have O-rich chemistry, 7 (out of 8) have SEDs classified as 0 or I (with AR Pup being the only exception with an SED of type-III), 7 of them are also known binaries (with the only exception being AI CMi, which does not have near-IR excess; see, e.g., SED in the Torun catalog), and the three RV Tau variables with detected CO 2–1 lines all show very narrow CO lines (with V_exp = 1.5 ∼ 5.6 km s⁻¹).

We summarize below the properties of the different groups of pAGB objects, as inferred from the above discussion. The key features of each CO-IR group are collected in Table 4 and illustrated in Figure 7.

Zoom In Zoom Out Reset image size

6.5.1. Group-I

6.5.2. Group-II

6.5.3. Group-III

6.5.4. Transition Region

Although the majority of the O-rich AGB stars (empty blue circles in the second row of Figure 3) are concentrated in a small region on both the CO-IR and IRAS C–C diagrams, some objects show significantly smaller R_CO25 ratios (CO-weak) or significantly redder C23 colors. The redder objects could be either post-thermal pulse objects with detached CSEs (see, e.g., Steffen et al. 1998), OH/IR star candidates, or mis-identified pAGB stars. Many of the CO-weak objects are semi-regular variables of which the CSEs are probably optically thinner and thus CO molecules could have been partially destroyed by photodissociation.

The 28 OH/IR stars (half-shaded black circles) scatter in a larger region with similar or smaller CO-IR flux ratios but redder C12 and C23 colors than the other O-rich AGB stars. This can be explained by cooler CO gas and lower average dust temperatures in the very thick CSEs of the OH/IR stars (Heske et al. 1990; Kastner 1992).

The O-rich AGB stars have slightly bluer C23 colors and about three times smaller R_CO25 ratios than the C stars on average. The redder C23 color of C stars was previously determined (e.g., in the study of IRAS C–C diagram by van der Veen & Habing 1988) to be due to the shallower emissivity of carbon-rich dust in the 12–100 μm region (Zuckerman & Dyck 1986b). The stronger CO line in C stars have been recognized since the CO line survey of Nyman et al. (1992). The most possible explanation is higher CO abundance in carbon stars than in O-rich AGB stars, because only part of the oxygen atoms are used to make CO molecules in the latter.

S stars on the R_CO25-C23 diagram (red filled circles in the top row of Figure 3) scatter in the similar regions as the C stars, indicating that the gas and dust properties of the CSE of S stars are closer to those of C stars than O-rich AGB stars. However, because S stars are expected to have much less dust in their CSEs due to the lock of most C and O atoms into gaseous CO molecules, their CO-IR flux ratios are intuitively expected to be even larger than that of typical C stars, which is not true in our Figure 3. The lower than expected CO-IR ratios of S stars could be explained by assuming that the CO molecules are not efficiently formed through ideal equilibrium chemistry (as discussed for C stars by Papoular 2008) and by assuming dust grains such as the recently proposed solid SiO dust (Wetzel et al. 2013) may still be efficiently formed around S stars.

Comparing the IRAS C–C diagrams in the upper right and middle right panels of Figure 3, one can see that the S stars also have different IRAS colors than the C- and O-rich AGB stars. Their C12 colors are generally bluer than the latter, while their C23 colors are similar to C stars but redder than O-rich AGB stars. Again, the solid SiO dust proposed by Wetzel et al. (2013) has the potential to naturally explain these color differences. The solid SiO grains have the 10 μm feature as normal silicates, but not the 18 μm feature. Compared with the silicate dust in O-rich AGB stars, the lack of the 18 μm feature of SiO dust in S stars results in a weaker IRAS 25 μm flux density, and thus bluer C12 and redder C23 colors of S stars than O-rich AGB stars, as shown above. Compared with the C-rich dust in C stars, the most salient feature of the SiO dust is the appearance of the emission feature at 10 μm which naturally causes the enhancement of the IRAS 12 μm flux density, and thus bluer C12 and comparable C23 colors of the S stars against C stars.