NEAR-INFRARED SPECTROSCOPY OF INFRARED-EXCESS STELLAR OBJECTS IN THE YOUNG SUPERNOVA REMNANT G54.1+0.3

NIR spectroscopic observations were performed for 7 of the 11 IR-excess stellar objects discovered by Koo et al. (2008) in the SNR G54.1+0.3 with the TripleSpec spectrograph on the 5 m Palomar Hale telescope on 2008 August 9. TripleSpec is a slit-based NIR cross-dispersion echelle spectrograph covering the entire NIR atmospheric window simultaneously with a spectral resolving power R of 2500–3000 using six (from 3 to 8) echelle orders mapped onto two adjacent quadrants of a Hawaii II HgCdTe 2 K × 2 K detector array from Teledyne, Inc. (Wilson et al. 2004; Herter et al. 2008). The slit width and length of TripleSpec are 1'' and 30'', respectively, and the broad spectral coverage of the spectrograph makes it ideal for conducting observations of new objects with little known information such as the IR-excess stellar objects in G54.1+0.3.

Figure 1 shows the positions of the IR-excess stellar objects in the AKARI 15 μm image numbered in decreasing order of their Spitzer MIPS 24 μm brightness (Koo et al. 2008). Table 1 lists the coordinates and magnitudes in the visible to MIR wave bands apart from those for Object 2, which has been detected only at λ ≳ 3.6 μm (Monet et al. 2003; Zacharias et al. 2004; Koo et al. 2008). The photometric data used in Koo et al. (2008) are based on the Spitzer Infrared Array Camera (IRAC) Galactic Legacy Infrared Mid-Plane Survey Extraordinaire (GLIMPSE)⁵ catalog, which somehow does not contain the Two Micron All Sky Survey (2MASS) NIR magnitudes of Object 6; these are instead obtained from the 2MASS All-Sky Point Source Catalog⁶ (PSC; Skrutskie et al. 2006) and added in Table 1.

Zoom In Zoom Out Reset image size

The seven objects (i.e., 1, 4, 5, 7, 8, 10, and 11) marked in red in Figure 1 were observed with TripleSpec in this work. However, the low signal-to-noise ratio (S/N) of Object 4, the faintest one in the K_s band, makes the reliable use of its spectrum difficult, and consequently, it has been excluded from further analysis. All the objects were observed with the typical AB nodding pattern, separated by 10'' along the slit direction, to effectively subtract the sky background emission. The total exposure time of each object was 600 s. The standard star HD 171623 (A0V) was observed at similar airmasses to those of the objects for standard star calibration.

We first conduct flat fielding and bad pixel correction before obtaining two-dimensional wavelength solutions of the dispersed images using the OH sky emission lines. For each frame of a dithered pair, we subtract the sky background of a given frame using the other frame and then combine the pair of sky-subtracted spectra from the two frames to produce the final spectrum. Although TripleSpec covers wavelengths from 0.78 to 2.4 μm using six echelle orders of 3–8, we only use orders 3–6 (i.e., ∼1.0–2.4 μm) due to the low S/Ns of the spectra obtained in orders 7 and 8. The S/Ns of the final spectra range between 5 and 30. The extracted spectra show the features of the telluric absorption lines convolved with the TripleSpec system responsivity (Figures 2(a) and (d)).

Zoom In Zoom Out Reset image size

For the standard source calibration, we are unable to find an appropriate standard spectrum of our calibration source HD 171623 (A0V) that can be used reliably to calibrate our TripleSpec spectra. Instead, we use the scaled model spectrum of Vega, which has the same spectral type and luminosity class as HD 171613, generated by the ATLAS12 stellar models (Castelli & Kurucz 1994) as the standard spectrum of our calibration source. The widths of the model H absorption lines of the Vega spectrum are larger than those of the observed lines of HD 171623; therefore, we replace those lines, including Brγ, Br11, Br12, Paβ, and Paδ, with a Voigt profile in a scaled model spectrum if the lines do not blend or overlap. In the case where the H and telluric lines are adjacent (e.g., Brγ), the telluric lines are modeled using a Gaussian profile and subtracted.

Figure 2 presents an example of this calibration procedure for the order-3 (2.03–2.40 μm) spectrum of Object 1. The observed spectrum of the standard source HD 171623 in Figure 2(a) exhibits many telluric absorption lines and broad CO₂ absorption bands near 2.06 μm. This is divided by the model spectrum (Figure 2(b)) to produce the spectrum in Figure 2(c) featuring the telluric lines and system responsivity. The final calibrated spectrum (Figure 2(e)) is obtained by dividing the observed spectrum (Figure 2(d)) by the spectrum in Figure 2(c).

Figures 3 and 4 present the obtained spectra of the IR-excess stellar objects after normalization by a third- or fourth-order polynomial in wavelength ranges of 0.99–1.11, 1.22–1.31, 1.51–1.72, and 2.01–2.25 μm, corresponding to orders 6, 5, 4, and 3, respectively. Since higher orders have low throughputs for grating-based spectroscopy, the order-6 spectra have lower S/Ns than the others. Also note that the order-3 spectra are heavily contaminated by numerous telluric lines. For all the spectra, photospheric absorption lines are dominant; in particular, the H Brackett series are pronounced, including six continuous transitions of the Br11–16 series. In increasing order of the wavelength, we identify the following absorption lines: He i 1.003, Paδ 1.005, He i 1.083, He i 1.092, and Paγ 1.094 μm (order 6; Figure 3, left); He i 1.279, Paβ 1.282, and He i 1.297 μm (order 5; Figure 3, right); Br16 1.556, Br15 1.571, Br14 1.589, Br13 1.611, Br12 1.641, Br11 1.681, and He i 1.701 μm (order 4; Figure 4, left); and He i 2.113 and Brγ 2.166 μm (order 3; Figure 4, right).

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

We measure the equivalent widths (EWs) of the identified lines with a high S/N by fitting a Gaussian profile, and these results are presented in Table 2. Note that the EWs of the He i 1.297 μm line of Object 11, Br11 of Object 7, and Brγ of Object 5 are large because of contamination by nearby telluric features. In the obtained spectra shown in Figures 3 and 4, we have not identified any obvious emission lines. Some emission-like features in the spectra (e.g., those around ∼1.27 and ∼1.57 μm) are residuals of strong telluric lines that have not been completely removed during the standard star calibration.

The obtained spectra (Figures 3 and 4) of the IR-excess stellar objects show absorption lines of strong H transitions and relatively weak He i transitions. This, together with the lack of He ii lines, indicates that these are stars of a late O- and early B-type between O8 and B3 (Hanson et al. 1996, 1998, 2005). This is consistent with the previous interpretation based on the JHK_s color–color diagram and SEDs of these objects (Koo et al. 2008). Also, the absence of the He i 2.059 μm and Brγ emission lines suggests that they are MS stars rather than supergiants (Hanson et al. 1996).

To determine their spectral types more precisely, we first conduct a thorough literature search of reliable stellar spectroscopic libraries. However, as far as we are aware, only limited studies have been performed for NIR spectroscopic studies of OB stars so far, and even those available studies are somewhat inadequate for spectral classification use due to their small sample sizes (e.g., Wallace & Hinkle 1997; Meyer et al. 1998; Wallace et al. 2000) or paucity of the identified lines useful for line ratio comparison (e.g., Hanson et al. 1996, 1998, 2005; Bik et al. 2005).

The photospheric H lines of OB stars, on the other hand, are known to be very sensitive to temperature (Lenorzer et al. 2004), and, consequently, the EWs of the H lines are expected to be correlated with the stellar temperatures. Figure 5 confirms a correlation between the two parameters in the case of Br11, which exhibits a tight correlation (with a correlation coefficient of −0.94) of

$$\begin{equation} {\rm EW}_{{\rm Br}11}\,({\rm \mathring{\rm{A}}}) = -0.23(\pm 0.02) \times (T\, {\rm (kK)})+9.29(\pm 0.46). \end{equation} \tag{ 1 }$$

Here, we use the results of Martins et al. (2005) and Schmidt-Kaler (1982) for the temperatures of OB stars and those of Hanson et al. (1998) for the line widths. The results of Meyer et al. (1998), which are based on much smaller sample sizes, also agree well with this relation. Note that among all the H lines in Table 2 Br11 is an ideal choice for this type of analysis because it is relatively well isolated and thereby free of contamination from nearby sky/telluric lines, although other photospheric H lines may be used for the same purpose as long as reliable measurements of their EWs are made. For instance, Figure 6 compares the temperatures of OB stars and the EWs of their Brγ lines available from Hanson et al. (1996), where the two parameters are still well correlated as

$$\begin{equation} {\rm EW}_{{\rm Br}\gamma }\,({\rm \mathring{\rm{A}}}) = -0.25(\pm 0.01) \times (T\, {\rm (kK)})+11.84(\pm 0.33). \end{equation} \tag{ 2 }$$

The increased scatter in Figure 6 between the two parameters compared with that in Figure 5 is mainly caused by larger errors in the measured EWs of the Brγ lines, which are near telluric lines. The temperatures of the IR-excess stellar objects obtained using Equation (1) are presented in Table 3 along with their spectral types ranging between O9 and B2.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Table 3. Stellar Parameters Derived from the Relation between the Equivalent Widths of Br11 and the Stellar Temperatures

Object	Temperaturea	Spectral	Distance	AV
	(K)	Type	(kpc)	(mag)
1	26000 ± 2000	B1 (B0.5–B1.5)	$4.6^{+0.4}_{-0.3}$	$7.4^{+\lesssim 0.05}_{-0.3}$
5	33000 ± 3000	O9 (O7.5–B0)	$8.1^{+1.9}_{-1.3}$	$7.5^{+\lesssim 0.05}_{-0.1}$
7	≳ 17000	≲ B3.5	≳ 3.3	≳ 7.8
8	23000 ± 2000	B1.5 (B1–B2)	$5.4^{+0.5}_{-0.7}$	$6.9^{+0.1}_{-\lesssim 0.05}$
10	22000 ± 3000	B2 (B1–B3)	$5.5^{+0.9}_{-1.2}$	$7.4^{+0.1}_{-0.2}$
11	23000 ± 4000	B1.5 (B1–B2.5)	$6.5^{+1.2}_{-1.7}$	$7.1^{+0.2}_{-0.2}$

Note. ^aThe effective temperatures of the MS stars are adopted from Martins et al. (2005) for O-type stars and those of Schmidt-Kaler (1982) for stars later than B0.

Download table as:  ASCIITypeset image

The uncertainties in the estimated temperatures of the IR-excess stellar objects in Table 3 (Column 2) represent those from the measurement errors of the EWs of the Br11 lines. The scatter in the correlation between the EWs and the stellar temperatures in Equation (1) is an additional source of uncertainty of ∼2000 K. The combined uncertainties of these two are roughly equivalent to one subclass in the determined spectral types (Column 3) of the IR-excess stellar objects. Furthermore, there is an intrinsic difference of approximately one spectral subclass around the late O and early B spectral types between the two stellar models of Martins et al. (2005) and Schmidt-Kaler (1982). We therefore believe that the spectral types of the IR-excess stellar objects in Table 3 have uncertainties equal to a total of two subclasses.

It is notable that the measured EWs (∼1 Å) of the He i lines (Table 2) are also consistent with the interpretation that the spectral types of the IR-excess stellar objects lie between late O- and early B-type stars (see Hanson et al. 1996, 1998 for details).

We estimate the photometric distance to the six IR-excess stellar objects based on the determined temperatures by fitting their SEDs with the ATLAS9 Kurucz stellar models (Castelli & Kurucz 2003). For this, we use wavelengths shorter than 5 μm in the SED fits, to concentrate on stellar emission, and the interstellar reddening law of R_V = 3.1 (Draine 2003). The best-fit distances and extinctions lie within 3.3–8.1 kpc and A_V = 6.9–7.8 mag, respectively, with mean values of 6.0 ± 0.4 kpc and 7.3 ± 0.1 mag (Table 3).

Figure 7 shows the distribution of the obtained distances and extinctions. The two parameters of Object 7 only have lower limits because the lack of a B-band magnitude makes it difficult to determine the upper limits—note that the B-band magnitude is critical in determining the temperature of OB-type stars in the SED fits. The large extinction (A_V ≳ 7.8 mag) together with the absence of the B-band magnitude indicates that Object 7 has a locally increased extinction. We discuss the extinction of the IR-excess stellar objects in Section 4.

Zoom In Zoom Out Reset image size

The distance to G54.1+0.3 was previously determined in several studies to be 5–10 kpc based on 21 cm H i absorption lines (Koo et al. 2008; Leahy et al. 2008) and $9^{+1.0}_{-1.5}$ kpc from a dispersion measure analysis of the central pulsar J1930+1852 (Camilo et al. 2002; Cordes & Lazio 2002). Two additional distances of 8 kpc and 6.2$^{+1.0}_{-0.6}$ kpc have been proposed from ¹³CO J = 1–0 observations of nearby molecular clouds (Koo et al. 2008; Leahy et al. 2008), although it is uncertain whether the clouds are indeed associated with the SNR (Lee et al. 2012). The 6.0 ± 0.4 kpc distance to the IR-excess stellar objects derived from the SED fits in this study is somewhat smaller but still consistent with previously reported values. Since our result is independent of the Galactic rotation model, we adopt 6 kpc as the distance to both the IR-excess stellar objects and G54.1+0.3 in the following discussion.

In the first scenario, the IR-excess stellar objects are interpreted to be massive YSOs whose formation was triggered by the progenitor of the SNR G54.1+0.3 (Koo et al. 2008). As Koo et al. (2008) have previously shown, the SEDs of the IR-excess stellar objects have a dip at around 8 μm with a strong MIR excess, similar to those of stars without substantial circumstellar material such as YSOs with a truncated disk or Herbig Ae/Be stars with a breakup inner disk (Hillenbrand et al. 1992; Malfait et al. 1998; D'Alessio et al. 2005). However, the NIR spectra (Figures 3 and 4) of the IR-excess stellar objects do not show any emission lines that are often found in Herbig Ae/Be stars (Hillenbrand et al. 1992; Malfait et al. 1998). This indicates that the IR-excess stellar objects are in a relatively later evolutionary stage of pre-MS stars with only little nearby circumstellar material remaining. In general, the disk lifetime of YSOs is well correlated with the mass of the central star, and for stars more massive than 7 M_☉, it is known that disk dispersion takes less than 1 Myr (Hillenbrand et al. 1992; Alonso-Albi et al. 2009). Hence, the IR-excess stellar objects are likely to be a few million years old, and their formation may have been triggered during the late MS or post-MS evolutionary stages of the progenitor. In the case of the latter, a potential delay in the triggered formation process due to local environment effects has been proposed (Koo et al. 2008).

The IR loop in G54.1+0.3, where the IR-excess stellar objects are embedded, was suggested to be composed of fast-expanding SN ejecta dust (Temim et al. 2010), although this does not preclude the existence of the local ISM and/or circumstellar material in the loop. If the ejecta dust is illuminated by the IR-excess stellar objects, then significant enhancement of the MIR emission can occur, and depending on the distance between the two, the 8 μm dip features seen in the SEDs of the latter (Figure 8) may arise. If the IR emission of the stellar objects is from SN dust rather than circumstellar dust, then the IR-excess stellar objects would be older than a few million years, and subsequently, the triggered formation by the progenitor star of G54.1+0.3 could have happened earlier, e.g., during their MS stages.

On the other hand, we note that confirmation of the IR-excess stellar objects as a result of triggered star formation is somewhat uncertain based on the available information from the existing observations. As described above, the spatial confinement of the IR-excess stellar objects in the IR loop surrounding the PWN in G54.1+0.3 indicates potential triggered star formation by the SN progenitor star. However, recent numerical simulations of the feedback effects on star formation (Dale & Bonnell 2011; Dale et al. 2012) suggest that the features such as bubble walls or pillars, in which young stars are embedded, do not always trace triggered star formation, although they may serve as a good indicator. According to the results, such geometrical distributions of stars can be produced without feedback, depending on environments (Dale & Bonnell 2011; Dale et al. 2012).

According to the second scenario, an SN exploded as a member of a cluster of stars that includes the IR-excess stellar objects, and the stellar cluster illuminates the SN ejecta dust to produce the strong MIR emission that is observed (Temim et al. 2010). In this case, the results of the NIR spectroscopy reported in this paper help constrain the mass and type of the SN progenitor as follows. First, the mass (M ∼ 17 M_☉) of an O9 star, which is the earliest spectral type of the IR-excess stellar objects (Table 4), gives the lower limit of the progenitor mass, and considering the continuous distribution of the initial mass function (e.g., Salpeter 1955), the progenitor mass cannot be much greater than that of an O9 star. In other words, it is unlikely that the progenitor was massive enough to be a Wolf–Rayet star of ≳35 M_☉ (e.g., Smith et al. 2011 and the references therein). Consequently, we can constrain the progenitor mass to be in the range of 18–35 M_☉, although the upper limit may be significantly lower than 35 M_☉.

The SN type of G54.1+0.3 has not yet been definitively confirmed. Chevalier (2005) previously ruled out the IIL/b possibility due to a lack of any observational evidence indicating the existence of strong circumstellar interactions and suggested that G54.1+0.3 is an SN IIP or Ib/c. If the progenitor star was a single star, then we believe that it is less likely to be an SN Ib/c, given the upper mass limit, which leaves us only with the IIP possibility. The mass range of SN IIP progenitors was first calculated to be 9–25 M_☉ by Heger et al. (2003) based only on single star evolutionary models. Smith et al. (2011), however, recently suggested a smaller range of 8–18 M_☉ after observationally accounting for contributions from both single stars and binaries in the entire core-collapse populations. This range of Smith et al. (2011) for the progenitor mass of SNe IIP places the G54.1+0.3 progenitor mass close to the upper limit, similar to that of the O9 star, which is the earliest existing member of the cluster as noted. Therefore, if the progenitor of G54.1+0.3 was indeed a member of the cluster to which the IR-excess stellar objects also belong and exploded as an SN IIP, the spectral type of the progenitor should be that of a late O-type star (e.g., O8). Finally, we note that in principle we cannot yet completely rule out the possibility of the G54.1+0.3 progenitor being a binary that exploded as an SN Ib/c, although we believe that this possibility is low since SNe Ib/c from binaries are much less abundant than SNe IIP.