GALACTIC CEPHEIDS WITH SPITZER. II. SEARCH FOR EXTENDED INFRARED EMISSION

P. Barmby; M. Marengo; N. R. Evans; G. Bono; D. Huelsman; K. Y. L. Su; D. L. Welch; G. G. Fazio

doi:10.1088/0004-6256/141/2/42

1. INTRODUCTION

Classical Cepheid variable stars are of key importance in the extragalactic distance scale and in studies of Galactic stellar and chemical evolution. But the very property that makes them useful as distance indicators—pulsation—also makes Cepheids difficult to model accurately. For several decades, a major problem in the understanding of Cepheids has been the discrepancy between masses estimated from stellar evolution theory and those estimated from pulsation theory (Christy 1968; Stobie 1969; Fricke et al. 1972). While the problem was partly addressed by revision of radiative opacities and changes in evolutionary tracks (Moskalik et al. 1992), the discrepancy persists at the 10%–20% level (Caputo et al. 2005; Keller & Wood 2006). Measurements of several Cepheid masses are now available from binary systems (Benedict et al. 2007; Evans et al. 2008, 2009), but the comparison with evolutionary and pulsational masses has not yet produced a definitive result.

There are a number of possible solutions to the "Cepheid mass discrepancy," summarized by Bono et al. (2006). Recent studies have come to differing conclusions on which might be responsible: Caputo et al. (2005) suggested that mass loss can account for the discrepancy, while Keller (2008) concluded that additional mixing in Cepheid main-sequence progenitors was responsible. Mass loss remains one of the less-understood parameters in stellar evolutionary theory (Willson et al. 2008; Vink 2008), and its relevance to Cepheid masses is still not completely clear. In the last several years, several lines of investigation have suggested that mass loss may be important in Cepheids. These include the discovery of extended near-infrared emission from a number of Galactic Cepheids (Mérand et al. 2006, 2007; Kervella et al. 2006), possibly from circumstellar shells, as well as theoretical modeling of pulsation-driven mass loss (Neilson & Lester 2008) and its application to Magellanic Cloud Cepheids (Neilson et al. 2009, 2010). Both the presence of shells around nearby Cepheids and the evidence for infrared excess in LMC Cepheids have implications for the stars' derived diameters and therefore for the distance scale. Systematic effects on distances are possible if Cepheid mass loss is substantial enough to affect spectral energy distributions and if it works differently in nearby and distant galaxies (e.g., is a function of metallicity). Neilson et al. (2010) concluded that mass loss in Cepheids "has an important effect on the structure of the IR P–L [infrared period–luminosity] relations" although the extent of the effect is not well characterized.

Attempts to estimate Cepheid mass-loss rates through infrared observations began about 25 years ago. McAlary & Welch (1986) used infrared excesses inferred from the IRAS Point Source Catalog to estimate mass-loss rates of $\dot{M} <10^{-9}\,\hbox{ to }10^{-8}$ M_☉ yr⁻¹, while Deasy (1988) combined ultraviolet and infrared observations to estimate upper limits of $\dot{M}<10^{-10}\hbox{ to }10^{-7}$ M_☉ yr⁻¹. Welch & Duric (1988) used radio observations to place similar upper limits on the mass loss in ionized gas. In general, these rates are too low to account for the mass discrepancy. Recently, the Spitzer Space Telescope has been used in a number of studies of Cepheids in the mid-infrared, including the period–luminosity relation (Ngeow & Kanbur 2008; Freedman et al. 2008; Marengo et al. 2010a). Kervella et al. (2009) combined ground-based infrared imaging and interferometric measurements with Spitzer data (including data from the program described here) to detect compact circumstellar envelopes around both ℓ Car and RS Pup; the latter was previously known to be associated with a reflection nebula.

To investigate the possibility of infrared excesses due to mass loss, we have obtained Spitzer (Werner et al. 2004) observations of a sample of 29 classical Cepheids. Our results on the mid-infrared period–luminosity and period–color relations for Cepheids are presented in Marengo et al. (2010a, Paper I); a detailed analysis of extended emission near δ Cep is presented by Marengo et al. (2010b). In Paper I, we determined that there was no evidence from IRAC colors for warm circumstellar dust. The present paper investigates the presence of cooler dust around this sample of Cepheids via a search for extended mid-infrared emission. An important consideration is that extended emission may be the result of recent star-forming activity near a Cepheid rather than mass loss from the Cepheid itself. However, distinguishing those two different causes is not straightforward, so a search for extended emission places an upper limit on the amount of mass loss. Except where mentioned otherwise, in all of the following analysis we use the stellar distances given by Fouqué et al. (2007).

2. OBSERVATIONAL DATA

The sample selection and first-epoch observations are described in detail by Marengo et al. (2010a). Briefly, 29 Galactic Cepheid variables and three non-variable supergiants were observed with both IRAC (Fazio et al. 2004) and MIPS (Rieke et al. 2004) as part of Spitzer Space Telescope General Observer program ID 30666. The IRAC observations described by Marengo et al. (2010a) were primarily made in subarray mode. The six brightest stars were observed in full-array mode so that fluxes could be measured by fitting the outer part of the point-spread function (PSF); four stars were observed in both full-array and subarray modes. The MIPS observations used the Photometry Astronomical Observing Template with the shortest possible observation time (48.2 s) at 24 μm and varying times (37.7–1300.2 s) at 70 μm. Three stars (S Nor, V Cen, and U Sgr) were observed in two adjacent fields of view at 70 μm to better cover the surrounding clusters. While our target SZ Tau (see Section 3.5) may also be in the cluster NGC 1647 (Turner 1992), after inspecting its 24 μm image, we changed the planned 70 μm observation to use only one field of view centered on the star. This allowed us to increase the observation time and, as a bonus, also provided a second epoch of 24 μm observations. We performed simple aperture photometry on post-basic calibrated data (BCDs) mosaics primarily from the pipeline version S16.1 output. For the three cluster stars, the 70 μm data were remosaiced using MOPEX to combine the data for the two fields of view. For two stars (RS Pup and GH Lup), the 70 μm mosaics were remade to remove negative sidelobes.

2.1. New IRAC Data and Reduction

To follow up earlier results, we obtained some additional Spitzer data in Cycle 5, on the dates 2008 October 5 and 6. All stars were observed in full-array mode with IRAC as part of Guaranteed Time program 50350 (PI: G. Fazio). This increased the field of view for stars which had previously been observed only in subarray mode and also provided deeper observations (2 s frames at each of 36 dither positions) for all stars. Each target fainter than K = 2 was observed with the IRAC full-frame 12 s High Dynamic Range (HDR) mode, using a 12-point Reuleaux small-scale dither pattern for a total integration time of 144 s in each band. HDR 12 s frames were used to better reveal the area closer to the saturated cores of the targets, which also helped the PSF subtraction process. For the six brightest sources, in order to avoid excess saturation, targets were observed with 2 s frames using IRAC full-frame non-HDR mode and the 36-point Reuleaux small-scale dither pattern, for a total integration time of 72 s in each band. The Reuleaux dither pattern was used to achieve the required total exposure time while obtaining the best possible spatial sampling of the targets in anticipation of the PSF subtraction process.

The BCDs for all observations were reduced using the custom post-BCD software IRACproc (Schuster et al. 2006), which generated mosaics for each source with the final exposure depth and outliers removed. The final images had a pixel scale of 0.8627 arcsec pixel⁻¹ ( $1/\sqrt{2}$ the original IRAC pixel scale) which provides a better sampling of the sources' PSF for a more accurate PSF subtraction. The PSF subtraction was also performed using a routine part of IRACproc, which has been extensively tested in other Spitzer programs (Marengo et al. 2006, 2010a; Luhman et al. 2007) to search for faint structures and point sources around bright stars.

2.2. MIPS Photometry

Although the MIPS observations used in this paper are the same as those in Marengo et al. (2010a), a major purpose of the present work is to examine extended emission around a subset of the stars in our sample. Accordingly, we have performed point-source fitting and subtraction on the 24 μm images of some targets to separate the contributions of the star and extended emission to the total flux density. The data were reprocessed with the MIPS instrument team Data Analysis Tool using standard procedures. The PSF used for the 24 μm subtraction was built from the observations of the A4 V star, τ₃ Eri (Su et al. 2008), which matches well with the theoretical PSF computed from STinyTim Program (Krist 2006) after proper smoothing. The subtraction was done by iteratively shifting the PSF to match the centroid of the target in sub-pixel accuracy, adjusting its scaling, and subtracting it from the target image until the residual inside r < 12''(∼ to the first bright Airy ring of the PSF) was minimized. The residual flux is generally less than 1% of the total flux of a typical point source on a clean background.

To estimate the non-point-source emission near stars showing extended emission in the MIPS bands, we used the results of point-source fitting and subtraction. We performed aperture photometry of the point and extended emission on images which had the field (but not the target) stars subtracted, using the largest apertures which could be contained in the image. From this we subtracted the flux measured by fitting and subtracting the target stars to derive the total non-point-source emission. This procedure depends on accurate background subtraction for the large-aperture photometry, particularly difficult in the 70 μm images, and the resulting uncertainties are large. An alternative approach which can constrain the physical conditions within regions of extended emission is used by Marengo et al. (2010b): measuring the surface brightness of the extended emission in small areas. Because the stars are heavily saturated in the IRAC bands, this is the only method available to investigate the extended emission visible in the IRAC images. In both IRAC and MIPS images, we measured the surface brightness in several small regions near stars with extended emission and subtracted backgrounds measured in regions located off the extended emission. The regions used are shown in the images of the relevant stars; the measured surface brightnesses are reported in Table 1.

Table 1. Surface Photometry of Extended Infrared Emission

Star/Region	5.8 μm	8 μm	24 μm	70 μm
GH Lup 1	0.07 ± 0.01	0.27 ± 0.01	0.16 ± 0.02	1.1 ± 0.1
GH Lup 2	0.02 ± 0.01	0.36 ± 0.01	0.29 ± 0.02	1.8 ± 0.1
GH Lup 3	0.13 ± 0.01	0.49 ± 0.01	0.51 ± 0.02	3.5 ± 0.2
GH Lup 4	0.10 ± 0.01	0.41 ± 0.01	0.46 ± 0.02	2.6 ± 0.1
RS Pup 1	0.24 ± 0.03	0.63 ± 0.02	0.89 ± 0.39	9.4 ± 0.8
RS Pup 2	0.18 ± 0.01	0.56 ± 0.01	0.82 ± 0.30	24 ± 1
RS Pup 3	0.27 ± 0.01	0.83 ± 0.03	1.99 ± 0.86	40 ± 1
RS Pup 4	0.33 ± 0.01	0.91 ± 0.01	3.10 ± 0.72	47 ± 1
RS Pup 5	0.23 ± 0.01	0.84 ± 0.03	3.14 ± 2.08	35 ± 2
S Mus 1^a	0.05 ± 0.01	0.10 ± 0.01	1.3 ± 1.1	2.0 ± 0.2
S Mus 2	0.08 ± 0.01	0.09 ± 0.01	0.8 ± 0.6	1.9 ± 0.1
S Mus 3	0.15 ± 0.01	0.16 ± 0.01	0.7 ± 0.3	0.9 ± 0.1
S Mus 4	0.07 ± 0.01	0.11 ± 0.01

Notes. All surface brightness values in MJy sr⁻¹. ^aFor S Mus, regions are not the same for IRAC and MIPS.

Download table as: ASCII Typeset image

3. ANALYSIS

Infrared colors were used in Paper I to demonstrate a lack of warm circumstellar dust in our sample of Cepheids. However, dust mass lost from Cepheids and then transported to larger distances would be expected to be cooler and possibly visible as extended infrared emission. One example of this phenomenon is the infrared nebula visible around δ Cep (Marengo et al. 2010b): here we complete the survey of our Cepheid and comparison star sample. None of the stars showed any extended emission in the 3.6 μm or 4.5 μm bands, so those bands are not discussed further here. Distinguishing between extended emission associated with a specific star and that due to nearby Galactic cirrus and/or star formation, especially at low Galactic latitudes, is not always straightforward. We expect that emission associated with an individual star should be centered on or near the star, with its surface brightness decreasing with increasing distance from the star. Unrelated infrared emission will not in general have these characteristics, although of course it could hide or confuse real low level emission. For brevity, we refer to unrelated infrared emission as "cirrus" below, although we recognize that it may not be true Galactic cirrus emission.

Of our sample of 29 Cepheids, 16 showed clear signs of cirrus in the IRAC 8 μm images: AQ Pup, BF Oph, DT Cyg, FF Aql, GH Lup, RS Pup, S Mus, S Nor, SW Vel, T Mon, U Car, U Sgr, V636 Sco, V Cen, VY Car, and X Cyg. Cirrus is also visible in the MIPS images of GH Lup, V Cen, VY Car, and U Sgr.⁹ As expected, the presence of cirrus is related to Galactic latitude: of the Cepheids in our sample with |b| < 10°, 15 of 23 appear to be surrounded by cirrus emission. The remaining star with apparent unrelated emission is DT Cyg, at b = −10 fdg 8. The list of cirrus sources includes three of the four stars we believed to be cluster members (S Nor, V Cen, and U Sgr); this classification is confirmed by Gieren et al. (1997), who also found that VY Car is a member of the Car OB 2 association. The extended emission near these stars could be associated with either their clusters or the general interstellar medium (ISM). Figure 1 shows some examples of stars surrounded by cirrus.

**Figure 1.** Images of some Cepheids with nearby extended infrared emission unrelated to the star. Left to right: DT Cyg, FF Aql, V Cen, and VY Car. Top row: PSF-subtracted IRAC 8.0 μm images, Bottom row: non-subtracted MIPS 24 μm images. All fields of view are 4 arcmin squares. In these and all following images, north is oriented up and east left. The black and white bands in the IRAC images are artifacts from both the bright source and the PSF subtraction process.
Download figure:
Standard image High-resolution image

As a quantitative test for excess infrared emission near the stars, we computed the average surface brightness around each star and compared it to a background measurement. The IRAC 8 μm images were used for this measurement as they have a larger uniform-exposure field of view than the MIPS images, and much better sensitivity to extended emission than the 5.8 μm images. The average surface brightnesses were measured in 1 arcmin radius apertures surrounding each star, with the PSF subtraction residuals and bright star artifacts masked. The background measurement was made in an annulus with inner and outer radii of 74 and 98 arcsec. Point sources near the target stars, imperfect PSF subtraction, and unrelated cirrus emission all complicate the comparison between low surface brightness emission near the stars and in their vicinities. To be conservative, we consider as having possible extended infrared emission only those stars for which the average surface brightness in the measurement aperture exceeds that in the background aperture by 20 times the quadrature sum of the surface brightness standard deviations. Since our measurement errors are influenced by the factors listed above and are unlikely to be Gaussian, we do not consider these detections to have true 20σ significance—the value chosen is intended as a starting point to guide the analysis.

We found no evidence for extended infrared emission around two stars which have been previously found to have circumstellar envelopes: Polaris (Mérand et al. 2006) and Y Oph (Mérand et al. 2007).¹⁰ Other stars with neither cirrus nor extended infrared emission are BB Sgr, U Aql, V350 Sgr, W Sgr, β Dor, η Aql, and ζ Gem. Five stars showed both cirrus and extended emission centered on or near the star: GH Lup, RS Pup, S Mus, T Mon, and X Cyg. Three further stars had little detectable cirrus but did have extended infrared emission centered on the star: δ Cep, ℓ Car, and SZ Tau.

The extended emission around δ Cep is discussed in detail by Marengo et al. (2010b) and briefly summarized here. The emission consists of a large-scale (0.1 pc) arched structure visible at 70 μm, with diffuse, filamentary emission at 5.8, 8.0, and 24 μm contained within the arch. The brightest region of this diffuse emission is between δ Cep and its hot companion HD 213307. The symmetry axis of the 70 μm emission near δ Cep is aligned with the star's velocity vector relative to the local ISM, suggesting a bow shock at least partly swept up from the ISM. The emission near the stars is hypothesized to be due to the interaction of their stellar winds. The mass-loss rates required for the creation of these structures are estimated to be in the range 5 × 10⁻⁹to6 × 10⁻⁸ M_☉ yr⁻¹.

The remaining stars with extended infrared emission are discussed individually below.

3.1. GH Lup

Images of GH Lup are shown in Figure 2. The most obvious component of extended emission is an arc-like structure to the east of the star, visible at 8, 24, and 70 μm. The structure is not centered on the star itself but roughly 20'' to the south (2.2 × 10⁴ AU at the star's distance) with a radius of approximately 32'' (3.6 × 10⁴ AU). Spatially coincident with the arc is a point source visible at 5.8 and 8.0 μm only, which we can identify with the K_s = 9.5 object 2MASS J15244147-5251282. Because the region is crowded, with copious diffuse emission, it is unclear whether the arc near GH Lup is related to the Cepheid, the nearby star, or neither. The stellar crowding also makes defining regions for surface photometry of the diffuse emission difficult; we chose the regions marked in Figure 2 to avoid stars in the 8 μm image and capture at least some of the extended emission. The resulting measurements are given in Table 1, with the colors of the extended emission plotted in Figure 3. Using a modified blackbody spectrum to assign color temperatures from the S₂₄/S₇₀ ratios yields 55 K < T < 81 K, depending on the emissivity index β for the modified blackbody. The S₈/S₂₄ ratios for GH Lup tend to be higher than those found for the other stars with extended emission. Together with the structure of the emission and the extensive nearby cirrus, this leads us to speculate that the emission near GH Lup is only moderately likely to be associated with the star itself (see Section 4).

**Figure 3.** Surface brightness ratios of extended emission near GH Lup, RS Pup, and S Mus, with emission near δ Cep (circles) and NGC 7023 (star) shown for comparison. Increasing values on the horizontal axis correspond roughly to increasing dust temperature, while increasing values on the vertical axis correspond to increasing PAH content.
Download figure:
Standard image High-resolution image

3.2. ℓ Car

Images of ℓ Car are shown in Figure 4. These data were also analyzed by Kervella et al. (2009), who failed to detect any nebulosity in the Spitzer images, but found evidence of small-scale (∼1'') extended emission in near- and mid-IR interferometric data, attributed to hot circumstellar dust. Our PSF-subtracted IRAC images show low surface brightness emission, at 5.8 and 8.0 μm, surrounding the star with a radius of ∼35'', or 1.7 × 10⁴ AU at a distance of 498 pc (Kervella et al. 2009). We do not detect any significant extended emission at 70 μm, and only a tentative excess at 24 μm, northeast of the nearly saturated core of the PSF-subtracted star. The 8.0 μm emission appears brighter northwest of the star, where an arc-like structure is detected at a distance of ∼30'' (1.5 × 10⁴ AU). Following a suggestion by the referee, we computed the components of the star's Galactic peculiar space motion and their projection onto the sky, using the procedure described by Marengo et al. (2010b). The star's position and proper motion were taken from the Hipparcos catalog (Perryman et al. 1997), the distance from Kervella et al. (2009), and heliocentric radial velocity of 3.6 km s⁻¹ from Nardetto et al. (2009). We find a space velocity of 24.9 km s⁻¹ along a position angle of 315° (east of north), with the direction shown as an arrow in the second panel of Figure 4. The vector is roughly aligned along the direction from the star to the arc-like structure, suggesting a possible connection between the two (e.g., a bow shock, reflection nebula, or light echo). Image artifacts from the PSF subtraction and the star make the connection difficult to secure, however. Artifacts also make it impossible to accurately measure the surface brightness of the extended emission and thus estimate the mass-loss rates that could be responsible for the observed nebulosity. Higher spatial resolution imaging covering a wider area would clearly be useful in further study of ℓ Car.

3.3. RS Pup

RS Pup is embedded in a well-known reflection nebula (Westerlund 1961; Havlen 1972) whose infrared emission has been noted by numerous authors (e.g., Gehrz 1972; McAlary & Welch 1986). PSF-subtracted images of RS Pup are shown in Figure 5. The morphology of the extended infrared emission varies with wavelength but the brightest region is to the northeast of the star. Most of the emission is within about 80'' of the star, a linear distance of 0.77 pc at the 1.992 kpc distance of RS Pup (Kervella et al. 2008), although the small field of view of the 70 μm image makes this difficult to pin down. The MIPS data, but not the full-frame IRAC data, were also analyzed by Kervella et al. (2009), so here we present the first image of the resolved 8.0 μm emission around RS Pup. Kervella et al. (2009) attributed the ≳20 μm resolved emission, as well as the infrared excess in the IRAS data, to an extended envelope with a radius of roughly 1 pc and a temperature of ∼40 K. Kervella et al. (2009) also detected resolved emission at much smaller angular distances to the star and attributed this to a "warm and compact component" at a distance of 300–3000 AU and a temperature of a few hundred K. They proposed that the warm component comes from RS Pup's mass loss, and the cold component is ISM gas compressed by the star's stellar wind. In Marengo et al. (2010b), a similar origin for the extended infrared emission around δ Cep was proposed.

We add to the analysis of RS Pup's extended emission by examining the colors of the extended emission. Using the surface brightness measurements in Table 1, Figure 3 plots the ratios of the extended emission around RS Pup and S Mus, with the values derived for δ Cep by Marengo et al. (2010b) for comparison. Using a modified blackbody spectrum to assign color temperatures from the S₂₄/S₇₀ ratios yields 50 K < T < 80 K, depending on the emissivity index β for the modified blackbody. These temperatures are a little lower than the values derived for δ Cep; however, compared to δ Cep, the RS Pup nebulosity is more distant from the central star. Like the δ Cep nebulosity, the RS Pup extended emission has S_8.0/S₂₄ ∼ 0.5, suggesting that some polycyclic aromatic hydrocarbons (PAHs) must be present (pure dust at 50 K < T < 80 K would imply a much lower ratio; Marengo et al. 2010b). We can also compute an average dust temperature for the RS Pup nebulosity using the total flux densities of the extended emission. At 70 μm, we used the largest aperture that will fit on the image (85'' radius) to derive a total flux density of 9.7 Jy. At 24 μm, we measured the flux density in a large aperture (72'' radius) and subtracted the PSF-fitting flux of the central source to derive an extended emission flux density of (620–330=) 290 mJy. Using a modified blackbody spectrum to assign color temperatures from the ratio of the two bands yields 44 K < T < 57 K, depending on β. These temperatures are a little higher than the estimate of 40 K given by Kervella et al. (2009), who extrapolated the 70 μm surface brightness profile to derive a larger total flux density.

The dust mass in the RS Pup nebula is best estimated with far-infrared photometry since, as Figure 5 shows, the stellar emission is insignificant at wavelengths of 70 μm and beyond. Because our 70 μm MIPS images have a rather small field of view, the difficulty of accurate background subtraction makes any attempt to estimate the total flux density (including that of Kervella et al. 2009) problematic. However, a rough estimate of total dust mass is possible and provides a comparison point for previous results. We use the formula given in Equation (2) of Evans et al. (2003), assuming that all of the 70 μm emission is from optically thin dust with T = 40–80 K and emissivity κ_ν = 3Q_ν/ρa in cm² g⁻¹. For comparison with McAlary & Welch (1986), we use κ = 19.0, scaled from the 100 μm value of the a = 0.1 μm "dirty silicate" dust model in Jones & Merrill (1976) via a Q_ν ∝ λ⁻² wavelength dependence. The derived total dust masses for RS Pup are in the range 1 × 10⁻³to1.4 × 10⁻² M_☉. (The range of values here accounts for the change in the Planck function, but not any temperature dependence of the dust emissivity, over the given temperature range.) The uncertainty due to temperature is compounded by the uncertainty in dust emissivity: for example, using the wavelength-dependent emissivity of amorphous fayalite given by Evans et al. (2003) would result in a value of κ nearly 15 times larger (and thus masses 15 times smaller) than given above. The 100 μm flux density from the IRAS Small Scale Structure Catalog (Helou & Walker 1988) should provide a better estimate of the total flux density of the RS Pup nebula. Using this measurement, McAlary & Welch (1986) estimated a total dust mass of 0.028 M_☉ (for their adopted distance of 1.778 kpc) or 0.035 M_☉ for the 1.992 kpc distance we use. The mass derived from the 70 μm flux density is the same order of magnitude, likely smaller because our observations do not capture the full extent of the RS Pup nebula.

As McAlary & Welch (1986) point out, the large size of the RS Pup nebula makes estimating a mass-loss rate from the nebular mass "extremely difficult." With a gas-to-dust ratio of 100, the McAlary & Welch (1986) dust mass converts to a total mass of 3 M_☉. An argument in favor of the idea that some of the nebula is swept-up ISM results from a simple comparison of the nebula mass with the helium-burning timescale of RS Pup (Kervella et al. 2009). Taking the mass estimate for RS Pup provided by Caputo et al. (2005), the helium-burning time inside the instability strip is of the order of 1.2 Myr. If all of the nebula consists of mass lost by RS Pup during its helium-burning phase, the implied mass-loss rate would be ≳10⁻⁶ M_☉ yr⁻¹. This would imply a total mass loss at least of the order of 10%. As Kervella et al. (2009) note, this would be sufficient to solve the pulsational/evolutionary mass difference for RS Pup. However, as noted by Neilson & Lester (2008), the observational mass-loss rate for RS Pup disagrees with their theoretical predictions ((1.6–6.5) × 10⁻⁹ M_☉ yr⁻¹) by several orders of magnitude. These predictions do appear to fit the observational data for Magellanic Cloud Cepheids (Neilson et al. 2009, 2010), which makes it more believable that at least some of the mass in the RS Pup nebula did not originally belong to the star. The observed S_8.0/S₂₄ ratio is too high for pure dust emission, but lower than typical for the typical PAH content of the ISM. Since Cepheid atmospheres are oxygen-rich and not expected to form PAHs, this suggests that both swept-up ISM (or pre-existing material containing PAHs) and stellar wind must comprise the nebula.

3.4. S Mus

Images of S Mus are shown in Figure 6. There are slightly extended emission features to both the north and south of the star visible at both 8 and 24 μm, with radial extents 35''–45'' ((2.9–3.7) × 10⁴ AU at a distance of 820 pc). There is also extended emission visible at 70 μm, to a radius of about 55'' (4.5 × 10⁴ AU), although the morphology is different at that wavelength. The surface brightness of this emission is faint compared to that around RS Pup, and the orientation of the PSF subtraction residuals makes computing S_8.0/S₂₄ and S₂₄/S₇₀ at the same location difficult. The S Mus point shown in Figure 3 is for the box marked "3" in the IRAC images and "2" in the MIPS images. Although uncertain, the colors imply an even lower PAH content that in RS Pup and δ Cep, and a somewhat higher color temperature (70–120 K). This star has a high-temperature companion (Evans et al. 2006) and the extended emission could be associated with that star rather than S Mus itself, or the companion could be responsible for heating the circumstellar material (as hypothesized for δ Cep; Marengo et al. 2010b).

The photometry of the extended emission around S Mus can be used to estimate the mass lost by the star. Large-aperture (35'' radius) photometry of the 70 μm emission yields a total flux density of 96 ± 12 mJy; subtracting the point-source flux density (measured in a 16'' radius aperture as for Paper I) of 39 ± 11 mJy yields a nebular flux density of 57 ± 16 mJy. We use the same dust mass estimation procedure as for RS Pup, a temperature of 70–120 K, and a stellar distance of 0.88 kpc to derive an implied dust mass (4–17) × 10⁻⁷ M_☉. With a gas-to-dust ratio of 100, the total implied mass would be approximately (4–17) × 10⁻⁵ M_☉. If we instead use κ = 56 as in Marengo et al. (2010b), the total mass agrees quite well with that derived for δ Cep. Given the photometric uncertainties, the agreement may well be coincidental; however, the order of magnitude comparison is intriguing. A mass-loss timescale can be estimated by dividing the distance to the emission (35'' is 0.15 pc at the distance of S Mus) by the escape velocity (100 km s⁻¹) to yield t ≈ 1500 yr, so the resulting mass-loss rate would be (3–10) × 10⁻⁸ M_☉ yr⁻¹. As for δ Cep, this estimate is consistent with the predictions of Deasy (1988) but well above those of Neilson & Lester (2008). There are two reasons why our value for the mass-loss rate is likely to be an upper limit and therefore larger than the Neilson & Lester (2008) predictions. (1) We assumed that all of the dust responsible for the extended emission comes from wind material as opposed to having been swept up from the ISM; as for RS Pup, the PAH content implies that ISM sweeping is likely relevant. (2) Using the escape velocity to estimate the nebula's age results in the minimum possible age, and hence a maximum for the mass-loss rate.

We can estimate the total mass lost for S Mus through its mass and crossing time for the instability strip, as above for RS Pup. We estimated pulsational and evolutionary masses through the relation provided by Caputo et al. (2005, Tables 2–5): using the near-infrared photometry from Laney & Stobie (1994, transformed to the 2MASS system using Koen et al. 2007) and optical photometry from Pedicelli et al. (2010), corrected for extinction using the reddening given by Fernie et al. (1995) and the reddening law of Cardelli et al. (1989). We used an intrinsic distance based on the near-infrared photometry and period–Wesenheit relations provided by Persson et al. (2004). This procedure results in a mean mass 7.0 ± 1.5 M_☉, and according to Bono et al. (2000) the corresponding crossing time of the instability strip is of the order of 0.14 Myr. Given the mass-loss rate above, this results in a small total mass loss of 0.006 M_☉. Using the recent estimate of the radius of S Mus by Groenewegen (2008) and the period–mass–radius relation given by Bono et al. (2001), the pulsation mass is lower, ∼5 M_☉, and the instability strip crossing time much longer, of order 1 Myr; however, the total mass lost would still be only 0.04 M_☉, too small to significantly affect the mass discrepancy.

3.5. SZ Tau

SZ Tau was the only Cepheid found in Paper I to have a statistically significant [24] − [70] color excess. Images of this star are shown in Figure 7. There is an extended object about 25'' × 8'' clearly visible in the PSF-subtracted 8.0 μm image, about 7'' to the northeast of the star. At the 512 pc distance to the star, 25'' × 8'' corresponds to linear scales of (1.3 × 0.4) × 10⁴ AU. The object is not clearly visible in the 5.8 μm band, due to PSF subtraction residuals; however it is visible at 24 μm and appears to be the source of most of the 70 μm luminosity. The object is resolved at 24 μm and has a similar appearance in two images in that band taken about a year apart, suggesting that it is not an instrumental artifact or a solar system object. Noriega-Crespo et al. (1997) detected a "ridge or filament" of similar shape but much larger angular size in IRAS imaging of Betelgeuse; they suggested that this structure is emission from the ISM, unrelated to the star. Turner (1992) concluded that SZ Tau is a first-overtone pulsator and noted several other unusual characteristics of this star (e.g., period changes), but it is unclear how these features could be related to the presence of extended emission.

The elongated shape of the emission near SZ Tau suggests a background galaxy. The infrared colors should yield further information on the object's nature, although photometry of the SZ Tau extended emission is complicated by the star itself. At 70 μm, the extended emission appears to be the dominant source of emission. When we measure flux density in a 35'' radius aperture centered on the 70 μm peak and a 39''–65'' background annulus, the resulting flux density (which may include some contribution from the star) is 74 mJy.¹¹ We measured the flux density of the SZ Tau extended emission at 24 μm by doing a large-aperture (35'' radius) measurement on the extended emission+star and subtracting the PSF-fitting star flux to give an extended emission flux density of (175–144=) 31 mJy. Measuring the 8 μm flux density is the most difficult since there is significant contamination from the PSF subtraction residuals. Measuring the emission on an image with the residuals masked in 35'' radius aperture with 35''–43'' background annulus gives a flux density of 10 mJy; applying the IRAC extended source photometric correction changes this to 7.4 mJy. Different approaches to residual masking and photometric aperture change the resulting flux density by ∼20%–30%, so the value is highly uncertain. This is hardly surprising since the extended emission is more than 100 times fainter than the star at 8 μm. Consistent with this result is the classification of the star's mid-infrared spectrum as normal, with no dust, based on data from the PHT-S instrument on the Infrared Space Observatory (Hodge et al. 2004).

We can ask whether the galaxy-like appearance of the extended emission near SZ Tau is consistent with its mid-infrared flux densities and colors. The object is rather bright for a galaxy: number counts at 24 μm (Shupe et al. 2008) show that there are only 3.5–4 galaxies per square degree brighter than about 15 mJy. Taken together, all of our 24 μm Cepheid images cover about 0.5 deg², so we might expect one or two such galaxies in our survey. Figure 8 compares the colors of the SZ Tau extended emission with some comparison samples. The color f_8.0/f₂₄ = 0.24 is consistent with the colors of both nearby and distant galaxies (Muñoz-Mateos et al. 2009; Hainline et al. 2009). However, the flux ratio f₂₄/f₇₀ = 0.42, while inconsistent with an unobscured stellar source, is rather high for a galaxy detected at 70 μm; such galaxies typically have 0.05 ≲ f₂₄/f₇₀ ≲ 0.11 (Frayer et al. 2006). The flux ratios and colors are more consistent with those of candidate young stellar objects (YSOs) detected in the Cepheus Flare (Kirk et al. 2009), although YSOs do not typically have lenticular morphologies. Planetary nebulae can have asymmetric shapes, but the SZ Tau emission has quite different colors (see Figure 8). The best candidates to explain the extended emission near SZ Tau are either an unusually shaped Galactic object (possibly a YSO) or a background galaxy with very weak cold dust emission. This mysterious object warrants further investigation. A more definitive conclusion on its nature could be reached with higher spatial resolution infrared observations, in which the flux from the star and extended emission are more clearly separated.

**Figure 8.** Flux density ratios for submillimeter-detected galaxies (Hainline et al. 2009), nearby galaxies (Muñoz-Mateos et al. 2009), Cepheus young stellar objects (Kirk et al. 2009), LMC planetary nebulae (Hora et al. 2008), and the extended emission near SZ Tau.
Download figure:
Standard image High-resolution image

3.6. T Mon and X Cyg

These stars show extended emission in the IRAC 8.0 μm band with only tentative detections in the MIPS bands; images can be seen in Figures 9 and 10. In the image of T Mon, there appears to be cirrus across the entire field of view. However, there is brighter nebulosity closer to the star, specifically to the north and south in an hourglass-like shape extending to about 45'' radius (6.2 × 10⁴ AU, or 0.3 pc, at a distance of 1389 pc). The emission is centered around the star, suggesting that it may be locally illuminated. The extended emission near X Cyg has the second-lowest significance (after GH Lup) of the stars with emission above background levels. It is brightest to the northwest of the star and concentrated within a radius of 30'' (3.5 × 10⁴ AU, or 0.16 pc, at a distance of 1163 kpc). For both stars, the 24 μm images show hints of enhanced brightness at similar positions to the 8 μm emission, but these positions are too close to the PSF subtraction residuals to make definitive conclusions or to measure a surface brightness. There is no corresponding detectable emission at 70 μm. With both stars having extended emission observed at only one wavelength, the detections must be regarded as tentative. Observations with the James Webb Space Telescope, particularly its coronographs, would provide an avenue to confirm or refute them. If the detections are confirmed, the properties of these stars may provide a challenge in relating them to any mass loss. Neilson & Lester (2008) predicted that mass-loss rates, while showing large scatter with period, are generally larger for long-period stars. T Mon is a long-period star with a companion (Evans et al. 1999), while X Cyg has a period of 16.4 d and no known companion hotter than type A5 (Evans 1992). However, a low-mass companion with an orbit in the plane of the sky is not ruled out for X Cyg and might be important in forming and/or shaping an extended nebula.

**Figure 9.** PSF-subtracted images of T Mon in the (left to right) 5.8 μm and 8.0 μm IRAC bands, and the 24 μm and 70 μm (not PSF-subtracted) MIPS bands. All images are 2.5 arcmin square.
Download figure:
Standard image High-resolution image

**Figure 10.** PSF-subtracted images of X Cyg in the (left to right) 5.8 μm and 8.0 μm IRAC bands, and the 24 μm and 70 μm (not PSF-subtracted) MIPS bands. The field of view is a 2.5 arcmin square.
Download figure:
Standard image High-resolution image

4. DISCUSSION AND CONCLUSIONS

Spitzer mid-infrared observations of Cepheids have produced no strong evidence for large amounts of warm circumstellar dust (Marengo et al. 2010a). Here we use the same observations to place limits on cooler dust. Of our sample of 29 Cepheids, two stars (RS Pup and δ Cep) have well-detected extended infrared emission, two (GH Lup and S Mus) have possible extended emission at 24 and 70 μm, and a further three (ℓ Car, T Mon, and X Cyg) show tentative evidence for extended infrared emission at shorter infrared wavelengths only.

Is there anything particularly special about the stars in which we do have (tentative) evidence for mass loss? Our observations of the nebulosity around δ Cep and S Mus suggest that the detectability of a dustless wind might be enhanced when interstellar matter around the Cepheid is swept up into a bow shock, and a hot companion heats the material. Compared to the emission around RS Pup and GH Lup, the S₂₄/S₇₀ ratio (a proxy for dust temperature) is higher near S Mus and δ Cep. Both of these stars have known hot companions, while RS Pup and GH Lup do not (although low-temperature companions cannot be excluded using current data). A counter-example might be provided by T Mon, which also has a hot companion, but does not have detectable extended emission at 24 or 70 μm or an infrared excess; however, T Mon's companion is cooler (type A0) than that of the other two stars (B3 and B7-8).

The surface brightness ratios S_8.0/S₂₄ are a rather crude tool for deriving detailed physical conditions, but can give some insight into the PAH content of the ISM near these stars. The low PAH content indicated by the S_8.0/S₂₄ ratios for RS Pup, S Mus, and δ Cep—much lower than in the NGC 7023 reflection nebula or the ISM in general—suggests that the ISM around these stars has been diluted to some extent by the stellar winds. The higher PAH content near GH Lup would imply less dilution, suggesting that the emission there may be unrelated to the star. We caution that, while the amount of dilution might be indicated by the S_8.0/S₂₄ ratio, high spatial resolution spectroscopy in which the stellar and extended emission are separated would provide much greater physical insight into the conditions near these stars.

The IRAC-only extended emission near ℓ Car, T Mon, and X Cyg also suggests that these stars are somehow unusual. However, we have been unable to find any clear distinction between these stars and the three which also show extended MIPS emission, or between all seven stars and our full Cepheid sample. Like RS Pup, ℓ Car and T Mon have long periods (>27 days) and low temperatures; GH Lup, S Mus, and δ Cep have periods <10 days, and X Cyg is intermediate with a period of 16.4 days. Six of the seven stars have super-solar metallicity (Romaniello et al. 2008; Andrievsky et al. 2005; Kovtyukh et al. 2005) but the difference in mean metallicity between these stars and the other stars in the sample is barely statistically significant (p = 0.025). Evans (1992) placed strong limits on the presence of hot companions in Cepheids: of the stars in our sample, 3 of the 7 with possible extended emission have such companions (S Mus, δ Cep, and T Mon), as do 5 of the 22 stars without extended emission (an additional 3/23 are radial-velocity binaries). With such small numbers it is not yet possible to say whether the presence of hot companions enhances the detection of extended infrared emission.

RS Pup is clearly different from all of the other Cepheids in our sample in terms of the extent and brightness of its infrared emission. One possible explanation is that this star has swept up more material, either by having been on the instability strip for longer (unlikely, since long-period stars evolve faster) or by being located in a region of higher density in the ISM. An alternative explanation is suggested by visual examination of a Schmidt plate taken by one of us (D.W.) in the region of RS Pup. The image reveals a number of near-identical dust clumps, with similar angular sizes, in the vicinity of the RS Pup nebula. If these clumps are at the same physical distance and are physically related, but only one of them has a Cepheid in it, it could be argued that they are all likely pre-existing nebulae—that is, not the result of Cepheid mass loss. The counter-argument would be that the ring-like structures observed in the surrounding nebula appear to be centered on RS Pup (Havlen 1972): this would be an extreme coincidence if the nebula were unrelated to the star.

Before our Spitzer observations, only two Cepheids had unambiguous evidence for circumstellar nebulae: ℓ Car and RS Pup. We have now doubled that number with the addition of S Mus and δ Cep; if extended emission around GH Lup, T Mon, and X Cyg is confirmed, this would increase the sample size even more. A rough estimate of the fraction of Cepheids with substantial dusty mass loss can be made as 4–7 of 29, or ∼14%–24%.¹² The true fraction of Cepheids with mass loss could be below this limit if the observed nebulosity is related to the gas and dust from which the star formed. This is however unlikely, because even for small space velocities (∼1 km s⁻¹, an order of magnitude less than the observed velocity of δ Cep and ℓ Car), in 50 Myr (the time it takes for an intermediate mass star to become a Cepheid) these stars will have moved away from their birthplace by several tens of parsecs. This is much larger than the size of the observed nebulosity around the stars in our sample.

Addressing the initial motivation for this study, it appears that mass loss, at least its dusty component visible in the infrared, can only partially account for the Cepheid mass-loss discrepancy (e.g., up to 1/5th of a 10% discrepancy in the case of δ Cep, see Marengo et al. 2010b). Our estimates of Cepheid mass loss (in the range 10⁻⁸ to 10⁻⁹ M_☉ yr⁻¹) are comparable to previous studies based on the detection of infrared and UV excess (e.g., Welch & Duric 1988; Deasy 1988), but the higher spatial resolution of our images helps to separate local emission from background contamination. The detection of shocked structures, as in the case of δ Cep and possibly ℓ Car, is a compelling indication of current mass loss in those stars. While the total mass lost over the Cepheid phase is small, the confirmation that these processes are active needs to be taken into account for a proper modeling of Cepheid evolution. The presence of significant Cepheid winds can affect hydrodynamic models of Cepheid pulsations and the way shock waves propagate in the Cepheid atmospheres. Finally, the observed level of circumstellar dust needs to be further explored, as it may have a significant effect on inferred Cepheid distances as shown by Neilson et al. (2010).

The authors thank H. Neilson for useful conversations and L. D. Matthews for assistance with computing the projected space velocity of ℓ Car. We thank the referee for helpful comments. This work is based on observations made with the Spitzer Space Telescope, which is operated by the Jet Propulsion Laboratory, California Institute of Technology under NASA contract 1407. Support for this work was provided by NASA through an award issued by JPL/Caltech. P.B. and D.L.W. acknowledge research support through Discovery Grants from the Natural Sciences and Engineering Research Council of Canada; P.B. also acknowledges support from an Ontario Early Researcher Award. N.R.E. acknowledges support from the Chandra X-Ray Center Grant NAS8-03060. D.H. was supported in part by the National Science Foundation REU and Department of Defense ASSURE programs under grant no. 0754568 and by the Smithsonian Institution.

Facilities: Spitzer (IRAC, MIPS) - Spitzer Space Telescope satellite

GALACTIC CEPHEIDS WITH SPITZER. II. SEARCH FOR EXTENDED INFRARED EMISSION

Article metrics

Permissions

Author affiliations

Dates

ABSTRACT

1. INTRODUCTION