LIFTING THE VEIL OF DUST FROM NGC 0959: THE IMPORTANCE OF A PIXEL-BASED TWO-DIMENSIONAL EXTINCTION CORRECTION

The GALEX FUV and NUV images are obtained from the Multi-Mission Archive at the Space Telescope Science Institute⁶ (MAST). NGC 0959 was observed in the GALEX All-sky Imaging Survey (AIS; Martin et al. 2005) and the GALEX Nearby Galaxy Survey (NGS; Gil de Paz et al. 2007). Since the galaxy is barely visible in the AIS images, we will use only the deeper (∼1695 s) NGS data for our analysis, focusing only on a 75 × 75 image section centered on NGC 0959. Since the average FWHM of stars near NGC 0959 is larger in the NUV image (e.g., Martin et al. 2005), we convolve the FUV image to match the PSF in the NUV.

Ground-based UBVR images were obtained by Taylor et al. (2005) with the direct CCD imager at the VATT at Mt. Graham International Observatory (MGIO) in Arizona. These flux-calibrated images are available through the NASA/IPAC Extragalactic Database⁷ (NED). For a detailed description of this data set, we refer the reader to Taylor et al. (2005, and references therein). The effective exposure times in U, B, V, and R are 1200, 600, 480, and 360 s, respectively. The native resolution and pixel scale are ∼13 (FWHM) and 037 pixel⁻¹, respectively. These images are registered, convolved, and resampled to match the orientation, PSF, and pixel scale of the GALEX NUV image.

The Spitzer/IRAC 3.6–8.0 μm pipeline-product images were obtained from the Spitzer Archive⁸ via Leopard. For each filter, the mosaiced image was created with an effective exposure time of 26.8 s. The native pixel scale is 12 pixel⁻¹, and the effective resolution (FWHM) ranges from ∼22 at 3.6 μm to ∼23 at 8.0 μm. Like for the VATT images, we match the orientations, PSF, and pixel scale of the IRAC images to those of the GALEX NUV image.

To visually (qualitatively) investigate the effect of our pixel-based extinction correction on an image of NGC 0959, we first construct two color composites of the galaxy, composed of the Spitzer/IRAC 3.6 μm (red channel), the ground-based V (green channel), and the GALEX FUV (blue channel) images. Figures 2(a) and (b) show the color composites before and after extinction correction. The image resolutions are matched to that of the GALEX NUV image. For easy comparison, both images were created using the same color stretch and with the IRAC 8.0 μm contours over-plotted.

Zoom In Zoom Out Reset image size

In Figure 2(a), there are several regions in the galaxy that appear much redder than other parts of the galaxy. These regions include: (a) the southern half of the galaxy, especially along the southern "edge" of the galaxy disk (as defined by our S/N_min = 3.0 requirement in each of the filters); (b) the strong 8.0 μm emission region that is running from NW to SE of the galaxy through its center, appearing especially redder at its northern end point; and (c) some localized regions at the northwestern and eastern edge of the galaxy disk. Since dust features are not resolved at GALEX resolution, we cannot tell whether these pixels in Figure 2(a) are red due to extinction, or are dominated by intrinsically red stellar populations. From the distribution of the estimated visual dust extinction A_V (Figure 1), however, we infer that these red pixels are indeed red due to intervening dust at these locations.

After applying our extinction correction, Figure 2(b) shows that many of the strikingly red pixels in Figure 2(a) indeed have become bluer. Especially regions (b) and (c) have become much bluer than before, while some of the regions (a) still seem to have relatively red colors compared to other parts of NGC 0959's disk. There are some other regions which have become much bluer as well—mostly located in the SW region and northern regions of the galaxy. These regions already had a relatively blue hue in Figure 2(a), but become much bluer and brighter after extinction correction. They likely are actively SF regions.

To see whether these red and blue regions in Figure 2 actually correspond to dust features and SF regions, we created a third color composite image with a higher spatial resolution from the ground-based UVR images at their native resolution and pixel scale of FWHM ≃ 13 (matched across UVR) and 037 pixel⁻¹. Figure 3 shows that the blue regions that become bluer and brighter are indeed likely SF regions. It also shows that the redder pixels in region (b) described above seem to be caused by thick dust lanes running from the NW to SE in the galaxy. It is hard to see whether regions (a) and (c) are caused by a dust lane or not, but regions (a) seem to be distributed around bluer clumps in Figure 3. The localized regions (c) do not appear to be unrelated background or foreground objects. Such objects usually have colors sufficiently different from those of the genuine galaxy pixels that they would appear as a distinct and separate branch or grouping of pixels in pCMDs and pCCDs. Since we found no such feature in the examined pCMDs and pCCDs for NGC 0959 (e.g., Figure 4), we treat these regions as parts of the galaxy.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

To study the effect of our pixel-based extinction correction in a more quantitative way, we examined pCMDs and pCCDs using various combinations of images from FUV through 8.0 μm for significant features or groupings in the pixel distribution. A distinct grouping of pixels in these diagrams should indicate that those pixels are dominated by the same or a similar mix of stellar populations. Among different diagrams examined (not shown here), a pCCD of (B − 3.6 μm) versus (FUV − U) color after extinction correction (Figure 4(a)) was selected, because it clearly shows distinct tracks and groupings of pixels. In particular, distinct red and blue sequences can be identified.

The uncertainties in the colors at the 25th and 75th percentiles are ∼0.01 and ∼0.02 mag in (FUV − U), and ∼0.05 and ∼0.08 mag in (B − 3.6 μm), respectively. Since the gap of ∼0.2 mag at the bluer end of (FUV − U) color between the two sequences in Figure 4(a) is much larger than the photometric uncertainties in the colors, this separation of pixels is not a random effect. The sequences are also not the result of differences in residual reddening, as can be seen by comparing the directions of the sequences with that of the reddening vector drawn in Figure 4(a). The red and blue sequences are connected in the 1.15 ≲ (FUV − U) ≲ 1.65 mag color range by pixels with intermediate colors. At colors redder than (FUV − U) ≃ 1.65 mag, the pixels again somewhat separate into sequences with redder and bluer (B − 3.6 μm) colors.

As can be seen in various spectral energy distribution (SED) models (e.g., Bruzual & Charlot 2003; Anders & Fritze-von Alvensleben 2003; Maraston 2005; Kotulla et al. 2009), the (FUV − U) color is very sensitive to age for the youngest stellar populations. Since most of the FUV flux is emitted by young, massive OB stars, combining a GALEX and an optical broadband filter provides strong age constraints (e.g., Kaviraj et al. 2007). The (B − 3.6 μm) color was selected empirically, while examining different combinations of colors. The IRAC 3.6 μm bandpass is commonly used as a stellar mass distribution tracer (e.g., Willner et al. 2004), because it is associated more with the distribution of redder and older stars. Since the optical B band is generally sensitive to younger stars, the (B − 3.6 μm) color can be used to distinguish mixtures of stellar populations with and without significant recent high-mass star formation. The combination of (FUV − U) and (B − 3.6 μm) colors in Figure 4(a), therefore, provides a powerful diagnostic of the recent star formation history averaged over a pixel.

Based on the distinct red and blue sequences, as well as transition regions identified in Figure 4(a), we separated pixels into six different groups (color-coded in Figure 4(a)). The group boundaries (selection criteria) are summarized in Table 1. The exact location of the boundaries between some of these pixel groups is somewhat arbitrary, but is motivated by the following considerations.

The Group I pixels (color-coded purple) reside on the blue part of the blue sequence, and have blue colors in both (B − 3.6 μm) and (FUV − U). The dominant stellar populations are very young and massive OB stars, while older stellar populations contribute little to the total flux in these pixels. The (B − 3.6 μm) and (FUV − U) colors become redder as these young stellar populations age, and the number of remaining OB stars decreases.

On the other extreme, the Group VI pixels (color-coded red) have red colors in both (B − 3.6 μm) and (FUV − U). These pixels are therefore dominated by old, quiescent, "red-and-dead" stellar populations and do not contain a detectable fraction of young, massive stars.

The Group II pixels (color-coded blue) have the same blue (FUV − U) color range as Group I, but redder (B − 3.6 μm) colors. The blue (FUV − U) color implies the presence of OB stars. The redder (B − 3.6 μm) color indicates that Group II pixels have a non-negligible contribution to the total light from older (underlying or superposed along the line of sight) stellar populations.

Group V pixels (color-coded orange) cover the same, red, (FUV − U) color range as Group VI, indicating that they too lack a detectable fraction of young, massive OB stars. Yet, their bluer (B − 3.6 μm) color suggests that the flux in these pixels is dominated by light from intermediate age stellar populations.

The Group III and Group IV pixels (color-coded green and yellow, respectively) are located between these extreme cases. The light in these pixels is likely dominated by stellar populations in transition between the extreme groups along either blue or red sequences, or between the blue and the red sequence, or represents a mixture of stellar populations (from unresolved adjacent regions or regions superposed along the line of sight) with different star formation histories.

Figure 4(b) shows the pCCD of (B − 3.6 μm) versus (FUV − U) color before the application of the extinction correction. For each pixel, the color coding was preserved from that in Figure 4(a). This allows us to track the effect of applying the extinction correction in this color–color space. Interestingly, as shown in the inset of Figure 4(b), once the color coding is removed, obvious features like the blue and red sequences of Figure 4(a) are no longer discernible. The Group I and II pixels, for example, which were clearly separated in Figure 4(a), have blended to form a continuous distribution in Figure 4(b). The main difference between Groups I and II, which both are characterized by the presence of OB stars, is the fraction of light contributed by older stellar populations. Group III and IV pixels, defined as having 1.15  ⩽  (FUV − U) <  1.65 mag after extinction correction, are affected differently by the extinction correction (as a comparison of Figures 4(a) and (b) shows). Like Group I and II pixels, Group IV pixels are found scattered out to much redder (B − 3.6 μm) colors before extinction correction. Most of the Group III pixels, on the other hand, can be located in the same color–color space in both panels of Figure 4. The same difference is seen between Groups V and VI. This indicates that Group III and V pixels have almost no measurable dust extinction, while Group IV and VI pixels are significantly obscured by dust. As a side note, the reduction of the scatter going from the observed to the extinction-corrected version of the (B − 3.6 μm) versus (FUV − U) pCCD is only possible if the extinction correction applied to each individual pixel was appropriate, at least to first order.

The pixel-averaged dust extinction estimated by our method in NGC 0959 (Figure 1) indeed varies from A_V = 0.0 to A_V ≃ 0.8 mag arcsec⁻². In Paper I, we also measured the average visual dust extinction across the entire galaxy, $\overline{A_V}$ = 0.064 ^+0.086_−0.049 mag arcsec⁻², and the azimuthally averaged radial extinction profile, A_V(R), which has a central value of A_V(R = 0) ≃0.25 mag arcsec⁻², but quickly drops to below A_V = 0.1 mag arcsec⁻² beyond a radius of ∼0.2 R₂₅, where R₂₅ denotes the major axis radius at the μ_B = 25.0 mag arcsec⁻² isophote (RC3; de Vaucouleurs et al. 1991). These average A_V values are much smaller than the peak pixel-based A_V estimate. Also, while the $\overline{A_V}$ and the radial profile do not become zero at any radius, with our method (see Paper I), about ∼55% of the analyzed pixels in NGC 0959 have A_V = 0.0 mag arcsec⁻² to within the photometric uncertainties.

Before we proceed to the following section, let us consider what might cause the difference in dust extinction between Groups I and III on the blue sequence. Dust and gas are easily removed by starburst heating (e.g., Mihos & Hernquist 1994, 1996) and stellar winds (e.g., Murray et al. 2005), and both mechanisms become stronger as the size of an OB association becomes larger. Therefore, if the dominant stellar populations in Group III pixels are indeed evolved Group I populations wherein OB stars have died off, we would expect them to suffer less extinction, unless some mechanism for rapid reformation and pervasive distribution of dust were at play. Figures 4(a) and (b) show that, once dust free, these populations do not become significantly dustier evolving from Group III to Group V, apparently ruling out such rapid reformation. Since we do not observe a similar strong drop in extinction going from Group II to Group IV, we conclude that most of the extinction in those pixels is not physically associated with the OB associations themselves, although they could still represent star formation "blisters" on the far side of large molecular cloud complexes.

To determine whether our pixel-based extinction correction can be used as a general method, or whether NGC 0959 is a special case, we applied the same method to NGC 7320 (SA(s)d at D = 14.0 ± 1.0 Mpc, sampling 102 × 102 pc² pixel⁻¹) and UGC 10445 (SBc at D = 20.0 ± 1.4 Mpc, 146 × 146 pc² pixel⁻¹), which, although more distant, are selected from our larger sample of 45 galaxies with FUV through mid-IR imagery as relatively close analogs. A more detailed analysis of the distribution of extinction and the intrinsic stellar populations of these galaxies will be performed in K. Tamura et al. (2010, in preparation; Paper III). Here we simply compare their (B − 3.6 μm) versus (FUV − U) pCCDs before and after the application of our pixel-based extinction correction, to help validate the current results for NGC 0959.

Figure 5 shows the distribution of the pixels of NGC 7320 and UGC 10445 that meet our S/N criteria in the same color–color space as in Figure 4. The main panels show the distributions after extinction correction and the insets show the distributions before extinction correction. For each galaxy, both the main panel and inset show the same ranges in color. The vertical dotted lines in the insets represent (FUV − U) = 0.0 mag. Reddening vectors corresponding to A_V = 0.2 mag arcsec⁻² are drawn in the lower right corners of the main panels. Since the area subtended by a single pixel is larger than was the case for NGC 0959, the blending of light from distinct stellar populations within a single pixel becomes more significant, resulting in less clear separation of blue and red sequences. Yet, different groupings of pixels, analogous to those defined in Figure 4(a) for NGC 0959, are still recognizable in Figure 5 for both galaxies.

Zoom In Zoom Out Reset image size

Having shown that application of our pixel-based extinction correction reveals significant groupings of pixels—dominated by different types or mixtures of stellar populations—in the (B − 3.6 μm) versus (FUV − U) pCCD, our next question becomes whether the spatial distribution of pixels belonging to these pixel groups will reveal meaningful large- and small-scale physical structures within NGC 0959. To address this question, and to study how each pixel group relates to the visual and physical properties of stellar populations, we plot the pixel groups defined in Figure 4(a) onto a two-dimensional pixel coordinate map (Figure 6). Each square in Figure 6 represents a 15 × 15 pixel, color-coded according to the pixel group it is a member of. Black dotted contours again trace the Spitzer/IRAC 8.0 μm polycyclic aromatic hydrocarbon (PAH) emission. We emphasize that each pixel group was defined without using any pixel coordinate information. In other words, this is the first time that the spatial distribution of pixels belonging to each of the selected pixel groups is revealed.

Zoom In Zoom Out Reset image size

First and foremost, Figure 6 demonstrates that the different pixel groups are not distributed randomly across the face of NGC 0959's galactic disk, but cluster in well-defined regions. No systematics in the data processing and process of defining the pixel groups would be expected to result in spatial artifacts larger than 3 × 3 pixels. Since most regions cover much larger contiguous areas, we conclude that also the smallest spatial groupings must be genuine. Since defining the different pixel groups was possible only in the extinction-corrected (B − 3.6 μm) versus (FUV − U) pCCD, a two-dimensional extinction correction is thus crucial to reveal regions with systematically different star formation histories.

Group I pixels (purple) are found in and around some of the bluest regions in the extinction-corrected color composite of FUV, V, and 3.6 μm images (Figure 2(b)), most of which are recognizable as such in the uncorrected, higher resolution, ground-based UVR color composite (Figure 3) as well. Since the stellar populations dominating these pixels are much brighter than those in other regions (Figure 3), these regions must indeed be vigorously forming stars.

Group II pixels (blue) are generally distributed in regions adjacent to the Group I pixels. Each of the regions (b) and (c) of Section 5.1 is found to belong to this pixel group, as well. In particular, much of a more or less linear structure that runs from the NW to SE of the galaxy through its center, with high 8.0 μm surface brightness (PAH emission) and signs of higher than average obscuration (Figures 1 and 3), consists of pixels that belong to Group II. These regions have the blue (FUV − U) colors characteristic of OB associations, but are redder in (B − 3.6 μm), because a larger fraction of the light is contributed by older stellar populations. These observations are consistent with the presence of a stellar bar. Although NGC 0959 has been studied before (e.g., Esipov et al. 1991; Taylor et al. 2005), to our knowledge this bar has not previously been reported. The dust-corrected color composite in Figure 2(b) also suggests that NGC 0959 has a non-negligible bulge or central condensation, that is partially obscured by dust in Figures 2(a) and 3. We suggest, therefore, that its morphological classification be changed from Sdm (RC3; de Vaucouleurs et al. 1991) or Sc/Irr (UGC; Nilson 1973) to SBcd.

An interesting contrast in average stellar population age is revealed by comparing the spatial distributions of Group III (green), Group IV (yellow), Group V (orange), and Group VI (red) pixels. Whereas Group III pixels are mostly found in the northwestern half of the galaxy, Group V and VI pixels are distributed predominantly toward the eastern and the southeastern periphery of the galaxy (although a smaller region of Group V pixels appears along the northern rim). The regions occupied by Group III pixels appear neither particularly blue (actively star forming) nor red (quiescent) in Figure 2(b), consistent with the idea that intermediate age (a few 100 Myr) stellar populations are the dominant contributors to the flux in these pixels. The extinction map (Figure 1) shows that these pixels suffer no, or at most, minimal extinction by dust. The Group V regions correspond to regions that have a smooth appearance without abrupt changes in surface brightness in Figures 2 and 3, and lacking any signatures of dust (Figures 1 and 3). Their red (FUV − U) colors and neutral (B − 3.6 μm) colors indicate that these must be regions with intermediate age (∼1–2 Gyr), largely unattenuated stellar populations in the outskirts of the galaxy. The Group VI pixels (red) in the southern half of the galaxy disk suffer some extinction (Figure 1), but the low surface brightness in these regions does not allow one to easily discern dust features in Figures 2 and 3. Even though these regions suffer some extinction, Figure 4(a) demonstrates that these pixels are red mostly because the light is dominated by older (likely older than a few Gyr) stellar populations. The Group IV pixels are distributed throughout the galaxy between the other populations, but perhaps mostly surrounding the Group II pixels. Like Group III, their (FUV − U) colors indicate the presence of relatively young stars and the absence of OB stars, but the redder (B − 3.6 μm) colors of Group IV pixels show that a larger fraction of the flux is contributed by older populations. The extinction map (Figure 1) shows that these pixels tend to suffer somewhat less attenuation than Group I and II pixels. The overall impression is that of a large-scale past star formation episode that started in the south or southeast and propagated toward the northwest, and that is unrelated to the ongoing massive star formation traced by Group I and II pixels, which presently is concentrated more toward the western half of NGC 0959.

The apparent ridge of somewhat higher extinction along the southeastern edge (as defined by our S/N ⩾ 3 criterion) of NGC 0959, so striking in the extinction map (Figure 1) might hint at the presence of either a warp in the galactic disk or an outer spiral arm delineated by a dust lane. In either case, no significant star formation must have been associated with that portion of the galactic disk for the past few Gyr.

Concerning the SE-to-NW stellar population gradient of NGC 0959 and its SE extinction ridge, it is worth noting that NGC 0959 is a member of the NGC 1023 galaxy group, which contains 12 other major galaxies (Tully 1980). The center of that group is located ∼1°.6 (∼240 kpc at D≃ 10 Mpc) south of NGC 0959. The group environment and/or gravitational interaction with NGC 0949—the group member nearest in projection on the sky—may well have played a role in producing the extinction ridge and the stellar population gradient in the disk of NGC 0959. We believe this certainly deserves further study.