A PILOT STUDY USING DEEP INFRARED IMAGING TO CONSTRAIN THE STAR FORMATION HISTORY OF THE XUV STELLAR POPULATIONS IN NGC 4625

NGC 4625 (D = 9.5 Mpc for H₀ = 70 km s⁻¹ Mpc⁻¹; Kennicutt et al. 2003) is a Magellanic dwarf spiral that could be interacting with the neighboring Magellanic dwarf NGC 4618⁶ or the newly discovered dwarf galaxy NGC 4625A (Gil de Paz et al. 2005). The high surface brightness inner disk is of one-armed spiral morphology and is only a few kiloparsecs (R₂₅ = 11 ×  095; Kennicutt et al. 2003) in diameter. It has a well mapped H i disk, coincident with the XUV emission, of nearly constant surface density from its optical radius to ∼5'  (Bush & Wilcots 2004; Gil de Paz et al. 2005; Kaczmarek & Wilcots 2012). A known low surface brightness optical disk, coincident with the extended H i, has been imaged in the B and R bands, but without the depth needed for precise color measurements in the outer disk (Swaters & Balcells 2002; Gil de Paz et al. 2005). Abundances indicate a discontinuity in metallicity at R₂₅ and a nearly constant outer disk abundance of ∼0.2 Z_☉ (Goddard et al. 2011).

IRAC observations of NGC 4625 were carried out during Spitzer Cycle 6 (PID 60072, PI S. Bush) using standard observing parameters. Each pointing within a 2 × 3 position map having 252'' spacing was observed with 30 × 100 s dithered exposures. We selected a cycling dither and a medium dither pattern amplitude. The observations were configured so that NGC 4625 was centered in the coverage map at 3.6 μm, the more sensitive of the two operating IRAC detectors. Corresponding images taken at 4.5 μm cover the optical disk, but not the entire field surrounding the galaxy that we used to calculate background emission, consequently we do not use them in our analysis. NGC 4625's companion, NGC 4618, intrudes onto a portion of the images.

The first steps in the IRAC data reduction, using the Basic Calibrated Data (BCD), were carried out by team members at the Center for Astrophysics. All BCD frames not containing the center of NGC 4625 were object-masked and median-stacked according to the array detector orientation; the resulting stacked image was then visually inspected and subtracted from the individual BCDs. This was done to eliminate long-term residual images arising from prior observations of bright sources; it also served to minimize gradients in the celestial backgrounds. Dust contamination can be an issue in the 3.6 μm band (Zibetti & Groves 2011), but in the absence of corresponding images taken at longer wavelength IR images (unavailable in the Spitzer warm mission phase, when these observations were taken) we cannot correct for dust contamination. Since we are primarily concerned with the low metallicity, low density outer regions of the galaxy, we would not expect this to significantly alter our results.

After these preliminaries, the images were combined into two spatially registered mosaics using IRACproc (Schuster et al. 2006). IRACproc augments the capabilities of the standard IRAC reduction software (MOPEX). The software was configured to automatically flag and reject cosmic ray hits based on pipeline-generated masks together with a σ-clipping algorithm for spatially coincident pixels. IRACproc calculates the spatial derivative of each image and adjusts the clipping algorithm accordingly. Thus, pixels where the derivative is low (in the field) are clipped more aggressively than are pixels where the spatial derivative is high (point sources). This avoids downward biasing of point-source fluxes in the output mosaics.

After mosaicing, a plane was fit to the binned image to subtract a recognizable diffuse background gradient. After masking the galaxies in the image, the data was binned in 10 × 10 pixel bins. The median of the resulting bins was fit with a linear fit in two dimensions following the procedure proposed by Westra et al. (2010, Appendix). Further details of the background subtraction are discussed in the Appendix.

NGC 4625 was imaged by the Galaxy Evolution Explorer (GALEX) as a part of the Nearby Galaxies Survey (Gil de Paz et al. 2007a) in the far-ultraviolet (FUV) and near ultraviolet (NUV) bands with a total exposure time of 3242.2 s. The exposures were processed by the GALEX pipeline (Martin et al. 2005, version 6.0.1) and retrieved from the NASA Mikulski Archive for Space Telescopes.⁷

Background counts in the UV images are very low, usually only a few counts per pixel. Consequently background pixel values follow a Poisson distribution and cannot be approximated by a symmetric, Gaussian distribution for background subtraction. Subtracting the mean of a Poisson distribution does not center it symmetrically around zero, leaving a risk that the noise will bias measurements of the signal. When the signal exceeds the background by many orders of magnitude, the background counts can be safely ignored, but faint outer disk features in these images are often only a few times brighter than the mean background. Consequently, careful background subtraction is needed for our study. We rebinned the UV images into pixels large enough that 99% of the newly defined pixels contained at least 10 counts. The resulting FUV pixels are 9'' × 9'' and the resulting NUV pixels are 3'' × 3''(6 × 6 and 2 × 2 original pixels respectively). The resulting pixel distribution is Gaussian and can be background subtracted. From this point on, only the rebinned image is used in this work.

The rebinned images were background subtracted using the same technique we used to subtract the background gradient in Section 2.2. A wide area surrounding both NGC 4625 and its companion NGC 4618 were masked, the rebinned images were now binned for the purposes of background subtraction, and a surface was fit to the median value of the bins following Westra et al. (2010, Appendix). The surface was fit iteratively, clipping 3σ outliers until the fit converged. The FUV image was divided into 3 × 3 pixel (27'' × 27'') bins and the median of the bins was fit with a plane. The NUV image was divided into 5 × 5 (15'' × 15'') pixel bins and fit with a surface of order two in both x and y to remove nonlinear, large scale variations in the background. Because the galaxy was masked, there is no danger that we subtracted the galaxy's emission with this fit.

H i maps of the NGC 4625 field were derived from observations taken with the Very Large Array in B and C configurations (Kaczmarek & Wilcots 2012). The moment map was created with natural weighting and 1800 cleaning iterations on the image file. A flux cut off of 2σ (0.73 M_☉ pc⁻²) was applied. The final synthesized beam was 1314 × 1302.

To calculate the radial profile of NGC 4625, we had to determine which, if any, of the sources spread across the image are a part of NGC 4625 and which are background galaxies. While in most galaxies 3.6 μm light is dominated by blackbody emission from older stellar populations, stars of all ages emit at 3.6 μm and younger AGB stars can be brighter in the 3.6 μm than older red giant stars. In environments possibly dominated by younger stellar clusters, such as the outer disks, the 3.6 μm emission could be dominated by AGB stars and it is possible that some of the sources in the image are hot young clusters. However, many of them must be high redshift galaxies (Fazio et al. 2004a; Ashby et al. 2013).

Due to the low resolution of the UV data (rebinned image; see Section 2) it is very difficult to identify 3.6 μm counterparts to UV clumps. Consequently, we could not separate young stellar clusters from high redshift galaxies based only on FUV, NUV and 3.6 μm emission. Optical imaging is not available to the depth of the 3.6 μm data, which means we could not use the optical colors to discriminate between background galaxies and NGC 4625 sources. It is consequently difficult to assess the contribution these sources have to the overall profile. Given the prevalence of background galaxies in the 3.6 μm (Fazio et al. 2004a; Ashby et al. 2013) and the lack of data to individually assess sources, we chose to mask all sources in the image and analyze only the smooth, unresolved profile of the galaxy. However, we note that we are likely masking a few sources that are a part of NGC 4625 and consequently our derived profile is a lower limit to the total 3.6 μm light.

We used the Astronomical Point source Extractor for MOPEX (APEX) to detect and mask sources. APEX subtracts a local median in 15 × 15 pixel windows from the image to remove any smooth variations in the image, including the unresolved emission from the galaxy, and selects groups of four pixels or more that are 3σ above the local noise value. Because the many background galaxies detected by IRAC are not point sources, we did not attempt to fit and subtract every source. Instead, all detected sources were masked at 3σ above the background noise. To be sure we were accurately accounting for the emission in the wings of true point sources we fitted PRFs by eye to sources obviously exhibiting diffraction spikes and subtracted the fitted PRFs from the image in addition to masking their cores. The masked image is shown in Figure 2. This mask is inappropriate for measuring the profile of the inner disk, because the high surface brightness center of the galaxy is detected as a source, so we used the profile only to analyze outer disk properties.

Zoom In Zoom Out Reset image size

Next, we measured the 3.6 μm profile in circular annuli. Apertures are centered on R.A. = 12^h41^m526 and decl. = +41°16'215 (J2000) and are 135 wide in order to encompass the beam of the lowest spatial resolution data we analyze in this study, the H i maps (Section 2.4). To evaluate the profile error, we chose a distance beyond which we believe all the flux is due to background sources and use apertures beyond that radius to evaluate the error in the background. We chose the radius corresponding to the minimum in the H i profile (13 kpc, derived in Section 3.2) as the edge of the galaxy and use the median of the flux measured in annuli beyond that radius as the background level. Error bars were derived by taking an estimate of background uncertainty, the standard deviation of the apertures from the median, and adding sampling errors in quadrature. Sampling error is the error associated with having a limited number of pixels to calculate the mean ($\sqrt{{\sigma }/{N}}$, where N is the number of pixels and σ is the standard deviation of their values).

The profiles are shown in Figure 3. The black solid line is the profile without any masking. Beyond 75'', the profile rapidly flattens because it is dominated by the contribution of background galaxies. The dashed line is the profile with sources masked by APEX. In the outer disk this unresolved emission falls off sharply. We refer to the masked profile as the "unresolved" profile.

Zoom In Zoom Out Reset image size

Also plotted in Figure 3 is the profile derived from the SINGS survey 3.6 μm data which is 1/7 as deep as our data (Muñoz-Mateos et al. 2009). Muñoz-Mateos et al. (2009) accounted for background galaxies by applying a mask based on IRAC colors. Without deep observations at 4.5 μm and 8 μm to compare to our images, this method was not available to us, so we cannot compare the masking methods directly. However, as their study primarily focused on the contents of the inner disk, their masking choice was less conservative than ours and many faint background galaxies remain in their images after masking (Muñoz-Mateos et al. 2009, left side of Figure 2). Consequently we believe our more conservative masking choice allows a better characterization of the profile of the outer disk and that the difference in masking most likely accounts for the difference between the profiles. Our conservative masking choice is inappropriate for the inner disk, so the Muñoz-Mateos et al. (2009) profile should be used to study the inner disk. Note that at the transition between the inner and outer disk, approximately 60'', our profiles agree to within the error bars.

To derive colors and gas surface densities, we also calculated FUV, NUV and H i profiles. ⁸ We first transformed the UV images and H i map to the same pixel scale and orientation as the 3.6 μm image. The differing resolutions of the 3.6 μm and UV images make it inappropriate to apply the APEX mask used for the 3.6 μm image to the UV images. The 3.6 μm image could be smoothed to the UV resolution prior to masking, but this makes the discrimination between 3.6 μm sources and the 3.6 μm smooth emission more difficult. However, the 3.6 μm bandpass encompasses many more redshifted background galaxies than the FUV and NUV, so background galaxies are not as prevalent in the UV. Consequently, in the UV images, we only masked other low redshift galaxies in the field (NGC 4618, NGC 4625A) and bright foreground stars. Then the profiles were measured with the same apertures as in used for the 3.6 μm profile (Section 3.1). To subtract the minor contribution from background galaxies, in each band we measured the median surface brightness in point-source masked annuli beyond 280'', and subtracted the median value of these annuli from all aperture values to give the final profiles shown in Figure 4. The error in the profiles is taken to be the standard deviation from this median added in quadrature to aperture sampling errors. The FUV and NUV profiles are consistent with the previous analysis by Gil de Paz et al. (2005), with an inner disk, a break, and a shallower decline in the outer disk. The H i profile shows a slow decline from the break around 50'' to 280''. Because the H i distributions of NGC 4625 and its neighboring galaxy, NGC 4618, overlap, the profile then begins increasing again at 280''. For our purposes, we defined the minimum in the H i profile as the edge of NGC 4625.

Zoom In Zoom Out Reset image size

The 3.6 μm profile is similar to the UV and H i profiles in that it shows a strong break at 60'' and a smoothly declining outer disk.⁹ However, the slope of the outer disk profile is much steeper than either the UV or H i outer disk slopes. To quantitatively compare the profiles, we show color profiles in Figure 5. While the FUV − NUV color is constant with radius, the FUV − 3.6 μm color profile drops steeply from edge of the inner disk until it is no longer detected. For comparison we plot the gas phase metallicity gradient measured by Goddard et al. (2011), which, like the multi-wavelength profiles, shows a break at the edge of the inner disk.

Zoom In Zoom Out Reset image size

To constrain the star formation history of the outer disk of NGC 4625, we compared the derived colors to stellar population models. These models predict the emission of a stellar population with a given star formation history, metallicity and initial mass function (IMF). We chose the canonical Bruzual & Charlot (2003) models, which we believe have the most accurate representation of the near-infrared (JHK) contribution of AGB stars. However, we note that the contribution of AGB stars to 3.6 μm emission is not well understood and to estimate the error in the models we also consider Bruzual (2007) models, which differ from Bruzual & Charlot (2003) models by up to a magnitude in the 3.6 μm.

We first compared model colors for two extreme star formation rate histories, a constant star formation rate and an instantaneous burst of star formation (simple stellar population), to our colors. We selected a Chabrier IMF (Chabrier 2003) with mass limits of 0.1 and 100 M_☉ and two stellar metallicities based on the gas phase metallicities derived for NGC 4625's outer disk (Goddard et al. 2011). Using the [N ii]/[O ii] metallicity indicator calibrated by Bresolin (2007), Goddard et al. (2011) measured that the outer disk abundance ranges from 0.4 Z_☉ to 0.2 Z_☉. The comparison is shown in Figure 6, where Bruzual & Charlot (2003) models are the solid lines and Bruzual (2007) models are the dashed lines. Extinction vectors for both a Calzetti et al. (2000) and a Fitzpatrick (1999) Milky Way dust extinction curve for E(B − V) = 0.1 (Gil de Paz et al. 2005) are shown.

Zoom In Zoom Out Reset image size

Figure 6 shows that the FUV − 3.6 μm color provides the stronger constraint on the age of the stellar population and the FUV − NUV color provides the stronger constraint on the star formation history. The FUV − NUV colors, particularly near the edge of the inner disk, are mostly too red to have had a constant star formation history, but toward the edge of the outer disk the FUV − NUV error bars overlap with the constant star formation rate models. Once a star formation history is chosen, the FUV − 3.6 μm color constrains the age. In either case, the declining FUV − 3.6 μm color indicates that the age of the outer disk decreases with radius. In a disk that has undergone a burst, the age of the disk declines with radius from approximately 100 Myr to 10 Myr. In a disk that is constantly star forming, the age declines from approximately 10 Gyr to 100 Myr.

Galactic extinction may alter NGC 4625's colors. The extinction vectors in Figure 6 indicate that observed extinction curves disagree on the amount of extinction in the FUV − NUV, which varies from almost none to 0.2 mag, but they agree fairly well on FUV − 3.6 μm extinction of 0.7–1.0 mag. Accounting for this extinction, the outer disk colors are more consistent with a single burst of star formation model that has a younger age, but are also more consistent with a constant star formation rate model. It could be argued that applying Calzetti et al. (2000) extinction makes the colors of the inner radii of the outer disk too red to concur with a burst of star formation, but toward the inner disk the population will undoubtedly begin to be a composite of the inner and outer disk stellar populations, making interpretation difficult.

Reddening within NGC 4625 could also alter the colors. Using the H₂ and H i measurements from Schruba et al. (2011) we made a back-of-the-envelope calculation of the amount of extinction expected using the solar neighborhood value of N(H)/E(B − V) = 5.8 × 10²¹ (Bohlin et al. 1978). Depending on the reddening curve used, there could be approximately 0.5–0.8 E(FUV − 3.6) at 60'' and approximately 0.2–0.3 E(FUV − 3.6) at 170''. The maximum difference between 60'' and 170'' is then 0.5 mag of reddening, which is quite small compared to the 3.5 magnitude drop in color. Consequently, we do not believe the radial gradient in color is due to a radial gradient in extinction.

Existing NUV−B and B − R color profiles (Gil de Paz et al. 2005) are not precise enough to constrain the star formation history far into the outer disk. Near the edge of the inner disk they agree with the UV and 3.6 μm colors presented here: slightly too red to be constantly star forming and consistent with single stellar populations whose age is loosely constrained between 1 Gyr and 100 Myr.

While the errors allow for multiple interpretations, taken at face value, the colors of the outer disk imply a stellar population younger than 1 Gyr that decreases in age with radius. In other words, we have not discovered a reservoir of stellar mass that cannot be explained by the young population that is also responsible for the UV emission. However, we have compared the entire FUV profile to the smooth component of the 3.6 μm profile, which may not be a fair comparison. As discussed in Section 3.1, without deeper optical data or higher resolution UV data to selectively mask 3.6 μm sources, it is very difficult to determine what portion of the 3.6 μm sources seen near the galaxy are counterparts to young UV clusters. By comparing the colors derived from the unresolved 3.6 μm profile to stellar populations models, we implicitly assumed that no 3.6 μm sources that belong to the galaxy were masked and that the UV and 3.6 μm emission are from the same stellar population. If the 3.6 μm profile is missing some emission due to excessive masking, we would have underestimated the age of the disk.

It is also possible, even likely, that we are detecting a "composite population" of a young outer disk stellar population superimposed on an older, fainter population that extends from the inner disk to the outer disk. This would cause redder colors with decreasing radius as seen in NGC 4625's profiles. Instead of assuming that the unresolved 3.6 μm emission and the UV emission are from the same stellar population, we could have assumed that we have masked all the counterparts to the UV emission, because they are not smooth, and that our smooth profile reflects only a different, older population. Of course, the truth is likely in between, but considering both allows us to determine which conclusions are robust to our assumptions.

In Figure 7 we have calculated how long the current star formation rate, derived from the UV profile (Salim et al. 2007), must have been sustained in order to build the mass profile, which is derived from the 3.6 μm emission assuming reasonable 3.6 μm mass-to-light ratios from Figure 6. In Figure 7, we have also shown how mass-to-light ratios evolve with age, using stellar populations models with a constant star formation rate until the present and stellar populations models with a constant star formation rate until 200 Myr ago. The second model is designed to exclude the 3.6 μm contribution from the UV-emitting population, which we have now assumed we fully masked. At a given radius, the length of time the disk is required to form stars at the current star formation rate to create the smooth 3.6 μm profile is determined by the age at which the mass-to-light ratio and age are consistent between the two panels of Figure 7. At 60'', using Bruzual & Charlot (2003) models, these are consistent at longer timescales, approximately 8 Gyr, with higher mass-to-light ratios, while at 170'', they are consistent at shorter timescales, less than a Gyr, with low mass-to-light ratios. Of course, if the star formation rate was lower in the past, star formation could have extended further into the past.

Zoom In Zoom Out Reset image size

While the exact age of the population is difficult to say, two conclusions seem robust to assuming one stellar population or assuming two independent stellar populations, one emitting in the UV and one emitting in the 3.6 μm. First, the mean age of the outer disk decreases with radius. Second, the star formation at current rates in the outermost disk is a recent event.

While a young outer disk for NGC 4625 certainly seems plausible, there are a few key uncertainties in this analysis. As discussed in Section 3.1, Zibetti & Groves (2011) have shown that dust can make a substantial contribution to the 3.6 μm emission, which would violate our assumption that the 3.6 μm emission is emitted by stars. However, it is unlikely that there is enough hot dust in the low density, low metallicity outer disk of NGC 4625 to significantly alter the 3.6 μm emission profile. The stellar populations models also contribute uncertainty. There is easily a factor of two difference in 3.6 μm mass-to-light ratio between Bruzual & Charlot (2003) and Bruzual (2007), and choice of the IMF in these models may significantly alter the predicted emission of younger populations.

It is well established that stars migrate radially from their birthplaces in a galaxy (Roškar et al. 2008a, 2008b) and studies of star formation suggest that at the gas surface densities observed in the outer disk of NGC 4625, star formation occurs with low efficiency (e.g., Martin & Kennicutt 2001). If the stellar content in the outer disk was not born there, our stellar populations analysis does not reflect the characteristics of the outer disk, but rather a scattered inner disk. However, it is unlikely that many stars could migrate as far as 5 kpc from where they were born in less than 150 Myr. Therefore, FUV − NUV colors are unaffected and suggest that NGC 4625's young outer disk was born in-situ, in a low-efficiency sub-threshold regime (Martin & Kennicutt 2001; Kennicutt 1989). However, it is possible that stellar migration from the inner disk has created a low-mass older outer disk component coincident with the more recent star formation. Deep optical imaging would further constrain whether the outer disk is a single young population of decreasing age with radius or a composite population of decreasing average age, but it will not reveal whether this population was born in situ or has migrated from the inner disk.

It is difficult to determine conclusively if NGC 4625 is interacting with either of its neighbors: NGC 4618, which is about half its mass (Bush & Wilcots 2004; Kaczmarek & Wilcots 2012), or the dwarf galaxy NGC 4625A. A recent interaction could have compressed gas in the outer disk and initiated star formation in the recent past (Cox et al. 2008; Bush et al. 2008, 2010). The velocity field of the outer H i disk is extremely regular, surprising for a galaxy that has undergone a recent interaction. In fact, parameterized H i morphology of XUV disks galaxies is rarely distinguishable from that of other disk galaxies (Holwerda et al. 2012). However, the star formation rate is very low in the outer disk of NGC 4625 and only a small fraction of the mass needs to be compressed by tidally induced spiral structure to create the UV emission observed (Bush et al. 2008, 2010). If any distortion of the disk by interaction-induced warping is present, it is undetectable in our mosaics, because of the face-on orientation of the galaxy. Several of the best-studied XUV disks show minor signs of interactions, such as extended low surface brightness features, nearby companions or warped outer H i disks, while having an intact inner disk (Chonis et al. 2011; Thilker et al. 2005; Mihos et al. 2013). This suggests the tantalizing possibility that XUV emission is an indication of "flyby" encounters or interactions with a minor perturber.

A young NGC 4625 outer disk is consistent with inside-out disk formation and it seems plausible that a recent interaction with a companion caused a burst of star formation in the existing outer gas disk. However, it is still difficult to determine how the extremely extended outer gas disk was assembled. The metallicity of the disk gives some clues to its origin. In the closed box chemical evolution model, which assumes a galaxy does not exchange any gas with its environment, the relationship between a galaxy's gas fraction and metal yield depends only on the mass of stars that have formed in it, and is independent of star formation rate and history. Given a metal yield y_eff, metal abundance Z and gas fraction μ, under the closed box model y_eff = Z/ln (μ) (Edmunds 1990). We calculate the metal yield required by the closed box model for the gas fraction observed in the outer disk of NGC 4625 assuming H i is the dominant gas mass and a range of 3.6 μm mass-to-light ratio. The effective metal yield is shown in Figure 8 with error bars reflecting the component profiles' errors. A reasonable metal yield would imply consistency with a closed box model and suggest that the metal content of the gas is entirely due to the formation and evolution of the presently observed stellar content. In NGC 4625's outer disk, depending on the mass-to-light ratio chosen the metal yield increases from around 0.01 at the edge of the inner disk to 0.3 near the edge of the outer disk. These metal yields are very high, consistent with other measurements of outer disk yields (Werk et al. 2011, 2010) and measured high metallicities in outer disks (Bresolin et al. 2009, 2012) and inconsistent with the closed box model. This suggests that most of the metal content in the outer disk was not created by in-situ star formation.

Zoom In Zoom Out Reset image size

Bresolin et al. (2012) and Werk et al. (2011) discuss scenarios that could enrich outer disks, including gas accretion and radial metal mixing through either turbulent processes or galactic outflows and re-accretion. They are unable to conclusively determine the mechanism for enriching outer disks and comment that a combination of processes could be at work. In the particular case of NGC 4625, it is plausible that it has exchanged some gas in an interaction with NGC 4618, increasing the size of the outer disk and enriching it while simultaneously initiating star formation. However, it is difficult to explain how a large portion of enriched gas could be exchanged without the corresponding exchange of evolved stars. NGC 4618 also has several large holes in its inner disk H i distribution, indicating past supernovae activity (Kaczmarek & Wilcots 2012). If the two galaxies share a common halo, it is possible supernovae in NGC 4618 have enriched the halo's intergalactic medium. Simulations have also shown that late accretion of halo gas can appreciably increase the gas in an outer disk at late times (Kereš & Hernquist 2009), but this is likely to be extremely low metallicity.