THE OPTICAL STRUCTURE OF THE STARBURST GALAXY M82. II. NEBULAR PROPERTIES OF THE DISK AND INNER WIND^*

In 2006 February and 2007 February observations using the Gemini-North Multi-Object Spectrograph (GMOS) Integral Field Unit (IFU; Allington-Smith et al. 2002) were obtained covering five regions near the center of M82 (programme IDs: GN-2006A-Q-38, and GN-2007A-Q-21, PI: L.J. Smith), with 0.3–0.8 arcsec seeing (5–14 pc at the distance of M82; see indicator on Figure 2). The IFU in two-slit mode gives a field-of-view of 7 × 5 arcsecs (which corresponds to approximately 50 × 35 pc at the distance of M82) sampled by 1000 hexagonal contiguous fibers of 02 diameter. A separate block of sky fibers, offset by 1 arcmin from the object field, provides simultaneous sky observations. We took two dithered exposures at each position with integration times between 1200 and 2400 s each, and used the R831 grating to give a spectral coverage of 6100–6790 Å with a dispersion of 0.34 Å pixel⁻¹. A spectral resolution of R ≃ 5200 was measured from single Gaussian fits to isolated arc lines. The locations of each field are shown on an HST/ACS F658N image in Figure 1.

Zoom In Zoom Out Reset image size

Basic data reduction was performed following the standard Gemini reduction pipeline (implemented in iraf⁷). After reduction and combining the dithered exposures the five final data cubes each contained 825 wavelength and flux calibrated, sky-subtracted spectra.

DensePak (Barden et al. 1998) was a small fiber-fed integral field array attached to the Nasmyth focus of the WIYN⁸ 3.5 m telescope. It had 91 3'' fibers arranged in an array of 30 × 45 Four additional fibers, offset by ∼60'' from the array center, served as dedicated sky fibers.

On 2001 14th April, we used DensePak to observe the inner disk of M82. The 860 line mm⁻¹ grating was used in two settings; four fields were observed in the wavelength range of 5820–6755 Å (dispersion 0.46 Å pixel⁻¹; R ≃ 6700), three of these positions were also observed in the wavelength range of 7745–9700 Å (dispersion 0.96 Å pixel⁻¹; R ≃ 3650). These settings allowed us to cover many of the brightest optical and near-IR (NIR) nebular emission and stellar absorption lines. A number of bias frames, flat fields and arc calibration exposures were also taken together with the science frames. Basic reduction was achieved using standard iraf packages. After reduction, the datafiles contained 86 wavelength-calibrated and sky-subtracted spectra. Figure 1 shows the position of the DensePak fields overlaid on the HST F658N image.

Following the methodology employed by Westmoquette et al. (2007a, 2007b), we fitted multiple Gaussian profile models to each emission line using an idl χ²-fitting package called pan (Peak ANalysis; Dimeo 2005), to quantify the gas properties observed in each IFU field.

Hα is detected in every fiber of both the GMOS and DensePak data sets. The high S/N and spectral resolution of our data allowed us to quantify the line profile shapes to a high degree of accuracy. In general, we find the emission lines to be composed of a bright, narrow component (hereafter referred to as C1; this was sometimes split into two narrow components, hereafter C1 and C3) overlaid on a fainter, broad component (hereafter C2). In some regions of our DensePak data, we can also identify an Hα absorption component of stellar origin (hereafter C4). A full description of the procedure we employed to fit the lines and automatically determine the number of Gaussian models to fit is given in Paper I, together with estimates of the fit uncertainties. It is prudent to recall that for the purposes of consistency in the fitting and analysis, where a triple-Gaussian fit was chosen we specified that the broadest component should be assigned to component 2 (C2), and after that, the brightest to be component 1 (C1), and the fainter narrow component to component 3 (C3).

Given the limitations of current IFU observing techniques and multi-component line-fitting methods, a rigorous analysis of the uncertainties associated with the results presented in this paper is currently not feasible. First, the flux calibration of IFU data is not just subject to the usual kinds of uncertainties found in spectroscopic data, but also to additional complications due to the complexity and design of the instrument. Furthermore, the combination and interpolation of the dithered IFU maps introduces additional uncertainties. These factors together mean that we cannot rigorously quantify the uncertainties in the calibration of the individual spectra. Second, although a reduced χ² value for a multi-component line fit can be calculated, we currently do not fully understand how to quantify the uncertainties associated with each individual Gaussian component fit to the line profile. However, we can estimate errors from the observed variances in observed parameters in regions where conditions are relatively stable.

As discussed in Paper I, we estimate that, for the fluxes, there is a 0.5%–10% error in C1 (with the range resulting mostly from the variation in S/N levels), 8%–15% on C2 and 10%–80% on the C3 flux, with the level depending on the relative flux in C3. The presence of a third fit component may slightly increase the errors in the C1 and C2 fluxes particularly where C3 is faint. When these estimates are carried through, we might expect similar percentage errors for the line ratio and electron density measurements. Due to the aforementioned difficulties, we do not attempt to include error maps or show error bars for each figure, but where specific measurements are quoted in the text we have made an effort to indicate approximate uncertainties. Finally, we have converted the fluxes of the Hα absorption component (C4) to equivalent widths (EWs), and we estimate that the uncertainties on the EWs lie in the range of ∼0.5–1 Å.

In Paper I, we presented the line width and radial velocity results from the GMOS and DensePak data sets. In the sections following this, we will present and describe the remainder of our results, including the line flux and flux-derived properties of the emission line gas.

Figures 2–4 show maps of the Hα fluxes for the three-line components, respectively. For reference, we have overplotted contours from the HST/ACS F439W (∼B band; Mutchler et al. 2007) image in order to highlight the distribution of stellar emission and therefore the locations and extents of regions A and C. We have also indicated the maximum size of the seeing disk experienced during these observations (08) with filled circle. We begin by noting that C1 and C2 are detected over almost all of the five positions, and C3 is identified in approximately one third of all spaxels. We also note that the typical EW of the Hα emission line in position 1 is 50 Å, meaning that we can safely ignore the effects of underlying stellar absorption.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

To a first-order approximation, the flux morphology in the three Hα line components is fairly similar. However, there are important differences. First, we note that the region around the cluster M82-A1 (northern corner of position 1; Smith et al. 2006) does not stand out in any of the three component maps (although it is clearly seen in the F439W contours). Second, the C2 and C3 fluxes appear to be brighter on the southern side of any prominent feature compared to C1. For instance, in position 1 (region A) the flux of C1 peaks to the north and east of region A, whereas C2 peaks in the center to southwest of region A (its flux, however, is strongly influenced by the varying contribution from C3). Furthermore, the peak in C1 flux coincident with the large Hα knot in position 3 is located slightly to the north compared to the C2 peak. In position 2, this offset trend is again seen, with the peaks in C2 and C3 being offset slightly to the south compared to C1.

In order to assess the importance of C2 in terms of its contribution to the total Hα flux, we have mapped the the C2/C1 flux ratio for all five IFU positions in Figure 5. Surprisingly, of the spaxels that have both a C1 and C2 detection, ∼60% exhibit C2/C1 flux ratios greater than unity (i.e., on average C2 contributes more to the total line flux than C1; see also Table 1). Ratios >10 most likely result simply from bad fits. However, true ratios in the range 5–10 are seen in the wind positions 4 and 5 (Figure 4(f) of Paper I shows an example line profile with a C2/C1 flux ratio in this range), and there are a significant number of spaxels with C2/C1 ratios of ∼2 in position 3 and parts of the south-west of position 1. The majority of spaxels with ratios <1 are found in positions 1 and 2 (regions A and C), whereas those exhibiting the lowest ratios (0.1–0.3) are found in position 4 (coincident with the regions containing the broadest C2 emission, i.e., the position 4 knot—see Paper I). In summary, although the C2/C1 flux ratio varies significantly and rapidly over the five IFU fields, the general trend is of increasing C2/C1 with radius. This is discussed further and in context with the rest of our findings in Section 5.3. Finally, we note that finding such a large number of high (>1) C2/C1 flux ratios is in sharp contrast with what was found in NGC 1569 by Westmoquette et al. (2007a, 2007b). There, ratios fell in the range 0.1–0.3 with an average 0.12.

Zoom In Zoom Out Reset image size

Figures 6–8 show maps of the spatial variation of electron density, n_e, in each of the three-line components, derived from flux ratios between Gaussian fits to the [S ii]λλ6717,6731 lines (assuming an electron temperature, T_e = 10⁴ K).

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

The density distributions in the three-line components are distinctly different. We begin by discussing the densities in C1 (Figure 6). Within our limited spatial coverage, there are two peaks in n_e, one in the northern corner of position 1 and one in the northern corner of position 2. Position 1 covers the eastern half of region A; the peak in n_e at ∼2000–3000 cm⁻³ is coincident with the region around cluster M82-A1 and the well-known major axis dust lane that bisects the whole nuclear region (see O'Connell et al. 1995). From here, n_e falls off rapidly to the south and west, meaning that in the core and western half of region A (position 3) the densities are significantly lower. The peak in position 2 is again located to the north of the complex core (region C), and reaches ∼2000–3000 cm⁻³. Following the same trend, the densities then fall off toward the south, but remain fairly constant at ∼1000 cm⁻³ throughout the rest of region C. These values are consistent with previous studies (e.g., W07c). The densities in positions 3, 4, and 5 fall off consistently toward the southeast (along the minor axis direction), almost reaching the low-density limit (∼100 cm⁻³) at the edge of our coverage. Interestingly, the density morphology in both the wind positions (4 and 5) follows similar linear minor axis-oriented patterns as observed previously in the FWHM and radial velocity maps (Paper I). In position 5, we associated these with a discrete outflow channel, but finding these type of patterns also in position 4 hints at the presence of multiple similar channels throughout the inner wind.

In C2, n_e peaks at ∼3000 cm⁻³, in the southwest of position 1 as it overlaps with position 3 (i.e., the core of region A). This is distinctly different to what we see in C1: the western half of region A is much denser in C2, whereas the eastern half is considerably less dense. This type of result immediately highlights the advantage of spatially resolved observations over long-slit data (e.g., W07c). A secondary peak in C2 densities of just over 1000 cm⁻³ is seen in the center of region C (position 2). Again the morphology of C2 densities in region C differs significantly to those in C1. The C2 densities in the wind positions (4 and 5) follow the same pattern as seen in C1 (falling off along the minor axis), but remain much higher for the same radii. The gas emitting C2 must therefore remain denser further out into the wind. It is worth noting here that the linear minor axis-oriented patterns are still visible, but appear to anti-correlate with those seen in C1. This is discussed further in Section 5.3.

The C3 density maps, tracing the density of the fainter component of any split lines, are noisy and very incomplete. There are, however, a number of points worth drawing attention to. C3 densities peak in the core of region C (at ≳2000 cm⁻³, but with large uncertainties), approximately coincident with the flux peak. The gas knot in the center of position 3 shows up clearly with densities between a few 100 and 1000 cm⁻³. And in positions 4 and 5, C3 is detected in a number of northwest–southeast strips with densities of a few 100 cm⁻³. This distribution, where the densities of the front and backsides of the expanding structures are clearly different, emphasizes that the structure of M82 is highly complex.

To summarize this section, the C1 densities do not follow the Hα emission morphology, but peak in the north of our spatial coverage, to the north of complexes A and C. This is coincident with the location of the major axis dust lane separating complexes A and C from complexes D and E (see Figure 1, Paper I). The C2 densities are much more consistent with the Hα flux distribution (i.e., the C2 densities peak within the SB complexes), but remain much higher further out into the wind compared to C1. Since C3 represents the weaker component of any split lines detected, it is interesting that its density distribution does not closely follow that of C1. Uncertainties on the C3 densities are, however, much higher than those for C1 or C2.

The forbidden/recombination line flux ratio of [S ii](λ6717 +  λ6731)/Hα can be used as an indicator of the number of ionizations per unit volume, and thus the ionization parameter, U (Veilleux & Osterbrock 1987; Dopita et al. 2000, 2006). U is nearly constant within a single Strömgren sphere, a property that appears to hold over large scales in M82. [S ii]/Hα is also particularly sensitive to shock ionization because relatively cool, high-density regions form behind shock fronts which emit strongly in [S ii] thus producing an enhancement in [S ii]/Hα (Dopita 1997; Oey et al. 2000). Maps of this ratio in the three-line components are shown in Figure 9 (left panels), where a single contour line at log[S ii]/Hα = −0.5 represents a fiducial threshold above which non-photoionized emission is thought to play a dominant role (Dopita & Sutherland 1995; Kewley et al. 2001; Calzetti et al. 2004). In environments such as this, the most likely excitation mechanism after photoionization is that of shocks.

Zoom In Zoom Out Reset image size

The ratio of [N ii]λ6583/Hα can also be used in a similar fashion to [S ii]/Hα, as a diagnostic of the gas excitation level and to search for the presence of shocks (Veilleux & Osterbrock 1987; Dopita & Sutherland 1995). In Figure 9 (right panels), we show maps of this ratio for all three components; here the contour level represents a fiducial non-photoionized threshold of log[N ii]/Hα = −0.1. These maps are qualitatively very similar to those of [S ii]/Hα, and comparing the two figures gives an idea of the significance of the variations and the uncertainties. To aid with this further, we have plotted the [S ii]/Hα and [N ii]/Hα points for each IFU position together in Figure 10, separated into three graphs by line component. The dashed lines represent the aforementioned non-photoionization thresholds.

Zoom In Zoom Out Reset image size

The gas excitation in C1 (Figure 9, top panels) remains very uniform throughout the majority of positions 1, 2, and 3 (regions A and C), with mean values of log([S ii]/Hα) = −0.96 ± 0.22, and log([N ii]λ6583/Hα) = −0.28 ± 0.23. The errors quoted are the standard deviations of the samples, and their similarity shows just how constant these two ratios remain over these two positions. These results, in terms of both the absolute values and constancy, are in good agreement with previous studies (Shopbell & Bland-Hawthorn 1998; Förster Schreiber et al. 2001; Smith et al. 2006; W07c).

The C2 [S ii]/Hα and [N ii]/Hα line ratios are both consistently lower than the average in two regions: complex A and the western half of complex C. For the most part, these low excitation regions are coincident with the corresponding C2 density peaks (Figure 7), suggesting that local changes in the ionization parameter play a strong part in setting the line ratios.

In components C1 and C2, high (non-photoionized) line ratios (both in [S ii]/Hα and [N ii]/Hα) are predominantly found in positions 4 and 5. This may indicate that, as the distance from the starburst complexes increases, shock excitation begins to play an increasing role in ionizing the gas. The phenomenon of increasing line ratios with radius has been found in many starbursting galaxies on both the small and global scales (Heckman et al. 1995; Martin 1997; Calzetti et al. 1999, 2004). Where the line ratios are within the photoionized regime, the radial gradient can be explained as a dilution of the radiation field from the centralized source, but in many cases shock excitation is required to explain the ratios observed that are outside the reach of pure photoionization.

Notably, one of the regions of high-line ratios seen in the GMOS data, identifiable in both the C2 maps, is coincident with the "position 4 knot" identified in Paper I, and discussed further in Section 5.4. The C1 [N ii]/Hα map also exhibits very low ratios (log([N ii]/Hα) < −0.6) in position 4 that are not coincident with any obvious features in the corresponding density map (Figure 6). In C3, high-line ratios are found to the center–east of position 3 and to the north and east of the region C flux peak in position 2.

In summary, the unprecedented spatial resolution of these line ratio maps shows that (1) the excitation of the gas within the starburst complexes is extremely uniform (and consistent with pure photoionization), despite the dramatic density gradients; and (2) regions of both high- and low-excitation gas are found within the inner wind flow, but only in very localized, distinct pockets. In C1, these regions do not correspond to anything seen in any of the other maps presented here or in Paper I, but the low-excitation regions in C2 are coincident with the density peaks.

Figure 11 shows the distribution of Hα flux in components 1, 2, and 3, and the EW (in Å) of component 4, together with contours representing the WIYN R-band image presented in Paper I. As expected, the flux map of C1 (the bright narrow component) closely follows what is found through Hα narrowband imaging: the bright nuclear starburst complexes stand out in position 1, whereas region B to the east of the disk (position 3) is conspicuously faint in Hα. The broad component, C2, is detected over parts of all the positions, but is brightest in regions A and C and only detected in a few spaxels in region B. Over all the regions where C2 is detected it is, on average, of equal brightness to C1 (see Table 1). However, the C2/C1 flux ratio is not constant across the face of the disk, as can be seen in Figure 12, where we plot a map of the C2/C1 flux ratio. With this logarithmic color stretch, it is clear that the ratio begins at low values near the disk midplane (0.3–0.5) and increases with height above and below the disk (up to ∼10).

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Component 3 (C3), the fainter narrow component, is detected over parts of the whole central starburst area, extending out to the north of the disk (position 4) and region B (position 3). As in the GMOS data, its intensity is on average ∼40% that of C1 (Table 1). C4, the absorption component, is observed over the whole of region B (in position 4 and a part of position 3) and to the southwest of cluster F in position 2. An absorption component is expected in these regions since they are located outside the current starburst (Smith et al. 2007) and hence contain an older stellar population dominated by A-type stars (de Grijs et al. 2001; Förster Schreiber et al. 2003; Konstantopoulos et al. 2008). The Hα flux absorbed is at a maximum in region B, where EWs of ∼3–4 Å are seen. An example line profile from region B is shown in Figure 5(d) of Paper I, clearly illustrating a broad and well-fitted absorption component.

Maps of the electron densities, n_e, measured from the [S ii] ratio (assuming T_e = 10⁴) are shown in Figure 13. Note that the velocities and FWHMs of the three [S ii] components are consistent with those of Hα (Paper I), meaning that the components identified in both lines physically correspond to one another, despite the lower S/N of the [S ii] line. In C1, the highest densities are found in the eastern half of region A, peaking at ∼1000 cm⁻³. The location of this peak agrees well with our GMOS data; its somewhat lower value can be attributed to the difference in spatial resolutions between the two data sets (i.e., the much finer spatial resolution of GMOS allows us to pick up the small-scale, localized regions of higher density gas). The peak density in region C is lower (at 500–700 cm⁻³; in good agreement with W07c), and densities remain at a level of a few 100 throughout the central starburst region, particularly on the northern side. Another peak is found in region B, with densities reaching 500–700 cm⁻³.

Zoom In Zoom Out Reset image size

In C2, densities rise to ≳1000 cm⁻³ in the spaxels covering the western half of region A and region E, in excellent locational agreement with our GMOS data (again the lower absolute values can be attributed to the coarser spatial sampling of DensePak). Densities remain higher to much larger radii in C2, both in the major and minor axis directions. In order to highlight this further, we have plotted the radial distribution (relative to the nucleus) of both C1 and C2 electron densities in the lower-right-hand panel of Figure 13. This clearly shows that the C2 densities remain much higher out to larger radii relative to C1. This intriguing finding extends the trend we found in the GMOS data to the entire disk, and is discussed in relation to the origin of C2 in Section 5.

O'Connell & Mangano (1978) were the first to recognize that ionized gas densities in the starburst core of M82 were unusually high, finding [S ii]-derived electron densities of ∼1800 cm⁻³ in regions A and C, and subsequent multi-angle long-slit observations showed that the gas densities rapidly fall off radially (Heckman et al. 1990). The minor axis decrease was accurately measured by McKeith et al. (1995) using deeper long-slit data. They found electron densities of ≳1000 cm⁻³ near the nucleus (where they only detect their strongest line component), and by ±500 pc the densities had fallen to ∼100–200 cm⁻³. W07c presented HST/STIS spectroscopy of the starburst core and the first evidence for how rapidly the densities vary with position: over only 2.5 arcsecs (45 pc) they found the densities to fall from ∼1800 cm⁻³ (near cluster A1; see also Smith et al. 2006) to ∼1000 cm⁻³ in the center of region A and 600–800 cm⁻³ in region C.

Our DensePak measurements thus agree very well with these studies. The coarse spatial resolution of DensePak explains why we do not see the highest densities mentioned in the literature and seen with our GMOS data. Having said that, our DensePak observations have the advantage of both wider two dimensional spatial coverage and high S/N, meaning we have been able track the density of both C1 and C2 over a significant proportion of the inner disk. Finally, overall the densities of both C1 and C2 stay higher on the northern side of the disk compared to that of the southern side. This can be understood as being a result of the 80° inclination of the disk, such that the northern side is nearest to the observer (see Figure 13 of W07c).

The main finding of this section is that although the C1 and C2 densities peak at approximately the same value in the center of the starburst, the C2 densities remain higher to much larger projected radii (i.e., both in the major and minor axis directions) than their C1 equivalents.

Figure 14 shows maps of the [S ii](λ6717+λ6731)/Hα and [N ii]λ6583/Hα line ratios. Since not all line components were detected in each line, we have summed the flux of all identified components for each line before calculating the line ratios.

Zoom In Zoom Out Reset image size

A significant range in values are seen in both maps, with elevated ratios in region B (position 3 and the northeast of position 4) and to the southwest of cluster F (position 2). However, a closer inspection of the spectra from these regions reveals that the majority of spaxels with these high ratios exhibit an Hα absorption component—we have marked spaxels where an absorption component is detected with a cross in order to highlight the correspondence. Since absorption is only present in recombination lines, the forbidden/recombination line ratios become biased toward artificially high values when the absorption is present. This highlights the need to be careful of absorption effects in line ratios, particularly for ratios derived from images, or simple single Gaussian fitting, where absorption cannot be accounted for.

Disregarding these Hα-absorption affected regions, the two line ratios remain fairly uniform (particularly [N ii]/Hα). The mean ratio of log([N ii]/Hα) for all spaxels not containing absorption is −0.3 ± 0.1, and similarly for log([S ii]/Hα) is −0.8 ± 0.3. These values are in very good agreement with the GMOS data (Section 3.3).

O'Connell & Mangano (1978) find a log([N ii]/Hα) ratio for the nuclear regions (including complexes A, C, and E) of −0.28 from long-slit data. Shopbell & Bland-Hawthorn (1998) find uniform and low values of log([N ii]/Hα) in the inner 1 kpc of the M82 outflow ranging from −0.5 to −0.2. Both these results are consistent with our measurements, and in very good agreement with W07c. The line-ratio maps of Shopbell & Bland-Hawthorn (1998) also show very high values further out in the disk (to the east and west of the nucleus), in the same place as we find exaggerated values due to Hα absorption. At minor axis heights of 1 kpc or more (>60''), Shopbell & Bland-Hawthorn (1998) find increasing values of [N ii]/Hα which they attribute to a dilution of the radiation field in the outer wind together with a possible increasing influence of shocks. However, our data do not reach this far in radial extent so we do not see this effect.

The electron densities we have mapped peak in the starburst core, fall off dramatically as we leave the disk, and exhibit significant local variations. The peak values (few 1000 cm⁻³) imply very large interstellar pressures (P/k > 10⁷). Although these results have been found before by a number of independent studies (e.g., Heckman et al. 1990; McKeith et al. 1995; W07c), our high-resolution, two-dimensional maps show that the peaks are very localized and that there is a large amount of structure within the density distribution in the starburst core.

Contrary to expectation, the C1 densities peak not in the starburst complexes themselves, but to the north, in the major axis dust lane. Recalling one of our conclusions from Paper I—that the emission of C1 is weighted toward the inner-most regions of the starburst (i.e., the base of the outflow)—we see that these density results support this idea, and indicate that the body of bright, relatively quiescent gas from which C1 is emitted, is not associated directly with the main starburst complexes themselves, but extends well beyond the complexes into the dense dust lane. C2, however, is at its densest within complex A (and to a lesser extent C); its overall distribution follows the flux morphology much more closely than C1. Its peaks are also of a few 1000 cm⁻³ implying large pressures. These results are again in support of our conclusions from Paper I, where we suggested that the C2 emission is weighted toward regions further out along the line of sight.

In this paper, we have also presented evidence at high spatial resolution for the impressive uniformity of the line ratios and excitations within complexes A and C (particularly in component C1), first highlighted by studies such as Förster Schreiber et al. (2001) and Beirão et al. (2008). We have also confirmed how this uniformity can be extended to the whole inner disk region covered by the DensePak fields of view. The unprecedented spatial resolution and spatial coverage of the line-ratio maps shows that (1) the excitation of the gas within the starburst complexes is extremely uniform (and consistent with pure photoionization), despite the dramatic density gradients; and (2) regions of both high- and low-excitation gas are found within the inner wind flow, but only in very localized, distinct pockets.

We will now use these density and excitation results, together with work from the literature, to build up a model to describe the ISM conditions within the starburst core. As described in the introduction, the modeling of observations of gas in the central regions made from the optical radio indicate a fragmented (shredded/fractious) ISM, consisting of a range of clouds from small/dense clumps with low-filling factors (<1 pc, n_e ≳ 10⁴ cm⁻³) to larger filling factor, less dense gas (e.g., Lord et al. 1996; Förster Schreiber et al. 2001; Fuente et al. 2008). These are bathed with an intense radiation field and embedded in an extensive high temperature (T ≳ 10⁶ K), X-ray-emitting ISM. Emission measure results from W07c suggest that the most compact clouds are found in the clump cores where the star formation is most intense and the gas pressures are high, whereas the cloud sizes in the inter-complex regions are larger. The observed near-constant state of the starburst (ionization parameter and line diagnostics) must therefore be a consequence of the small cloud sizes, allowing the gas conditions to respond quickly to any changes. This is rather akin to the thermalisation of a large number of individual particles in a plasma. To this picture, we are now able to add that the denser clouds are not only bathed in UV but are also interacting with an extensive, very hot, and presumably turbulent ISM phase that is responsible for the X-ray emission. The left-center panel of Figure 16 contains a schematic representation illustrating these ideas. Since we associate the broad component C2 with emission from turbulent mixing layers (TMLs) on the surface of interstellar clouds, the existence of so many small, compact, high-density clumps embedded in a hot, fast-flowing wind must make for an ideal environment for the copious emission of broad lines. These surface wind-clump interaction layers must also be where mass is evaporated and stripped off the clouds by the intense radiation field and wind flows. This suggests that the majority of the mass loaded into the wind must originate in these clumps, where the density of the gas being stripped is highest. This argues strongly against C2 originating from scattered light from the nucleus. A further implication of the existence of so many dense clouds in complexes A and C is that there must still be a significant gas supply within the starburst core, despite the intense star formation, heavy mass loading and wind outflows.

In our model, C2 is only emitted from TMLs on cloud surfaces in direct line of sight with a wind source, whereas C1 can be emitted from any 10⁴ K gas located within the Strömgren radius of the starburst. This includes the static ionized layers beneath the TMLs (bottom panel of Figure 16), ionized gas clouds which are not in a direct line of sight with an ionizing/wind source (i.e., wind shadowed—ionizing radiation is easily scattered—an example of a cloud in a wind shadow is illustrated in the bottom-left corner of the center-left panel in Figure 16), and general diffuse ionized gas. Furthermore, the excitation of the gas within the starburst complexes is extremely uniform, and consistent with pure photoionization. This can be explained through the well-mixed nature of the clusters and clouds, which transforms the presumably very inhomogeneous distribution of ionizing photons from the clusters into a uniform escaping radiation field. A helpful analogy might be to liken the surfaces of the starburst complexes to the surface of the Sun, where there are many gas phases (temperatures and densities), a strong density gradient, chaotic dynamics and morphology, and denser pockets of gas in the form of loops and filaments (shredded shells/bubbles) entrained in the outflow/wind, all being accelerated outward.

As a final note to this section, we turn our attention to the ∼7 Myr old super star cluster M82-A1 (Smith et al. 2006). In that study we found M82-A1 to have followed a non-standard evolutionary path for a star cluster, since it is surrounded by a large, pressurized H ii nebula that is stifling its wind (see also Silich et al. 2007). If the H ii shell forms a full sphere around the cluster and this stage is not simply transient, then it is difficult to explain how the superwind is powered if one assumes M82-A1 to be representative of a typical cluster in region A. However, our GMOS maps show that M82-A1 is in fact embedded in one of the highest C1 density peaks in the starburst core. Recalling that the inclination of the disk is such that on the southern side we are seeing up into the starburst (the northern side is partially obscured by the front of the disk; see e.g. W07c), we can now see that the rapid fall in densities from the north of our IFU fields into the cores of complexes A and C shows that the conditions toward the south of the complexes are much more favorable to the escape of the individual cluster winds. This implies that clusters toward the outside surfaces of the complexes may in fact be the primary drivers of the superwind.

As we move away from the plane of the disk, there are a number of results that indicate that emission from C2 becomes increasingly more important (compared to C1). First, the C2/C1 flux ratio increases (all the way through our GMOS fields out to the extent of our DensePak data; Sections 3.1 and 4.1). Since the C2/C1 ratio reflects the proportion of the volume occupied by TML versus standard photoionized gas, this indicates that within the energy injection zone (thought to be defined by the minor axis velocity inflection points; McKeith et al. 1995; Shopbell & Bland-Hawthorn 1998; W07c) the turbulent motions must play an increasing role with radius. Second, the electron densities in C2 remain higher to larger radii (Sections 3.2 and 4.2), suggesting that away from the dense core of the galaxy, then density in the TMLs is preferentially higher than in the narrow-line-emitting gas. Although the majority of the mass loading might take place within the starburst complexes, the region of mass loading must extend far out into the inner wind. Thirdly, the line widths of C2, although already high in the nuclear regions, also increase with height above the disk (Paper I), peaking at the location of the McKeith et al. (1995) minor axis velocity inflection points. Finally, it is possible that these effects are also partly due to the weakening or dilution of the C1 emission with height. What happens to these relationships beyond the energy injection zone (beyond the minor axis velocity inflection points) is something that should be investigated.

On average, the C2/C1 flux ratios are considerably higher than what was found in NGC 1569 (Westmoquette et al. 2007b). Since C2 is produced at the interface layers of interstellar clouds, the significantly larger gas mass and number of clusters in M82 has resulted in C2 being more prominent and may mean that a larger percentage of mass is being loaded into the outflow.

In Paper I, we identified a discrete outflow channel in IFU position 5 characterized by expanding, narrow-line-emitting walls surrounding a cavity of broad-line-emitting gas. With the additional results that we have presented here, we can explore the characteristics and implications of this outflow channel further.

Figure 15 shows Hα position–velocity plots extracted from seven evenly spaced (every 2'') pseudo-slits positioned parallel to the major axis, stepping down through GMOS IFU position 5. Also plotted are the C1 and C2 electron densities extracted from the same pseudo-slits (the C3 densities have been omitted for clarity, but are consistent with those of C1; see Figures 6 and 8). A coherent, expanding cylinder, as indicated by the overlaid dashed ellipses, can be traced in the narrow-line (C1 and C3) kinematics extending along the whole length of position 5. These velocity ellipses show that the structure, while >7'' (120 pc) in length, increases slightly in physical width from the north to south from ∼2''–3'' (35–50 pc), and maintains an approximately constant (subsonic) expansion velocity of ∼60 km s⁻¹. This is indicated in a schematic cartoon of the outflow channel in the right-hand panel of Figure 16. The velocities of the broad C2 line suggest that it originates within the space between the two walls, with central velocities at rest with respect to the C1–C3 expansion. The [S ii]-derived electron density plots demonstrate how the walls of the channel are defined by peaks in the C2 densities (700–1000 cm⁻³) and troughs in the C1 densities (∼200 cm⁻³), whereas the center of the channel is defined by higher C1 densities (≳500 cm⁻³) that decrease with radius. What is not shown in this plot is that the widths of C2 appear to be broader (∼250 km s⁻¹) near the channel walls and narrower (≳150 km s⁻¹) in the center (Paper I).

Zoom In Zoom Out Reset image size

Thus, in our model the channel cavity contains the hot outflowing wind fluid. As it flows outward it interacts with the cooler, denser walls of the channel and with clumps of entrained material to produce broad-line emission, while the walls themselves emit primarily the narrow lines C1 and C3. Finally, we note that the linear minor axis-oriented patterns observed in the density, FWHM, and radial velocity maps (Section 3.2; Paper I) hint at the existence of multiple similar channels within the southern outflow, as is predicted by models such as those of Tenorio-Tagle et al. (2003).

The presence of organized channels which funnel shock-heated gas outward from starburst complexes has a number of important implications. These structures resemble scaled-up versions of the cluster wind channels described by Tenorio-Tagle et al. (2003), and as in their models, will tend to collimate hot gas outflows. The collimated outflows are a possible source for the well defined loop structures of ionized gas seen in Hα emission above the main starburst complexes in M82. Indeed hints of the channel can be seen in the Hα morphology (see e.g., Figure 1 in Paper I) in the form of enhanced emission following the channel walls. Without the line ratio and kinematical evidence, however, it would have been impossible to distinguish from any other formation in this chaotic region. Channeled flows will also influence the mass loading process, and it seems likely that some of the gas within such columns could experience minimal mass loading. In this case, they will improve the efficiency in launching a galactic wind in that some hot gas can flow out of the disk relatively unimpeded. In addition, outflow channels can provide low opacity pathways for the escape of UV radiation from starburst complexes which in turn can help ionize the large quantities of extraplanar gas associated with this galaxy.

In Paper I, we identified a small region in GMOS IFU position 4, coincident with a bright, point-like object on HST/ACS imaging, with distinctly blueshifted (by ∼140 km s⁻¹) and broad (≲350 km s⁻¹) Hα emission that we referred to as the "position 4 knot." From the data presented in this paper, we see that this knot does not show up in either the flux or density maps. However, a small number of spaxels near the location of the knot exhibit a slight enhancement in [S ii]/Hα C2 and a significant enhancement in [N ii]/Hα C2 compared to the surroundings (Figure 9). To confirm these results, we extracted and summed the spectra from 10 spaxels centered on the knot; the resulting spectra and multi-Gaussian line fits are shown in Figure 17. The broad (FWHM ∼ 330 km s⁻¹), blueshifted (−140 km s⁻¹) component is clearly identifiable in all five emission lines. The line component properties are given in Table 2. The electron densities derived from the [S ii] line ratios indicate that the gas from which the broad component is being emitted is denser than the narrow-line gas (∼1000 cm⁻³ compared to ∼500–900 cm⁻³). The line-ratio enhancements described above do not show up in the summed spectrum since they have been diluted by the surrounding spaxels.

Inspection of archived HST/ACS High Resolution Channel (HRC) images (Prop ID: 10609, PI: Vacca) reveals one other similar nearby point-like source (hereafter source 2), the location of which, together with the position 4 knot, is identified on a F330W (∼U), F435W (∼B), F550M (∼V), and F814W (∼I) HRC color composite in Figure 18. Measurement of their sizes (using ishape; Larsen 2004) show that they are both unresolved (FWHM ≲ 006) in each of the ACS/HRC images, including F658N (∼Hα). For both objects, we find no evidence for variability (>0.2 mag) in any of the aforementioned filters over the epoch range 1995 September to 2006 March (the range covered by all the HST WFPC2 and ACS images available in the archive). Absolute F555W magnitudes (M_V) and colors are given in Table 3. No evidence of the object is found in any of the available HST/NICMOS NIC1/2 images (F160W, F164N, F187N, F190N).

Table 3. Point Source Photometry, where U = F330W, B = F435W, V = F555W, I = F814W

Object	Coordinates		V	MV	U − V	B − V	V − I	B − I	U − I	U − B
	(J2000)		(mags)
1 (pos4 knot)	09h55m5406	69°40'4435	22.16	−5.22	1.14	0.98	1.40	2.38	2.45	0.16
2	09h55m5303	69°40'3960	20.46	−6.92	0.68	0.62	0.79	1.41	1.47	0.06

Download table as:  ASCIITypeset image

Considering our spectroscopic and photometric evidence and their unresolved nature in the HST imaging, there are a number of plausible explanations for these sources. We can immediately dismiss the position 4 knot as being a background object based on its radial velocity, it only having a difference of ∼140 km s⁻¹ relative to the systemic velocity of M82. Furthermore, since it does not exhibit an enhanced density compared to its surroundings (Figure 7), does not show up in any of our IFU line or continuum flux maps, and does not correlate with any known radio source at this location (Huang et al. 1994; Wills et al. 1997; McDonald et al. 2002), it is unlikely to be a supernova remnant.

We find that the most likely explanation is that the position 4 knot is a hot stellar object immersed in an unresolved ionized circumstellar nebula. The FWHM of the broad component then represents the expansion of the nebula of ∼165 km s⁻¹. We derive an absolute magnitude of M_V ∼ −5.2 which increases to ∼−7 if we assume a V-band extinction of ∼2 mags, suggesting a massive star. The photometry shows that the star is fairly red but this may be due to a dusty circumstellar nebula. The high [N ii]/Hα ratio suggests that the nebula may be enriched in nitrogen. All these findings lead us to speculate that the position 4 knot may be a luminous blue variable (LBV) star (Humphreys & Davidson 1994). These objects are usually surrounded by nitrogen-rich circumstellar material which has been ejected as a result of an unstable event in the past. These stars can spend many years in a hot, quiescent phase and have similar absolute magnitudes to the position 4 knot. The large radial velocity of the emission from this object is intriguing; it might indicate that the star was ejected from the cluster in which it was formed, thus explaining how such a star is found so far from the main star-forming complexes. Measuring a projected distance of the star from the nearest star-forming region is ∼5''–6'' (∼90–100 pc), and even assuming a conservative estimate of the age of the LBV (a 60 M_☉ mass LBV would be ≲4 Myr old), the predicted velocity is quite plausible (<50 km s⁻¹).

Source 2, for which we only have photometry measurements, is brighter and redder than the position 4 knot (Table 3). Although it is also unresolved on the HST/ACS HRC images and also does not correlate with any known radio sources. Spectroscopic measurements, well within the capabilities of 8–10 m class telescopes, are needed to ascertain the true nature of these objects.