A SURVEY OF DIFFUSE INTERSTELLAR BANDS IN THE ANDROMEDA GALAXY: OPTICAL SPECTROSCOPY OF M31 OB STARS

Previous analyses of luminous B-type supergiants in M31 found abundances comparable to those in the solar neighborhood (e.g., Trundle et al. 2002). The GMOS spectra were therefore classified with respect to Galactic standards (Walborn & Fitzpatrick 1990; Lennon et al. 1992). Spectral types and stellar radial velocities are presented in Table 1; those stars with previous spectral classifications are listed for comparison with our new classifications in Table 2. Figure 2 shows a selection of the blue GMOS spectra, illustrating the spectral sequence of the B-type supergiants.

Zoom In Zoom Out Reset image size

Reported stellar radial velocities are the mean measurements of the line centers of selected H, He, and metal absorption lines. The velocity for 4029.71 is a notable outlier of the Field 2 results and the spectrum shows some evidence for a potential second component (e.g., He ii λ4200, although in the opposite velocity sense to what might be expected)—it seems likely this is a composite spectrum of more than one star.

Reddenings (E_B−V) were calculated using the intrinsic colors of Wegner (1994), supplemented by Johnson (1966) for later spectral types. The foreground reddening is discussed in Section 3.4.

Observational details of the 34 program stars are summarized in Table 1. We adopt abbreviated target names based on the minutes and seconds of right ascension in the LGGS catalog. Thus, stars belonging to each of the three fields are distinguished by the first two digits of their abbreviated names (39 for Field 1, 40 for Field 2, and 44 for Field 3).

Of the stars observed, two are late-type foreground stars (4405.89 and 4409.71) and one (3945.82) is a Wolf–Rayet in M31, and therefore unsuitable for our DIB survey. To investigate the DIBs in M31, only the sightlines passing through a significant quantity of ISM were selected for further analysis. Each spectrum was examined for the presence of DIBs, by reference to the survey of Hobbs et al. (2008). No DIBs were detected toward stars with a foreground-corrected reddening of E^M 31_B−V<0.05 mag. This is as expected given the spectral noise, and the assumption that M31 DIB strengths are similar to those typically observed in the MW: for the typical S/N of the spectra near 5780 and 6283 Å (100 and 80, respectively), using the approximate minimum DIB W detection threshold equation of Hobbs et al. (2008; W_max = 1.064× FWHM/(S/N)) yields respective lower detection limits of 22 and 63 mÅ for these DIBs. Scaled to a reddening of E_B−V = 0.04, the mean Galactic λλ5780 and 6283 DIB W/E_B−V values are 21 and 47 mÅ, respectively (using data from Herbig 1993, Thorburn et al. 2003, Megier et al. 2005, and Cordiner 2006), which is consistent with their non-detection in our spectra at M31 reddenings of 0.04 and less. The sample of 14 stars with E^M 31_B−V⩾0.05 is listed in Table 3, with the addition of three targets (3956.94, 4032.28, and 4406.87), which were used as telluric standards for Fields 1, 2, and 3, respectively.

Sky-subtracted spectra of the regions covering the sodium D lines are shown in Figure 3. At the resolution of the GMOS spectra (100 km s⁻¹ in this wavelength region), the interstellar Na D absorption lines in M31 and the MW can be modeled using single Gaussian interstellar cloud components. Both lines of the Na D doublet were fitted simultaneously using the vapid routine (Howarth et al. 2002) to produce models of the Galactic and M31 Na i distributions in each sightline. Fitting both lines of the sodium doublet simultaneously helps to constrain the Doppler b parameter in data at this relatively low spectral resolution. The resulting interstellar Na i cloud radial velocities, Doppler b parameters, and column densities are summarized in Table 3. Monte Carlo error estimates were generated by refitting each cloud model to 100 different replicated spectra, each with random Poisson noise added, and taking the ±68th percentiles of the resulting (free) parameter ranges.

Zoom In Zoom Out Reset image size

The Doppler shift of Field 3 is comparable to the instrumental resolving power, and as a result, the MW and M31 absorption components are blended which hinders the accurate derivation of the individual component parameters. To obtain physically realistic results for 4409.52 and 4412.17, the velocity and Doppler b parameter of the MW component had to be held fixed during the fitting (at v = −23.3 km s⁻¹ and b = 19.9 km s⁻¹, which are the averages for the MW components toward the other stars in Table 3). Toward the telluric standard 4406.87, there was no detectable M31 component of Na i; the only absorption detected is at a mean radial velocity of −32.0^+5.2_−4.2 km s⁻¹ which is unlikely to be in M31 and is consistent (in velocity and column density) with the other MW foreground components observed toward the stars in Fields 1 and 2. An upper limit on the M31 Na i column density of 2.8 × 10¹¹ cm⁻² was derived for this sightline assuming the radial velocity and Doppler width of the gas to be the average of those values measured toward the other stars in Field 3. It must be noted, however, that all of the Field 3 Na i fit parameters are subject to additional uncertainty due to the likely contamination of the spectra by sky-subtraction residuals. Strong telluric Na i emission features dominate these GMOS spectra at velocities ±100 km s⁻¹ with respect to the D-line centers, and their imperfect subtraction during data reduction thus leads to potential uncertainties in the spectra over this velocity range.

The observed sightlines pass through the halo of M31 and are expected to intercept multiple interstellar clouds with component structure that is unresolved in our spectra. Thus, the derived (single-component) interstellar cloud models provide no information about the detailed kinematics of the M31 ISM except for the overall velocity width and peak radial velocity of the Na i gas distributions. The column densities derived from these broad, single-component fits should generally be considered as lower limits due to the likely presence of unresolved saturated structure in the Na D lines.

Since sodium lines can be present in the background stellar spectra, we investigated the possible stellar Na D contributions to our observed spectra. Theoretical spectra for a range of stellar types spanning those observed were calculated using synspec (Hubeny & Lanz 2000) and Kurucz (1993) model atmospheres with solar metallicity. Stellar contributions to the Na D line equivalent widths were found to be negligible compared with the interstellar component for all the stars listed in Table 3; only in cooler stars (of type A and later) does the stellar contribution become important.

The measured M31 interstellar sodium radial velocities match the velocities of nearby M31 stars and planetary nebulae, as can be seen in Figure 4. This demonstrates with little ambiguity that these clouds are located in the main disk of M31.

Zoom In Zoom Out Reset image size

The mean foreground Na i column density toward our surveyed M31 stars is 3.2 × 10¹² cm⁻². Sembach et al. (1993) measured the Na i column density in lines of sight through the Galactic halo, probing gas that should be similar to the MW foreground ISM in our sightlines. The correlation between Na i and E_B−V permits an estimate of the foreground reddening. Performing a least-squares fit to the Sembach et al. (1993) data yields a linear relationship: ${E_{B-V}}=(N({\rm Na \textsc {i}})+1.41\times 10^{11})/3.79\times 10^{13}$, from which we derive a foreground reddening of E_B−V = 0.088 mag. However, there is considerable scatter in the Na i versus E_B−V data—the mean rms error on E_B−V is 0.053—and there is likely to be contamination of our Na i profiles near v = 0 due to sky-line subtraction residuals. These factors make the derivation of foreground reddening from Na i quite uncertain. From the Leiden/Argentine/Bonn (LAB) H i survey¹¹ (Kalberla et al. 2005; Hartmann & Burton 1997), the spectra obtained at the points nearest the centers of our three fields give foreground H i column densities of 5.75 × 10²⁰ cm⁻² (Field 1), 6.07 × 10²⁰ cm⁻² (Field 2), and 6.68 × 10²⁰ cm⁻² (Field 3), assuming optically thin emission. Using Equation (7) of Burstein & Heiles (1978) (which assumes a constant gas-to-dust ratio) to calculate the foreground reddening, the average value is E_B−V = 0.068 mag. The modified Burstein & Heiles relation utilized by the NED database¹² results in E_B−V = 0.0625 mag, which is in good agreement with the foreground reddening toward M31 estimated by Schlegel et al. (1998) from the median Galactic dust emission in the surrounding regions (E_B−V = 0.062 mag). We therefore adopt an MW foreground reddening of 0.06 ± 0.02, which is used in the calculation of the foreground-corrected reddenings (E^M 31_B−V) given in Table 1. Quoted errors on E^M 31_B−V include an uncertainty of 0.02 in the foreground reddening, plus the uncertainty on the intrinsic stellar B−V values (equal to the uncertainty of a single spectral subclass).

As shown in Figure 5, the λ5780 DIB is detected in several of the observed sightlines at a wavelength matching the Galactic λ5780 rest wavelength of 5780.37 Å (see, for example, Galazutdinov et al. 2000). This absorption feature most likely originates in the Galactic halo. The average foreground λ5780 equivalent width is 40 ± 10 mÅ, as measured from the co-added spectrum of all the stars from Fields 1 and 2 listed in Table 1 (excluding the Wolf–Rayet and those sightlines in Field 3 with blended MW and M31 Na i cloud components). Using a Gaussian fit, the mean radial velocity of the λ5780 carrier gas was found to be 0.5 ± 7.3 km s⁻¹. The mean Galactic λ5780 equivalent width per unit reddening is [W(5780)/E_B−V]_mean = 527 mÅ mag⁻¹, with a standard deviation of 230 mÅ mag⁻¹ (see Section 3.2 for details of the literature data used; the outliers toward ρ Oph and the Orion Nebula have been excluded). Using this ratio, the average MW foreground λ5780 equivalent width measured toward M31 corresponds to a foreground reddening of E_B−V = 0.08^+0.08_−0.04 mag, which is consistent with the value of 0.06 adopted in this study.

Zoom In Zoom Out Reset image size

We searched the spectra of each of the sightlines with foreground-corrected reddening E^M 31_B−V⩾0.05 for the DIBs listed in the survey of Hobbs et al. (2008). For the λλ6283 and 6993 DIBs, which are partially obscured by overlapping telluric lines, corrections were performed by division with spectra from lines of sight with negligible E^M 31_B−V selected from the same field (designated with superscript "b" in Table 1). The λλ4428, 5705, 5780, 5797, 6203, 6269, 6283, 6379, 6613, 6660, and 6993 DIBs were detected in at least one sightline each, with Doppler shifts matching the M31 Na i and H i velocities (H i data are presented in Section 3.6). The λ5849 and λ6445 DIBs were tentatively detected in the spectrum of 3944.71 only. The λλ4501, 4726, 6196, 6376, and 6532 DIBs were too weak to be detected in any of the spectra, which is consistent with the assumption of MW-like strengths (per unit reddening), given the S/N and low resolving power of our spectra. Severe contamination by telluric absorption prevented analysis of the λλ6886, 6919, 7224, and 7334 DIBs. Equivalent widths were measured by integration across a linear- or quadratic-fitted continuum. Errors on DIB equivalent widths measured this way were determined by σΔλ, where σ is the rms noise in the continuum and Δλ is the FWHM of each DIB. As a result of the relatively low S/N and low dispersion of the spectra compared with other contemporary Galactic DIB studies (e.g., McCall et al. 2010), additional uncertainties in the measured DIB equivalent widths arise due to the difficulties in establishing the true full widths of the DIBs over which to perform the integration. For example, in Figure 5 the presence of overlapping weak stellar/noise features at the wings of the λ5780 DIB observed toward 3944.71, 4034.80, and 4412.17 causes depressions in the continua that cannot be distinguished from the DIB by eye, and thus artificially boost the measured equivalent widths. Conversely, the broad, shallow, Lorentzian-like wings of the λ6283 DIB (shown in Figure 6) are lost in the noise, resulting in likely underestimation of the full width of this feature. Therefore, to provide a second, more accurate measurement of the λλ5780 and 6283 radial velocities and equivalent widths that takes the complex profiles of the DIBs into account, Galactic DIB template spectra were fitted to the observed M31 spectra.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

The Galactic DIB templates were derived from high-resolution (R = 58, 000), high S/N (∼2000) spectra of β¹ Sco (see Cordiner 2006), shifted to the interstellar rest frame. These were convolved with a Gaussian interstellar cloud model, then with the GMOS spectral point spread function (assumed Gaussian), and rebinned to the dispersion of the GMOS spectra. The interstellar cloud model was optimized using the same nonlinear least-squares algorithm employed by the vapid code (Howarth et al. 2002). This technique has the benefit that the statistical uncertainties in the measurements of the equivalent widths and velocities of weak DIBs are minimized, but has the drawback that the intrinsic M31 DIB profiles and rest wavelengths are assumed to closely match those of the MW DIB standard star β¹ Sco. If this were not the case, unknown systematic errors could occur, but there is no evidence to suggest that M31 DIB profiles differ from those of β¹ Sco. Profiles of some Galactic DIBs have been shown to exhibit variations in substructure (Galazutdinov et al. 2002). So far, however, all observed variations have been small and are negligible compared to the size of the GMOS resolution element and the level of noise in our spectra. The reader is referred to Section 3.5.2 of Cordiner (2006) for a more detailed discussion of the merits of the DIB fitting technique.

The fitted and integrated equivalent widths are shown in Table 5 and match each other to within the error bars. However, for four out of seven cases (for which the DIBs are not affected by obvious blends, denoted by superscript b), the fitted values of W(5780) are systematically less than the integrated values. For W(6283), the fitted values are generally greater than the integrated values. These differences are as expected given the increased accuracy afforded by using known DIB profiles to better constrain their full widths, as explained above.

Error estimates for the λ5780 and λ6283 equivalent widths and velocities were derived by the same Monte Carlo technique as used for the Na i spectral modeling; 500 different replicated spectra were produced for each DIB with (random) Poisson noise added to each, the σ value which was taken to be equal to the rms noise of the local continuum. Errors due to the uncertainties inherent in the spectral continuum rectification were accounted for by multiplicatively scaling each replicated spectrum by a random factor of between 1 − σ and 1 + σ. The DIB model parameters were refitted for each replicated spectrum and individual parameter errors were obtained from the ±68th percentiles of the respective parameter ranges.

As can be seen from the data in Table 4, the error estimates on the DIB equivalent widths derived from Monte Carlo fitting (in the column labeled as "Fit.") are typically larger (by ∼50%) than those derived using σΔλ (column labeled as "Int."). Although our Monte Carlo errors do not take account of interloping stellar features (these are considered separately; see below), they are expected to be more reliable than the σΔλ errors, which assume a somewhat arbitrary ±1σ error on the continuum placement, as described by McCall et al. (2010), and do not take into account the non-Gaussianity of the DIB profiles, which is especially important for λ6283 with its very broad, Lorentzian-like wings.

Table 4. λ5780 and λ6283 DIB Equivalent Widths and Velocities, M31 H i 21 cm Column Densities, Gas-to-dust Ratios, Interstellar NUV Radiation Field Strengths, and Spitzer 8/24 μm Flux Ratios for Observed Sightlines

Sightline	W(5780)		W(6283)
	Fit	Int.	Fit	Int.	v(5780)	v(6283)	$N({\rm H\,\mathsc{i}})$	G	INUV	8/24 μm	W(5780)/EB−V	W(6283)/EB−V
	(mÅ)	(mÅ)	(mÅ)	(mÅ)	(km s−1)	(km s−1)	(1021 cm)−2	(1021 cm−2 mag−1)	Draine	Flux	(Å mag−1)	(Å mag−1)
3936.51	84+33−32	89 (19)	279+96−85	217 (55)	−510.7+36.8−36.0	−498.5+63.9−49.3	2.8	20	0.27	1.16	1.2+2.8−0.7	4.0+8.9−2.3
3944.71	318+26−24	359 (17)	671+79−71	544 (48)	−512.4+7.5−7.0	−524.8+13.6−15	4.5	7.1	0.48	0.63	1.0+0.2−0.2	2.1+0.6−0.4
3945.35	202+22−24a	302 (14)a	587+69−74	414 (52)	−515.4+12.2−11.7	−510.6+15.6−15.4	4.2	6.4	0.35	0.94	0.6+0.2−0.1	1.8+0.5−0.4
3958.22	211+33−48	207 (35)	439+98−89	419 (60)	−517.9+18.1−16.7	−511.9+26.4−25.4	4.0	8.5	0.35	1.78	0.9+0.4−0.3	1.9+0.9−0.6
3959.98	<29	<42	117+90−60	124 (54)	...	−481.3+81.6−51.6	2.6	25	0.59	0.71	<2.2	2.2+13.8−1.6
4030.28	<41	<28	<42	<112	...	...	2.8	16	3.1	0.90	<0.9	<0.9
4034.61	120+39−41	165 (25)	375+105−89	405 (65)	−514.9+33.3−35	−601.3+31.0−30.4	2.3	4.0	2.9	1.20	0.4+0.2−0.2	1.3+0.7−0.4
4034.80	215+24−26	238 (22)	383+76−78	326 (54)	−584.5+14.2−13.7	−543.5+25.3−27.4	1.3	2.9	1.9	1.15	1.0+0.4−0.3	1.7+0.9−0.6
4037.92	<39	<49	177+122−82	123 (68)	...	−600.1+43.4−49.1	3.3	10	0.97	1.38	<0.3	1.1+1.3−0.6
4408.04	<30	<40	<243	<142	...	...	1.6	15	0.79	1.16	<2.3	<2.7
4408.79	146+74−48a	<88	189+169−115	171 (104)	...	−93.8+39.0−47.0	1.2	3.4	0.83	1.05	0.8+1.0−0.4	1.1+1.9−0.7
4409.52	<38	<57	<65	<168	...	...	1.4	11	0.59	1.26	<3.2	<5.4
4412.17	190+17−16	256 (10)	491+83−81	423 (53)	−88.1+8.2−8.2	−82.8+17.7−22.3	2.0	2.3	1.1	0.90	0.8+1.0−0.4	1.1+1.9−0.7
4417.80	79+55−41	<57	194+148−92	221 (76)	−99.3+57.8−48.6	−107.0+42.3−55.6	1.7	6.6	0.48	0.76	0.6+1.2−0.4	1.6+3.0−1.0

Notes. Fitted and continuum-integrated DIB equivalent widths are given, labeled Fit. and Int., respectively. The errors (±) on the integrated equivalent widths are σΔλ and are given in brackets, with upper limits calculated as 2σΔλ. Errors on the gas-to-dust ratios ($G=\frac{1}{2}$ $N({\rm H\,\textsc {i}})$/E^M 31_B−V) are ±100% due to probable contamination by background H i (see Section 3.6). ^aDIB equivalent widths probably affected by line blending.

Download table as:  ASCIITypeset image

The spectral regions spanning the λ5780 and λ5797 DIBs are shown in Figure 5 for the 14 stars in our M31 reddened sample (plus three telluric standards). The spectral regions covering λ6269 and λ6283 are shown in Figure 6 (these spectra have been divided by the spectra of the telluric standards from each respective field). M31 components of λ5780 are detected in 9 sightlines and λ6283 in 11 sightlines. The MW foreground λ6283 components are effectively removed during telluric division because of their presence in the telluric standard spectra.

A total of 11 different DIBs (λλ4428, 5705, 5780, 5797, 6203, 6269, 6283, 6379, 6613, 6660, and 6993) were detected in M31 toward the three most reddened targets 3944.71, 3945.35, and 4412.17 (data presented in Table 5 and Figures 5–13). Figures 8–13 also show reference MW DIB spectra (from Cordiner 2006), as well as telluric standard spectra to illustrate the likely level of contamination by stellar and telluric features.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

To account for possible additional uncertainties in the DIB equivalent widths given in Table 5 due to the presence of overlapping stellar lines, we have estimated the possible stellar line contributions. The table shows the total equivalent widths of stellar features that might be overlapping the DIBs, calculated from synthetic B0 Ia and B9 Ia stellar spectra (at temperatures of 28,000 and 10,000 K, respectively), using the method described in Section 3.3. The line list used in the spectral modeling includes transitions of the following elements and their ions: H, He, C, N, O, Na, Mg, Si, Ca, Cr, Mn, and Fe. The calculated stellar line contributions to measured DIB equivalent widths are generally small (except for the very broad λ4428 DIB which is contaminated by the O ii blend near 4415 Å) and in most cases are negligible compared with the errors arising from continuum placement and statistical noise.

As a result of the close blending between the MW and M31, interstellar gas components due to the low Doppler shift of Field 3 (discussed further in Sections 3.3 and 3.6), there may be some ambiguity as to which galaxy the Field 3 DIBs originate. The main discriminator we use to identify the location of the DIB carriers is their radial velocity compared with the Na i component structure. The separation of the MW and M31 DIB velocities is clear in all cases apart from 4412.17 and 4417.80, for which the DIB velocities appear to be intermediate between the M31 and MW Na i components. However, a number of factors lead us to conclude that the DIBs observed in these sightlines originate in M31: (1) the DIB velocities match the peak M31 H i velocities (see Section 3.6), whereas we detect a negligible contribution to the DIB equivalent widths at radial velocities corresponding to H i in the MW (at v_LSR < 45 km s⁻¹; see Braun et al. 2009); (2) LAB survey data toward the targets in question (Kalberla et al. 2005; Hartmann & Burton 1997) show a relatively small foreground H i component, such that the $W(5780)/N({\rm H\, \textsc {i}})$ ratios toward 4412.17 and 4417.80 would be very large if this DIB originated in the MW halo and inconsistent with the values observed previously in the MW ISM (see Figure 8 of McCall et al. 2010); (3) assuming the Galactic halo obeys the average MW DIB W versus E_B−V relation, the equivalent widths expected from the M31 foreground reddening of E_B−V = 0.06 ± 0.02 are only 32 ± 11 mÅ for λ5780 and 71 ± 24 mÅ for λ6283, whereas the measured DIB equivalent widths toward 4412.17 of 190 ± 17 mÅ and 491 ± 82 mÅ, respectively, are consistent with the M31 reddening (E^M 31_B−V) in this sightline; and (4) the Na i spectrum of 4406.87 in Figure 3, which shows no evidence for M31 Na i absorption, gives another indication of the properties of the foreground (MW) neutral gas in Field 3. The quantity and velocity of this foreground gas are very similar to the amount of gas in the other fields where the MW λ5780 components (which can be easily separated from the M31 DIBs) are clearly very weak. Assuming the foreground gas chemical properties are not grossly variable on the scale of the ∼1° that separates the GMOS fields, this strongly suggests that the foreground DIBs in Field 3 would be similarly as weak. Small-scale structure in the Galactic halo could produce variations in the relative abundances of atoms, dust, and DIB carriers, but given the reasons outlined above, it is highly unlikely that the Field 3 DIBs originate in chemically peculiar, compact, high-velocity Galactic clouds.

Dr. Robert Braun kindly made available data cubes and opacity-corrected H i column density maps from the Braun et al. (2009) 21 cm survey of M31, performed using the Westerbork Synthesis Radio Telescope and the Green Bank Telescope. These data are at a spatial resolution of 18'' × 15'', which corresponds to a linear resolution of 70 × 56 pc at the distance of M31. The H i spectra have a velocity resolution of 2.3 km s⁻¹. From these data, we used bi-linear interpolation to extract opacity-corrected column densities and H i spectra for our targets (shown in Figure 14).

Zoom In Zoom Out Reset image size

Opacity-corrected H i column densities and gas-to-dust ratios $G=\frac{1}{2}$$N({\rm H\,\textsc {i}})$/E^M 31_B−V are given in Table 4; the gas-to-dust ratio has thus been calculated assuming that half of the total H i column lies in front of the target stars. However, this fraction is uncertain, so the errors on G are ±100%.

The average gas-to-dust ratios are 13.4, 8.2, and 7.7 (in units of 10²¹ cm⁻² mag⁻¹), for Fields 1–3, respectively, with an overall average of 9.8. Previous measurements of the M31 gas-to-dust ratio lie in the range ≈ (5–15) × 10²¹ cm⁻² mag⁻¹ (van den Bergh 1975; Bajaja & Gergely 1977; Kumar 1979; Lequeux 2000). These values may be compared with the nearby Galactic value of 5.8 × 10²¹ cm⁻² mag⁻¹ (Bohlin et al. 1978), which was derived from H i absorption lines and therefore does not suffer from the problem of background contamination. It is difficult to draw firm conclusions from these data due to the uncertainties in the fraction of background H i in M31, but it seems reasonable to conclude that the gas-to-dust ratio of M31 is broadly similar to that of the MW. Although many of the observed sightlines appear to have gas-to-dust ratios significantly greater than Galactic, this may be caused by an observational bias toward stars located on the near side of the M31 disk; our derived gas-to-dust ratio would be an overestimate in such cases. Accounting for errors, the gas-to-dust ratio is lower toward 4412.17 than the average MW value. This sightline also clearly exhibits relatively weak DIBs per unit reddening.

Figure 14 shows that the velocities of the interstellar clouds as probed using Na i lines correspond imperfectly with the peak H i velocity in many cases. This may be due to the presence of background H i clouds that do not contribute to the Na i absorption spectra. However, for three of the Field 3 stars, the Na i velocities do not match the H i gas at all, with Na i velocities at least 20 km s⁻¹ redward of the H i. For 4412.17 and 4417.80, the Na i velocities are similarly displaced from the DIB velocities (see Section 3.5), but the DIBs match very closely with the positions of the H i peak velocities. Given that DIB carriers, Na i and H i are correlated (e.g., Hobbs 1974; Ferlet et al. 1985; Herbig 1993; McCall et al. 2010), and their velocities are generally closely matched in the diffuse ISM, there is likely to be a problem with the measured M31 Na i cloud velocities in Field 3 (probably for the reasons related to sky subtraction discussed in Section 3.3). The measured Na i velocities would be offset to the red (as observed) if the MW component is not as strong as was derived in the vapid model fits or if the sky-line subtraction was incomplete. To resolve this situation would require high-resolution Na i spectra in order to separate the M31 and MW components. The H i gas in Field 3 spans a relatively narrow range of radial velocities (from about −120 to −60 km s⁻¹) and should be considered as a more reliable indicator of the velocity of the interstellar matter in these sightlines. The resulting velocities from Gaussian fits to the H i emission profiles are shown for comparison with the M31 DIBs in Field 3 in Figures 5–11.

CO peak velocities from radio observations by Nieten et al. (2006) are consistent with those of the H i gas. For Field 3, CO matches better with H i than Na i, which provides further evidence that our measured M31 Na i velocities for 4408.04, 4412.17, and 4417.80 are probably in error.

In addition to the stars in Table 1, three H ii regions were observed serendipitously in Field 3, spatially coincident with the GMOS slits of the stars 4408.79, 4408.94, and 4417.12. Using the relative strengths of the emission lines in these spectra, it is possible to derive an estimate for the interstellar metallicity in the vicinity of Field 3. The usefulness of these spectra is limited because they are not flux calibrated and the blue and red wavelength regions were recorded in separate exposures. Nevertheless, an estimate of the O/H abundance ratio can be obtained from the relative N ii 6583 Å and Hα line fluxes (see Pettini & Pagel 2004).

The emission spectra were extracted from regions at the edges of the GMOS slits where the stellar flux was negligible, sky-subtracted using sky spectra from nearby slits showing no nebular emission, and continuum-subtracted using low-order polynomial fits. The final spectra are shown in Figure 15. The logarithmic flux ratios (log ([N ii]/Hα)) for 4408.79, 4408.94, and 4417.12 are −0.54, −0.63, and −0.61, respectively, which correspond to oxygen abundances 12 + log [O/H] = 8.59, 8.53, and 8.51 (all with 1σ errors of ±0.18), using Equation (2) of Pettini & Pagel (2004). The average value, 8.54 ± 0.18, is similar to that in the solar neighborhood, where the oxygen abundance of H ii regions is in the range 8.45–8.50 (see Pilyugin et al. 2003 and references therein).

Zoom In Zoom Out Reset image size

Our derived [O/H] ratios add to the limited data already available in the literature on the metallicity of M31. Trundle et al. (2002) reviewed the previous work on oxygen abundance measurements of H ii regions in M31 and identified that there is some uncertainty in the metallicity gradient across M31. Measured oxygen abundance gradients range from −0.013 to −0.027 dex kpc⁻¹, depending on the method used to derive [O/H]. Extrapolating these gradients gives 12 + log [O/H] = 8.70–9.21 at the center of M31, compared with 8.90 in the center of the MW. The three H ii regions we observed are approximately 23' from the center of M31, which corresponds to 5.2 kpc (assuming a distance of 783 kpc to M31; Holland 1998). The average oxygen abundance in the three H ii regions (8.54) lies toward the lower end of the range of values previously reported at this radius in M31.

Fields 1 and 2 lie at radial distances (r) of 63' (14.3 kpc) and 41' (9.3 kpc), respectively, from the center of M31. Using the metallicity relation of 12 + log [O/H] = 8.7–0.013r, which was deduced from the study of B star abundances by Trundle et al. (2002), we find oxygen abundances of 8.51 and 8.58 for these two fields. Thus, there is no evidence to suggest that the metallicity of the ISM along any of our surveyed sightlines is significantly different from solar.

Near-ultraviolet interstellar radiation field strengths (NUV ISRFs) were calculated for the M31 ISM in each sightline, using the NUV GALEX mosaic of M31 (imaged at a central wavelength of 2271 Å) from Thilker et al. (2005). The NUV ISRF strengths (I_NUV) are given in Table 4 in units of the Galactic NUV ISRF of Draine (1978), integrated over the aperture size and NUV wavelength response function of the GALEX telescope. Maps of I_NUV for each of our three fields were computed from the GALEX image and are shown in Figure 17.

To calculate these I_NUV maps, individual pixel fluxes were background-subtracted and corrected for foreground extinction using A_NUV = 8.0 × E_B−V (Gil de Paz et al. 2007), assuming an MW foreground extinction of 0.06 mag (see Section 3.4). The M31 extinction was derived for each field from the average of the E^M 31_B−V values of the stars listed in Table 1. For each sightline in the ISRF maps, I_NUV was calculated at a point at a line-of-sight distance r halfway into the foreground M31 cloud, using the extinction–distance relation derived by Vergely et al. (1998) for the solar neighborhood (dA_V/dr = 1.2 mag kpc⁻¹, with R_V = A_V/E_B−V = 3.1), and assuming that the observed NUV flux is emitted from the image plane (at a radial distance of 783 kpc). The NUV photons were assumed to be subject to the same level of extinction as a function of distance as in the solar neighborhood; consequently, the calculated I_NUV are dominated by the flux from OB stars in the immediate vicinity of the interstellar clouds of interest; stars at distances greater than 1 kpc make a negligible contribution to the total NUV flux at a given point in space. The NUV ISRF strengths thus calculated are only approximate because they take no account of the fact that the shape of the extinction law, and the extinction as a function of distance in M31, may be different from that observed in the MW. The M31 foreground extinction is treated as a uniform slab of material, so any small-scale inhomogeneities in the M31 ISM are neglected.

Within Fields 1–3, respectively, the interstellar UV field strengths (I_NUV) averaged over all pixels are 0.24, 0.77, and 0.34 (in units of the Draine field), with an average value of 0.45. Evidently, even though these fields are relatively rich in OB stars, the NUV ISRF is still significantly weaker than that of the MW. This is consistent with the conclusions of previous authors (see Section 1). Field 2 has the greatest I_NUV values due to the presence of OB78 (NGC 206)—the richest OB association in M31. Indeed, the sightlines toward the DIB target stars in this field have NUV field strengths which are comparable to or greater than the (Draine 1978) Galactic value. On the other hand, most of the stars in Field 1 toward which DIBs are observed have I_NUV < 0.5, including the star in our sample 3944.71, toward which the M31 DIBs are strongest.

For the three M31 fields presented here, we obtained reduced infrared Spitzer Space Telescope images in the IRAC 8 μm and MIPS 24 μm bands from the surveys of Barmby et al. (2006) and Gordon et al. (2006). The flux in the 8 μm band is dominated by PAH emission from the 7.7 μm (and 8.6 μm) aromatic bands (see, for example, Draine & Li 2001 and Flagey et al. 2006), whereas the flux in the 24 μm band predominantly arises from interstellar dust. The ratio of 8 μm to 24 μm fluxes may therefore be used as an approximate measure of the relative PAH emission intensity per unit dust mass (e.g., Flagey et al. 2006; Engelbracht et al. 2008; Gordon et al. 2008), and as such, provides an indication of the abundance and excitation level of PAHs in the ISM.

The 8 μm IRAC images are shown for our three fields in Figure 17. The interstellar emission flux was measured for each of the sightlines in Table 4 from the background-subtracted Spitzer images using a ring median filter of radius 5 pixels and width 2 pixels in order to exclude the flux from stellar point sources. To measure the 8/24 μm interstellar flux ratio, the 8 μm images were convolved and rebinned to the lower spatial resolution of the 24 μm images using the convolution kernel provided by Gordon et al. (2008).

In the plane of the MW, the interstellar 8/24 μm flux ratio (R) varies between about 0.1 and 2.0, with a mean value of 1.6 (Flagey 2007). Similarly, in nearby spiral galaxies, R also varies between about 0.1 and 2, with considerable scatter within individual galaxies (Bendo et al. 2008). The ratio R correlates with metallicity, which has been interpreted as being due to variation in the strength of PAH emission relative to the 24 μm dust emission, at least partly as a result of differences in PAH abundances (Engelbracht et al. 2005). However, PAH emission correlates better with ionization of the ISM than the metallicity (Gordon et al. 2008; Bendo et al. 2008), so the 8 μm flux may be a better indicator of the degree to which the PAHs are ionized rather than their overall abundance. R also tends to be lower inside H ii regions and higher in denser, dustier gas associated with photodissociation region boundaries.

In our GMOS fields, R values range between 0.63 and 1.78, with mean values of 1.05, 1.16, and 1.04 in Fields 1–3, respectively. Thus, there is nothing particularly unusual about the PAH emission in these fields compared with other galaxies, although the mean values are somewhat lower than observed in the MW. This is perhaps indicative of a reduced level of ionization or processing of the aromatic material in our M31 fields as a result of the lower UV radiation fields that are typical of M31.

The strength of the 8 μm emission discussed here is at variance with previous studies (Cesarsky et al. 1998; Pagani et al. 1999) that identified anomalously weak aromatic mid-IR emission bands in the bulge and nucleus of M31. Thus, the PAH emission traced by the 8 μm flux, at least in our observed fields (which are away from the nucleus/bulge), is apparently not as weak as previously thought and implies that the PAHs in M31 may in fact not be dissimilar to those found in the MW and other spiral galaxies. It must be noted, however, that the NUV ISRF in our observed fields may be greater than is typical for M31 due to the concentration of OB stars, which could give rise to increased 8 μm emission in these regions relative to the average conditions in M31.