5 Nebular abundances

   
5.1 Oxygen abundances: Direct method

The direct or standard method of obtaining oxygen abundances from emission lines is applicable to any galaxy where [O III] $\lambda 4363$ is detectable and for which the doubly ionized O⁺² ion is the dominant form of oxygen (Osterbrock 1989). A summary of the "standard'' method by which oxygen abundances are derived can be found in Dinerstein (1990). Computations were performed with SNAP. The relative abundances of singly- and doubly-ionized oxygen and the total oxygen abundance by number are computed using the method described by Lee et al. (2003a). An O⁺²/H abundance was computed using an O⁺² temperature, derived from the intensity of the [O III] $\lambda 4363$ and [O III] $\lambda\lambda 4959,5007$ lines, and an O⁺/H abundance was computed using an O⁺ temperature derived using Eq. (2) from Izotov et al. (1997b).

Direct ([O III] $\lambda 4363$) abundances were obtained for three galaxies (A1346-358, IC 1613 H II#37, and IC 5152 H II#A) and are listed in Table 9. Errors in direct oxygen abundances were computed from the maximum and minimum possible values, given the errors in the line intensities; errors in reddening and temperature are not included. For the remaining galaxies, secondary techniques using the bright emission lines of ionized oxygen are utilized to derive oxygen abundances.

 

 

Table 9: Derived nebular abundances. Column (1): H II region. Column (2): Oxygen abundance from [O III] $\lambda 4363$ measurements; lower limits to the oxygen abundance were obtained from 2$\sigma$ upper limits to the [O III] $\lambda 4363$ flux. Column (3): Oxygen abundance d. Column (4): Oxygen abundance derived using bright-line method by Pilyugin (2000). Columns (5) and (6): Nitrogen-to-oxygen abundance ratios: derived with O⁺ temperatures from Table 8 and with the bright-line method, respectively. Columns (7) and (8): Neon-to-oxygen abundance ratios: derived with O⁺² temperatures from Table 8 and with the bright-line method, respectively. NOTES: ^a The Pilyugin value is uncertain, but upper branch abundances appear to be correct; see text in Sect. 6.1.16. ^b Computed using the method of Thurston et al. (1996). ^c  $I(\hbox {[N II]$\lambda \lambda 6548,6583$ })$ is likely a lower limit. ^d  $I(\hbox {[Ne III]$\lambda 3869$ })$ likely too low. ^e Pilyugin value is uncertain. Oxygen abundances derived with $I(\hbox {[N II]$\lambda 6583$ })/I(\hbox {H$\alpha $ })$calibrations (van Zee et al. 1998b; Denicoló et al. 2002) are consistent with lower branch values; the value listed here is the average. ^f [N II] $\lambda 6583$ not measured; cannot break degeneracy in bright-line method. ^g Same as note e, but see also Sect. 6.1.15.

	12+log(O/H)
	Direct	Bright-Line		log(N/O)		log(Ne/O)
H II Region	[O III]$\lambda$4363	McGaugh	Pilyugin	Direct	Bright-line	Direct	Bright-line
(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)
Centaurus A group dwarfs
A1243-335 H II#A	...	7.87	7.69	...	-1.62	...	...
A1334-277 ap1	...	7.45	7.34	...	-1.04	...	...
A1346-358 H II#A	$8.19 \pm 0.06$	8.26	8.22	$-1.37 \pm 0.08$	-1.44	$-0.46 \pm 0.05$	-0.50
DDO 161 ap1	>7.76	8.08	8.03	...	-1.91	...	...
NGC 5264 ap1	...	8.66	8.54 a	...	-0.57 b	...	...
Sculptor group dwarfs
AM0106-382 ap1	...	7.54	7.71	...	-1.22	...	...
AM0106-382 ap2	...	7.58	7.56	...	-1.47	...	...
AM0106-382 ap3	...	7.61	7.59	...	-1.58	...	...
ESO347-G017 ap1	>7.80	7.89	7.78	...	-1.30	...	-0.52
ESO347-G017 ap2	...	7.92	7.76	...	-1.31	...	-0.53
ESO347-G017 ap3	...	7.96	8.03	...	>-1.86 c	...	...
ESO348-G009 ap1	...	7.89	8.07	...	-1.60	...	...
UGCA 442 H II#2	>7.48	7.81	7.88	...	-1.41	...	...
Other southern dwarfs
A0355-465 H II#B	...	8.23	8.01	...	-1.61	...	-0.48
ESO358-G060 ap1	>7.26	7.38	7.26	...	-1.24	...	>-1.11 d
IC 1613 H II#13	>7.61	7.90	7.89	...	-1.40	...	...
IC 1613 H II#37	$7.62 \pm 0.05$	7.88	7.71	$-1.13 \pm 0.18$	-1.35	$-0.60 \pm 0.05$	-0.62
IC 2032 ap1	...	7.96	7.98	...	-1.37	...	-0.21
IC 5152 H II#A	$7.92 \pm 0.07$	7.91	7.80	$-1.05 \pm 0.12$	-1.09	$-0.69 \pm 0.08$	-0.77
NGC 2915 ap1	...	8.29	8.33 e	...	-1.71	...	...
NGC 2915 ap2	...	8.21	8.35	...	-1.25	...	...
NGC 3109 H II#6 ap1 f	...	...	...	...	...	...	...
NGC 3109 H II#6 ap2 f	...	...	...	...	...	...	...
NGC 3109 H II#6 ap3 g	...	7.50	7.40	...	-1.37	...	...
NGC 3109 H II#6 ap4 g	...	8.07	8.13	...	-1.36	...	...
NGC 3109 H II#6 ap5 g	...	7.64	7.52	...	-1.28	...	...
NGC 3109 H II#6 ap6 g	...	7.60	7.51	...	-1.20	...	...
NGC 3109 H II#6 ap7 g	...	7.85	8.08	...	-1.38	...	...
Sag DIG H II#3	...	7.44	7.33	...	-1.63	...	...

   
5.2 Oxygen abundances: Bright-line method

In the absence of [O III] $\lambda 4363$, the bright-line or empirical method was used to compute oxygen abundances. The method is so called because the oxygen abundance is given in terms of the bright [O II] and [O III] lines. Pagel et al. (1979) suggested that the ratio

$$R_{23} = \frac{I({\hbox{[O II]$\lambda 3727$ }}) + I({\hbox{[O III]$\lambda\lambda 4959,5007$ }})}{I({\hbox{H$\beta$ }})}$$ (1)

could be used as an abundance indicator. For the dwarf galaxies in the present work, the calibrations by McGaugh (1991,1994) and Pilyugin (2000) were used to derive oxygen abundances.

McGaugh (1991,1994) produced a set of photoionization models using R₂₃ and

$$O_{32} = \frac{I({\hbox{[O III]$\lambda\lambda 4959,5007$ }})}{I({\hbox{[O II]$\lambda 3727$ }})}$$ (2)

to estimate the oxygen abundance. However, R₂₃ is not a monotonic function of the oxygen abundance. At a given value of R₂₃, there are two possible choices of the oxygen abundance as shown in Fig. 3. Each filled circle represents an H II region from dIs in the control sample with measured [O III] $\lambda 4363$. Model curves from McGaugh (1991) are superimposed. A long-dashed line marks the approximate boundary below (above) which the lower branch (upper branch) occurs. Despite the fact the oxygen abundances for H II regions in the control sample of dIs range from about one-tenth to about one-half of the solar value, these H II regions are clustered around the "knee'' of the curves, where ambiguity is greatest about the choice of the appropriate branch in the absence of [O III] $\lambda 4363$.

intensity ratio can discriminate between the lower and upper branches (McCall et al. 1985; McGaugh 1991,1994). The strength of the [N II] $\lambda 6583$ line is roughly proportional to the nitrogen abundance and the [N II]/[O II] intensity ratio is relatively insensitive to ionization. McGaugh (1994) has shown that in galaxies ranging from sub-solar to solar metallicities, [N II]/[O II] can vary by one to two orders of magnitude and that [N II]/[O II] is roughly below (above) 0.1 at low (high) oxygen abundance. A plot of the [N II]/[O II] intensity ratio versus R₂₃ is shown in Fig. 4.

Figure 3: Oxygen abundance versus bright-line indicator, R23. The filled circles indicate H II regions from the control sample of nearby dIs (Lee et al. 2003a), whose oxygen abundances were obtained directly from measurements of the [O III] $\lambda 4363$ emission line. Crosses indicate southern dwarfs in the present sample for which [O III] $\lambda 4363$ was measured. Calibration curves for log(O/H) against log R23(McGaugh 1997, private communication) are plotted for four different values of the ionization parameter (log u; see McGaugh 1991). These curves were derived using a standard stellar initial mass function for a cluster of ionizing stars with an upper mass limit of ${\cal M}_{\rm upp}$ = 60 $M_{\odot }$ (solid lines) and 100 $M_{\odot }$ (short-dash lines); see also McGaugh (1991). The horizontal long-dash line marks the approximate boundary below (above) which the lower (upper) branch occurs. [O III] $\lambda 4363$ abundances are consistent with lower branch values. The error bars at the lower right indicates typical uncertainties of 0.1 dex for the R23 indicator and 0.1 dex for the direct oxygen abundance.

While most H II regions in the present sample lie in the lower branch regime, they lie to the left of the locus of points defined by H II regions in nearby dIs. This is expected for metal-poor galaxies as ionization effects become more prominent (McGaugh 1991). The corrected [N II]/[O II] was used to determine the branch for computing the oxygen abundance in both the McGaugh and Pilyugin calibrations.

Figure 4: [N II]/[O II] discriminant versus bright-line indicator, R23. Filled circles indicate H II regions from the control sample of nearby dIs with [O III] $\lambda 4363$ detections (Lee et al. 2003a). Crosses indicate H II regions for the present sample of southern dwarfs; crosses surrounded by open circles indicate [O III] $\lambda 4363$ measurements. The error bars at the lower right indicate typical uncertainties of 0.1 dex for the R23 indicator and 0.2 dex (at most) for the [N II]/[O II] ratio. The dotted lines mark the regions in the diagram occupied by high surface brightness (HSB) spiral galaxies (McCall et al. 1985) at the upper left and low surface brightness (LSB) galaxies (McGaugh 1994) towards the lower right. As suggested by McGaugh, the horizontal dashed line marks the approximate boundary below (above) which the lower (upper) branch of the bright-line method is selected to determine a unique value of an oxygen abundance.

For the McGaugh (1997, private communication) calibration, analytical equations for the oxygen abundance are given in terms of $x \equiv \log~R_{23}$ and $y \equiv \log~O_{32}$. The expressions for lower branch and upper branch oxygen abundances are

$$12 + \log~({\rm O/H})_{\rm lower} 12 - 4.93 + 4.25x - 3.35\sin ~(x)$$ =

$$- 0.26y - 0.12\sin ~(y),$$ (3)

$$12 + \log~({\rm O/H})_{\rm upper} 12 - 2.65 - 0.91x + 0.12y ~ \sin ~(x),$$ = (4)

respectively, where the argument of the trigonometric function is in radians.

Pilyugin suggested a new calibration for the bright-line method. His method at low metallicities accounts for the systematic uncertainties in the R₂₃ method, whereas at high metallicities, he obtains a relation for the oxygen abundance as a function of the intensities of the bright [O II] and [O III] lines. For lower branch and upper branch abundances, we use Eq. (4) from Pilyugin (2000), and Eq. (8) from Pilyugin (2001a), respectively.

Bright-line oxygen abundances are derived and listed in Table 9. Figure 5 shows how the different determinations of the oxygen abundance vary with O₃₂ and R₂₃. Differences in derived oxygen abundance between the direct ([O III] $\lambda 4363$) and bright-line McGaugh methods, between the direct and bright-line Pilyugin methods, and between the two bright-line methods are shown. The separations among the three methods appear to increase with increasing O₃₂. The difference between the McGaugh and Pilyugin calibrations (indicated by asterisks) appears to correlate with log O₃₂; this effect is also observed by Skillman et al. (2003). IC 1613 H II#37 with the largest measured O₃₂( $\log~O_{32} = 0.998$) exhibits the largest overall discrepancy among the direct, McGaugh, and Pilyugin methods $\rm (12{+}log(O/H) = 7.62$, 7.88, and 7.71, respectively).

   
5.3 Nitrogen-to-oxygen and neon-to-oxygen abundance ratios

Based upon observations of H II regions in spiral and dwarf galaxies, nitrogen appears to be both a primary and secondary product of nucleosynthesis. It remains uncertain, however, whether nitrogen is produced mostly from short-lived massive stars or from longer-lived intermediate-mass stars. An extensive review of the possible origins for nitrogen is discussed by Henry et al. (2000).

Measurements of the nitrogen-to-oxygen ratio, N/O, have been used to differentiate between the different origins for nitrogen. It has been suggested that N/O can be used as a "clock'' to measure the time since the last burst of star formation (e.g., Garnett 1990; Skillman et al. 1997,2003). This scenario works if bursts of star formation are separated by long quiescent periods, if the delivery of nitrogen into the interstellar gas is delayed relative to oxygen, and if there is no significant metals loss. The result is that N/O values are low at a given O/H if a burst of star formation has occurred recently, whereas N/O values are high after a long quiescent period.

For low-abundance H II regions, $\log({\rm N}/{\rm O}) \approx \log({\rm N}^+/{\rm O}^+)$is a good approximation (e.g., Garnett 1990). Assuming that $T_e({\rm N}^+) = T_e({\rm O}^+)$, the nitrogen-to-oxygen abundance ratio is

$${\rm N}/{\rm O} = \frac{I({\rm [N\;II]}\lambda~6583)}{I({\rm ... ... O}^+))} {j({\rm [N\;II]}\lambda~6583;~n_e,~T_e({\rm O}^+))},\$$ (5)

where j is the volume emissivity of the given line at density n_e and temperature T_e. In the absence of [O III] $\lambda 4363$, a bright-line oxygen abundance was derived using the McGaugh calibration and an O⁺²temperature was obtained using the correlation between oxygen abundance and temperature in Fig. 7 of McGaugh (1991). Subsequently, an O⁺ temperature was estimated using the data in Table 1 of Vila-Costas & Edmunds (1993). The nitrogen-to-oxygen abundance ratio was then computed by using the appropriate line intensities with Eq. (9) from Pagel et al. (1992).

Neon is a product of $\alpha$-processes in nucleosynthesis occurring in the same massive stars which produce oxygen. As a result, the neon-to-oxygen ratio, Ne/O, is expected to be constant with oxygen abundance. Assuming that doubly-ionized neon is found in the same zone as doubly-ionized oxygen and that $T_e({\rm Ne}^{+2}) = T_e({\rm O}^{+2})$ and that $\log({\rm Ne}/{\rm O}) \simeq \log({\rm Ne}^{+2}/{\rm O}^{+2})$, the neon-to-oxygen abundance ratio is

$${\rm Ne}/{\rm O}\!=\! \frac{I({\rm [Ne\;III]}\lambda~3869)}{I... ...\lambda~3869;~n_e,~T_e\left({\rm Ne}^{+2}\right)\right)}\cdot\$$ (6)

In the absence of [O III] $\lambda 4363$, a bright-line temperature for the O⁺²zone is obtained (see previous paragraph) and the neon-to-oxygen abundance ratio is computed using the appropriate line intensities with Eq. (10) from Pagel et al. (1992).

Table 9 lists N/O and Ne/O abundance ratios. For the three galaxies with [O III] $\lambda 4363$ detections, N/O and Ne/O values derived using the direct method generally agree with those derived with the bright-line method, except that the direct value of log(N/O) for IC 1613 H II#37 is 0.2 dex larger than the bright-line value. A few of the Ne/O values are higher than those derived for blue compact dwarf galaxies: $\rm log(Ne/O) \approx -0.7$ (Izotov & Thuan 1999), but [Ne III] $\lambda 3869$ flux measurements for galaxies in the present sample may be overestimated from noisy spectra.