THE RECENT STAR FORMATION IN NGC 6822: AN ULTRAVIOLET STUDY

We used GALEX images in FUV (λ_eff = 1539 Å, FWHM ≈270 Å), and NUV (λ_eff = 2316 Å, FWHM ≈615 Å) with resolution 42 (FUV), and 53 (NUV; Morrissey et al. 2007), corresponding to ∼12 pc at the distance of NGC 6822. The images are sampled with 15 pixels.

The GALEX images of NGC 6822 were taken on 2005 August 20 as part of the NGS program, with exposure times of 4654 s (FUV) and 6198 s (NUV). The data were downloaded from the MAST archive. The 12 diameter GALEX field of view is centered at R.A. =19^h44^m5737, decl. =−14°47'33''.32, near the center of the galaxy (R.A. =19^h44^m578, decl. =−14°48'11'', FK5 2000). NGC 6822, with an optical diameter (at ∼25 mag arcsec⁻²) of 156 (Karachentsev et al. 2004), is contained in the central portion of the GALEX image, which is shown in Figure 1.

Zoom In Zoom Out Reset image size

We identified the SF regions using the GALEX FUV image, which unambiguously reveals the young, hot massive stars not heavily embedded in interstellar dust. We followed the general method of Kang et al. (2009), adapted to the case of NGC 6822. We defined contours of regions with FUV surface brightness ⩾3σ above the background. An important issue in defining extended source contours and measuring their flux is the background estimate. Several approaches were used to find the best method for background evaluation (see also the discussion in Kang et al. 2009). For the purpose of source detection only, we constructed a background image applying a two-step circular median filter (64 pixels diameter, ≈1.5 arcmin) to the FUV intensity map ("int" file). The first pass of the filter identifies pixels which belong to localized peaks via masking pixels brighter than the local median background estimate. The second pass of the filter operates only on the final list of non-peak pixels to obtain a background image less biased by substructure than a one-pass median filter. The diameter of the median filter was chosen to provide a background image where measurements of the background for individual sources are closest to the median flux density of the intensity map images in 6 pixel wide annuli around the sources, from here on the "local background." The background image produced using the adopted median filter gives sky estimates slightly lower (by about 0.17/0.18 mag arcsec⁻² for FUV/NUV) than the local background. For comparison, the background image provided by the pipeline gives a background estimate lower than the local background by 0.52/0.60 mag arcsec⁻² on average for FUV/NUV. A background-subtracted image was constructed, subtracting our background image from the intensity map image, and used for source detection.

Source contours were defined to enclose contiguous pixels with FUV flux more than 3σ above the background image. This threshold corresponds to an FUV surface brightness of 0.0015 counts s⁻¹ pixel⁻¹ (26.8 AB mag arcsec⁻²) on the background-subtracted image, or average 0.0025 counts s⁻¹ pixel⁻¹ (26.2 AB mag arcsec⁻²) on the intensity map image.

The effect of the threshold choice on the source-contour definition is illustrated in the top panels of Figure 1, which show SF regions 75, 57, 27, and 20 defined for thresholds of 1σ (green), 3σ (light blue), and 4σ (dark blue). A low threshold of 1σ or 2σ would cause regions such as 27 and 20 (OB8 and OB6, from here on "OB" designations are from Hodge 1977) to merge, and the main body of the galaxy to appear as one large region. If a threshold higher than 3σ is used, sparse associations like region 75 (OB15) split into several sources or into individual stars (see also Kang et al. 2009 for more discussion on the procedure and the choice of parameters).

The sources defined using the 3σ threshold follow the distribution of blue stars as shown by resolved stellar photometry from HST imaging (Bianchi et al. 2001; Bianchi & Efremova 2006) and ground-based data (Massey et al. 2007b). To exclude artifacts, in the initial list we rejected sources with areas less than 16 arcsec² (7 pixels).

We further restricted the analysis sample to sources larger than 150 arcsec² (≈860 pc²), in order to exclude single stars (mostly foreground) and background objects, and to examine SF complexes massive enough that stochastic effects will not be significant in deriving ages and masses by model analysis (Section 3). Stochasticity may affect the comparison of integrated star cluster photometry with stellar population models, as was first pointed out by Girardi et al. (1995). Quantitative assessment of this effect is still a matter of debate, given that more factors are relevant in such analysis, including initial mass function (IMF), metallicity, and extinction. For example, Fatuzzo & Adams (2008) estimate that for clusters with ≳1000 stars the IMF is sampled well enough so that their UV flux is close to model predictions for integrated populations (but they also point out that the exact limit may vary with IFM). Their analysis concerns statistical distributions of bound, spherical, zero-age stellar clusters. Such limit corresponds to ⩾5 stars more massive than 10 $\cal M_{\odot }$ i.e., earlier than spectral type B2V (with the parameters adopted by these authors). We will return to this point again later.

We choose to restrict the analysis sample with the area cut of 150 arcsec² after examining the distribution of stars with spectral type earlier than B2V⁸ inside the FUV-defined contours. In Figure 2, we plot the FUV magnitude versus the areas of the SF regions: those containing ⩾5 blue massive stars are shown with dots. A cut by an area of ⩾150 arcsec² includes 94% of these regions in the sample, and very few regions containing less than five blue stars (20%). We examined the alternative option of a brightness cut, which is often used in studies of more distant galaxies. Such criterion would either include fewer SF regions with five or more blue stars, or more regions with less than five blue stars, in the analysis sample. For example, a brightness cut at FUV ⩽ 20.2 mag includes in the analysis sample 94% of the regions with ⩾5 blue stars (the same fraction as our area cut of ⩾150 arcsec²), but 44% of the selected regions would contain less than five blue stars. Therefore, a cut by area better satisfies our requirement of a minimum number of blue stars within a source contour, including in the analysis sample as many as possible of the clusters having ⩽5 massive stars, and as few as possible clusters with <5 massive stars. A brightness or luminosity cut would also strongly be affected by extinction or extinction correction (see Section 3.2). We point out that this criterion may not necessarily be the best choice for more distant galaxies where similar data would give a lower spatial resolution, or for galaxies where star formation is less sparse. In the specific case of NGC 6822, such criterion, more precise than a luminosity cut for our purpose, could be tested and tuned given the vicinity of the galaxy and the detailed information on its stellar population. Finally, a cut by area may eliminate young compact clusters and may be undesirable in disk galaxies, for example, where young compact star clusters abound (e.g., Bianchi et al. 1999 and Chandar et al. 1999 for M33; Hodge et al. 2010 and Y.-B. Kang et al. 2011, in preparation for M31). In NGC 6822, there are very few such compact clusters and their exclusion would not change our results. This work aims to detect unbound OB associations and SF complexes, not compact star clusters. Another advantage of the area cut is that it effectively excludes foreground stars.

Zoom In Zoom Out Reset image size

The resulting analysis sample includes 77 FUV-defined sources with area ⩾150 arcsec² and brightness ≲26.8 mag arcsec⁻², within a 02  radius (1.72 kpc) of the center of NGC 6822. The 02 radius is 1.5 times the optical semimajor axis of the galaxy (78 at ∼25 mag arcsec⁻²) given by Karachentsev et al. (2004). Hα emission is detected out to a radius of R_Hα = 1.65 kpc (Hunter & Elmegreen 2004; see also Section 2.3), and the H i disk (de Blok & Walter 2006) also exceeds the optical size of the galaxy (see also Bianchi 2011). Our sample extends to a slightly larger area than that of Melena et al. (2009).⁹

The areas of the selected SF regions range from 150 to 5400 arcsec² (860–3 × 10⁴ pc²). Table 1 gives identification, coordinates of the "centroids" (the median α and δ values of the pixels included in the contours), and areas of the FUV-defined regions, ordered by increasing R.A. The contours are shown in blue in Figure 1 over the GALEX FUV image. In the next section, we describe the photometry measurements, which are used in Section 3 to derive ages and masses.

For photometric measurements, we used the intensity map ("int") images (in units of counts s⁻¹ pixel⁻¹) generated by the GALEX pipeline dividing the count map by the relative response map.¹⁰ We measured the FUV and NUV flux of each SF region within its FUV-defined contour and the local background. The background was measured over an area defined by smoothing the source contour and expanding it by 3 pixels (inner background contour) and 9 pixels (outer contour), i.e., creating a 6 pixel wide "annulus" around the source, which follows its shape. The median of the flux pixel⁻¹ in the background region, excluding portions of nearby sources falling in the background annulus, was then subtracted from every pixel inside the source contour. The conversion from (counts  s⁻¹) to magnitudes in the AB photometric system was performed using zero points, ZP =18.82 mag (FUV) and 20.08 mag (NUV; Morrissey et al. 2007). We calculated the photometric errors as $\Delta {\rm mag}\approx \break 2.5/ \ln (10)\times \frac{N}{S}$ = $1.09\frac{N}{S}$, where S is the flux from the source in the aperture and N is the quadratic sum of all the noises affecting the image. S is expressed by $S=F\times \rm{EPADU}$, where F is the flux in counts, and $\rm{EPADU}$ is the conversion factor from ADU to e⁻, for GALEX EPADU = 1. We consider the Poisson noise of the photon flux, and the background fluctuations to dominate, so we used the following expression to estimate the noise: $N=\ssty\sqrt{N_{{\rm source}}^2 + N_{{\rm background}}^2 + A\sigma ^{2}}=\ssty\sqrt{S + S_{{\rm background}} + A\sigma ^{2}}$, where A is the area of the source (in pixels) and σ is the standard deviation among the pixels in the background annulus (in counts). The resulting FUV and NUV magnitudes and their errors are listed in Table 1.

The total flux from the selected SF regions is 45% of the integrated FUV flux from NGC 6822 ($F_{\rm FUV\rm tot}=1.05\pm 0.07\times 10^{-9}$ erg s⁻¹ cm⁻² or $\rm FUV_{{\rm tot}}=12.1\, \pm \, 0.07$ mag AB) and 35% of the total NUV flux ($F_{\rm NUV\rm tot}=1.09\, \pm\, 0.04\times 10^{-9}$ erg s⁻¹ cm⁻² or $\rm NUV_{{\rm tot}}=11.7\pm 0.04$ mag AB), measured from the pipeline sky-subtracted image in an aperture of 02 radius. The flux not included in our SF regions comes from smaller sources excluded by our area cut (10% of $F_{\rm FUV\rm tot}$), from older diffuse populations, and from scattered emission from SF regions. The fraction of flux included in the selected SF sites is lower in NUV than in FUV because foreground stars and diffuse light from older populations are more conspicuous at longer wavelengths.

We also used the publicly available CTIO Hα image from the survey of Massey et al. (2007a) to define contours of Hα emitting regions. The Hα image has an exposure of 300 s, a scale of 027 pixel⁻¹, and a resolution of 09 (2.2 pc at the distance to NGC 6822). We used the V and R images from the same survey (Massey et al. 2007b) to correct the Hα image for continuum, by subtracting from it a linear combination of the V and R images, scaled to match the intensity of the continuum sources. We define contours of Hα emitting regions using a threshold of 3σ above the background, corresponding to a surface brightness of 3 × 10⁻¹⁷ erg s⁻¹cm⁻² arcsec⁻². The Hα-defined contours are drawn in red in Figure 1. They generally follow the Hα contours given by Hodge et al. (1988, 1989), who defined them in a similar way using a threshold of 2 × 10⁻¹⁷ erg s⁻¹cm⁻² arcsec⁻².

We used the calibration factor of 1 count s⁻¹ = 1.8 × 10⁻¹⁶ erg s⁻¹ cm⁻² given by Massey et al. (2007a) for emission-line sources. We did not attempt to correct for [N ii] emission line contamination, which we expect to not exceed a few percent of the flux. The average $F_{[\rm N\,{\mathsc {ii}}]\lambda 6584}$/$F_{\rm {H\alpha }}$ in the H ii regions measured by Pagel et al. (1980) is about 6%; adopting $F_{[\rm N\,{\mathsc {ii}}]\lambda 6548}$/$F_{[\rm N\,{\mathsc {ii}}]\lambda 6584}$ ≈ 1/3, we derive $F_{[\rm N\,{\mathsc {ii}}](\lambda 6548+\lambda 6584)}$/F_Hα ≈ 8%. Since both [N ii]λλ 6548, 6584 fall in the wings of the 50 Å wide filter passband centered at Hα, we expect the actual contribution from [N ii] emission to be of the order of 2%–3%.

The Hα measurements are listed in Table 2. In the last column, we give the cross-identifications with Hα sources previously defined by Hubble (1925; H followed by a roman number), Hodge (1977; Ho followed by an arabic number), Kinman et al. (1979; K followed by a Greek letter), Killen & Dufour (1982; KD followed by an arabic number), and Hodge et al. (1988; HK followed by an arabic number). For regions with Hα surface brightness >1 × 10⁻¹⁶ erg s⁻¹cm⁻² arcsec⁻², our measurements agree within ∼25% with those by Hodge et al. (1989), after de-correcting the latter for extinction with A$_{\rm {H\alpha }}=0.9$ (O'Dell et al. 1999). The total Hα flux from the Hα emitting regions, $F_{\rm {H\alpha }}=1.8\pm 0.2\times 10^{-11}$ erg s⁻¹ cm⁻² (the sum of all measurements in Table 2), is higher by about 2% than the total flux from Table 1 of Hodge et al. (1989), most of the difference coming from faint H ii regions not covered by the imaging of Hodge et al. (1988). We further checked our Hα calibration and continuum-source subtraction by measuring the flux from the H ii regions Hubble V and Hubble X surrounding OB8 and OB13 in 42''×42'' square apertures, similar to the ones used by O'Dell et al. (1999): our measurements agree within 10% with the values given by these authors.

Flux at UV wavelengths is very sensitive to extinction by interstellar dust, and in order to derive the age and mass of the SF regions from model analysis, we first estimated the amount of reddening in each.

We used information from resolved stellar photometry, and estimated the E(B − V) of the hot massive stars (selected with (B − V)₀ < −0.2 and M_V < −2.5, i.e., earlier than about ∼B2V) within each contour. For several OB associations, E(B − V) values are available for individual stars, derived by Bianchi et al. (2001) and Bianchi & Efremova (2006) from HST multi-band photometry (from UV to optical). For the most massive stars in six OB associations, we also have VLT spectroscopy (B. V. Efremova et al. 2011, in preparation), which confirms the results from HST photometry. For the regions without HST photometry, we derived E(B − V) using the CTIO U, B, V photometry of Massey et al. (2007b), with the standard "Q-method" (e.g., Bianchi & Efremova 2006; Kang et al. 2009), and by comparing the observed (U − B), (B − V) colors with progressively reddened stellar model colors.

We accounted for extinction in each SF region using the median E(B − V) value of the massive stars it contains. The values range from E(B − V) = 0.22 (purely foreground extinction) to 0.66 mag, with a mean of E(B − V) = 0.36 mag, and are given in Table 1. The mean E(B −  V) values are similarly distributed, ranging from E(B − V) = 0.21 to 0.51 mag, with an average of E(B − V) = 0.37 mag. The typical 1σ scatter (also given in Table 1) around the mean E(B − V) in individual SF regions is 0.13 mag. The wide range of extinction values in the SF regions (not uncommon in SF galaxies) underscores the importance of accurate extinction corrections, particularly relevant in the UV regime, for deriving ages and masses. Other works adopt a generic assumption, for example, Melena et al. (2009) used a constant extinction of 0.27 mag for all their sample regions in NGC 6922, corresponding to E(B − V) =0.05 mag of internal extinction in addition to the E(B − V) = 0.22 mag foreground reddening. Our results derived for individual regions (Table 1) indicate that a higher value is more typical. The model magnitudes plotted in Figure 3 illustrate how such assumptions affect the derived ages and masses; more model plots showing the effects of extinction can be found in Bianchi (2011).

We assume a foreground extinction of E(B − V = 0.22 mag, consistent with the minimum E(B − V) estimated in this work, and with previous estimates by Bianchi et al. (2001), Bianchi & Efremova (2006), and Massey et al. (2007b). Any additional extinction is considered to originate within NGC 6822.

While the derived E(B − V) values are mostly based on optical photometry of the stars, and do not depend significantly on the type of dust, the selective extinction A_λ/E(B − V) at UV wavelengths is known to strongly vary with environment. Therefore, the correction of UV magnitudes, and the resulting ages and masses, strongly depend on the adopted extinction curve (e.g., Bianchi 2011). In the analysis that follows, we consider four different types of internal extinction, found in the MW and in known low-metallicity SF environments: (1) MW-type extinction with R_V = 3.1. In this case, the GALEX (FUV–NUV) color is basically reddening free (Bianchi 2011, and references therein); (2) the average extinction curve derived by Misselt et al. (1999) for LMC stars outside the 30 Doradus region (from here on, "LMC"), which gives an average color excess ratio for hot stars (T_eff>10, 000 K) of $E(\rm FUV-\rm NUV)/\mbox{$E(B-V)$}\approx 1$; (3) the extinction curve in the LMC 30 Doradus region (from here on "LMC2") derived by Misselt et al. (1999), yielding $E(\rm FUV-\rm NUV)/\mbox{$E(B-V)$}\approx 2$; and (4) the extremely UV-steep extinction curve derived by Gordon & Clayton (1998) for SMC stars (from here on, "SMC"), which yields $E(\rm FUV-\rm NUV)/\mbox{$E(B-V)$}\approx 5$.

We derive the ages of the SF regions from their integrated FUV–NUV colors compared with the SSP models for low-metallicity populations (see below), and masses from the age and UV magnitudes, accounting for reddening. We compared results obtained by adopting the different extinction curves mentioned in the previous section with information from resolved stellar photometry and Hα, in order to assess what type of selective extinction is more appropriate.

We found that a uniform extinction type is not adequate for the whole sample of SF regions in NGC 6822. If we assume "average MW" extinction with R_V = 3.1 for the whole sample (as adopted, e.g., by Hunter et al. 2010), the measured colors imply ages too old for several regions which show Hα emission and which appear to be a few Myr old in H-R diagrams from HST photometry (Bianchi et al. 2001). On the other hand, if we use a UV-steep extinction curve, "LMC2" for example, to deredden all SF regions, the (FUV–NUV) color for part of the sample is overcorrected, such that it appears unrealistic when compared with model predictions at any age. We found that different extinction curves are needed to bring the ages from GALEX integrated measurements in agreement with results from resolved studies for a subsample of well-studied regions. Regions 27, 57, 75, 19, 20 (approximately corresponding to OB8, OB13, OB15, OB7, and OB6 as defined by Hodge 1977), and 52, are included in the HST photometric studies by Bianchi et al. (2001) and Bianchi & Efremova (2006); $\rm{VLT}$ spectroscopy of the most massive stars confirms the ages derived from HST photometry. By comparing results from integrated measurements with resolved stellar photometry of these well-studied SF regions (outside the central part of the galaxy, where measurements are not complicated by significant diffuse older populations), and with information from Hα emission, we derived a criterion for choosing the type of extinction curve, and apply it to the rest of the sample. Specifically, the age information for the best-studied SF regions, with spectroscopy available for the hottest stars, was based on the presence (or absence) of O-type stars, W-R type stars, or B supergiants, as well as on the photometric H-R diagram of their stellar population, and we ruled out extinction curves giving very discrepant results from the GALEX color. For the FUV-bright regions clearly associated with Hα emission, we ruled out extinction curves yielding ages significantly older than 10 Myr from integrated UV photometry. Finally, UV-steep extinction curves were excluded in the cases where they would yield an intrinsic FUV–NUV color bluer than any stellar population model at any age. While derivation of the actual extinction curve would require UV spectroscopy of several stars in each region, and it is not possible with broadband photometry, the representative known curves examined give sufficiently different results (Table 3) that some of these assumptions could definitely be excluded in many cases. Some consistent trends within the subsample of SF regions with information on their stellar content allowed us to define general criteria.

Table 3. Ages and Masses of the FUV-defined SF Regions Assuming Different Extinction Curves

IDb	MW with Rv = 3.1a		LMCa		LMC2a		Extinction Curve
	Age (Myr)	Mass (103$\cal M_{\odot }$)	Age (Myr)	Mass (103$\cal M_{\odot }$)	Age (Myr)	Mass (103$\cal M_{\odot }$)	Adoptedc
EB-FUV 1	186+15−26	240.+355.−143.	47+49−34	46.+74.−28.	3+38−3	1.9+3.4−1.2	LMC
EB-FUV 2	61+1−2	7.7+5.5−3.2	25+20−14	2.7+2.1−1.2	5+22−4	0.4+0.3−0.2	Rv3.1
EB-FUV 3	28+1−7	6.0+15.9−3.8	5+21−5	0.8+2.2−0.5	2+25−2	0.7+2.1−0.4	Rv3.1
EB-FUV 4	37+1−2	18.+12.−7.	6+12−2	1.7+1.3−0.8	2+3−2	1.3+1.1−0.6	LMC
EB-FUV 5	56+2−5	11.+10.−5.	11+17−7	1.6+1.5−0.8	2+4−2	0.6+0.6−0.3	Rv3.1
EB-FUV 6	123+1−8	560.+949.−278.	76+45−41	320.+596.−166.	46+75−43	180.+361.−96.	SMC
EB-FUV 7	71+1−8	13.+46.−6.	44+25−40	7.5+30.2−3.9	19+51−19	2.6+11.6−1.4	Rv3.1
EB-FUV 8	37+1−2	8.8+11.5−4.7	13+23−9	2.4+3.4−1.3	4+32−4	0.6+0.9−0.4	Rv3.1
EB-FUV 9	8+1−3	1.0+2.1−0.7	...	...	...	...	Rv3.1
EB-FUV 10	46+17−28	66.+257.−53.	⋅⋅⋅	⋅⋅⋅	⋅⋅⋅	⋅⋅⋅	Rv3.1
EB-FUV 11	51+1−3	12.+15.−7.	16+28−12	2.8+4.0−1.6	3+33−3	0.4+0.7−0.3	Rv3.1
EB-FUV 12	223+6−8	120.+42.−32.	90+25−20	37.+14.−10.	26+21−16	8.6+3.5−2.5	Rv3.1
EB-FUV 13	114+2−14	48.+72.−29.	48+44−33	18.+29.−11.	11+63−11	3.0+5.3−1.9	LMC
EB-FUV 14	53+1−9	14.+42.−11.	11+42−10	2.2+7.2−1.7	2+51−2	0.7+2.7−0.6	Rv3.1
EB-FUV 15	129+1−14	260.+492.−164.	64+64−44	120.+244.−76.	27+101−26	43.+98.−29.	Rv3.1
EB-FUV 16	129+14−34	58.+154.−42.	30+58−27	11.+33.−8.	2+51−2	1.0+3.2−0.7	Rv3.1
EB-FUV 17	90+2−3	16.+7.−5.	33+13−15	5.0+2.5−1.7	4+10−2	0.4+0.2−0.2	Rv3.1
EB-FUV 18	67+1−1	6.0+6.7−2.5	46+20−28	3.9+4.9−1.7	27+39−24	2.0+2.8−0.9	Rv3.1
EB-FUV 19	118+5−18	330.+546.−204.	44+45−35	100.+192.−68.	5+58−5	7.6+15.4−5.1	LMC2
EB-FUV 20	54+1−5	110.+217.−68.	22+31−18	39.+81.−24.	5+48−5	6.2+14.3−4.0	LMC2
EB-FUV 21	45+1−2	4.6+9.9−2.1	22+22−18	2.0+4.6−0.9	6+38−6	0.4+1.0−0.2	Rv3.1
EB-FUV 22	40+1−4	6.7+12.6−3.6	15+24−12	1.8+3.9−1.0	4+35−4	0.5+1.1−0.3	Rv3.1
EB-FUV 23	64+4−12	43.+91.−29.	8+36−6	3.7+8.5−2.6	...	...	Rv3.1
EB-FUV 24	46+1−2	4.0+5.9−2.0	22+24−17	1.6+2.6−0.8	6+40−6	0.3+0.5−0.2	Rv3.1
EB-FUV 25	100+2−15	410.+977.−290.	39+53−34	140.+366.−101.	5+78−5	11.+32.−8.	Rv3.1
EB-FUV 26	64+3−8	32.+42.−18.	11+27−8	3.9+5.5−2.3	...	...	Rv3.1
EB-FUV 27	132+4−21	1200.+2570.−822.	50+66−39	400.+932.−280.	8+88−8	45.+115.−33.	LMC2
EB-FUV 28	44+1−2	7.4+11.0−3.9	18+26−13	2.4+3.9−1.3	5+38−5	0.5+0.9−0.3	Rv3.1
EB-FUV 29	130+6−30	820.+2381.−610.	45+77−40	250.+795.−188.	5+102−5	17.+62.−14.	Rv3.1
EB-FUV 30	154+1−26	830.+2405.−580.	66+89−53	310.+1011.−224.	20+135−20	77.+277.−57.	Rv3.1
EB-FUV 31	29+3−10	18.+38.−12.	3+12−3	1.4+3.3−1.0	...	...	Rv3.1
EB-FUV 32	17+1−1	0.5+0.1−0.0	17+0−3	0.5+0.1−0.0	17+0−6	0.5+0.1−0.0	Rv3.1
EB-FUV 33	157+1−2	28.+16.−10.	98+39−27	16.+10.−6.	60+55−33	9.0+6.0−3.6	Rv3.1
EB-FUV 34	109+1−16	390.+1276.−223.	61+47−50	200.+721.−118.	30+77−30	90.+361.−55.	Rv3.1
EB-FUV 35	244+9−44	390.+1139.−291.	106+136−74	130.+418.−99.	39+194−39	40.+142.−31.	Rv3.1
EB-FUV 36	143+1−10	41.+70.−23.	80+62−42	21.+39.−13.	44+98−41	11.+21.−6.	Rv3.1
EB-FUV 37	129+1−1	8.9+1.5−0.7	122+6−12	8.4+1.5−0.7	114+14−23	7.8+1.5−0.6	Rv3.1
EB-FUV 38	36+1−16	13.+54.−10.	4+32−4	1.0+5.0−0.8	...	...	Rv3.1
EB-FUV 39	197+7−37	1500.+5370.−1179.	77+127−65	460.+1822.−367.	18+185−18	83.+369.−68.	Rv3.1
EB-FUV 40	28+1−3	2.3+3.9−1.2	7+19−5	0.5+0.9−0.3	3+24−3	0.2+0.4−0.1	Rv3.1
EB-FUV 41	64+20−31	38.+215.−33.	3+40−3	1.3+8.6−1.2	...	...	Rv3.1
EB-FUV 42	303+7−16	100.+97.−49.	168+76−68	40.+42.−21.	73+109−54	15.+17.−8.	Rv3.1
EB-FUV 43	24+4−7	6.0+11.2−3.9	2+4−2	0.8+1.6−0.5	...	...	Rv3.1
EB-FUV 44	51+1−1	7.5+3.5−2.4	25+15−10	3.2+1.6−1.1	6+18−3	0.5+0.3−0.2	Rv3.1
EB-FUV 45	24+2−4	1.7+1.4−0.8	3+3−2	0.2+0.2−0.1	...	...	Rv3.1
EB-FUV 46	84+1−9	220.+711.−100.	54+28−47	130.+485.−64.	33+49−33	76.+305.−38.	Rv3.1
EB-FUV 47	48+1−6	7.4+14.1−4.9	6+31−4	0.6+1.2−0.4	...	...	Rv3.1
EB-FUV 48	60+1−1	5.1+1.8−1.4	48+11−8	3.9+1.5−1.1	39+20−22	3.1+1.3−0.9	Rv3.1
EB-FUV 49	82+2−7	12.+14.−7.	27+27−22	3.5+4.3−1.9	3+30−3	0.3+0.4−0.2	Rv3.1
EB-FUV 50	374+9−26	460.+672.−274.	221+121−99	200.+321.−123.	107+192−85	75.+132.−48.	Rv3.1
EB-FUV 51	119+2−33	34.+159.−25.	46+75−43	11.+60.−9.	6+115−6	0.9+5.2−0.7	Rv3.1
EB-FUV 52	226+8−17	130.+129.−65.	100+64−42	46.+49.−24.	39+63−34	15.+17.−8.	LMC
EB-FUV 53	313+7−13	80.+57.−33.	175+60−53	32.+25.−14.	78+84−43	12.+10.−5.	Rv3.1
EB-FUV 54	113+4−17	34.+64.−22.	44+49−36	11.+24.−8.	5+69−5	0.9+2.0−0.6	Rv3.1
EB-FUV 55	123+1−10	15.+25.−9.	67+54−39	7.6+14.2−4.5	36+85−34	3.7+7.6−2.3	Rv3.1
EB-FUV 56	156+1−6	68.+67.−33.	88+54−36	35.+37.−18.	48+80−41	17.+20.−9.	Rv3.1
EB-FUV 57	64+1−8	440.+1041.−288.	25+38−21	140.+379.−98.	5+58−5	19.+55.−13.	LMC2
EB-FUV 58	143+1−7	95.+123.−50.	84+57−39	52.+74.−29.	48+94−43	27.+41.−15.	Rv3.1
EB-FUV 59	97+1−11	6.0+28.3−2.2	71+25−63	4.1+22.1−1.6	50+46−50	2.7+16.6−1.1	Rv3.1
EB-FUV 60	84+1−4	150.+230.−76.	51+32−33	87.+142.−45.	27+56−25	41.+72.−22.	Rv3.1
EB-FUV 61	184+1−12	11.+24.−6.	124+59−73	6.6+15.4−3.4	74+108−69	3.6+9.4−2.0	Rv3.1
EB-FUV 62	31+1−1	3.3+0.5−0.5	17+3−3	1.5+0.3−0.2	8+7−3	0.6+0.1−0.1	LMC
EB-FUV 63	110+1−2	7.5+5.3−3.1	76+32−25	4.8+3.7−2.1	50+57−34	3.0+2.5−1.3	Rv3.1
EB-FUV 64	16+1−10	1.7+11.7−1.3	3+13−3	0.3+2.6−0.3	⋅⋅⋅	⋅⋅⋅	Rv3.1
EB-FUV 65	47+1−2	8.6+8.5−4.3	18+24−12	2.6+2.8−1.4	5+31−5	0.5+0.6−0.3	Rv3.1
EB-FUV 66	157+1−6	250.+590.−65.	130+25−77	200.+516.−55.	101+54−94	150.+439.−44.	LMC2
EB-FUV 67	21+1−1	1.0+0.7−0.2	16+4−10	0.7+0.5−0.1	11+9−8	0.4+0.4−0.1	Rv3.1
EB-FUV 68	2+1−2	0.5+1.9−0.4	⋅⋅⋅	⋅⋅⋅	⋅⋅⋅ .	⋅⋅⋅	Rv3.1
EB-FUV 69	45+1−7	6.8+14.5−4.6	6+31−4	0.5+1.2−0.4	⋅⋅⋅	⋅⋅⋅	Rv3.1
EB-FUV 70	39+1−1	2.3+2.7−0.0	39+0−24	2.3+2.9−0.0	39+0−35	2.3+3.2−0.0	Rv3.1
EB-FUV 71	40+1−5	14.+26.−9.	9+30−6	2.1+4.4−1.4	3+36−3	0.8+1.8−0.5	Rv3.1
EB-FUV 72	49+1−2	8.5+7.1−3.9	16+21−11	2.2+2.0−1.0	3+17−3	0.4+0.4−0.2	Rv3.1
EB-FUV 73	70+1−1	8.5+3.0−2.2	57+12−11	6.6+2.5−1.8	46+24−22	5.1+2.1−1.5	Rv3.1
EB-FUV 74	123+1−2	14.+10.−6.	84+37−27	9.0+6.8−3.9	56+65−38	5.6+4.6−2.5	Rv3.1
EB-FUV 75	32+1−6	93.+331.−42.	13+17−11	30.+120.−14.	5+25−5	8.9+40.2−4.4	LMC2
EB-FUV 76	116+1−14	15.+57.−5.	84+30−67	10.+45.−4.	60+54−60	6.8+33.7−2.7	Rv3.1
EB-FUV 77	6+1−1	0.2+1.1−0.0	5+0−5	0.2+1.2−0.0	5+0−5	0.2+1.3−0.0	Rv3.1

Notes. ^aFor all sources, a foreground extinction of E(B − V = 0.22 was applied with MW-type dust, and the additional internal extinction (from the total E(B − V) given in Column 9 of Table 1) using three different dust types: "Rv3.1" stands for average MW extinction with Rv = 3.1 (Cardelli et al. 1989); "LMC" indicates the average LMC extinction for stars outside LMC2 by Misselt et al. (1999); "LMC2" indicates that the extinction curve by Misselt et al. (1999) was used. ^bCorresponding to the IDs in Table 1 and the blue labels in Figure 1. ^cThe results for the extinction curve given in this column were adopted in the analysis. Only for source 6, we adopted an SMC-type extinction curve, which gives age = 4⁺¹¹⁸₋₄ Myr and mass = 11⁺⁴³₋₈ × 10³ $\cal M_{\odot }$.

Download table as:  ASCIITypeset images: 1 2

For the youngest SF regions, a UV-steep extinction curve brings ages from integrated measurements into agreement with resolved H-R diagram results. These regions are associated with strong Hα emission (plotted with circles in Figures 3 and 4) and typically have high FUV surface brightness, i.e., the SF is intense and the complex is very compact. Therefore, we assumed UV-steep extinction curves to correct for internal extinction of regions with surface brightness higher than 25.0 mag arcsec⁻², and MW-type extinction for the rest of the sample. Among the high surface brightness SF regions, we found "LMC2'-type extinction to be preferable for sources with $\rm FUV<17.5$ mag, and the less steep "LMC" extinction to be better for sources fainter (in FUV) than this limit.

Zoom In Zoom Out Reset image size

We stress that these criteria were derived ad hoc, to bring the results from integrated measurements into agreement with detailed studies of a subsample. HST photometry of sample regions for a wider sample of Local Group galaxies will be used to expand this comparison (Bianchi et al. 2010, 2011). However, the results are not surprising: for example, in a study of SF sites in M51, Calzetti et al. (2005) found starburst-like extinction to be applicable only to sites with the strongest star formation.

Two metallicity values, Z = 0.004 and Z = 0.008, were explored, encompassing the estimated metallicity of young stars in NGC 6822, Z = 0.006 (Muschielok et al. 1999; Venn et al. 2001). Models with metallicity Z = 0.008 yield slightly younger ages and consequently lower masses of the SF regions. The effect of metallicity on the derived age and mass varies with age and extinction type, as discussed by Kang et al. (2009; see their Figure 12), Bianchi (2009; in particular Figure 9), and Bianchi (2011).

Ages and masses of the SF regions, derived from integrated FUV, NUV photometry using models with metallicity Z = 0.004, and assuming three different extinction curves for internal extinction, are given in Table 3; the last column indicates which values are adopted in our analysis. These values are plotted in Figure 4. The gap in the age distribution at 10 Myr is due to a slight degeneracy of the UV color in the range 6–11 Myr, due to RSGs emission (Fall et al. 2009). The uncertainty in the derived ages caused by this effect is smaller than the uncertainty introduced by the scatter in E(B − V), and it does not affect our overall results significantly. We stress that the masses of the individual SF regions should not be interpreted in terms of the cluster mass function for two reasons. First, we defined irregular, mostly unbound, complexes. Second, we used a constant threshold throughout the galaxy, to derive source contours, in the interest of adhering to an objective criterion and a consistent flux limit. This choice inevitably may cause some sparse regions to break into subcomponents (but their ages, and the masses of each, would not be affected), and more diffuse regions to merge.

The magnitude limit of our sample, from the 3σ detection threshold of 26.8 mag  arcsec⁻² and the area cut of ⩾150 arcsec², translates (using the SSP models) into a mass detection limit increasing with age, shown with a black line in Figure 4 for a foreground reddening E(B − V = 0.22 mag. The actual mass limit of the sample is higher because most sources have a higher reddening. As can be expected, the detection limit causes incompleteness for low masses at old ages.

For comparison, we also estimated the masses of the FUV-selected SF regions from resolved stellar photometry. The mass of each SF region was derived by extrapolating the number of stars above 10 $\cal M_{\odot }$ (corresponding to about B2V, and chosen with M_V <−2.5 and (B − V)₀ < −0.2) and up to the most massive star still on the MS, to the interval 0.1–100 $\cal M_{\odot }$. We assumed an IMF with α = 2.3 in the range $0.5\,\cal M_{\odot } < M < 80$ $\cal M_{\odot }$, and α = 1.3 for $0.1\,\cal M_{\odot } < M < 0.5$ $\cal M_{\odot }$ after Kroupa (2001). The masses derived from the H-R diagrams agree with those from integrated UV photometry for SF regions younger than 10 Myr. For older regions, the H-R diagrams give lower masses than the integrated measurements, by up to a factor of 20, probably because the most massive stars have evolved. The most discrepant regions have large areas and are in the main body of the galaxy, where the background is higher; they may be the result of merging of nearby regions expanding with age.

The possible contribution by foreground MW red dwarfs to the integrated GALEX flux of the SF regions was also estimated, since the density of foreground stars is significant in the direction of NGC 6822. We used our stellar model grids to estimate the possible contribution to the FUV and NUV fluxes from foreground stars of intermediate colors (0.1 < (B − V)₀ < 1.2 and M_V < −2; see, for example, Figure 6 of Bianchi & Efremova 2006). The derived potential effect on the FUV–NUV color is very small (−0.014 mag on average) and does not affect the results.

The main concerns to be addressed when deriving ages and masses of SF complexes by comparison with SSP model colors are: (1) the degree of coevality of the stellar complex and applicability of the SSP assumption and (2) stochastic effects from the top-IMF small number statistics. The latter affects both the analysis of integrated measurements and of resolved stellar counts. However, it affects only the small mass clusters. As we explained above, we restricted our analysis sample so to include sources with ⩾5 massive stars, in order to minimize problems of stochasticity. More importantly, we do not interpret our results in terms of mass distributions of individual clusters; instead, we add the masses of SF regions in broad age bins (next section) in order to obtain the total stellar mass formed at different epochs. In this way, we derive information on global star formation with broad time resolution, and stochastic effects on individual cluster masses average out. As for the assumption of "SSP" (or instantaneous star formation of each region), we tested the results by also using models with exponential SFH, decaying over short timescales: the results showed no appreciable difference. The measured FUV–NUV color of most sources is incompatible with CSP ("continuous star formation stellar populations") model colors of any age, ruling out the CSP assumption often used to derive global galaxy SFR, as not applicable to the individual SF regions in our sample. Strict coevality is not observed even in bona fide globular clusters, the epitome of "single age" stellar population, therefore some degree of uncertainty is carried in all works by this assumption, which is, however, the most compatible with the observed properties of our young population sample.

The SF regions have ages ≲400 Myr, as derived in the previous section from UV photometry, due to the FUV selection, and their masses range from 2.0 × 10² $\cal M_{\odot }$ to 1.5 × 10⁶ $\cal M_{\odot }$, when individual extinction correction is applied to the sources, as explained in Section 3.2.

We added the UV-derived masses of the SF regions in four age ranges to estimate the average SFR within these time intervals. We find: SFR =1 × 10⁻² $\cal M_{\odot }$ yr⁻¹ (2–10 Myr), SFR =1.5 × 10⁻² $\cal M_{\odot }$ yr⁻¹ (10–100 Myr), SFR =4.4 × 10⁻² $\cal M_{\odot }$ yr⁻¹ (100–200 Myr), and SFR =1.4 × 10⁻² $\cal M_{\odot }$ yr⁻¹ over the intire 2–100 Myr range. The results are shown in Figure 5; the uncertainties, shown as gray boxes, take into account the photometric errors and the E(B − V) scatter within the SF regions, which is typically one order of magnitude larger than the photometric errors. The uncertainties are up to a factor of four of the derived values. Additional (smaller) uncertainties may arise from the assumption that the stellar population in each SF region has formed in a single burst, and from the adopted IMF. Coevality is supported by the H-R diagrams of seven young SF regions, studied with HST (Bianchi et al. 2001; Bianchi & Efremova 2006), but it may be more questionable for older larger regions, which may result from the merging of subcomponents dissolving with time. Stochastic effects for low-mass complexes, as discussed previously, could affect individual masses, but average out when masses of several clusters are summed in broad age bins. Moreover, the total mass is dominated by the most massive SF regions.

Zoom In Zoom Out Reset image size

Figure 5 illustrates also how the adopted type of interstellar extinction affects the results. It shows the SFR estimated using the extinction correction as derived in this work for individual regions, as well as by assuming that the total reddening for all sources is from Milky Way-type dust (green lines), as was adopted by, e.g., Wyder et al. (2007) and Hunter et al. (2010). The resulting SFR is significantly lower in the most recent epoch, because the most massive SF regions are shifted to older ages when we use this extinction curve.

For older ages, an incompleteness at low masses sets in, driven by our flux detection limit (see Figure 4). Our FUV-flux threshold translates (using our SSP models) into mass limits of ∼60 $\cal M_{\odot }$, ∼170 $\cal M_{\odot }$, and 2900 $\cal M_{\odot }$ at 5 Myr, 10 Myr, and 100 Myr, respectively, if only foreground extinction was present. The actual limit is higher since most young stellar populations have additional internal extinction. The average detection limit in the 2–10 Myr bin is about 100 $\cal M_{\odot }$. The cluster mass function derived by Lada & Lada (2003) for embedded clusters in the solar vicinity has an exponent ≈−2 down to cluster masses of about 50 $\cal M_{\odot }$, then it drops. If Lada & Lada's results for solar vicinity were applicable to this dwarf irregular galaxy, the contribution of small mass clusters (50–100 $\cal M_{\odot }$) to the total mass would be about 10% in this age bin, where we detect clusters with masses up to 4 × 10⁴ $\cal M_{\odot }$. However, star formation in NGC 6822 is patchy and may not resemble that of a massive spiral galaxy (Bianchi et al. 2001, 2010, 2011); in any case, this estimate should be regarded as an upper limit since the majority (∼90%) of the embedded clusters are expected to merge and become part of larger OB associations or field stars before they are 10 Myr old (Lada & Lada 2003). At older ages, the mass function for embedded clusters is no longer applicable: according to Lada & Lada (2003), after 100 Myr, 94% of the embedded individual clusters have merged into large OB associations (if there was no cluster disruption, and if the above mass function were applicable, our detection limit would miss 45% of the clusters' mass at this epoch: this approximate figure can be taken as a conservative upper limit). The sum of the masses of our SF regions with ages between 100 and 200 Myr yields SFR ⩾0.044 $\cal M_{\odot }$ yr⁻¹ in this period, in agreement with the study of stellar populations by Gallart et al. (1996), who found SFR =0.04 $\cal M_{\odot }$ yr⁻¹ in this epoch (adopting E(B − V) = 0.24 mag). However, our value should be considered as a lower limit on the SFR because the FUV detection threshold corresponds to a high-mass detection limit for old populations. In addition, the uncertainty factors in the UV-based method, discussed previously, become significant at ages older than ∼100 Myr.

We point out that our photometry of individual regions aims at isolating young SF complexes, and the older, diffuse galaxy population is subtracted from the flux. Therefore, it would not be appropriate to compare the total flux from the SF region photometry with integrated galaxy models assuming a global SFH (see also Section 2.3).

We also estimated the SFR of NGC6822 in very recent epochs from the Hα emission (Section 2.4). The total Hα flux from H ii regions measured in this paper (the sum of the flux of the sources in Table 2) is $F_{\rm {H\alpha }}=1.8\pm 0.2\times 10^{-11}$ erg s⁻¹ cm⁻². After correcting the flux of the individual sources for reddening, using E(B − V) values from the associated or nearest FUV sources, and A$_{\rm {H\alpha }}/\mbox{$E(B-V)$}=2.5$, the total unreddened flux is $F_{\rm {H\alpha }}(\rm unreddened)= 4\pm 1 \times 10^{-11}$ erg s⁻¹ cm⁻², corresponding to a luminosity of $\log L_{\rm {H\alpha }}=39.1\pm 0.2$ erg s⁻¹. The uncertainty takes into account photometric errors and E(B − V) scatter within individual SF regions. Most of the Hα emission (∼70%) comes from Hα regions 5, 15, and 26 (see Figure 1 and Table 2). These include the H ii regions Hubble V and Hubble X, where an excellent agreement was found (under the assumption of optically thick gas) between the Hα luminosity and the number of ionizing photons estimated using resolved HST photometry (Bianchi et al. 2001; Bianchi & Efremova 2006) and VLT spectroscopy (B. V. Efremova et al. 2011, in preparation). The Hα luminosity translates into SFR =0.008 ± 0.003 $\cal M_{\odot }$ yr⁻¹ using the calibration by Panuzzo et al. (2003), which is based on the same SSP models we used to analyze the UV fluxes, the difference between the case (models) with and without dust being less than the uncertainty. Using other calibrations, we obtain similar results: for example, SFR =0.01 ± 0.003 $\cal M_{\odot }$ yr⁻¹ if we use the calibration by Hirashita et al. (2003), with f = 1. Among our Hα-defined H ii regions, there is one (FUV source EB-FUV 52, Hα source EB-Hα 25) which includes only two blue stars (an O-type star and an early B-type supergiant, according to our VLT spectroscopy). This gives an indication of the sensitivity of our Hα source detection. In general, Hα-based SFRs should always be regarded as a lower limit, due to the possibility of leakage of ionizing photons.

Hα emission traces the hottest, most massive stars, and this estimate is in good agreement with the UV-derived SFR in the recent 2–10 Myr, as we would expect. Our result is also in agreement with previous estimates of SFR based on Hα luminosity: SFR =0.01 $\cal M_{\odot }$ yr⁻¹ by Hunter & Elmegreen (2004), and SFR =0.016 $\cal M_{\odot }$ yr⁻¹ by Cannon et al. (2006).

Star formation more recent than 1–2 Myr is embedded in dust and detectable by 24 μm dust emission rather than in the UV, where the stellar flux is still heavily obscured. We estimated the SFR from the 24 μm emission using the Spitzer/MIPS measurements by Cannon et al. (2006), and the second term in Equation (D10) of Leroy et al. (2008). We derived SFR (24 μm) =0.02 $\cal M_{\odot }$ yr⁻¹; this value is shown with a blue line in Figure 5, over a very short time interval, since IR dust emission typically traces the youngest populations, where the massive stars have not yet dissipated the dust of the parent cloud. Therefore, the far-IR detected star formation should complement the stellar mass of young populations detected from FUV measurements.