3 Results of the photometry

Contour plots in K and the light distribution in each band, for each galaxy, are shown in Fig. 1. North is left and East is down. The outermost isophote corresponds to a signal-to-noise ratio of 1.75 (about 22 mag arcsec^-2 for each BCDG) and the spacing of the contours is 1. in units of signal-to-noise ratio. The bar represents 1. kpc scale, at the assumed distance using H₀ = 80 Mpc km^-1 s^-1. The other plots show the color distributions for J-K, J-H, and H-K, in individual graphs for clarity. The last 3 graphs show the light distributions for J (stars), H(squares) and K (diamonds), as functions of the equivalent radius (in arcsec), of the reduced scales r^1/4, and in logarithmic scale.

 

 

Table 2: Results of the near infrared surface photometry

Name	mK	$\mu_{\rm eff}$[K]	$< \mu_{\rm eff} >$[K]	$r_{\rm eff}$[K]	0.25[K]	C21[K]	mK(8'')
	J-K	[J]	[J]	[J]	[J]	[J]	J-K(8'')
	H-K	[H]	[H]	[H]	[H]	H-K(8'')
Haro 14	10.80	21.68	18.53	14.0	6.3	2.23	13.05
	0.96	19.77	18.75	10.0	5.42	1.84	0.78
	0.47	19.12	18.15	9.4	4.9	1.95	0.18
Mk 996	11.64	20.03	18.64	10.0	3.6	2.81	13.13
	1.16	20.00	18.75	6.2	3.0	2.05	0.89
	0.74	19.55	18.32	6.2	2.6	2.37	0.42
Mk 600	14.39	19.50	18.93	3.2	1.84	1.75	15.19
	-1.04	21.16	20.22	9.6	4.8	2.0	0.0
	-1.23	21.05	19.97	9.2	5.0	1.84	0.02
Fairall 301	10.99	20.43	19.66	21.64	9.81	2.21	13.50
	1.25	20.71	19.98	16.35	6.42	2.55	0.85
	0.57	21.24	19.62	14.11	5.83	2.42	0.29
Tol 0610-387	12.36	22.32	19.78	12.2	5.9	2.05	14.48
	0.60	23.09	20.85	14.8	6.6	2.25	0.62
	-0.28	22.68	20.06	15.7	8.9	1.76	0.18
Tol 0645-376	13.45	20.31	19.62	6.87	3.66	1.88	14.62
	1.16	21.20	21.04	12.24	6.47	1.89	1.23
	-0.36	22.23	20.52	7.72	3.06	2.52	0.42
Tol0954-293	12.90	20.27	19.26	7.50	3.14	2.39	14.96
	0.77	20.57	20.52	9.59	3.80	2.52	0.98
	-0.25	21.89	19.55	9.41	3.55	2.65	-0.02
Tol 0957-279	10.56	20.05	18.95	19.0	10.7	1.77	13.55
	1.34	22.09	20.65	22.4	9.8	2.28	1.05
	1.15	20.28	19.64	15.4	6.6	2.32	0.03
Tol 3	10.78	18.38	17.39	8.3	4.2	1.99	12.25
	0.37	19.39	18.18	10.2	5.2	1.95	3.57
	-0.38	19.60	18.18	14.4	6.0	2.42	0.55
UM 465	10.37	18.30	17.42	10.29	7.74	1.33	12.06
	1.64	19.14	18.75	8.64	4.11	2.10	4.01
	1.19	19.60	18.23	9.93	4.51	1.98	1.03
UM 465B	12.43	18.71	17.83	4.81	3.37	2.03	13.56
	2.14	20.52	19.61	4.84	2.05	2.36	2.02
	1.52	20.58	19.37	4.07	2.05	1.98	1.41
UM 461	14.16	20.33	19.71	5.14	2.88	1.78	15.05
	0.99	21.07	21.31	10.49	6.06	1.73	1.26
	-0.68	22.16	20.58	6.82	3.49	1.95	0.56

Table 2 shows the results of the surface photometry:

Column 1: Object Name.
Column 2: Asymptotic K magnitude, J-K and H-K asymptotic colors, derived by isophotal integration and curve of growth methods.
Column 3: Effective surface brightness in K, J and H, in mag arcsec^-2.
Column 4: Mean effective surface brightness in K, J and H within the half light radius $r_{\rm eff}$.
Column 5: Half light radius in arcsec in K, J and H.
Column 6: Radius at a quarter of the total luminosity in arcsec.
Column 7: Concentration index in K, J and H, defined as the ratio between r_0.25 and $r_{\rm eff}$ (de Vaucouleurs & Aguero 1973).
Column 8: Integrated magnitude in K, J and H within a 8''aperture, for comparison with previous results (see below).

The magnitudes, and colors presented in this table, and on the plots of Fig. 1 are not corrected for galactic extinction.

3.1 Note on infra-red morphology

In Papers I and II, we found that the BCDGs could be classified into two groups depending on their light profiles: those dominated by an exponential law, and those dominated by an r^1/4 law. In the near-IR, two morphological types of BCDGs stood out: those with structures (e.g. Tololo 3, Fig. 1i) and those without (e.g. Mk 996, Fig. 1b). In the case of the former, most off-centered structures are associated with the star formation regions. Whether those structures would remain or not after the starburst has faded, is unclear. The presence of resolved sub-structures is mostly independent of the light profiles. The substructures seen in our BCDGs are generally smoother in the near-IR than in the optical bands (PapersI and II).

Some BCDGs which have regular envelopes in the optical do not have regular envelopes in the IR (e.g. Tololo 0610-387, Mk 600 Figs. 1e and 1c respectively). These irregularities could be due to the detection limit in the near-IR because the envelopes in the optical of these BCDGs are faint (below 24 mag arcsec^-2 in the Bband).

The main point is that BCDGs, in the near-IR, still do not look like dEs. The internal substructures are still visible, and obviously associated with the star forming regions. The case of Tololo 0610-387 reveals that these structures could in part be due to the presence of red supergiant stars, while in other BCDGs the infrared excess is most probably due to the contribution of the emission lines such as Br$\gamma$.

3.2 Color distributions

3.2.1 Integrated colors

The observed mean asymptotic color of our sample is red, $J-H=0.74\pm 0.47$, $J-K=0.95\pm 0.78$, $H-K=0.21\pm 0.85$and $B-K=3.03\pm 0.89$. The large spreads are due primarily to Mk 600, UM 461 and Tol 0645-376. These galaxies show very peculiar structures similar to those of mergers and gravitationally interacting galaxies. They are very blue in the visible spectrum, clearly dominated by the starburst emission. For the other BCDGs, comparison with infrared colors of the different stellar types (Johnson 1966) indicates the presence of a dominating evolved stellar population, likely red giants of K and M types ($J-K\sim$ 0.64, and $B-K\sim 3.40$ for a K0 red giant star). Figure 2a shows the color distribution of our sample.

Since the only available other photometric measurements for BCDGs have been done using an 8'' aperture (Thuan 1983; Hunter & Gallagher 1985), we simulated aperture photometry on the images of our BCDGs in order to compare our data with the results of those authors. We have carefully followed their procedure, i.e. searching for the brightest knot appearing on the visible image of the BCDGs and centering the aperture on it. The distribution of the colors in 8'' is shown in Fig. 2b, the arrows representing the mean values obtained by Thuan (1983) and Hunter & Gallagher (1985). The mean colors of our sample are even redder than theirs: $(J-K)_{8''}=1.44\pm 1.19$, $(J-H)_{8''}=1.02\pm 0.96$ and $(H-K)_{8''}=0.42\pm 0.43$. These high values are due to three of the BCDGs: Tol 3, UM 465A and UM 465B. Tol 3 presents two bright knots of star formation, the brightest in the visible being somewhat redder, it is centered with respect to the outer isophotes (both in the visible and in the infrared). This knot appears to be the nucleus of the host galaxy, and is composed of a significant fraction of evolved stars, both young red supergiants and old red giants (Terlevich et al. 1990). The same remarks apply to UM 465B because only the color indexes involving K band are affected. The red colors of UM 465A could be due to the presence of a dust lane visible across the central parts, as indicated by the high Balmer deficit observed: H$\alpha$/H$\beta$ = 4.10 (Terlevich et al. 1991). When the three objects are removed from the statistics, our mean values compare statistically well with those of Hunter & Gallagher and Thuan.

Figure 2: a) Asymptotic-color distributions: open B-K, loose stripes J-H, dashed J-K and filled H-K; b) integrated colors in an 8'' aperture: open B-K, loose dashed J-H, tight dashed J-K, and filled H-K. The arrows represent the mean values obtained by Thuan (1983) and Hunter & Gallagher (1985)

We have studied the relations existing between the asymptotic color indexes. J-K, B-K and H-K are strongly correlated with each other, while J-H appears to be constant for the r^1/4 BCDGs studied here, independent of the values of the other color indexes. Figures 3a and 3b show the different relations. On each diagram, we have indicated the location of a A0V star and of a K0III red giant.

Figure 3: a) Color-Color diagrams in the infrared: B-Kvs. J-K, B-K vs. H-K and J-K vs. H-K. The locations of A0V and K0III spectral type star are plotted for comparison; b) color-color diagrams for B-K, J-K and H-K as functions of J-H (sensitive to metallicity). The locations of A0V and K0III spectral type star are plotted for comparison

Figure 4: a) Color-Color diagrams showing the excess in the Kband. The open circles represent the exponential BCDGs, the filled lozenges represent the r1/4 BCDGs. The lines plot the Color-Color diagrams of ZAMS (dashed) and red giant (continuous) stars for solar metallicity; b) B-K versus B-R color diagram. The open circles represent the exponential BCDGs, the filled lozenges represent the r1/4 BCDGs. The dashed line represents the color evolution of a single stellar population including the onset of the AGB stars (>108 years). The continuous line represents the same stellar population including the onset of the RGB stars (>109 years). The data were taken from Bressan et al. (1998)

The two photometric classes, r^1/4 (filled lozenge) and exponential (empty circle) BCDGs, show different relations. As seen in Figs. 3a and 3b, the excess in K is independent of the J-Hcolor index for the r^1/4 BCDGs. This is probably due to the fact that the J-H index is very sensitive to metallicity and age. If r^1/4 BCDGs are the products of mergers and strong gravitational interactions, then we expect to observe very mixed populations of stars and metallicities, including intermediate age components. On the other hand, if exponential BCDGs are the products of slow accretion of gas without significant disturbance of the system, we expect to observe a gradient in the relative importance of the young starburst with respect to the evolved stars. This is characteristic of the degeneracy encountered for the predictions of integral colors in stellar population synthesis models (Charlot et al. 1996).

3.2.2 Asymptotic giant branch stars

While the mean colors of our sample are generally compatible with the presence of cosmologicaly ($\sim$10¹⁰ years) old stars, we note that, in many cases, the individual colors show a significant excess in the K band with respect to the optical bands. Figure 4a shows the optical-near-IR color-color diagram, the filled lozenges are the r^1/4 BCDGs, the open circles are the exponential BCDGs. The dashed line represent the zero age main sequence while the plain line represents the red giant branch stars, both for a solar metallicity. The BCDGs show a systematic excess in the K band of at least 1 mag with respect to the optical colors. This excess is seen with the effective colors. The star forming regions do not show this systematic excess (see next section). The optical images do not show any significant extended nebular emission either (Paper II). The K excess extends over the whole galaxy, the stars responsible for this emission should thus be distributed more or less smoothly throughout the galaxy: i.e. they have had time to diffuse in the gravitational potential. It takes between 10⁸ to 10⁹ years for a dense star cluster to diffuse within the galactic potential. The only stars that have a life time large enough to still be shining strongly in the near-IR after 10⁸ years are the intermediate mass stars (i.e. with masses ranging between $2~M_\odot$and $10~M_\odot$). More specifically, asymptotic giant branch (AGB) stars. Following Girardi & Bertelli (1998), our excess in the IR color could be partly due to the AGB population, allowing us to set a lower limit to the age of our BCDGs of at least 10⁸ years.

The question now is whether AGB stars alone, with ages ranging between 10⁸ to 10⁹ years, are enough to account for the total Kemission. Again, following Girardi & Bertelli (1998), after 10⁹ years, there is a second increase of the IR colors due to the onset of the red giant branch stars (RGBs). Figure 4b shows B-K as function of B-R color-color diagram. As for Fig. 4a the filled circles are the r^1/4 BCDGs and the open circles are the exponential BCDGs. The dashed line represents the color evolution of a single purely AGB population, the plain line represents the evolution of a single stellar population including the low mass stars below $2~M_\odot$, both for low metallicity stellar populations. The tracks are derived from Bressan et al. (1998). The authors only give B-V and V-K colors as functions of time. To derive B-Rwe applied a systematic V-R of 0.3 typical for BCDGs (Telles & Terlevich 1997). After 10⁹ years the red giant star population becomes dominant. Unless there is a continuous production of AGB stars which would require a significant underlying continuous star formation in some of our BCDGs, RGB stars have to be present in order to reproduce our excess in the near infrared. The presence of a diffuse population of AGBs is also consistent with the apparent lack of metallicity gradients of the gas in many BCDGs, and a very low continuous star formation (Legrand et al. 2000), as they would contribute to enrich homogeneously the interstellar medium over large scales.

3.2.3 Star forming regions

We have studied the colors of the star forming regions in our 12 BCDGs. For each "blob'', its spatial extent was determined from the B-R map (Doublier et al. 1997, 1999). Variable-aperture photometry was performed on each blob, and the contribution of the underlying galaxy was systematically removed. We have corrected the colors both for Galactic extinction (Burstein & Heiles 1982) and from internal extinction derived from the spectroscopic data available in the literature (Terlevich et al. 1991; Izotov et al. 1994; Stasinska et al. 1996; Storchi et al. 1994). When the spectroscopic data were not available, we assumed a mean internal extinction for the H II regions of E(V-B)=0.2 mag. The metallicity is assumed to be the same for all star forming regions inside a given galaxy, since mixing is probably very efficient across the small scale-lengths of the BCDGs (Marconi et al. 1994); also measurements in Tol 3 show that the ratio between oxygen and hydrogen is very similar from one star forming center to another (Stasinska & Leitherer 1996), even though they are some 300 pc apart.

Figure 5 shows the color-color diagrams for the star forming knots for the index (J-K) versus (B-R) and (B-K), and for the index (B-K) versus (B-R). The solid line represents the variation of the color-color indexes as functions of time for a young starburst with a solar metallicity and a Salpeter Initial Mass Function (Cerviño & Mas-Hesse 1994). At lower metallicities (dashed line), the evolution is slightly shifted toward the lower-left corner since the metallicity affects the colors of the stellar populations such that at lower metallicity the stars appear bluer.

Figure 5: Color-Color diagram for the star forming regions. Each symbol represent one galaxy: circle Fairall 301, square Haro 14, triangle Mk 600, cross (+) Mk 996, cross ($\times$) Tol 0610-387, asterisk Tol 0645-376, star Tol 0954-293, crossed square (+) Tol 0957-279, crossed square ($\times$) Tol 3, lozenge UM 461, filled hexagon UM 465A and filled lozenge UM 465B. The solid line represents the evolution of a starburst of solar metallicity, with a Salpeter IMF between 0 and 20 Myr from the population synthesis models of Cerviño & Mas-Hesse (1994). The dashed line represents the evolution of the same starburst with a lower metallicity of $Z=0.08z_{\odot }$

Each knot associated with a given BCDG is represented with the same symbol. For the majority of BCDGs, the different knots are clustered in small areas of the diagram. Taking into account the observational errors, this indicates that the star forming regions are probably of the same age (coeval) and thus derive from the same "starburst'' event. However, for some BCDGs, the knots occupy different locations in the diagram. In the case of Mk 600, the differences in the location of the two knots indicate that the dominating populations in the knots differ in evolution.

   

Table 3: Mean values of the photometric parameters in the near infrared

	N$^{\rm o}$	$< \mu_{\rm eff} >$*	$< \mu_{\rm eff} >$	$< \mu_{\rm eff} >$	< $r_{\rm e}>$*	< $I_{\rm e}>$(K)	< B-K >*
		J mag/arcsec2	H	K	(K) kpc	$L_{\odot K}$/pc2
r1/4	5	19.77 (1.24)	19.32 (0.91)	18.71 (0.82)	1002 (1185)	4406 93762089	2.8 (1.1)
Exponen.	7	19.97 (0.99)	19.37 (0.89)	18.89 (0.97)	688 (344)	3732 91201527	3.2 (0.7)

(*): Standard deviation of the sample is in parenthesis.

Regions showing an excess both in R and K bands compared to the Jband are most probably dominated by red giants. Since their positions in the BCDGs appear to be centered with the outer isophotes, they could be the nuclei of the galaxies (Tol 3, UM 465B and UM 461). An excess seen only in the K band probably indicates the presence of young red supergiant stars, or dust-enshrouded star clusters.

A more complete analysis of the stellar populations in the star forming regions, and in the host galaxies will be given in a following paper.

3.3 Compactness index and mean surface brightness

The compactness index as defined by de Vaucouleurs & Aguero (1973) and Fraser (1977), $C_{21}=\frac{r_{0.5}}{r_{0.25}}$, varies little in our small sample: $<C_{21}>(K)=2.02\pm 0.37$, $<C_{21}>(J)=2.16\pm 0.26$ and $<C_{21}>(H)=2.15\pm 0.30$ (Fig. 6). The scatter in K is only due to the exponential BCDG UM 465 (face-on) with C₂₁(K)=1.33 and the r^1/4 BCDG Mk 996 with C₂₁(K)=2.89. Both galaxies represent almost the perfect photometric cases in K band (<1.9 for the spirals and >2.5 for the ellipticals, Fraser 1977). The otherwise small scatter in the values of C₂₁(K) around $<C_{21}>\ \sim$ 2.0 indicates that the starburst does not play a significant role in the stellar densities in our BCDGs. In the optical, we found that exponential BCDGs and r^1/4 had significantly different C₂₁values. Moreover, since the values for the concentration indexes are intermediate between "spirals'' and "ellipticals'' (Fraser 1977; de Vaucouleurs & Aguero 1973), it shows that BCDGs have more complex structures than previously believed.

Figure 6: Distribution of the compactness index C21 in each J, H and K band

The situation, however, is different in the J and H bands which are more sensitive to the young and intermediate age stars. C₂₁(H) shows a spurious bimodal distribution possibly due to the small sample size. The C₂₁(J) shows a flatter distribution.

Both bands are sensitive to the young main sequence stars of the starbursts, but also to the intermediate age population. In addition, the BCDGs for which C₂₁(H) is larger than 2 have only one star forming region, while the BCDGs for which C₂₁(H) smaller than 2 have at least 2 star forming regions. This could be simply a structural effect.

Similar remarks apply to the distribution of the mean surface brightness with the effective radius (Fig. 7: $<\mu_{\rm eff}>_K\ =18.81\pm 0.87$ mag arcsec^-2( $<I_{\rm eff}>\ =4~10^3\, L_\odot$/pc²), $<\mu_{\rm eff}>_J\ =19.88\pm 1.05$ mag arcsec^-2 and $<\mu_{\rm eff}>_H\ =19.35\pm 0.91$ mag arcsec^-2. Table 3 presents the mean values of our measured photometric parameters for the r^1/4 BCDGs and the exponential BCDGs, the photometric types being defined from the B band. Within the observed dispersion, the infrared values in both photometric classes are very similar, while they were clearly different from class to class in the visible (Paper II). If the near infrared properties reflect what we can expect the BCDGs to look like after the burst has faded, then their mean photometric properties will not convey enough information to make out a clear dynamical classification. It follows that once the starburst has faded away, the remaining stellar distribution will not allow a strict photometric classification of the light profiles.

 

 

Table 4: Absolute magnitudes and radii in the near infrared

Name	MK	MH	MJ	$R_{\rm eff}$[pc] (K)	$R_{\rm eff}$[pc] (J)	$R_{\rm eff}$[pc] (H)
Haro 14	-19.62	-19.15	-18.65	821	588	555
Mk 600	-16.12	-17.35	-17.16	197	586	562
Mk 996	-19.91	-19.17	-18.75	992	613	610
Tol 0610-387	-19.03	-19.29	-18.43	1113	1346	1432
Tol 0957-279	-18.04	-16.89	-16.70	483	571	391
Tol 3	-18.48	-18.86	-18.11	286	350	497
Fairall 301	-18.65	-18.08	-17.40	891	581	673
Tol 0954-293	-19.15	-19.40	-18.38	934	1171	1194
Tol 0645-376	-21.36	-21.72	-20.20	3049	3427	5430
UM 465A	-19.91	-18.72	-18.27	567	492	476
UM 465B	-17.85	-16.32	-15.71	265	225	267
UM 461	-15.60	-16.28	-14.61	223	296	455

3.4 Brightness distributions

As seen from a comparison between Fig. 1 of this paper and Fig. 1 of Paper II, the light profiles in the infrared are generally consistent with the light distributions observed in the B and R bands. In some peculiar cases, however, the brightness profiles in J and K differ significantly from the ones in the visible. Mk 996 shows a light distribution in J clearly dominated by an exponential law, while the profiles in H, K, Band R are dominated by an r^1/4 law. This galaxy thus presents two dynamically different components of different stellar population. The "exponential'' component is clearly dominated by the starburst, while the r^1/4 component appears to be the old part of the galaxy. This behaviour strongly suggests a decoupling between the different dynamical components of the galaxies. The internal exponential component could have been created from slow gas infall, while the r^1/4 old component could originate from the formation of the galaxy itself, i.e. during isothermal gravitational collapse, or an equal mass merger.

In Fig. 1, considering all objects with respect to what is seen at visible wavelengths, the structures are less pronounced, the central peak is less significant in the near infrared than in the visible. The color profiles show a striking property though: the J-K (open circles) and J-H profiles (filled lozenge) of the r^1/4 BCDGs decrease then increase with increasing radius, while the H-K profiles (open square) remain monotonic (e.g., Mk 996). The J-K and H-K profiles of the exponential BCDGs decrease monotonically while J-H decreases then increases with increasing radius (e.g. Haro 14). This implies that the distribution of the stellar populations varies from one photometric type to the other, i.e. the young stars being more concentrated in the center of the r^1/4 BCDGs than in the exponential BCDGs.

The colors of the central regions do not vary much from one BCDG to another ( $(J-K)_{\rm c} \sim$ 1.0, $(J-H)_{\rm c} \sim$ 0.5, and $(H-K)_{\rm c} \sim$ 0.), except for Mk 996, which has a dust lane obscuring partly the nucleus (Thuan et al. 1996), and Mk 600 which is very blue. The B-K color profiles reach a limiting value at the outermost Kisophote, similar for all our BCDGs: $B-K \sim$ 4.6, while the central value is similar for the all the BCDGs: $B-K \sim 2.6\pm 0.1$ (NOT corrected for Galactic extinction). Correcting the observed values for Galactic extinction, the $<B-K>\ \sim 1.7$ mag corresponds to the values measured for the star forming regions (see Fig. 7).

Figure 7: Distribution of the mean effective surface brightness in each J, H and K band

Figure 8: a) Magnitude - radius relation at the effective radius $r_{\rm eff}$, in pc in each band J, H and K. The continuous line represents the empirical -5 relation, the dotted line represents the least square fit to the data. The slope of the fit is indicated in the upper right corner; b) magnitude - radius relation at r0.75 in pc, in each band J, H and K. The continuous line represents the empirical -5 relation, the dotted line represents the least square fit to the data. The slope of the fit is indicated in the upper right corner

   

Table 5: Relation between absolute magnitude and radius in the bands J, H and K

	A	$\sigma$
r0.5
J	-4.88(0.46)	-3.87(0.13)
H	-4.60(0.37)	-4.27(0.09)
K	-4.64(0.33)	-3.90(0.31)
r0.75
J	-4.86(0.38)	-3.96(0.11)
H	-4.82(0.28)	-4.69(0.12)
K	-4.58(0.29)	-4.75(0.13)

A: slope of the relation.
$\sigma$: slope of the relation after 1 sigma rejection.