We have applied the commonly used SFR estimators to our reference sample of star forming galaxies. Given that the estimators are all for the same IMF and stellar models we do not expect these aspects to introduce any scatter. Figure 1 shows the SFR(H), SFR(OII) and SFR(UV) plotted against the SFR(FIR). Clearly the sample shows a correlation plus a large scatter.
Figure 1. Standard SFR estimators vs. SFR(FIR). No extinction corrections were applied to the data. The solid line represents equal values.
To simplify the analysis and simultaneously make use of the fact that SFR(FIR) is probably the best SFR estimator available, we will use in what follows FIR normalized SFR, i.e. the SFR relative to SFR(FIR). The FIR normalized SFR(H), SFR(OII) and SFR(UV) are:
This is better seen in the distribution histograms of the normalized SFR as shown in Figure 2. As reference we included in parts a,d,g of the figure the normalized SFR computed using the observed luminosities. The central and right columns show the dust extinction corrected ratios using the MW and the Calzetti extinction laws respectively. The corrections were applied following the common methodology and are described in Appendix A.
Figure 2. Histograms of the SFR rates given by the different tracers normalized to the SFR(FIR). In the left panels (a, d, g) no corrections were applied to the data. In the central panels (b, e, h) we corrected the H, [OII]3727 and UV continuum by using the MW extinction curve. In the right panels (c, f, i) the H, [OII]3727 and UV continuum were corrected with Calzetti's extinction law. The median and standard deviation are given for each case. The number of objects is 29.
Our main conclusion is that irrespective of the extinction law applied, the SFR(H) is close to the SFR(FIR) while both SFR(OII) and SFR(UV) show a clear excess. The excess is much larger for SFR(UV) than for SFR(OII) suggesting a wavelength dependent effect, probably an extinction over-correction. Bearing in mind that our reference sample has a large fraction of low metallicity and low extinction galaxies this result suggests that applying these standard methods to estimate SFR will systematically overestimate the SFR in samples at intermediate and high redshifts where either SFR(OII) or SFR(UV) are used. This result is in apparent contradiction with what has been found and shown in Madau-type plots in recent years, where the SFR obtained from UV and optical data are much lower than that obtained from mm and sub-mm observations at intermediate and high redshifts. In order to reach agreement between both determinations, fixed (and somehow arbitrary) amounts of extinction have been applied to the UV/optical data, because at the moment, the intermediate and high redshift samples do not allow a reliable determination of the dust extinction. It is worth noting that Steidel et al. (1999) applied a fixed correction to their sample that is close to the average UV. On a positive note we should indicate that the application of the reddening corrections reduce considerably the scatter in all three normalized SFR estimators as we will show below.
It is important to clarify the origin of the detected excess in the extinction corrected OII and UV. There is one effect that is not normally taken into account, namely that the presence of an underlying young stellar population with deep Balmer absorptions will bias the observed emission line ratios towards larger Balmer decrement values, mimicking the dust extinction effect. In the next section we will recalculate the SFR for the different tracers but including an estimation of the effect of an underlying population. We will also estimate the effect of photon escape in UV.