An I-band calibration of surface brightness fluctuation measurements at blue colours

Context. The surface brightness fluctuation (SBF) method is a powerful tool for deriving distances to galaxies for which single stars cannot be resolved. Up to now, the method has been calibrated mainly at red colours due to the intrinsic faintness of blue early-type galaxies. Aims. In this paper we address the I-band calibration of the SBF method at blue colours, the regime of dwarf elliptical galaxies (dEs). Methods. We present deep and wide-field VI photometry of the central Fornax cluster obtained at Las Campanas Observatory. With these data we perform an SBF analysis of 25 dEs in the range -16.5 < M-V < -11.2 mag, 0.8 < (V - I)(0) < 1.10 mag. Our colour calibration is accurate to better than 2%. For the calibration analysis we exclude eight dEs whose SBF measurement was affected by poor seeing (FWHM similar or equal to 1). Results. Our SBF data are inconsistent at the 3 sigma level with a colour independent absolute SBF magnitude M-I, presenting a problem for SBF models that predict such a flat relation. There is a weak indication in our data (1.8 sigma) to favour a two-branch calibration over a one-branch calibration with broad scatter in the M-I - (V - I) plane. We obtain the following one-branch empirical SBF calibration: M-I = -2.13(+/- 0.17) + 2.44(+/- 1.94) x [(V - I)(0) - 1.00] mag. We deduce a 0.34 +/- 0.14 mag cosmic scatter of M-I, which is significantly larger than found at redder colours. This agrees with those theoretical SBF models that predict MI is more sensitive to age-metallicity variations in the blue than in the red. We find evidence that the fainter galaxies in our samples contain younger and more metal-rich stellar populations than the brighter ones. The application of our empirical calibration to published SBF measurements of Hydra and Centaurus cluster dEs leaves the distances to both clusters unchanged.