The radio/X-ray domain of black hole X-ray binaries at the lowest radio luminosities

We report on deep, coordinated radio and X-ray observations of the black hole X-ray binary X IE J1118+480 in quiescence. The source was observed with the Karl G. Jansky Very Large Array for a total of 17.5 h at 5.3 GHz, yielding a 4.8 +/- 1.4 mu Jy radio source at a position consistent with the binary system. At a distance of 1.7 kpc, this corresponds to an integrated radio luminosity between 4 and 8 x 10(25) erg s(-1), depending on the spectral index. This is the lowest radio luminosity measured for any accreting black hole to date. Simultaneous observations with the Chandra X-ray Telescope detected XTE J1118+480 at 1.2 x 10(-14) erg s(-1) cm(-2) (1-10 keV), corresponding to an Eddington ratio of similar to 4 x 10(-9) for a 7.5 M-circle dot black hole. Combining these new measurements with data from the 2005 and 2000 outbursts available in the literature, we find evidence for a relationship of the form l(r) = alpha+beta l(X) (where l denotes logarithmic luminosities), with beta = 0.72 +/- 0.09. XTE J1118+480 is thus the third system together with GX339-4 and V404 Cyg for which a tight, non-linear radio/X-ray correlation has been reported over more than 5 dex in l(X). Confirming previous results, we find no evidence for a dependence of the correlation normalization of an individual system on orbital parameters, relativistic boosting, reported black hole spin and/or black hole mass. We then perform a clustering and linear regression analysis on what is arguably the most up-to-date collection of coordinated radio and X-ray luminosity measurements from quiescent and hard-state black hole X-ray binaries, including 24 systems. At variance with previous results, a two-cluster description is statistically preferred only for random errors less than or similar to 0.3 dex in both l(r) and l(X), a level which we argue can be easily reached when the known spectral shape/distance uncertainties and intrinsic variability are accounted for. A linear regression analysis performed on the whole data set returns a best-fitting slope beta = 0.61 +/- 0.03 and intrinsic scatter sigma(0) = 0.31 +/- 0.03 dex.