Supermassive black holes and their host spheroids II. The red and blue sequence in the M_(rm BH) - M_(rm *,sph) diagram

In our first paper, we performed a detailed (i.e. bulge, disks, bars, spiral arms, rings, halo, nucleus, etc.) decomposition of 66 galaxies, with directly measured black hole masses, $M_{BH}$, that had been imaged at $3.6~\mu m$ with Spitzer. Our sample is the largest to date and, for the first time, the decompositions were checked for consistency with the galaxy kinematics. We present correlations between $M_{ BH}$ and the host spheroid (and galaxy) luminosity, $L_{sph}$ (and $L_{gal}$), and also stellar mass, $M_{*,sph}$. While most previous studies have used galaxy samples that were overwhelmingly dominated by high-mass, early-type galaxies, our sample includes 17 spiral galaxies, half of which have $M_{BH} < 10^7~M_\odot$, and allows us to better investigate the poorly studied low-mass end of the $M_{BH} - M_{*,sph}$ correlation. The bulges of early-type galaxies follow $M_{BH} \propto M_{*,sph}^{1.04 \pm 0.10}$ and define a tight red sequence with intrinsic scatter $\epsilon = 0.43 \pm 0.06~dex$ and a median $M_{BH}/M_{*,sph}$ ratio of $0.68 \pm 0.04\%$, i.e.~a $\pm 2\sigma$ range of 0.1-5%. At the low-mass end, the bulges of late-type galaxies define a much steeper blue sequence, with $M_{BH} \propto M_{*,sph}^{2-3}$, indicating that gas-rich processes feed the black hole more efficiently than the host bulge as they coevolve. We additionally report that: i) our Sersic galaxy sample follows a less steep sequence than previously reported; ii) bulges with Sersic index $n<2$, argued by some to be pseudo-bulges, are not offset to lower $M_{BH}$ from the correlation defined by the current bulge sample with $n>2$; and iii) $L_{sph}$ and $L_{gal}$ correlate equally well with $M_{BH}$, in terms of intrinsic scatter, only for early-type galaxies - once reasonable numbers of spiral galaxies are included, the correlation with $L_{ sph}$ is better than that with $L_{gal}$.