G. Devaucouleurs, SPECIFIC FREQUENCIES OF GLOBULAR-CLUSTERS IN ELLIPTIC GALAXIES - A NEW TEST OF THE EXTRAGALACTIC DISTANCE SCALE, The Astrophysical journal, 415(1), 1993, pp. 33-39
The specific frequencies of globular clusters in elliptical galaxies,
S, being the ratio of their (extrapolated) total number, N(t), to the
absolute luminosity of the galaxy, depend on the adopted distance scal
e. If the distance is derived from the redshift and some assumed value
of h = H/100, the optimum value, h(m), may be derived for a sample of
normal E galaxies covering a sufficient range of distances by finding
the value of h (or of log h) which minimizes either the dispersion si
gma(S) or sigma(log S) or, better, the relative dispersion F(S) = sigm
a(S)/[S]. Application to a sample of nine E galaxies for which Harris
and van den Bergh calculated S for h = 0.50, 0.75, and 1.00, gives H(m
) = (85 +/- 4) km s-1 Mpc-1. This is also the value which optimizes th
e agreement between the different methods and minimizes [S] = 4.7 +/-
0.7. The mean Hubble ratio predicted by the original short scale (EDS
VII) for the same nine objects was [H] = 87 +/- 3 (Appendix A). Becau
se all nine objects are in the north Galactic hemisphere where the app
arent Hubble ratio is known to be about 20%-30% less than in the south
Galactic hemisphere, the corresponding all-sky average could be as hi
gh as H = 95, in close agreement with the short-scale value. An altern
ative use of the errors by means of linear relations and correlation c
oefficients with the Hubble modulus HM makes little difference to the
solutions. The frequency function of 28 solutions is very nearly Gauss
ian with mode [H(m)] = 86 +/- 2 and dispersion sigma(H(m)) = 12 (Appen
dix B). A suggestion to use weighted means is tested, but although agr
eeing in the mean (84.5 +/- 6) with the unweighted solutions, it leads
to a much larger scatter and is contraindicated (Appendix C). A large
r, more precise and more homogeneous collection of counts, magnitudes,
and extinction corrections over a bigger range of distances will be n
ecessary to better evaluate the errors and potential precision of the
method.