Many of the oldest similar to 3.5 Ga komatiites have low Al2O3/TiO2 and dep
leted HREE, features that are attributed to garnet fractionation at some st
age during their formation. It is not known, however, whether this fraction
ation took place at the time of melting, or well before, perhaps during the
crystallization of an early magma ocean. We report here the Hf isotope com
positions of four Al-depleted komatiites from the 3.45 Ga Barberton Greenst
one Belt in South Africa and six Al-undepleted komatiites from 2.7 to 1.9 G
a belts in Canada and Zimbabwe. We also measured the Hf isotope composition
s of five tholeiites from Barberton and two tholeiites from the 1.9 Ga Cana
dian Belt, to provide information about the composition of ambient Precambr
ian mantle. AU the komatiites have high initial Hf-176/Hf-177 ratios. All i
nitial epsilon(Hf) values are positive, in the range of +2.6 to +7.8. Almos
t this entire range is also present in the 3.45 Ga Barberton komatiites (+2
.6 to +7.3). epsilon(Hf)(T) correlates with epsilon(Nd)(T) and defines a cr
ude linear array that coincides with compositions of both modern oceanic ba
salts and of Phanerozoic to Archean juvenile crustal rocks. The composition
of Bulk Silicate Earth [J. Blichert-Toft, E Albarede, Earth Planet. Sci. L
eu. 148 (1997) 243-258] plots below this array, at slightly lower epsilon(H
f) TO explain the positive initial epsilon(Hf) values of 3.45 Ga Barberton
komatiites, the source must have had high Lu/Hf. This implies a source enri
ched in garnet, a result contrary to earlier models in which the low Al2O3/
TiO2 and low Lu/Hf of this type of komatiite was thought to be inherited di
rectly from mantle depleted in garnet. The low Al2O3/TiO2 and low Lu/Hf,re
features of the primary komatiite magmas, not of their mantle source, and r
esult from garnet fractionation during komatiite formation. To explain the
bias to high initial epsilon(Hf) values of the komatiites and other terrest
rial rocks, we consider two alternatives: (a) a reservoir with low Lu/Hf fo
rmed very early in Earth history; (b) accepted values for the Kf isotopic c
omposition of Bulk Silicate Earth are inappropriate and the actual composit
ion has higher Hf-176/Hf-177 and lower (LU)-L-176/Hf-177. A problem with th
e latter interpretation is that the proposed composition Lies off, the acho
ndrite meteorite isochron, which implies that the Earth has a different age
, or accreted from material with composition different from that of meteori
tes. (C) 1999 Elsevier Science B.V. All lights reserved.