We report new Hf (and Nd) data for more than 100 sedimentary samples, recen
t to Archean in age, from a wide range of depositional environments. These
data document the behavior of Lu-Hf and Sm-Nd isotopic systems in the globa
l sedimentary system. In conjunction with existing data for mantle-derived
rocks, we now have reasonable constraints on coupled Hf-Nd isotopic behavio
r in the crust and mantle. Lu/Hf and Hf isotopic compositions are strongly
fractionated between muds and sands in passive margin sediments due to conc
entration of low Lu/Hf, low Hf-176/Hf-177, Hf-rich zircons in mature sands.
In active margin settings, Lu-Hf fractionation due to the 'zircon effect'
is minor due to the less weathered and more juvenile character of the sedim
ents. Nd isotopic compositions are not highly fractionated by sedimentary s
orting because heavy minerals, also rich in REEs, do not fractionate Sm-Nd
efficiently. The lack of a large and systematic fractionation at active mar
gins means that no significant Hf-Nd decoupling occurs here. This is import
ant because sediments at active margins are the most likely to be recycled
to the mantle. Hf-Nd isotopic data for all terrestrial samples fall along a
single coherent trend (epsilon(Hf) = 136 epsilon(Nd) + 2.95) which we call
the 'terrestrial array'. This array is composed of two complementary compo
nents: a mantle array (epsilon(Hf) = 133 epsilon(Nd) + 3.19, defined by all
oceanic basalts; and a crustal array (epsilon(Hf) = 1.34 epsilon(Nd) + 2.8
2), defined by sediments, continental basalts, granitoids, and juvenile cru
stal rocks. The similarity of the crustal and mantle arrays indicates that
no large-scale Hf-Nd decoupling occurs between the crust and mantle. The co
herency of the terrestrial Hf-Nd array implies mixing within the mantle, du
e to stirring processes, and also within the crust, due to homogenization b
y collective sedimentary processes. in addition, tight Hf-Nd covariation ma
y also imply that efficient crust to mantle recycling has modulated isotopi
c correlation in the silicate Earth. All Hf-Nd arrays, including the terres
trial array, lie significantly above (2-3 epsilon(Hf) units) the BSE (bulk
silicate Earth) reference. This would appear to require a hidden reservoir
in the Earth, heretofore unsampled, to balance the Hf-Nd isotopic compositi
on of the terrestrial array. However, the discrepancy between the terrestri
al array and BSE may simply be due to differences in the way the CHUR (chon
dritic uniform reference) values were determined for the Lu-Hf and Sm-Nd is
otope systems. (C) 1999 Elsevier Science B.V. All rights reserved.