Numerical simulation of the rupture process is usually performed under an a
ssumption of scale invariance of the friction process although heterogeneou
s fault properties are shown by both direct observations of surface crack g
eometry and slip inversion results. We investigate if it is possible to def
ine an effective friction law for a finite fault with a small-scale heterog
eneity, that is, with a distribution of narrow segments with a resistance t
o rupture higher than the rest of the fault. We consider a model where the
local boundary condition corresponds to a linear slip-dependent friction la
w. We define the effective slip-dependent friction law by analogy with the
theoretical spectral solution for the initiation phase in the case of a hom
ogeneous infinite fault. We use finite difference simulations to test the v
alidity of this approach. The results show a surprisingly good agreement be
tween the calculations for the complete heterogeneous fault model and for a
homogeneous fault with an effective friction law. The time of initiation a
nd the average of the slip velocity on the fault are well predicted by the
effective model. The effective friction law exhibits a nonlinear slip depen
dence with an initial weakening rate different from the one of the local la
ws. This initial weakening rate is related to the geometry of the heterogen
eity and can be obtained by an eigenvalue analysis. The effective law shows
a kink at a slip that corresponds to the average slip on the fault for whi
ch the stress concentration of the strong segments is sufficient to trigger
their rupture. While based on a rather simple model of a fault, these resu
lts indicate that an effective friction can be defined and used for practic
al purposes. The heterogeneity of a fault tends to decrease the initial wea
kening rate of the local weak patches. Since the initial weakening rate con
trols the initiation duration, this last point indicates that the duration
of initiation expected from actual heterogeneous faults is much larger than
the one deduced from small-scale laboratory measurements. The actual fract
ure energy is not conservative in the rescaling of the friction law.