Geoscience Reference
In-Depth Information
Garnet
10
3
T / Tm
0.9
0.8
0.7
0.6
10
−
5
s
−
1
YAG
GGG
YIG
Gross
Gross-Andra
Alm-Pyro
10
−
2
Spe
Pyro-Alm-Spe
10
2
10
−
3
4
5
6
10000/T(K)
7
8
9
(a)
10
−
2
10
−
4
YAG
GGG
YIG
Gross
Gross-Andra
Alm-Pyro
10
−
5
Spe
Pyro-Alm-Spe
1.1
1.2
1.3
1.4
Tm / T
1.5
1.6
1.7
10
−
3
1.0
1.1
1.2
1.3
(c)
T m / T
(b)
Fig. 4.15
Crystal structure-creep strength systematics (normalization of strain-rate is not made in these studies,
but the Debye frequency varies only modestly among different materials). (a) A comparison of creep strength of
various garnets as a function of temperature (Karato
et al
., 1995a). (b) A comparison of creep strength of garnet after
normalization (Karato
et al
., 1995a). (c) crystal structure-plasticity systematics (modified from Karato, 2011a).
Reproduced with permission of Elsevier.
these results,
Recently, new results on clean olivine aggre-
gates were published showing substantially dif-
ferent creep strength (Faul & Jackson, 2007). In
these clean samples, prepared by the sol-gel syn-
thesis process fromoxides, the rate of deformation
by diffusion creep is substantially lower than that
in San Carlos olivine. It is possible that this dif-
ference is caused by a small amount of melt
that exists in nominally ''melt-free'' San Car-
los olivine aggregates as suggested by Takei and
Holtzman (2009a). However, other factors such
as the impurity content may also be the cause for
such a difference.
Markedly different results were published by
Jin
et al
. (1994) who reported that the addition of
ε
(
φ
)
=
ε
(0)
·
exp (
αφ
)
(4.19)
where
φ
is the melt fraction and
α
is a nondi-
mensional constant (
∼
20-30 (
α
≈
25 for diffusion
30 for dislocation creep))
7
.With1%
of melt, strain-rate is enhanced by
creep and
α
≈
∼
30%.
7
This is an empirical formula without strong theoretical
basis. The asymptotic behavior for
φ
→
0 does not agree
with a model of equilibriummelt geometry reviewed by
Kohlstedt (2002), and the another asymptotic formula
for
φ
→
1 does not make sense because, at that limit,
the strain-rate must agree with that of a viscous fluid
but this formula does not have the viscosity of liquid
(see Karato, 2008).
