Geos596F Subduction Zones: An Integrated View
Wednesday, September 22
Flat-Slab Subduction Zones (J. Guynn, notes by M. Anderson, edited by Zandt)
It seems that questions surrounding flat-slab subduction revolve around two dominant
areas: a) How does the flat slab affect the overlying mantle and crust? and b) Why is the
slab flat? This week we discussed the papers in order: Kay & Mpodozis (2001), van
Hunen et al. (2002), and a brief touch on Saleeby (2003); this summary is therefore
appropriately divided. We spent much time in discussion clarifying Kay's model for
explaining LREE trends in porphyry deposits in the forearc of the South American
subduction zone. This summary starts with the consensus and the moves into additional
questions.
Kay's paper presents a petrologic model that explains crustal thickening evident in
Sm/Yb measurements on porphyry deposits along arc segments of the South American
subduction zones. This model attempts to tie crustal thickening with stages of slab
flattening, steepening, and the associated inferred fluxes of water and melt into the crust.
Figure 3 gives a schematic summary of the steps involved in slab flattening and figure 6
gives a schematic representation of what is happening in these same stages in the crust.
Figure 4 gives the steps and results when the slab steepens again. The main steps are as
follows:
1. Normal subduction leads to arc volcanism through
relatively thin crust (figure 3, bottom; figure 6, panel 1).
Water escapes the system mostly through magmatism,
but plenty of water is stored in the system.
2. The slab starts to flatten and magmatism starts to
slacken (figure 3, middle; figure 6, panel 2). Because of
water and magmas stored at lower crustal levels, the
crust is weak and thickens. This increases pressure at
lower crustal levels, which triggers eclogitization that
releases water from amphiboles that can migrate to
higher levels. This process produces mineralization.
3. When the slab is flat, water is still released from the
slab, but because of cold temperatures and inhibited
magmatism, it is stored in the mantle in serpentinite, and
a little enters the lower crust (figure 3, top; figure 6,
panel 3). Most of the water has been released from the
lower crust and as a result the dominant phase is the
anhydrous phase garnet. Mineralization stops at this
point.
4. When the slab steepens (figure 4), asthenospheric
mantle enters and heats the super-hydrated former
mantle wedge, which produces a rush of volcanism
(ignimbrite flare-up). Apparently this re-hydration also
weakens the lower crust and is a catalyst for more
shortening and thickening of the crust.
Continuing questions and discussion:
1. In Mexico there is a flat slab and mineralization goes all the way inland, not just
in a narrow band of porphyry as in South America. How does this fit with the
above model? There are not good crustal thickness estimates for this region of
Mexico, so correlations with crustal thickness is difficult.
2. According to an alternative model, porphyries are thought to be a product of
adakitic volcanism and adakitic volcanism implies slab melting. This does not fit
with the above model. Kay would argue, along with others here at the U of A that
there is a significant mantle component of porphyries which cannot be explained
if it is the result of slab melting only. Also, there is a question about the proper
identification of adakites, therefore some "adakites" associated with porphyries
may not signify slab melting at all.
3. The geophysicists want to know whether the connection between Sm/Yb ratios
and crustal thickness is well-accepted in the geochemistry community. The
geochemistry contingent at this meeting thinks so. Sm is known to go into garnet
and garnet is an anhydrous mineral that occurs in larger quantities in the lower
crust of thicker crustal columns. There is some question about how quantitative
this measure of crustal thickness is. In addition, figure 3 of Saleeby (2003) shows
4.
5.
6.
7.
mantle xenoliths with garnet and cpx in one suite. This calls into question how
representative these petrologic phases are of distinct crustal thicknesses.
The progression of mineralization shown in figure 6 is certainly simplified. Are
there any complications to the simplified petrologic picture that could upset the
model?
Does a flat slab always correlate with thickened crust? It seems so given the
observations in the Laramide western U.S. and in the two places in South
America. Again, we're not sure about Mexico.
The ages of thickening of the crust that Kay calls on to fit her model are
controversial. Therefore the crustal thickening timing is still somewhat of a
question. What, if anything can we do or observe to make this more clear?
Kay's model focuses on weakening of the crust as the mechanism for enhanced
crustal thickening during flat-slab subduction episodes, therefore she does not
comment much on how the flat slab affects stresses within the crust. This is still a
big question with regard to flat slabs, both in magnitudes of stresses within the
crust and how stresses are transmitted differently from slab to crust in a normal
vs. flat-slab subduction geometry. Are stresses increased during flat-slab
subduction, and if so, how does this affect deformation?
Next we moved into discussions of why the slab is flat. Jerome mentioned that early
work by Jarrard (his 1986 paper is referenced in van Hunen, 2002) showed that there
is no correlation between age of the subducting slab and slab dip. This may be the
reason why people moved to talking about oceanic plateaus as a mechanism for slab
buoyancy in flat-slab subduction zones. There seems to be correlation between
subduction of oceanic plateaus and flat subduction from observations world-wide
(Lara mentioned work by Gutscher). In addition, it seems that some oceanic plateaus
are so thick that they resist subduction completely. Some examples are: the OntongJava plateau, some accreted terranes in Canada, and possibly Cyprus.
The paper we briefly discussed, van Hunen (2002), makes a
thermal/geodynamic/petrologic model of flat subduction that seems to work pretty
well. His model relies on a kinetic delay of the transformation of basalt to eclogite in
the crust of the subducting plateau that makes the slab buoyant. Three areas of
discussion resulted from his conclusions:
1. He concludes that the maximum length of subducting slab that can be flat is
300-400 km. Yet in figure 6, this limit does not show in the trends of flat-slab
length vs. time. We concluded that we'll have to believe that he carried out
the numeric experiments for long enough time frames to see this leveling off.
In addition, there is a big difference between this 300-400 km limit and
lengths of flat slab inferred to have underlain such subduction zones as the
Laramide in the western U.S. (up to 1500 km length; Saleeby figure 1). He
explains the discrepancy by requiring active overriding of the upper plate over
the subducting plate to supplement the buoyancy factor. This was not a part
of his models and is therefore a possibility.
2. There may be other possibilities not discussed by van Hunen. We discussed
the possibility that there may be other mechanisms, yet undetermined,
responsible for flat slab subduction. While a plateau subduction may be
likely in the Laramide example (we may be able to correlate reconstructed
plate geometries with a very large plateau in the eastern Pacific), people infer
that flat subduction may have existed at two distinct depths under the western
U.S. In the north, it was at 100 km (Saleeby figure 3) while in the south it
may have been as shallow as ~30 km. This ultra-shallow flat subduction does
not fit with van Hunen's models involving basalt-to-eclogite transformation
because in his models, the slab flattened at a depth of 100 km. Steve
mentioned a caveat: models inferring this super shallow slab rely on the deep
subduction of forearc sediments (discussed last week); however, there are
alternative models that might be able to explain deep sediment subduction that
don't involve an ultra-shallow flat slab.
3. The big unresolved observation in the end is: we know very little about the
reaction kinetics that may allow for delayed transformation of basalt to
eclogite. Maybe we can clear this up with some consideration of the literature
addressing this question directly (next week).