Throughout the ultrahigh-pressure (UHP) metamorphic unit of the Dora-Maira massif, western Alps, pyrope megablasts contain
the typical assemblage clinochlore–kyanite–talc–rutile ± phlogopite ± ellenbergerite as prograde inclusions. In the upper
part of the UHP unit in Val Gilba, some megablasts (XMg=0.89–0.98) contain in addition polymineralic inclusions consisting
of various combinations of enstatite, gedrite, sapphirine, clinochlore, talc, magnesiostaurolite and rare corundum or spinel.
We present evidence that these assemblages developed from cracks running across the megablasts, and are therefore of late
origin, post-dating the highest-pressure stage. Enstatite (XMg=0.94–0.99) contains 0.7 to, typically, 3 wt% Al
2O
3, but up to 8.4 wt% in the presence of sapphirine. Sapphirine (XMg=0.96–0.998, Be-free) shows the largest compositional variations,
with Si contents ranging from 1.7 to at least 2.1 atoms pfu, thereby clearly exceeding the 2:2:1 stoichiometry. The late-stage
talc contains up to 4 wt% Al
2O
3, 0.35 wt% Na
2O and 0.6 wt% F; gedrite 1.1–2.9 wt% Na
2O and up to 0.36 wt% F. The successive development within pyrope of alternative hydrous assemblages involving first enstatite
plus an Al-rich phase (kyanite, sapphirine, magnesiostaurolite) ± clinochlore, then a gedrite compositionally close to pyrope,
then talc plus an Al-rich phase (sapphirine, corundum), is a clear record of decompression. However, the temperature conditions
implied under the assumption of high H
2O activity are 100 to 150 °C higher than, and so inconsistent with existing constraints on the decompression path. These constraints
are in particular the stability of talc + phengite in the matrix assemblage during decompression, and the absence of regional
evidence for a granulite-facies event. This inconsistency can only be alleviated if H
2O activity inside the garnet megablast was (or became) considerably reduced with respect to that in the matrix. Fluid influx
into an opening fracture in garnet, sealing of the fracture by breakdown products of pyrope and continued evolution under
closed-system conditions may have led to increasing solute concentration and such low H
2O activity
within the garnet megablast, driving the microsystem toward fluid-absent conditions. Micrometre-size inclusions of Ca-sulfate and
crandallite-type compounds in minerals of these reactive areas may be evidence for such residual brines and suggest that these
were phosphate- and sulfate- rather than halide-dominated. This finding is additional evidence for the very local control
that fluid composition and H
2O activity may have on the occurrence of granulite-facies assemblages, regardless of temperature. It highlights the role of
deformation (here fracturing) in triggering reactions in otherwise unreactive systems. It also shows how carefully inclusion-
to-host relationships have to be considered, post-growth reaction within the host being more common than hitherto reported.