We investigated deformation processes within a lower crustal shear zone exposed in gabbros from Arnøya, Norway. Over a distance of ∼1 m, the gabbro progresses from nominally undeformed to highly sheared where it is adjacent to a hydrous pegmatite. With increasing proximity to the pegmatite, there is a significant increase in the abundance of amphibole and zoisite (which form at the expense of pyroxene and calcic plagioclase) and a slight increase in the strength of plagioclase lattice-preferred orientation, but there is little change in recrystallized plagioclase grain size. Phase diagrams, the presence of hydrous reaction products, and deformation mechanism maps all indicate that the water activity (aH2O) during deformation must have been high (∼1) in the sheared gabbro compared with the nonhydrated, surrounding host gabbro. These observations indicate that fluid intrusion into mafic lower crust initiates syn-deformational, water-consuming reactions, creating a rheological contrast between wet and dry lithologies that promote strain localization. Thus, deformation of lower continental crust can be accommodated in highly localized zones of enhanced fluid infiltration. These results provide an example of how fluid weakens lower continental crust lithologies at high pressures and temperatures.
 We examined sheared gabbros from Arnøya, Norway (Figure 1), to investigate the effects of hydration on rheology in lower continental crust lithologies. Strain in the lower crust is generally localized in mylonites [e.g., Ramsay, 1980; Rutter and Brodie, 1988]. Viscous deformation processes have been well studied for several lower crust minerals such as olivine [e.g., Kohlstedt and Goetze, 1974; Mei and Kohlstedt, 2000] and plagioclase [e.g., Tullis and Yund, 1987, 1992; Rybacki and Dresen, 2004; Stünitz et al., 2003]; however, processes associated with deformation in polycrystalline material, fluid infiltration, and the evolution of creep processes with increasing strain localization and hydration are less well constrained. These processes influence the mechanical behavior of the lower crust [e.g., Mackwell et al., 1998].
 Under lower crustal conditions, the addition of small amounts of water changes the phase equilibrium [e.g., Stünitz and Tullis, 2001]. The growth of new phases can modify the overall rock strength through their own inherent strength [e.g., Mitra, 1978; Steffen et al., 2001], or, because they are generally finer grained, by promoting a transition in deformation mechanism from dislocation to diffusion creep (e.g., Rutter and Brodie , see Stünitz and Tullis  for a comprehensive list of weakening mechanisms). The suite of gabbro samples we analyzed progress from nominally undeformed to mylonitized with increasing proximity to a hydrated pegmatitic dike, which provided a fluid source for alteration of the gabbro country rock. Some studies suggest that grain size reduction in dry oceanic crust [Mehl and Hirth, 2008] and/or contrasting bulk lithologies [Homburg et al., 2010] are critical for strain localization. In contrast, the results presented here indicate that the effects of fluid alone may be sufficient to cause localization in a locally hydrated lower continental crust. Motivated by the pronounced effects of hydration on rock strength and strain localization, we combine microstructural and phase equilibrium constraints to (1) assess how fluid influences the active deformation mechanisms, (2) ascertain if and how mineralogical changes related to fluid infiltration weaken the rock, (3) constrain the relative timing of reactions, hydration, and deformation, (4) understand how these processes influence the rheology of the lower crust, and (5) provide a geologic context to which to compare weakening mechanisms associated with tectonic tremor at the base of the San Andreas Fault.
2. Tectonic Setting and Outcrop Observations
2.1. Tectonic Setting
 We collected samples of gabbro that intruded into amphibolite- to granulite-grade metasediments from the island of Arnøya, Norway (Figure 1). Arnøya is located in the southwest of the Seiland Igneous Province, which was emplaced during the Neoproterozoic [Roberts et al., 2010]. The tectonic setting of Arnøya is somewhat ambiguous; it has been mapped as part of both the Precambrian crystalline rocks of the middle allochthon Kalak nappe complex [Gorbatschev, 1985; Roberts, 2003] and the uppermost allochthon of the Laurentian margin [Gee et al., 2008]. Our field and microstructural work show that the gabbro on both Arnøya and the neighboring island of Kågen are part of the same intrusion complex. The Kågen gabbro is considered a part of the Vaddas nappe [Zwaan, 1988], but the distinction between Vaddas and Kalak nappes is not clear in this region [Lindahl et al., 2005] and is presently under investigation. Regardless of the specific tectonic province, the Arnøya/Kågen gabbro shows a migmatitic zone of contact metamorphism, which formed during the emplacement of the gabbro and deformation of the adjacent felsic and metapelitic rocks. As a result, migmatites and pegmatites are found in the immediate vicinity of the gabbro, intrude the gabbro, and are deformed together with the gabbro.
2.2. Outcrop Observations
 Both undeformed and sheared rocks of differing lithologies outcrop along a 1 km stretch of exhumed lower crust on Arnøya. A quarry exposes fresh olivine gabbro that exhibits varying degrees of strain localization associated with syntectonic pegmatitic dikes. This gabbro unit is surrounded by metaigneous amphibolite (beige unit in Figure 1b), which represents the margin of the gabbro deformed during later-stage nappe emplacement. The amphibolite is encompassed by deformed metasedimentary rock (pink unit in Figure 1b). The gabbro was initially mapped as two different units: a core of olivine gabbro and an outer shell of olivine-poor gabbro [Roberts, 1973]; however, subsequent investigators mapped this as one undivided gabbro/metagabbro/amphibolite/diorite unit [Sigmond et al., 1984].
 Coarse-grained, anastomosing pegmatitic dikes composed of quartz, feldspar, muscovite, biotite, kyanite, and pyrite crosscut the gabbro throughout the outcrop and are typically surrounded by alteration zones of varying widths (Figures 1c and 1e). These pegmatites range in thickness from millimeters to tens of centimeters and extend laterally, and in some cases vertically, for several meters. In many locations, the pegmatites are sheared, indicating that they intruded prior to or during deformation (Figure 1d).
 We examined the mineralogy and microstructures of samples that record progressive strain localization associated with a 0.1 m wide pegmatite. The alteration zone associated with the pegmatite is 0.5 m wide within the gabbro (Figure 1f). We analyzed the following rocks illustrated in Figure 1: nominally undeformed and unmetamorphosed olivine gabbro (A7D); moderately metamorphosed and nominally undeformed olivine gabbro (A7E—sampled at the edge of the alteration front in Figure 1e); deformed, amphibole-bearing gabbro (A7F—sampled adjacent to the pegmatite shown in Figure 1); and mylonitized and completely metamorphosed gabbro with associated felsic layers (A7B, Figure 1d). The first three samples span an approximately 1 m section above the shear zone; the mylonite sample was obtained approximately 2 m north of the other three samples because the shear zone was buried underground beneath the sheared gabbro A7F (Figure 1c). The mylonitized felsic material in sample A7B is either the continuation of the same pegmatite separating A7E and A7F or a different pegmatite composed of the same mineralogical constituents. The gabbro develops a strong foliation with increasing proximity to the shear zone: nominally undeformed gabbros A7D and A7E are not foliated, sheared gabbro A7F is strongly foliated, and A7B is mylonitized. This high strain zone extends approximately 0.5 m away from the pegmatite into the gabbro. Thin sections were prepared for microstructural observations by cutting samples parallel to lineation and perpendicular to foliation (if present).
3.1. Electron Microprobe
 We analyzed major and trace element compositions of plagioclase, amphibole, and pyroxene in all samples using a Cameca SX-100 electron microprobe with a 10 nA beam current; a 15 kV acceleration voltage; a focused beam size of 2 µm for pyroxene, amphibole, and zoisite; and a defocused 15 µm beam for plagioclase. The compositions of adjacent amphibole and plagioclase grains were measured for thermometry. We also analyzed host/daughter grains of recrystallized plagioclase and pyroxene to compare their compositions to those in the nominally undeformed gabbro.
3.2. Electron Backscatter Diffraction
 We used electron backscatter diffraction (EBSD) data to examine crystal orientations and lattice-preferred orientations (LPOs) of plagioclase, pyroxene, and amphibole as a function of distance from the shear zone and pegmatite fluid source. We used a JEOL 845 SEM at 20 kV and a probe current of 6 × 10−9 amps to collect the EBSD data. Thin sections were polished with 0.06 µm colloidal silica and carbon coated to reduce charging in the SEM. We obtained one data point per grain from ∼1.46 mm spaced transects across each of the four thin sections to produce pole figures. Owing to the large grain size in gabbros A7D and A7E, more than one point per grain occasionally may have been analyzed on adjacent transects. We processed the data with Oxford/HKL Technology's Channel 5 software and quantified the strength of the LPO using the M-index, which is a measure of fabric strength (ranging from 0 to 1) based on the distribution of uncorrelated misorientation angles [Skemer et al., 2005]. Statistically robust comparisons can be made for data sets that include at least 150 grains [Skemer et al., 2005], which we acquired for plagioclase in all samples, clinopyroxene in the nominally undeformed gabbro (A7D and A7E), and amphibole in the deformed gabbro (A7F and A7Bb). Pole figures are contoured with a halfwidth of 8.5° and a global maximum of seven times a uniform distribution.
3.3. Grain Size
 We quantified the grain size of all phases using the line intercept method on crosspolarized photomicrographs. We measured at least 200 grains of both plagioclase and mafic minerals, both parallel and perpendicular to the lineation (maximum and minimum elongation directions) in each sample. To convert these measurements to a grain diameter, we multiplied the arithmetic mean of the grain size distribution by the geometric correction factor of 1.5 [Gifkins, 1970].
4.1. Sample Description and Petrography
 Table 1 lists the modal abundances of the dominant minerals in the four samples. Farthest from the shear zone, the lack of foliation, generally well-preserved olivine grains, and undeformed corona textures indicate that this part of the gabbro remained nominally undeformed. The predominant mineralogy of the nominally undeformed gabbro (A7D) is plagioclase, clinopyroxene, and olivine with some orthopyroxene (Figure 2a). Trace phases include spinel, amphibole rims on the clinopyroxene, and igneous biotite; trace amounts of chalcopyrite and zoisite overprint the plagioclase. Kyanite has been found in other samples of this gabbro body [Nasipuri et al., 2011]. Large plagioclase grains (>1 mm) exhibit subgrains, undulose extinction, and generally lobate grain boundaries (although some are straight; Figure 2b). Smaller plagioclase grains (<200 µm) have abundant twins, no undulose extinction, and ∼120° triple grain junctions. Large clinopyroxene grains have generally straight exsolution lamellae and thin rims of smaller, recrystallized clinopyroxene, amphibole, and orthopyroxene. The olivine grains exhibit minor subgrain formation, minor late stage serpentinization, and have nominally undeformed corona textures consisting of finer-grained clinopyroxene, orthopyroxene, and trace spinel (Figure 2b).
Table 1. Modal Abundances of Phases in the Gabbro Suitea
Modal abundances presented are the nominally undeformed and unmetamorphosed gabbro (A7D), the nominally undeformed and slightly metamorphosed gabbro (A7E), the deformed and metamorphosed gabbro (A7F), and the mylonitized gabbro (A7B).
 At the edge of the alteration front (sample A7E), the gabbro has almost the same mineral composition as A7D (Figure 2c). Although still minor compared to the other phases, amphibole rims on the pyroxene grains are slightly thicker than in sample A7D, and there is little orthopyroxene. The olivine grains have nominally undeformed corona textures of fine-grained pyroxene, amphibole, and trace spinel, again suggesting that this part of the gabbro experienced only minimal deformation (Figure 2d). Large plagioclase grains (>1 mm) all have lobate grain boundaries, undulose extinction, and generally less twinning than A7D; the twins that are present are commonly warped. The smaller plagioclase grains show undulose extinction and lobate grain boundaries (Figure 2d). Many triple grain junctions exhibit dihedral angles of ∼120°. Larger pyroxene grains (>1 mm) have undulose extinction and warped lamellae. Smaller pyroxene grains appear generally strain free and commonly have 120° grain junctions.
 With increasing proximity to the pegmatite and shear zone, the gabbro develops a strong foliation defined by a layering of plagioclase and amphibole (sample A7F, Figure 2e). There is a loss of olivine and pyroxene (particularly orthopyroxene) and a transition from amphibole as a minor phase rimming clinopyroxene to amphibole as the dominant mafic phase. Olivine is no longer present in sample A7F; presumably it was entirely consumed by the corona-forming reactions. Zoisite is abundant, and there is a trace amount of quartz and opaque minerals. Amphibole grains show uniform extinction and form triple grain junctions of 120° (Figure 2f). Recrystallized plagioclase grains have 120° grain junctions and straight (nonlobate) grain boundaries. Some plagioclase porphyroclasts exhibit undulose extinction and minor twinning. Compositional layering is defined by bands of predominately plagioclase and zoisite parallel to bands of predominantly amphibole (Figure 2f), which defines the foliation. Zoisite grains are highly elongate and define a lineation within the well-developed foliation. The few remaining clinopyroxene grains are recrystallized, surrounded by amphibole, and exhibit warped exsolution lamellae.
 In the core of the shear zone, mylonitic gabbro and recrystallized felsic bands exhibit a well-developed foliation defined by compositional layering of plagioclase and amphibole, both of which also exhibit a shape-preferred orientation (SPO) (Figure 2g). This sample has a relatively coarse-grained, nearly monophase plagioclase band with minor quartz, zoisite, and amphibole (A7Bmono, Figure 2h) and a polyphase, finer-grained region with abundant amphibole, plagioclase, and zoisite with minor amounts of clinopyroxene, quartz, and opaque minerals (A7Bpoly, Figure 2i). Plagioclase porphyroclasts have mechanical twins, undulose extinction, and tails of recrystallized grains, which have some twinning, and generally form 120° triple junctions. Larger zoisite grains are elongate and define a lineation within the foliation and smaller zoisite grains are randomly oriented. The relicts of the coronae observed in the nominally undeformed samples are also highly deformed and consist of amphibole and clinopyroxene.
4.2. Phase Composition and Reactions
 The compositions of plagioclase, amphibole, and pyroxene are summarized in Table 2. In this gabbroic suite, crosscutting relations show that two main reactions alter the mineralogy. Initially, olivine reacts with plagioclase to form fine-grained coronae of clinopyroxene, orthopyroxene, and minor spinel:
Table 2. Major Element Averages for Plagioclase, Amphibole, and Clinopyroxene for All Four Thin Sectionsa
N indicates the number of grains analyzed. Error in parentheses is one standard deviation (for example, single digits equal deviation in hundredths). There are no clinopyroxene grains in A7Bmono. Low amphibole n values for A7D and A7E reflect the lack of amphibole in these samples.
Avg. An content
 Later, concurrent with the intrusion of the pegmatites, these newly formed phases react to produce hydrous minerals:
 Reaction (2) produces changes in bulk rock mineralogy and mineral phase composition. With increasing proximity to the fluid source and shear zone, plagioclase becomes more albitic. Average anorthite contents of plagioclase in the moderately undeformed samples A7D and A7E are An68 and An70, respectively, while the average in the deformed sample A7F is An45 (Table 2; supporting information, Figure S1).1 The average composition of recrystallized plagioclase grains in the most deformed, polyphase region (A7Bpoly) is more calcic (An55) than plagioclase in the sheared, felsic, monophase zone (A7Bmono, An43). A7Bpoly recrystallized plagioclase grains have anorthite contents of An50, and the porphyroclasts are more calcic (An62). In the gabbro closer to the shear zone and pegmatite, amphibole generally contains less Al and Na and more Mn and Si than amphibole in the nominally undeformed gabbro. Pyroxenes are higher in Si and Ca and lower in Ti and Al in the sheared and hydrated gabbro.
4.3. Grain Size
 Overall, there is little change in recrystallized grain sizes with increasing strain. Plagioclase recrystallized grain size (average of grains < 1 mm) ranges from 300 to 285 µm (Figure 3) between the nominally undeformed (A7D) and the felsic monophase mylonite (A7Bmono) samples. The average plagioclase grain size is ∼170 µm in the mylonitized polyphase band (A7Bpoly). The average plagioclase aspect ratio is ∼1 for A7D, A7E, A7F, and A7Bmono. The average aspect ratio is greater (1.3) in the mylonitized, polyphase sample (A7Bpoly).
 Between the nominally undeformed (A7D) and slightly metamorphosed (A7E) gabbros, the recrystallized grain size of mafic phases (average of grains < 1 mm) decreases from 310 to 251 µm. Average mafic grain size increases in the sheared gabbro (A7F) to 280 µm (amphibole and clinopyroxene) and then decreases to 150 µm (amphibole and clinopyroxene) in the mylonitized polyphase gabbro (A7Bpoly). Average mafic grain aspect ratio increases from ∼1 in A7D to 1.2 A7F and A7Bpoly.
4.4. Lattice-Preferred Orientation
 Pole figures for plagioclase and amphibole are shown in Figures 4 and 5, respectively. In general, the samples exhibit a weak plagioclase LPO, with the exception of the mylonitized felsic band in the most deformed sample (A7Bmono). The plagioclase fabric strengths, as quantified by the M-index [Skemer et al., 2005], remain relatively weak with increasing proximity to the shear zone (Figure 4). Due to their large grain sizes, plagioclase grains in the undeformed gabbro A7D (M-index of 0.06) and A7E (M-index of 0.05) were likely counted more than once, so M-index values for A7D and A7E are overestimates. The plagioclase monophase region (A7Bmono; M-index of 0.07) has a maximum in  parallel to the foliation plane and a maximum in (100) close to foliation normal, initially suggesting slip on the (100)  system; however, the  Burgers vector is much larger (1.3 nm) than those in other common slip directions (0.7 nm for  and 0.8 nm for ), making slip in the  direction highly unlikely [Marshall and McLaren, 1977]. Instead, comparison of the LPO with the kinematics of the shear zone (see boxed inset in Figure 4) indicates slip could have occurred on ( ) plane in the [ ] direction (Burgers vector = 0.7 nm) [Marshall and McLaren, 1977; Olsen and Kohlstedt, 1984]. Slip in the [ ] direction has been inferred for recrystallized plagioclase in naturally deformed gabbros [Kanagawa et al., 2008] and mafic dikes [Pearce et al., 2011]. Irrespective of slip system, the pole figures show that there is a pronounced plagioclase LPO in the deformed monophase region, whereas the LPO in the other samples—including the directly adjacent deformed polyphase bands—is weak.
 The LPOs are stronger for amphibole (Figure 5). The M index for amphibole in the two deformed gabbros is 0.05 (A7F) and 0.10 (A7Bpoly), respectively. There were not enough amphibole grains in A7D, A7E, or A7Bmono to obtain a statistically robust M index or pole figure.
 Our outcrop observations and analytical results indicate that strain localization in the gabbro is promoted by the addition of fluid. In the following sections, we first provide an analysis of the deformation conditions and active deformation mechanisms. We then explore the implications of our results for understanding (1) the role of strain localization on the strength of the lower continental crust and (2) the deformation conditions under which low-frequency earthquakes and tectonic tremor occur.
5.1. Deformation Conditions
5.1.1. Pressure and Temperature
 To estimate the temperature during deformation, we employed Holland and Blundy's  HbPlag thermometer using microprobe analyses of adjacent rims of hornblende and plagioclase. The equilibrium assemblage used by Holland and Blundy  includes quartz and magnetite; thus, we use data from samples A7B and A7F because A7D and A7E have no quartz. At 1 GPa, HbPlag (using edinite-tremolite exchange vectors) predicts a temperature range of 610°C–710°C in A7B and 620°C–680°C in A7F (supporting information, Figure S2). At 0.5 GPa, HbPlag predicts a temperature range of 600°C–670°C in A7B and 615°C–670°C in A7F (supporting information, Figure S2). The pressure during deformation is constrained to lie between 0.7 and 0.9 GPa based on (1) the presence of kyanite in the syn-tectonic pegmatites, which is stable at pressures of 0.7–0.9 GPa between temperatures of 650°C and 750°C; (2) pressure estimates of 0.85 GPa obtained from analysis of corona microstructures associated with reaction (1) in the nominally undeformed gabbro [Nasipuri et al., 2011]; and (3) the presence of zoisite in the deformed assemblage, which is stable at temperatures less than approximately 650°C at a pressure of 0.9 (see section 5.1.2). Based on these analyses, we use a temperature range of 650°C ± 50°C at a pressure of 0.9 GPa (supporting information, Figure S2) to evaluate the applicability of flow laws and assess the water activity during deformation (see section 5.1.2).
 We interpret that HbPlag coupled with associated phase equilibrium provides estimates for the temperature and pressure during deformation (rather than cooling) based on the following observations. Amphibole, which forms by hydration reaction (2), is only present in significant amounts in the deformed samples. The foliation in these samples is defined by amphibole-rich layers exhibiting LPOs and SPOs, indicating that amphibole formed synkinematically. Plagioclase in monomineralic layers also exhibits an LPO consistent with the kinematics of the shear zone. The temperatures are minimum estimates, as the phases used in the thermometer could have reequilibrated at temperatures slightly lower than those of the main deformation event. However, we do not observe any zoning in the amphibole or plagioclase grains within the deformed samples.
5.1.2. Water Activity and Fluid Abundance
 To constrain water activity (aH2O) during deformation, we explored the effects of water on phase stability with a pseudosection generated using PerpleX [Connolly, 1990]. A temperature versus aH2O pseudosection for the system NCFMASH (Na2O, CaO, FeO, MgO, Al2O3, SiO2, H2O) is shown in Figure 6a for a pressure of 0.9 GPa. We determined the bulk composition for this calculation from the initial composition and modal abundance of phases in the nominally undeformed olivine gabbro (plagioclase, clinopyroxene, orthopyroxene, olivine, and spinel). Figure 6b shows the influence of temperature and aH2O on modal abundance for this bulk composition, and Figure 6c shows modal variation with temperature and water content. The locations of the samples in Figure 6a are based on the temperatures determined using HbPlag and aH2O constrained by the modal abundances of the rocks summarized in Table 1. The presence of synkinematic zoisite indicates that the aH2O ≈ 1 (1.13 wt % water) in the shear zone (there is no constraint on the maximum amount of free water present); the lack of amphibole in the adjacent, nominally undeformed gabbro, indicates aH2O < ∼0.1. This difference in aH2O indicates that a strong gradient in aH2O existed during deformation.
 In general, the results shown in Figure 6 accurately predict the phases present in the entire sample suite. Although olivine is stable at the high temperatures of gabbro crystallization, it is unstable at 0.9 GPa and temperatures less than 900°C, consistent with the observations of the corona texture (representing reaction (1)) observed in the nominally undeformed gabbro. The preservation of these corona textures in the gabbro farthest from the fluid source (A7D and A7E) provides further evidence of low aH2O outside of the shear zone [e.g., Passchier and Trouw, 2005]. A trace amount of igneous biotite is present in the nominally undeformed gabbro. We did not add K in the pseudosection shown in Figure 6, but calculations with a trace amount of K demonstrate that biotite is stable at aH2O between 1 and ∼0.04 at 900°C and to even lower aH2O at lower temperatures. Thus, biotite does not further constrain the aH2O. At low aH2O, the model also predicts the presence of small amounts of garnet, which we do not observe. Thus, either the reaction kinetics was too sluggish to stabilize garnet or the predicted modal abundance of garnet is low enough to be within the resolution of thermodynamic calculations.
 The presence of strongly aligned zoisite in A7B and A7F indicates that aH2O ≈ 1 during deformation. Within the range of temperatures predicted by HbPlag, at 0.9 GPa, zoisite is stable at 650°C (Figure 6a). At lower pressures, zoisite is not stable within the temperature range predicted by HbPlag. For example, at 0.75 GPa, zoisite is only stable at temperatures below 600°C, which is below the predicted temperature range. The thermal stability of zoisite is even lower at lower aH2O. These observations corroborate the pressure estimates described above.
 The presence of rare zoisite grains and amphibole rims in the nominally undeformed gabbro likely results from late-stage alteration. In addition to their scarcity, amphibole and zoisite grains are randomly oriented and heterogeneously distributed. Zoisite grains are small and typically found near healed cracks in plagioclase grains, and as discussed, amphibole appears only as thin rims. If zoisite was formed in the nominally undeformed gabbro during deformation, the temperature would have been <550°C at aH2O < 0.1, but such temperatures are significantly lower than those predicted by HbPlag thermobarometry. The trace hydrous phases in the nominally undeformed country rock may have formed during the last stages of the deformation event as minor fluid migrated away from the source or during a later overprinting event. Regardless, these phases are rare, randomly oriented, and not deformed, suggesting that there was not enough fluid to stabilize them during the main deformation event.
5.1.3. Differential Stress and Strain Rate
 We estimated stress during deformation using plagioclase-recrystallized grain size piezometry [Twiss, 1977]: σ = Bd−0.68 where σ is differential stress in MPa, d is grain size in microns, and B is the preexponential constant. For this analysis, we use the average recrystallized grain sizes in the most deformed gabbros (A7F and A7B). The differential stress shown in Figure 7 is based on the grain size of the monophase plagioclase bands in A7F and A7Bmono. The grain size in the monophase bands provides a lower estimate for stress because postkinematic annealing may have augmented grain size. Such annealing is mitigated in the mixed polyphase bands where second phase grain pinning limits grain growth, thereby promoting diffusion creep. Because the plagioclase in the finer-grained region (A7Bpoly) likely deformed via diffusion creep (see section 5.2), we used stress estimates predicted by the slightly coarser-grained, monophase region (A7Bmono) in the same thin section to estimate the stress in A7Bpoly [cf., Mehl and Hirth, 2008].
 We combine our calculated values of temperature, water activity, and stress to estimate the strain rate in the shear zone using a deformation mechanism map (Figure 7). We use plagioclase flow laws to construct the map, based on (1) the presence of abundant recrystallized plagioclase and lack of pervasive recrystallization in other phases and (2) lack of flow laws for amphibole. Constant strain rate contours are shown in Figure 7 for wet anorthite at a pressure of 0.9 GPa and a temperature range of 600°C–700°C using a flow law with the form:
 where ε is the strain rate, n is the stress exponent, m is the grain size exponent, fH2O is water fugacity, r is the water fugacity exponent, Q is the activation energy, P is pressure, V is the activation volume, R is the gas constant, T is temperature, and A is the preexponential constant. We assume that the total strain rate is the sum of the strain rates for both dislocation and diffusion creep. We use experimental creep parameters from Rybacki et al. , which require modest extrapolations in water fugacity, grain size, and temperature. For the strain rate contours in Figure 7, we used V = 24 cm3/mol for both the diffusion creep and dislocation creep regimes. The value for V in the dislocation creep regime is poorly constrained because of the limited pressure range under which creep data have been acquired. Rybacki et al.  estimate V = 36 ± 20 cm3/mol for wet anorthite, based primarily on experiments in the diffusion creep regime. An uncertainty in V in the range between 24 and 54 cm3/mol results in approximately a factor of 2 uncertainties in differential stress at a pressure of 0.9 GPa, which is equivalent to approximately 1 order of magnitude change in strain rate (gray bar in Figure 7). For V estimates of 38 and 54 cm3/mol to be applicable to this gabbro suite, the stresses would have to be significantly higher than those predicted by grain size piezometry and/or temperatures would have to be significantly lower than those predicted by HbPlag thermometry.
 The deformation mechanism map indicates that the shear zone strain rate was ∼10−13 s−1, which is reasonable—if not low—for a shear zone. Using this strain rate and the width of our shear zone (∼0.1 m), we obtain a shear displacement rate (x) of 10−13 s−1 = x/0.1 m, where x = 10−14 m/s, which is 5 orders of magnitude slower than typical plate rates. Although strain is difficult to quantify, we estimate that strain in the most deformed sample (A7B) is probably greater than 10 based on the observation that the foliation is approximately parallel to the shear zone. At a strain rate of ∼10−13 s−1, it would take ∼3 Myr to deform to a strain of 10. We estimate a strain of ∼10% in the nominally undeformed sample farthest from the fluid source (A7D) based on the observation of recrystallized plagioclase grains. Thus, the total strain in the mylonite is at least 2 orders of magnitude larger than that in A7D (thus, a strain rate of <10−15 s−1 in A7D).
 Extrapolation of experimental data indicates a small change in water activity can promote a large change in strain rate (Figure 8). Knowing the approximate temperature, aH2O, pressure, stress, and strain rate during deformation, we compare the deformed samples (circles) and the undeformed samples (gray-shaded box) to flow laws for wet anorthite [Rybacki et al., 2006], wet 50-50 mixtures of anorthite and clinopyroxene [Rybacki and Dresen, 2000], dry anorthite [Rybacki et al., 2006], and dry 50-50 mixtures of anorthite and clinopyroxene [Rybacki and Dresen, 2000]. With an aH2O ≈ 0.1 the nominally undeformed gabbro is hydrated enough to deform at a strain rate significantly faster than that predicted by the dry anorthite flow law (Figure 8), consistent with the formation of recrystallized plagioclase grains.
5.2. Microstructural Constraints on Deformation Mechanisms
 Microstructural observations, such as subgrain boundaries, indicate the activation of crystal plasticity in all samples. The presence of an LPO is often interpreted to indicate that dislocation creep is the dominant deformation mechanism [e.g., Wenk and Christie, 1991]. In our sample suite, plagioclase in the monophase zone of the mylonite (A7Bmono) has one of the strongest fabric strengths (M-index of 0.07) and an LPO consistent with the kinematics of the shear zone, likely produced by slip on the ( ) [ ] system (Figure 4). The plagioclase LPO in sheared gabbro A7F (M-index of 0.09) suggests that dislocation creep was also active in this part of the rock.
 In contrast, the plagioclase LPO in the polyphase region of the mylonite (A7Bpoly) is weaker (M index of 0.05) with weak maxima that do not correlate in any straightforward way to the kinematics of the shear zone (Figure 4), both of which suggest that diffusion creep was active in this part of the rock. This interpretation is consistent the deformation mechanism map, which indicates that plagioclase in the polyphase zone of the mylonite deformed in the diffusion creep regime near the boundary between dislocation and diffusion creep at 650°C, while the monophase zone deformed in the dislocation creep regime (Figure 7).
 Laboratory studies on plagioclase indicate that LPOs may also form during diffusion creep [e.g., Barreiro et al., 2007]. Strong LPOs have also been observed in olivine + orthopyroxene aggregates deformed in the diffusion creep regime [Sundberg and Cooper, 2008]. However, based on the observed microstructures of the samples with a stronger LPO, we believe that the plagioclase LPO in A7F and A7Bmono results from dislocation creep. The weaker LPO in A7Bpoly may be a result of either diffusion creep or weakening of a preexisting LPO during a switch to diffusion creep [Lapworth et al., 2002].
 The amphibole LPO in sample A7Bpoly is strong (Figure 5); however, we observe no crystal plastic substructures in these amphibole grains. Following the conclusions of Berger and Stünitz , we infer that the amphibole LPO reflects metamorphic recrystallization and rigid grain rotation during deformation. Strain is localized where there is a dramatic increase in the amount of amphibole. Indeed, the macroscopic texture of the most sheared sample indicates that the strain in the polyphase region (A7Bpoly) is similar to that in the monophase regions (A7Bmono). This observation suggests that the diffusion creep rate of the wet amphibole + plagioclase rock is similar to that of pure plagioclase (i.e., Figure 8). Therefore, the diffusion creep rate in the wet amphibole + plagioclase rock must be significantly greater than that predicted by the two-phase (wet 50-50 mixture of anorthite and clinopyroxene) flow law. Enhanced weakening in amphibole + plagioclase aggregates observed by Kenkmann and Dresen  was interpreted to result from diffusion and dislocation accommodated grain boundary sliding. The effect of synkinematic amphibole growth on rock strength has not been quantified in laboratory experiments.
5.3. Strain Localization Mechanisms and Timing of Reactions Relative to Localization
 Our analyses indicate that strain localization in the mafic assemblage on Arnøya was directly linked to fluid infiltration. Many studies suggest that grain size reduction is the dominant driver for strain localization [Rutter and Brodie, 1988; Stünitz and Tullis, 2001; Platt and Behr, 2011]. While we do observe a reduction in the number of large (>1mm) porphyroclasts between the undeformed and the mylonitized gabbro, the recrystallized plagioclase grain size in the monophase regions of the mylonitized gabbro is the same within error of that in plagioclase-rich zones of the nominally undeformed host rock (Figure 3). The smallest recrystallized grains in the mylonite are observed in the polyphase zone (A7Bpoly), where microstructures suggest that deformation occurred via diffusion creep (Figure 4). However, in the directly adjacent (and coarser-grained) region of the mylonite, the microstructures indicate deformation via dislocation creep (A7Bmono). Thus, we conclude that fluid infiltration from the pegmatite promoted crystal plasticity in plagioclase, contributed to the growth of amphibole, and enhanced diffusion creep rates in the polyphase regions of the mylonite, all of which contributed to strain localization.
 The extent and duration of fluid-induced strain localization was limited by the volume and transport kinetics of fluid available from the pegmatite source. The combination of deformation microstructures and phase stability relations suggests that a gradient in aH2O was maintained during deformation. The stability field of the mineral assemblage outside of the shear zone (sample A7E) at 650°C is constrained to a small window of aH2O ≈ 0.1 (Figure 6a), while the assemblage in the highly strained regions indicate aH2O ≈ 1. The lack of deformation in the gabbro at the boundary of the hydration front further supports our conclusion that strain localization was directly linked to the influx of fluid into the gabbro.
 The diffusion rate of hydrous species (H2O and OH) in plagioclase is not well constrained. Intriguingly, the length scale of hydration indicated by the hydration front in our samples, exhibited in both the variations in phase stability with changing aH2O and textural observations (i.e., the preservation of corona textures), is consistent with data for H diffusion in plagioclase derived from the kinetics of dehydration [Johnson, 2003] and the timescale of deformation (∼3 Myr) estimated by the magnitude of strain and strain rate (supporting information, Figure S3). Because of the steep gradient in chemical potential that existed during deformation, we conclude that the hydration front migrated outward through the gabbro until either the fluid phase was consumed or temperatures were low enough to inhibit the diffusion of hydrous species. This gradient would juxtapose a weaker, wet rheology with a strong, dry rheology; this strength contrast would promote strain localization in the weaker rock.
5.4. Viscosity and Strength of the Lower Continental Crust
 Deformation of the sheared gabbro at a strain rate of 10−13 s−1 and a stress of 20 MPa corresponds to an effective viscosity of 2 ×1020 Pa·s. This viscosity estimate is similar to the non-Newtonian viscosity predicted for lower continental crust in the southwest United States by Freed et al.  and Freed and Bürgmann , who examined the far-field surface displacement after the 1999 Hector Mine earthquake (Figure 9). The results of these modeling studies suggest that the viscosity of the lower crust is approximately an order of magnitude greater than that of the deeper (and higher temperature) upper mantle. Additionally, some researchers conclude that observations of earthquakes in the deep continental crust beneath Tibet indicate that the lower crust is strong and the underlying mantle is weak [Jackson, 2002; Maggi et al., 2000; Bürgmann and Dresen, 2008, and references therein]. However, viscosity estimates based on field observations of sheared rocks [e.g., Homburg et al., 2010] and lithospheric elastic thickness estimates [e.g., Burov and Diament, 1995; Burov and Watts, 2006] suggest that the lower crust is much weaker than the underlying mantle. Our results indicate that strain localization resulting from hydration allows the bulk lower crust host rock to remain strong while deformation is accommodated in weaker regions of high aH2O over small spatial scales. These results reinforce the importance of the combined effects of deformation and fluid on both mafic and felsic assemblages previously described by other researchers [e.g., Austrheim, 1987; Austrheim and Boundy, 1994; Menegon et al., 2011].
5.5. Field and Seismic Strength Estimate Discrepancies: A Comparison With Tectonic Tremor
 The pressure and temperature conditions estimated for our gabbro suite are very similar to those where tectonic tremor occurs on the deep extension of the San Andreas Fault (SAF) [Shelly, 2010; Nadeau and Dolenc, 2005]. Tectonic tremor beneath the SAF is hypothesized to occur where fluid is abundant [Shelly et al., 2006; Ide et al., 2007; Shelly, 2010], which is similar to the scenario that led to strain localization in our gabbro suite. However, the stress estimates predicted for our gabbro suite (∼20 MPa) are significantly higher than those estimated during the low frequency earthquakes at the base of the crust along the SAF, which are triggered by tidally induced shear stresses of 1.77 × 10−4 MPa (differential stress of ∼3.5 × 10−4 MPa, Thomas et al. [2009, Figure 9]). Thus, while our viscosity estimates are also similar to those predicted for the lower crust in this region (Figure 9), some other processes must be responsible for weakening beneath the SAF. There are several possible explanations for the stress estimate discrepancy: (1) the rocks beneath the SAF may not be controlled by plagioclase rheology. (2) There may be more water present beneath the SAF than in our gabbros. Although our phase stability diagrams indicate that aH2O ≈ 1 and water was present to at least 1.13 wt % in our gabbro samples, there is no upper bound on the amount of a free fluid phase (i.e., Figure 6c). (3) While our study shows the importance of fluid on strain localization, our shear zone is much smaller than the plate-scale size of the SAF. Although the processes associated with fluid weakening may be similar at all scales, a larger spatial extent of a fluid alteration zone (and associated deformation) on a large, mature shear zone could manifest in a significantly weaker lithology compared to a smaller, less mature shear zone. This discrepancy between field and seismic stress estimates highlights the importance of further field and laboratory study to reconcile the geodetic and seismic observations with those from exhumed lower crust rocks.
 This study of deformation processes in a mafic, hydrated, lower continental crust shear zone demonstrates the critical role of fluid and synkinematic, amphibole-forming metamorphic reactions for strain localization. Our results suggest that water-enhanced reactions reduce the strength of the lower crust. The correlation of phase assemblages (which constrain water activity) and deformation textures indicates a sharp aH2O gradient was preserved during deformation. This gradient juxtaposes a weaker, wet rheology with a strong, dry(er) rheology, and the strength contrast promotes strain localization in the weaker rock. Strain localization also correlates with a dramatic increase in the amount of amphibole present. While the effect of amphibole growth on rheology is not well constrained by laboratory data, our observations indicate a significant enhancement of diffusion creep in these rocks relative to predictions based on extrapolation of flow laws for wet gabbroic lithologies. Our results, which are consistent with lower continental crust viscosity estimates of Freed et al. , suggest that deformation in the lower crust may be accommodated dominantly in highly localized zones that experienced fluid infiltration, thereby allowing other parts of the mafic lower continental crust to remain relatively undeformed.
 We thank Luca Menegon for help with fieldwork and paper discussions, Bill Collins for assistance with thin sections, Joseph Boesenberg for assistance with the microprobe, and Anthony McCormick for assistance with the SEM. We thank Erling Ravna for directing us to the interesting gabbro outcrops on Arnøya. We are grateful for the helpful reviews provided by Jane Selverstone and Georg Dresen. This research was supported by National Science Foundation grants EAR-0810188 and EAR-1220075 and the Mount Holyoke Alumnae Association Hannum-Warner Fellowship.