Pristine microstructures in pseudotachylytes formed in dry lower crust, Lofoten, Norway

Feldspar-rich pseudotachylytes from the island of Moskenesøya, Lofoten, formed in dry granulites under lower crustal conditions during the Caledonian orogeny. The central parts of the pseudotachylytes, where the cooling rates were slowest, are characterized by microlites and spherulites of plagioclase and K-feldspar. K-feldspar surrounding plagioclase is consistent with crystallization from a melt during cooling instead of devitrification as the origin of the spherulites. Very thin (a few micrometres wide) injection veins, which experienced very rapid quenching, contain amorphous or cryptocrystalline material. The preservation of this material and of the fine-grained microstructures shows that, under fluid-absent conditions, recrystallization and reactions are slow and the original microstructures of the pseudotachylytes can be preserved. This article is part of a discussion meeting issue ‘Understanding earthquakes using the geological record’.


Introduction
Seismic faulting can lead to frictional melting and the formation of pseudotachylytes: dark veins with a glassy to finely crystalline matrix. Pseudotachylytes are most common on faults in the seismogenic regime of the upper and middle crust [1], but have also been found in exhumed lower crustal rocks where they are taken as evidence for intermediate-depth earthquakes (e.g. [2][3][4][5][6][7][8][9]). The origin of these earthquakes is somewhat enigmatic because at the high temperatures and confining pressures in the lower crust, deformation would be expected to be viscous (e.g. [10]). To overcome the high confining pressure and allow for frictional failure, stresses have to be very high or local weakening mechanisms must be active. Sufficiently high stresses cannot be sustained over orogenic timescales and would have to be transient, potentially caused by earthquakes in the shallower crust [11]. In the absence of high stresses, possible weakening mechanisms include thermal runaway [12,13], plastic instabilities [9,14] and dehydration-or reaction-induced embrittlement (e.g. [15]).
Pseudotachylytes originating at all depths are rare in the rock record compared to the expected number of earthquakes [1]. One possible reason for this observation is the presence of fluid in most fault zones: Sibson & Toy [1] suggested that only dry, intact, crystalline rocks produce pseudotachylytes. By contrast, when fluids are present in a fault, thermal pressurization of those fluids during seismic slip would inhibit melt generation. Another explanation for the scarcity of pseudotachylytes is that their fine-grained or amorphous nature makes them susceptible to deformation, recrystallization and alteration, so that they are easily overlooked in the rock record [16]. The introduction of fluids during or after faulting in particular causes alteration of the mineral assemblage in the pseudotachylyte. In the absence of fluids and deformation, pristine pseudotachylyte microstructures can be preserved. This has been the case for pseudotachylytes in feldspar-rich rocks on Moskenesøya in Lofoten, Northern Norway. The microstructures of the Moskenesøya pseudotachylytes, which formed under lower crustal conditions during the Caledonian orogeny, are remarkably well preserved. A general description of these pseudotachylytes has been provided by Dunkel et al. [17], who interpreted them to be the result of transiently high stresses in the lower crust based on the absence of ductile deformation and the anhydrous mineralogy (see §3b). Here, we describe the microstructures of the pseudotachylytes in detail via scanning electron microscopy (SEM) and transmission electron microscopy (TEM) and associated analytical techniques and discuss their development.

Methods
Thin sections were prepared perpendicular to the faults that show pseudotachylyte veins.
A Hitachi SU5000 FE-SEM (Field Emission Scanning Electron Microscope; Department of Geosciences, University of Oslo) was used for backscatter electron imaging and electron backscatter diffraction (EBSD) analysis. Crystallographic orientations of the minerals in the samples were determined by indexing of EBSD patterns, as acquired with a Bruker e-Flash detector. For EBSD mapping, the FE-SEM was used at low vacuum with uncoated samples, an accelerating voltage of 20 kV and a working distance of 20 mm. The step size used for EBSD mapping was 0.3 µm. EBSD patterns were indexed automatically using the Esprit v. 2.1 software from Bruker. Qualitative energy-dispersive spectrometry (EDS) element maps, which were acquired concurrently to the EBSD maps, were used to assist the phase identification. Visualization of the crystallographic orientation, calculations of the orientation distribution functions (ODFs) and plotting of pole figures were conducted using the Matlab toolbox MTEX (v. 5.5.1; http://mtex-toolbox.github.io) [18][19][20]. For the ODF calculations, the de la Vallée Poussin standard orientation distribution function was used, with a halfwidth of 10°.
Three in situ TEM lamellae were prepared using a Helios 600i focused ion beam (FIB)-SEM machine. Initially, a 15 × 2 × 2 µm (length × width × thickness) layer of platinum was deposited on top of the region of interest using the FIB operating at 30 keV and a beam current of 300 pA. The frontal and back trenches were created with the Ga-FIB operating at 30 keV and 15 nA

Background (a) Geological setting
The Lofoten-Vesterålen islands are part of the Norwegian Caledonides and consist of an Archean to Paleoproterozoic metamorphic complex of para-and orthogneisses. This sequence was intruded under granulite-facies conditions at approximately 1870-1770 Ma by an Anorthosite-Mangerite-Charnockite-Granite (AMCG) suite that makes up about 50% of the archipelago [21,22].
During the Caledonian orogeny (ca 490-390 Ma), western Baltica was partially subducted under Laurentia; however, the relation of the Lofoten block to Baltica and the Caledonides is unclear. In Early Carboniferous plate reconstructions, where east Greenland is welded to the Norwegian margin, Lofoten has a central location in the northern parts of the restored orogeny. It records very few Caledonian deformation structures [5], which is probably due to limited fluid availability and thus the high strength of the anhydrous granulites [23].

(b) Previous work
The pseudotachylyte samples discussed in this study were collected at roadcuts along the east coast of Moskenesøya (figure 1a). The occurrence and chemical and mineralogical composition of these pseudotachylytes were described by Dunkel et al. [17]: they occur as thin (ca 1-15 mm), dark veins in massive, undeformed granulites of mangeritic and gabbroic composition. They often have injection veins or form pseudotachylyte breccias (figure 1b). The host rocks contain mainly plagioclase and K-feldspar, variable amounts of pyroxenes (orthopyroxene in the mangeritic, ortho-and clinopyroxene in the gabbroic varieties) and amphibole. Ti-bearing magnetite is common in both rock types. Small amounts of quartz are limited to the mangeritic rocks. The pseudotachylytes reflect the mineralogical composition of their hosts, with the addition of some garnet, which only occurs as rare overgrowths in the host rocks and is slightly more common in the pseudotachylytes.
Because this mineralogy did not allow for conventional geothermobarometry, the pressure conditions during pseudotachylytes formation were estimated using elastic geobarometry on quartz inclusions in garnet [17]. Other, hydrous pseudotachylytes and shear zones in Lofoten record temperatures of ca 650-750°C. Assuming similar temperatures for the Moskenesøya pseudotachylytes, pressures were in the range 0.8-1.2 GPa, indicating lower crustal formation conditions for the pseudotachylytes. Additionally, the absence of hydrous phases alone suggests that faulting did not occur in the shallow crust, where fluid infiltration usually accompanies or follows an earthquake.
Neither the host nor the clasts in the pseudotachylyte show evidence of pre-seismic ductile deformation. Therefore, Dunkel et al. [17] excluded thermal runaway or plastic instabilities as possible mechanisms for the formation of these lower crustal earthquakes. Dehydration embrittlement is not considered a viable mechanism either because of the lack of biotite, which would have formed under hydrous conditions. In the absence of such weakening mechanisms, Dunkel et al. [17] suggested transiently high stresses in the dry, strong granulite-facies rocks.

Results
All mineral phases present in the host rock (plagioclase, K-feldspar, pyroxene, magnetite, amphibole, quartz, see §3b) also occur as angular to subrounded lithoclasts in the pseudotachylyte. The main newly grown phases in the pseudotachylyte matrix are plagioclase, K-feldspar, garnet and magnetite. In the following, we describe the micro-and nanostructures of those phases in detail.

(a) Pseudotachylyte microstructures
Representative samples from two localities are shown in figures 2 (locality ØL) and 3 (HAM). The feldspars exhibit a variety of fine-grained microstructures in the pseudotachylyte matrix (see below), but some large feldspar clasts (diameters on the order of 100 µm) have been observed as well. Magnetite occurs mainly as small grains (diameter of a few micrometres). Rare larger grains have diameters comparable to the magnetite grains in the host rock. Pyroxene and amphibole occur as small angular to slightly embayed clasts (ca 10 µm in diameter). Quartz is rare and has been observed as both larger clasts and as part of the pseudotachylyte matrix. Garnet has a dendritic to snowball habit with occasional euhedral overgrowths (diameter few 10 s of µm), with inclusions of the other pseudotachylyte phases (figure 2c).    ( figure 5). In some pseudotachylyte veins, a zonation with microlites/spherulites in the centres and equant grains towards the edges can be observed (figures 3d and 4d). However, in other pseudotachylytes, no such zonation is visible, which might be due to the complex shape of the pseudotachylyte network.
In the microlitic or spherulitic domains, plagioclase laths are often surrounded by K-feldspar ( figure 4a,d). In some cases, however, the two feldspars are intergrown and the microlite bundles or spherulites as a whole are rimmed with K-feldspar (figures 4c and 5c). Even though the       common, larger plagioclase crystals have high dislocation density cells, which can be partially organized in subgrain walls with similar dimensions to the recrystallized grains just adjacent to the large grain (figure 7d). Nanometre-wide pseudotachylyte veins intruded the polycrystalline host and partially corroded the plagioclase in contact with the melt (figure 7c). In the bright-field TEM mode, plagioclase (figure 8a) and the small Fe-Ti oxides in the pseudotachylyte (figure 8b) display some dislocations, as shown by the variable contrast, whereas K-feldspar is largely strain free. Locally, nanometre-scale rounded crystals of plagioclase and garnet occur within an amorphous portion of the TEM lamella (figure 8b), which has a similar composition to point 4 in figure 8.
EDS analysis of the pseudotachylyte veins from TEM lamella 3 show plagioclase (figure 9, points 1 and 3) and K-feldspar (figure 9, point 2) microlites/nanolites, with aspect ratios on the order of 5 : 1 and no clear preferred orientation relative to the vein walls (figure 9a-c). Ti-rich magnetite with grain sizes from 100 to 400 nm is also common in the pseudotachylytes (figure 9c). The matrix has a K-feldspar-like composition (figure 9, point 4).

Discussion
In the following, we will interpret the evolution of the studied fault rocks over the very short time interval from brittle failure to complete solidification of the frictional melt based on the petrographic observations described above and discuss the implications of the preservation of the described microstructures.

(a) Fragmentation, melting and flow
During dynamic rupture and subsequent frictional slip, the host rock was fragmented, producing the clasts of feldspar, quartz, magnetite, pyroxene and amphibole that are later preserved as lithoclasts in the pseudotachylyte (e.g. Figure 3a  Note that the little peaks of Ga and Cu are not related to the sample, but due to FIB sample preparation (Ga) and the TEM lamellae grid (Cu). An acicular feldspar in (b) is divided, perpendicular to its long axis, into plagioclase (point 1) and K-feldspar (2). Point 3 shows another plagioclase analysis; 4 is an analysis of the matrix between the acicular crystals, which has an alkali-feldspar-like composition.
(b) Crystallization of feldspar-the role of nucleation rate and undercooling rocks, but also very common in pseudotachylytes [29]. While the standard view of spherulites is often single, acicular crystals radiating outwards from a common centre, there are a multitude of possible morphologies, including densely branched spherulites (e.g. [30]). We suspect that this is the case here and that many of the small feldspar grains visible in the spherulites (figure 5c,d) are connected in three dimensions. Their origin can be devitrification of glassy material or rapid crystallization during quenching. In the present case, rapid crystallization from a melt is the most probable mechanism because of the spatial relation between plagioclase and K-feldspar (see below).
During direct crystallization from a melt, the presence or absence of equant grains, single microlites and spherulites can be used as rough indicators of cooling rates, but those microstructures also depend on other factors such as composition and water content [31]. In general, large undercooling favours the formation of surface instabilities during crystal growth because of the associated low diffusivity and fast growth rate [32]. In experimental crystallization of plagioclase at 0.5 GPa, Lofgren [33] observed spherulitic growth at undercooling from 150°C to 400°C. However, those experiments were conducted under water-saturated conditions, whereas we assume that the melt was anhydrous (and at higher pressure).
The cooling rate was lower in the centres of the pseudotachylyte veins or pockets than at the edges, in contact with the wall rock. This scenario resulted in a higher degree of resorption of feldspar clasts in the vein centres, causing a decrease in the amount of wall rock fragments from vein margin to centre. The higher degree of undercooling in the outer parts of the veins also led to a higher nucleation rate. Combined with the higher number of clasts that can provide sites for heterogeneous nucleation, this resulted in the growth of many small, equigranular feldspar grains. However, the occurrence of nanolites observed in TEM in ØL-1A (figure 9) shows that structures indicative of fast crystallization can be present even when not visible at lower (SEM) magnification. What appears as equigranular grains in SEM observations may be aggregates of nanolites in a cryptocrystalline or amorphous matrix. Toward the centres of the pseudotachylytes, the nucleation rate and number of clasts was lower, while the growth rate was still high, leading to the formation of microlites and spherulites of plagioclase visible even in optical microscopy (figures 3b and 5a,b).
The intergrown plagioclase and K-feldspar spherulites rimmed by K-feldspar (figure 5c) indicate eutectic crystallization of both feldspars followed by the growth of pure K-feldspar. In other cases, K-feldspar occurs around plagioclase laths (figure 3f ). K-feldspar surrounding plagioclase indicates the direct growth of the two feldspar phases from a melt, because this spatial separation is unlikely to have originated during devitrification. A similar separation of plagioclase and K-feldspar has been observed in pseudotachylytes formed in the Adamello tonalites under shallower conditions, with K-feldspar-rich rims around plagioclaserich spherulites [34].

(c) Preservation of pseudotachylyte microstructures
The microstructures described above, including amorphous and/or cryptocrystalline material are clearly inherited from the time interval of dynamic rupture, frictional sliding and cooling towards ambient temperatures. There is no evidence for reactions (including hydration reactions) after the quenching of the frictional melt. Incipient deformation may be recorded in some pseudotachylytes by the elongated shape of clasts (figure 2) and dislocation movement (figure 7d). Similarly pristine lower crustal pseudotachylytes have been described from the Musgrave Ranges in Australia [2,35], the Ivrea Zone [4] and Nusfjord East in the Lofoten area [6]. This preservation is remarkable considering the long history of these rocks after their formation, including the uplift to lower P, T-conditions. Thermal models suggest that the cooling of a thin pseudotachylyte melt to ambient temperature lasts for a maximum of seconds to minutes (e.g. [36,37]). The absence of recrystallization and deformation after the formation of the pseudotachylyte demonstrates that in dry systems at lower crustal temperatures, solid-state nucleation and growth processes are