The Annapurna Detachment (AD) is a low-angle (∼20°–30° dip), north-dipping normal fault and ductile high-strain shear zone in calc-mylonites, and forms part of the South Tibetan Detachment (STD) that runs along 1800 km length of the Himalaya. The AD separates kyanite and sillimanite grade gneisses and tremolite + clinopyroxene ± hornblende-bearing marble–calc-silicates of the Greater Himalayan Sequence (GHS) below from unmetamorphosed Palaeozoic–Mesozoic sedimentary rocks of the Tethyan sedimentary zone above. It was active at ca. 22–18 Ma during south-vergent ductile extrusion (channel flow) and exhumation of the Himalayan mid-crust footwall. Restoration of the STD system suggests around 80–100 km of southward extrusion of the footwall gneisses relative to the Tethyan hanging-wall rocks. Folds in the hanging wall of the AD were formed prior to normal faulting, but axial planes are curved into alignment with the shear zone suggesting extrusion of the metamorphic footwall rocks beneath a passive roof fault. North-vergent recumbent backfolds in the Nilgiri–Tukuche peaks were enhanced by backsliding during footwall extrusion, although this does not indicate “orogenic collapse,” lowering of surface elevation, or decreasing crustal thickness because new material was continually being underthrust from the south. Axial planes of backfolds are curved and progressively rotate from subvertical in the north to subhorizontal immediately above the AD. Low-angle normal faults in the Himalaya were active during the Early Miocene, concomitantly with thrusting at deeper structural levels along the Main Central Thrust (MCT) zone. The passive normal faults and ductile shear zone were initiated at low angles aided by partial melting and ductile flow within the GHS. They do not indicate alternating periods of extension and compression but were active in a wholly compressional environment. The Channel Flow model for the Greater Himalaya and the passive roof fault model for the low-angle normal faults adequately explain all geological field structural and metamorphic criteria.

Rock mechanics theory suggests that active extensional normal faults can only have dip angles of >30° (Anderson, 1951). Almost all earthquakes on active normal faults occur on faults with dips at angles higher than 30°. Earthquakes associated with normal faulting generally nucleate at depths ca. 10–20 km around the brittle-ductile transition and their ruptures propagate upwards toward the Earth's surface. As extension proceeds faults may tilt toward shallower angles of dip as a result of isostatic adjustment and flexure, when they become inactive as a new steeper active fault develops, usually in the hanging wall (Buck, 1988). In many classical rift valleys such as the Gulfs of Corinth and Evvia (Greece) for example, steep, active, normal faults with dips ∼60°–80° occur along the flanks of the rift axis with older, inactive normal faults further away from the rift axis at shallower angles of dip (e.g., Jolivet et al., 1996; Ring et al., 2001; Papanikolaou and Royden, 2007). Mapping of some Pliocene-active rifts in Tibet has also led to a similar evolutionary model with initial high-angle normal faults evolving with time into low-angle detachments with increasing extension and isostatic footwall rebound (Kapp et al., 2008).

Many actively extending continental areas in the world expose low-angle normal faults, most notably in the Basin and Range Province, western United States (e.g., Wernicke, 1981; Davis, 1983; Yin, 1989), in metamorphic core complexes such as those in Greece and the Aegean Sea (e.g., Lister et al., 1984; Ring et al., 2001; Chéry, 2001) and along numerous passive continental margins. Seismic reflection images and earthquake source parameters in the Woodlark basin east of Papua New Guinea show that a magnitude 6.2 earthquake occurred at a depth of 5.2 km on a low-angle normal fault that dips at 25°–30° (Floyd et al., 2001). Morley (2009) has recently shown that many examples of low-angle normal faults in Cenozoic rift basins of Thailand were initiated at low angles (20°–30° dips) and not by isostatic rotation of high-angle faults. Despite abundant geological evidence for low-angle normal faulting in the continents, only a few if any earthquakes presently occur on active low-angle normal faults (Axen, 1999; Abers, 2009). It has recently been suggested that some low-angle normal faults may be active in the brittle regime but aseismically creeping as opposed to seismic stick-slip motion (Hreinsdottir and Bennett, 2009).

Classic Andersonian theory assumes that orientations of principal stresses are vertical and horizontal. Initiation of a low-angle normal fault, however, requires the stress tensor in the brittle upper crust to have inclined principal stress axes and therefore major shear stress in the vertical plane (Westaway, 1999). This extreme situation requires special conditions where dramatic variation occurs across the extending region. Westaway (1999) suggested that shear stresses of ∼100 Mpa could occur near the base of ∼10-km-thick brittle upper crust due to the combined effects of lower crustal flow and loading. Lower crustal flow would cause horizontal shear traction along the base of the brittle upper crust (e.g., Yin, 1989). This proposed geometry, vertical rheological differences, and stress conditions are similar to that seen along the Himalaya, where a low-angle, north-dipping normal fault and ductile shear zone (South Tibetan Detachment [STD]) bounded the upper part of a southward extruding partially molten mid-crustal layer (Greater Himalayan Sequence of high-grade metamorphic rocks, migmatites, and anatectic granites) during the Miocene (e.g., Burg et al., 1984; Burchfiel and Royden, 1985; Searle, 1986; Searle and Rex, 1989; Grujic et al., 2002; Searle et al., 2003, 2006, 2007; Law et al., 2004; Grujic, 2006; Godin et al., 2006a; Cottle et al., 2007).

Whereas the Basin and Range province, the East African rift, and the Aegean–Greece regions are all undergoing whole-lithosphere extension, the Himalaya has been under continued north-south lithospheric compression since the India–Asia collision in the Lower Eocene ca. 50 Ma (Green et al., 2008), which continues to this day. Thus, despite some geometric similarities in the field relationships, the South Tibetan Detachment low-angle normal faults are not similar to the Basin and Range low-angle normal faults. Some earlier interpretations of the low-angle normal faults along the Himalaya include “orogenic collapse” (e.g., Dewey, 1988) and alternating periods of contraction and normal faulting or the transference of motion from the Main Central Thrust (MCT) to the Main Boundary Thrust (MBT) (England and Molnar, 1993). However, detailed fabric analysis combined with U-Pb dating of gneisses and leucogranites shows that MCT-related thrusts and STD-related low-angle normal faults and their respective ductile shear zones were active simultaneously (e.g., Hodges et al., 1992, 1996; Searle et al., 1992, 1999, 2007, 2008; Hodges, 2000, 2006; Law et al., 2004). Models showing early Himalayan compression followed by a phase of extension are discounted because timing of folding and thrusting and present day GPS show that the Himalaya has been under continuous compression since at least 50 Ma. It is suggested that these faults do not indicate lowering of surface elevation or decreasing crustal thickness (“orogenic collapse”). On the contrary, during active Miocene Channel Flow and motion along both the MCT and the STD, the Tibetan Plateau and the Himalaya continued to rise and thicken due to continued underthrusting of Indian plate material from the south.

This paper reviews and describes the STD low-angle normal faults in the Annapurna– Dhaulagiri Himalaya in Nepal and the spectacular north-vergent folds above the detachment in the Dhaulagiri and Annapurna ranges of Nepal (Fig. 1). Three cross sections are presented along the Modi Khola–Machhapuchare transect and across the Nilgiri–Annapurna and Dhaulagiri ranges, together with one structural profile along the Kali Gandaki GHS profile. It is concluded that the STD formed a passive roof (“stretching”) fault (Means, 1990; Searle et al., 2003, 2006; Law et al., 2004; Cottle et al., 2007) during the Miocene southward-extrusion of the mid-crustal GHS footwall (Channel Flow) and that it was active during the Late Miocene at similar low angles to that seen today. Subhorizontal mid-crustal channel flow and low-angle normal faulting along the brittle-ductile transition resulted from partial melting within the core of the GHS. Eocene–Oligocene folds were refolded during Miocene extrusion of the GHS beneath. The spectacular recumbent folds on Nilgiri and Tukuche peaks were enhanced by north-vergent drag folding during southward-directed channel flow and exhumation of the GHS along the footwall of the STD. The passive roof fault model, combined with the Channel Flow model, shows how low-angle normal faults can operate in regions of crustal convergence, shortening and thickening, and require no crustal or lithospheric extension at all.

Extensive field-based research in central Nepal by Bordet et al. (1971), LeFort (1975), Pêcher (1989), and Guillot et al. (1993, 1995) included publication of the comprehensive 1:200,000 scale geological map of the Annapurna–Manaslu–Ganesh Himalaya (Colchen et al., 1980, 1986). This map summarized all the stratigraphic data for the Tethyan sedimentary zone but failed to recognize the major low-angle normal faults of the STD system. Caby et al. (1983) working in the Annapurna region and Burg et al. (1984) in the south Tibet region were the first to map a normal fault along the position of the STD. Detailed studies by Brown and Nazarchuk (1993), Vannay and Hodges (1996), Godin et al. (1999a, 1999b, 2001), and Godin (2003) mapped out these structures along the Kali Gandaki, and Hodges et al. (1996) mapped at least three north-dipping normal faults along the Modi Khola and in the Annapurna Sanctuary region.

At deeper structural levels evidence of ductile “extension” in high-grade metamorphic rocks is known from the Greater Himalayan Sequence (GHS). Coleman (1996) correlated the Chame Detachment in the Marsyandi valley with the STD, but Searle and Godin (2003) pointed out that pressure-temperature (P-T) conditions of rocks above and below are similar (580–560 °C; 4–5 kbar; Schneider and Masch, 1993) and reinterpreted the Chame Detachment as a ductile shear zone wholly within the GHS. Searle and Godin (2003) and Searle et al. (2008) reinterpreted the original Colchen et al. (1986) location of the STD and MCT faults based on high strain zones and metamorphic grade of rocks above and below (Fig. 1). They mapped a major low-angle ductile shear zone and normal fault at higher structural levels (Phu Detachment) wrapping around the top of the Manaslu leucogranite, which they correlated with the main STD. This placed the Manaslu leucogranite entirely in the footwall of the STD, in common with all other Himalayan leucogranites (Searle et al., 2009), rather than intruding across it into the Tethyan sedimentary zone above (Guillot et al., 1993; Harrison et al., 1995, 1999).

The GHS sections along the Kali Gandaki (Vannay and Hodges, 1996; Godin et al., 2001; Larson and Godin, 2009), Modi Khola (Hodges et al., 1996), Marsyandi (Coleman, 1996; Coleman and Hodges, 1998), Nar (Searle and Godin, 2003; Gleeson and Godin, 2006; Godin et al., 2006b), and Burhi Gandaki river valleys (Larson et al., 2010) have now also been mapped in greater detail. Figure 2 shows the structural correlations of all the low-angle normal faults along the Dhaulagiri, Annapurna, and Manaslu massifs. The location of the MCT has always been hotly disputed. Initial locations placed the MCT along or close to the kyanite isograd (LeFort, 1975; Colchen et al., 1980, 1986), while other authors used lithological differences (Gansser, 1964), differences in Nd isotopes (DeCelles et al., 2000; Robinson et al., 2001; Richards et al., 2005) or detrital zircon ages (Parrish and Hodges, 1996), or combinations of these (Martin et al., 2005) to determine the location of the MCT. Searle et al. (2008) mapped the MCT at lower structural levels based on strain gradients and metamorphic fabrics and suggested the base of the inverted metamorphic sequence was the most logical place to map the thrust, in common with the western Himalaya (Stephenson et al., 2000, 2001; Searle et al., 2007). Thus all rocks above the MCT were affected by some degree of Tertiary Himalayan metamorphism, whereas those below were largely unaffected by Tertiary metamorphism.


A continuous Palaeozoic and Mesozoic sedimentary succession is exposed in the Annapurna and Dhaulagiri ranges and the Kali Gandaki River, which cuts in between both mountains, providing an excellent transect across the entire North Indian plate margin in this region of the Nepal Himalaya (Fig. 3). The upper crustal rocks of the Nepal Himalaya comprise ∼10–12 km total thickness of Cambrian to Lower Cretaceous sedimentary rocks (Bordet et al., 1971; Colchen et al., 1980; Godin, 2003). Although not exposed at the present-day erosion surface in the Annapurna region, it is likely that there is a considerable thickness (∼2–3 km) of Proterozoic black shale or low-grade metasedimentary rocks beneath the Cambro–Ordovician limestones comparable to the Haimanta Series in the western Himalaya and the Cheka Formation in the Eastern Himalaya. The base of the sedimentary sequence in the Annapurna–Dhaulagiri range comprises up to 2500 m stratigraphic thickness of thick-bedded carbonate rocks increasing in metamorphic grade toward the base. The Sanctuary and Annapurna Yellow (also called Larjung) Formations, comprising biotite-grade calcareous and semipelitic schists and metacarbonates, are assumed to be Cambrian in age because they underlie stratigraphically dated Ordovician limestones of the Nilgiri Formation (Bordet et al., 1971; Colchen et al., 1980, 1986). These massive bedded micritic limestones are overlain by ∼400 m of pink sandstones and quartzites (North Face quartzites) and are nearly 2000 m thick on the peaks of Nilgiri and Dhaulagiri. Over 2000-m-thick upper Palaeozoic black shales, limestones, and arenaceous sandstones make up the Sombre, Tilicho Lake, and Thini Chu Formations. Mesozoic sediments are represented by Triassic limestones and calcareous shales, Jurassic black shales with a rich ammonite and belemnite fauna, and Cretaceous greywackes and calcareous sandstones (Colchen et al., 1980; Gradstein et al., 1992).

The tectonic pseudo-stratigraphy of the highly metamorphosed and deformed GHS below the STD (Fig. 3) can also be compared to the stratigraphy of the unmetamorphosed rocks above. GHS structural unit I kyanite-grade pelites and unit II dominantly calc-silicates with less common biotite-hornblende gneisses have Neoproterozoic–Cambrian protolith ages (Gehrels et al., 2003). The Ulleri augen gneiss has zircon ages of 1.83 Ga and intrudes Kuncha pelites, which also have Proterozoic deposition or protolith ages (DeCelles et al., 2000). Palaeoproterozoic orthogneisses of similar age (1.8 Ga) have also been reported from the Ama Drime massif, north of Everest, the northern extremity of the GHS rocks in the Nepalese Himalaya (Cottle et al., 2009), showing that similar protolith ages extend across the entire Himalaya. No Archean rocks are known from the Himalaya, so the Indian Shield Archean basement rocks that once underlay the Lesser, Greater, and Tethyan Himalaya must have underthrust northwards beneath the Lhasa block of southern Tibet. GHS structural unit III augen gneisses have Cambrian (ca. 500 Ma) U-Pb zircon ages and are interpreted as metamorphosed Cambrian granites (Hodges et al., 1996). GHS units IV and V (Larjung unit) are thought to be metamorphosed equivalents of the Cambro–Ordovician Annapurna and Nilgiri Formations.

The Annapurna–Dhaulagiri Himalaya were initially mapped in detail by Colchen et al. (1980, 1986) and later by Godin (2003). Hodges et al. (1996), Larson and Godin (2009), and Larson et al. (2010) mapped the GHS rocks south of Annapurna and Dhaulagiri. Figure 2 shows a compilation map of the STD zone in the Dhaulagiri–Annapurna range, including the Kali Gandaki and Modi Khola transects. Three cross sections across the Annapurna–Machhapuchare (Fig. 4), Nilgiri (Fig. 5), and Dhaulagiri (Fig. 6) ranges are also presented balanced for the known stratigraphic thicknesses, but bearing in mind the ductile style of folding in the deeper structural levels. The stratigraphic thicknesses used follow the work of Colchen et al. (1980, 1986) except where balancing constraints suggest differences. For example, the thickness of the Cambrian Sanctuary and Annapurna Yellow formations must be considerably thicker than suggested by Colchen et al. (1980, 1986).

The Annapurna–Machhapuchare section (Fig. 4) follows that initially published by Hodges et al. (1996). The Nilgiri cross section (Fig. 5) is similar to that presented in Godin (2003, fig. 3a) and Kellett and Godin (2009, fig. 7) with a few critical differences. Godin (2003) mapped a rootless F1 isoclinal fold, the Fang nappe, initially mapped by Colchen et al. (1986), that was refolded during F2. My interpretation differs in that the Fang fold (Fig. 7) is part of the main north-directed backfolds and is directly along strike with the Marpha fold above the Annapurna Detachment. Another difference is the recognition of a minor north-directed backthrust between the Nilgiri fold and Marpha fold. The Dhaulagiri cross section (Fig. 6) follows a similar line to that of Godin (2003, fig. 3c). On Godin's section, the Annapurna Detachment dips to the NNE at ∼34°, necessitating a much greater thickness of Cambrian–Ordovician limestones at depth than stratigraphy allows, whereas on my cross section it dips at 28° beneath Dhaulagiri, flattening to ∼5° beneath Syang peak and Jomoson to the north.

The STD runs the length of the Himalaya and forms the upper structural boundary of the Greater Himalayan Sequence (GHS) of metamorphic rocks, migmatites, and Miocene leucogranites (e.g., Hodges, 2000, 2006; Brown and Nazarchuk, 1993; Searle et al., 1997a, 2003, 2006; Godin, 2003). In the Annapurna Range three low-angle normal faults, the Hiunchuli, Machhapuchare, and Deurali Detachments (Fig. 3), have been mapped, each placing low-grade rocks on top of high-grade rocks (Hodges et al., 1996). These authors interpreted the low-angle normal faults as gravitationally driven compensation structures that were active simultaneously with thrusting along the Main Central Thrust (MCT) system at the base of the GHS. In the Nilgiri and Dhaulagiri ranges, west of the Annapurnas, these multiple normal faults appear to merge into one very large ductile shear zone, termed the “high-strain zone,” over 1500 m thick by Godin et al. (1999a, 1999b), with later low-angle brittle normal faults cutting through the top (Annapurna Detachment).

Deurali–Chame Detachment

The oldest structure in this region appears to be the structurally lowest normal fault, the Deurali Detachment, which strikes WNW and dips at ∼30° NNE. It was active at the same time as the highest MCT-related thrust fault, the Chomrong thrust, which places kyanite-grade gneisses with in situ partial melts south over lower amphibolite facies rocks. U-Pb geochronology shows that both faults were active at ca. 22.5 Ma (Hodges et al., 1996; Coleman, 1996; Coleman and Hodges, 1998). The southern limit of Tertiary Himalayan metamorphism coincides with another south-vergent thrust, the Chrandrakot thrust (Fig. 4), and this is where Searle et al. (2008) mapped the MCT. The Deurali Detachment is marked by a 1.2-km-thick mylonite zone with abundant top-to-north kinematic indicators and NNE-plunging stretching lineations. In the Marysandi valley Coleman (1996) and Coleman and Hodges (1998) mapped a similar >300-m-thick mylonite zone termed the Chame Detachment, which they correlated with the STD, in a structurally equivalent place to the Deurali Detachment. However, as pointed out by Searle and Godin (2003), rocks above the Chame Detachment are high-grade metamorphic rocks (diopside + K-feldspar + tremolite marbles, calc-silicates, and sillimanite gneisses) with P-T conditions similar to rocks beneath the detachment, and therefore this fault is not the same as the STD as mapped by others elsewhere along the Himalaya. The Deurali–Chame Detachment is interpreted as a low-angle ductile shear zone with top-to-north or base-to-south motion wholly within the GHS metamorphic sequence. The STD separating rocks affected by Tertiary Himalayan metamorphism below, from rocks largely unaffected by Himalayan metamorphism above, lies structurally higher in the section along the Machhapuchare Detachment.

Machhapuchare Detachment

The Machhapuchare Detachment appears to mark the main northward limit of Tertiary Himalayan metamorphism and therefore marks the position of the main STD in this region. It strikes WNW, dips at ∼25° NNE, and separates very low-grade or unmetamophosed Cambrian limestones above from amphibolite facies marbles and calc-silicates intruded by Miocene leucogranite dykes below (Hodges et al., 1996). To the north and east, the Machhapuchare Detachment can be correlated with the Phu Detachment, the main branch of the STD that wraps around the upper levels of the Manaslu leucogranite and the Chako dome (Searle and Godin, 2003; Godin et al., 2006b; Gleeson and Godin, 2006). The Phu Detachment marks an abrupt decrease in metamorphic grade structurally upwards. Below the fault in the Chako dome (Nar valley), sillimanite gneisses and diopside-bearing calc-silicates have been intruded by a network of leucogranite dykes and sills emanating from the western margin of the Manaslu leucogranite, whereas above the fault are unmetamophosed Palaeozoic sedimentary rocks (Searle and Godin, 2003; Gleeson and Godin, 2006). Godin et al. (2006b) dated one folded and boudinaged dyke using U-Pb zircon at 20.05 ± 0.06 Ma. In the Nar valley the age of the Phu Detachment is constrained as between 18 and 16 Ma (Searle and Godin, 2003; Godin et al., 2006b), the U-Pb age of the youngest phase of the Manaslu leucogranite (Harrison et al., 1999). In the Modi Khola (Fig. 4) the age of motion on the Machhapuchare Detachment is constrained as younger than 18.5 Ma, the U-Pb age of leucogranites in the footwall (Hodges et al., 1996).

Hiunchuli Detachment and Fang Nappe

In the Annapurna Sanctuary region two more north-dipping normal faults structurally above the Machhapuchare Detachment have been mapped around the peaks of Machhapuchare (6993 m) and Ghandharba Chuli (6248 m). On the Colchen et al. (1980) map, these rocks are all mapped as Cambrian Sanctuary–Annapurna Yellow Formations. The tight syncline of the Machhapuchare fold is bounded below by the Hiunchuli Detachment and above by the other unnamed north-dipping normal fault. The Machhapuchare fold can be traced along strike to the west into the southward closing “Fang nappe,” a SSW-verging isoclinal fold that has been interpreted as refolded by the Nilgiri north-vergent fold (Colchen et al., 1980, 1986; Godin et al. (1999a, 1999b, 2001). The critical refolded fold closure crops out near the summit of the Fang (7647 m; Fig. 7) on the Colchen et al. (1980, 1986) map. Three-dimensional mapping of this structure and extrapolation of the fold axis plunge (Fig. 4) shows that the Machhapuchare–Fang fold pair is probably continuous to the synclinal fold separating the Nilgiri and Marpha north-vergent folds in the Kali Gandaki (Fig. 5). If this is correct then there is no need to infer a special deformation phase (the earliest south-vergent fold phase D1 of Godin et al. (1999a, 1999b; 2001) only for this one “Fang nappe” structure.

Annapurna Detachment

The Deurali, Machhapuchare, and Hiunchuli Detachments mapped along the Modi Khola and in the Annapurna Sanctuary (Hodges et al., 1996) merge to the west into one thick ductile shear zone (Fig. 8). Caby et al. (1983) first mapped a low-angle normal fault in the Kali Gandaki, but Brown and Nazarchuk (1993) described it in more detail and correlated it with the STD as defined elsewhere along the northern Himalaya (Burchfiel et al., 1992). Brown and Nazarchuk (1993) and Godin et al. (1999a, 1999b, 2001) mapped the Annapurna Detachment as following the upper contact of unit 3 orthogneisses of the GHS and below the calc-silicates and marbles of the Larjung Formation, which they regarded as the base of the Tethyan sedimentary sequence. However, these rocks are upper amphibolite facies marbles and calc-silicate gneisses containing the assemblage Cal + Qtz + Bt + Ms + Pl + Kfs + Scp + Cpx ± Grt ± Hbl (Vannay and Hodges, 1996) and in places show partial melt or migmatitic textures (Fig. 9A). In this study the “Larjung unit” belongs to the GHS, not to the base of Tethyan series, and the Annapurna Detachment lies structurally above. The Larjung unit calc-silicates are also intruded by garnet + muscovite + tourmaline leucogranite dykes, one of which has a U-Pb monazite age of 22.5–23.0 Ma (Godin et al., 2001). These dykes are not continuous upwards into the unmetamorphosed Tethyan zone north of Larjung (Fig. 2). Melting is a very passive process with schistosity in the blocks broken up by leucogranite melt showing parallel fabrics with the host gneisses (Fig. 9B). This upper amphibolite facies assemblage is comparable in grade to the kyanite-bearing pelites of GHS unit 1, which also contain in situ leucosome melt pods (Fig. 9C). U-Pb monazite dating of kyanite-grade metamorphism and partial melting indicate Oligocene ages between 35 and 31.5 Ma (Godin et al., 2001). The new placement of the Annapurna Detachment in this study now maps it along the clear planar north-dipping low-angle normal fault that separates the unmetamorphosed limestones of the Nilgiri–Tukche folds above from the high-grade tremolite + diopside marbles and calc-silicates intruded by leucogranite dykes beneath (Fig. 8). The age of this low-angle normal fault must be <22.5 Ma from the U-Pb ages reported by Godin et al. (2001). The contact corresponds to a major metamorphic break and is marked by a thick series of platy calc-mylonites dipping at 20°NNE (Fig. 9D). Tight to isoclinal folds are common in the ductile shear zone that underlies the low-angle normal fault.

Kyanite is stable in metapelites of the GHS from the basal shear zone of the Annapurna Detachment structurally down to Dana in the Kali Gandaki transect, a structural thickness of ∼10 km (Fig. 8). U-Pb monazites growing in equilibrium with peak kyanite-grade assemblages yielded ages spanning 35.0–31.5 Ma, thereby providing clear evidence of Oligocene metamorphism (Godin et al., 2001). Kyanite-bearing partial melts show that in situ melting occurred at depths of ∼35–40 km during the Oligocene. The common and widespread Miocene sillimanite Eo–Himalayan metamorphism, so prevalent elsewhere along the Himalaya, is greatly reduced in the Kali Gandaki profile, although this may be a result of the greater volume of calcareous metasediments rather than pelitic rocks in the Kali Gandaki. Thus the mid-crustal channel flow may have originated from deeper levels than shown in the models of Beaumont et al. (2001, 2006).

Also lacking in the Dhaulagiri and Annapurna sections are the large leucogranite intrusions common elsewhere along the Himalaya. To the east, however, along strike within the upper levels of the GHS is the 4–5-km-thick Manaslu leucogranite, one of the larger such intrusions along the Himalaya. The Chako dome (Searle and Godin, 2003; Godin et al., 2006b; Gleeson and Godin, 2006) is an updomed part of the GHS, exposing high-grade marbles, calc-silicates, and pelites that have been intruded by numerous leucogranite dykes and sills, almost certainly emanating from the Manaslu intrusion to the east. The Phu Detachment cuts all the metamorphic rocks in the Chako dome and is interpreted as also cutting the top of the Manaslu leucogranite (Fig. 10). Guillot et al. (1993) interpreted the upper contact of the Manaslu leucogranite as intruding and cross-cutting the north vergent Naike fold in the sedimentary rocks above. The Naike fold has north-vergent geometry (Guillot et al., 1993) very similar to the Nilgiri fold in the Kali Gandaki, and the contact with the leucogranite is here interpreted as a later low-angle normal fault and ductile shear zone.

The lower levels of the GHS along the Kali Gandaki (Fig. 8) are composed of thick quartzite units (Dana, Gandrung quartzites), pelites of greenschist facies grade (Kuncha Formation), and the Proterozoic Ulleri augen gneiss. The earlier position of the MCT as mapped by Colchen et al. (1980, 1986) was along the base of the kyanite-bearing pelites above Dana. Searle and Godin (2003), Searle et al. (2008), and Larson and Godin (2009) all mapped the MCT at a lower structural position along a prominent zone of mylonites beneath the greenschist facies rocks. Larson and Godin (2009) confirmed this lower position, previously mapped as the Ramgarh thrust, by quartz c-axis fabric data. Lesser Himalayan rocks beneath this zone are unmetamorphosed mainly Proterozoic sedimentary rocks.

Spectacular three-dimensional (3-D) outcrops of the Annapurna Detachment are exposed along the Kali Gandaki gorge and along the south and east faces of Dhaulagiri (Fig. 11A). The AD along the base of Dhaulagiri is a major zone of ductile shearing. The gently north-dipping Cambrian–Ordovician limestones on Dhaulagiri appear to have slipped northwards during southward extrusion of the footwall metamorphic rocks, aided by low frictional stresses of the calc-mylonite glide horizon. North-vergent folds are also beautifully exposed in the Tukuche–Dhampus range to the north of Dhaulagiri (Fig. 11B). A giant kilometer-sized north-vergent fold is also spectacularly exposed on the west face of Nilgiri peak (Fig. 12A) with the trace of the axial surface curving around the summit of Nilgiri South (Fig. 2). This isoclinal backfold has a north-vergent backthrust along its lower limb, exposed in the cliffs above Marpha village (Fig. 12B). Fold axial planes are steep at high structural levels (Fig. 12C) and curve into alignment with the gentle north-dipping Annapurna Detachment, a low-angle normal fault that has been correlated with the STD. The fold closures show tight accommodation folding in the cores of the folds (Fig. 12D).

Three models have been proposed to explain the north-vergent backfolds above the STD in Nepal (Godin et al. (1999a, 1999b; Godin, 2003; Kellett and Godin, 2009). The first (Model 1) relates the folds to gravity-induced “collapse” folding above the Annapurna Detachment when folds were formed during uplift and extrusion of the footwall GHS (e.g., Caby et al., 1983; Burchfiel et al., 1992; Hodges et al., 1996). The second (Model 2) proposes that the north-vergent folds were formed during an early contractional deformation, predating initiation of the STD, and were later cut and progressively transposed by the high-strain zone associated with the Annapurna Detachment (Brown and Nazarchuk, 1993; Godin et al. (1999a, 1999b; Searle and Godin, 2003). The third (Model 3) proposes that the backfolds were drag folds formed as a result of southward translation and associated ductile shearing of the underlying GHS (Searle et al., 1997b; Carosi et al., 2007; Kellett and Godin, 2009). This model is similar to the passive roof fault model for the STD as proposed by Searle et al. (2003, 2006, 2007) and Law et al. (2004) in which the STD acts as a stretching fault (Means, 1990). In this model hanging-wall rocks remain fixed during STD motion while footwall rocks are actively extruding from beneath. However, folds in the hanging wall become progressively reoriented with axial planes dragged into alignment, with increased south-directed footwall shearing and extrusion during channel flow.

In the western Himalaya, similar but smaller scale backfolds have been mapped above the Karsha Detachment part of the STD system in Ladakh (Searle et al., 1997b). The Karsha Detachment dips at ∼20° north and places Permian and younger sedimentary rocks on top of older Palaeozoic and Permian Panjal volcanics. Fold axes are progressively rotated from north vergent to upright to south vergent toward the north above the Karsha Detachment (Searle et al., 1997b; Fig. 5) in a manner very similar to that seen along the Kali Gandaki. The north-vergent backfolds along the Kali Gandaki profile clearly show progressive steepening of axial planes toward the north, similar to the mapped pattern in the Hidden Valley north of Dhaulagiri (Kellett and Godin, 2009). At Nilgiri (Figs. 5 and 12A), fold axial planes are horizontal and show gradually steeper dips to the north. On Tukche and Dhampus peaks (Figs. 6 and 11B), axial planes are very clearly curved into alignment with the Annapurna Detachment toward lower structural levels. At Jomoson and Kagbeni in the north (Figs. 2 and 11B), folds have more upright axial planes. The geometry of the folds and their curved axial planes could indicate that some sort of backsliding may have occurred during extrusion and uplift of footwall GHS rocks beneath the Annapurna Detachment, but the drag fold model (Model 3) appears to be the most likely mechanism to explain the geometry of these folds (Kellett and Godin, 2009).

Structural Evolution

Godin et al. (1999b, 2001) and Godin (2003) suggested that the Tethyan sedimentary sequence had been affected by three phases of folding (D1, D2, and D4) with contrasting vergence, interspersed with two major extensional events (D3 and D5). Evidence for the early south-vergent folds (D1) comes solely from the “Fang nappe” as mapped by Colchen et al. (1980). However, down-plunge projection of this tight to recumbent fold exposed on Fang peak (7647 m) at high elevations projects westward directly into the Marpha fold along the base of the Nilgiri backfold in the Kali Gandaki to the west, and eastward directly into the Machhapuchare syncline in the Modi Khola (Hodges et al., 1996; Fig. 9A). It is suggested that these structures are all refolded folds belonging to the later phase of low-angle normal faulting. D2 of Godin et al. (1999a, 1999b) and Godin (2003) are the NE-verging backfolds above the STD. These authors proposed that these folds were part of an Oligocene NE-verging fold phase that has no correlative anywhere else along the Himalaya, and were later cut by the low-angle normal fault, the Annapurna Detachment. However, fold axial planes curve into alignment with the Annapurna Detachment and the mylonites in the ductile shear zone immediately beneath, and are not cut by the fault, indicating instead that the folds were refolded during later southward extrusion of the footwall gneisses of the GHS (Kellett and Godin, 2009).

The structural scenario preferred here, following Searle et al. (1997b) and Kellett and Godin (2009), involves a continuum of crustal shortening during the Early Eocene–Oligocene that resulted in folding of the upper crust across the Tethyan Himalaya (Fig. 13). Early folds were then refolded during Miocene extrusion of the GHS (Channel Flow) when active thrusts and ductile shear zones along the base and active normal faults and ductile shear zones along the top of the GHS channel were moving synchronously. Thus fold axial planes curve into alignment with the Annapurna Detachment and steepen toward the north. The youngest deformation phase in the Tethyan Himalaya involved establishment of north-south aligned steep normal faults associated with relatively minor east-west extension of the upper crust and formation of the Thakkhola graben.

Mustang–Thakkhola Graben Faults

The Mustang–Thakkhola graben is the largest of a series of at least 10 regularly spaced north-south (or NNE-SSW) aligned, east-west extending graben systems across southern Tibet (Armijo et al., 1986). Normal faults bounding these graben strike at right angles to the STD low-angle normal faults and underlying ductile shear zone. North-south aligned steep normal faults cut through the rocks of the Tethyan sedimentary zone but do not cut GHS rocks to the south across the STD. Up to 1000 m thickness of fluvial and lacustrine deposits are preserved in the Mustang–Thakkhola graben (Hagen, 1968; Fort et al., 1980; Colchen, 1999). The oldest basin fill sediments are 11–9.6 Ma, and southward axial drainage was established with the Thakkhola Formation from 7 Ma (Garzione et al., 2000, 2003). Spectacular perched terraces along the upper Kali Gandaki and in Mustang demonstrate multiple episodes of basin damming, filling, and then breaching during the Neogene uplift of the Himalaya to the south. The Kali Gandaki River is clearly an antecedent river formed by southward flow off the Tibetan plateau prior to uplift of the Greater Himalaya. The river presently cuts a gorge up to 6800 m deep in between the high peaks of Dhaulagiri (8167 m) to the west and Nilgiri (7063 m) and Annapurna (8091 m) to the east.

The Mustang–Thakkhola graben is asymmetric with one major fault, the Dangardzong fault, striking N20°E and dipping ∼60°ESE, along the western margin and a series of inward-dipping minor faults along the eastern margin. Hurtado et al. (2001) presented 14C ages for river terraces that suggested a 17.2 ka minimum age for latest movement along the Dangardzong fault. Hurtado et al. (2001) suggested that the Dangardzong fault cuts and offsets the Annapurna Detachment (as mapped by Brown and Nazarchuk, 1993) but is cut by an even younger, active strand of the STD, the Dhumpu Detachment. Mapping during this study, however, suggests that the Dangardzong fault cannot be traced south of Marpha, and does not cut the Annapurna Detachment. No trace of the young Dhumphu Detachment of Hurtado et al. (2001) was found and it seems unlikely that any strand of the STD low-angle normal faults is presently active. In the Marysandi and Nar valleys, the STD has been folded about east-west aligned large buckle folds (Godin et al., 2006b), again suggesting that the fault has not been active recently.

Origin of the Backfolds

A major phase of crustal shortening and folding along the north Indian plate margin occurred during the Eocene–Oligocene following India–Asia collision and prior to initiation of the South Tibetan Detachment (Godin et al., 2001; Godin, 2003). In the Annapurna–Nilgiri–Dhaulagiri Range, north-vergent backfold axial planes are progressively rotated into alignment with the STD along lower structural levels. Backfolds were enhanced by the later southward extrusion of GHS gneisses in the footwall of the STD during the Miocene (Channel flow). Upright fold axial planes were then progressively dragged into alignment with the southward flow of footwall rocks (Fig. 13).

Projection of the Annapurna Detachment along the Kali Gandaki section (Fig. 11A) and its restoration suggests that the AD may cut up-stratigraphic section to the north, the opposite of that expected of a low-angle normal fault. However, the folds in the hanging wall were undoubtedly formed prior to motion along the STD faults so the restored fault profile cuts already folded rocks. The earlier restored sections across the Annapurna and Manaslu sections (Searle and Godin, 2003) showed the STD-related low-angle normal faults cutting through previously backfolded folds, as in the model of Godin et al. (1999a, 1999b). However, if the backfolds were largely formed later during southward flow of the middle crust (Channel flow) and subsequent northward tilting, as proposed by Kellett and Godin (2009) and here, a more likely restoration might be similar to that shown in Figure 13.

Four main observations suggest that the huge Nilgiri–Tukche backfolds (Fig. 12A) immediately above the Annapurna Detachment were affected by some degree of “collapse.” First, fold axial planes are not cut by the STD but instead curve into alignment with the high strain zone. Second, fold axial planes progressively steepen to the north from flat-lying at Nilgiri in the south to subvertical at Kagbeni in the north. Third, the backfold style is more akin to gravity-driven collapse style folds, not to a foreland fold-thrust belt style. Fourth, everywhere else along the Himalaya Eocene–Oligocene folds are south-vergent to upright. There seems to be no reason why north-vergent folds were present at this time only immediately above the STD in the Dhaulagiri–Annapurna ranges. However, no lowering of surface elevation or “collapse” need be inferred. North-vergent folding in the Nilgiri range probably contributed to thickening locally as a result of southward extrusion of the GHS footwall and northward tilting of the entire mid-crustal slab or channel.

Geometry of the STD Faults and Ductile Shear Zones

Along the Modi Khola at least four north-dipping normal faults have been mapped, each putting low-grade rocks above high-grade rocks (from the base the Chame, Deurali, Machhapuchare, and Hiunchuli Detachments; Hodges et al., 1996). These faults progressively young to the north and are contemporaneous with similar splays of thrust faults along the Main Central Thrust system in the south. Age constraints suggest that deformation progressed with time structurally downward along the MCT zone and structurally upward along the STD normal faults. The STD normal faults and MCT thrust faults together with their respective ductile shear zones do not represent rapid alternations between shortening and extension (“compensation faults”; Hodges et al., 1996). Rather, normal and thrust faults were active at the same time, facilitating southward extrusion of the GHS mid-crustal channel (Searle and Rex, 1989; Grujic et al., 2002; Searle and Szulc, 2005; Searle et al., 2003, 2006, 2008; Jessup et al., 2006).

In the Kali Gandaki Gorge and Dhaulagiri Range to the west, these STD normal faults merge into one very large ductile shear zone up to 1500 m wide (Godin et al., 1999b), capped along the top of the calc-mylonite zone by a brittle low-angle normal fault, here interpreted as the position of the Annapurna Detachment. The AD was originally mapped as separating GHS unit 3 augen gneisses below from Tethyan sedimentary rocks above (Brown and Nazarchuk, 1993; Godin et al. (1999a, 1999b). However, the rocks above this horizon are upper amphibolite facies tremolite + diopside ± hornblende bearing calc-silicates and marbles intruded by tourmaline-bearing leucogranite dykes that are interpreted here as the upper part of the GHS. Rocks within the shear zone grade from highly sheared high-grade marbles below up to unmetamorphosed Cambrian–Ordovician limestones-dolomites forming an easy glide horizon for the north-vergent backfolds above. Later low-angle brittle faults cut the upper part of the ductile shear zone. Godin et al. (2006a) defined two low-angle normal faults separating this shear zone. Major jumps in P-T conditions across the fault show that exhumation of footwall gneisses was accomplished by footwall extrusion bringing the high-grade, partially molten gneisses toward the surface beneath a passive roof fault. The AD forms, in effect, the uplifted and exhumed Miocene brittle-ductile transition.

STD Initiated at Low-Angle as a Passive Roof Fault

The STD system of low-angle normal faults effectively juxtaposes three different rocks that were previously at different structural depths (Fig. 14). The original depth of burial of the deepest GHS kyanite-grade gneisses (10–12 kbars equivalent to ∼35–38 km) beneath the STD prior to movement along the STD is shown (green box on Fig. 15A), assuming the same angle of dip of the STD normal faults. Pressure estimates of the Larjung unit V clinopyroxene–tremolite– K-feldspar marbles are ∼9 kbar equivalent to ∼28 km depth of burial (yellow square on Fig. 15A) in the restored section. The depth of burial of the Cambrian limestones immediately above the STD at the base of Dhaulagiri is constrained at ∼2.8–3.4 kbar, the burial depth of the 10–12 km thick Cambrian–Cretaceous sedimentary overburden (black circle on Fig. 15A). During exhumation of the GHS by southward ductile flow, the kyanite gneisses were initially exhumed by simple shear up to 9 kbar where they were welded to the base of the Larjung unit marbles. The two units were then exhumed further along the STD (Annapurna Detachment) footwall such that both units moved up to immediately beneath the Cambrian limestones at 10–12 km depths (Fig. 15B). By ∼18 Ma all three rocks were adjacent to one another across the Annapurna Detachment ductile shear zone. Assuming the same angle of dip of the STD as measured today, the relative offset and exhumation of the footwall rocks must have been a minimum of 80–100 km.

The low-angle normal faults could not have been originally at a steeper angle because the crust along the hanging wall would have to have been much thicker, up to 80–100 km given the offsets shown on Figure 15. During the exhumation of the GHS footwall rocks beneath the Annapurna Detachment in the Early Miocene (∼22–18 Ma) the hanging wall remained fixed while the footwall was moving rapidly up to the Earths’ surface. Ar-Ar ages show rapid cooling during this period across the GHS (Searle and Godin, 2003; Godin et al., 2006a). The geometry of the STD corresponds to that of a passive roof fault or stretching fault. The STD low-angle normal faults were operating within a wholly compression environment and show no net extension at all.

Channel Flow and the STD

The Himalayan Channel Flow model proposes that gravitational potential difference between the double thickness crust of Tibet and the normal thickness crust of the underthrusting Indian plate resulted in the southward directed crustal flow of a mid-crust layer bounded by low-angle ductile shear zones associated with the MCT along the base and the STD along the top (Beaumont et al., 2001, 2006; Grujic et al., 2002; Searle and Szulc, 2005; Searle et al., 2003, 2006, 2007; Jessup et al., 2006; Godin et al., 2006a; Grujic, 2006). This model was based on field, structural, thermobarometric profiles and age constraints from numerous profiles across the GHS together with geophysical data from the INDEPTH project in Tibet (Nelson et al., 1996). Two end-member types of crustal flow are the Couette flow model and the Poiseuille (“pipe-flow”) flow model. With Couette flow, the velocity distribution due to the pressure gradient shows that the boundaries of the channel are actively moving (Godin et al., 2006a). These profiles would show a jump in P-T conditions across the bounding ductile shear zones below (MCT) and above (STD). In the Poiseuille flow model, maximum velocities in the channel are in the center with the margins of the channel, both MCT below and the STD above, not moving. This model is therefore likely to show a pure shear flattening of isograds rather than a jump in P-T conditions across the bounding ductile shear zones. This latter situation might be relevant in STD sections that do not show clear brittle faults along the top such as that in the Dzakaa Chu section, north of Everest (Cottle et al., 2007) and in the southern klippen of the Bhutan Himalaya (Long and McQuarrie, 2010).

Low-angle normal faults and ductile shear zones with top-to-north (or better described as base-to-south) kinematic indicators underlie the 10–12-km-thick Cambrian to Cretaceous sedimentary sequence along the northern Himalaya. In the Annapurna–Dhaulagiri ranges of Nepal at least four low-angle normal faults cut the Cambrian–Ordovician limestones and marbles. These faults merge to the west into one thick ductile shear zone, the Annapurna Detachment with calc-mylonites separating relatively unmetamorphosed limestones affected only by burial metamorphism (∼2–2.8 kbar maximum) above from kyanite-grade gneisses that were buried to depths of more than 35 km (10–12 kbars pressure) below. Assuming the angle of dip of the AD remained the same the geological offsets along the detachment are around 80–100 km, similar to the STD in the Everest region (Searle, 2003; Searle et al., 2003, 2006; Jessup et al., 2006, 2008; Cottle et al., 2007).

Folds in the hanging wall of the AD were formed prior to movement along the low-angle normal faults, but spectacular north-vergent backfolds in the Nilgiri range were enhanced by backsliding during southward extrusion of the GHS ductile deforming rocks. Fold axial planes are curved to the south into alignment with the low-angle normal fault along the top of the ductile shear zone in the AD. Low-angle normal faults in the Himalaya do not indicate “orogenic collapse,” decreasing crustal thickness, or lowering of surface elevation because there was a continual influx of Indian plate material underthrust beneath the Himalaya from the south which had the effect of continually jacking-up the Himalaya, a process that continues to this day along the active MBT–Main Frontal thrust system of faults. The passive roof fault model for the STD, combined with the Channel Flow model, shows how low-angle normal faults can operate in regions of crustal convergence, shortening and thickening, and require no crustal or lithospheric extension at all.

My thanks to Laurent Godin, Kyle Larson, Rick Law, and Kip Hodges for stimulating discussions in the field, to Laurent Godin and Rick Law for detailed comments on an earlier draft of the paper, to Mike Murphy and an anonymous person for reviews, and to Suka Ghale and the team from Gorkha for trekking logistics. This work was initially funded by NERC grant NER/KS2000/00951 and later by NSF grant EAR 0711207 to R.D. Law and M.P. Searle.