We use structural, petrographic, and geochronological data to examine processes of exhumation of partially molten crust in the Late Devonian–early Carboniferous Chandman dome in the Mongolian tract of the Central Asian Orogenic Belt. The dome is composed of a magmatite-migmatite core and a low-grade metamorphic envelope of early Paleozoic metasediments and Carboniferous clastics. Its tectonic evolution can be divided into three main stages. The oldest fabric is a subhorizontal foliation, S1, in migmatites that is subparallel to the magmatic foliation in granitoids and to the greenschist facies schistosity in the enveloping metasediments. This event is interpreted as a result of horizontal deep crustal flow at depths of 20–25 km. The S1 layering was subsequently transposed into a new foliation, S2, or affected by open to close upright F2 folds that are locally truncated by steep walls of diatexites, suggesting influx of partially molten crust into fold cores. The shallow-dipping magmatic foliation in granitoids is locally reworked by vertical magmatic to gneissic S2 fabrics. Syn-S2 metamorphic assemblages and synkinematic to postkinematic cordierite point to exhumation of the migmatites and granitoids from 20–25 km to ∼10 km, and concomitant isobaric heating of the surrounding upper crust. New 40Ar/39Ar ages of 350–340 Ma from both the high-grade core and the metamorphic mantle overlap with previously published crystallization ages of 360–340 Ma, suggesting that magmatism and cooling in the upper crust are partly synchronous. Late syn-D2, S2-parallel leucogranite sheets crosscutting both the magmatic core and the mantling migmatites either exploit S2 or crosscut horizontal S1 fabrics; they are interpreted as brittle expulsion of magma during ongoing syn-D2 exhumation. We suggest that the partially molten crust and magmas rose vertically into the upper crust, along steep planar fabrics that are parallel to the axial fold plane of a crustal detachment folding, without contribution of buoyancy forces. In order to test that crustal-scale detachment folding can exhume partially molten crust, we apply an analogue model with temperature dependent rheology of the lower crust represented by a partially molten wax layer overlain by an upper crustal sand layer. It is shown that the fold core initially filled by low-viscosity partially molten wax rapidly migrates upward during fold lock-up, enhancing upward extrusion of magma and migmatites along the fold axial plane. The exhumation of the lower crust wax is facilitated by erosional unroofing of the upper crustal sand above the hinge of the antiform. In Chandman, localized siliciclastic lower Carboniferous basins rimming the dome attest to this erosional phenomenon. Using a simple geometrical analysis we show that detachment folding can explain magma collection in an orientation perpendicular to the main shortening direction, and episodic emplacement of magmas during amplification of the antiform. In our view, the detachment folding model provides a new model for the exhumation of a weak zone above a rigid floor (basement from which the fold is detached) and its vertical extrusion related to locking of the fold and post-buckle flattening. This model helps elucidate steep retrograde pressure-temperature-time paths along steep fabrics, overlapping ages from different geochronometers, and emplacement of voluminous syntectonic magmas.

Phanerozoic accretionary orogens are characterized by a large heterogeneity of composition, structure, and metamorphic grade compared to collisional orogens (e.g., Cawood et al., 2009). A common feature of these orogens is the lateral segmentation of middle crust into zones as wide as several kilometers of vertically foliated low-grade rocks alternating with regions of horizontally foliated granitoids and migmatites. This structural pattern is well developed in the archetypal accretionary system, the Paleozoic Terra Australis orogen (Cawood, 2005; Gray and Foster, 2004), where low-grade regions alternate with zones of high-temperature–low-pressure (HT-LP) metamorphic belts and granitoids marked by peak temperatures of ∼700–800 °C at 4–6 kbar (Flood and Vernon, 1978).

The Altai orogen represents a highly metamorphosed core of the second largest Paleozoic accretionary orogen, the Central Asian Orogenic Belt (CAOB; or Altaids, Fig. 1A; Şengör et al., 1993) that shows characteristic alternation of low- and high-grade zones both in Mongolia (Fig. 1B) and in the adjacent Chinese Altai (Jiang et al., 2015). Here, low-grade late Cambrian to Early Ordovician turbidites and volcanic rocks alternate with zones of granitoids and rocks showing HT-LP metamorphism and migmatization of Silurian to Devonian age (ca. 450–350 Ma; Burenjargal et al., 2014; Sun et al., 2008). Studies (Broussolle et al., 2015; Jiang et al., 2015) have shown that the tectonic evolution of these migmatitic and granitic complexes is related to their vertical transfers to supracrustal levels and formation of narrow magmatite-migmatite domes. This deformation style may characterize hot and remobilized accretionary complexes where partially molten lower crust is juxtaposed against overlying unmetamorphosed or low-grade supracrustal rocks in the form of migmatite-granite domes (Broussolle et al., 2015; Jiang et al., 2016).

The mechanical and thermal roles of magma in orogens have been extensively studied (e.g., Brown and Solar, 1998; Reichardt and Weinberg, 2012). Hollister and Crawford (1986) suggested that syntectonic magmas and anatectic melts weaken the crust and concentrate deformation, and may therefore act as melt-lubricated shear zones facilitating, for example, thrusting of deep crust over upper crustal rocks. Somewhat counterintuitively, in convergent systems these magmatic bodies typically form tabular subvertical magmatic sheets oriented at a high angle to the direction of active horizontal contraction (Ingram and Hutton, 1994; Lehmann et al., 2013). Rosenberg and Handy (2005) showed that even incipient melting (<7% melt) could lead to crustal-scale localization of deformation. Thompson et al. (1997a, 1997b) explored theoretically the implications in terms of heat and material extrusion of deep-crustal rocks along a rheologically weak vertical tabular zone that is laterally bounded by two rigid and converging indenters. Nevertheless, these studies do not explain both the crustal architecture of narrow migmatite-granite zones juxtaposed to almost unmetamorphosed sediments and the relative contribution of the mechanical behavior of magma, lateral shortening, and gravity on the exhumation of hot lower crust in remobilized accretionary complexes such as the Altai orogen in Mongolia and north China (Jiang et al., 2016). In theory, low-viscosity, partially molten rocks can rise either due to diapirism (gravitational upwelling), detachment folding (a form of tectonically driven extrusion), or diapiric detachment (a combination of the two), as discussed extensively by Soula (1982) and Soula et al. (1986), synthetized by Whitney et al. (2004), and quantified by Burg et al. (2004); however, in detail these mechanisms in accretionary systems remain unexplored. This calls for a need to study and understand the possible magma-assisted exhumation mechanisms accounting for the crustal architecture of the Altai orogenic core in particular, and partially molten metasedimentary accretionary complexes in general.

To this end, we have studied the Chandman massif, located in the southwestern Mongolian part of the CAOB (Economos et al., 2008). This massif exhibits a lateral association of Devonian–Carboniferous granitoids and gneisses in the core, mantled by mid-crustal migmatites and peripheral medium- to low-grade metasedimentary sequences. Based on detailed field observations, pressure-temperature (P-T) data, and 40Ar/39Ar geochronology, we suggest a conceptual model of the role of magma in formation of gneiss domes and accretionary systems. Then, using analogue wax modeling, we propose a new model of magma-assisted localized exhumation of lower crust in core of a crustal-scale detachment fold together with the progressive influx of magma into the core of a growing antiform, followed by a vertical extrusion stage linked to massive post-buckle flattening of a weak core of folded domain. Using a simple geometrical analysis we show that detachment folding can explain magma collection in a perpendicular orientation with respect to the main shortening direction as well as episodic emplacement of magmas during amplification of the antiform.

The evolution of the CAOB lasted from ca. 600 Ma to ca. 270 Ma and was dominated by accretion of magmatic arcs, backarcs, accretionary complexes, and Mesoproterozoic continental blocks (Fig. 1A; see reviews of Wilhem et al., 2012; Xiao et al., 2015). Two major geological domains form the eastern tract of the CAOB in Mongolia, separated by the Main Mongolian Lineament: the Caledonian and Hercynian domains (Zonenshain, 1967). The northern Caledonian domain is formed by a Mesoproterozoic continental basement rimmed by a belt of accreted Neoproterozoic–early Paleozoic ophiolites (and eclogites), oceanic basins, and magmatic arcs (Dijkstra et al., 2006; Jian et al., 2014; Štípská et al., 2010). The southern Hercynian domain is essentially built up of late Cambrian sedimentary and volcanic accretionary material: the so-called Altai magmatic arc-accretionary complex in the adjacent Chinese Altai region (Xiao et al., 2009). It is covered by Silurian–Devonian passive margin sequences (Ruzhentsev et al., 1985) and composed of oceanic magmatic arc and backarc rocks to the south (Demoux et al., 2009; Lamb and Badarch, 2001), all deformed during Late Devonian to Carboniferous.

The Altai magmatic arc-accretionary complex extends from an area west of the Siberian craton in Russia to northeast China and southern Mongolia, where it is traditionally known as the Mongolian or Gobi-Altai zone (Fig. 1B). Some suggest that an early Neoproterozoic continental crust constitutes the basement of the Altai magmatic arc-accretionary complex (Wang et al., 2009); others deny the existence of a Precambrian basement (Sun et al., 2008), and suggest that the high-grade rocks are metamorphosed equivalents of sediments within a major early Paleozoic accretionary complex (Jiang et al., 2012). The Mongolian and Chinese tract of this complex zone was affected by Devonian–Carboniferous arc-type magmatism, migmatization, and deformation (Cai et al., 2011; Kröner et al., 2010; Yuan et al., 2007), related to prolonged oceanic subduction. Hanžl et al. (2016) and Jiang et al. (2016) suggested that this accretionary event produced a horizontally stratified continental crust made up of accretionary sediments and interlayered Devonian granitoids. Lehmann et al. (2010), Zhang et al. (2015), and Jiang et al. (2015) showed that the geology of the Altai Mountains in Mongolia and China is characterized by an alternation of migmatite-granite zones with low-grade domains consisting of Paleozoic sediments (Fig. 1B). These studies also confirm that the migmatites and foliated granitoids were exhumed in cores of narrow domal structures during the Devonian. Forward gravity modeling and the geochemical study of Guy et al. (2015) suggested that the partially molten lower crust was underthrust by a south-dipping block made of the Cambrian and Proterozoic basement of the Lake zone (now crops out in the north; Fig. 1B) before the late Carboniferous. This led Guy et al. (2015, fig. 11 therein) to propose a model of a crustal wedge underthrust by a rigid basement promontory during the Late Devonian–Carboniferous convergence. The entire Altai accretionary system and the adjacent domains of Mongolia (Guy et al., 2014) and Beishan (Tian et al., 2013) were subsequently overprinted by regional-scale steep, close to isoclinal, folding (e.g., Lehmann et al., 2010; Guy et al., 2014; Xiao et al., 2015). This event is related to late Paleozoic to Mesozoic oroclinal closure of an originally north-south–trending ribbon continent (the Tuva-Mongol orocline; Fig. 1A) and massive shortening of the Mongolian CAOB between the Siberia and north China cratons (Edel et al., 2014; Xiao et al., 2015).

In Mongolia, the Altai magmatic arc-accretionary complex mainly consists of Cambrian–Ordovician quartz-chlorite schists, siliceous thinly laminated rocks, strongly altered metabasalts, epidote-actinolite schists, and massive metadiabase porphyries (the so-called Tugrug formation; Kröner et al., 2010). There is an overlying thick sequence of siliciclastic rocks with subordinate marbles of Ordovician to presumably Silurian age. The overlying Devonian rocks consist of clastic sediments and coral-rich limestones, accompanied by various volcanic rocks. The youngest unit is an early Carboniferous flysch, forming intermontane basins (Markova, 1975). All primary stratigraphic contacts between these formations were reactivated in the Chandman area (Rauzer et al., 1987). The Paleozoic units are intruded by numerous calc-alkaline Late Devonian to early Carboniferous granitoids, coeval with high-grade metamorphism (Burenjargal et al., 2014; Hrdličková et al., 2008), and alkaline Permian plutons.

Similar to other accretionary orogens, the Mongolian Altai shows clear alternations of deep-crustal, partially molten rocks with low-grade to unmetamorphosed units (map of Fig. 1B). In the studied section, the high-grade rocks form two large-scale domal structures forming the northern and southern parts of the Mongolian Altai (cross section in Fig. 1B). The structurally deepest Cambrian–Ordovician rocks overlain by Devonian platform carbonates in the center form a low-grade synform that is locally 30–40 km wide. The Carboniferous siliciclastic basin overlies Devonian rocks along the southern edge of northern high-grade domain, the Chandman massif.

The Chandman massif is a 50-km-long and as much as 10-km-wide west-northwest–east-southeast–trending structure composed of igneous rocks, mantled by migmatites that form the magmatite-migmatite core, and peripheral medium- to low-grade metasedimentary rocks that form a metamorphic envelope (Fig. 2; Lehmann et al., 2010). The rocks ∼2.5 km away from the igneous core are of lower greenschist metamorphic grade. The northern boundary of the Chandman massif is the major Cenozoic left-lateral Bogd thrust fault (Fig. 1B; Cunningham, 2010).

The magmatic rocks of the magmatite-migmatite core (Fig. 2) are mostly porphyritic granodiorites and granites, with minor leucogranite sheets and scarce diorites (Hanžl and Aichler, 2007). Most of the granitoids are metaluminous with minor peraluminous varieties (Economos et al., 2008). Common xenoliths are meter to locally kilometer in size and consist of amphibolite, migmatized paragneiss and mica schist, rare quartzite, and calc-silicate lenses (Hanžl and Aichler, 2007). Migmatites range from predominant metatexites to diatexites. Along the eastern part, numerous meter thick concordant to discordant sheets of medium-grained muscovite-bearing leucogranites penetrate the magmatic rocks and migmatites (Fig. 2A). The metamorphic envelope consists of nonmigmatized medium- to low-grade metasedimentary rocks of Cambrian–Ordovician age (the Tugrug Formation) and Devonian to early Carboniferous very low grade rocks (Fig. 2).

Reported crystallization ages of granitoids (map of Fig. 1B) are 350.4 ± 1.7 Ma for a strongly foliated granite gneiss, 340.9 ± 2.5 Ma for a weakly deformed granite (both SHRIMP, sensitive high-resolution ion microprobe, U-Pb zircon; Kröner et al., 2010), and 345 ± 2 Ma for a granodiorite (laser ablation–inductively coupled plasma–mass spectrometry U-Pb zircon; Hrdličková et al., 2008). Migmatization was dated as 356 ± 1 Ma and subsolidus deformation reveals an age of 347 ± 4 Ma (both obtained using laser ablation−split stream, LASS, U-Pb monazite; Broussolle et al., 2015). Available geochronological data point to migmatization slightly predating plutonism and synchronous protracted magmatism and deformation.

Three ductile deformational events developed across all the rock types characterize the tectonometamorphic evolution of the Chandman area (Fig. 2; Broussolle et al., 2015; Lehmann et al., 2010). D1 is developed as a bedding-parallel S1 subhorizontal greenschist facies schistosity in the metamorphic envelope, and as a high-grade gneissosity S1 in the magmatite-migmatite core. The D2, Late Devonian east-west shortening is characterized by upright folding, development of axial planar foliation S2 associated with emplacement of F2-axial planar sills in migmatites, and magmatic to solid-state deformation of granitoids in the core of the Chandman massif. A north-northeast–south-southwest, Permian to Mesozoic D3 shortening strongly affected the entire Chandman massif; it is responsible for the present-day west-northwest–east-southeast–elongated shape of the Chandman massif. Meter-scale, sharp to subangular, open to tight steep-hinge asymmetric folding of the S2 foliations in the magmatite-migmatite core are outcrop expressions of the D3 deformation event (Fig. 3A). F3 long limbs are parallel to a west-northwest–east-southeast–striking, spaced, and disjunctive foliation that is locally developed in the magmatic-migmatite core (Fig. 3B; Lehmann et al., 2010, fig. 9E therein), and to sporadic spaced, slaty cleavage S3 in the metamorphic aureole (Fig. 3C). Lehmann et al. (2010) interpreted the north-south–striking, steep F3 short limbs to reflect the orientation of S2 structures prior to the D3 overprint. At the scale of the Chandman massif, the attitudes of the S2 trajectories (Fig. 2B) allow the subdivision of the Chandman massif in a central domain of broadly north-south–striking S2 foliations, surrounded by zones where S2 is predominantly west-northwest–east-southeast striking: these are the core and limbs of a regional-scale F3 fold, respectively. Consistently, back rotation of the eastern and western limbs of the Chandman regional-scale F3 fold to their pre-D3 position shows that the Chandman massif was originally an elongated dome-like structure in a broadly north-south direction with an across-strike zonation made of a metamorphic envelope formed by peripheral low- to medium-grade metasedimentary rocks. These rocks pass into mantling migmatites along both margins and to magmatic rocks forming the core of the massif. Based on heterogeneous deformation of the Chandman massif, we describe three informal domains represented by the hinge and the western and eastern limbs of the regional F3 fold separately (Fig. 2C).

In order to explain the thermal structure and exhumation mechanism of the magmatite-migmatite core of the Chandman massif, we have carried out a detailed structural study of the migmatites and granitoids and compared it with the structural record of the upper crustal units of the metamorphic envelope. The eastern limb of the F3 regional fold best preserves the original relationships between the metamorphic envelope and the high-grade core rocks (Fig. 4A). Cenozoic faults reactivated the contacts between the magmatite-migmatite core and the metamorphic envelope in many places of the western F3 limb. The structural successions in both the metamorphic envelope and the high-grade core are described separately, with the goal of resolving the interactions between regional deformation and flow of partially molten rocks and magmas. We first report on the metamorphic evolution of the high-grade core, mantling migmatites and the metamorphic envelope. A compilation of P-T data from Broussolle et al. (2015) is combined with our own observations of occurrences of metamorphic minerals and associated isograds in metasedimentary rocks (Figs. 4 and 5). We then describe the structures in both the metamorphic envelope and the high-grade core from the envelope toward the core.

Metamorphic Evolution

The syn-D2 metamorphic isograds are parallel to the contact between the metamorphic envelope and the magmatite-migmatite core and range from andalusite in the peripheral low-grade schists to sillimanite in the mantling migmatites (Fig. 4A). The granitoids locally contain syn-D2 andalusite- and garnet-bearing xenoliths as well as ubiquitous postkinematic cordierite (Fig. 4A). The distribution of isograds indicates an inward-directed increase in pressure and temperature in the schists of the metamorphic envelope that is also compatible with systematic grain coarsening.

Metasedimentary rocks of the metamorphic envelope are characterized by the growth of syn-S2 andalusite (Fig. 5A) and late-S2 to post-S2 cordierite in the western limb of the F3 fold and the growth of syn-S1–S2 garnet followed by syn-S2 growth of cordierite in the eastern limb (Fig. 5B) (Broussolle et al., 2015). The P-T diagrams (Fig. 4B) show that the metamorphic envelope of the northeast part of the Chandman massif recorded pre-D2 to syn-D2 nearly isobaric heating from 500 to 530 °C to 530–570 °C at 2.5–3 kbar, whereas its southern part recorded syn-D2 heating reaching 540–620 °C at 3–4 kbar (Broussolle et al., 2015).

The migmatites are characterized by a syn-S1 assemblage of garnet, sillimanite, and ilmenite (Fig. 5C) and a syn-S2 assemblage of garnet, sillimanite, biotite, plagioclase, and K-feldspar (Fig. 5D). Post-tectonic cordierite, muscovite, and chlorite overgrew the S2 assemblage. The simplified P-T diagram (Fig. 4B) shows that the D1 deformation in the sillimanite stability field at 6–7 kbar and 700–750 °C is followed by D2 decompression to the cordierite stability field at 3–4 kbar and 680–750 °C (Broussolle et al., 2015). These conditions are consistent with P-T conditions of 2.9–3.7 kbar and 725–775 °C of magma crystallization from a granodiorite sample (Al-in-hornblende barometry, hornblende and plagioclase thermometry; Economos et al., 2008). The data indicate that the migmatites were exhumed from deep structural levels during D2 deformation while the metamorphic envelope recorded only isobaric pre-D2 to syn-D2 deformation (Fig. 4B).

Deformation in the Low-Grade Rocks of the Metamorphic Envelope

Although being both elongated in the typical west-northwest–east-southeast–trending direction (map in Fig. 1B), the Ordovician to Carboniferous metasedimentary sequence shows a different structural record than the rocks of the Cambrian–Ordovician (Tugrug) metasedimentary formation. F3 upright buckle folds affect the bedding of the Ordovician to Carboniferous metasedimentary rocks, and are associated with a pervasive west-northwest–east-southeast–striking steep S3 slaty cleavage. The D2 overprint is weak in these rocks, and is only depicted by locally steeply plunging S0-S3 intersection lineations (Lehmann et al., 2010, fig. 4 therein) and rare east- or west-dipping bedding planes (Lehmann et al., 2010, fig. 9D therein). In contrast, the Cambrian–Ordovician schists display mostly D1 and D2 structures and only scarcely and heterogeneously developed spaced S3 cleavage (Figs. 2A, 2B). F2 upright and symmetric folds are open to close and parallel with mineral, aggregate, or corrugation lineation L2 (Fig. 2C). Locally, the S1 layering is entirely transposed by axial planar S2 schistosity (e.g., Figs. 2A, 2B).

Deformation of the Mantling Migmatites

We use the term metatexite for migmatites that show continuous S1 layering at a mesoscale; diatexites are migmatites that do not show continuous S1 layering. The leucocratic parts of the rock are termed leucosome or leucocratic veins or dikes if they crosscut the host-rock layering. We use the proportion of leucosome at the mesoscale as indication of the minimum percentage of former melt that crystallized in the rock. It is a minimum percentage because melt loss is likely to occur in a rock that is undergoing both strain and a prograde P-T path above the solidus (Powell and Downes, 1990; White and Powell, 2002). Leucosomes typically consist of plagioclase (Pl) + quartz (Qz) + potassium feldspar (Kfs) ± hornblende (Hbl) ± biotite (Bt) (abbreviations after Whitney and Evans, 2010) with grain sizes varying from medium to very coarse grained. Leucosomes show sharp boundaries surrounded by fine- to coarse-grained melanocratic selvages.

Low-strained domains devoid of the D3 deformation best preserved the migmatitic structures. The most common are metatexites with preserved subhorizontal or shallow-dipping S1 foliation folded by upright F2 folds or crosscut by steep leucogranite veins. Because the deformation intensity and style are strongly controlled by melt proportion (e.g., Brown and Solar, 1998; Schulmann et al., 2009), we further describe separately the structures from low to high leucosome proportions in migmatites.

At ≤30% leucosome proportion, the migmatites show an interconnected network of leucosomes and leucocratic veinlets parallel or oblique to the S1 foliation (Fig. 6A). Locally, the S1 foliation is affected by F2 folds (Fig. 7A) that vary from open buckle folds for lower leucosome fractions (≤20%) to close and strongly flattened folds for higher leucosome proportions (∼30%). Folds are often disharmonic, with hinges truncated by axial planar S2-parallel subvertical dikes (Fig. 7A). In some outcrops, migmatites exhibit an S2-parallel stromatic foliation, marked by vertical intercalations of millimeter- to decimeter-wide leucosomes and mica-rich bands and relics of isoclinal or intrafolial folds.

Higher leucosome proportions (≥30%) occur in metatexites located tens of meters from the core magmatic rocks. These metatexites form an S1 multilayer sequence deformed by close F2 folds with cuspate hinges that are locally transposed by meter-wide zones of diatexites (Figs. 7B, 7C). These steep axial planar diatexite sheets enclose aligned biotite schlieren and synmigmatitic rootless folds that are parallel to the regional S2 foliation (Fig. 7B). In many outcrops, the S1 leucosomes and S2-parallel leucogranite dikes are in compositional and structural continuity (Fig. 7D). Leucosomes in S1 indicate that segregation along S1-parallel planes occurred prior to the F2 folding and the melt migration along S2-parallel sheets during D2, but before leucosome crystallization (Fig. 7D). The diatexite sheets pass gradually to core magmatic rocks.

In a few places, weakly foliated diatexites surround foliated and steeply oriented tabular rafts of metatexites and refractory amphibolites (Fig. 6B). Here, the internal S1 stromatic foliation in the rafts shows either absence of internal folding or is affected by F2 upright folds. These rafts have internal structure similar to that of the migmatites in areas of lower melt proportions, and therefore are interpreted as fragments of metatexites that have been disaggregated during the D2 deformation event in regions of more advanced anatexis.

Deformation of Granitoids of the High-Grade Core

The granitoids generally reveal two subperpendicular planar magmatic fabrics, both marked by preferred orientation of biotite-rich schlieren, magmatic crystals, enclaves, and xenoliths (Figs. 8A, 8B). In some places, the older, gently dipping magmatic foliation Sm1 is rotated into cuspate-shaped, upward-pointing centimeter-scale upright folds associated with the second and steep axial planar magmatic foliation Sm2 (Fig. 8C). The orientations of both magmatic fabrics are parallel with the S1 and S2 metamorphic foliations in migmatites and low-grade rocks of the metamorphic envelope (Figs. 2A, 2B). The shallow-dipping magmatic fabric Sm1 does not carry a visible mineral lineation, whereas the steep S2-parallel magmatic foliation has a moderately plunging mineral lineation (Fig. 2C). Locally, the D1 horizontal fabrics are entirely transposed by D2 zones marked by a magmatic, submagmatic, and subsolidus amphibolite to greenschist facies fabric. The S1-parallel magmatic foliation is only locally preserved, mainly in the westernmost part of the western F3 limb (where S2 zones are as wide as 1 m), whereas the D2 deformation dominates along the eastern limb and in the hinge of the F3 regional fold (Figs. 2A, 2B).

Domains devoid of subsolidus D2 deformation preserve the original interaction of magmatic rocks of different composition. The magmatic contacts are commonly steep, parallel to the regional S2 orientation and/or alignment of magmatic minerals (Figs. 8D, 8E). These contacts are represented by meter-wide mingling and mixing zones between magmas of different mineralogy (Figs. 8D, 8E). These observations indicate that the bulk of the Chandman intrusive rocks was emplaced during the activity of the D2 deformation event. However, some magma batches intruded after the D2 deformation, as exemplified by an undeformed diorite vein crosscutting the S2 foliation of a granite gneiss (Fig. 8F). One of these gneisses was dated as 350.4 ± 1.7 Ma (U-Pb zircon SHRIMP; Kröner et al., 2010).

Decimeter- to hectometer-sized septa of host migmatites and Tugrug quartzitic and amphibolite schists are present along the margins of the magmatic core and in the western F3 limb. They are commonly subvertical and restricted to domains of S2-parallel magmatic foliations (Figs. 2A, 2B). The septa display internal close to isoclinal upright folds affecting the stromatic S1 layering with axial planes parallel to the magmatic Sm2 foliation (Fig. 8G). Septa composed of refractory rocks deflect the magmatic foliations. Their internal foliation is not folded by F2 folds and is either parallel or sharply discordant with the septa margins. This indicates that D2 did not affect these refractory layers at the time of their incorporation into the magmatic rock and that they only underwent rigid body rotation during later D2 deformation. Therefore, horizontal shortening D2 occurred during and/or outlasted assimilation of the country rocks.

Structural Pattern of the Intrusive Leucogranite Dikes in the High-Grade Core

Hundreds of decimeter- to meter-wide steep muscovite-bearing leucogranite dikes intrude the migmatites and the magmatic core of the Chandman massif (Figs. 2A, 2B). These dikes show sharp contacts with their host rocks (Fig. 9) and display a strong magmatic to solid-state foliation S2 (Fig. 9A). Although D2 did not fold the leucogranite veins, they have a shallow-plunging mineral lineation parallel to the regional L2 lineation.

Six types of contact relationships between the leucogranite dikes and sills and their host rocks were recognized. The leucogranite dikes emplaced in the mantling migmatites: (1) truncate the S1 stromatic layering at a high angle; (2) cut the F2-folded S1 stromatic foliation at an acute angle (Fig. 9A); (3) are in petrographic continuity with the stromatic S1 fabric (e.g., Fig. 7D); (4) are parallel to S2 and are locally boudinaged (Fig. 9B). In the core magmatic rocks, the leucogranite dikes: (5) are concordant to the S2 foliation or (6) truncate the host magmatic or solid-state foliations at an angle of ≤30°. All these relationships suggest that the muscovite-bearing leucogranite dikes were injected during progressive D2 deformation. The dikes are commonly strongly gneissified (Fig. 9A), indicating that they accommodate the D2 strain while the host rock remained mostly static and did not continue to fold after the dike emplacement. At the dike margins, the locally folded stromatic S1 layering forms decimeter-wide margin-parallel higher strained D2 zones, suggesting some amount of D2 strain localization into the host rock during or after melt crystallization (Fig. 9A).

Previously 40Ar/39Ar ages of 365 ± 18 Ma and 330 ± 11 Ma on white micas were obtained from samples of the peripheral greenschist facies metasedimentary envelope and mantling migmatites, respectively (Lehmann et al., 2010). However, large errors characterize these data, precluding the unraveling of rapid tectonic events. Here we present 6 new 40Ar/39Ar ages on 100–300 µm grain fractions from the metamorphic envelope (sample M680A), mantling migmatites (sample M717A), and the Chandman magmatic core (samples M635B and M776B) with the aim of characterizing the thermal history of the entire massif. Single-grain 40Ar/39Ar analyses of white mica and amphibole of samples M680A, M717A, and M635B were performed by laser step-heating at Geoazur laboratory (University of Nice, Sophia-Antipolis, France). Multigrain (0.2 mg) 40Ar/39Ar analyses of white mica and biotite of sample M776B were performed by laser step-heating at Spectrum (Central Analytical Facility of the Faculty of Science, University of Johannesburg, South Africa). Ages were calculated using Isoplot v. 4.15 (Ludwig, 2012). The decay constants of Steiger and Jäger (1977) and the atmospheric 40Ar/36Ar ratio of 298.56 ± 0.31 (Lee et al., 2006) were used (for detailed analytical procedures see Appendix A in the GSA Data Repository Item1). The 40Ar/39Ar step-release spectra are presented in Figure 10 and the data are shown in Tables DR1A and DR1B. The map in Figure 1B shows the location of the analyzed samples.

Metamorphic Envelope

In the metamorphic envelope, white mica was separated from an andalusite-bearing mica schist (sample M680A) located in the western F3 limb, ∼2 km south of the contact with the core magmatic rocks (Fig. 1B). In the outcrop, the mica schist displays a weakly developed spaced cleavage S3 that is not associated with any metamorphic growth (Fig. 3A). This rock consists of fine-grained white mica, quartz, chlorite, andalusite porphyroblasts, and large opaque minerals. Although the individual white mica and chlorite are pervasively oriented into the S2 metamorphic foliation, the alternation of quartz- and white mica–dominated layers marks the F2-folded S1 compositional layering. The plateau age of 347.4 ± 4.9 Ma was calculated from 9 consecutive steps that together compose 100% of the total 39Ar released (Fig. 10A).

Mantling Migmatites

In mantling migmatites, white mica was separated from an S2-parallel leucogranite sill (sample M717A) intercalated with a biotite-bearing diatexite with stromatic layering at the contact between the magmatic rocks and the migmatitic paragneisses. The locality is in the hinge region of the regional F3 fold (Fig. 1B). The leucogranite sill consists of quartz, K-feldspar, plagioclase, white mica, and rare chlorite. Large magmatic white mica crystals (0.2–1 mm) are aligned parallel to the margins of the sill. There is also finer grained white mica associated with zones of fine-grained feldspar and quartz that together define the margin-parallel greenschist S2 foliation. Seven consecutive ages on a large grain that compose 92.7% of the total 39Ar released provide a plateau age of 353.3 ± 3.7 Ma (Fig. 10B).

Magmatic Core Granitoids

In the Chandman magmatic core, amphibole was dated from a granodiorite (sample M635B) and white mica and biotite from a post-D2 pegmatite dike (sample M776B). Amphibole was separated from a foliated coarse-grained granodiorite in the eastern F3 limb (Fig. 1B). The granodiorite consists of plagioclase, amphibole, biotite, quartz, titanite, and minor epidote. A strong shape preferred orientation of euhedral plagioclase defines the magmatic foliation. The plateau age of 354.0 ± 5.3 Ma was calculated from 3 consecutive steps that constitute 99.6% of the total 39Ar released (Fig. 10C).

White mica and biotite were separated from a post-D2 pegmatite dike (sample M776B) cutting through a migmatite septum in the granodiorite, located in the western limb of the F3 regional fold (Fig. 1B). The sample contains quartz and coarse-grained, euhedral, and poorly oriented white mica, K-feldspar, plagioclase, and biotite. Two duplicates of white micas were measured. Analysis M776B-1 gave a plateau age of 344.8 ± 1.9 Ma calculated from 100% of the total 39Ar released (Fig. 10D). Analysis M776B-2 gave a pseudo-plateau integrated age of 345.5 ± 1.9 Ma calculated from 14 consecutive steps that constitute 97.7% of the total 39Ar released (Fig. 10E). The biotite plateau age of 333.9 ± 2.2 Ma was calculated from 10 consecutive steps that together contain 95.3% of the total 39Ar released (Fig. 10F).

In general, samples with an S2 fabric from the magmatic core and mantling migmatites gave identical 40Ar/39Ar ages of ca. 354 Ma, regardless of the dated mineral (amphibole or white mica), whereas samples from post-D2 magmatic rocks gave younger ages of ca. 345 Ma (white mica) and ca. 334 Ma (biotite). A sample from the mica schist metamorphic envelope with a well-developed greenschist facies S2 gave an 40Ar/39Ar age of ca. 347 Ma (white mica), slightly younger than 40Ar/39Ar ages from the nearby D2-deformed marginal migmatites and intrusive rocks.

Origin and Evolution of the Chandman Massif

We have studied in detail three different parts of the Chandman massif, the low-grade metamorphic envelope, the mantling migmatites, and the granitoid core, representing different crustal levels. The structural observations combined with petrology and geochronology allow the division of the mesoscopic structural evolution into three main stages: (1) horizontal crustal flow; (2) upright folding associated with vertical melt transfer; and (3) expulsion of melt along vertical dikes.

The oldest fabric is represented by a ubiquitous subhorizontal foliation S1 in migmatites (Fig. 6A) subparallel to sheets of granitoid (Fig. 8A) indicating horizontal flow of partially molten rocks and granitoids at deep crustal levels. The S1 layering was subsequently folded by open to closed upright folds (Fig. 7A) either transposed into the vertical S2 foliation or truncated at a high angle by steep walls of diatexites (Fig. 7B). Within the magmatic rocks, large volumes of granitoids exhibit a shallow-dipping magmatic foliation (Figs. 2A and 8A) locally reworked by S2 vertical fabrics (Figs. 8B, 8C). Altogether, these indicate that vertical channels parallel to axial planes of F2 folds progressively reorganized originally horizontal migmatites and granitoids (Fig. 7). The upward-cuspate migmatite structures (Figs. 7B, 7C) resemble the funneling network leucosomes of Weinberg et al. (2013) and the vertical cuspate structures mapped by Kisters et al. (1996), both facilitating efficient vertical melt and solid material extrusion. The steep S2-parallel leucogranite sheets crosscutting both magmatic core and mantling migmatites either exploit S2 or crosscut horizontal S1 fabrics (Fig. 9). The brittle character of the leucogranite dikes might reflect the embrittlement of melt-bearing wall rocks in response to a high strain rate during dike propagation, although the consistent orientation of the dikes (Fig. 2B) indicates that brittle failure was controlled by large differential stress rather than by elevated melt pressure (Cosgrove, 1997). In addition, the lack of migmatitic breccias rules out deformation at high strain rates. We consistently favor a model where the magmatite-migmatite core was solidified and cooled before the leucogranite dike intrusion. Away from the core, in the Ordovician and Carboniferous metasedimentary sequence, the D2 overprint is much weaker.

We report six new 40Ar/39Ar ages from the Chandman massif. For all the 40Ar/39Ar dated samples, volume diffusion of radiogenic Ar calculated by Warren et al. (2012) suggests that the argon content has been reset at inferred peak P-T conditions for both the metamorphic aureole and the mantling migmatites (Broussolle et al., 2015). The Chandman massif records protracted magmatism from 363.0 ± 3.1 Ma to 341 ± 3 Ma (Kröner et al., 2010). The timing of migmatization during D2 deformation is constrained by the 238U/206Pb monazite age of 356 ± 1 Ma for an S2-parallel stromatic migmatite (map of Fig. 1B; Broussolle et al., 2015). Syn-D2 to post-D2 cooling of the mantling migmatites is recorded by the 238U/206Pb monazite age of 347 ± 4 Ma (map of Fig. 1B; Broussolle et al., 2015), and the new 40Ar/39Ar white mica age of 353 ± 4 Ma (Fig. 10B). New 40Ar/39Ar white mica age of 347 ± 5 Ma for an andalusite mica schist of the metamorphic envelope is interpreted as a syn-D2 to post-D2 cooling age following peak metamorphism in the andalusite zone (Fig. 10A). Together with the new 40Ar/39Ar age of 354 ± 5 Ma for an S2-parallel magmatic amphibole from a core granodiorite (Fig. 10C), these ages imply rapid cooling during D2. They may also suggest that deformation and exhumation in the core and eastern F3 limb of the Chandman massif ended ca. 350 Ma. However, the zircon crystallization age of 341 ± 3 Ma from a D2 granitoid (Kröner et al., 2010) rather suggests that protracted and localized syn-D2 magmatism heterogeneously reheated a system that was already cooled at upper crustal levels at temperatures below any detectable Ar resetting in 40Ar/39Ar step-release spectra in samples from adjacent areas. Localized late syn-D2 magmas may have been the source of heat for the syn-D2 to post-D2 cordierite overgrowth in both the exhumed high-grade core and the laterally juxtaposed low-grade schists (Fig. 4A). A post-D2 white mica and biotite from the same pegmatite sample from the westernmost part of the massif yield 40Ar/39Ar plateau ages of 345 ± 2 and 334 ± 2 Ma, respectively (Figs. 10D–10F); a higher 40Ar closure temperature of white mica (∼400 °C) versus biotite (∼300 °C) can explain the difference in age (Harrison et al., 1985, 2009). This implies that D2 deformation in the westernmost part of the Chandman massif stopped before ca. 345 Ma.

The structural and geochronology interpretations discussed here allow development of a large-scale tectonic model of the Chandman massif. The D1 deformation produced subhorizontal metamorphic foliations in the upper crust at a pressure of ∼3 kbar (corresponding to depth level of ∼10 km), and in the partially molten deep crust at ∼7 kbar (corresponding to a depth of 20–25 km; Fig. 4B; Broussolle et al., 2015). Hanžl et al. (2016) reported a similar layered horizontal architecture from the nearby Tseel terrane that is interpreted to result from a vertical shortening event of Middle Devonian age. The D2 deformation is marked by isothermal decompression in the high-grade core and rapid cooling at depth level of ∼10 km, while the low-grade metamorphic envelope recorded only isobaric heating and cooling at the same level (Fig. 4B; Broussolle et al., 2015). The available geochronological data (monazite U-Pb dating of Broussolle et al., 2015) suggest that the hot migmatites reached middle crustal levels ca. 355 Ma and together with the metamorphic envelope cooled ca. 345 Ma (Fig. 4B); published U-Pb zircon ages from various types of granitoids point to continuous magmatic activity within this time span and until ca. 340 Ma (Hrdličková et al., 2008; Kröner et al., 2010). All these data can be explained by a model of localized exhumation of partially molten crust and granitoids in the core of a crustal-scale antiform.

Chandman Massif—Result of Detachment Folding?

The architecture and proposed tectonic evolution of the Chandman massif are similar to those of injection folds described previously in numerous high-grade terrains (the Archean in Scandinavia or Greenland) by Beloussov (1959, 1961). These structures are characterized by the accumulation of ductile material in anticlinal cores due to the flow of a deep ductile unit beneath brittle hanging-wall rocks. Kruger and Kisters (2016) suggested that detachment folding in suprasolidus conditions triggered accumulation and segregation of granitic magmas in the core of an upright, regional-scale anticline developed above a basal detachment; tightly folded cover sequences overlying voluminous leucogranites and a weakly deformed basement support their detachment model. An alternative scenario might be transtensional folding, which can also exhume deep crust in the core of large-scale upright folds (e.g., Fossen et al., 2013). However, the absence of S2 shear zones, L2 tectonics, and F2 hinge-parallel stretching in Chandman rules out this possibility. Instead, the following features of the Altai domes can be correlated with detachment fold structures: (1) alternations of narrow zones of high-grade, partially molten crust, rheologically weak and originally deep-crustal with low- to medium-grade supracrustal rocks (Figs. 1B and 11B); (2) relatively weaker deformation and limited burial of upper crust (i.e., the Ordovician to Devonian cover sequence and the upper part of the Tugrug formation) compared to the high-grade core of the Chandman dome or similar structures in the Chinese Altai (Jiang et al., 2015; this work); and (3) existence of an horizontal anisotropy in both the mid-crustal migmatites and the weakly metamorphosed or unmetamorphosed upper crust prior to formation of magmatite-migmatite domes (e.g., Jiang et al., 2015; Hanžl et al., 2016). These features are traditionally attributed to the mechanical behavior of salt during detachment folding (Blay et al., 1977) typically developed in the Pyrenees or Zagros Mountains (e.g., O’Brien, 1957). The prerequisite for such a behavior is a rigid basement overlain by a weak ductile horizon represented by a low-viscosity layer commonly represented by the source salt, and located beneath a brittle upper crust. We will argue below that these conditions are relevant to deformation and exhumation of partially molten crust and formation of the Chandman massif.

The main prerequisite of the detachment folding model is the existence of a rigid substratum, which is not exposed in the Mongolian Altai. However, its presence is inferred from two geological and geophysical studies (Guy et al., 2015; Jiang et al., 2016). Guy et al. (2015) and Jiang et al. (2016) showed that the Ordovician sedimentary precursor forming the Chinese and Mongolian Altai accretionary complex was genetically related to the erosion of the Lake zone and underlying Paleoproterozoic basement to the north; they proposed that the Altai sedimentary accretionary complex was subsequently molten and stretched during a Middle Devonian regional-scale extensional event (Hanžl et al., 2016), producing a layer of rigid granulite residuum beneath weak migmatized sediments (Jiang et al., 2016, fig. 12 therein). We argue that during the Devonian, the Altai accretionary belt was formed by three layers represented by a deep, rigid granulite residuum, an intermediate, weak partially molten layer (made of Devonian–Carboniferous granitoid intrusions and mantling migmatites), and a strong, weakly metamorphosed upper crust (located away from the contact-metamorphic aureoles of granitoid intrusions; Fig. 1B; Guy et al., 2015; Jiang et al., 2016). The granulite residuum layer inferred from geophysical and geochemical modeling seems to represent the best candidate to form the rigid substratum from which the low-viscosity migmatite layer could be detached during the Carboniferous Altai orogeny.

The similarity of the Chandman dome with detachment folds implies a sequence of evolutionary stages marking the interaction between large-scale folding and a ductile layer (granitoid magmas and migmatites) deformation. The three stages discussed above involve (1) formation of a syn-D1 subhorizontal low-viscosity layer formed by migmatites and granitoids; (2) opening of saddle reef structure at the top of the anatectic front associated with the influx of migmatites and granitoid melts from the surrounding regions during the early stages of D2 fold amplification (Fig. 11A); (3) locking of the antiform accompanied by massive shortening of the weak magmatite-migmatite core and the draining of magma to upper crustal levels (Fig. 11B). Leucogranite dikes were emplaced at this stage in an orientation generally parallel to S2 (Fig. 2B). This contrasts with the synmagmatic detachment folding identified by Kruger and Kisters (2016) where a significant number of dikes are parallel to the detachment fold profile. Kruger and Kisters (2016) suggested that these leucogranite sheets accommodate hinge-parallel stretching. We therefore do not think that hinge-parallel stretching is a major component of the D2 deformation in the Chandman massif.

The intermontane basins of early Carboniferous age in the Gobi-Altai and Lake zones occur between high-grade domains in wide synforms (Fig. 1B; Hanžl and Aichler, 2007; Markova, 1975; Rauzer et al., 1987). The 358 Ma detrital zircons (Kröner et al., 2010) suggest that effective erosion removed the overlying rocks of the Chandman massif and reached its high-grade core during the growth of the antiform. The detachment model explains well the limited burial in the metamorphic envelope at the time of upward movement of the magmatite-migmatite core (Fig. 4A).

The question arises whether the structural and petrological data presented here may reflect laterally forced buoyancy-driven ascent of high-grade rocks and magmas called detachment diapirism or pure detachment folding without the contribution of density inversion (Burg et al., 2004). In principle, both mechanisms are not mutually exclusive and can account for the quasi-symmetrical architecture of the core of the Chandman dome and the P-T-time paths related to the exhumation of partially molten crust (e.g., Maierová et al., 2014).

Nevertheless, we argue that the detachment folding without (or with negligible) contribution of diapiric rise of migmatite core is more likely in the studied accretionary complex. The calculated density difference between the partially molten paragneisses and the overlying sediments is too low (Jiang et al., 2016, fig. 9 therein) to activate Rayleigh-Taylor instability in reasonable time scales (Gerya et al., 2001). Existing geochronological data (Broussolle et al., 2015; Kröner et al., 2010; this work) suggest that the transition from horizontal flow to vertical rise of the Chandman core was very rapid, implying dominance of horizontal shortening rate over gravity (Burg et al., 2004; Lexa et al., 2011; Maierová et al., 2014). The structural observations presented for the Chandman dome are compatible with the presence of climbing folds in the core of the anticline parallel to upright folds in surrounding low-grade mantle, which are features favoring a detachment folding mode (Burg et al., 2004, fig. 14 therein). The metamorphic mantle reveals neither structural nor petrological evidence, such as burial and heating, typical of marginal synclines (Warren and Ellis, 1996; Petri et al., 2014).

Applying Equations 20 and 21 of Burg et al. (2004) for a wide range of parameters and plotting the resulting B and Bdet numbers into their phase diagram (fig. 5 of Burg et al., 2004), it can be shown that the Chandman dome should originate by the mechanism of detachment diapirism driven by gravity and lateral shortening. This is valid for a small density contrast (1 kg m–3), a high viscosity difference typical for competent and cold upper crust and hot partially molten lower crust (η12 > 1000), and a range of natural shortening rates. However, in the studied example, the upper crust is clearly brittle, ruling out even a rough estimation of its viscosity. In addition, the exhumed migmatized rocks, which probably lost a significant amount of melt, have to be denser compared to the highly porous Ordovician to Devonian sediments forming the bulk of the upper crust. These arguments, together with the structural observations listed here, indicate that a detachment folding model is most appropriate to explain both tectonometamorphic evolution and exhumation history of the Chandman dome.

Mechanical Interactions of Magma and Host-Rock Deformation during Detachment Folding

The presented model is related to questions that need further explanation and quantification. (1) Can large-scale detachment folding exhume partially molten crust to supracrustal levels without the contribution of diapirism, and what are the mechanisms of exhumation? (2) Can detachment folding explain magma collection in a perpendicular orientation with respect to the main shortening direction and the episodic emplacement of magmas during amplification of the antiform?

Exhumation of Partially Molten Crust and Detachment Folding

We address the first question here by means of a conceptual analogue model that was created to understand the amplification dynamics of crustal-scale detachment folds. The entire experimental setup was designed to simulate the deformation of hot crust above a rigid layer and in front of a moving indentor (Fig. DR1 in the Data Repository Item). In this model, the upper brittle crust corresponding to the Ordovician to Devonian metasedimentary units is represented by fine-grained quartz sand while the partially molten middle to lower crust (migmatites and magmatic rocks) was simulated by heated paraffin wax. Appendix B in the Data Repository Item gives details of the analogue setup, materials used, and scaling.

Our experiment started after 5 h of heating by shortening of modeled partially molten wax and sand domain above a movable heated plate (Fig. 12). During the experiment, at 11% shortening a broad bulge formed by the buckling of the rigid upper crust and molten wax accumulated underneath. In the next stage (11%–29% shortening), the bulge amplified into a large buckle fold with subvertical limbs, an arcuate hinge zone, and a cusp-shaped triangular core zone filled with melt (fold 1 in Fig. 12). Formation of a second fold to the right followed the amplification of the first fold (fold 2 in Fig. 12). A wide synform formed by the upper crust started to separate the two antiforms. During progressive shortening (until 35%), the two folds further amplified (folds 1 and 2 in Fig. 12) and partially molten wax was injected along axial planes from the triangular zones toward the fold hinges. The fold limbs became attenuated while the crestal part of the fold hinges and the upper crust were thinned by normal faulting and erosion (transfer of sand from hinges to synforms). Late amplification of folds 1 and 2 was accompanied by bulging and incipient folding of the multilayer closer to the backstop of the model (fold 3, Fig. 12). The progressive formation of detachment folds presented here is fundamentally similar to the sequential growth of buckle folds simulated by Blay et al. (1977).

The presented model shows that detachment folding effectively exhumes the deep, partially molten crust to supracrustal levels via two main exhumation mechanisms: (1) forced injection of low-viscosity molten material along the fold axial plane during the post-buckle flattening stage and (2) the localized erosion and extensional unroofing of the upper crust above the antiforms. Both mechanisms are supported by structural and metamorphic records in the migmatite-granite core and by presence of the localized siliciclastic lower Carboniferous basin rimming the southern margin of the Chandman dome. The whole finite geometry of the model corresponds well with the general structure of the Mongolian Altai marked by two antiforms cored by high-grade rocks surrounding wide upper crustal synform (Fig. 1B).

The entire experiment was carried out in 1 G conditions, a low-density contrast (wax density = 810 kg m–3, sand density = 1460 kg m–3) and a high rate of box shortening that implies an ∼30 × 106 times faster background velocity than a diapiric rate (Equation 18 of Duretz et al., 2011; Appendix C in the Data Repository Item). This indicates the negligible influence of gravity and a clear dominance of lateral pure shear on the growth of detachment folds in the wax and sand experimental set up.

Geometrical Model of Volume Evolution of the Core of the Detachment Fold

The second question addresses the problem of melt collection and its flow within the core of the detachment fold. We argue that our structural data rule out syn-D2 stretching parallel to F2 fold hinges, allowing us to study the geometric evolution in the core of a detachment fold in a vertical section orthogonal to F2 folds only. These calculations are used to explain the behavior of magma during the amplification of a detachment fold, features that cannot be discussed using the crustal-scale numerical (Burg et al., 2004; Maierová et al., 2014) or analogue models presented here. The main parameters controlling deformation of the magmatite-migmatite core is its volume change as a function of fold amplification and lateral shortening (Fig. 13; calculations in Appendix D in the Data Repository Item). In our model, the initial prefolding setting is characterized by a crustal thickness H, a ductile layer with thickness h, and an initial wavelength of anticline W0 (Fig. 13A).

During the first stage of shortening (to 50%; stage 1, Fig. 13), the rapid fold amplification is accompanied by an influx of melt into the core of the antiform expressed as the difference between volume generated in the fold core and the demise of volume in source area (ΔV with positive slope; dashed black curve in Figs. 13B, 13C). Upright folding of the originally horizontal anisotropy and flow of low-viscosity material in the core of the antiform deep in the crust characterize this stage. This can explain well the connection of S1 leucosomes and S2 leucogranite dikes (Fig. 7D), and vertical extrusion of diatexite walls along the axial planes of F2 folds (Figs. 7B, 7C). Folding of the horizontal foliation connected with the vertical flux of magma is possible because the entire region is dilated, but at the same time progressively horizontally shortened. This may provide an alternative explanation to the mechanical paradox of axial planar leucosomes forming perpendicular to the maximum shortening direction (Vernon and Paterson, 2001). It is important that the core of the antiform undergoes first upright folding deep in the crust, and the entire folded region is exhumed during later stages of fold growth (Fig. 13B).

The second evolutionary stage (stage 2, Fig. 13) is marked by a decrease of the rate of fold amplification (limb rotated by 60°; Fig. 13C), which is associated with post-buckle flattening (e.g., Price and Cosgrove, 1994). The progressive fold closure is manifested by a decrease in the difference between the volumes of magma in the core of the antiform compared to the source (Figs. 13B, 13C). This configuration results inevitably in overpressurization of the melt in the core of a box-shaped fold hinge area (Fig. 13A). This late stage of fold amplification is connected with the rise of overpressurized ductile walls, which may break and intrude the anticlinal crest along predisposed zones of weakness (e.g., Bonini, 2003, fig. 7 therein). The minimum diapiric pressure conditions necessary for viscous fluid to intrude a preexisting fracture in the crest of a detachment fold were calculated (e.g., by Acocella et al., 1999, or Schultz-Ela et al., 1993), suggesting that diapiric buoyancy forces are not needed in this context, in contrast to the diapiric model of injection folding of Beloussov (1961). In the Chandman massif, massive gneissification of granitoids, deformation of migmatites associated with further vertical extrusion of weak material, and rapid cooling of the system characterize this stage. The laterally compressed, solidified, and cooled magmatite-migmatite core is reintruded by steep granitoid dikes (e.g., leucogranite veins; Fig. 9) that are expelled from deeper crustal levels and exploit the axial planes of F2 folds and the S2 fabric. At this stage, the system is characterized by the loss of buoyant melt that is actively drained upward (Fig. 11B).

This study shows that the vertical extrusion tectonics explaining exhumation of deep-crustal low-viscosity rocks, as introduced by Thompson et al. (1997a, 1997b), is valid only for the late stages of detachment folding following rotation of limbs into a steep position. The rigid floor depth artificially introduced by Schulmann et al. (2003) as a necessary prerequisite for vertical elevation of weak rocks in a vertical tabular zone is here represented by a rigid basement from which the folded domain is detached. In fact, the detachment fold model represents an excellent link between the collection of rheologically weak material above the rigid floor in the fold core during the first rapid folding stage and the vertical extrusion of this material during the second post-buckle flattening stage. It is in particular the first stage that was missing in the Thompson et al. (1997a, 1997b) and Fossen and Tikoff (1998) models, and implies that the folding is the best mechanism initiating vertical extrusion of deep-seated and rheologically weak material (e.g., Štípská et al., 2004).

Geodynamic Implication of Altai Detachment Folding

An important question is whether the Devonian–Carboniferous deformation of the entire Altai arc-accretionary complex can be interpreted as a giant fold-thrust belt typical of thin-skinned tectonics (Coward, 1983; Glen and VandenBerg, 1987). The studies of Kozakov et al. (2002), Burenjargal et al. (2014), and Hanžl et al. (2016) show that the central part of the Altai complex was a scene of an unprecedented HT-LP event associated with important crustal melting, resulting in the intrusion of numerous crustal granitoids at 380–370 Ma. It can be proposed that the Altai accretionary complex in Mongolia was extended, heated, and partially molten ca. 380–370 Ma prior to the arrival of the early Paleozoic–Proterozoic indenter at 360 Ma (the Lake zone, Fig. 1B), which resulted in formation of large detachment folds at 350–340 Ma above a rigid granulite substratum. The model presented here is not dissimilar to the tectonic switching scenario marked by alternations of extensional and contractional events proposed to explain Paleozoic tectonic evolution of the Lachlan orogen in eastern Australia (Collins, 2002). Like the Altai, the Lachlan orogen is interpreted as a typical thin-skinned orogen resulting from the detachment of folded upper crust from the deep crust (Gray and Foster, 2004; Gray et al., 1991).

We suggest an exhumation model of crustal-scale detachment folding involving exhumation of partially molten lower crust and sequential intrusions of magmatic sheets along the axial planar foliations of F2 folds over a time period of ∼20 m.y. Magma-lubricated steep transfer zones distributed in the hot crust accommodated the 3–3.5 kbar upward entrainment of the mid-crustal anatectic material.

The detachment fold model presented here is characterized by the extrusion of deep, partially molten crust above a granulite residuum layer in the form of a ductile wall through an upper crustal lid. The fold core is initially filled by underpressurized and buoyant magma that becomes rapidly overpressurized and extruded upward as the fold locks, similar to salt plugs in fold-thrust belts.

Such a mechanism probably creates the lateral heterogeneous crustal pattern described in Phanerozoic accretionary orogens (Gray and Foster, 2004; Lehmann et al., 2010). The heterogeneous exhumation of partially molten hot crust is a feature typical of hot metasedimentary accretionary complexes that mechanically operate as crustal-scale thin-skinned fold-thrust belts, and our work offers a viable mechanism explaining this process.

This work was financially supported by the Czech Science Foundation (grant 17–17504S to Schulmann) and by funding from the Ministry of Education of the Czech Republic (grant LK11202). We thank Judith A. Kinnaird and the THRIP (Technology and Human Resources for Industry Programme) funding (grant TP2010072800046) for funding Lehmann’s visit to Prague. This research is partially funded by the DST-NRF (Department of Science and Technology–National Research Foundation) Centre of Excellence for Integrated Mineral and Energy Resource Analysis. We thank Otgonbaataar Djorsuren for field assistance, Arnaud Broussolle for discussions on metamorphic assemblages, and Pavel Hanžl for discussions on Chandman geology. George Henry is acknowledged for improving the English. We thank an anonymous reviewer, Aaron Yoshinobu, Felix Gervais, and associate editor Jean Bédard for constructive and helpful comments on an earlier version of this manuscript, and an anonymous reviewer and science editor Kurt Stüwe for useful reviews.

1GSA Data Repository Item 2017331, a summary of the analytical methods for 40Ar/39Ar geochronology (Appendix A and Table DR1), the experimental set-up, and preparation, description of analogue material and scaling relations of the analogue modelling (Appendix B, Figure DR1, and Tables DR2 and DR3), a calculation of relative contribution of diapiric velocity versus lateral, background velocity (Appendix C), and the analytical solution describing geometric evolution of the detachment folding (Appendix D), is available at http://www.geosociety.org/datarepository/2017, or on request from editing@geosociety.org.