Born in the Pacific and raised in the Caribbean: construction of the Escambray nappe stack, central Cuba. A review

During the Late Jurassic and Early Cretaceous the North and South American continental plates rifted apart, leading to the formation of oceanic crust in the so-called Proto-Caribbean area and the Gulf of Mexico (e.g., Pindell & Kennan, 2009; Pindell et al., 2012; Boschman et al., 2014). Much debate has arisen on the nature of the evolving intra-oceanic plate boundary connecting the east-dipping subduction zones of the North and South American Cordillera at the mouth of this widening seaway. Further debate involves the question of how an east-dipping subduction zone can realistically evolve into the west-dipping juvenile oceanic island-arc system that will become the intra-oceanic Great Caribbean Arc (GCA; Pindell et al., 2012) migrating through the space between the two American plates and subducting the Proto-Caribbean oceanic crust (Pindell & Kennan, 2009; Pindell et al., 2012; Boschman et al., 2014; Lidiak & Anderson, 2015). Here we follow the seminal suggestion of Pindell et al. (2012), i.e., given the relative plate motions, the intra-oceanic plate boundary should have been a trench-transform-trench system. The GCA can evolve from the bridging transform segment without the necessity of a debatable flip. Multiple collisions of the eastward-migrating Caribbean arc with the surrounding continental crustal domains occurred, in part because of changes in the direction of plate movements, such as the opening of the South Atlantic. In passing, the northern part of the GCA interacted with the continental block of the southern Yucatán platform (e.g., Pindell & Kennan, 2009; Martens et al., 2012; Solari et al., 2013; Maldonado et al., 2016). In addition, García-Casco et al. (2008) suggested that this northern GCA segment collided with and overrode a terrane they named Caribeana, an elongated submarine promontory with stretched continental crust and overlying sediments projecting off the Yucatán continental margin. Thus continental material was entrained into the migrating intra-oceanic GCA. As a consequence, the northwestern part of the arc was thrust onto Yucatán and subsequently onto the Bahamas platform of North America, producing the Late Cretaceous to Early Paleogene North Caribbean Suture Zone (NCSZ, e.g., Stanek et al., 2006, 2009; Fig. 1a). Owing to the eastward motion of the Caribbean plate and the formation of the Cayman transform fault, the northwestern, Cuban part of the NCSZ was left behind. The eastern part of the NCSZ, which was dismembered by strike-slip faults, was displaced eastward together with the present-day Caribbean plate and now forms the Antillean islands from Hispaniola to Puerto Rico. In this way, remnants of the subduction-accretionary complex of the GCA are spread out along the entire northern boundary of the Caribbean microplate (e.g., Iturralde-Vinent, 1994; Pindell et al., 2005; Iturralde-Vinent et al., 2008, 2016; Fig. 1a), but collisional features are best preserved on Cuba.

The subduction-related metamorphic rocks shown in Fig. 1a are one of the keys to understanding the geological history of the Caribbean (e.g., Pindell et al., 2005, 2012; García-Casco et al., 2008; Boschman et al., 2014). Two types of occurrences can be distinguished. In the first, serpentinite mélanges of the intra-oceanic stage of the GCA are exposed. Here, tectonic (corner flow) and rheological models (e.g., Cloos, 1982; Cloos & Shreve, 1988; Gerya & Stöckhert, 2002; Gerya, 2010) provide the possibility of explaining the mass-flow dynamics of rock units in the wedge-shaped channels of such subduction zones. Thus numerically predicted pressure-temperature-time (P-T-t) paths (e.g., Gerya et al., 2002) can be compared to those derived from field evidence and petrological studies (Krebs et al., 2008, 2011). These intra-oceanic GCA occurrences yield valuable information on the thermal structure, petrology, kinematics and evolution of intra-oceanic subduction zones in general, and sections of the GCA subduction zone in particular (e.g., Harlow et al., 2004; Krebs et al., 2008, 2011; Brueckner et al., 2009; Flores et al., 2013; Hertwig et al., 2016).

In the second type of occurrence (see Fig. 1a), the oceanic subduction-accretionary complexes of the GCA collided with and generally overrode the North American continental margin. Platform-type sedimentary rocks of the margin were subducted and often tectonically interspersed and reassembled in the subduction environment of the GCA. In these HP-occurrences the pre-collisional process of intra-oceanic subduction can be studied and traced from the initial interaction with the North American continent up to the final collision.

The nappe stack exposed in the Escambray Mountains (Fig. 1b) of Cuba represents a key area for studying the second type of subduction-related rocks mentioned above that involves collision with the continental margin (e.g., Iturralde-Vinent et al., 2016). In fact, a wealth of data-rich literature on the Escambray locality exists, but it is scattered across various publications, monographs, technical reports and theses, of which many are difficult to retrieve. It is the purpose of this review to provide an annotated compilation of information from Russian, Spanish, German, and English sources. Given that field access to key areas in the Escambray Complex has been politically restricted for quite some time, a review of such data can provide a solid foundation for future research thrusts.

A major problem in studying the subduction-related metamorphic rocks of the Caribbean has been that both the early and peak stages of subduction and collision processes are often poorly dated. The Ar–Ar, K–Ar, and Rb–Sr methods predominate, but these date growth of minerals only in low-temperature metamorphic rocks. In other situations only cooling ages are obtained. Constraints on subduction initiation can be inferred from the protolith ages of HP/LT rocks, or from the age of volcanism associated with subduction, but dating the peak metamorphic conditions (i.e., maximum subduction) is problematic. Zircon U–Pb dating can be ambiguous because it is not always clear whether the zircon was inherited from a protolith or actually grew at HP conditions. Other more promising methods such as Sm–Nd or Lu–Hf have not yet been widely applied (e.g., Krebs et al., 2008; Brueckner et al., 2009). To augment our review and add a robust time perspective to the existing data, we also present new Lu–Hf, U–Pb, Rb–Sr, and ⁴⁰Ar/³⁹Ar dates for key samples of metamorphic rocks of the Escambray Complex. From this framework, we then develop a comprehensive model for the formation of the Escambray nappe stack in the context of regional Caribbean tectonic history.

The laboratory methods and primary data sets as well as guiding principles and assumed closure temperatures for geochronology used in this paper are described in the Supplementary Materials (see Appendix S10, freely available online on the GSW website of the journal, https://pubs.geoscienceworld.org/eurjmin).

In central Cuba (Las Villas Region), the NCSZ trends ESE parallel to the island axis, and the Central Cuban Main thrust separates the Cuban fold-and-thrust belt of mostly non-metamorphic sediments to the north from ophiolitic and island-arc volcanic units to the south (Fig. 1b). The ophiolite sequence in the Las Villas region (locality 6b in Fig. 1a) is a mélange-like structure of highly deformed and serpentinized ultramafic rocks surrounding rare blocks of eclogite (e.g., García-Casco et al., 2002), shown as the “HP mélange zone of central Cuba” in Fig. 1b, and folded Paleogene clastic sediments (Somin & Millan, 1981). To the south, the non-metamorphic volcano-sedimentary sequences of the GCA are folded into a megasyncline and thrust northwards onto the serpentinite mélange. South of the lowlands underlain by these non-metamorphic GCA sequences, the Escambray Mountains expose the largest contiguous occurrence of metamorphic rocks along the northern Caribbean Plate boundary, extending for about 1800 km² (Figs. 1 and 2).

Morphologically, the Escambray Mountains can be divided into the western Trinidad dome (TD) and the eastern Sancti Spiritus dome (SSD), which expose a tectonic nappe stack comprising the tectonically upper Mabujina Amphibolite Complex (MAC; Somin & Millan, 1981), and several underlying nappes that together constitute the subduction-related Escambray Complex s.s. (EC). The present distribution of exposed units of the EC and the MAC is shaped by young, mostly post-Miocene brittle faults (Fig. S9 in Supplementary Material), especially the prominent North Escambray Fault (Fig. 2), which forms a scarp delimiting the uplifted metamorphic sequences to the south (Dublan & Álvarez Sánchez, 1986). All metamorphic units are unconformably covered by Eocene or younger sediments.

The first complete description of the EC was published by Somin & Millan (1981) and Millán & Somin (1985a), who revealed an inverted metamorphic zonation, with the tectonically lowermost and lowest-grade greenschist-facies nappes exposed in the core of the TD, surrounded by tectonically overlying higher-grade rocks. Millán & Somin (1985a) and Dublan & Álvarez Sánchez (1986) constructed a lithostratigraphic column based on rare observations of fossil remnants and lithological correlation with the non-metamorphic Mesozoic stratigraphic section of western Cuba (Millán & Myczynski, 1978; Somin & Millan, 1981; Stanik et al., 1981). Millán & Somin (1985b) and Millán Trujillo (1997) reinterpreted the complex in terms of tectonic nappes and thrusts and reassembled the various lithostratigraphic formations into tectonic units. The scheme of four fundamental tectonic super units of Millán Trujillo (1997) has been used as a template in the summaries presented by Auzende et al. (2002), Blein et al. (2003), Schneider et al. (2004), García-Casco et al. (2006a, 2006b), Despaigne-Díaz et al. (2016, 2017), and Cruz-Gámez et al. (2016). However, the inherent complexities of this interdependent and interwoven lithostratigraphic and lithotectonic nomenclature then led to the introduction of a restructured nappe designation of the SSD (e.g., Stanek, 2000) based on pressure-temperature-time (P-T-t) paths. In the following, we will show why the observed distinct P-T conditions of metamorphism, in combination with augmented geochronological data, lead to this logical alternative subdivision of the nappe stack (Fig. 2). This in turn sets the stage for a model of the geotectonic evolution of the Escambray Massif.

“Unit I” or the first super unit of Millán Trujillo (1997) is found only in the western Trinidad dome (“lgf” in Fig. 2). The metamorphic grade is low and sedimentary structures are preserved. The rock types are mainly metasediments (metacarbonate, calc-schist, metapelite, metasandstone, marbles, metachert) and basic metavolcanic rocks. Mylonitized serpentinite bodies are common. Initial descriptions of rock types, mineral assemblages, and observed metamorphic grade were provided by Somin & Millan (1981) and Millán & Somin (1981, 1985a, 1985b). The general structure and lithostratigraphy were refined by Millán Trujillo & Álvarez-Sánchez (1992). Recently, Despaigne-Díaz et al. (2016, 2017) presented a structural analysis from the lowest greenschist-facies nappes of Unit I up to the overlying eclogite-bearing Montforte klippe (Fig. 2). New microanalytical mineral data and ⁴⁰Ar/³⁹Ar data (see Sect. 4.4) allow Despaigne-Díaz et al. (2016, 2017) to identify three stages of structural evolution that they correlated with processes of subduction and exhumation.

Mineral assemblages in Unit I generally comprise phengite, chlorite, epidote, albite, and titanite, with actinolite important only in metabasic rocks and calc-schists. Quartz, calcite, and garnet rich in spessartine and grossular may be present (Somin & Millan, 1981; Millán & Somin, 1981, 1985a, 1985b; Despaigne-Díaz et al., 2016, 2017). Despaigne-Díaz et al. (2016) suggest the P-T path shown in Fig. 3. Temperatures are estimated to be 300–400 °C from the typical greenschist mineral assemblages. Pressures are estimated from the celadonite content of phengite directly obtained as Si atoms per formula unit (apfu) from the mineral analyses. These pressures only represent minimum values, however, because phengite does not occur within the critical univariant assemblage with quartz, biotite, and K-feldspar. At these low temperatures and Si values of >3.4 apfu, the calibration of Massonne & Szpurka (1997) can be considered to be the most realistic (Massonne, pers. comm., 1997; Simpson et al., 2000). Although Despaigne-Díaz et al. (2016) base their estimate on an Si content of 3.65 apfu, their tabulated values indicate no contents exceeding 3.5 apfu in the greenschist units. The P-T path exceeds this isopleth considerably in pressure; it lies in the high-pressure greenschist-facies (HPGS) P-T field, but the critical presence of lawsonite has not been documented. Figure 3 indicates that the metamorphic grade in the greenschist units of the TD is distinctly lower than in the Pitajones nappe of the SSD.

Despaigne-Díaz et al. (2016, 2017) noted the rare incipient growth of biotite and magnesio-hornblende oriented in the main S₂-foliation in some metabasic rocks, which they correlate with the exhumation stage. They ascribe this feature to a short-lived heating event during exhumation, such as might be expected when nappes of different temperatures are juxtaposed in a nappe stack (e.g., Rötzler et al., 1998).

The major contact of Unit I with the overlying Gavilanes unit (i.e., the Montforte Nappe) is marked by ductile, strongly sheared serpentinite horizons and lenses as well as intense folding (Despaigne-Díaz et al., 2016).

The Pitajones Unit (“upper greenschist-facies” unit in Fig. 2) was described and defined (Grevel, 2000; Stanek, 2000; Stanek et al., 2006) from detailed study in the SSD and eastern TD. The predominant rocks are monotonous carbonate and quartz-mica schists with intercalations of massive metacarbonates in the tectonically lower part (Fig. S11-1). The upper part exhibits carbonate-mica schists with tectonic slivers or boudins of greenschist and metagabbro, as well as massive black and grey marbles. Strongly deformed greenschist-facies mylonites and carbonate-mica schists mark the boundary to the overlying Gavilanes Nappe (Millán & Somin, 1981, 1985a; Dublan & Álvarez Sánchez, 1986; Millán Trujillo, 1997; Grevel, 2000; Stanek et al., 2006).

Auzende et al. (2002), Schneider et al. (2004), García-Casco et al. (2006b, 2008) and Despaigne-Díaz et al. (2016, 2017) described this rock unit as “blueschists” or “lawsonite blueschists,” which is misleading because no sodium amphiboles are known from anywhere in this unit (Somin & Millan, 1981; Grevel, 2000; see also Fig. S9, layer “Mineral assemblages (Somin & Millan, 1981)”). The misconception stems from the traditional mingling of lithostratigraphic nomenclature and rock descriptions. Millán & Somin (1981) defined a “Felicidad Member” in the TD as an integral part of a stratigraphic sequence later called the La Chispa Fm. (Millán & Somin, 1985a, 1985b), which is widespread in the outcrop area of the Pitajones Unit. In its type locality (see Fig. 2, locality 1), the Felicidad Fm. consists of metavolcanic greenschists (Ab + Act + Chl + Ep ± WM ± Qz ± Cal ± Lws). Despite the distinct differences in metamorphic grade, Somin & Millan (1974, 1981) also correlated lithostratigraphically a local occurrence of jadeite-bearing blueschist (see Fig. 2, locality 2) with the Felicidad Fm., leading to a later erroneous, generalized assumption that the La Chispa Fm. and Pitajones Unit as a whole also contain widespread glaucophane and jadeitite. However, the jadeite-bearing blueschists of locality 2 are allochthonous river deposits not part of the in situ La Chispa Fm. or Pitajones Unit (Grevel et al., 1998, 1999; Maresch et al., 2012).

The lawsonite-bearing greenschist assemblage described by Somin & Millan (1974, 1981) is also found along the northern boundary of the Pitajones in the SSD (Somin & Millan 1981; Grevel, 2000). Elsewhere in the Pitajones, Somin & Millan (1981) and Grevel (2000) identified pseudomorphs with relict lawsonite in metagabbros, so that lawsonite may have originally been more widespread. The lawsonite–actinolite association has not received much attention, but Bousquet et al. (2008) recognized a high-pressure greenschist (HPGS) facies (Fig. 3) with this assemblage as a transitional zone between greenschist- and blueschist-facies P-T conditions. The presence of lawsonite + actinolite is a function of the effective bulk composition. However, the existence of a HPGS facies and the general P-T conditions proposed by Bousquet et al. (2008) have also been corroborated by HP-LT isochemical P-T pseudosections calculated for seventeen different metabasic bulk compositions by Willner et al. (2009, 2016).

The P-T paths for the Pitajones Unit in Fig. 3 summarize presently available data. Mineral compositions corroborate that the HPGS was reached in the SSD. Actinolite of the Pitajones Unit in the SSD reaches 0.5 Na apfu in the B-site (Grevel, 2000), thus approaching the sodium-calcium amphibole winchite in composition. Winchite is typical for HPGS P-T conditions (Willner et al., 2009). Note that actinolite in Unit I (Sect. 3.1.1) in the TD reaches only Na^(B) = 0.22 apfu (Despaigne-Díaz et al., 2016). Grevel (2000) found that Si in phengite can reach 3.5 apfu, indicating minimum pressures typical for HPGS conditions (Fig. 3, Sect. 3.1.1). Lawsonite is replaced by clinozoisite, chlorite, and albite, pointing to a subsequent return to lower-pressure, greenschist-facies conditions (Millán & Somin, 1985a; Grevel, 2000). Late retrograde actinolite and phengite are essentially Na-free and low in Si, respectively (Grevel, 2000).

The maximum temperatures varied between 400 and perhaps 550 °C (schematically indicated in Fig. 3 by the subparallel P-T loops), with temperatures higher in the SSD than in the TD. Although garnet has not been reported from Pitajones metabasic rocks of the TD (e.g., Somin & Millan, 1981), it is common in metagabbros of the SSD, where it is almandine-dominant (Grevel, 2000). In garnet-bearing samples, actinolite can also be Al- and Na-rich, approaching barroisite in composition. Thus temperatures of over 500 °C were reached in the SSD and the P-T path entered the epidote-amphibolite field at high pressures of at least 7–8 kbar (Fig. 3), the lower pressure stability limit of barroisite (e.g., van Staal et al., 2008). Conventional geothermometers using garnet–amphibole, garnet–phengite, and/or amphibole–plagioclase pairs yield temperatures ranging from 410 to 520 °C for metagabbros and 480 to 520 °C for calcareous mica schists (Grevel, 2000). Tröss (1998) studied the carbon isotope fractionation between coexisting carbonate and carbonaceous material in samples of impure marble. She obtained a temperature of 383 ± 22 °C (averaged from a range of 363–435 °C) for ten samples from an outcrop of impure marble at the Hanabanilla dam site in the TD (locality 4 in Fig. 2). This bulk method depends strongly on the homogeneity of sample splits and the calibration used, and errors are difficult to quantify, but the relatively low temperature calculated was also corroborated by Beyssac (pers. comm., 2013), who obtained 382 ± 17 °C via Raman spectroscopy of carbonaceous material (e.g., Beyssac et al., 2002) on five of these same samples. This latter method yields information on the highest temperatures reached by a sample. Tröss (1998) obtained 464, 554 and 556 °C for three samples from the Pitajones unit in the SSD. The corresponding Raman-based results on the same samples are 550 ± 21, 537 ± 29 and 547 ± 23 °C, respectively (Beyssac, pers. comm., 2013).

The Gavilanes Unit or Nappe was described and defined (Stanek et al., 1998; Grevel, 2000; Stanek et al., 2006) on the basis of detailed investigations in the SSD and eastern TD. It coincides approximately with Unit III of Millán Trujillo (1997), but the tectonic contacts in the SSD have been more precisely determined and augmented by kinematic analyses (Stanek et al., 2006). This unit consists of stretched layers of quartz- and carbonate-mica schist with tectonic intercalations and slivers varying in size from tens to hundreds of meters of marble, high-grade antigorite serpentinite, and unambiguously high-pressure rocks such as eclogite, blueschist, and deerite-bearing metachert, as well as omphacite- and glaucophane-bearing garnet–mica schist (Somin et al., 1975; Somin & Millan, 1981; Millán & Somin, 1985a, 1985b; Millán, 1987; Grevel, 2000; Auzende et al., 2002; Schneider et al., 2004; Somin et al., 2005; Stanek et al., 2006; Grevel et al., 2006; Maresch et al., 2012). Locally, serpentinite and talcose rock of blackwall origin can coat these tectonic slivers. Eclogitic rocks can be found both as in-sequence bodies in the mica schists and marbles (Figs. S11-1e and S11-1f) as well as blocks within mélange-like serpentinite masses, which themselves have been tectonically incorporated into the schist sequences (Grevel, 2000; Schneider et al., 2004; Somin et al., 2005; Stanek et al., 2006). At the map scale, the Gavilanes Unit is a tectonic mélange with a matrix of quartz- and carbonate-mica schist (Stanek et al., 2006).

In general, the observed kinematic indicators of metamorphic fabrics in the Gavilanes unit as well as the greenschist-facies units of the EC show a tectonic transport top to N-NE (Stanek et al., 2006; Despaigne-Díaz et al., 2016, 2017) during the formation of the Escambray nappe stack.

The variety of mineral assemblages found in eclogite, blueschist, garnet–mica schist, and ferruginous metachert allows application of different conventional thermobarometers as well as multivariant mineral-equilibria calculations to estimate P-T conditions (Grevel, 2000, 2006; Schneider et al., 2004; Somin et al., 2005; García-Casco et al. 2006a). In keeping with the nature of a mélange, individual samples may yield distinctly different P-T conditions of metamorphism. According to the above authors, the peak metamorphic conditions recorded in eclogitic rocks lie between 580 and 630 °C at 15–20 kbar. One sample from the SE margin of the SSD yields 25 kbar (Grevel, 2000). Metasedimentary omphacite- and glaucophane-bearing garnet–mica schists yield 530–610 °C at 16–23 kbar (Grevel, 2000). Where significant segments of the P-T paths can be constructed, these are shown in Fig. 3. Prograde paths are not well defined, but García-Casco et al. (2006a, 2006b) suggest isobaric prograde paths for some eclogites. Grevel (2000) also found tight hair-pin loops in others. All P-T paths show a steep retrograde branch from 8 to 13 °C/km, suggesting exhumation during active subduction (Schneider et al., 2004; García-Casco et al., 2006a; Grevel et al., 2006; Stanek et al., 2006). All P-T paths converge at 300–400 °C and 5–10 kbar. These conditions correlate well with the field of coexistence (Fig. 3) of pumpellyite and lawsonite (Willner et al., 2009, 2016), which form late unoriented overgrowths (Grevel, 2000). A steep retrograde path is also indicated by deerite-bearing metacherts (Grevel, 2000). The well-established stability field of this mineral (Lattard & Le Breton, 1994), together with a temperature of ~470 °C obtained by oxygen-isotope thermometry on magnetite–quartz (Grevel et al., 2006) indicate that the metacherts must have passed through conditions exceeding 15 kbar at this temperature.

As in the Pitajones Unit, Tröss (1998) obtained temperatures from carbon-isotope fractionation in impure marble. Ten samples yielded a mean of 524 °C with a large standard deviation of 40 °C. Raman thermometry on carbonaceous material also yields similar results of 548 ± 35 °C for 13 samples (Beyssac, pers. comm., 2013). In both cases, the temperatures reached by the impure marbles appear to be significantly lower than in the eclogites of the Gavilanes Unit.

Dark amphibolites, commonly with large garnet crystals and minor intercalations of serpentinite and metasediment, are found in both the SSD and the TD. These rocks were named the Yayabo Formation by Millán & Somin (1981) and described by Somin & Millan (1974, 1981), Somin et al. (1975), and Millán & Somin (1976). Somin & Millan (1981) inferred a common tectonic and metamorphic history for the Escambray Complex and the Yayabo Formation. However, detailed study of the large, 2 × 10 km² main occurrence in the northeastern SSD (Fig. 2) has shown that these rocks differ from the Gavilanes Unit in age, metamorphic grade, and P-T path (Fig. 3), classifying the Yayabo Unit as a separate nappe (Grevel, 2000; Stanek, 2000; Stanek et al., 2006).

The rocks of the Yayabo Nappe vary from coarse-grained, garnet- and zoisite-rich amphibolites to fine-grained amphibolites rich in white mica. The dominant mineral assemblage is blue-green Amp + Ab + Ep/Zo + Rt and/or Ttn ± Grt ± Chl ± WM (mainly Ph, rare Pg) ± Qz (Somin & Millan, 1981; Grevel, 2000). The Si content of phengite varies from 3.1 to 3.4 apfu (Grevel, 2000). Amphibole (nomenclature of Hawthorne et al., 2012) cluster near the compositional boundary between the sodium–calcium and calcium amphibole subgroups (see also Fig. 4). The mineral assemblage is typical of the epidote-amphibolite facies (Fig. 3), and the occurrence of sodium–calcium amphibole attests to the high-pressure part of this facies (e.g., Ernst, 1979, 1988; Banno et al., 2000; van Staal et al., 2008). Estimates of P-T based on multivariant mineral equilibria (Grevel, 2000) are hampered by uncertainties in amphibole solid-solution models. Nevertheless, calculations are relatively consistent at 12–14 kbar and 580–650 °C, lying in a critical P-T range where the epidote-amphibolite, blueschist, eclogite and amphibolite facies impinge on one another (Fig. 3). An isobaric prograde path extending from ~520 °C is indicated by multivariant mineral calculations with inclusion assemblages in garnet (Grevel, 2000). The retrograde path shown in Fig. 3 is constrained by the observed re-equilibration of Yayabo rocks at lower-pressure epidote-amphibolite conditions.

Small bodies and layers of dark amphibolite similar to rocks of the Yayabo Unit are also found sheared into schists of the Gavilanes Unit along the eastern margin of the SSD and the northern and southern margins of the TD and SSD (Millán, 1978; Millán & Somin, 1981, 1985a, 1985b; Millán Trujillo, 1997). The tectonic style, together with the more or less parallel orientation of the dominant fabric in both amphibolites and carbonate-mica schists, suggests common ductile deformation. Thus these small bodies likely evolved within the Gavilanes Unit from a protolith similar to that of the Yayabo Unit. Millán (1978) reported partial replacement of “hornblende” by glaucophane, riebeckite, and winchite in Yayabo-like rocks, but the location of such samples was not given. On the other hand, Dobrezov et al. (1987) also mentioned retrograde transitions from their “eclogite type II” to Yayabo-like amphibolite. Whether these small occurrences of Yayabo-like rocks within the Gavilanes Unit actually correlate with the 2 × 10 km² main block will require further detailed investigation.

A very heterogeneous series of ortho-amphibolite with minor hornblende–plagioclase–biotite gneiss surrounding the EC was called the “Mabujina Series” by Somin & Millan (1981), on the basis of outcrops along the Mabujina River and the settlement of the same name. The MAC is exposed in two strips up to 10 km in width (Fig. 2) to the north of the TD (south of the town of Manicaragua) as well as to the northeast of the SSD (west of the city of Sancti Spiritus). Small outcrops at the southern and western rim of the EC (Somin & Millan, 1981; Millán & Somin 1981; Millán Trujillo, 1996), as well as similar rocks tectonically overlying the EC itself (Dublan & Álvarez Sánchez, 1986), have been interpreted to indicate a complete tectonic cover of the EC by the MAC. These small oucrops, north- and south-dipping metamorphic foliations along the northern and southern margins, respectively, and the serpentinite lenses along the contact are the main arguments for placing the Mabujina Unit on top of the tectonic nappe stack (Somin & Millan, 1981; Hatten et al., 1988; Millán Trujillo, 1996, 1997; Stanek et al., 2006, 2009). The northern outcrop area of the MAC (Fig. 2a) consists of fine- to coarse-grained basaltic metavolcanic rocks and metagabbros with metamorphic compositional layering. Abundantly interlayered sequences of ultramafic (La Lima), dioritic (El Quirro) and granodioritic to tonalitic (Piedras) gneisses were interpreted as multiple intrusions. Somin & Millan (1981) described massive amphibolites with relics of clinopyroxene as well as porphyritic metabasalts embedded in sheared matrices. South of Manicaragua, a stretched, apparently fault-bounded belt of only low greenschist-facies metavolcanic rock sequences has been called the Porvenir Formation by Dublan & Álvarez Sánchez (1986) and Millán & Somin (1985a). The northeastern outcrop area of the MAC (Fig. 2a) consists of similar amphibole-rich rocks, smaller amounts of granitoid gneiss, thinly bedded metavolcanic rocks (Millán & Somin, 1975; Rojas-Agramonte et al., 2011), and narrow strings of serpentinite bodies (Stanik et al., 1981). The MAC is cross-cut by the Cumanayagua granitoids in the north (Dublan & Álvarez Sánchez, 1986) and the Las Tozas igneous massif in the east near Sancti Spiritus (Stanik et al, 1981; Lang et al., 1986). Rojas-Agramonte et al. (2011) grouped all post-kinematic granitoids together, both in the MAC as well as in the Cretaceous island-arc unit to the north, as the so-called Manicaragua Batholith. Numerous small felsic pegmatite bodies, which cross-cut the metamorphic foliation of the amphibolites, show indications of a late but only minor tectono-metamorphic overprint (Somin & Millan, 1981; Dublan & Álvarez Sánchez, 1986; Grafe et al., 2001). These intrusions are important constraints for the timing of the metamorphic overprint of the MAC (see also Fig. 7). The contacts between the MAC and surrounding Cretaceous island-arc-related volcanic rocks have been mapped as being tectonic, because of the presence of low-temperature shear zones and cataclastic rocks (Mlcoch, in Dublan & Álvarez Sánchez, 1986).

The metamorphic overprint of the MAC has been described by Somin & Millan (1981), Stanik et al. (1981), Mlcoch (in Dublan & Álvarez Sánchez, 1986), Millán Trujillo (1996), and Grevel (2000). The highest-grade rocks are found in a zone near the contact with the EC and represent the tectonically lowest level of the MAC (Stanik et al., 1981; Millán Trujillo, 1996). The dominant mineral assemblage in amphibolites of the northern MAC is Hbl s.l. + Pl(An_30-70) ± Cpx, whereby the clinopyroxene is diopside and/or augite and considered to be a magmatic relic. In the eastern MAC, amphibolites can contain almandine garnet or biotite as well (Mlcoch, in Dublan & Álvarez Sánchez, 1986; Millán Trujillo, 1996; Grevel, 2000). Millán Trujillo (1996) also mentioned incipient migmatization. In this context, it is surprising that none of the above authors reported any evidence of a contact-metamorphic influence of the kilometer-scale Manicaragua intrusion. Further away from the EC, the most widespread mineral assemblage in metabasic MAC rocks is Hbl s.l. + Pl(An_10-50) ± Grt ± Ep (Mlcoch, in Dublan & Álvarez Sánchez, 1986; Millán Trujillo, 1996; Grevel, 2000), which is typical for the transition from the epidote-amphibolite to the amphibolite facies (e.g., Spear, 1993). Relics of magmatic clinopyroxene are commonly observed enclosed in garnet (Grevel, 2000), and magmatic plagioclase (An_50-70) is strongly saussuritized and replaced by epidote, white mica and sodic plagioclase (Mlcoch, in Dublan & Álvarez Sánchez, 1986; Millán Trujillo, 1996; Grevel, 2000). Towards the north and east, still further away from the EC, a transition to greenschist-facies assemblages is observed (Millán & Somin, 1975; Dublan & Álvarez Sánchez, 1986). Incipient retrograde metamorphism to greenschist-facies assemblages is common in all zones (Stanik et al., 1981; Millán Trujillo, 1996; Grevel, 2000).

The high variance of the mineral assemblages involved and the uncertainties in identifying equilibrium assemblages makes P-T estimates in the MAC difficult. Most are based mainly on conventional two-mineral exchange equilibria, yielding only temperatures. Mlcoch (in Dublan & Álvarez Sánchez, 1986) estimated 640 °C and 700–750 °C in the highest-grade zones, and 450–550 °C in meta-quartz-diorite and 550–650 °C in amphibole-pyroxene gneiss further away from the EC contact. In the northeastern MAC (Grevel, 2000), the Am–Pl thermometer of Holland & Blundy (1994) yielded remarkably high temperatures of 700–750 °C for an assumed pressure of 7.5 kbar. In the eastern MAC (Grevel, 2000) the Mabujina assemblages are more variable. Amphibolites with Pl(An_30-50) and garnet yield high temperatures of ~730 °C at an assumed 7.5 kbar as above for Am–Pl pairs, but most samples contain more sodic plagioclase, yielding 590–690 °C. In one sample of garnet amphibolite, application of the Kohn & Spear (1990) barometer yielded 8.5–8.8 kbar at 616 °C for an assumed Am + Grt + Pl equilibrium assemblage. The above rather broad P-T constraints are summarized in Fig. 3.

Grevel (2000) studied a critical profile in the Yayabo River (locality 3 in Fig. 2), in which the MAC can be followed via a semi-ductile mylonitic shear zone into rocks of the Yayabo Unit. The shear zone involves intercalations of rocks from both the Yayabo and Mabujina Units. Distinguishing these units in the field is difficult; rocks of both units are dark, fine- to coarse-grained amphibolites, in part with distinctive garnet porphyroblasts up to several cm in diameter. The Yayabo-type rocks show L-type tectonic features and compositional layering (Stanek et al., 2006). The assemblage is Am + Ab + Ep/Zo + Chl + Ttn ± Rt. In Mabujina-type rocks, epidote and chlorite are subordinate and ilmenite is the Ti-rich accessory phase. Most importantly, the Mabujina-type rocks contain albitized relics of magmatic plagioclase and rare inclusions of magmatic clinopyroxene in garnet. In the mylonitized shear zone separating the two units, the amphibolites are greenish and very fine-grained. Epidote, chlorite, and albite accompany abundant amphibole. Figure 4 shows that amphibole composition can be used to distinguish Yayabo and Mabujina rocks and also to estimate the P-T conditions of the shear zone. Yayabo amphibolites outside the shear zone contain sodium–calcium amphiboles as part of their primary assemblage, reflecting the high-pressure metamorphic origin of the Yayabo Unit (see Sect. 3.1.4). Mabujina rocks contain conspicuous brownish relics of pargasite and tschermakite, reflecting their magmatic source. In mylonitized amphibolites of the shear zone, magnesio-hornblende is the dominant amphibole, and growth of magnesio-hornblende can be observed in rocks of both the Yayabo and Mabujina in proximity to the shear zone, indicating convergence of pressure-temperature conditions. The assemblage Mg–Hbl + Ab + Ep + Chl of the shear-zone rocks indicates that the shear zone was already active at the P-T conditions of the epidote-amphibolite facies, and shows that the Yayabo and Mabujina Units were juxtaposed at these conditions (Fig. 3).

Grevel (2000) estimated retrograde P-T conditions in and near the shear zone from the secondary assemblage Am + Ep + Chl + Ab (Triboulet, 1992), which led to relatively consistent temperatures of 510–550 °C at 3–5.5 kbar.

Stratigraphic correlation of the metasedimentary units of the EC with the non-metamorphic Lower Jurassic to Middle Cretaceous sequences of the eastern continental margin of Yucatán is generally accepted (Millán & Myczynski, 1978; Somin & Millan, 1981; Stanik et al., 1981, Pszczółkowski, 1978, 1999; Somin et al., 1992; Iturralde-Vinent, 1994). In contrast, the nature and age of the protoliths of the high-grade metabasic sequences of the MAC, the Yayabo Unit, and the metabasic blueschists and eclogites of the Gavilanes unit have been the subject of a long-lasting debate. The metagabbros and -basalts of the MAC are considered to represent subduction-related basement of the GCA, i.e., the Cretaceous island arc exposed in central Cuba (Millán & Somin, 1985b; Dublan & Álvarez Sánchez, 1986; Stanek et al., 2009), but have also been interpreted as part of the upper ophiolite sequence of Cuba (Haydoutov, 1986). Blein et al. (2003) suggested provenance of the MAC as a tectonic sliver from western Mexico, on the basis of geochemical and isotope data and similarities with the composite Jurassic-Early Cretaceous Guerrero terrane. Inherited zircons in the MAC rocks could indicate a Neocomian proximal setting near SW-Mexico (Rojas-Agramonte et al., 2011). Schneider et al. (2004) found distinct geochemical differences among eclogites of the Gavilanes Unit and concluded that some were derived from a calc-alkaline, arc-like source, whereas others represented an N-MORB source.

Here we present new geochemical data of metabasic rocks from the Pitajones, Gavilanes, Yayabo and Mabujina Units, in order to compare these with literature compilations, to distinguish the different protoliths from each other, and to compare our interpretation with those of previous studies by Blein et al. (2003), Schneider et al. (2004), and Cruz-Gámez et al. (2016). The metabasic rocks of the greenschist-facies Unit I (data from Cruz-Gámez et al., 2016) represent metabasalt and metadiabase in which magmatic relics are still preserved (Cruz-Gámez et al., 2016). Fourteen eclogitic rocks from the Gavilanes Unit were subdivided on the basis of field relationships (see also Dobrezov et al., 1987). The first group are in-sequence in the regional metasedimentary matrix and the second group (here called exotic) are eclogite blocks within serpentinite or a talc-schist (blackwall) mantle. The serpentinite or blackwall bodies themselves are in tectonic contact with the metasedimentary matrix. The data set includes three eclogites previously studied by Schneider et al. (2004) and two by Cruz-Gámez et al. (2016). A suite of twelve garnet amphibolites sampled along the Yayabo River represents the Yayabo Unit. Four samples from the MAC consist of three metagabbros from the Yayabo River section and one amphibolite from the Jicaya Valley of the northern rim of the EC. Five subduction-related gabbros from the Camagüey section of the Cretaceous island arc (unpublished data of K.P. Stanek) of central Cuba are included for comparison. All data are given in Table S1 (Supplementary Material).

Along with the MORB-normalized element data plots in Fig. 5, two reference patterns for subduction-unrelated basaltic magmas are shown: OIB (alkaline, plume-related) and E-MORB (low-grade melting, rift-related basalts). In addition, data from Dilek & Furnes (2011) were used to construct patterns for subduction-unrelated, continental-margin basalts and for subduction-related, tholeiitic (IAT) volcanic-arc basalts. The Yayabo garnet amphibolites follow the E-MORB REE pattern without any significant negative or positive HFSE anomalies. The three greenschist samples of Unit I of the western TD (Cruz-Gámez et al., 2016) have significantly higher Th/Zr than the Yayabo amphibolites and no Nb–Ta depletion. In contrast to the Yayabo element trend, small negative spikes of Hf and Ti can be recognized. The greenschists are enriched in LREE compared to E-MORB at similar HREE contents and lie between E-MORB and OIB. The Mabujina metagabbros mimic the pronounced subduction-related “Th/Nb subduction-input proxy” (Pearce, 2014) of the Cretaceous island-arc gabbros from Camagüey. Both also show Zr, Hf, and Ti depletions similar to those of volcanic arc basalts. The metabasic rocks from the Gavilanes Unit indicate a depletion of Nb, Hf, and Ti only. The trace-element patterns of the two Gavilanes eclogite types are almost parallel between Nb and Ti. However, there is a significant difference in HREE_(n) content. The in-sequence eclogites are enriched in HREE_(n), comparable to continental-margin basalt, whereas the exotic eclogites show a distribution of HREE_(n) similar and parallel to subduction-related melts (island-arc gabbro).

The available geochemical data provide evidence that the metabasic rocks within the various units of the EC and MAC evolved from different protoliths. The garnet amphibolites of the Yayabo Unit likely originated as an E-MORB protolith not related to any subduction process. On the other hand, the amphibolites and metagabbros of the MAC appear to be former juvenile products of a subduction zone, a conclusion that was also reached by Blein et al. (2003). The overlap in the geochemical constraints for the eclogitic rocks from the Gavilanes unit and the greenschists of Unit I makes it difficult to reach an unequivocal conclusion. Considering the sedimentary nature of the host rocks of the in-sequence eclogites and greenschists, the most probable protolith could be a continental margin basalt from a magma-poor, continental-rifting setting (e.g., Whitmarsh et al., 2001). The protolith of the exotic eclogite type could be a volcanic-arc basaltic rock.

A summary of available geochronological data for the EC and MAC has been compiled in Table 1 (see also Table S8 and Fig. S9 in the Supplementary Material). There are a number of ages quoted in older Cuban and Russian literature, mainly K/Ar data, without clear indications of sample locality and/or rock type (Somin & Millan, 1981; Dublan & Álvarez Sánchez, 1986; Mossakowski et al., 1986; Hatten et al., 1988; Dublan et al., 1988). These data are referenced in Iturralde-Vinent et al. (1996), but excluded from Table 1.

First attempts to date the peak metamorphism of HP-metamorphic rocks from the Gavilanes Unit using U–Pb in zircon were made by Hatten et al. (1988) and Somin et al. (1992). These authors studied clogite from Loma de los Guapos (locality 5 in Fig. 2), where short, prismatic to spherical zircons yielded concordant ages between 106 and 102 Ma and were interpreted to have formed during HP metamorphism. However, these early results and their interpretation led to fundamental problems. Additional dating performed in the last 30 years (see Table 1) yielded cooling ages indicating that exhumation should be younger than 80 Ma, so that the time lag between a metamorphic peak at 106–102 Ma and exhumation appeared unrealistically long. For this reason, we carried out both additional U–Pb SHRIMP and Lu–Hf analyses (see Sect. 4.2) on the classical Loma de los Guapos eclogite occurrence. The U–Pb SHRIMP measurements on Sample S153 yielded a lower intercept at 104.7 ± 1.6 Ma (Fig. 6a) and corroborated the results of Hatten et al. (1988) and Somin et al. (1992). Only some very thin unstructured rims along the zoned cores yielded younger ages of 88–90 Ma, indicating that some resetting of the U–Pb isotope system must have occurred (see Appendix S3, S4). Despite the agreement in age, the internal zoning of the zircons in sample S153 and their Th/U values suggest a magmatic rather than a metamorphic origin.

Additional U–Pb dating from other localities has led to widely different results. Zircons from eclogite sample S145 from the eastern SSD have a lower intercept age of 148 ± 5 Ma (TIMS, Fig. 6d; Grafe, 2000), but SHRIMP data on the same zircon fraction gave slightly discordant ages in the range 170–150 Ma (Fig. 6c), and suggest that HP metamorphism partly reset younger rims on older cores. The 148 Ma TIMS “age” of Grafe (2000) could thus be the result of disturbed U–Pb systematics by Pb loss of old components. The protolith age of sample S145 should be near 170 Ma, in agreement with the majority of the core ages (see Appendix S3, S4).

For an eclogite from the northern SSD, Somin et al. (2005) reported a mean age of about 245 Ma (CA-TIMS) and an age range of 270−140 Ma (SHRIMP). The authors interpreted the results as the age spectrum of detrital zircons from a sediment derived from Central America. Permo-Triassic to Grenville ages were also measured for assumed detrital zircons in eclogites and blueschists from the northern SSD (Rojas-Agramonte et al., 2006). Another sample considered to be a HP-metasediment is S232, a glaucophane–garnet-bearing mica-schist with discordant zircons allowing an interpretation as apparently detrital zircons of Pan-African to Middle Proterozoic age (Fig. 6e; Grafe, 2000). The upper intercept of assumed discordia lines for different zircon groups anchored at 148 Ma (Grafe, 2000) show a broad “age” spectrum of inherited components. For the present study, a compositionally finely layered eclogite from the northwestern part of the SSD (sample S357) was also investigated by SHRIMP and yielded two groups of zircons (Fig. 6b). Rounded, short prismatic crystals have concordant ages of 2474 ± 30 Ma (see Appendices S3 and S4). The second group comprises strangely zoned euhedral crystals with “ages” of about 15 Ma, which cannot be interpreted in a geological context. Proterozoic zircons have also been found in western Cuba (Rojas-Agramonte et al., 2008), indicating possible provenance of these minerals from the Guayana craton.

Considering all available U–Pb data from zircon of HP-metamorphic rocks in the Gavilanes Unit, likely ages for the basic igneous protoliths of the eclogites would be 104 Ma and 170 Ma. Metasedimentary HP-rocks yield a broad age range of detrital zircons from Pan-African to Paleoproterozoic ages. All zircons have homogeneous, metamorphic rims, but the metamorphic peak temperatures of the Gavilanes Unit obviously did not reach the T_c of zircon of about 850 °C to completely reset the U–Pb isotope system. Most of the U–Pb zircon data should be interpreted as mixed dates ranging between the protolith age and the time of metamorphism-induced multi-stage Pb-loss.

A single fraction of titanite yielded an age of 70.9 ± 4.9 Ma (Grafe, 2000). The age was obtained by U–Pb TIMS on metamorphic titanite of the glaucophane–garnet-bearing mica-schist (sample S232, Fig. 6f), and was corrected with the common lead content of metamorphic phengite.

Initial attempts to date orthogneiss in the MAC using U–Pb in zircon were made by Bibikova et al. (1988), who found an Early Cretaceous metamorphic overprint on the basis of 118–108 Ma dates (interpreted as intrusion ages) on felsic MAC gneisses, and a 93 Ma date obtained on an undeformed quartz diorite. More recently, extensive U–Pb SHRIMP dating (Rojas-Agramonte et al., 2011) indicated that the protolith ages of tonalitic and trondhjemitic orthogneisses of the northern MAC range between 112.1 ± 2.1 and 132.9 ± 1.4 Ma (Table 1). These rocks are intruded by a suite of foliated and metamorphosed granitoid rocks (93.1 ± 0.7 and 93.5 ± 0.5) Ma, and only slightly to completely unfoliated igneous rocks (88.7 ± 0.7 to 84.2 ± 0.8 Ma). In the eastern MAC segment (northeast of the SSD), metabasaltic to meta-andesitic sequences yield protolith ages of about 93 ± 1.0 Ma, and deformed and metamorphosed felsic plutonic rocks have ages of 92.8 ± 0.7 Ma (Rojas-Agramonte et al., 2011; Table 1). Quartz monzonite samples of 89–87 Ma age are described as “poorly foliated” to “foliated” (Rojas-Agramonte et al., 2011), and apparently underwent late deformation. The first undeformed granitoid rock reported by these authors in the eastern MAC yields an age of 83 Ma. Thus the Mabujina Unit seems to be not only lithologically heterogeneous (see Sect. 3.2), but also exhibits a significant range in protolith age. Nevertheless, the timing of deformation and metamorphic overprinting in this complex “unit” is a key constraint for the interpretation of the metamorphic history of southern central Cuba (e.g., Rojas-Agramonte et al., 2011).

The field relationships of different types of orthogneiss and foliated igneous rocks have been mapped in detail in the valley of the Rio Jicaya (Fig. 7; Nr. 6 in Fig. 2) of the northern MAC (Grafe et al., 2001, Stanek et al., 2006). Grafe et al. (2001) reported Rb–Sr and ⁴⁰Ar/³⁹Ar ages for rocks from this area. A folded sequence of amphibolite and tonalitic-trondhjemitic gneiss with subhorizontal WNW-ESE trending shear indicators is intruded by foliated granodiorite and associated granite pegmatite. The granodiorite intruded the hinge of the folded amphibolite-facies foliation, whereas the pegmatite dikes run parallel to the main foliation. The metagranodiorite exhibits a coarse undulating metamorphic foliation of white mica and biotite; the feldspar is broken and subgrains have been rotated into the foliation. The granite pegmatite appears almost unfoliated, with only the coarse white mica showing some kinking (Grafe et al., 2001).

For the present study, new U–Pb SHRIMP ages have been obtained (Fig. 8; Table 1, Appendix S2). For a trondhjemitic orthogneiss (sample S142 in Fig. 7), the 130.9 ± 3.1 Ma age is considered to date the magmatic crystallization of the metatrondhjemite (Fig. 8a) and corroborates the results published by Rojas-Agramonte et al. (2011). Five zircon fractions were also separated from the foliated metagranodiorite F012 and the granite pegmatites F009 and F010 (Fig. 8). The U-rich rims of the short, prismatic zircons in the metagranodiorite sample F012 (see Appendix S4 of the Supplementary Material for CL images) yield a 89.5 ± 1.6 Ma crystallization age, but the finely zoned cores yield ages of about 123 Ma, similar to the ages determined by Rojas-Agramonte et al. (2011) for other trondhjemitic orthogneisses (F77) of the northern MAC. The needle-shaped zircons of sample F009 (fraction F009A) yield rim ages of 93.9 ± 0.7 Ma, and the short- to long-prismatic, well-zoned zircons of fraction F009B have rim ages of 90.4 ± 1.3 Ma (Figs. 8c and 8d). The needle-like, unzoned zircons of fractions F010 A and B are concordant and yield ages of 91.8 ± 1.0 Ma and 90.6 ± 1.1 Ma, respectively (Figs. 8e and 8f). These intrusion ages shift the time of the latest ductile tectono-metamorphic overprint down to ~90 Ma.

An undeformed and post-metamorphic granodiorite (sample S301; Fig. 8g) from the adjacent Cumanayagua Intrusion (Manicaragua batholith in the sense of Rojas-Agramonte et al., 2011; see Sect. 3.2) yields an age of 89 ± 1 Ma by multigrain zircon analysis (Grafe, 2000). This result also coincides with the data of Rojas-Agramonte et al. (2011), indicating post-orogenic intrusions into the amphibolite-orthogneiss sequences in the central part of the MAC between 89 and 84 Ma.

The core-rim relationships of the investigated zircons are noteworthy. There are no inherited zircons among those of the trondhjemitic gneiss S142. The zircons of pegmatite F009 show 90–94 Ma rims and uniform cores of 219.4 ± 2.6 Ma. Only a few zircons have Proterozoic “detrital” ages of ~1900 Ma. The cores of zircon grains in metagranodiorite sample F012 and metatrondhjemite S142 have Th/U values of ~0.32 (Fig. 8h). The 90–Ma rims of all investigated zircons have Th/U values of only 0.1–0.03. All Th/U values suggest a magmatic origin of the zircons. It seems likely that the metagranodiorite melt was derived from a proto-trondhjemitic rock similar to that of sample S142.

The results in Sect. 4.1 underscore that dating the peak or near-peak of HP-metamorphism in the Escambray Complex by U–Pb in zircon is problematic without additional detailed geochemical and petrological investigations (Rubatto et al., 1998). For this reason, five eclogite samples from the Gavilanes Unit and one garnet amphibolite from the Yayabo Unit were selected for Lu–Hf dating (see also Krebs et al., 2007; Stanek et al., 2009). For further comparison, two of the five eclogite samples from the Gavilanes Unit were taken from the classical Loma de los Guapos locality where previous extensive U–Pb dating had been done (Hatten et al., 1988; Somin et al., 1992).

The supplementary information needed to interpret the Lu–Hf results is summarized for eclogite S357 as a typical example in Fig. 9, and the analogous data for all samples are given in the Supplementary Material (Appendix S5, Table S6). Sample S357 was taken from a sequence several meters thick consisting of eclogite and blueschist lenses in a matrix of quartz-rich mica schist from the northwestern margin of the SSD (Sopimba locality). The eclogite is also compositionally layered. Garnet, clinopyroxene, and white mica (mainly paragonite and minor phengite) occur as porphyroblasts in a fine-grained matrix of epidote, subhedral amphibole, and clinopyroxene. The omphacite crystals have overgrown epidote, titanite, and rutile, and are compositionally zoned. Garnet includes all matrix minerals (Fig. 9). Multivariant equilibria calculations (Grevel, 2000) using the TWQ method and zoned garnet plus adjacent omphacite and paragonite inclusions yielded 14.5–15.5 kbar at 500–540 °C for the prograde path and 18.5 kbar at 600–620 °C for the metamorphic peak. The assemblage Barr + Ep + Pg + Qz gave a retrograde coordinate of 11.5 kbar at 470 °C. The Lu–Hf isochron is defined by three garnet fractions, omphacite, phengite, and the whole rock, and yields an age of 69.9 ± 1.6 Ma (Fig. 10e).

Eclogite S145 from the eastern border of the SSD (Buenos Aires locality) exhibits a compositional layering of omphacite and glaucophane mantled by green amphibole. The matrix minerals are chlorite, phengite, and epidote. Two garnet types are abundant in the rock: large, inclusion-filled crystals several millimeters in diameter and a smaller inclusion-free type. The maximum P-T conditions of 25 kbar at 520 °C are the highest recorded in the EC and were calculated using idiomorphic garnet in equilibrium with phengite and omphacite (Grevel, 2000). It was possible to use all mineral fractions (garnet 1–4, omphacite 1 and 2, whole rock) for the construction of the Lu–Hf isochron, resulting in an age of 71.0 ± 0.6 Ma (Fig. 10a).

Outcrops of eclogitic rocks near the Rio Higuanojo reservoir at the southern margin of the SSD were also investigated. The fine-grained matrix of foliated eclogite slices (S150) in the country rock consists of anhedral omphacite and subhedral epidote, with phenocrysts of euhedral garnet and rare sodium amphibole. The outer edges of the eclogite bodies have been retrograded at greenschist-facies conditions. The blue amphibole is mantled by secondary green amphibole and quartz. Only the cores of the up to 5 mm diameter garnet crystals contain inclusions of quartz, rutile, epidote, amphibole, omphacite, and white mica, whereas the outer zones of the garnet are inclusion-free. The seven-point Lu–Hf isochron of garnet (1–3), omphacite 2, phengite, green amphibole, and the whole rock yields an age of 69.95 ± 0.85 Ma (Fig. 10b).

The classical eclogite body of Loma de los Guapos in the northern Trinidad Dome had been previously sampled by several groups of geochronologists for U–Pb and Ar–Ar dating (Hatten et al., 1988; Somin et al., 1992, Renne in Draper & Nagle, 1988). The eclogite body is mantled by serpentinite and forms a large block in carbonate-mica schists and marbles. Samples S153 and S154 are medium- to coarse-grained eclogites with a groundmass of omphacite, zoisite, epidote, sodium, sodium–calcium and calcium amphibole, as well as white mica (paragonite and phengite). Actinolitic amphibole, in part coarse-grained and conspicuously emerald-colored in hand specimen, has locally overgrown the relict calcium amphibole, omphacite and garnet. The latter forms crystals up to 10 mm in size, with omphacite, epidote, amphibole, titanite, and rutile inclusions in the core. The rim of the garnet has overgrown earlier deformation features and contains inclusions of quartz, rutile, epidote, and omphacite. The two rocks yield identical ages, with the bright pink garnet fractions (2–4) from the less-magnetic fraction, omphacite (1, 2), amphibole, and whole rock of S153 yielding 69.6 ± 0.8 Ma (Fig. 10c), and garnet, omphacite, and whole rock of S154 yielding 70.27 ± 0.67 Ma (Fig. 10d). The dark red garnet (1) from the more magnetic fraction in S153 is considered to be an earlier prograde component.

The five Lu–Hf isochrons on eclogite thus yield a uniform age of 70 ± 1 Ma, the interpretation of which depends on the spatial distribution of Lu within porphyroblasts and the timespan over which garnet grew. For example, Lapen et al. (2003) showed that early sequestration of Lu into the cores of growing garnet porphyroblasts could bias Lu–Hf dates toward the onset of garnet growth relative to Sm–Nd dates. In the present study, Lu concentration peaks were observed in garnet cores (e.g., S150), closer to the rims (e.g., S145), or in both locations (S154). However, recasting Lu content as a function of volume rather than position along grain diameter (Scherer et al., 2000; e.g., Fig. 7 of Lagos et al., 2007) reveals that for our samples, Lu–Hf generally records the mean age of garnet growth (i.e., the time when roughly half of the garnet volume had been reached). The time interval required for total garnet growth can be estimated from studies of the GAC subduction zone by Krebs et al. (2008, 2011). Assuming from these data that garnet volume increased at a constant rate over 1 Myr or less (see also the P-T path in Fig. 9a), the outermost rims of these grains grew ~0.5–0.6 Myr or less after the mean garnet age. Thus for these metabasic rocks, Lu–Hf dates the prograde and near-peak metamorphism. Hence, an asynchronous overprint of the HP-metamorphic rocks of the Gavilanes unit of the EC, as discussed for instance by Schneider et al. (2004), seems unlikely.

The garnet amphibolite sample S321 (Yayabo Unit) was taken several hundred meters west of the major shear zone in the river outcrop of the Rio Yayabo (Stanek et al., 2006; locality (3) in Fig. 2) discussed in Sect. 3.3.2. This HP-metabasic rock (Fig. 3) contains a high proportion of zoisite, which occurs as idiomorphic needles up to several millimeters long. Amphibole and zoisite are oriented almost in parallel, producing a typical L-tectonite. The very fine-grained matrix consists of amphibole, anhedral epidote, and titanite. Garnet contains inclusions of quartz, epidote, titanite, and phengite. In addition, zoisite contains inclusions of amphibole, albite, and apatite, indicating the late growth of zoisite during peak-metamorphic conditions. The Lu–Hf isochron age, defined by garnet (1–3), amphibole (1, 2), zoisite, and whole rock fractions is 80.8 ± 2.4 Ma (Fig. 10f), considerably older than the Lu–Hf ages of eclogites from the Gavilanes unit. Based on garnet growth analysis analogous to that for eclogite, this age can be interpreted to date prograde and near-peak metamorphism.

No Rb–Sr or Ar–Ar data are available for the Pitajones unit, but Despaigne-Díaz et al. (2016) have recently published ⁴⁰Ar/³⁹Ar data on greenschist-facies rocks of Unit I of Millán Trujillo (1997) from the TD. These authors separated phengite from mica schist, calc-schist and metacarbonate and, on the basis of petrographic observation, grouped these mica fractions into S₁ and S₂ foliation types reflecting an earlier prograde D₁ and a later retrograde D₂ deformation. The weighted mean fusion ages for seven samples are given in Table 1 and were then recalculated by Despaigne-Díaz et al. (2016) to weighted mean averages of 60.5 ± 0.6 Ma for D₁ and 56.32 ± 0.40 Ma for D₂.

As far as the Gavilanes Unit is concerned, the five new eclogite Lu–Hf ages of ~70 Ma presented in this study provide a robust constraint for interpreting previous K–Ar, ⁴⁰Ar/³⁹Ar, and Rb–Sr results. Given that the Lu–Hf ages represent prograde or near-peak metamorphic conditions, it is plausible to assume that all K–Ar ages older than ~70 Ma are unrealistic, and have probably been affected by excess argon to some unknown degree. Considering that the Lu–Hf ages are essentially identical, it is reasonable to expect that the cooling (exhumation) histories of the HP Gavilanes rocks should show similar patterns as well. Table 1 summarizes 12 available Rb–Sr and ⁴⁰Ar/³⁹Ar determinations that range between 64.5 and 71 Ma for four samples of eclogite and a glaucophane–garnet–phengite schist. The age differences can reasonably be related to different argon release temperatures among the white micas (muscovite, phengite or paragonite; Forster & Lister, 2014) and amphibole, as well as the grain-size dependency of T_c for the Rb–Sr and ⁴⁰Ar/³⁹Ar ages. Schneider et al. (2004) provided a very instructive map of ⁴⁰Ar/³⁹Ar spot fusion analyses in a 1-mm phengite grain from eclogite sample LV66. The ten spot ages obtained vary from 66.6 to 72.2 Ma. The aggregate age from the sample is 68.2 ± 0.6 Ma. Schneider et al. (2004) had already noted that, due to the high closure temperatures, their Rb–Sr ages on phengite and ⁴⁰Ar/³⁹Ar ages on amphibole ranging between 70 and 65 Ma should represent cooling ages very close to peak-metamorphic conditions. The new Lu–Hf results support this assumption. To more precisely constrain the cooling path, we performed a high-resolution ⁴⁰Ar/³⁹Ar step-heating experiment on coarse-grained muscovite from a marble boudin in the Gavilanes Unit (sample S010, Fig. 11c, Table 1, Appendix S7), obtaining an age of 67.6 ± 0.4 Ma, which essentially corresponds to the average of the other tabulated cooling ages.

The Lu–Hf age of 80.8 Ma obtained on garnet amphibolite S321 from the Yayabo Unit indicates that this unit underwent distinctly earlier metamorphism than the Gavilanes Unit. To corroborate this result, we performed a high-resolution ⁴⁰Ar/³⁹Ar step-heating experiment on white mica from a quartz-mica vein cross-cutting Yayabo amphibolites near the S321 locality (sample S204, Fig. 11b, Table 1, Appendix S7) and obtained an age of 74.6 ± 0.5 Ma.

Grafe et al. (2001) dated the cooling path of the MAC for late-orogenic pegmatites cross-cutting the amphibolites (see Table 1). Depending on the textural context of the separated mineral fractions, the Rb–Sr isochrons were interpreted in terms of “grain-size-dependent isotopic reequilibration” as representing either “near-crystallization from melt” or later “metamorphic recrystallization” ages. Considering in addition the U–Pb ages of about 90 Ma from zircon in the pegmatites of the Rio Jicaya (see Sect. 4.1.2), the slightly younger Rb–Sr dates were interpreted as cooling ages reflecting the different closure temperatures of white mica and biotite. The coarse white-mica fractions yielded “near-crystallization from melt” ages of about 87–84 Ma for the northern MAC segment and ~82–80 Ma for the eastern MAC segment. The ⁴⁰Ar/³⁹Ar ages on white mica (~73 Ma) corroborate the Rb–Sr isochron ages on fine-grained white mica and biotite of ~74 Ma (assuming closure temperatures of ~350–300 °C for both; Grafe et al. 2001). Intermediate dates between 74 and 84 Ma were interpreted as meaningless “mixed ages” (Grafe et al., 2001), as subsequently corroborated by a ⁴⁰Ar/³⁹Ar step-heating experiment (sample F010, this study) on coarse-grained white mica. The weighted plateau age of 81.5 ± 0.5 Ma (see Fig. 11a, Appendix S7) reflects a closing temperature of about 375 °C, which coincides with the youngest Rb–Sr ages on white mica and is in the range of the total gas age of 73.8 ± 0.5 Ma on very fine-grained white mica from nearby sample F012.

In the eastern MAC segment, there are no U–Pb data available on the pegmatites, but the U–Pb ages of the youngest unfoliated granitoid rocks are ~83 Ma, rather than ~90 Ma as observed in the northern segment (Table 1). Correspondingly, the Rb–Sr cooling ages of coarse- and fine-grained white mica in the pegmatites (Grafe et al., 2001) also show younger cooling ages compared to the northern MAC segment. The coarse-grained minerals yield younger ages of ~82–80 Ma, close to the age of crystallization of the pegmatites, whereas the ages determined on fine-grained white mica are quite similar to those of the northern segment (Table 1). Thus, the eastern MAC appears to have been exhumed about ~8–9 Ma later than the northern segment. Only after ~74–72 Ma is a common exhumation path indicated.

Figure 12 shows pressure-temperature (-time) and pressure-time diagrams in which the P-T paths of Fig. 3 are augmented by the new and reviewed age data of Sects. 4.1–4.3. The Mabujina, Yayabo, Gavilanes and greenschist-facies units not only display distinctly different P-T paths, but also differ in the timing of peak metamorphism and exhumation rate.

Peak metamorphism and the related ductile deformation in the MAC must have occurred at ~90 Ma in Mabujina North and ~83 Ma in Mabujina East because the oldest undeformed granitoid magmatic rocks intruding the amphibolites in these areas are younger. The protoliths of all metamorphosed intrusive rocks are older (Table 1; Rojas-Agramonte et al., 2011). Burial did not exceed 30 km. However, the earlier history of the MAC is speculative. The U–Pb zircon protolith ages of orthogneisses and amphibolites are as old as 133 Ma (Table 1), but we do not know whether these protoliths were exhumed and then reburied, or whether they remained at depth and began exhumation at ~90 Ma. We note that the 133–112 Ma magmatic suite is found only in the northern Mabujina segment and not in the eastern part. There also appears to be a difference in the exhumation history of these two segments after 90 Ma. Although the P-T paths are essentially parallel, the northern MAC segment had reached upper crustal levels of less than 10 km depth by 81 Ma, whereas the eastern segment reached the same level some 5–8 Myr later. Exhumation rates thus vary from ~2 to ~1 mm/yr.

Burial depths in the Yayabo Unit correspond to intermediate levels of 50 km, and the exhumation path contrasts with that of the other units by being near-isothermal at temperatures equivalent to the epidote-amphibolite and amphibolite facies (Figs. 3 and 12a). Peak metamorphic conditions were reached at ~80 Ma, i.e., ~10 Myr after the MAC and ~10 Myr before the Gavilanes Unit. The time lag of ~6 Myr between peak metamorphism and cooling to below 400 °C is shorter than in the MAC, but clearly longer than in the case of the Gavilanes Unit. The exhumation P-T path is also distinctly different and more “indirect,” in that an initial near-isothermal exhumation is followed by a more isobaric leg before the closure temperature of white mica is reached. Mean overall exhumation rates of ~5 mm/yr are indicated, but the rate for the near-isothermal segment must have been higher, because the second leg should correspond to the lower rates of the MAC. It is important to stress that the Yayabo Unit had already been exhumed to shallow crustal levels several million years before the HP rocks of the Gavilanes Unit had reached maximum metamorphic conditions.

Rocks in the Gavilanes Unit record clockwise, hairpin-like P-T loops. Maximum burial varies from ~65 to ~80 km for various parts of this mega-mélange. Peak metamorphic conditions were reached at ~70 Ma. Exhumation and cooling from peak metamorphic conditions down to 375 °C occurred in only ~2 Myr. Although this time interval seems short, exhumation of the Gavilanes Unit occurred along very cool exhumation paths ranging from 8° to 13 °C/km (Sect. 3.1.3; Fig. 3). Temperatures below 400 °C would have been reached at still considerable depths. Even for eclogites with peak temperatures exceeding 600 °C, such cooling could be achieved at initial exhumation rates of ~15 mm/yr, a value that is not unreasonable when subducted continental material is involved (see Agard et al., 2009). It is interesting to note that eclogite S145, for which maximum pressures of 25 kbar have been calculated (Sect. 3.1.3; Grevel, 2000; Fig. 3), underwent the longest exhumation and consequently yields the youngest ⁴⁰Ar/³⁹Ar age of 64.5 ± 0.4 Ma on phengite. This time interval translates to an exhumation rate of ~10 mm/yr.

The greenschist-facies units of the EC show relatively open clockwise loops. Maximum depths of burial are less than 30–35 km, and peak metamorphic conditions were reached at ~60 Ma, about 10 Ma later than for the Gavilanes Unit. The ⁴⁰Ar/³⁹Ar data of Despaigne-Díaz et al. (2016) suggest exhumation rates of ~3 mm/yr.

The metamorphic complexes at the Northern Caribbean plate margin not only reflect the subduction history of the Caribbean Arc, but some of their constituent rock sequences also bear witness to an origin outside the present Caribbean realm and before the formation of the Caribbean plate itself. The initiation of subduction along the east-facing subduction zone of the Caribbean Arc was estimated to have occurred between 135 and 120 Ma (Pindell et al., 2006, 2012; Rojas-Agramonte et al., 2011; Boschman et al., 2014; Lidiak & Anderson, 2015). These estimates are corroborated by the ages obtained directly on subduction-related metamorphic rocks, which are as old as 104–120 Ma (García-Casco et al., 2008; Krebs et al., 2008, 2011; Cárdenas-Párraga et al., 2012), and by numerical simulation modeling of the dynamics of the Great Arc intra-oceanic subduction zone (Krebs et al., 2008). The protoliths of the Escambray Complex consist of volcano-sedimentary sequences (Gavilanes and Pitajones Units), considered to belong either to the Jurassic-to-Middle-Cretaceous sedimentary cover of the southeastern Chortis-Yucatán block (Iturralde-Vinent, 1994, 1996), or to oceanic crust (Yayabo Unit, Somin & Millan, 1981). The onset of this sedimentation and the related minor volcanic activity was linked to the Mesozoic rifting of Pangea (Stanek, 2000; Pindell et al., 2005; Pindell & Kennan, 2009). The U–Pb ages of zircon from HP-metamorphic rocks of the Gavilanes Unit clearly predate the initiation of the Great Arc of the Caribbean at 135−120 Ma. Only one date from the Loma de los Guapos exotic eclogite represents a protolith age (104.7 Ma) that could be interpreted as part of the oceanic crust of a fore-arc basement. This Middle Cretaceous basic protolith could have been incorporated into the Gavilanes mélange during the collision of the GCA with – and the subduction of – continental margin sequences, presumably at ~80 Ma. The geotectonic position of the protoliths of the heterogeneous amphibolite- to greenschist-facies Mabujina Amphibolite Complex, now located on top of the tectono-metamorphic nappe pile in southern central Cuba, is still a matter of debate. There are several interpretations of this rock unit: as the basement or lowermost part of the Cretaceous island arc (Somin & Millan, 1981; Dublan & Álvarez Sánchez, 1986; Millán Trujillo, 1996; Iturralde-Vinent, 1998; Grafe et al., 2001; Stanek et al., 2006), or as part of an ophiolite sequence (Haydoutov, 1986). Alternatively, Blein et al. (2003) suggested that the MAC represented an exotic sliver of southern Mexican continental crust. The protoliths of the mafic igneous rocks of the MAC, intruded by 120–130 Ma granodiorite to trondhjemite, are considered to be Jurassic to Cretaceous tuffs on the basis of poorly preserved fossils (Dublan et al., 1988). Again, these rocks must predate the opening of the Caribbean oceanic gap.

The metamorphic P-T-t-paths of the EC and MAC units are a direct result of the geotectonic development of the northwestern part of the GCA and the westernmost part of the North Caribbean Suture Zone (Pindell et al., 2012). The P-T-(-t) and P-t diagrams (Fig. 12), together with the geological and geochemical data (Fig. 5), show that the EC and MAC must represent an amalgamation of tectono-metamorphic units with contrasting provenance and metamorphic history.

As pointed out in Sect. 4.4, the sequential timing of peak metamorphism and stacking of the Escambray units can now be clarified to a high degree (Fig. 12c). The Yayabo Unit and the eastern segment of the MAC came together after ~80 Ma and before ~75 Ma. The missing link is that no direct analogous evidence exists for Mabuijna North, but the similarities in P-T-path for MAC North and MAC East suggest a reasonably similar scenario, even if initial exhumation was earlier in MAC North. The Gavilanes Unit joined MAC + Yayabo at ~68 Ma or shortly thereafter. The HP-metamorphic rocks of the Gavilanes and Yayabo (with Mabujina East) Units joined the Pitajones Unit and other greenschist-facies units after the D2 event dated by Despaigne-Díaz et al. (2016) at 56 Ma. It is possible that the proximity of “warmer” HP units and cooler greenschist-facies units might explain the “short-lived heating event” described by Despaigne-Díaz et al. (2016).

Two points of reference provide a basic template for the geotectonic reconstruction presented below and allow the dated tectono-metamorphic events to be followed back in time (see also Fig. 13). First, the time of arrival of the subduction-accretionary front of the GCA at the southern edge of the Bahamas platform is constrained by stratigraphic data on sediments in westernmost Cuba (Guaniguanico terrane at ~62–63 Ma; Pszczółkowski, 1994, 1999). Balancing the western Cuban fold-and-thrust belt (Saura et al., 2008) leads to an estimated shortening of 200 km and shifts the original site of sediment accumulation back to 20°N (present coordinates).

The second point of reference is the extinction of subduction-related magmatism in the western GCA at 72–75 Ma, which indicates the collision of the GCA with the southern North American continental margin at Yucatán (Hall et al. 2004; García-Casco et al., 2008; Stanek et al., 2009; Ratschbacher et al., 2009; Martens et al., 2012; Solari et al., 2013; Maldonado et al., 2016, 2018). This collision implies shifting of the extinct arc and its fore-arc complex over a distance of about 900 km between 72 and 46 Ma (the distance from point B to A in Figs. 13b and 13a, respectively), leading to a velocity of ~34 mm/yr. Krebs et al. (2008) estimated an orthogonal subduction velocity of 15–24 mm/yr for the active GCA, averaged over the life span between 120 and 55 Ma. This average velocity is used in our model for the reconstructed events before the 75 Ma Yucatán collision. In line with the direction of plate movement during the evolving GCA (Pindell et al., 2012), the 90-Ma magmatic events post-dating metamorphism in the Mabujina Unit (Table 1) were thus placed back to the eastern margin of the Chortis block, which at this time was still part of the southwestern active margin of southwestern Mexico (Rogers et al., 2007; Flores et al., 2015; Keppie, 2012).

Our model can be summarized in terms of six time frames, five of which are shown in the time panels of Fig. 13.

The pre-90 Ma situation: We follow the suggestion of Pindell et al. (2012) and assume that subduction in the west-dipping GCA began in the Early Cretaceous between 135 and 120 Ma, after rifting of North and South America in the Late Jurassic. The mafic protoliths of the MAC must themselves be older than the metamorphosed 132–112 Ma granitoid suite of intrusions (Sect. 5.2). These protoliths must have been buried to at least the granitoid intrusion depths of 8–15 km (Petford et al., 2000; Guillot et al., 1995). Epidote-amphibolite- to amphibolite-facies metamorphism and ductile deformation affected both. The metamorphic P-T conditions of the MAC are not well constrained (Sect. 3.2); the pressures appear to lie between those typical of orogenic collision tectonics (Barrovian) and those expected for a magmatic arc basement (Buchan or Abukuma). Nevertheless, the geochemical data (Sect. 3.3) suggest that the amphibolites and metagranitoid rocks are subduction-related. In fact, the P-T conditions to be expected under continental magmatic arcs are not well understood because local contact metamorphism associated with intruding plutons can mask the regional background P-T gradient. Within error, the MAC P-T conditions approach the background P-T gradient derived by Rothstein & Manning (2003) for the eastern Peninsular Ranges batholith of Baja California, Mexico, suggesting an origin in a Mid-Mesozoic magmatic arc system developed on protoliths such as those described by Baumgartner et al. (2008) as the Mesquito Composite Oceanic Terrane (MCOT) along the western margin of Chortis. The MCOT consists of serpentinites, basic volcanic and intrusive rocks, and Mesozoic radiolarites. The reoriented MCOT (after initial eastward motion of the Caribbean plate) fits the estimated position of the MAC well (position E in Fig. 13e). The youngest felsic gneisses (112 Ma) and the oldest unfoliated granitoid rocks intruding the northern MAC (as old as 90 Ma) limit the amphibolite-facies overprint to 110–90 Ma (position E in Fig. 13e). The provenance of the Mabujina protoliths from an Oaxaca-like terrane was already suggested by Blein et al. (2003) and discussed by Rojas-Agramonte et al. (2011). However, on the basis of lithology (Rogers et al., 2007; Baumgartner et al., 2008; Ratschbacher et al., 2009; Flores et al., 2015), the MAC protoliths most probably originated from Mesozoic tectonic units like the MCOT of southwestern Mexico. In any case, the MAC appears derived from a Pacific margin source. Rocks similar in age are scattered throughout the MCOT (Baumgartner et al., 2008), in the Siuna mélange at the eastern edge of the Chortis block (metabasites with ~139 Ma phengite ages: Flores et al., 2007, 2015), in the Guatemala suture zone (144–126 Ma ages of HP-rocks of the serpentinite mélanges of the Guatemala suture zone: Brueckner et al., 2009), and in the northern serpentinite mélange of Las Villas in central Cuba (eclogite blocks with Early Cretaceous ⁴⁰Ar/³⁹Ar cooling ages: García-Casco et al., 2002). Other exotic blocks could be related to the continental basement of Chortis or Yucatán. The 172 Ma granite of the Socorro Complex included in the accretionary complex of western central Cuba (Somin & Millan, 1981; Renne et al., 1989) could have been decoupled from a Rabinal-like complex (Solari et al., 2013) in Guatemala (southern Yucatán block) and tectonically transported to its present location. The 220 Ma inherited zircon cores in the syn- to post-orogenic pegmatites in the MAC show possible precursor relationships with a migmatitic thermal event at this time in the Rabinal Complex and similar units of the southern Maya-Chortis block (Ratschbacher et al., 2009). A similar Triassic protolith age was mentioned from the El Guayabo gneiss in western Cuba (Somin et al., 2006). Here, the metamorphic overprint is as young as the peak metamorphism of the Gavilanes eclogites. Llanes-Castro et al. (1998) reported a Jurassic paleontological age from basalts of the non-metamorphic ophiolitic sequence in Las Villas. It appears likely that, at least from 130 to 80 Ma, both the oceanic sequences from the MCOT and the continental material from the Chortis – Yucatán blocks were involved in tectonic transport by the GCA.

90 Ma: Beginning of intrusion of granitoid rocks produced by GCA activity into the gneisses and amphibolites of the MAC; the MAC, which had been inherited from the active western margin of the Chortis block, had become part of the basement of the GCA. The relocation of the MAC from a pre-120-Ma arc complex to the GCA can be explained by a sinistral transpressional scrape-off.

80 Ma: The GCA continued to move to the NNE under transpressional conditions along the southeastern rifted continental margin of Chortis. From 90 to 81 Ma, the MAC was slowly exhumed to a mid-crustal level and cooled to the K–Ar closure temperature in muscovite of about 350 °C (see Fig. 12). At this time, the E-MORB-like mafic protoliths of the Yayabo Unit were subducted as part of the proto-Caribbean oceanic crust. Slivers of this oceanic crust reached the HP-metamorphic peak and were transformed to the garnet amphibolites of the Yayabo Unit at about 80 Ma (point D in Fig. 13d).

75 Ma: The subduction-accretionary complex of the GCA reached the southern passive continental margin of the Yucatán block. The colliding fore-arc of the GCA interacted with the southern Yucatán continental block in complex manner (e.g., Martens et al., 2012; Solari et al., 2013; Maldonado et al., 2016, 2018). The Early Cretaceous oceanic HP-complex of the eclogite-hosting North Motagua serpentinite mélange of still disputed origin (Brueckner et al., 2009; Pindell & Kennan, 2009; Ratschbacher et al., 2009; Pindell et al., 2012) was thrust onto the southern Yucatán margin (75–70 Ma phengite ages from eclogite: Harlow et al., 2004). The southern margin of Yucatán was overridden by the GCA, reached eclogite-facies conditions and was again rapidly exhumed at 75.5 Ma (Martens et al., 2012; Maldonado et al., 2016) and back-thrust onto the Yucatán margin. The clastic material from the uplifted collisional units was deposited in the Sepur fore-deep basin along the southern edge of the Yucatán block (Pindell et al., 2005); on the Chortis block, activity along the Colon fold-thrust belt stopped (Rogers et al., 2007).

Magmatic activity in the GCA ceased after shallowing of the subduction angle (Hall et al., 2004; Stanek et al., 2009). The uplifted magmatic axis was deeply eroded, as indicated by the development of a Late Campanian peneplain in central Cuba and the formation of extensive carbonate platforms (Iturralde-Vinent, 1996). Extensional basins were filled by volcano-sedimentary material (Iturralde-Vinent, 1994, 2014; Hall et al., 2004). Following the predicted relative plate movement between the GCA and North America (Pindell & Kennan, 2009; Pindell et al., 2012), oblique to the southern Yucatán-Chortis margin, the collision and thrusting could have led to decoupling of the volcanic axis from the fore-arc by sinistral intra-arc tear faults.

By 75 Ma, the Yayabo Unit had been exhumed and juxtaposed with the Mabujina Unit, presumably during movement along the proposed sinistral intra-arc tear fault (point C in Fig. 13c). The formation of the observed ductile shear zone between both units (see Sect. 3.2) should have begun by 78–75 Ma, following the “near”-isothermal exhumation of the Yayabo amphibolites. The sedimentary sequences along the southern Yucatán continental margin started to become involved in the subduction zone of the GCA.

70 Ma: The protoliths of the Gavilanes Unit were subducted to variable depths of 50–80 km and simultaneously reached their HP-metamorphic peak at ~70 Ma. The Yayabo and Mabujina Units were situated at mid-crustal levels in the GCA and cooled below 300 °C. The Gavilanes Unit underwent rapid exhumation and cooling, so that these rocks were amalgamated with the combined Yayabo-Mabujina Units at 300–200 °C, as shown by the observed contacts which indicate brittle to semi-ductile conditions.

The uniform age of 70 Ma for peak HP-metamorphic conditions in the Gavilanes Unit indicates a synchronous metamorphic development of a generally metasedimentary unit. The basic protoliths of both the in-sequence and exotic eclogitic rocks in the Gavilanes Unit must have already been incorporated into the protoliths of the Gavilanes Unit before the peak metamorphic overprint. This protolith sequence of the Gavilanes Unit must have been accreted to the fore-arc of the approaching GCA at least by the time of latest magmatic events at about 75 Ma. The depositional area of the Gavilanes Unit would have been about 800 km south of the present Escambray Mountains along the margin of the southern Yucatán block. The in-sequence metabasic rocks of the Gavilanes Unit presumably correlate with Oxfordian (rift-related?) magmatism at the continental margin with an age of 150–160 Ma. Similar mafic volcanism was mapped in western Cuba as the El Sabalo Formation in the Sierra del Rosario Range (Pszczółkowski, 1994). This age fits the protolith age of eclogite S145 (150–170 Ma: Table 1) and could mark rifting-related magmatism at the southern Yucatán continental margin. However, the geochemical data and Mesozoic protolith ages of zircon of other metabasic rocks in the EC suggest that at least some of the metabasic rocks (S153: 104.7 Ma, Table 1) and serpentinized abyssal peridotites (Hattori & Guillot, 2007) were incorporated into the metasedimentary sequence by tectonic erosion deeper in the subduction channel before reaching maximum HP-conditions.

After 70 Ma, subduction of the Proto-Caribbean oceanic crust of the present Yucatán basin occurred at a shallow angle. The north-facing accretionary front involved both the shelf sediments to the east off Yucatán (future Pitajones Unit and greenschist-facies Unit I) and abyssal Cenomanian to Turonian sedimentary sequences (Placetas and Camajuaní sediments in Fig. 14b; Iturralde-Vinent, 1996; Millán Trujillo, 1997; García-Casco et al., 2008; Stanek et al., 2009). The Placetas and Camajuaní sediments formed – together with the Bahamas shelf sediments – a composite accretionary prism exposed to the north of central Cuba (see Fig. 2 in Iturralde-Vinent et al., 2008; J. Pindell, pers. comm., 2015). Following this train of thought, the concept of Caribeana proposed and discussed by García-Casco et al. (2008), i.e., narrow strips of continental crust extending into the spreading Proto-Caribbean oceanic crust, is a permissible but not indispensable prerequisite for explaining the origin of HP-metamorphic complexes in central Cuba.

56 Ma: The westernmost part of the accretionary prism had reached the continental margin sequences to the east of the Yucatán block. The southern sections of these continental margin sediments were subducted to HP-greenschist-facies conditions (point A in Fig. 13a), stacked under the older metamorphic units (e.g., Despaigne-Díaz et al., 2016, 2017) and transported to the north to form the present Cangre belt, the phyllitic sequences of the Pinos complex, and the tectonically lowermost greenschist-facies units of the EC (Iturralde-Vinent, 1996; Pszczółkowski, 1999; García-Casco et al., 2008; Stanek et al., 2009). The first extensional foreland and piggy-back basins began to form around the exhuming metamorphic complexes and thrust fronts (Iturralde-Vinent, 1994). The EC was exhumed and exposed at the southern margin of the Bahamas platform after the mid-Eocene (45 Ma) because pebbles of HP-metamorphic rocks first appear in stratigraphic sections of the basins surrounding the Escambray by the Upper Eocene (Kantshev et al., 1978).

We have carried out an in-depth review of 40 years of petrological and geochronological research on the nappe stack of the Escambray Complex archived in publications, technical reports, monographs and theses written in Spanish, Russian, German and English. This wealth of data has been augmented with our own new geochronological and geochemical data on critical aspects. As a result, it has become possible to pin down the sources of the protoliths with more confidence, with some originating in the Pacific realm in the Jurassic before the Caribbean even existed, while others formed during the rifting and spreading process accompanying the separation of the North and South American continental plates. Furthermore, it has become possible to better track the individual journeys of these rock sequences along the northern margin of the Caribbean, to monitor events of tectonic juxtaposition, and to define the sequence and timing of final nappe construction in the present Escambray Complex. On the other hand, this detailed knowledge also provides critical constraints that can be used to refine models of regional tectonics in the northern Caribbean margin, where many scattered remnants of the GCA exist.

Guillermo Millán Trujillo (IGP, La Habana) guided us in the field and provided unedited material. Thanks are due to C. Münker, (Münster, now in Köln), who helped to develop the Lu–Hf-method in Münster, K. Mezger (Münster, now in Bern) for helpful discussions, B. Idleman (Lehigh University) for SHRIMP dating. A.D. Renno (Freiberg) provided electron microprobe facilities. We thank L. Franz, an anonymous reviewer and the editors of EJM for numerous detailed comments that helped us to improve our presentation of this difficult project. This work was financially supported by Deutsche Forschungsgemeinschaft (DFG), projects STA 362/3-1 and MA 689/9-1, and the Instituto de Geología y Paleontología (La Habana, Cuba).