Provenance of Eocene–Oligocene sediments in the San Jacinto Fold Belt: Paleogeographic and geodynamic implications for the northern Andes and the southern Caribbean

The Paleogene period in the northern Andes and the onshore Colombian Caribbean basins was characterized by an increase in morphotectonic activity (Van der Hammen, 1961; Restrepo-Moreno, 2009; Villagómez et al., 2011; Spikings et al., 2015; Restrepo-Moreno et al., 2019) that modified regional paleogeography, triggering the production and transport of detrital materials and the generation of synorogenic basins (Ayala et al., 2012; Mora-Bohórquez et al., 2017; Mora et al., 2017). Sedimentary sequences in this region keep a record of uplift, exhumation, erosion, sediment production, and deposition in the northern Andes (Hernández and Jaramillo, 2009; Nie et al., 2012; Bernal-Olaya et al., 2015; Mora et al., 2018; Cardona et al., 2018). Additionally, some of these units have been the focus of research because the high number of oil and gas seeps reveals an active petroleum system (e.g., Barrero et al., 2007). The San Jacinto Fold Belt (SJFB) consists of a NE-trending, relatively low topographic structure (elevation <1000 m) that is located in the Caribbean plains of northwestern Colombia (e.g., the northern termination of the Central and Western Cordilleras). This geographic position coincides with the transition between the northern Andes and the South Caribbean deformed belt (Figs. 1 and 2). Throughout the Cenozoic, this region has been strongly influenced by the convergence of the South American and Caribbean plates and by the collisional history of the Panamá-Chocó block (Pindell and Kennan, 2009; Farris et al., 2011; Montes et al., 2015).

Previous geological studies have proposed different tectonic models for the evolution of the onshore Caribbean region of Colombia. For instance, Caro and Spratt (2003) proposed a paleorift tectonic setting during the Late Cretaceous, which was inverted, along with the Sinú fold belt, during the Cenozoic. In contrast, Rossello et al. (2011) questioned the Pacific origin of the Caribbean plate and suggested a passive continental margin during the Late Cretaceous–Paleogene, which was deformed by thin-skinned faulting and gravity sliding during the middle Miocene. Currently, a widely accepted model posits that the SJFB basin originated as an accretionary prism that formed in the Late Cretaceous and was subsequently deformed during the Oligocene (Duque-Caro, 1972, 1980; Dewey and Pindell, 1985; Toto and Kellogg, 1992; Guzmán et al., 2004; Cardona et al., 2012). This last model implies the accretion of oceanic terranes (Cardona et al., 2012; Mora et al., 2017; Mora-Bohórquez et al., 2017) and the formation of a forearc basin related to the SJFB and the Lower Magdalena Valley (LMV) (Mantilla-Pimiento et al., 2009; Mora et al., 2018; Mora-Bohórquez et al., 2017).

The SJFB and adjacent LMV are composed of sedimentary rocks spanning the Late Cretaceous to Pliocene, although they exhibit regional unconformities (Flinch, 2003; Guzmán et al., 2004; Cardona et al., 2012; Gómez et al., 2015) (Fig. 2). During the Late Cretaceous, sedimentation began with fine-grained marine deposits of the Cansona Formation (Guzmán et al., 2004). Subsequently, the collision of the Caribbean Plateau against the northern South American margin induced the first tectonic event, which is recognized by an early Paleocene unconformity and an increase in coarse-grained terrigenous deposits of the upper Paleocene–lower Eocene San Cayetano Formation (Guzmán et al., 2004; Mora et al., 2017). During the early to middle Eocene, the increase in morphotectonic activity generated another unconformity, which separates the San Cayetano Formation from the Chengue and Toluviejo Formations (Guzmán et al., 2004; Mora et al., 2017). Sedimentation in the Oligocene–Miocene interval is characterized by thick successions of sandstones interlayered with mudrocks and some coal layers that belong to the Ciénaga de Oro Formation (Guzmán et al., 2004; Mora et al., 2018). During the Neogene tectonic inversion of the SJFB and the LMV, the sedimentary environments shifted from shallow marine (Tubará Formation) to continental (Sincelejo Formation) (Guzmán et al., 2004).

Previous investigations on the provenance and paleogeography of the SJFB have been focused on Paleogene rocks. Abreu (2009) suggested that middle Eocene sedimentary rocks from the northern segments of the SJFB (e.g., Pendales Formation conglomerates) formed as a result of erosion of sedimentary sequences such as the Cansona and San Cayetano formations as well as from Cretaceous igneous rocks similar to those of the Aruba and Antioqueño batholiths (Restrepo-Moreno et al., 2009; van der Lelij et al., 2010). The erosional pulse was associated with an Eocene morphotectonic event (pre-Andean orogenic phase sensuVan der Hammen, 1961) reported in many locations of the Colombian Andes (Restrepo-Moreno et al., 2009, 2019; Restrepo-Moreno, 2009; Villagómez et al., 2011; Spikings et al., 2015) and the Caribbean (Rojas-Agramonte et al., 2006). Based on multiple analytical techniques, including petrography, heavy mineral analysis, conglomerate clast count, geochronology, and geochemistry, Cardona et al. (2012) proposed a model for the origin of the Paleocene–Eocene rocks in the SJFB. According to these authors, the metamorphic (schists and gneisses) and ultramafic lithics within the sandstones analyzed were associated with crystalline massifs adjacent to the SJFB, such as those reported in the basement of the LMV and the Central Cordillera, while the volcanic and plutonic components identified in the SJFB were attributed to a Late Cretaceous, intra-oceanic, volcano-magmatic arc that was accreted to the continental margin during the Paleogene. Similar conclusions were advanced by Silva-Arias et al. (2019). Recently, Mora-Bohórquez et al. (2017) provided new borehole data from Paleogene sedimentary rocks in the central and northern segments of the SJFB showing identical zircon U-Pb and ε_Hf signatures to those reported by Cardona et al. (2012), suggesting the existence of crust of oceanic affinity that served as a source of detrital materials to the Paleocene rocks presently exposed in the SJFB. From a paleogeographic standpoint, it is important to note that distal sources from southern segments of the Central and Eastern Cordilleras, with sediments being transported along the proto–Magdalena River, are unfeasible. A series of important paleo-highs such as Magangué and Cicuco (see cross-section B-B′ in Fig. 1) segmented the Magdalena basins, maintaining a separation between the Middle Magdalena Valley (MMV) and the LMV until the Miocene (Caballero et al., 2013). On the other hand, the Cauca River depression, which has possibly been established since Eocene times (Marín-Cerón et al., 2015), may have served as a route to the transport of sediments to the intramontane Amagá Formation and also into more distal positions to the north (e.g., deposits of the SJFB).

This contribution is a study of the central and southern portions of the SJFB (Fig. 2; Supplemental Table S1¹). In this region, we analyzed samples from two wells (ANH–San Antero–1X and ANH–Tierralta–2X wells), which were provided by the Agencia Nacional de Hidrocarburos (ANH), as well as outcrop samples. The sedimentary rocks analyzed include the San Cayetano, Toluviejo, and Ciénaga de Oro formations. In this research, we combine previous paleogeographic information (e.g., Alfaro and Holz, 2014; Bernal-Olaya et al., 2015; Mora et al., 2017; Mora et al., 2018) with new sedimentology, provenance, and biostratigraphy data sets with the aim of generating better-supported paleogeographic models.

The ANH–San Antero–1X well drilled 661 m of cores through the San Cayetano Formation (Fig. 3). It was described in detail in the Colombian national core repository (Litoteca Nacional, Piedecuesta, Colombia) by Universidad de Caldas geologists. The ANH–Tierralta–1X well has a depth of ∼2652 m. This borehole intersected the Toluviejo, San Jacinto, and Ciénaga de Oro formations (Fig. 3) (Mora et al., 2018). Two cored intervals were analyzed for biostratigraphy: the lower core located between depths of 2118–2027 m, and the upper core between 1130 and 731 m (Fig. 3). The lithological description of this well was based on the well logs. The outcrops of the Ciénaga de Oro Formation were described and sampled from newly exposed outcrops in the San Jerónimo anticlinorium, and along the Montería–Planeta Rica road cuts (Fig. 2). The names of the stratigraphic units were based on Guzmán et al. (2004) and Mora-Bohórquez et al. (2017) (Fig. 2).

Our provenance study encompassed petrography (113 samples), heavy mineral analysis (12 samples), zircon U-Pb geochronology (nine samples), and zircon typology (12 samples). These data were used to establish sedimentary source areas and reconstruct paleogeographic environments as well as to refine maximum depositional ages. Petrographic analyses of sandstones were performed using the Gazzi-Dickinson method (Ingersoll et al., 1984). Three hundred (300) points per thin section were counted to evaluate the modal composition and textural parameters such as grain sorting and roundness. The classification scheme by Folk (1954) was implemented. Tectonic provenance classification was carried out on sandstones with matrix-cement and organic-matter content <30% (Dickinson, 1985). The analysis of high-density accessory minerals in the sandstones provides fundamental provenance information (Pettijohn et al., 1973; Morton and Hallsworth, 2007). Heavy minerals in the sand fraction between ∼65 and 250 μm were concentrated using conventional gravimetric and magnetic susceptibility procedures (Mange and Wright, 2007). Mineral identification of up to 300 grains per sample was accomplished using the ribbon method (Mange and Wright, 2007). Heavy minerals were grouped by chemical stability according to Morton (1985) as follows: ultrastable and stable (zircon, tourmaline, rutile, apatite, garnet, and staurolite), moderately stable (epidote, titanite, and clinozoisite), and unstable (muscovite, biotite, calcite, serpentine, pyroxene, hornblende, and spinel).

As part of the mineral separation process, we obtained zircon concentrates for U-Pb geochronology and typology analyses. Detrital zircons were mounted in resin plugs, imaged by cathodoluminescence, and U-Pb dated by the laser ablation–inductively couple plasma–mass spectrometry (LA-ICP-MS) method (Košler and Sylvester, 2003; Gehrels, 2011) using a Nu-Plasma ICP-MS equipped with three ion counters and 12 Faraday detectors. For some samples, >200 crystals were mounted and >100 grains were ablated and dated. In samples with low zircon yields, the total number of recovered grains (60–100) was mounted and dated.

Heavy mineral and U-Pb zircon provenance can be complemented by the typological analysis of zircon populations (Konsa and Puura, 1999; Capuzzo and Bussy, 2000; Anani et al., 2012; Osorio-Granada et al., 2017). The morphology of this ubiquitous, refractory mineral is controlled by temperature and chemical conditions during crystallization (Pupin, 1980), making it useful for tracking the different igneous rock types present in source regions (Dabard et al., 1996; Anani et al., 2012). We documented the relative development of prismatic and pyramidal faces according to the classification scheme by Pupin (1980). Populations of types and subtypes were built, and these represent specific regions in the index T (IT) and index A (IA) space in the typology plot. IT corresponds to the development of prismatic forms and is controlled by crystallization temperature, while IA corresponds to the degree of development of pyramidal forms controlled by an anticorrelation between aluminum and alkalis (Pupin and Turco, 1972; Loi and Dabard, 1997). Morphology and typological characterization of zircons were accomplished on the same samples used for the heavy mineral analyses. Typological classification of each zircon followed the scheme proposed by Pupin (1980). At least 100 zircon grains per sample were classified with observations on general morphology (roundness, elongation, etc.) and chromatic (color, transparency, etc.) parameters. For our typological analysis, zircon crystals were scanned under reflected and transmitted light using a Nikon LV100 Tri-Polar and PET SMZ1500 petrographic stereomicroscope coupled with a high-resolution Nikon DS-F11 camera. Additionally, scanning electron microscopy (SEM) was executed to enhance crystalline face identification using an ESEM-Quanta 250 system under backscatter conditions.

Micropaleontological data, derived from pollen and spores (29 samples), calcareous nannofossils (30 samples), and foraminifera (31 samples), were generated and analyzed to establish a stratigraphic age and for paleoenvironmental interpretations. Slides for calcareous nannofossils were prepared following the decanting method of Flores and Sierro (1997). The slides were analyzed using a Nikon Eclipse LV100 optical microscope at 1000× magnification. Biostratigraphy was based on the Agnini et al. (2007) and Martini (1971) biozones. For pollen and spore analyses, we used the standard technique described by Traverse (2007). The slides were scanned using a high-resolution Nikon Eclipse 80i microscope at 40× and 100× magnification to identify biostratigraphic markers and paleoenvironmental indicators. Biostratigraphic data were compared to the biozones defined for the Cenozoic of northwestern South America by Jaramillo et al. (2011). Foraminifera samples were prepared following the methodology of Thomas and Murney (1985) and analyzed using a high-resolution Nikon PET SMZ1500 optical stereo microscope. Foraminifera biostratigraphic zones were based on Kaminski and Gradstein (2005) and Berggren and Pearson (2005).

All of the petrographic, heavy mineral, zircon typology, and biostratigraphic analyses were performed at the Instituto de Investigaciones en Estratigrafía (IIES), University of Caldas (Manizales, Colombia). Isotopic analyses for zircon U-Pb dating were performed at the Center for Isotope Geoscience, Department of Geological Sciences, University of Florida (Gainesville, Florida, USA). Detailed information about the methodology can be found in Table S1 (footnote 1).

Rocks of the ANH–San Antero–1X well were correlated with the San Cayetano Formation (cf. Guzmán et al., 2004). In this well, the formation can be divided into two intervals (Fig. 3). The lower interval (∼661–110 m) is primarily made of thick beds of gray and dark-green, massive, muddy oligomictic breccias composed of sedimentary clasts (mudrocks and fine-grained sandstones), which are interbedded with medium to thick beds of massive mudrocks and poorly sorted, fine- to coarse-grained sandstones. Ripple lamination, wavy, flaser, lenticular, convolute, parallel, and cross-bedding are observed (Fig. 3). Plant remains and horizontal burrows are common. The upper interval (∼110–0 m) is composed of thick beds of gray and greenish, fine- to coarse-grained, poorly to moderately sorted sandstones, which are massive and laminated (e.g., cross-bedding, ripple lamination, wavy, flaser, lenticular, parallel, and convolute bedding), interbedded with medium beds of gray mudrocks. The sandstones are commonly amalgamated and contain abundant plant remains (Fig. 3).

The lithological description for the ANH–Tierralta–1X well is based on borehole logs and divided into three intervals (Fig. 3). The first interval (∼2134–1945 m; partially cored) was correlated with the Toluviejo Formation (cf. Guzmán et al., 2004). It is mostly composed of laminated dark-gray mudrocks, which are calcareous in some cases, interbedded with skeletal packstones and mudstones with abundant foraminifera. Glauconite and disseminated pyrite are common. The following interval (∼1945–1219 m; cuttings) was correlated with the San Jacinto Formation (cf. Guzmán et al., 2004). It is composed of thick beds of conglomeratic quartz-rich sandstones and sandstones interbedded with gray mudrocks, followed by gray-brown mudrocks interbedded with thin beds of packstones. Foraminifera, glauconite, and disseminated pyrite are locally present. The upper interval (∼1219–914 m; partially cored) was correlated with the Ciénaga de Oro Formation. It is composed of laminated sandstone beds (quartzarenites, subarkoses, and sublitharenites), forming a thickening-upward sequence. Pyrite and glauconite, as well as plant remains, are common. In the upper part of this interval (∼949–914 m; cored), the sequence is dominated by gray packstones with abundant foraminifera and bivalves (Fig. 3).

The studied outcrops of the Ciénaga de Oro Formation are characterized by medium to thick beds of light-yellow quartzarenites, subarkoses, and sublitharenites interbedded with thin beds of gray massive and laminated mudrocks (Figs. 4 and 5). Thick and very thick beds (>1 m) of coal are locally present. The sandstones are arranged in tabular or lens-shaped beds and are structureless or laminated (e.g., cross-bedding, parallel and ripple lamination). Heterolithic beds with wavy, lenticular, and flaser bedding and load structures are present. Gastropods, bivalves, and trace fossils such as Thalassinoides and Ophiomorpha are locally abundant.

We analyzed a total of 84 sandstone samples from the ANH–San Antero–1X well (lower Eocene San Cayetano Formation). Based on the monocrystalline and polycrystalline quartz content, four intervals could be recognized in the San Cayetano Formation (Fig. 3). They are dominated by lithic arkoses, sublitharenites, and subarkoses, with lesser amounts of quartzarenites and feldspathic litharenites (Fig. 5). Texturally, most of the sandstones are fine grained with subangular and moderately sorted clasts. The main detrital constituents correspond to monocrystalline and polycrystalline quartz (42%–74%) followed by rock fragments (3%–41%) and feldspar (0.3%–3.7%). The lithic fragments are dominated by quartzites (0.7%–15.3%), felsic plutonic rocks (0.7%–11%), volcanics (0.3%–10%), schists (0.3%–6.3%), mudrocks (0.3%–5.7%), and sandstones (1%–3%). Other components such as micas (1%–12%), cherts (0%–5%), calcareous foraminifera (0%–2%), glauconite and organic matter (<1%), and accessory minerals (0.3%–5.3%) are also present. In the tectonic discrimination diagrams, these samples fell in the recycled orogen field, i.e., the quartz and transitional fields of Dickinson (1985) (Fig. 5; Table S3 [footnote 1]).

Nineteen (19) sandstone samples from the ANH–Tierralta–2X well and nine outcrop samples from the Ciénaga de Oro Formation were analyzed. Most of the samples are quartzarenites and sublitharenites, while a restricted proportion comprises subarkoses (Figs. 3, 4, and 5). They vary from coarse- to very fine-grained sandstones, with subrounded to subangular grains and moderate sorting. The primary constituents correspond to monocrystalline and polycrystalline quartz (48%–71%) followed by lithics (1%–25%) and feldspar (1%–5%). The lithic fragments include quartzites (0.3%–7.3%), shales (0%–8%), schists (0%–1%), volcanics (0%–0.3%), and acid plutonic rocks (0%–0.1%). Micas (1%–10%), organic matter (5%), cherts (0%–3%), foraminifera (0%–1.5%), glauconite (0%–0.3%), and accessory minerals (0%–1.3%) are also present. Based on these features, the sandstones of the Ciénaga de Oro Formation in both the ANH–Tierralta–2X well and outcrops fall mainly within the recycled orogen field, in the quartzose and interior craton fields of Dickinson (1985) (Fig. 5; Table S3 [footnote 1]).

Heavy minerals (Fig. 6; Table S4 [footnote 1]) from the San Cayetano Formation display the lowest content of zircon (∼8%) and the highest content of apatite (∼24%) and hornblende (∼8%) with respect to the other analyzed samples. Furthermore, the absence of rutile and the presence of garnet are notable. In the Toluviejo Formation, zircon and apatite are in similar proportions (∼18%), and pyroxene (∼11%), rutile (8%), and spinel (∼3%) are present. Another aspect of the heavy mineral fraction in the Toluviejo Formation is the absence of moderately stable mineral phases (epidote, titanite, and clinozoisite) and the abundance of muscovite (∼40%). Finally, in the Ciénaga de Oro Formation, heavy minerals are dominated by zircon (∼30%–40%), rutile is present in all samples, and the content of apatite (5%) decreases relative to the other formations. Moderately stable minerals (epidote, titanite, and clinozoisite) are common in all of the Ciénaga de Oro samples. For unstable phases, the most abundant mineral is muscovite (∼25%) while serpentinite is present in all of the samples. Other identified minerals include garnet, staurolite, and biotite.

We analyzed nine samples from the Paleogene units of the SJFB for zircon U-Pb dating (LA-MC-ICP-MS [MC = multicollector]). Specifically, we dated one lower Eocene sample from the San Cayetano Formation (ANH–San Antero–1X well), two upper Eocene–Oligocene samples from the Toluviejo and Ciénaga de Oro Formations (ANH–Tierralta–2X well), and six Oligocene samples from outcrops of the Ciénaga de Oro Formation (detailed U-Pb analytical results are presented in Table S5 [footnote 1] and Fig. 7). In general, there is not significant variation in the zircon ages analyzed from these three formations. The population with the highest probability is the Late Cretaceous (ca. 70–90 Ma), which occurred in all of the samples analyzed, followed by the Permian–Triassic (230–290 Ma) and Precambrian (ca. 900–1500 Ma) populations. An age population with relatively low probability is represented by the Late Jurassic (ca. 157–162 Ma) in the Ciénaga de Oro and San Cayetano Formations (Fig. 7).

According to the taxonomy of Pupin (1980), 17 zircon subtypes were identified (Figs. 8 and 9; Table S6 [footnote 1]). The main morphotype was S7 (>20%) followed by S2, S3, and S6 (11%–20%). Other typologies were present in lower amounts (<2%, 2%–10%). In general, the zircons have low values of IA (∼355–449) and intermediate to low IT values (∼387–459). Elongation and roundness (Gärtner et al., 2013) indicate that the Precambrian and Permian–Triassic zircons are mainly short to long stalky, and fairly rounded to rounded, while the Cretaceous zircons are poorly and very poorly rounded with a prevalence of columnar and/or prismatic shapes (Fig. 8). The typologies in the San Cayetano Formation show a similar trend, although an apparent equilibrium between the S8 and S3–S5 subtypes seems feasible, thus shifting the populations toward the apex of the distribution of granitic rocks in the typology diagram (i.e., relatively high values of IA and low IT). Samples from the Toluviejo and Ciénaga de Oro Formations depart from the previous trends by showing a clearly bipolar distribution, with two well-contrasted morphological populations that combine the S7–S8 subtypes with the more equant and less complex distribution of the crystal phases characteristic of subtype S20, close to field 7 (tholeiitic continental granites) in the IA-IT diagram (Figs. 9A and 9B). Ciénaga de Oro Formation outcrop samples yield a broad typological distribution centered around the S7 to S8 subtypes and IA and IT values close to 300 and 400, respectively, in the vicinity of field 2 (monzogranites-granodiorites).

A total of 21 samples from the ANH–San Antero–1X well were used for the calcareous nannofossil analysis. Four samples yielded recovery where Sphenolithus radians, Toweius spp., and Sphenolithus orphanknolli at ∼182 m depth were recognized. The co-occurrence of these taxa is associated with early Eocene biozones NP11 and NP13 (Martini, 1971; Agnini et al., 2007). Foraminifera identified in 28 samples included poorly preserved agglutinated benthic species such as Nothia excelsa, Nothia robusta, Psammosphaera irregularis, Rzehakina epigona, Praesphaerammina cf. P. gerochi, Popovia beckmanni, Saccammina complanata, Saccammina cf. S. grzybowskii, Spiroplectammina cf. S. trinitatensis, and Spiroplectammina spectabilis. These species were distributed over a wide range between biozones E8 and E16, which suggest a Late Cretaceous–late Eocene age (Kaminski and Gradstein, 2005). Palynological analyses of 71 samples gave 1531 palynomorphs, among which we identified Bombacacidites gonzalezii, Brevitricolpites macroexinatus, Cyclusphaera scabrata, Foveotriporites hammenii, Mauritiidites franciscoi minutus, Mauritiidites franciscoi pachyexinatus, Spirosyncolpites spiralis, Tetracolporopollenites maculosus, Apiculatasporites aff. cingulatus, Laevigatosporites granulatus, and Polypodiaceoisporites fossulatus. The occurrence of F. hammenii (Fig. 10) between ∼148 m and 429 m suggests an Eocene age (Jaramillo et al., 2011). Thus, the age proposed for the ANH–San Antero–1X well rocks is early Eocene. It is constrained in its lower part (∼633 m) by the presence of S. spiralis, whose first appearance datum is located in the early Eocene (ca. 55.6 Ma; Jaramillo et al., 2011), and in its upper part (∼182 m depth) by the early Eocene calcareous nanofossil assemblage (biozones NP11 and NP13) (Fig. 3).

Samples from the lower core of the ANH–Tierralta–2X well had moderately preserved calcareous nannofossils, which included Reticulofenestra bisecta, Reticulofenestra lockeri, Reticulofenestra reticulata, and Reticulofenestra umbilicus at depths between ∼2071 and 2117 m. This association indicates a biostratigraphic range between the upper NP16 and NP19 biozones of Martini (1971), which is equivalent to the middle–late Eocene (Perch-Nielsen, 1985; Agnini et al., 2007). The foraminifera are characterized by poor recovery and preservation. Some of the species identified at depths between ∼1944 and 2119 m include the Eocene benthic association Gyroidinoides subangulatus, Bulimina jacksonensis, and Cibicidoides micrus (Bolli et al., 1994; Holbourn et al., 2013). Cretaceous reworked species were also identified in this interval (e.g., Gyroidina praeglobosa, Gyroidinoides quadratus, Eggerina subovata, and Praebulimina reussi). Palynological species recognized at ∼2122 m include a poor and badly preserved assemblage composed partly by Cyclusphaera scabrata, Bombacacidites nacimientoensis, and Ranunculacidites operculatus. The latter species (Fig. 10) has its first appearance datum in the late Eocene (ca. 41 Ma; Jaramillo et al., 2011). Thus, the proposed age for this segment is late Eocene based on the nannofossils and the palynological data. As noted previously, the presence of dinoflagellate cysts and calcareous marine microfossils indicates coastal and shallow marine environments, with a considerable supply of terrestrial organic matter.

Additionally, we analyzed seven samples from the upper core of the ANH–Tierralta–2X well between the depths of ∼900 and 1100 m. However, the recovery of calcareous nannofossils is very poor (Table S2 [footnote 1]). We observed that the nannofossil abundance and preservation changed drastically relative to the lower portion of the core. A variation in the species was also identified. The presence of Sphenolithus predistentus, R. bisecta, and Cyclicargolithus abisectus at ∼930 m suggests the NP24 and NP25 late Oligocene biozones (Martini, 1971; Perch-Nielsen, 1985). Palynological analyses at ∼942 m revealed an assemblage with Cicatricosisporites dorogensis and Cricotriporites guianensis (Fig. 10), which indicate a middle Eocene to early Oligocene age (ca. 48–33 Ma) (biozones T06–T08; Jaramillo et al., 2011). Based on these data and the stratigraphic relationship with the lower part of the well, an Oligocene age is proposed for this interval, but it requires better biostratigraphic control as more samples are available (Figs. 2 and 3).

Outcrop samples from the Ciénaga de Oro Formation were also examined for micropaleontology. The foraminifera and calcareous nannofossils recovered were very poorly preserved, which precluded the identification of the formal biozones. Pollen analysis of 20 samples yielded C. dorogensis, C. scabrata, Concavissimisporites fossulatus, Foveotriletes ornatus, Perisyncolporites pokornyi, Foveotricolporites etayoi, Lanagiopollis crassa, Perfotricolpites digitatus, Polypodiisporites usmensis, Psilatricolporites costatus, T. maculosus, Retistephanoporites crassiannulatus, Retitrescolpites irregularis, Striatopollis catatumbus, Ulmoideipites krempii, and Venezuelites distinctus (Fig. 11). The coexistence of C. fossulatus, F. etayoi, and C. dorogensis suggests an Oligocene age (Jaramillo et al., 2011). These data agree with the study performed by Dueñas (1980) in the same area, which reported Magnaperiporites spinosus and Crassiectoapertites columbianus, whose first appearance datums correspond to the early Oligocene (Jaramillo et al., 2011).

Biostratigraphic ages could not be correlated with geochronological U-Pb zircon dates because the minimum zircon ages are ca. 70 Ma. The lack of zircon ages <70 Ma can be explained by the relatively low number of dated zircons (Gehrels, 2014), the granulometry of the dated sedimentary rocks (Ibañez-Mejia et al., 2018), the overwhelming amount of Cretaceous and Permian–Triassic zircons from potential source areas, and/or the lack of sources with younger ages.

In the lower interval (∼661–110 m) of the ANH–San Antero–1X well (San Cayetano Formation), the presence of massive, muddy oligomictic breccias indicates gravity-flow processes (e.g., debris flows) associated with escarpments or slope failures that can be triggered by earthquakes or by rapid sediment deposition (Talling et al., 2013). The clast composition indicates the erosion of older sedimentary rocks, mainly fine-grained sandstones and mudrocks. Dinoflagellates and calcareous microfossils indicate marine environments influenced by rivers, as confirmed by the presence of plant remains, pollen, and spores. The presence of pollen grains associated with plants that currently live in humid tropical zones (e.g., Bombacacidites, Mauritiidites, Tetracolporopollenites), together with pteridophyte spores and abundant fungal remains, suggests warm humid conditions. In the upper interval (∼110–0 m), the dominance of amalgamated sandstones and matrix-supported conglomerates, as well as the plant remains and marine and terrestrial palynomorphs, suggest nearshore environments. Some of the poorly preserved benthic foraminiferal assemblages, which suggest a bathyal environment (Kaminski and Gradstein, 2005; Berggren and Pearson, 2005), could be related to a reworking of older sedimentary units.

For the Toluviejo and San Jacinto Formations in ANH–Tierralta–2X well, the dominance of mudrocks, some of them calcareous, as well as marine and terrestrial microfossils suggests a low-energy marine environment with continental input. The progressive increase in sandstones towards the upper part (∼1219–949 m, Ciénaga de Oro Formation) imply nearshore conditions, probably associated with a delta front. The uppermost part of the studied sequence (∼949–914 m), dominated by gray packstones with abundant foraminifera and bivalves, indicates shallow marine conditions.

In the Ciénaga de Oro Formation (outcrop samples), the presence of coal beds, channel sandstones, marine microfossils, pollen and spores, mangrove pollen (Lanagiopollis crassa, related to extant Pelliciera rhizophorae), and marine palynomorphs in conjunction with the local abundance of mollusk shells and ichnofossils (e.g., Ophiomorpha) point to deltaic and shallow marine conditions (see also Dueñas [1983] and Guzmán et al. [2004]). Additionally, the presence of Tetracolporopollenites sp. (Sapotaceae), Retitrescolpites irregularis (Amanoa, Leguminosae), Striatopollis catatumbus (Leguminosae), Retitricolpites simplex (Euphorbiaceae), Bombacacidites sp. (Bombacaceae), Perisyncolporites pokornyi (Malpighiaceae), Mauriitia spp., and abundant fungal and pteridophyte spores suggests a lowland tropical humid forest. For instance, modern Mauriitia occupies floodplains and gallery forests (Hofmann, 2002) in the Llanos Orientales basin (Marchant et al., 2002).

In our study, the nature of the sandstones was related to a recycled orogen in the transitional and quartz subfields (Fig. 5). In general, the recycled orogen character is indicative of mixed source areas located in proximal and/or moderately distal positions (Dickinson, 1985). This model is supported by the poor to moderate sorting and low degree of roundness of the sediments. Another important factor to consider is the climate. As mentioned above, the palynological evidence indicates wet tropical conditions, which agrees with paleogeographic and paleoclimate models for the region (Erlich et al., 2003; Jaramillo et al., 2010). In this sense, petrographic evidence, such as the occurrence of feldspar, the content of apatite, and the presence of unstable mafic and ultramafic minerals, suggests that sediments for San Cayetano Formation came from igneous and metamorphic massifs located close to the depositional site.

Paleocene strata recorded, for the first time, syntectonic deposition and erosion of nearby source areas (Ayala et al., 2012). This means that the sedimentary cover above those basement blocks was very thin and was rapidly eroded. Additionally, the identification of Cretaceous reworked microfossils in the Toluviejo Formation can be associated with the erosion of preexisting sedimentary deposits of the SJFB (e.g., Cansona Formation and clasts of the San Cayetano Formation conglomerates; Fig. 2) and the Western Cordillera (e.g., Penderisco Formation; Álvarez and González, 1978). The petrographic evidence of this work differs from that of previous studies (e.g., Cardona et al., 2012; Mora et al., 2017) where the upper Paleocene–lower Oligocene sandstones are mainly lithic arkoses related to transitional and dissected arc fields. This compositional variation is the result of a combination of factors such as lateral variation of sedimentary environments, local variations of the source area, and variations of distance from the source area (Fig. 12).

For the Ciénaga de Oro Formation, the quartz contents increased while plutonic and metamorphic lithics decreased with respect to San Cayetano Formation, thus reflecting changes in the composition of the source area. The presence of tourmaline, muscovite, rutile, clinozoisite, staurolite, hornblende, biotite, and garnet is indicative of the erosion of granitoids and low- to medium-grade metamorphic rocks, such as those reported in the Central Cordillera and/or crystalline basement of the LMV. In contrast, highly unstable minerals such as spinel and serpentine, which are typical of ultramafic source rocks, were derived from the basement of the SJFB (e.g., Planeta Rica and Cerro Matoso peridotites) or from some segments of the Western Cordillera and the structural complexes linked to the Romeral fault system (e.g., Arquía and Quebradagrande complexes; Maya and González, 1995) (Figs. 1 and 12).

Petrographic results display a modal variation that progresses from the relatively high feldspar content in the lower Eocene (lithic arkoses, subarkoses, and sublitharenites) of the San Cayetano Formation (ANH–San Antero–1X well) to a quartz-rich sediment (quartzarenite, subarkose) in the Oligocene Ciénaga de Oro Formation (ANH–Tierralta–2X well and in the outcrop samples; see Fig. 5). This shift in compositional maturity may be due to increased weathering and erosion associated with climatic variations (Smith et al., 2008; Wan et al., 2009). Alternatively, since these two boreholes are separated more than 120 km, the changes in composition and/or geographic location of the source areas can explain shifts in compositional maturity as quartz-rich sources became more prevalent.

Zircon U-Pb age populations, especially the Late Cretaceous (ca. 70–90 Ma) population, are associated with plutonic-volcanic rocks (Fig. 12) located in the LMV basement (e.g., Bonga-1, Coral-9, and Cicuco-22 plutons; see Silva et al., 2017; Mora-Bohórquez et al., 2017), northern Central Cordillera (Antioqueño, Ovejas, and Sabanalarga batholiths, and Altavista stock; Leal-Mejía, 2011; Villagómez et al., 2011; Restrepo-Moreno, 2009), and Western Cordillera (Mistrató pluton; Gómez et al., 2015). The Permian–Triassic age population (230–236 Ma) can be correlated with metamorphic units present in the Central Cordillera (e.g., Cajamarca and Puquí complexes; Maya and González, 1995; Restrepo et al., 2011; Villagómez and Spikings, 2013) and the metamorphic and plutonic rocks found in the LMV basement (Silva et al., 2017; Mora-Bohórquez et al., 2017; Vinasco et al., 2006; Montes et al., 2010). Precambrian-population (ca. 900–1500 Ma) zircons are related to Grenvillian-age rocks that crop out in the northern Central Cordillera (e.g., San Lucas Gneiss; Leal-Mejía, 2011). Furthermore, they can be related to reworked zircons coming from Cretaceous–Cenozoic sedimentary rocks in either the Central Cordillera or Western Cordillera. Finally, the Late Jurassic population (ca. 157–162 Ma) can be linked to magmatic units of the Central Cordillera (e.g., Segovia batholith; Leal-Mejía, 2011) and common volcanic rocks located between the Central Cordillera and the Santander massif (e.g., Norean Formation; Leal-Mejía, 2011). Jurassic rocks in the Sierra Nevada de Santa Marta (e.g., Aracataca, Pueblo Bello, and Atanquéz batholiths; Tschanz et al., 1974) show older Jurassic zircon ages than those identified in the analyzed rocks and can be discarded as a potential source.

Finally, our zircon typological analysis indicates that the igneous source rocks correspond mostly to calc-alkaline monzogranites and granodiorites (Fig. 9), which can be associated with some plutonic rocks described in the Central Cordillera and LMV basement (Silva et al., 2017; Mora-Bohórquez et al., 2017; Vinasco et al., 2006; Montes et al., 2010).

Recent investigations on the structure of LMV basement rocks (Silva et al., 2017; Mora-Bohórquez et al., 2017) reported the presence of Cretaceous intrusive igneous units with positive ε_Hf values (e.g., Bonga-1 pluton) intruding the western basement of the LMV. The nature of this basement is unknown. However, the highly positive values of ε_Hf suggest no continental crust contamination, which is an argument that was used to suggest that the basement of the LMV was be composed of mafic crust related to the Quebradagrande Complex (Fig. 12; Mora-Bohórquez et al., 2017). In contrast, reported Triassic and Permian igneous and metamorphic rocks suggest the continuation of the continental crust of the Northern Andean block toward the eastern part of the LMV basement (Montes et al., 2010; Mora-Bohórquez et al., 2017; Silva et al., 2017; Vinasco et al., 2006).

The integration of provenance data available from the SJFB (Abreu, 2009; Cardona et al., 2012; Mora et al., 2017; this work) and the LMV (Montes et al., 2010; Mora-Bohórquez et al., 2017; Silva et al., 2017; Vinasco et al., 2006) allows us to propose an early Eocene–late Oligocene paleogeographic model. We suggest that the main detrital sources correspond to the LMV and the Central Cordillera basements, while smaller contributions of detrital material were sourced from crystalline massifs of the Western Cordillera. We do not consider the Permian–Triassic and Late Cretaceous rocks of the Sierra Nevada de Santa Marta as potential sources for the SJFB or the LMV because Piraquive et al. (2018), using thermochronological data, demonstrated that these rocks had not yet been exhumed during Oligocene–Miocene times.

The model we propose is corroborated by the isotopic and geochronological correlation (ε_Hf and U-Pb values) between the plutonic zircons of the Cretaceous and Permian–Triassic realms of the LMV basement and the Central Cordillera relative to the Paleogene sandstones and conglomerate clasts in the SJFB basin (Mora et al., 2017; Mora-Bohórquez et al., 2017; Cardona et al., 2012). Uplift and exhumation pulses during the Eocene–Oligocene have been identified for the northern portion of the Central and Western cordilleras (Restrepo-Moreno et al., 2009; Restrepo-Moreno, 2009; Cochrane et al., 2014) as well as the basement of the LMV (Mora-Bohórquez et al., 2017; Silva et al., 2017). These upheavals help explain the production of detrital materials carried to the Caribbean by rivers (proto–Cauca River and shorter drainages), and the area occupied by the modern SJFB was covered mainly by terrigenous fluvial, coastal, and marine sediments. Finally, further sources (Eastern Cordillera and Santander Massif) are discarded because regional basement-involved uplifts (e.g., Cáchira high in the northern MMV) did not allow fluvial connection between the MMV and LMV basins until the middle Miocene (Reyes et al., 2004; Mora et al., 2018; Caballero et al., 2013; Horton et al., 2015).

During the Oligocene, the sediments were deposited in deltaic and shallow marine environments, such as those represented by the Ciénaga de Oro Formation (Duque-Caro, 1972; Guzmán et al., 2004). The predominance of Cretaceous zircon populations suggests significant contributions from LMV and/or the Central Cordillera basement rocks (Fig. 13).

Further research involving more specific and high-resolution data sets (e.g., stratigraphic, geochemical, geochronology, isotopic) is necessary to fully characterize and differentiate basement rocks in the northern Andes, and thus, to increase the accuracy of provenance models. Major uncertainties remain concerning the provenance and paleogeographic models described here due to the following reasons:

1. There are no reported zircon ε_Hf isotopic data for the Magangué magmatic arc (84–70 Ma) proposed by Silva et al. (2017). This precludes comparisons against available data from the northern Central Cordillera (e.g., Antioqueño batholith), which exhibit slightly negative to slightly positive ε_Hf values (e.g., Restrepo-Moreno et al., 2009; Restrepo-Moreno, 2009; Cochrane et al., 2014) relative to the positive ε_Hf signatures of LMV basement granitic bodies, such as those obtained from samples of the Bonga-1X well (Mora-Bohórquez et al., 2017). In the transition from the Central to the Western Cordillera, there is a variety of Cretaceous granitoids with contrasting compositions for which ε_Hf data are nonexistent. Positive ε_Hf values are reported for Cretaceous zircons in granitic clasts from Paleogene conglomerates and detrital zircons in the SJFB (Cardona et al., 2012; Mora et al., 2017).

2. The lack of data from wells reaching the litho-structural blocks of oceanic affinity intruded by the Cretaceous plutons generates uncertainty about the extent of the Quebradagrande Complex below the LMV (Mora-Bohórquez et al., 2017).

The biostratigraphic data obtained for the studied borehole and outcrop samples in the central and southern portion of the SJFB allowed us to constrain the age of the sampled units, varying from lower Eocene to upper Oligocene. The sedimentary rocks of the San Cayetano Formation, in the ANH–San Antero–1X well, accumulated during the early Eocene, while the Toluviejo, San Jacinto, and Ciénaga de Oro Formations, in the studied interval of the ANH–Tierralta–2X well, accumulated during the late Eocene–late Oligocene. The outcrop samples from the Ciénaga de Oro Formation indicate an Oligocene age. These sediments were deposited mostly in tropical, humid, coastal and shallow marine environments.

The studied rocks can be related to a recycled orogen. The composition of the sandstones changed over time. While the San Cayetano Formation (lower Eocene) shows a wider modal composition (Fig. 5), from lithic arkoses to quartzarenites, the sandstones of the Ciénaga de Oro Formation (Oligocene) are mainly quartzarenites. Heavy mineral assemblages indicate variation between the three analyzed stratigraphic units. Changes here reported may be related to climatic shifts, the variations in the composition of the source area and/or change in its geographic location, and transport distance, because the two wells and outcrop samples are at different localities (>120 km apart).

Zircon U-Pb geochronology and zircon typologies do not show such marked variations compared to petrography and heavy minerals. These data indicate that the sediments of the SJFB basin came from the erosion of felsic igneous, low-grade metamorphic, and mafic to ultramafic rocks. Our lines of evidence suggest that that the source of the studied sedimentary rocks can be related to the basements of the LMV, Central Cordillera, and, in a lesser percentage, the Western Cordillera.

This project was conducted with financial support from Agencia Nacional de Hidrocarburos (ANH, Colombia) project “Estratigrafía del Cretácico Superior–Paleógeno del sector Sinú–San Jacinto, Caribe Colombiano: Aporte al conocimiento de su evolución geológica y sistemas petrolíferos”. Special thanks are due to Jairo Alonso Osorio, José William Garzon, Carlos Rodríguez, Edgar Bueno, and José Fernando Osorno (ANH) for their technical and administrative support. The Instituto de Investigaciones en Estratigrafía, Universidad de Caldas (Colombia), and the Departamento de Geociencias y Medio Ambiente, Universidad Nacional de Colombia (Medellín), provided important logistic and institutional support. We also express our gratitude to the Center for Isotope Geoscience, University of Florida (Gainesville), and for their support during sample preparation and analyses. Finally, we greatly appreciate the comments and suggestions from Dr. Cesar Witt (Université de Lille 1, Villeneuve d’Ascq, France) and Dr. German Bayona (Corporación Geológica ARES, Bogotá, Colombia), whose thorough revision helped to improve the quality of this manuscript.