This paper reports the discovery of glacial deposits of likely Siderian–Rhyacian age (2.58–2.06 Ga) in South America (Carajás Basin, Brazil), thereby expanding the potential reach of Paleoproterozoic glaciations to the Amazonian craton for the first time. Glacially derived diamictites are stacked within a hitherto unrecognized ∼600-m-thick siliciclastic succession, here named the Serra Sul Formation. Well-preserved textures, with evidence of glaciotectonism and ice rafting, indicate deposition in a coastal subglacial to glacial-fed submarine fan system, in which the immediately underlying units (banded iron formation and volcanic rock) were the main source and bedrock. The Serra Sul diamictite may be correlated with any of the known Paleoproterozoic glaciations, or with none of them.


Paleoproterozoic glaciations (ca. 2.45–2.22 Ga; Rasmussen et al., 2013) are recorded in almost every continent, but they have not yet been reported in South America. These glaciations, also called the Huronian Glaciation Event (Tang and Chen, 2013), occurred at the time of the Great Oxidation Event (GOE), and it is suggested that they were triggered by the supercontinent breakups at the end of the Archean (Aspler and Chiarenzelli, 1998; Bleeker, 2003; Strand, 2012). While some studies have argued that they may represent an early snowball Earth episode (e.g., Evans et al., 1997; Hoffman, 2013), others have suggested that there may have been a series of more typical glaciations occurring at different times and in different places (e.g., Young, 2014).

In the Carajás Basin, Amazonian craton (Brazil), intervals of matrix-supported breccia were previously reported as a particular rock of the Serra Sul succession, and considered a Neoarchean magnetite-rich breccia of the lowermost banded iron formation (BIF) unit (Cabral et al., 2013). Strangely, these rocks were never described in the Neoarchean succession (Grão-Pará Group), which is generally documented as a set of ca. 2.75 Ga BIF and mafic to felsic volcanic rock (Trendall et al., 1998) affected by Cu-Au mineralization at ca. 2.57 Ga (Tallarico et al., 2005). This was partially due to the lack of an accurate stratigraphic framework for the Carajás Basin that has caused several units to be neglected, misinterpreted, or even overestimated.

The initial investigation of structures and textures from drill cores revealed that the strata of the Serra Sul succession are inverted. This is generally observed on the borders of the Carajás Basin that were deformed by regional folds associated with the latest tectonic event that affected the area, the Transamazonian cycle at ca. 2.0 Ga (Machado et al., 1991). In fact, the breccia intervals occur within a siliciclastic succession hitherto unrecognized in the basin, hereafter formally named the Serra Sul Formation. This formation unconformably overlies the Neoarchean units and unconformably underlies the siliciclastic deposits of the Águas Claras Formation of ca. 2.06 Ga (Mougeot et al., 1996), suggesting Paleoproterozoic (2.58–2.06 Ga) age constraints (Fig. 1).

In this study, based on a detailed description of well-preserved delicate primary textures throughout the Serra Sul Formation (see Items DR1 and DR2 in the GSA Data Repository1), we attempted to resolve this issue by producing sedimentological and stratigraphic evidence that helps to identify the origin of these breccia. The results provide important evidence of the paleoenvironmental conditions on the Amazonian craton in the early Paleoproterozoic.


A sedimentological and stratigraphic study based on detailed descriptions of drill cores from throughout the Serra Sul Formation was undertaken. To cover as much of the lateral stratigraphic information as possible, detailed measured sections were taken throughout an ∼600-m-thick succession from drill cores located in three different areas of the Carajás Basin (a, b, and c in Fig. 1; see Items DR3, DR4, and DR5). For each of these areas, a simplified sedimentary log was created and organized in the north-south section toward the basin depocenter (Fig. 2). Small samples (up to 10 cm long) were collected from the drill cores, and polished thin sections were made in order to verify the microstructures and compositions of the diamictite matrices and clasts. Scanning electron microscope (SEM) analyses were performed to investigate the microtextures on quartz grain surfaces. For this analysis, small pieces of core (n = 2) were gradually broken and quartz grains (n = 100) were randomly selected, then fixed in an electrically conductive carbon double-stick tape, coated with gold, and imaged using secondary electrons at the Geological Survey of Brazil (Belém).


Foliated to Massive Diamictite Facies Association

The association of foliated to massive diamictite facies comprises diamictite interbedded with thin beds of black shale (as much as 3 m thick). There are uninterrupted intervals of diamictite, 150 m thick, in the stratigraphic section (Fig. 2, log a; see Item DR1). The diamictite has a foliated matrix that tends toward a massive aspect. The BIF and mafic to felsic volcanic-subvolcanic rock clasts vary in size from pebbles to boulders, and vary in shape from angular to well rounded. Faceted, flattened, rotated, boudinaged, sheared, and lens-shaped clasts are widespread (Figs. 3A and 3B; see Item DR6). Fractures occur preferentially along the borders of clasts and rarely cross-cut them; they are even found perpendicular to the foliation and filled with quartz and dolomite. In some clasts, the inflection of the foliation resulted in the development of asymmetrical and symmetrical pressure shadows. The fine-grained and dark-colored matrix is composed of micrometer-scale fragments (up to 100 μm) of quartz and magnetite, and the alignment of these two minerals defines the foliation (Fig. 3C). Isolated bullet-shaped and boulder-sized clasts occur in black shale beds, sandwiched by diamictite intervals (Fig. 3D).

Rhythmite Facies Association

Another facies association found within the Serra Sul Formation is composed of only rhythmite that constitutes a monotonous succession as much as 300 m thick at the base of the stratigraphic section (Fig. 2, logs b and c; see Items DR2 and DR7). The rhythmite is composed of intercalated normally graded, fine-grained sandstone and mudstone stacked into fining-upward cycles as much as 10 cm thick; the lithological contact is generally marked by soft-sediment deformation structures and rip-up clasts.

Conglomerate-Sandstone-Rhythmite-Diamictite Facies Association

Conglomerate and sandstone interbedded multiple times with rhythmite and diamictite compose the final facies association observed in this study (Fig. 2, log b; see Item DR2). Normally graded to massive conglomerate and sandstone are stacked into 0.30–30-m-thick fining-upward cycles. In the stratigraphic section, they reach uninterrupted intervals of as much as 150 m thick. In sandstone, the grains are mainly composed of iron chert. Conglomerate clasts are angular to well-rounded fragments of BIF, iron chert, and volcanic rock ranging from pebble to cobble size, with cross-laminae and planar laminae weakly developed. Clast- to matrix-supported conglomerate has a sandy to granular matrix (Fig. 3E; see Item DR8). Rhythmite and diamictite occur interbedded with one another. The massive diamictite reaches as much as 10 m thick and consists of a mud-rich matrix in which pebble- to cobble-sized and angular to well-rounded BIF, iron chert, and volcanic rock clasts occur scattered. Isolated pebble-sized clasts occur with high dip angles and deformed underlying laminae (Fig. 3F), and coarse-grained clasts are mound forming in concave-down beds (Fig. 3G). Faceted clasts, including elongated pentagonal ones, also occur scattered within the muddy matrix (Figs. 3H and 3I; see Item DR9). Fractures, steps, and grooves occur on quartz grain surfaces (Fig. 3J; see Item DR10).


Diamictite Textures: Tectonic versus Sedimentary

The observation that foliated diamictite is restricted to certain beds and that it occurs interlayered with massive diamictite and black shale beds indicates an intrastratal deformation at the time of deposition, and the coeval occurrence of deformed and preserved clasts in the same interval suggests a highly heterogeneous deformational setting (e.g., Menzies, 2012; Busfield and Le Heron, 2013, 2018). The observation that fractures are restricted to the borders of the clasts and rarely cutting them, are preferentially oriented perpendicular to foliation, and are filled with quartz and dolomite that replaced the original matrix minerals indicates that these structures were formed during sedimentation and subsequently infilled by secondary minerals.

Similar to the interpretation of Cabral et al. (2013), our observations indicate that the pervasive deformation fabric present in some diamictite intervals is derived from synsedimentary deformation, certainly linked to subglacial mechanisms, rather than tectonic deformations. The observation in drill cores that the succession is only tilted and that no shear bands are present strengthens this hypothesis, as does the fact that no evidence of metamorphism was recorded in the thin section. Ultimately, the occurrence of well-preserved delicate microtextures such as steps, striations, and fractures on the surfaces of quartz sand grains confirms that the Serra Sul diamictite is unmetamorphosed or was metamorphosed to a very low grade (cf. Machado et al., 1991; Pinheiro and Holdsworth, 1997).

Data Interpretation and Model

The occurrence of foliated to massive diamictite with faceted, fractured, flattened, sheared, boudinaged, and rotated clasts with pressure shadows suggests that these rocks were formed in subglacial environments and recorded primary ice-contact sedimentation (i.e., tillite), where high strain rates led to progressive deformation (Menzies et al., 2006; Busfield and Le Heron, 2013, 2018; Le Heron et al., 2017). The quartz- and magnetite-rich matrices of these diamictites suggest that they were derived from a rock flour produced by intense abrasion of the bedrock. The observation that volcanic rock and BIF clasts penetrate, puncture, and deform the underlying laminae of mudstone suggests that these clasts are dropstones. The occurrence of beds composed of coarse-grained clast mounds surrounded by fine-grained sediments further suggests that these beds are dumpstone (i.e., rock formed from the melt-out of debris contained in floating dirt-laden icebergs; Thomas and Connell, 1985).

The large number of faceted clasts embedded in mudstone beds, some of which have elongated pentagonal shapes and high dip angles, indicates that they were dropped over a muddy substrate and associated with rapid burial by fine sediments without subsequent reworking. Similarly, the fairly common occurrence of faceted clasts sandwiched by mudstone indicates a hydrodynamically incongruent sedimentary setting (Le Heron, 2015; Le Heron et al., 2017). The grooves observed on the surfaces of the quartz grains are straight, parallel, and V-shaped, suggesting that they are striations produced by glacial erosion (Whalley and Krinsley, 1974; Mahaney et al., 1996; Immonen, 2013). On the other hand, the rhythmic centimeter-scale intercalation between fine-grained sandstone with normal gradation and mudstone suggests deposition by low-density turbidity currents, and the meter-scale intercalations of massive to normally graded conglomerate and sandstone suggests deposition by the types of high-density turbidity currents associated with submarine fan systems (Pickering et al., 2015).

A glacially influenced coastal to marine system is envisaged as the depositional environment that accommodated the sediments of the Serra Sul Formation (Fig. 4). This system was developed over and supplied by bedrock mainly composed of Neoarchean BIF and volcanic rock belonging to the Grão-Pará Group. This hypothesis is strongly supported by the observation that quartz and magnetite compose the matrices of subglacial diamictite, and by the fact that BIF, iron chert, and volcanic rock are the main constituents of the clasts. The surface that separates the Serra Sul deposits from the Neoarchean strata represents a great unconformity that may be used for correlation inside and outside the basin. Ice overloading on the continent, combined with the lateral expansion of glaciers, promoted high rates of deformation of the bedrock. This was likely strengthened by the presence of a previously formed regolith cover that greatly favored the glaciotectonic deformation and had a composition similar to that of the bedrock. Short-lived marine transgressions into the continent associated with the retreat of glaciers may explain the occurrence of black shale beds repeatedly interbedded with subglacial diamictite at the base of coastal deposits.

In the marine setting, submarine fan systems were washed and supplied by glacial meltwater, which promoted the resedimentation of glaciogenic debris in deep waters. The simultaneous release of ice-rafted debris (IRD) from icebergs in distal waters and deposition of fine-grained particles generated massive diamictite beds with dropstones and dump structures. The alternation of sedimentation by submarine fan systems and ice rafting explains the multiple intercalations of IRD-bearing diamictite and sandy and/or conglomeratic facies that is observed throughout the succession (Fig. 2, log b). Although gravity-flow deposits constitute a greater part of the succession, IRD-bearing intervals of glaciogenic diamictite are sharply distinct. The interbedding of these deposits testifies to subaqueous sedimentation within the basin. The similar composition of clasts in the subglacial (Fig. 2, log a) and debris-flow diamictite (Fig. 2, logs b and c) suggests a genetic correlation between these environments and that the debris-flow deposits are glaciogenic (e.g., Le Heron et al., 2017). The thick beds of rhythmite at the base of the succession (Fig. 2, logs b and c) suggest a prolonged sedimentation by low-density turbidity currents in the deepest parts of the basin, although the absence of IRD-bearing intervals in these deposits does not necessarily indicate ice-free conditions (e.g., Le Heron, 2015).

Ultimately, the depositional system hypothesized for the Serra Sul Formation is similar to those interpreted for Paleoproterozoic glaciations worldwide (e.g., Miall, 1985; Chen et al., 2019). The Serra Sul succession represents the first large-scale siliciclastic sedimentation after the predominantly chemical sedimentation that characterized the Neoarchean of the Carajás Basin. Compared to the Neoproterozoic Sturtian glaciation, in which iron formation was deposited simultaneously with diamictite in sub–ice shelf settings (Lechte and Wallace, 2016), the deposition of diamictite in the Serra Sul system was completely decoupled from the deposition of the BIF. Although the expressive rift-related volcanism of the Carajás Basin represented by the Grão-Pará Group (Olszewski et al., 1989; Martins et al., 2017) may be related to the breakup of the southern supercontinent at the end of the Archean, and may have also been one of the triggers of Serra Sul glaciation, this volcanic event predates by nearly 200 m.y. (2.75–2.58 Ga) the onset of this cooling episode.


A core-based sedimentological and stratigraphic investigation in the Carajás Basin (Brazil) revealed the occurrence of likely Siderian–Rhyacian (2.58–2.06 Ga) glacial strata within a succession hitherto unrecognized, newly named the Serra Sul Formation. We suggest that Amazonia underwent glaciation in the early Paleoproterozoic, which may or may not be correlated with any of the known Paleoproterozoic glaciations. The Serra Sul diamictite represents the first reported occurrence of Paleoproterozoic glacial deposits in South America, expanding the potential reach of glaciations of this time period to the Amazonian craton for the first time.


We thank Sérgio Huhn, Fernando Matos, and Luiz Costa of Vale S.A. (Parauapebas, Brazil) for making the drill cores available for the research; the Geological Survey of Brazil (Belém, Brazil) and the PROPESP/UFPA for supporting this work; Werner Truckenbrodt for his constructive comments on an earlier version of the manuscript; and Roberto Araújo Filho and Alexandre Ribeiro for fieldwork assistance. We are also very grateful to Daniel Le Heron, Paul Hoffman, Kari Strand, and an anonymous reviewer for reviewing our manuscript; and to James Schmitt for his editorial work.

1GSA Data Repository item 2019397, complete set of core-based sedimentary logs, borehole locations, table of lithofacies, and supplementary figures, is available online at http://www.geosociety.org/datarepository/2019/, or on request from editing@geosociety.org.
Gold Open Access: This paper is published under the terms of the CC-BY license.