Abstract

Interpretation of recent, high-quality seismic data in the Gulf of Mexico (GOM) has led to competing hypotheses regarding the basin’s rift to drift transition. Some studies suggest a fault-controlled mechanism that ultimately results in mantle exhumation prior to seafloor spreading. Others suggest voluminous magmatic intrusion accommodates the terminal extension phase and results in the extrusion of volcanic seaward dipping reflectors (SDRs). Whereas it has been generally accepted that the plate motions between the rift and drift phases of the GOM are nearly perpendicular to each other, it has not been greatly discussed if the breakup mechanism plays a role in accommodating the transition in plate motion. We have developed a plate kinematic and crustal architecture hypothesis to address the transition from rift to drift in the GOM. We support the proposition of a fault-controlled breakup mechanism, in which slip on a detachment between the crust and mantle may have exhumed the mantle. However, we stress that this mechanism is not exclusive of synrift magmatism, though it does imply that SDRs observed in the GOM are not in this case indicative of a volcanic massif separating attenuated continental and normal oceanic crust. We support our hypothesis through a geometrically realistic 2D potential field model, which includes a magnetic seafloor spreading model constrained by recent published seismic data and analog rock properties. The 2D model suggests that magnetic anomalies near the continent-ocean transition may be related to removal of the lower continental crust during a phase of hyperextension prior to breakup, ending in mantle exhumation. The kinematics of breakup, derived from recent satellite gravity data and constrained by our spreading model and the global plate circuit, suggests that this phase of hyperextension accommodated the change in plate motion direction and a diachronous breakup across the GOM.

Introduction

Early attempts to reconstruct the supercontinent Pangea presented “plate tectonicists” with a problem: Bringing the Americas into contact with Africa by closing the Atlantic requires southern Mexico and South America to occupy the same space. In consequence, for as long as workers have been attempting to improve our reconstructions of Pangea, we also have had to reckon with the tectonic history of the Gulf of Mexico (GOM) and the Caribbean.

Dietz and Holden (1970) address the Mexico-South America space issue by eliminating the isthmus of Panama, and placing the Yucatan, Honduras, and the Greater Antilles into the space left between North America, Africa, and South America. Although the solution was based on a rather selective conservation of area and geometric convenience, aspects of it persist in modern models, perhaps most importantly including an intra-American origin for the Yucatan peninsula (GOM).

Pindell and Dewey (1982) advance the concept of reconstructing the Yucatan into the GOM. They create additional space for South America by segmenting southern Mexico into blocks based on rock type, age, and structural trends and move these blocks to restore hypothesized offsets. They then rotate the Yucatan block into the GOM, suggesting that a spreading center and accretion of oceanic crust could accommodate the motion. A Pacific origin for the Caribbean Plate also was discussed, which eliminated some overlaps in the Pangea reconstruction. These two fundamental paradigms, rotation of the Yucatan into the GOM and a Pacific origin of the Caribbean, have remained the foundation for an abundance of studies and reconstructions seeking to refine our understanding of the GOM-Caribbean region for the past several decades (e.g., Marton and Buffler, 1994; Bird et al., 2005; Hudec et al., 2013; Nguyen and Mann, 2016; Pindell et al., 2016).

The “rotation” kinematics of the GOM drift phase, in which the Yucatan rotates counterclockwise away from North America accommodated by seafloor spreading, is for all intents and purposes the accepted model. Unequivocal support for the “rotation” phase came most recently in estimates of the gravitational potential from satellite altimetry (Sandwell et al., 2014) that clearly demonstrate the presence of oceanic fracture zones, broadly consistent with “rotational” kinematic models. An earlier “rift phase,” in which the Yucatan is “translated” away from North America, also is postulated in many models (Pindell, 1985; Kneller and Johnson, 2011; Hudec et al., 2013). This “two-phase” opening of the GOM is unusual, however, in that the seafloor spreading direction is essentially perpendicular to the rift direction postulated by most authors.

The structures accommodating the terminal phase of extension and continental breakup in the GOM, and the transition in direction between rift and drift phases, are not well understood. In the eastern GOM, two stark alternatives have been suggested: (1) An “outer marginal detachment” (OMD), separating the continental crust and mantle, allowed extension to be accommodated by slip of the crust along the mantle surface prior to breakup. This hypothesis has been termed outer marginal collapse by Pindell et al. (2014). (2) Alternatively, it has been proposed that the terminal phase of synrift extension could be accommodated by a magmatic process, in which large volumes of igneous material are added to the crust to accommodate extension with accompanying extrusive volcanics that form seaward-dipping reflectors (SDRs) (Imbert and Philippe, 2005; Eddy et al., 2014). These two mechanisms generally are thought to represent end members of the rifting process, with mantle exhumation taking place on “cold” margins with little to no active plume influence or excess mantle heat (Iberia-Newfoundland, Sibuet et al., 2007), whereas magmatic breakups occur where an active plume is participating in rifting or mantle temperatures are elevated due to recent past tectonic history (e.g., south Atlantic Margins, Jackson et al., 2000).

This study seeks to test and refine the prevailing tectonic hypotheses that describe the geologic history of the GOM. Primarily, we seek an internally consistent set of plate rotations, limit of oceanic crust (LOC), and potential fields model of the crust that differentiates between oceanic crust, normal extended continental crust, and the limits and nature of the transitional terrane formed by the terminal phase of extension. The potential fields model of the crust is geometrically constrained by recent published refraction and reflection profiles (Eddy et al., 2014; Pindell et al., 2014) and constrained in terms of its magnetization by well-established magnetic properties for oceanic crust and exhumed mantle (Gee and Kent, 2007). The plate kinematic model is determined from analysis of potential-field data (Bankey et al., 2002; Sandwell et al., 2014) in the context of a global plate model (Zahirovic et al., 2015).

Data and methods

Gravity and derivation of a pole of rotation

A variety of geophysical data has been used to derive poles of rotation for the Yucatan in the past. Pindell et al. (2016) use a set of aeromagnetic anomalies to update rotations for the Yucatan, resulting in a wandering pole during seafloor spreading. Hall and Najmuddin (1994) also interpret magnetic anomalies in the GOM to derive a pole of rotation, which was later reinforced by kinematic reconstructions of gravity data by Bird et al. (2005). Hudec et al. (2013) cite the geometry of the Tamaulipas margin as a primary constraint on their rotations for the Yucatan. Nguyen and Mann (2016) are the only authors, to our knowledge, to explicitly invoke the geometry of the fracture zones apparent in the gravity data of Sandwell et al. (2014) to derive a pole for the Yucatan. However, the kinematic criterion of fit or method of pole derivation is not discussed. Here, we explicitly demonstrate the pole of rotation implied by the geometry of the GOM fracture zones, using classic methods of tectonic analysis.

The fracture zone anomalies in the gravity data of Sandwell et al. (2014) can be seen in the free air anomaly (FAA) and, perhaps better, in the vertical gradient of the FAA (Figure 1, the vertical gravity gradient [VGG]). To derive a pole of rotation from these features, we follow the basic algorithm of fracture zone analysis laid out by Cox and Hart (1986) on a data set extracted using a semiautomatic method of picking fracture zones described here. We begin by interpreting a set of lines, wherein each line is collinear with one of three fracture zone anomalies (Figure 2a and 2b). We then use an automatic grid picking algorithm that generates points at local grid peaks (Blakely and Simpson, 1986) on the VGG. Using a buffer operation, we select the automatically picked grid peaks within 10 km of our fracture zone interpretation and extract them from the set of all the grid peaks. We are thus left with a set of semiautomatically picked grid cells that define the fracture zone locations. We discard our initial guiding interpretation, and we pass a moving window (approximately 100 km) over the fracture zone anomalies to provide a manageable subset of points for tectonic analysis (Figure 2c). Each point on the fracture zone and its adjacent point define a short great circle segment that represents a portion of the fracture zone. We calculate the local azimuth of these great circle segments, and then we plot the perpendicular great circle arcs. Finally, we calculate the intersection of all such arcs and the centroid of all the intersections to find a pole of rotation derived explicitly from the gravity data (Figure 3). The entirety of the analysis is performed in spherical coordinates.

Kinematics from the global plate circuit

The timing of rotation of the Yucatan about our derived pole is based on considerations from a modern and robust reference global plate model (Zahirovic et al., 2015). To estimate the age at which seafloor spreading ceases in the GOM, we consider the interactions between the Yucatan Plate (YUC) and the stable South American Craton (SAM). We treat the kinematics of SAM, the North American Craton (NAM), and the African Craton (AFR) as described by Zahirovic et al. (2015) as boundary conditions for this analysis.

Initially, we fix the Yucatan plate to North America. We then rotate SAM toward NAM and YUC according to the global plate circuit. Once the plate polygons of YUC and SAM begin to overlap, either YUC must retrodeform, rotate (by undoing seafloor spreading), or both to accommodate further reconstruction of SAM. We estimate this contact to occur at approximately 154 Ma. This age is sensitive to how the plate boundaries of YUC and SAM are interpreted, which is especially challenging for SAM, which has been modified by Cenozoic emplacement of the Caribbean Plate. For that reason, we treat this as an initial “candidate age” for the breakup between SAM and YUC. We also observe that appealing only to deformation of YUC at an age close to 154 Ma is not realistic because pushing SAM farther back without rotating the Yucatan would imply significantly greater amounts of extension in the southern YUC compared to the northern YUC. This sort of differential stretching is not observed in crustal thickness estimates for the region (Laske et al., 2013). Thus, we assume that some component of rotation (seafloor spreading in the GOM) also is necessary at this age. Refining the timing of seafloor spreading is discussed in the next section.

Magnetic anomalies and crustal modeling

We derive a geometrically realistic spreading model for oceanic crust in the GOM, intended to constrain the LOC and characterize the crust at the ocean-continent transition. Our model is constructed as a 2D forward model, with multiple profile segments that form an arc parallel to the spreading direction based on our derived pole location (an Euler pole flow line, Figure 4). The model must be built in this orientation to cross perpendicular to the presumptive magnetic chrons in the oceanic crust.

To ensure that our model is geometrically realistic, we chose a model location that crosses the recently published seismic refraction (GUMBO 3) and reflection (Fugro 533) profiles published by Eddy et al. (2014). Our model intersects GUMBO 3 and Fugro 533 at the location of a distinct set of en echelon anomalies (EEA, Figure 4). At that location, the sediment thickness, top of basement, and top of normal density mantle are tied to their corresponding velocity layer representations in GUMBO 3, within 10%, which is the often-stated approximate depth uncertainty of refraction velocity models (e.g., Christensen and Mooney, 1995). Away from this crossing point, the geometry of the crust is projected along strike from GUMBO 3 and Fugro 533 onto our transect. The sedimentary section is a simple depth-density function based on regional and publicly available well data, hung from the water bottom in 1 km bins. Salt thickness also is projected from Fugro 533 onto our transect; however, at this regional scale, it has a minimal impact on the gravity field. Qualitatively, uncertainty in the actual geometry of the crust along our profile increases away from its crossing point with the seismic results of Eddy et al. (2014). Deviations in our model from the results of Eddy et al. (2014) were made to satisfy the gravity and magnetic data while recognizing that our interpretation of these along-strike changes is subject to a greater degree of nonuniqueness than near the crossing point.

The oceanic crust is modeled with a simple uniform thickness (8 km) based on the typical thickness observed on GUMBO 3 and with a flat Moho as observed on Fugro 533. A topographic low of approximately 2 km is included at the extinct spreading ridge (ESR) as observed on Fugro 533. The magnetization of oceanic crust has been discussed by many authors and is reviewed by Gee and Kent (2007). The ocean’s characteristic igneous layers, pillow basalts, dikes, and gabbros, have somewhat distinct magnetic characteristics. Most importantly, these layers generally are expected to be dominated by remanent, rather than induced, magnetizations with intensities ranging between 1 and approximately 25 A/m. The pillow basalts tend to have the most intense magnetization, but also lower curie temperatures (200°C–300°C) due to titanium replacement in magnetite. Dikes and gabbros, by contrast, contain nearly pure magnetite with curie temperatures closer to 580°C. Thus, even under thick sediment loads, these layers are expected to retain strong remanent magnetizations that dominate the magnetic anomaly, which can cause characteristic “seafloor stripes,” if other conditions also are met. For our model, we assume a remanent direction based on the apparent polar wander path for North America (Torsvik et al., 2012) calculated at the location of the extinct spreading center (29N, 86W), at an age of approximately 154 Ma (normal polarity: Dec = −19.7, Inc = 16.7; reversed polarity: Dec = 160.3, Inc = −16.7). We also limit the remanent magnetization intensity of the layers to the geometric mean reported by Gee and Kent (2007) to allow for some intensity decay, potentially due to burial and heating, over geologic time. Our initial model derives chrons from the geomagnetic polarity timescale spanning an age range of 145–170 Ma (Gradstein et al., 2012), to cover a wide possible range of oceanic ages. To determine the “best” magnetic model for the oceanic crust, we translate and dilate the spreading portion of the model to satisfy the anomalies observed in the GOM. This process is akin to adjusting the spreading rate.

The magnetic data used here are composed of track line measurements, stored by NOAA’s National Center for Environmental Information, and gridded and released by Bankey et al. (2002). Because our spreading model is constructed along an arbitrary line, a grid extraction of the anomalies is required. In its available state, the gridded data contain multiple along-line artifacts that make it difficult to use for magnetic modeling and tectonic analysis. We releveled the data following a simple algorithm: (1) calculating a coarse grid from all of the line data, (2) calculating the misfit from each line and the coarse grid on a line-by-line basis, (3) computing a low-level polynomial function to the misfit between each line and the coarse grid, adding each line’s misfit function to the line data (thus reducing the corrugation between lines, while preserving the mid- to high-frequency signal), (4) regridding the data and identifying tares (potentially time transient anomalies) and removing them from the database, and (5) regridding the cleaned database. As a result, many along-line artifacts that previously illustrated the ship tracks from port of departure and across the GOM have been removed.

Results

Euler pole for oceanic spreading in the GOM

The Euler pole derived from our analysis of the gravity anomaly data is located at 21.6240, −82.1671 (Figure 3). We calculate flow lines using our pole and the wandering pole of Pindell et al. (2016) (Figure 5). In a qualitative sense, flow lines using both poles reproduce the trajectory of the fracture zone anomalies with one obvious caveat. At the northwest end of the fracture zone anomaly, the flow line from this study deviates from the interpreted fracture zone, and at the southwest (SW) end of the same anomaly the flow line from Pindell et al. (2016) deviates. Inspection of the westernmost fracture zone anomaly clearly shows it to be subparallel to the adjacent two eastern anomalies at its tips. This could be due to a poorly organized initial spreading center in the western GOM. Apart from this, both sets of flow lines produce a satisfying fit to the data. The flow lines generated from the rotations of Pindell et al. (2016) demonstrate that the change in the pole location evoked by those authors produces a subtle kink in the expected fracture zone geometry (Figure 5b). This kink is not observed in the gravity anomaly, but it may be too subtle to have a measurable effect using the satellite data, if it exists. For the eastern two fracture zone anomalies, the kink causes their flowlines to deviate significantly from the observed anomalies at the tips. This kink also implies a reorganization of the spreading center mid-drift, or nonideal spreading from the same spreading center. Because the moving pole is not necessary to fit the data, and it would lead to an overly complicated spreading model, we adopt our single pole for the purpose of generating a magnetic spreading model. A simple set of GPlates files (polygons and rotations) are included as supplementary files, and the complete rotation parameters are summarized in Table 1.

Magnetic model and implications for drift kinematics

Eddy et al. (2014) show that the ESR in the GOM is a significant topographic low. The magnetic anomaly associated with the ESR, however, is a significant magnetic high. Assuming a source limited to oceanic crust, and magnetization intensities based on analogs (Gee and Kent, 2007), a relatively long normal polarity interval at and around the ESR is necessary to match the magnetic high. In the mid-Jurassic, several “long” normal-polarity chrons occur starting near 154 Ma and younger. Recalling from the “Data and methods” section that our initial spreading model contained all of the chrons from 145 to 170 Ma, we observed our best initial fit to the data with chron M24n (153.6 Ma) at the ESR and a full spreading rate of 2.4 cm/yr, which would classify the GOM midocean ridge as a slow spreading center (Dick et al., 2003). This fit was achieved assuming uniform magnetization intensity in the different chrons and varying only the direction of magnetization between normal and reversed. Other models were attempted in which the younger long-polarity chrons (e.g., M22An, 151.9 Ma) were placed at the spreading ridge, but the presence of the long older chrons on the oceanic crust landward of the ESR produced significant anomalies that are not observed in the data. To complete the 2D transect, we place chron M24n on either side of the ESR, mirroring the reversal pattern to the end of the line. Once this fit was obtained, we allowed the magnetization intensities within each chron to be varied within the distribution of intensities reported by Gee and Kent (2007) to refine the final fit (Figure 4).

Our initial spreading model intentionally stretched well inboard of previous estimates of the LOC, underlying a significant positive magnetic anomaly, with en echelon and conjugate equivalents that we call here the “en echelon anomaly” (EEA, Figures 1 and 4). Whereas our spreading model provides a good fit to the data outboard of this anomaly, it does not provide a good fit to the EEA. The age of oceanic crust that would underlie the EEA contains multiple short polarity intervals that do not produce a magnetic high, even if the magnetizations of the intervals are selectively enhanced to favor one polarity.

Seismic profiles in the northeast GOM have been used to demonstrate the presence of a crustal scale detachment, or “outer marginal detachment” (OMD) (Pindell et al., 2014) at the transition between continental and oceanic crust. The detachment is suggested by Pindell et al. (2014) to expose lithospheric mantle, creating an “exhumed mantle high” between the attenuated continental crust and oceanic crust. To test this idea, we model the EEA using analog magnetic properties for serpentinized mantle (Gee and Kent, 2007; parameters listed in the table accompanying Figure 4) with a depth of serpentinization of 3–4 km thick. This body provides a good fit to the data, but limitations of the interpretation are presented in the “Discussion” section. The boundary between our serpentinized mantle body and oceanic crust coincides with the outboard edge of the EEA anomaly. Thus, we interpret the LOC to coincide with the outboard edge of this anomaly and its en echelon equivalents.

To complete our model, we make a simple assumption of a magnetically strong lower continental crust and modestly magnetic upper continental crust, in keeping with typically inferred mafic and felsic compositions, respectively. A magnetically weak upper crust overlying exhumed mantle, without the lower crust in between, is in keeping with the outer marginal collapse conceptual model of Pindell et al. (2014) and satisfies the large magnetic low that is continuous along-strike landward of the EEA. Geodynamically, this interpretation also is consistent with the discussion presented in Nirrengarten et al. (2016) in which outcrop and seismic examples of mantle exhumation can be explained by removal of the ductile lower crust and the brittle behavior of the upper crust and mantle, with the mantle surface acting as a mechanical detachment.

Inboard of the ocean-continent transition, we model a relative crustal thick underlying the Southern Plateau, identified in Fugro 533 (Eddy et al., 2014; our Figure 4). This portion of the model has a depth to basement and depth to Moho consistent with the reflection and refraction results published by Eddy et al. (2014). We take care to note that we are not inferring that the area underlying the Southern Plateau has undergone any kind of thickening. By contrast, the Southern Plateau is likely a remnant basement high, in between two regions of crust that have undergone relatively greater extension. The outboard region eventually was stretched until breakup occurred. The inboard region is known as the Apalachicola Basin and is modestly well imaged by Fugro 533 (Eddy et al., 2014). We have modeled the Apalachicola Basin as an area of attenuated crust, with a shallower Moho (between 3 and 5 km) than one would infer from the velocity model of GUMBO 3. There are a few reasons why our crustal geometry here is reasonable for the purposes of this regional-scale model: (1) The velocity structure underneath the Apalachicola Basin is only partly constrained by raypaths at Moho depths on GUMBO 3 (on its outboard edge). (2) Fugro 533 demonstrates Moho reflection segments that are mostly shallower than the Moho interpretation from the velocity model on GUMBO 3; thus, there is some uncertainty in the depth to Moho from the seismic results themselves. (3) The relevant section of our line is along-strike from GUMBO 3, and some variation in crustal geometry and Moho depth is not unreasonable to provide a regional-scale gravity fit. Our model of the along-strike equivalent of the Apalachicola Basin is intended only as a regional-scale approximation that satisfies the gravity data and is loosely constrained by seismic results in the region.

Limit of oceanic crust and reconstruction

Our estimate for the distribution of oceanic crust is based on a combination of rigid kinematic “rules” and our magnetic spreading model. To map the oceanic crust, we seed Euler pole flow lines at the ends of spreading ridge segments and calculate the position of the fracture zone resulting from our rigid kinematics every 1 Myr. We define the LOC where these flow lines intersect the outboard edge of the EEA (Figure 6). This strategy, which assumes symmetric spreading, allows us to estimate the LOC under the Sigsbee escarpment, where salt obscures the seismic and gravity constraints on the structure of the crust (Figures 1, 6, and 7). Our magnetic spreading model suggests a youngest age for oceanic crust sometime during chron M24n (152.9–153.6 Ma) and a full spreading rate of roughly 2.4 cm/yr. For numerical convenience in the plate model, we have rounded the age of youngest oceanic crust up at the ESR to 154 Ma and calculated the age of older oceanic crust accordingly. Based on our interpreted LOC, the GOM proceeded to oceanic spreading diachronously, with the western “compartment” achieving breakup ca. 166 Ma, the central compartment ca. 164 Ma, and the eastern compartment ca. 160 Ma. Due to rounding, the actual breakup ages may be slightly younger (within 1 Myr).

Our segmentation of the GOM into three compartments (Figure 6) is based on the potential applicability of our 2D model to map view interpretation of the magnetic anomalies along strike. The central compartment contains anomalies consistent with the profile modeled in Figure 4. The Gulf Coast Magnetic Anomaly (GCMA) and anomalies north of it can be interpreted as a function of the geometry of the deformed crust, produced during necking of the upper and lower crust during rifting. In this context, the GCMA, at least at long wavelengths (>100 km), is the result of a thick block of continental crust, including a magnetic lower crust, coincident with the Southern Plateau High as observed in the seismic results of Eddy et al. (2014). Outboard of this, the lower crust is removed, leaving only a weak magnetic upper crust in an outer trough coinciding with a broad magnetic low. This block of upper crust is suggested to rest directly on the mantle, which will be increasingly serpentinized basinward, as it is unroofed. This serpentinized mantle is the proposed source of the EEA. Outboard of this, the GOM magnetic anomalies are explained by our spreading model.

In the western compartment, the distinct pattern of anomalies discussed in the central compartment is not as clear. Potential along-strike equivalents of the magnetic lows and highs representing removal of the lower crust and exhumed mantle are observed, but they are not as sharp, and other anomalies obscure the pattern. Published refraction and reflection data in the western compartment (e.g., Van Avendonk et al., 2015) also are somewhat indistinct compared to the results of Eddy et al. (2014) in the central compartment, perhaps owing to the presence of a complex salt canopy. Therefore, we provide only an estimate of the LOC in the western compartment. An analysis of crustal structures in the western compartment is beyond the scope of this study.

Finally, the eastern compartment contains clear equivalents to the anomalies discussed in the central compartment, but the symmetry of them across the GOM is not present. The EEA equivalent is offset on the North American side and has no Mexican equivalent. It also dies out along strike leaving only a deep magnetic low. This pattern may suggest a transition from a symmetric breakup mechanism in the central compartment to an asymmetric compartment in the east. Reduced extension rates eastward toward the Euler pole may have prevented sufficient extension from fully necking out the lower crust and isolating a block of upper crust and exhumed mantle at the eastern end of this compartment.

We reconstruct the magnetic anomaly grid using our rigid plate model to evaluate the fit of conjugate anomalies (Figures 811). EEA highs reconstruct precisely adjacent to each other in the northeast GOM (central compartment), supporting the kinematics of our model (Figure 11). Potential conjugate matches in the west GOM also are inferred, although, as discussed, the anomalies are less pronounced in that region. Long linear magnetic anomalies on the northern Yucatan Peninsula reconstruct to be collinear with NE-trending anomalies on the North American craton, with a slight offset near the beginning of seafloor spreading (Figure 10). At the beginning of the rotation phase, the anomalies are collinear and slightly overlapping, suggesting these anomalies likely represent deformation due primarily to the “translation” phase of the GOM rift (169 Ma, Figure 11).

Discussion

Crust at the continent-ocean transition in the northeast GOM

Our approach to potential fields modeling of the crust in the northeast GOM is of course, not the only possible solution. It does have the advantage, however, of leveraging thickness and depth of the source layer from refraction and reflection data, a known location of the spreading center, and following the expected behavior of oceanic crust from analog studies. The magnetic intensities and remanent directions are all consistent with rock analogs, rock ages, kinematic history (spreading rate), and paleolatitudes discussed in this paper. Our approach has an additional advantage: In attempting to push the oceanic magnetization toward shore, we can observe where the assumptions consistent with oceanic crust no longer satisfy the data, and thus a different crustal type, and the LOC, is implied.

The LOC proposed in our study is the conclusion that we draw with the highest relative confidence. Our estimate of the LOC is farther outboard than that of Eddy et al. (2014; and our Figure 4), who place the LOC at the edge of the GCMA. Whereas those authors did compare a magnetization profile with their transect, they did not provide a potential fields model linking their interpreted crustal architecture to the anomalies. We show that the anomalies observed along our profile are consistent only with typical oceanic crust recording a seafloor spreading remanence up to the outboard edge of the EEA. The EEA requires a wide magnetized body not consistent with oceanic crust of probable Middle Jurassic Age (many short chrons whose anomalies would effectively cancel each other out). Although our LOC is a geometric idealization based on the location of the EEA and Euler flow lines, the reconstruction of the anomalies along these boundaries demonstrates a good qualitative match of the conjugate margins (Figures 811).

Our model is in contrast to some other studies that propose an asymmetric spreading history for the GOM. Filina et al. (2020), for example, illustrate a series of gravity anomalies (lows), in a high-pass Bouguer residual of the satellite gravity anomaly, that lie between the known ESR and northern GOM margin. Given an LOC interpretation that differs from our own (with more oceanic crust on the northern side), these anomalies could explain an asymmetric distribution of oceanic crust about the ESR. The contrast between our model and that of Filina et al. (2020) stems primarily from the interpretation of the LOC. Our interpretation of the magnetic anomalies suggests symmetric oceanic crust about the ESR. Filina et al. (2020) demonstrate a reasonable interpretation of the gravity that suggests an asymmetric distribution of oceanic crust. In order to reconcile these differences, potential-fields modeling constrained by seismic data spanning the U.S. and Mexico sides of the GOM may be necessary. For the time being, there is room for valid alternative hypotheses.

The nature of the crust at the continent ocean transition (COT) is an interpretation that we provide in light of our potential-fields modeling results, the seismic results of previous authors (Eddy et al., 2014; Pindell et al., 2014) and geodynamic considerations that we will discuss in this section (Jackson et al., 2000; Pérez-Gussinyé and Reston, 2001; Nirrengarten et al., 2016; Buck, 2017).

In our view, the current geophysics is best understood with a geologic model in which the terminal phase of extension along the northeast GOM is accommodated by fault mechanics that results in hyperextension of the upper crust along a detachment separating the crust and mantle, ultimately resulting in exhumation of the mantle. We prefer this model to a model of extension accommodated by magmatic intrusion of the entire crust, with coeval subaerial flood basalts resulting in SDRs, namely, because it explains a broader subset of observations made in the region, which we will discuss below. Pindell et al. (2014) term this fault-controlled hyperextension model the “outer marginal collapse.” We will adopt this term here to describe our iteration of this hypothesis, although it may not be strictly what the originating authors had in mind. We should also preview that synrift volcanism and outer marginal collapse are not mutually exclusive, neither in the original formulation of Pindell et al. (2014) or in this study. In fact, we will demonstrate that synrift volcanics in the GOM may constitute reflectors that dip toward the sea in our preferred model, without implying that crustal extension was accommodated by intrusion of significant volumes of igneous material added to the lower and upper crust.

It is worth noting that the GOM is not the only margin where a debate ensues between magma-rich and magma-starved end members for crustal hyperextension. For example, in the Campos Basin of Brazil, Norton et al. (2016) propose a magma-rich transition to oceanic crust based on the presence of SDRs adjacent to a region where some seismic results could be interpreted to suggest the absence of synrift stratigraphy and the beginning of oceanic crust. On the other hand, Zalán et al. (2011) and Brune et al. (2014) provide seismic evidence and thermomechanical models (respectively) that support the presence of exhumed mantle in the Campos Basin. Zhang et al. (2017) eventually publish high-quality 3D seismic data that convincingly argued that the SDRs observed by Norton et al. (2016) are more likely synrift stratigraphy forming a rollover anticline against a major rift fault, outboard of which additional synrift stratigraphy could be imaged; that is, SDRs in this case did not uniquely identify the transition to oceanic crust. At the same time, it is well known that the Campos Basin contains abundant synrift volcanics (Rangel and Carminatti, 2000). Thus, even potentially “magma-starved” margins that achieve hyperextension through crustal slip on the mantle may contain intrusive and extrusive volcanics. It is this hybrid hypothesis that we will develop further, following a discussion of the current observations and models.

Eddy et al. (2014) provide several arguments supporting the interpretation of a magma-rich breakup mechanism in the GOM that should be considered. The first to consider is the observation of steeply seaward dipping reflectors near the COT along Fugro 533. The reflections they identify as SDRs underlie a region where strong positive and negative magnetic anomalies occur. The reflections themselves appear to lack either the stacked or progressively shallowing dipping morphology of SDRs observed in the south Atlantic or Australian margins as reviewed by Buck (2017), but they do coincide with step changes in the Jurassic unconformity level, suggesting they may be structural rather than intrusive/extrusive in nature. By contrast, Eddy et al. (2014) observe much steeper, stacked reflections with increasingly shallower dips well inboard of the LOC in the Apalachicola Basin, that also underlie a distinct and profound magnetic high (100s of nT). These are more consistent with typical SDR models, and the large magnetic high over the basement low they occupy attests to their potential volcanic nature. Indeed Liu et al. (2019) model these reflections in the Apalachicola Basin with high magnetic susceptibilities and densities, providing a satisfying fit to the potential fields data. Thus, it seems likely that the inboard Apalachicola Basin could have a significant synrift volcanic fill, but these reflections are not located at the transition to oceanic crust. A lack of strong magnetic anomalies in the along-strike equivalent of the Apalachicola Basin intersected by our model may also suggest these basalts have limited aerial extent.

Eddy et al. (2014) produce a velocity model along GUMBO 3 that suggests a “high-velocity” lower crust (VP 6.5–7.5 km/s) near the OCT, which could be consistent with magmatic underplating. However, these velocities also are not outside the range observed in lower continental crust in extended terranes or shields and platforms near depths of 30 km (Christensen and Mooney, 1995). Ocean-bottom seismic records along GUMBO 3 record arrivals that are interpreted as PmP (Moho) arrivals, which are not typically interpreted on the Iberian margin in the region of known exhumed mantle (e.g., Dean et al., 2000; Davy et al., 2016). However, Minshull et al. (2014) do discuss wide-angle reflections occurring at Moho depths in the region of exhumed mantle on the Iberian Margin and importantly demonstrate the potential for alternative velocity models to those previously published by Dean et al. (2000). Neither does Fugro 533 provide a clearly migrated Moho from the oceanic crust through the transition zone back toward the continent in the GOM COT. Finally, Liu et al. (2019) model the crust in this transitional region from the potential fields’ perspective. Their model, which uses high susceptibilities and densities for interpreted SDRs and lower crustal intrusions, produces a satisfying fit to the gravity data but a calculated magnetic high that exceeds the observed magnetic anomalies. Given these uncertainties and room for alternative models, it seems too soon to rule out the possibility of a mechanical detachment on the crust-mantle boundary in the GOM.

On the other hand, Pindell et al. (2014) observe landward-dipping reflections in the northeast GOM linking the basement at the OCT to the subcontinental Moho. This family of reflections has been dubbed the “outer marginal detachment” (OMD), and similar dipping reflections are observed on the Iberian margin (e.g., Minshull et al., 2014). Although, on the Iberian margin, the changing polarity of dipping reflections in the region of serpentinized mantle may be suggestive of a more complex pattern of deformation than has yet been proposed for the GOM. Pindell et al. (2014) suggest that the OMD represents a crust-mantle contact, along which the crust slipped as the margin proceeded to break up. Should sufficient slip occur along this boundary, the upper mantle could eventually be exposed. Pérez-Gussinyé and Reston (2001) show that serpentinization of the mantle is a likely necessity to allow slip along the crust-mantle boundary, because serpentinization lowers the coefficient of friction sufficiently for the contact to act as a mechanical detachment. Nirrengarten et al. (2016) also show that a mechanical detachment on top of the mantle explains observations of transitional crust in the eastern Swiss Alps in addition to the Iberian margin where exhumed mantle is known. Thus, the proposition of Pindell et al. (2014), that a mechanical detachment separates the crust and mantle in the GOM seems plausible and would require some degree of serpentinization for slip to occur. As a consequence of serpentinization, this “exhumed” mantle also would have a lower density, seismic velocity, and stronger magnetization than the adjacent unserpentinized mantle.

Our 2D model is built to demonstrate the validity of the exhumed mantle hypothesis from the potential fields’ perspective. Referring to Figure 4, inboard, necking of continental crust produces magnetic anomalies associated with lateral changes in crustal thickness caused by rifting. Specifically, the GCMA is explained by a relatively thick crustal block referred to in the profile of Eddy et al. (2014) as the Southern Plateau. The relative basement high and thickness of the magnetic lower crust contribute to the anomaly. A broad magnetic low outboard of the GCMA is consistent with removal of the more magnetic lower crust as occurs in some thermomechanical models prior to hyperextension (e.g., Huismans and Beaumont, 2011) and, as suggested for analog examples by Nirrengarten et al. (2016), leaving the upper crust to contact the mantle. Gradual replacement of the lower crust with partially serpentinized mantle along an OMD explains the increasing gravity and could be consistent with seismic velocities in the same area reported by Eddy et al. (2014). The EEA is explained by the presence, thickness, and relative elevation (high) of the exhumed mantle body, which we interpret to constitute the basement step up imaged in Fugro 533.

Kinematic models for crustal hyperextension in the GOM

We have represented our working understanding of the spectrum of potential mechanisms to achieve hyperextended crust in the GOM in a series of conceptual forward kinematic models (Figure 12). These models represent an attempt to integrate the studies discussed in this paper, which contain field and seismic observations and conceptual, physical, and thermomechanical models into a simple kinematic framework using fault-controlled simple shear in the upper crust and pure shear ductile deformation in the lower crust. These models are thus imperfect. However, some features of these models may be useful advancements of the outer marginal collapse concept. The magma-starved and magma-rich end members begin with simple shear-distributed deformation of the upper crust and ductile necking of the lower crust (Figure 12a). In the magma-starved end member, continued thinning of the crust near the rift axis and removal of the ductile lower crust allows exhumation, hydration, and cooling of the upper mantle, leading to serpentinization and the establishment of the OMD (Figure 12b). The inferred geometry of the contact between the necked lower crust and the underlying mantle mimics the geometry of the OMD (Moho). Continued extension detaches on the OMD, which now is a contact with serpentinized mantle near the rift axis. Away from the rift axis where the mantle is not serpentinized and acting as a detachment, shear must become distributed in the upper mantle. The terminal phase of extension “drags” the upper crust down the OMD, creating significant subsidence at the rift axis. The magma-rich end member accommodates continued extension through magmatic addition to the lower crust, with extrusive equivalents filling topography in the center of the rift basin (Figure 12c). Counter regional faults at the center of the basin cause the igneous fill to dip toward the proto-ocean.

It seems improbable that crustal hyperextension and mantle exhumation would not lead to some component of decompression melting in any rift environment. Certainty in the updip GOM synrift section, volcanics have been observed (reviewed in part by MacRae and Watkins, 1995) and are inferred in the nearby Apalachicola Basin from the results of Eddy et al. (2014) and Liu et al. (2019). Thus, our preferred conceptual kinematic model is a hybrid of the magma-rich and magma-poor end members (Figure 12d and 12e). In this model, the geometry of the rift is controlled by the kinematics of faulting along the OMD, but a component of extrusive magmatism that filled the earlier rift topography is deformed toward the center of the rift basin, causing, interestingly enough, SDRs. Though our kinematic models do not include sediment accumulation, these “SDRs” could have clastic and volcanic components, not unlike the SDRs of the Campos Basin discussed by Norton et al. (2016). Thus, these SDRs do not imply significant accommodation of extension through magmatic addition to the crust.

Drift and breakup kinematics of the Gulf of Mexico

Our rigid reconstruction of the GOM restores conjugate anomalies at the margins of the oceanic crust (Figure 10). It also suggests diachroneity of the breakup in the GOM, with the western GOM proceeding to seafloor spreading first (approximately 166 Ma) and the eastern GOM proceeding to seafloor spreading subsequently, with the final breakup occurring potentially as late as 160 Ma. This implies that, during the “rotation phase” of GOM tectonics, seafloor spreading and continental extension were occurring along strike from each other. Thus, the transition between rifting directions associated with the “translation” phase, and rifting directions associated with the “rotation” phase, occur prior to breakup for a large part of the GOM.

To account for a mid-synrift change in the extension direction, we again rely on a discussion of the interaction of the South American and Yucatan plates. The spreading rate that we have derived for the GOM oceanic crust is consistent with the motion of SAM away from the Yucatan; that is, it ensures that SAM slowly pulls away from the Yucatan until the two plates are decoupled at ca. 154 Ma, and seafloor spreading ceases in the GOM, as suggested by the magnetic spreading model. A significantly faster spreading rate would cause YUC to keep up with SAM, not allowing rifting between the two plates, or would cause YUC to overtake SAM, which is an implausible scenario. The implication of the relative motions between the YUC and SAM is that as the Yucatan rotates, it slowly breaks up with SAM, along a rift propagating from northeast to southwest (Figures 811). If we assume that the velocities of YUC and SAM do not change much during this phase, we can extrapolate the motion back until this rift is “healed,” or the Yucatan and South American plate boundaries are parallel and overlapping. This occurs at roughly 169 Ma (Figure 11). At this point, we “end” the rotation phase of the Yucatan (in the reconstruction sense), and inherit the SAM plate motion to describe the extension direction for the “translation” phase of the early synrift, allowing roughly equal amounts of overlap between the YUC, NAM, and SAM in the final Pangea reconstruction. In short, we suggest that the transition from translation to rotation occurs around 169 Ma, before breakup has occurred anywhere in the paleo-GOM.

Our suggested prebreakup timing of the transition from translation to rotation implies a terminal phase of rift-related deformation, with the main extension direction nearly perpendicular to the previous direction from the early synrift. The pronounced magnetic lows on each side of the GOM’s central and eastern compartments have a sharp inboard edge that trends perpendicular to the rotation direction. We have associated this anomaly with removal of the lower crust, which results in a tapering of continental crust to less than 10 km thick. We therefore imply that removal of the lower crust is one of the primary ways in which hyperextension occurs in the GOM and is associated with the “rotation” directed extension of the terminal rift phase. We also have implied in our 2D model that some component of brittle extension in the upper crust should be parallel or subparallel to the rotation direction.

Conclusion

Long-wavelength magnetic anomalies in the GOM can be explained using analog rock properties for oceanic crust of the appropriate age and a two-layer continental crust having undergone outer marginal collapse leading to exhumation of the mantle. The distribution of anomalies is consistent with a simple rigid plate kinematic model derived from tectonic analysis of gravity anomalies that suggests symmetric oceanic spreading about identified mid-ocean ridges, in a central compartment of the GOM. Prior to breakup, symmetric and conjugate crustal structures and resulting anomalies were produced in a phase of hyperextension that removed the lower crust and exhumed the mantle in the central compartment of the GOM. Toward the east, where the extension magnitude and rate wane, the interpreted exhumed mantle and their related anomalies become asymmetric and eventually taper out. Toward the west, where the extension magnitude is greater and the extension rate faster, the crustal structure is more uncertain, due to a less clear anomaly pattern and more ambiguous seismic results. However, assuming symmetric spreading, a reasonable LOC can be estimated. The structures that accommodate breakup in this “western compartment” require further study.

Acknowledgments

We thank our Chevron colleagues P. Lovely, C. Rivero, and B. Cabote for the thoughtful conversations on geodynamics and structures in the GOM and helpful suggestions on the manuscript. We also are grateful to three anonymous reviewers and the associate editor who provided constructive feedback on our manuscript. Most of the analyses in this paper were performed using the open-source software GPlates and licensed versions of LCT’s “2MOD,” Seequent’s “Oasis Montaj,” Esri’s “Arc GIS,” and Petroleum Expert’s “MOVE.” We have provided our “releveled” magnetic grids and GPlates files as supplementary files to to this paper (S1, S2, S3, and S4). These supplementary files can be opened with the GPlates open-source software, are public domain, and are not subject to copyright.

Data and materials availability

Data associated with this research is in the public domain and properly referenced within our manuscript.

Daniel Minguez earned his Ph.D. (2015) in earth and environmental science from Lehigh University, Pennsylvania. He joined Chevron’s Energy Technology Company thereafter. At Chevron, he performs fundamental research in tectonics, structural geology, potential-fields geophysics, and paleomagnetism. He also serves an applied-science role for exploration and subsurface characterization around the world.

E. Gerald Hensel received his master’s degree (1982) in geophysics from the University of Washington, Seattle. He started his career with Chevron Overseas Petroleum that same year as a potential-fields geophysicist. He worked for Chevron in potential fields and related disciplines, including software support, application programming, database, and velocity analysis. He has had opportunities to work on domestic and international projects. He retired in 2018 and is active in professional and nonprofessional interests.

Betty Johnson is leader, Basin Framework Technology at Chevron. She received her B.S. (1978) in physics from Harvey Mudd College. She previously worked as a potential-fields geophysicist at Unocal. At Chevron, she manages the Basin Framework team and technology development program.

Freely available online through the SEG open-access option.