We report new strain analyses of mechanically twinned calcite in veins hosted by Neogene (13.6–4.3 Ma) sedimentary and volcanic rocks recovered from the Terror Rift system in the southern Ross Sea, Antarctica, by the ANDRILL (ANtarctic geological DRILLing) McMurdo Ice Shelf (MIS) Project. Strain analyses of the ANDRILL MIS AND-1B drill core samples yield prolate and oblate ellipsoids with principal shortening and extension strains ranging from –7% to 9%, respectively. The majority of samples show ≤25% negative expected values, indicating homogeneous coaxial strain characterized predominantly by subvertical shortening. We attribute the subvertical shortening strains to mechanical twinning at relatively shallow depths in an Andersonian normal faulting stress regime induced by sedimentary and ice sheet loading of the stratigraphic sequence and characterized by low stress magnitudes. Oriented samples yield a northwest-southeast average extension direction that is subparallel to other indicators of Neogene extension. This northwest-southeast extension is consistent with strain predicted by Neogene orthogonal rifting in a north-northeast–trending rift segment, as well as models of right-lateral transtensional rifting. The overall paucity of a noncoaxial layer-parallel shortening signal in the AND-1B twin populations favors orthogonal extension in the Neogene Terror Rift system, but could also be due to spatial partitioning of strain in a transtensional rift regime.


The Antarctic continental interior has undergone repeated rifting events within the West Antarctic Rift system (Fig. 1) since the creation of the Antarctic plate by ca. 80 Ma during Gondwana breakup (LeMasurier, 1990; Tessensohn and Wörner, 1991; Boger, 2011). The rift system is flanked by the Transantarctic Mountain uplift along the edge of the East Antarctic craton (ten Brink et al., 1997), shows multiple episodes of faulting (Davey and Brancolini, 1995; Salvini et al., 1997), and is marked by Miocene to active volcanoes (LeMasurier, 1990). Deformation within the rift system is of widespread interest because of its significance for understanding the breakup of Gondwana and constraining Mesozoic to Cenozoic global plate circuits (Steinberger et al., 2004; Granot et al., 2010, 2013). Rocks within the rift system have been the subject of geophysical studies (Bosum et al., 1989; Behrendt et al., 1991, 1996; Damaske et al., 1994; Davey and Brancolini, 1995; Salvini et al., 1997; Luyendyk et al., 2001, 2003; Finn et al., 2005; Karner et al., 2005; Hall et al., 2007; Henrys et al., 2007; Fielding et al., 2008), outcrop-based structural analyses (Wilson, 1995; Storti et al., 2001, 2008; Rossetti et al., 2000, 2002, 2003, 2006; Läufer et al., 2003) and thermochronological work (Fitzgerald, 1992, 2002; Lisker, 2002; Lisker and Läufer, 2013). However, there are few structural analyses of rift basin strata sampled by drilling (Wilson and Paulsen, 2000, 2001; Millan et al., 2007; Wilson et al., 2007), despite the fact that such data have the potential to provide important insights into the geodynamic evolution of rifting. The intent of this paper is to present new strain data from Neogene sedimentary and volcanic rocks recovered by the ANDRILL (ANtarctic geological DRILLing) McMurdo Ice Shelf (MIS) AND-1B drill core from the Terror Rift of the West Antarctic Rift (Fig. 1) to better define strain patterns and their relation to existing kinematic models for the evolution of the rift system.


Rifting within the West Antarctic Rift system commenced during Mesozoic breakup of Gondwana and has continued episodically into the Neogene (Cooper et al., 1987; Tessensohn and Wörner, 1991; Wilson, 1993; Encarnación et al., 1996; Cande et al., 2000; Rocchi et al., 2002, 2003; Henrys et al., 2007; Tonarini et al., 1997; Fielding et al., 2008; Sutherland, 2008; Wilson and Luyendyk, 2009; Granot et al., 2010, 2013). Tectonic models for the rift system typically invoke phases of extension in the Cretaceous and Cenozoic (Cooper et al., 1987; Tessensohn and Wörner, 1991; Davey and Brancolini, 1995; Salvini et al., 1997; Fitzgerald, 2002; Karner et al., 2005; Huerta and Harry, 2007; Bialas et al., 2007; Siddoway, 2008; Wilson and Luyendyk, 2009). Seismic studies indicate at least two distinct periods of Cenozoic rifting (Paleogene and Neogene) within the western Ross Sea. The youngest phase produced a Neogene fault system superimposed on the Cretaceous(?) to Paleogene Victoria Land rift basin in the western Ross Sea known as the Terror Rift (Fig. 1; Cooper et al., 1987; Davey and Brancolini, 1995; Salvini et al., 1997; Hall et al., 2007; Henrys et al., 2007; Fielding et al., 2008). Neogene faulting within the Terror Rift has occurred concomitant with the eruption of an extensive alkali volcanic province along the Transantarctic Mountain rift flank and offshore localities collectively known as the McMurdo Volcanic Group (Bosum et al., 1989; LeMasurier, 1990; Kyle, 1990; Behrendt et al., 1991, 1996; Damaske et al., 1994). Active volcanism characterizes the Terror Rift today, and interpretations of marine seismic data indicate that faulting is locally as young as the Pliocene to Pleistocene (Hall et al., 2007; Henrys et al., 2007).

Previously workers have attributed the presence of transtensional fault arrays in the southern and northern Victoria Land sectors of the Transantarctic Mountain rift flank to right-lateral oblique rifting during the Cenozoic (Wilson, 1995; Salvini et al., 1997; Storti et al., 2001, 2008; Rossetti et al., 2000, 2002, 2003, 2006). Other geodynamic models for Mesozoic–Neogene rifting call for simple orthogonal extension (Busetti et al., 1999; Trey et al., 1999; Cande et al., 2000; Karner et al., 2005; Davey and De Santis, 2006; Davey et al., 2006; Wilson and Luyendyk, 2009; Granot et al., 2010, 2013). Storti et al. (2008) challenged Cenozoic orthogonal rifting models, and argued for a Paleogene inception of the Terror Rift, which, according to their model, encompasses reverse, strike-slip, and normal faulting in the Transantarctic Mountain rift flank uplift and offshore regions of the western Ross Sea (Fig. 1). In this model, Paleogene to recent faulting accommodates right-lateral oblique extension induced by a transfer of shear from right-lateral transform faulting along the mid-ocean ridge in the Southern Ocean (Salvini et al., 1997; Storti et al., 2007).

The debate surrounding the nature of rift kinematics in the western Ross Sea exists because spatial and temporal patterns of strains recorded within the rift system are poorly documented. Kinematic studies of fault arrays within the Transantarctic Mountains rift flank are hampered because faults in these areas typically occur in rocks that are no younger than Jurassic (Warren, 1969). Seismic studies of fault systems in the submarine rift basins are hampered because of limited information on fault kinematics. There have, however, been drilling projects that have recovered deformed Cenozoic strata from the western sector of the rift system (Fig. 1).

One of the most important drill cores yet obtained of Neogene strata filling the Terror Rift is the ∼1285 m ANDRILL MIS AND-1B drill core (Naish et al., 2009). The AND-1B drill core was recovered in the Windless Bight region of the Ross Ice Shelf in 2006–2007 (Figs. 1 and 2; Naish et al., 2007). The core includes interbedded late Miocene to Pleistocene sedimentary (diatomite, diamictite, sandstone, and shale) and volcanic rocks (Fig. 3; Krissek et al., 2007; McKay et al., 2009; Di Roberto et al., 2010; Cody et al., 2012; Ross et al., 2012; Williams et al., 2012; Wilson et al., 2012). Systematic fracture logging of the core identified ∼1400 natural fractures (i.e., preexisting fractures in the rock intersected by coring) that are dominated by normal faults and calcite veins (Figs. 4 and 5; Wilson et al., 2007). Steep extension veins folded by compaction (Fig. 5A), fault breccias (Fig. 4B), and unfolded extension veins (Fig. 5B) indicate deformation of a continuum of sublithified to lithified rock during burial and lithification of the stratigraphic sequence (Wilson et al., 2007; Millan, 2013). Natural fractures are present in the highest core that the Core Structure Measurement Group was able to log (i.e., not retrieved in plastic liners) at the ANDRILL McMurdo Ice Shelf Drill Site Laboratory, with steep conjugate faults occurring ∼42 m below seafloor (mbsf) (Wilson et al., 2007). There is a zone of high fracture density in an interval between ∼125 and 300 mbsf that is estimated to range in age from ca. 1.6 to 3 Ma (Wilson et al., 2012). Natural fractures are ubiquitous in the core below ∼450 mbsf (Wilson et al., 2007), consistent with the presence of normal faults detected on seismic sections below the Ri seismic reflector (Figs. 2 and 3; Horgan et al., 2005) that has been interpreted as a regional ca. 4.3–3.6 Ma unconformity in the Victoria Land Basin (Fielding et al., 2008; Wilson et al., 2012).

The natural fractures recovered in the AND-1B core assume regional significance in the problem of West Antarctic Rift system tectonics because they are rare examples of rift structures that crosscut Neogene strata and can be directly studied. This paper focuses on Neogene strains recorded by mechanically twinned calcite found along veins, faults, clast margins, and in cavities at depths ≥473 mbsf within the AND-1B core. Calcite acts as an extremely sensitive strain gauge by twinning at low differential stresses (≤20 MPa) (Lacombe and Laurent, 1996; Ferrill, 1998; González-Casado et al., 2006). This property makes twinning strain analyses a powerful tool that has been used to study paleostrain and stress patterns of rock cements and secondary veins found in outcrops and drill cores from relatively undeformed sectors of plate interiors (Lacombe et al., 1990; Craddock et al., 1993; Craddock and Pearson, 1994; van der Pluijm et al., 1997), as well as orogenic belts (Engelder, 1979; Teufel et al., 1984; Craddock et al., 1988, 2000, 2007; Kilsdonk and Wiltschko, 1988; Evans and Dunne, 1991; Lacombe et al., 1994; González-Casado and García-Cuevas, 1999; Craddock and Relle, 2003; González-Casado et al., 2003) and rift systems (Friedman and Heard, 1974; Lomando and Engelder, 1984; Craddock et al., 1997).

Other than one study of calcite twinning strains in cements and veins within early Paleozoic rocks of the Ellsworth Mountains (Craddock et al., 1998), there have been no other calcite strain analyses reported from Antarctica. In the case of the Terror Rift, calcite twinning analyses allow us to evaluate the problem of Neogene rifting strains in a new way, namely by determining whether mechanical twins in calcite recovered from the faulted stratigraphic section record strain patterns expected if deformation occurred within Andersonian strike-slip or normal faulting stress regimes. If deformation of the calcite occurred due to horizontal compression associated with strike-slip tectonism, then we expect mechanical twinning to have recorded subhorizontal maximum shortening strains. In contrast, deformation due to the lithostatic loading associated with a normal faulting stress regime predicts subvertical maximum shortening strains. Calcite strain analyses can also yield important information about the direction of extension within the Terror Rift at the time of twinning. Here we provide new constraints on strains associated with deformation within the Terror Rift, through calcite twinning strain analyses performed on 19 samples collected from the AND-1B core.


We collected samples from the AND-1B core with the purpose of studying fracture fills to supplement fracture logging conducted at the AND-1B drill site during coring operations. We refer to the samples by their depths (in mbsf) in the core. From this sample suite we selected 19 samples of vein calcite from a depth range of 473 to 1279 mbsf that appeared to have sufficient numbers of twinned calcite grains to determine three-dimensional (3-D) strain ellipsoids using the calcite strain-gauge technique (Groshong, 1972, 1974). Figure 3 shows the locations of the samples within the AND-1B core. Our sample suite includes twinned calcite along striated faults (Fig. 4A; samples 911.97 and 1085.71), cemented fault breccias (Figs. 4B and 6C; samples 1279.2 and 1279.57), anastomosing webs of extension veins that likely mark shear zones within the core (Fig. 4C; samples 905.25, 918.74, 920.83, 930.04), extension veins subsidiary to a normal fault (Fig. 4D; sample 1265.66), and an ∼55° dipping fracture classified as a possible fault based on its dip angle during structural logging (sample 842.01). The sample suite also includes 6, ∼1–2-mm-thick steeply dipping (>75°) folded extension veins (Figs. 5A and 6D; samples 961.29, 1115.62, 1115.66, 1119.36, 1193.53, and 1274.34), an unfolded extension vein (Figs. 5B and 6B; sample 674.82), a calcite-lined open cavity in a volcanic flow (Fig. 6A; sample 648.13), and a calcite-rimmed sedimentary clast margin (sample 473.84). All of the samples were oriented with respect to the up direction and an arbitrary north direction marked by a red scribe line drawn along the length of the core (e.g., Figs. 4B, 4C, 5A, and 5B).

We followed standard procedures for conducting the optical universal stage microscope measurements of c-axis and twin orientations, twin and grain thicknesses, and the number of twins within mechanically twinned calcite grains (Turner and Weiss, 1963; Evans and Groshong, 1994). We measured and analyzed thin and thick twin sets and assigned an average thickness of 0.5 μm for thin twins. Calcite has three possible twin glide horizons that can be activated if the principal stresses have suitable orientation and sufficient magnitude to achieve the critical resolved shear stress (∼10 ± 5 MPa) required to induce mechanical twinning in optimally oriented grains (Turner et al., 1954; Jamison and Spang, 1976; Ferrill, 1998; Lacombe and Laurent, 1996; Lacombe, 2001; González-Casado et al., 2006). The calcite strain-gauge technique can yield accurate strain tensor results for as few as five twin sets in the absence of a complex strain history and measurement errors. However, in practice, more robust ellipsoid results are assured with the measurement and analysis of higher numbers of twin sets, which helps maximize 3-D coverage of twin optic orientations (Groshong, 1974; Groshong et al., 1984). In cases where we had ample sample material from the drill core (n = 5; samples 473.84, 674.82, 905.24, 1115.66, and 1279.57), we analyzed twins in as many as 25 grains in 2 orthogonal thin sections (Groshong et al., 1984). For the remaining samples (n = 14), we measured twins in as many as 50 grains in a single or multiple parallel thin sections.

We used the CSG99 strain-gauge software (Evans and Groshong, 1994) to conduct calcite strain-gauge analyses on the optically cleaned data sets. The first step in a calcite strain-gauge analysis involves determination of a bulk strain tensor using all of the data collected (referred to as ALL) from a calcite aggregate (Groshong, 1972, 1974; Groshong et al., 1984). The technique determines the expected value of strain for each twin set given the calculated bulk strain tensor, as well as the deviation of the expected and measured values of strain for each set (Groshong, 1974). These initial results are, in turn, used to guide the removal of 20% of the data with the largest magnitude deviations between expected and measured strains from the entire data set. This cleaning procedure improves strain ellipsoid accuracy by reducing data scatter that is likely related to inhomogeneous strains and possible measurement errors (Groshong, 1974; Teufel, 1980; Groshong et al., 1984). The remaining data set (largest deviations removed; referred to as LDR) is then used for a new strain analysis. Twin sets within grains that are expected to twin due to the calculated bulk strain tensor are known as positive expected values (PEV), whereas twin sets within grains that are not expected to twin due to the calculated bulk strain (i.e., they have the wrong sense of shear) are referred to as negative expected values (NEV) (Groshong, 1974). LDR strain-gauge analyses that yield >40% NEV could signal noncoaxial or inhomogeneous strains (Groshong, 1974; Teufel, 1980). In such a case, the original data (i.e., the entire data set prior to LDR cleaning) can be divided into PEV and NEV data sets and analyzed separately to evaluate the possibility of noncoaxial deformation (Groshong, 1974; Teufel, 1980).

Calcite grain c-axes are not uniformly distributed in our data sets (they range from having 5–10 sigma distributions on Kamb contour plots); this is probably related to our limited numbers of orthogonal thin sections, the limited range of view offered by the universal stage (Stauffer, 1966), and possibly to mineral growth within fractures (Bons et al., 2012). We evaluated the impact of the presence of the c-axis concentrations on the strain analyses by reducing c-axis concentrations by culling 0%–63% of twin sets whose grains contributed to optic axis concentrations until optic axes had ≤5 sigma distributions on Kamb (1959) contour plots. These optically culled data sets (referred to as OCLDR, OCPEV, and OCNEV) were then analyzed following the analytical procedure outlined here.

Strain ellipsoids for four samples (905.24, 911.97, 920.83, and 930.04) were reoriented with respect to north by matching features in core imagery with correlative features observed in oriented borehole televiewer imagery (see Jarrard et al., 2001a; Paulsen et al., 2002). Borehole televiewer logging could not be conducted deeper than 1018 mbsf because these borehole depths exceeded the length of the logging cable (Morin et al., 2007). Four samples (1265.66, 1274.34, 1279.20, and 1279.57) from these depths were therefore reoriented with respect to north by rotating the average bedding dip direction of their respective intact core intervals to match the average in situ bedding dip direction (320°) determined from oriented intact core intervals. The remaining samples (n = 11) and their respective strain ellipsoid results are oriented with respect to the up direction in the core, but not with respect to north. The AND-1B borehole is within ∼2° of vertical (Morin et al., 2007), indicating that the maximum principal strain axes of the strain ellipsoids and the c-axes of the data sets can be viewed in an Earth surface frame of reference.

Stereonets in Figure 7 show the three principal strain axes and c-axis distributions for the individual sample analyses of the ALL, LDR, PEV, and NEV data sets, as well as the optically culled OCLDR, OCPEV, and OCNEV data sets. Figure 8 shows cumulative stereonet plots of the principal shortening (e1) and extension (e3) strain axes determined from the individual sample analyses. Table 1 provides the details of the calcite strain data analyses for the raw data set (ALL). Table 2 provides the details of the calcite strain data analyses for the LDR, PEV, and NEV data sets, as well as the optically culled OCLDR, OCPEV, and OCNEV data sets. Tables 1 and 2 also show differential stresses inferred from average twin densities using the method of Rowe and Rutter (1990).


Twinned and untwinned calcite grains occur within fracture fills in the samples and range in morphology from fibrous to elongate blocky and blocky calcite spar (Figs. 6A–6D). Grain sizes measured perpendicular to twins in the raw data set (ALL) range from ∼85 µm to 780 µm. Twin densities of the 6 folded vein samples are generally higher than the majority of the twin densities in the remaining 13 samples (e.g., 61 twins/mm versus 34 twins/mm average in ALL; Table 1). Thick twins and bent twins tend to be more common in the folded vein sample suite. Overall, twin thicknesses (thin and thick), twin morphologies, and twin densities (e.g., 5 twins/mm to 71 twins/mm in ALL) suggest that twinning generally occurred at temperatures <200 °C (Burkhard, 1993; Ferrill, 1991; Ferrill et al., 2004), consistent with AND-1B borehole logging studies that indicate a 76.7 °C/km contemporary temperature gradient at the drill site (Morin et al., 2010). Differential stresses inferred from average twin densities range from 68 to 265 MPa for the raw data set (ALL) (Rowe and Rutter, 1990), but these values are likely unreliable overestimates because the twin density technique is not applicable to low-temperature (<200 °C) deformation (Burkhard, 1993; Ferrill, 1998).

Mechanical twinning of calcite due to the application of noncoaxial vertical and horizontal loads predicts high NEV percentages (>40%) following LDR analysis (Teufel, 1980), but LDR data for 18 of 19 samples show ≤25% negative expected values (Table 1). Collectively, these results indicate homogeneous coaxial strain and are consistent with a single or multiple approximately coaxial deformation episodes (Teufel, 1980). The sole exception is sample 473.84 (∼30% NEV for LDR data), for which we divided and analyzed the PEV and NEV data separately.

LDR strain analyses of our samples yield strain ellipsoid results that are similar to the strain results produced by analysis of the raw data set (ALL). LDR strain analyses yield prolate (58%), oblate (32%), and plane (10%) strain ellipsoids with principal shortening (e1) and extension (e3) strains ranging from –6.7% to 8.6%, respectively (Table 2). LDR data from a majority of the samples (n = 17 of 19) and the PEV data split for sample 473.84 yield similar strain ellipsoid orientations (subvertical maximum shortening and subhorizontal maximum extension axes) (Table 2; Figs 7 and 8). The only two analyses that did not yield subvertical shortening directions include the LDR results for sample 648.13 and the LDR/NEV data split for sample 473.84, both of which show shallow maximum principal shortening and stretching axes. The principal shortening directions (e1) yielded by the LDR data are similar to the e1 directions yielded by the raw data set (ALL) (i.e., they are also dominated by subvertical e1 axes). The LDR data e1 axes show less scatter with respect to the e1 axis population yielded by the raw data (ALL), as would be expected because of the increased accuracy produced by the reduction of inhomogeneous strains and possible measurement errors (Groshong, 1974; Teufel, 1980; Groshong et al., 1984).

Optic axis populations of the optically culled data (OCLDR) (Fig. 7) can be divided into three types based on their distribution. Type 1 samples (1115.66, 1265.66, 1274.34, and 1279.57) have optic axis populations that show nearly random distributions. Type 2 samples (473.84, 648.13, 674.81, 905.24, 911.97, 920.83, 961.29, 1085.7, 1193.53, and 1279.2) show good 3-D optic axis distributions that have weak girdle or broad bimodal patterns. Type 3 samples (842.01, 918.74, 930.04, 1115.62, 1119.36) show limited to fair 3-D optic distributions with girdle patterns for relatively small populations of optic axes (≤17). These optic axis distributions are generally similar to other twin analyses of calcite veins that have yielded interpretable paleostrain and stress results (Kilsdonk and Wiltschko, 1988; Craddock et al., 2007; Lacombe et al., 2009).

The optically culled OCLDR, OCPEV, and OCNEV data sets yield strain ellipsoid axis magnitudes (–6% to 8%) and orientations that are virtually the same as the results yielded by the LDR, PEV, and NEV analyses, regardless of differences in optic axis distribution types (i.e., types 1–3) (Table 2; Figs. 7 and 8). The optic axis concentrations present in some of the samples therefore appear to have a negligible impact on the first-order results of the LDR, PEV, and NEV strain analyses, which predominantly show subvertical principal shortening and subhorizontal principal extension directions.

Experimental studies indicate that the best-constrained principal strain axis is the axis that has a magnitude with opposite sign with respect to the other two principal axes (Groshong et al., 1984). Of 19 samples, 10 yield positive (extensional) e2 axes as a result of LDR, OCLDR, or PEV analyses, meaning that their subvertical negative (shortening) e1 axes are the best-resolved strain axes. The nine remaining samples yield negative (shortening) e2 axes as a result of LDR or OCLDR analyses, meaning that their shallow e3 (extension) axes are the best-resolved strain axes. The subvertical e1 (shortening) axes of two of these nine samples (1115.62 and 1279.2) are independently confirmed by strain results from samples 1115.66 and 1279.57, which come from the same two intact core intervals and show subvertical e1 (shortening) axes that are the best-constrained strain axes. The subvertical e1 (shortening) axes for sample 1115.62 and four other of the nine samples (961.29, 1119.36, 1193.53, and 1274.34) are independently supported by subvertical shortening implied by folding of their respective veins around subhorizontal axes.

Reorientation of LDR and OCLDR strain ellipsoids for samples 905.24, 911.97, and 930.04, with respect to north yields east-west to northwest-southeast maximum principal extension axes, whereas sample 920.83 yields a north-south maximum principal extension (e3) axis (Fig. 7). Reorientation of LDR and OCLDR strain ellipsoids for samples 1265.66, 1274.34, 1279.20, and 1279.57 with respect to north yields east-northeast–west-southwest to northwest-southeast maximum principal extension (e3) axes.


Deformation Timing

All of the calcite samples studied herein were collected below a basin-wide unconformity (i.e., the Ri seismic reflector) that is ca. 4.3–3.6 Ma at 440 mbsf in the AND-1B core (Figs. 2 and 3). This correlates with the ca. 13 Ma to ca. 4 Ma stratigraphic interval that is dominated by normal faulting in marine seismic data of the Victoria Land Basin, as well as normal faults and opening mode veins in the AND-1B core (Wilson et al., 2007, 2012; Hall et al., 2007; Henrys et al., 2007; Fielding et al., 2008; Millan, 2013). Twinned calcite along or proximal to striated faults (samples 911.97, 1085.7, and 1265.66) suggests that twinning might have occurred, at least in part, concomitant with faulting of the AND-1B stratigraphic sequence (e.g., Laurent, 1987). Age constraints permit faulting, tensile fracturing, and mechanical twinning of fracture fills to be as old as the ca. 13 Ma to ca. 4 Ma stratigraphic intervals in which they are hosted. However, faulting as young as Pliocene to Pleistocene age has been documented elsewhere for the Terror Rift (Hall et al., 2007; Henrys et al., 2007; Fielding et al., 2008), and is possible for fractures (and mechanical twins) within the AND-1B core. The greater abundance of veins below the Ri unconformity and the interpretation that folded veins formed during compactional volume loss influenced by diagenetic reactions and collapse of early cement framework (Millan, 2013) argue for a late Miocene to Pliocene age for vertical shortening.

Stress Magnitudes

The ∼10 ± 5 MPa critical resolved shear stress that is generally thought to be required for mechanical twinning of calcite (Turner et al., 1954; Jamison and Spang, 1976; Ferrill, 1998; Lacombe and Laurent, 1996; Lacombe, 2001; González-Casado et al., 2006) would require a minimum differential stress of ∼20 ± 10 MPa to cause twinning on optimally oriented twin glide horizons. The lack of precise age constraints for the time of twinning within our samples precludes definitive reconstruction of the lithostatic loads at the time of twinning. However, a conservative figure for the present effective maximum vertical stress (lithostatic stress with an assumed hydrostatic pore pressure removed) at the depths from which we collected our samples (473–1279 mbsf with an additional ∼900 m water depth) has been determined, using core grain density measurements and calculated fractional porosities, to range from 3.7 to 12.5 MPa (Niessen et al., 2013). Cyclic vertical loading of the crust by the repeated grounding of continental ice sheets has also occurred since ca. 12 Ma in the Ross Sea (McKay et al., 2009; Naish et al., 2009) and probably increased the vertical stress during the Last Glacial Maximum by ∼6.5 MPa at the drill site (Niessen et al., 2013), yielding effective vertical stresses ranging from ∼10 to 19 MPa. Cumulative loads like these would place our samples within the depth window for twinning in an ambient stress field characterized by vertical uniaxial compression or low minimum horizontal stresses. It is also possible that twinning reflects greater loads or smaller assumed pore pressures than held by the assumptions, or stress concentrations associated with faulting and/or bulk vertical compaction of the sequence. Stress concentrations would permit twinning at shallower depths earlier in the burial history, but even in this case, the subvertical shortening strains are still ultimately related to the force imposed by the lithostatic load.

Stress Regime and Rift Kinematics

The principal stress axes at shallow depths (<2 km) within the crust are generally expected to have subvertical (SV) and subhorizontal orientations (Anderson, 1951; Engelder, 1993). The vertical stress in the crust is due to the gravitational force exerted by the column of overlying rock (i.e., the lithostatic pressure). The vertical stress is expected to induce a subvertical maximum principal shortening strain as it drives the rearrangement of grains during porosity loss associated with burial (Ramsay and Wood, 1973). Mechanical twinning of calcite occurs under relatively low differential stresses (≤20 MPa; Lacombe and Laurent, 1996; Ferrill, 1998; González-Casado et al., 2006) and when lithostatic loads are sufficient and SV > SH (the maximum horizontal stress), twinning is expected to record a subvertical shortening strain (Friedman and Heard, 1974; Lomando and Engelder, 1984, Craddock et al., 2000; González-Casado et al., 2006). The first-order results yielded by twinned calcite in the AND-1B samples record a relatively uniform strain pattern characterized by subvertical shortening and shallow (layer parallel) extension, indicative of twinning in a normal faulting stress regime (SV > SH). Relatively small bedding dips (∼10° for stratigraphic intervals deeper than ∼900 mbsf; depositional ages ca. 9 Ma or older) and differences in optical axis distributions appear to have had little impact on these first-order results. The subvertical shortening strains indicated by the majority of the calcite data are therefore consistent with the normal faulting stress regime indicated by the dominance of normal faults within the AND-1B core and demonstrated by marine seismic sections across the Terror Rift to the north of Ross Island (Hall et al., 2007; Henrys et al., 2007).

Twinning due to subvertical shortening (compaction) appears to be substantiated by variations in twin densities. Rock mechanics studies show that twin densities will increase with increasing strain at low temperatures (<350 °C) (Rybacki et al., 2013). The correlation of high twin densities with folded veins is therefore consistent with the accruement of greater twinning strains in veins that formed earlier in the compaction of their hosting stratigraphic intervals (Wilson et al., 2007; Millan, 2013). Lower twin densities in some or in all of the remaining samples might therefore reflect twinning later in the burial history when smaller porosity reductions were insufficient to cause buckling of preexisting veins. All of the folded veins yield subvertical shortening strains, which are consistent with the subvertical shortening implied by the folds. This, coupled with the ≤25% NEV yielded by the folded veins, suggests that possible inhomogeneous strains associated with twin plane rotation during folding were successfully reduced and/or that twinning occurred late with respect to folding (Groshong, 1974). However, we cannot preclude the possibility of some twin plane rotation in the folded vein samples, and the folded vein strain ellipsoid results should therefore be viewed with this in mind.

The azimuths of the long axes of strain ellipsoids yielded by the ALL, LDR, and OCLDR data sets yield virtually indistinguishable 106° (27° alpha-95), 104° (28° alpha-95), and 109° (30° alpha-95) Fisher mean directions, respectively (Fig. 8). Excluding sample 1274.34 because of possible twin plane rotations during folding has negligible impact on the Fisher mean directions for the ALL (109°, 31° alpha-95), LDR (106°, 33° alpha-95), and OCLDR (113°, 35° alpha-95) data sets. The anomalous long axis (e3) of the strain ellipsoid for sample 920.83 could be a signal of radial extension during burial compaction (Millan, 2013). Alternatively, it could mark an artifact related to the limited number of oriented samples, errors, or noise that could cause a large strain ellipsoid rotation because of small differences in strain magnitude between the intermediate (approximately east-west; e2) and maximum (approximately north-south; e3) axes of the strain ellipsoid.

The dominant east-southeast–west-northwest average extension direction yielded by individual analyses is consistent with the west-northwest and east-southeast dip directions of north-northeast–striking faults interpreted on marine seismic sections at depth at the drill site (by T.J. Wilson and S. Henrys) as well as the dominant northeast strike of folded veins from the AND-1B core (Millan, 2013). Northwest-southeast extension is subparallel to the west-northwest (∼288° azimuth) deviation of the AND-1B borehole from vertical (from 700 to 1024 mbsf), which might reflect updip prodding of the drill bit by a population of fractures with a dominant southeast dip direction within this section of the borehole (Morin and Wilkens, 2005; Morin et al., 2010). Northwest-southeast extension is also subparallel to the northwest-southeast extension direction implied by the elongation of the ca. 1.3 Ma Hut Point and younger than 7.7 Ma White Island volcano complexes, Pleistocene to contemporary volcano alignments at Mount Morning (Paulsen and Wilson, 2009), and faults hosted in Oligocene strata in the Cape Roberts drill core (Fig. 9; Wilson and Paulsen, 2001).

Northwest-southeast extension indicated by the calcite twin data is oblique to the overall north-northwest strike of longitudinal faults in the Terror Rift system north of Ross Island. However, the northwest-southeast extension direction is at a high angle to a northeast-striking rift segment that has been inferred in the area of the drill site and to the south (Johnston et al., 2008) in the same general region where the Transantarctic Mountain Front rift border fault is inferred to change trend and step westward across a regional accommodation zone (Wilson, 1999). Northwest-southeast extension is consistent with strain predicted by orthogonal rifting in a northeast-trending rift segment, as well as tectonic models that invoke Neogene right-lateral transtensional rifting.

Layer-parallel shortening in samples 473.84 (NEV) and 648.14 could reflect twinning due to horizontal compression within an Andersonian strike-slip faulting stress regime, but local inhomogeneous strain-stress fields are also possible. The nearly orthogonal flip between the e1 and e3 axes for PEV and NEV data from sample 473.84 makes polyphase deformation dubious (Burkhard, 1993). The overall paucity of a noncoaxial layer-parallel shortening signal in the AND-1B twin populations indicates that the horizontal compressive stresses hypothesized by transtensional kinematic models for the rift system have been absent or of insufficient magnitude to cause a widespread noncoaxial strain overprint. The general lack of accompanying horizontal shortening in the twin data suggests that the Neogene Terror Rift system lacks a strong longitudinal strike-slip component (i.e., it favors orthogonal extension), but it is possible that the virtual absence of horizontal shortening could be due to spatial partitioning of strain within a transtensional rift system.

Given the young age for the calcite twinning strains, it is noteworthy that there is a rough correlation between the average extension direction recorded by the calcite and the present-day extension direction (070°–110°) predicted to be induced at 100 km depth by geodynamic modeling of mantle traction forces and plate motion in the Ross Sea region (Faccenna et al., 2008). The calcite strain results reported herein are currently limited to a single location within the rift system, precluding the delineation of rift-wide strain patterns and any possible changes with respect to time. In order to fully understand the spatial and temporal patterns of Neogene deformation within the Terror Rift and their relation to potential drivers such as mantle flow, there is a clear need for continued acquisition of oriented strain data from the Terror Rift and adjacent sectors of the West Antarctic Rift system.

We thank members of the ANDRILL (ANtarctic geological DRILLing) MIS (McMurdo Ice Shelf) Project downhole logging team for their work that enabled core reorientation studies after drilling and the ANDRILL MIS Science Team for their contributions to the project. We also thank Richard Groshong and John Craddock for helpful reviews that improved this manuscript, Mark Evans for providing a copy of the calcite strain-gauge software, Rick Allmendinger for providing his stereonet program, and the many ANDRILL support personnel that made this project possible. Paulsen acknowledges the University of Wisconsin Faculty Sabbatical Program for release time to work on this project and thanks the Antarctic Research Centre at Victoria University of Wellington (New Zealand) for hosting this work. The ANDRILL Program is a multinational collaboration between the Antarctic Programs of Germany, Italy, New Zealand, and the United States. Antarctica New Zealand is the project operator, and has developed the drilling system in collaboration with Alex Pyne at Victoria University of Wellington and Webster Drilling and Enterprises Ltd. Scientific studies are jointly supported by the U.S. National Science Foundation, New Zealand Foundation for Research Science and Technology, Royal Society of New Zealand Marsden Fund, the Italian Antarctic Research Programme, the German Research Foundation (DFG; grant LA 1080/6 to Läufer), and the Alfred Wegener Institute for Polar and Marine Research (Helmholtz Association of German Research Centres). Antarctica New Zealand supported the drilling team at Scott Base; Raytheon Polar Services supported the science team at McMurdo Station and the Crary Science and Engineering Laboratory. The ANDRILL Science Management Office at the University of Nebraska-Lincoln (Nebraska, USA) provided science planning and operational support.