The nature and history of the cratonic lithosphere beneath the southern margin of Laurentia are poorly understood due to the paucity of basement exposure in the Gulf Coast region. One place where we are afforded the opportunity to study the deep lithosphere of southern Laurentia is within the Balcones igneous province of Texas. Mantle xenoliths occur in Late Cretaceous alkali basaltic magmas erupted through the remnants of the Appalachian-Ouachita structural belt. Geochemical signatures in the form of enrichments in fluid-mobile trace elements (e.g., La) relative to fluid-immobile trace elements (e.g., Nb) are preserved in these mantle xenoliths. We interpret these signatures to represent metasomatism by subduction-related fluids, suggesting that the mantle xenoliths represent fragments of continental lithospheric mantle that served as the upper plate during a convergent episode. These xenoliths may thus represent fragments of Laurentian lithospheric mantle that were stabilized by collisional processes during the continent-continent collisions associated with the Grenville orogeny and subsequent Paleozoic orogenies. These observations suggest that some of the original continental lithosphere was preserved beneath the orogenic belt during continent-continent collision and did not undergo wholesale removal or delamination. Alternatively, the fluid signatures may have been caused by Farallon flat-slab hydration of the overriding lithosphere, but such a scenario seems unlikely because flat subduction is thought to have culminated in the early Cenozoic and hence after the eruption of these xenoliths.

The southern margin of the Laurentian craton has experienced orogeny in both the Proterozoic (Grenville) and Paleozoic (Ouachita) (Hatcher, 1989). Southern Laurentia has not been central in plate reconstructions, mainly due to a lack of basement exposure. However, a growing body of literature is attempting to delimit the extent and nature of lower lithospheric structures in the Gulf Coast region and extend tectonic models to include this region (Dunn, 2008; Gao et al., 2008; Mosher et al., 2008; Poole et al., 2005). Here, we present new geochemical data on mantle xenoliths from Knippa, Texas, in order to better understand the relationship between the subcontinental lithospheric mantle and the overlying crustal basement of Laurentia along its southern structural edge. We integrate the geochemical data from the mantle xenoliths with geologic studies of the crustal basement, as well as geophysical studies of the deep crust and lithosphere, to generate possible scenarios for the origin of mantle metasomatism and discuss the implications of these scenarios for the evolution of the deep lithosphere beneath the Laurentian craton.

Lithospheric mantle xenoliths (spinel lherzolites) were collected from the Vulcan open-pit quarry in the town of Knippa at 29.29°N, 99.65°W, located in Uvalde County, Texas (Fig. 1). The xenoliths are hosted in a fine-grained, olivine nephenilite. There are also fine- to coarse-grained alkali basalts, olivine nephelinites, nepheline basanites, and phonolites in the area (Miggins et al., 2004). Series of Late Cretaceous basalts were erupted through Cretaceous shallow-marine sediments, forming mostly hypabyssal laccoliths, plugs, sills, and dikes, along with accumulations of ash and tuff that settled in water (Ewing and Caran, 1982). Miggins et al. (2004) dated the eruption of these basalts between 80 and 82 Ma based on 40Ar/39Ar chronology. The volcanoes have intruded through the buried Paleozoic structural belt (delimited by the trend of a potential field trough), where Mesozoic sediments overlie Proterozoic Laurentian basement (Flawn et al., 1961).

The Knippa basalts appear to be related to a much larger volcanic province that stretches from south-central Texas east through Oklahoma and Arkansas. Late Mesozoic alkali igneous activity occurs along the Ouachita belt in Texas, Arkansas, and Oklahoma (Byerly, 1991; see Fig. 1). Spencer (1969) recognized that these widely spaced volcanic centers were probably related temporally and tectonically, referring to the collection of volcanic centers as the “Balcones igneous province” because the trend of the volcanic rocks and hypabyssal intrusive laccoliths and plugs follows the buried Paleozoic Ouachita structural belt, which in Texas follows the Balcones escarpment. Magnetic surveys have since confirmed the extent of the Balcones igneous province (Ewing and Caran, 1982; Smith et al., 2001, 2008). In Figure 1, the locations of magmatic intrusions are reproduced from Ewing and Caran (1982). The locations of the Balcones and Luling faults are also shown. The broad zone of Cretaceous igneous activity in Texas is roughly bounded by the Balcones and Luling fault zones, and it lies inside the trend of a regional gravity anomaly. The exact origin of these basalts remains unclear and is a subject for further study. It has been suggested that sediment loading of the Gulf of Mexico could have concentrated flexure of the lithosphere near a structural hinge such as the Ouachita structural belt (Miser, 1934), but flexure alone is not sufficient to cause melting because flexure (without erosion) does not lead to decompression of the mantle. Griffin et al. (2008) has suggested instead that the igneous province represents small-degree melts associated with decompressing asthenospheric mantle in response to lithospheric extension.

The xenoliths are clinopyroxene-poor lherzolites and harzburgites, containing variable proportions of olivine (ol), orthopyroxene (opx), clinopyoxene (cpx), and spinel (sp). The xenoliths occur as angular fragments within the basalt. Limited reaction between the host basalt and the xenoliths can be seen in the form of brownish, 1-mm-thick rims around the mantle xenoliths. The unreacted interiors of these xenoliths, however, still retain the original olivine-green color characteristic of unreacted peridotites.

Nine xenoliths were analyzed for this study. Modal abundances were difficult to calculate accurately because of the small size of the xenoliths (0.75–2 cm in diameter) and their coarse-grained nature (>1 mm). Modes were thus estimated by two methods. The first approach was point counting. However, many clinopyroxene (cpx) grains were much smaller than the average grain size of the xenoliths, and this resulted in either an overestimate or underestimate of cpx mode. To better assess cpx mode, we estimated the areal proportion of cpx in thin section (areal estimates and point counts of olivine [ol] and orthopyroxene [opx] modes were similar). The resulting mineral modes are 70–72% ol, 21–25% opx, 3–8% cpx, and ~2% spinel (sp) for samples 02A, 02F, 05A, 05B, 11B, and 15B. These modal mineralogies are similar to the dominant types of residual peridotite xenoliths worldwide. Two samples (01A and 08A) have <3% cpx and are classified as refractory harzburgites. Sample 06E is anomalously low in opx (~10%), but, in every other respect, it is similar to the other lherzolites; this may be due to poor statistical sampling.

Texturally, the xenolith grains are coarse and equigranular to slightly tabular, and they have many well-developed triple junctions. Sample 08A is the only sample to show textural evidence for deformation: alignment of spinels and orthopyroxenes defines a weak foliation, and spinels occur as slightly elongated blebs. Clinopyroxenes of varying texture occur. Most cpx grains are heavily “corroded” with spongy or turbid textures, which we interpret to be due to flash-heating and decompression during ascent in the host basalt. Some clinopyroxenes have only a spongy rim, and this rim is diffuse when sharing an edge with orthopyroxene. Only the occasional large clinopyroxenes (~1.5 mm) have clear cores without any distinctive turbid textures. Orthopyroxenes have well-defined edges and occur in clusters. They are light green in plain light, and cleavage fractures are visible in plane polarized light. The olivine is equigranular to tabular shaped and coarse grained (up to 3 mm). The olivines are sometimes cut by thin serpentinized veins, which are associated with postemplacement weathering.

Xenoliths were chosen for size and freshness, cut out of the host basalt, and made into 200-μm-thick, doubly polished sections for analytical work (as well as 30-μm-thick, doubly polished thin sections for petrographic analysis). Samples were analyzed for major and trace elements using laser ablation–inductively coupled plasma–mass spectrometry (LA-ICP-MS) using a ThermoFinnigan Element 2 ICP-MS at Rice University. Minerals in the thick sections were ablated using a 213 nm wavelength NewWave laser, pulsing at 10 Hz with a 55 μm spot size at 20 J/cm2. Minerals analyzed were selected and targeted with the laser microscope viewfinder to avoid grain boundaries, inclusions, and cracks. We performed one medium- (major elements) and one low-mass-resolution (trace elements) analysis on the core of each mineral analyzed. For analyses, 10–12 cycles of background signal were recorded prior to firing the laser, and 30–40 cycles of signal were recorded during ablation. Drift effects and decay of the signal were corrected for using 25Mg as an internal standard (Longerich et al., 1996). Three external basalt standards BHVO2 g, BIR1 g, and BCR2 g were simultaneously used for concentration calibrations. The ICP-MS and laser gases (helium) were tuned before analysis of standards and samples to attain a sensitivity of >600,000 cps for 15 ppm La in BHVO2 g (10 Hz, 55 μm spot size). Major-element concentrations were estimated from medium-resolution measurements by assuming all major and minor elements (Si, Mg, Ca, Fe, Mn, Ti, Na, K, Al) were oxides and that the oxides summed up to 100%. The resulting Mg content was then used as the internal standard for converting low-mass-resolution trace-element signals to concentrations.

Figure 2 is a typical example of the raw data. Signal intensities relative to Mg were constant during ablation for all elements except Rb, Cs, and Ba. Constant Mg-normalized signal intensities for a given element indicate homogeneity of that element on the time scales (and hence ablation depth) of ablation (~200 μm). Variable Mg-normalized signal intensities with time indicate heterogeneity. For example, in Figure 2A, the signal for Cs is orders of magnitude higher in the first half of ablation interval. Condie et al. (2004) have shown that large ion lithophile elements (LILEs) tend to accumulate along grain boundaries and in microcracks. We assume this to be the cause for variable intensities of Cs, Rb, and Ba, as opposed to surface contamination associated with sample preparation, because elevated Cs, Rb, and Ba signals also occur as pulses during ablation, not just at the start of ablation. These occasional pulses of Cs, Rb, and Ba are most evident in those cpx grains that have a “spongy” texture and thus may be related to microcracks or inclusions. No anomalous pulses were observed for Th, U, or Pb, which are often thought to behave similarly to Cs, Rb, and Ba.

Signal quality was ultimately assessed by examining time-resolved plots of normalized elemental intensities. Figure 2B shows that internally normalized ratios are constant and do not cross each other, indicating homogeneity for most elements. In the case of Cs, Rb, and Ba, their normalized intensities are not constant and therefore cross other normalized elements. We eliminated that part of the signal that was not parallel to the other elements. Nb and Zr in spinels suffered from CrAr isobaric interferences (Eggins et al., 1998) and were therefore discarded (in any case, Nb and Zr contents in peridotitic spinels are generally low and do not contribute a significant budget to the whole-rock Nb and Zr contents).

In Table 1, we present the modes, major oxide weight percents, and trace-element concentrations in ppm for mineral phases in each sample. These values reflect the average of all measurements run on the respective phase for that sample. In Table 2, we report the mineral modes, major oxide weight percents, trace-element concentrations in ppm, and calculated equilibration temperatures of the whole rock. Whole-rock compositions were estimated by summing up the mineral compositions according to their estimated proportions. In the case of 02F and 06E, spinels were too small to be analyzed, so the average composition of spinels in the other samples was used to reconstruct the major oxides for the bulk rock. We did not account for spinel in our trace-element reconstructions for the bulk rock because of its negligible effect on bulk-rock chemistry for all trace elements (except for Cr).


Temperatures of equilibration were estimated from Mg-Fe exchange between clinopyroxene and orthopyroxene using the method of Wells (1977) and Brey and Köhler (1990) after normalizing mineral major-element oxide compositions to 100 wt%. For the Brey and Köhler thermobarometer, a reasonable range of pressures for spinel lherzolites (2.0–2.5 GPa) was assumed. Our calculations indicate equilibration temperatures between 860 and 1142 °C using the Well's thermometer and between 820 and 1090 °C with the Brey and Köhler thermometer (one sample gave an anomalously low Brey and Köhler temperature of 786 °C) (Table 2). The average temperature of equilibration is 978 °C according to the method of Wells (1977) and 957 °C according to the method of Brey and Köhler (1990). A direct estimate of pressure was not possible because of the lack of appropriate mineral assemblages (e.g., garnet) for barometry.

Crude limits on pressure were determined by using the thermobarometric temperatures and fitting them to model continental geotherms (Fig. 3). The present-day surface heat flow in the region is ~50 mW/m2 (Pollack et al., 1993; Woodruff and Foley, 1985), resulting in a depth of ~70 km for the minimum calculated temperature of 786 °C. However, given the large range in temperatures, these xenoliths could have equilibrated at pressures and depths approaching 3 GPa, but, due to the lack of garnet in these samples, they are unlikely to come from depths greater than ~85 km (Klemme, 2004). If heat flow was appreciably greater in the Cretaceous (70 mW/m2), then the depth of equilibration could have been as shallow as 40 km, although this is unlikely given the passive nature of the margin at that time. We thus conclude that these xenoliths equilibrated at ~950 °C and at depths between 70 and 80 km. The pressure and temperature values of xenoliths appear to be associated with lithosphere that is thinner and warmer than typical Archean cratonic lithosphere, but significantly cooler than Phanerozoic lithospheric mantle, such as in the Basin and Range Province.

Major- and Minor-Element Chemistry

Mineral chemistries show similarities to continental lithospheric mantle samples. For example, Mg# ranges (molar Mg/[Mg + Fe]) of olivine, clinopyroxene, orthopyroxene, and whole rock are 0.89–0.91, 0.91–0.94, 0.89–0.91, and 0.89–0.91, respectively, and are slightly higher than that of primitive-mantle olivines (~0.89) but not as high as Archean cratonic mantle olivines (0.91–0.93). Al2O3 and Na2O contents of pyroxenes are low, and the Ni contents of olivines are high, which are typical of continental peridotite xenoliths. These high Mg# values, high Ni contents, and low Al contents all suggest that these minerals are derived from refractory peridotites, such as harzburgites and clinopyroxene-poor lherzolites, though the possibility of some amount of refertilization cannot be excluded.

Whole-rock compositions, estimated by reconstructing mineral modes and compositions, are also consistent with a refractory origin. Compared to fertile lherzolite, such as model primitive-mantle compositions, the Knippa peridotites have high whole-rock MgO and low CaO, Na2O, and Al2O3 values (Figs. 4A–4C). Error bars account for variations in estimated modal abundances, which far outweigh errors associated with mineral compositions. In a plot of CaO versus MgO (Fig. 4A), the Knippa peridotites fall within the same trend as seen in abyssal peridotites and peridotite xenoliths from Phanerozoic and Proterozoic continental lithosphere. There is no evidence from mineral modes or whole-rock chemistry to indicate that any of these samples are anomalously rich in Si, as is seen in some Archean cratonic mantle peridotites. The low Ca, Al, and Na contents and the negative correlation of these oxides with MgO are consistent with the Knippa peridotites representing the residues of previous melt depletion.

Trace-Element Chemistry

Primitive mantle–normalized (McDonough and Sun, 1995) trace-element spidergrams for reconstructed whole rocks are presented in Figure 5A and are plotted in order of increasing compatibility in the solid peridotite residue. Included compositions in these whole-rock reconstructions are the trace-element compositions of clinopyroxene, orthopyroxene, and olivine, but not spinel. The modal abundance of spinel is small, and the concentrations of trace elements in spinel are also small, so the effect of spinel on bulk trace-element compositions is negligible (Cr is an exception: spinel dominates the whole-rock Cr budget). In some cases, we were not able to measure the highly incompatible element abundances in olivine (e.g., light rare earth elements [LREEs]) because they were below detection limits, but the abundances of these elements are so low in olivine that they do not contribute to the bulk-rock composition. Thus, for the most part, the whole-rock trace-element budget is controlled by orthopyroxene and clinopyroxene.

A key feature of the whole-rock spider-grams is the depletion in moderately incompatible trace elements, such as the heavy rare earth elements (HREEs), relative to primitive mantle. In general, moderately incompatible elements, such as the HREEs, are more resistant to metasomatic disturbances (Canil, 2004; Lee et al., 2007a); hence, the HREE depletions likely reflect melt depletion. The HREE depletion also correlates with major-element indices of depletion, as can be seen by the correlation of Ca versus Lu (Fig. 4D).

Melt depletion, however, should result in positively sloping REE spidergrams (e.g., LREE depletion relative to HREEs) due to more efficient extraction of highly incompatible elements. However, “U-shaped” trends are seen, as exemplified by the relative enrichment in LREEs. High La/Sm and Sm/Yb ratios in our samples are reminiscent of metasomatized mantle xenoliths in subarc or post-Archean continental settings, as well as enriched magmas such as arc magmas (Fig. 6). They do not match the highly LREE-depleted signatures of oceanic abyssal peridotites (Niu, 2004), nor do they match typical mid-ocean-ridge basalts (MORBs), which are also depleted in LREEs relative to HREEs. Carbonatite-metasomatized xenoliths (Yaxley et al., 1991) have much steeper LREE enrichment patterns than the Knippa peridotites. These observations suggest that these xenoliths were cryptically re-enriched in incompatible trace elements after the melt depletion event that gave rise to their refractory major-element compositions.

We also examined the high field strength element (HFSE) abundances: Nb, Ta, Zr, Hf, and Ti. It is generally thought that depletion in these elements relative to the REE partner that exhibits the most similar bulk partition coefficients during anhydrous peridotite melting is a sign of hydrous fluid metasomatism because the HFSEs are considered to be much less fluid mobile than the REEs. Relative depletions in these elements occur in our clinopyroxene grains, but these relative depletions in clinopyroxene do not by themselves imply metasomatism by HFSE-depleted fluids/melts, since such depletions commonly occur in clinopyroxene grains from other mantle xenolith suites (much of the whole-rock HFSE budget is also housed in orthopyroxene). Thus, whole rocks often show little HFSE depletion even though clinopyroxenes are HFSE depleted. However, this is not the case in the Knippa samples. In all but two of the most refractory samples (01A and 08A), the reconstructed whole-rock HFSEs show relative depletions. For example, whole-rock Nb and Ta are depleted relative to La; Zr and Hf are depleted relative to Sm; and Ti is depleted relative to Eu. These HFSE depletions are typically found only in arc magmas or in mantle xenoliths that have been metasomatized by fluids (Kelemen et al., 1993). Because of the low fluid solubilities of HFSEs compared to REEs, fluid metasomatism typically yields HFSE-depletion signatures, whereas no HFSE anomalies are generated during silicate melt metasomatism.

Large ion lithophile elements (LILEs), such as Ba, Cs, and Rb, were more difficult to reconstruct because of their low concentrations in silicate minerals. We observed enrichments in these elements in some clinopyroxenes. Figure 7 shows the primitive mantle–normalized concentrations of trace elements in clinopyroxenes from sample 05A. As discussed in the petrography section, two textures of clinopyroxenes exist, the more common being the “turbid” clinopyroxene. Turbid clinopyroxenes show variable but high concentrations of Cs, Rb, and Ba. The clear-cored clinopyroxenes of 05A (two shots, a and b) have much lower concentrations of LILEs. In addition, we also found that the rims of clinopyroxenes showed enrichment in these elements compared to the core; thus, it is not clear what fraction of these enrichments is due to ancient fluid-metasomatic events or more recent disturbances associated with the magmatic precursor to the host magma.

AssessingTrace-Element Equilibrium

An important question regarding trace-element compositions of mantle xenoliths is whether the compositions are actually representative of the mantle or if they represent recent perturbations associated with contamination by the host magma. One way to address this problem is to assess whether the mineral chemistries have equilibrated. If the mineral chemistries have been recently disturbed, they should be out of equilibrium. Using extension of lattice strain theory to mineral-mineral element partitioning, Lee et al. (2007a) and Agranier and Lee (2007) showed that when equilibrium exists between mantle pyroxenes, there is a nearly log-linear variation in the ratio C(opx)/C(cpx) with increasing ionic radius of the lanthanide series. Samples that have been contaminated or recently metasomatized show deviations from this log-linear trend. Figure 8 demonstrates that there is, for the most part, a log-linear variation in the concentrations of REEs in orthopyroxene over clinopyroxene for the Knippa peridotites. These observations indicate that, at least for the REEs, equilibrium has been achieved, and the reconstructed whole-rock REE patterns are probably representative of the mantle. While we have not been able to do the same for the HFSEs, we assume that the high degree of equilibrium seen in the REEs hints that the HFSEs are also in equilibrium.

Geochemical Signature of Subduction

The Knippa xenoliths have whole-rock and mineral chemistries characteristic of the sub-continental lithospheric mantle. Equilibration temperatures of 800–1100 °C are too cool to be associated with hot plumes or asthenospheric mantle. Interpolating from model geotherms, these temperatures suggest that the xenoliths derived from depths of ~70–80 km, which implies that the base of the thermal boundary layer beneath Knippa could have been as thick as ~100 km (taking ~1300 °C to represent the temperature at the base of the thermal boundary layer). The enrichments of the LREEs and LILEs over middle and heavy REEs are inconsistent with fertile asthenosphere or oceanic mantle origin and instead require that these peridotites were cryptically metasomatized sometime after they were formed by melt depletion (as in Fig. 6). One important factor is the depletions in HFSEs relative to the REEs; these negative HFSE anomalies suggest the involvement of fluids during metasomatism and are often taken to imply an arc origin (Hofmann, 1988; Kelemen et al., 1993; Ryerson and Watson, 1987). Similar features have been seen in mantle xenoliths from the Sierra Nevada (Lee, 2005) and Canadian Cordillera (Peslier et al., 2002).

We envision the origin of the geochemical signatures as follows. Melt was originally extracted from these peridotites, resulting in low Ca, Al, and HREE contents and high Mg# and MgO. Based on the depletions in Ca, Al, and HREE contents with respect to an assumed fertile mantle starting composition, we estimate that these peridotites had ~15%–20% melt extracted (using experimental compositions of Walter [Walter et al., 1995, 2003] and parameterizations from pMELTS; Ghiorso and Sack, 1995). Such a melting event should generate much more depletion in LREEs and HFSEs relative to the HREEs (Fig. 5B). Thus, after the original melting event, small degrees of fluids or melts had to be added into the residual peridotite. In Figure 5B, we demonstrate the different effects of mixing melt-depleted peridotites with melts having no arc signature (MORB or ocean-island basalt [OIB]) and melts/fluids with a HFSE depletion. For the REEs, only small amounts of melt/fluid addition (0.5%–2.0%) are needed to match the LREE enrichment levels (note that HREEs remain unaffected). However, only the addition of HFSE-poor melts/fluids can generate the negative HFSE anomalies in the Knippa peridotites. The Knippa peridotites most likely represent fragments of subcontinental lithospheric mantle that were metasomatized by arc-like fluids. In contrast, the relatively flat REE patterns and lack of pronounced HFSE anomalies in samples 01A and 08A suggest metasomatism by silicate melts rather than flu-ids. These samples were metasomatized a second time (by silicate melts), which overprinted an original arc signature. Alternatively, these samples were never metasomatized by fluids. The metasomatic history of the lithospheric mantle beneath Knippa appears to be heterogeneous. However, what is important here is that many of the xenolith samples preserve an arc or subduction signature, begging the question of when this signature was imparted. Next, we consider the origin of fluid metasomatism in the context of various tectonic scenarios to better understand the relationship between the lithospheric mantle and crustal basement in the area.

When Was the Subduction Signature Imparted?

Tectonic Models

In this section, we discuss several scenarios for the origin of the subduction metasomatic signatures seen in the Knippa xenoliths. Since Knippa is thought to occur on the structural edge of the Laurentian craton (Buffler and Sawyer, 1985; Gao et al., 2008; Hatcher et al., 1989; Keller et al., 1989; Mosher, 1998; Viele and Thomas, 1989), two orogenic events that could have imparted subduction signatures are the Grenvillian and Ouachitan orogenies. We will also discuss the possibility that low-angle subduction of the Farallon plate could have also imparted subduction or fluid signatures. We provide a brief tectonic summary of each of these tectonic episodes and then discuss the tectonic episode that is most likely to have been responsible for the metasomatism seen in the Knippa xenoliths.

Scenario 1: Grenville Orogeny (Proterozoic)

The Grenville orogeny, occurring ca. 1 Ga, is recognized worldwide, and is associated with the formation of the supercontinent Rodinia (e.g., Dalziel, 1991; Hoffman, 1991; Moores, 1991; Rivers, 1997; Tollo et al., 2004; Torsvik et al., 1996). Precambrian lithosphere in Texas is found in the Llano Uplift, ~150 km northeast of Knippa. The main tectonic features in the Grenville province are continental arc magmatism and intense collisional deformation. A seismic study of the Grenville province in Ohio revealed a west-dipping suture, which is evidence for subduction beneath the Lauren-tian craton (Culotta et al., 1990). In addition, the oldest terrane of the Llano Uplift—the 1366–1232 Ma Valley Spring gneiss—is interpreted to be a Laurentian continental arc terrane based on structural and petrologic relationships (Mosher, 1998). The tectonic features seen in Quebec also indicate that the Laurentian craton was intruded by continental arc magmatism prior to collisional orogenesis (Rivers et al., 1993; Tollo et al., 2004). These observations place a convergent-destructive plate boundary along the southern and eastern margin of Laurentia, recording a progression from continental arc magmatism to arc-continent collision and finally continent-continent collision (Mosher, 1998; Mosher et al., 2008). It has been suggested that the subduction polarity in the Llano area inverted toward the south along an E-W–trending margin sometime before collision (Mosher et al., 2008), but overall NW-vergent subduction affecting a wide area of the continent dominated the Grenvillian event in southern Laurentia and could have metasomatized the mantle in the Knippa region. Alternatively, the existence of an exotic island arc in the Llano Uplift (Roback, 1996) and the inversion of subduction polarity at the end of the Grenville event toward the south could also imply that the Knippa region was attached to a “southern continent,” which served as the overriding plate of a Grenville-aged subduction zone with a southward polarity (S. Mosher, 2009, personal commun.). Under either scenario, the presence of continental arc magmatism in Proterozoic Texas and the general association of the crustal Grenville province with arc magmatism could thus explain the observed fluid-metasomatized signature of the Knippa mantle xenoliths.

Scenario 2: Ouachita Orogeny (Paleozoic)

During the Paleozoic, the closing of the Iapetus and Rheic Oceans assembled the super-continent of Pangea. Models for the Iapetus and Rheic Oceans are debated, but the following interpretation is preferred: structural evidence for periods of both east- and west-directed subduction have been proposed for the assembly of eastern Laurentia (Hatcher et al., 1989). Plutonic rocks occur throughout the eastern Appalachians, particularly toward the north (Hatcher et al., 1989), but little evidence exists that the subduction regime was actually “continental” in the southern Appalachians. In addition, based on the high burial depths of the Laurentian margin beneath the exotic Avalon terrane in the southeastern Appalachians, it is likely that Laurentia served as the underlying plate along much of the final suture with Gondwana during the Paleozoic orogenies (Keller and Hatcher, 1999). The east-dipping suture of Gondwana to Laurentia following closure of the Rheic Ocean is recognized in both seismic-reflection and magnetic anomaly data, marking a major crustal boundary in Florida (Buffler and Thomas, 1994).

In contrast to the eastern margin, the southern margin of Laurentia appears to have interacted with the Rheic Ocean only because North America rotated clockwise to close the Iapetus (Dalziel, 1997; Murphy and Nance, 2008; Scotese et al., 1990), and it has experienced a more limited accretionary history (Viele and Thomas, 1989). The Rheic Ocean is proposed to have closed primarily beneath Baltica on one side and South America on the other (Murphy et al., 2008). The polarity of subduction has been debated for the Ouachita belt, but several lines of evidence make a compelling argument for south-directed subduction. The Ouachita orogenic belt in southern Laurentia is devoid of arc materials, or volcanic strata of any kind other than fine ash falls (Wickham et al., 1976), while strong evidence exists for long-lived continental arc magmatism on the opposite (Andean) margin (Dalziel et al., 1994) and in Mexico's Acatlàn Complex (Cameron et al., 1992; Jones et al., 1986; McKee et al., 1988). The sedimentary and structural history of the Ouachita belt shows that a deep-water accretionary wedge was thrust over the Laurentian continental shelf and its Proterozoic basement ahead of arc terranes with Gondwanan affinities above south-directed subduction (Poole et al., 2005). Furthermore, the sedimentary package of the Ouachita belt records an increasing influence in time of volcanic source material from the southwest in deep-water shelf sediments, suggesting the approach of an overriding arc from the southwest above south-directed subduction (Mack et al., 1983; Wickham et al., 1976). A geophysical cross section of the Gulf Coast region by Mickus and Keller (1992) recognized a south-dipping suture south of the metamorphic core of the Ouachita belt. The lack of metamorphism in the Ouachita zone is evidence that the suture for the Rheic Ocean in this part of Laurentia was far removed (Nance and Linneman, 2008).

To summarize, most of the geologic and geophysical evidence suggests that the subduction associated with the southern Alleghenian and Ouachita orogenies was directed away from the margin of southern Laurentia; that is, southern Laurentia during this time represented the under-thrust plate, not the overriding plate. If so, it would seem unlikely for the subduction-related metasomatic signatures in the Knippa xenoliths to be associated with Paleozoic orogeny.

Scenario 3: Flat Subduction of Farallon Plate during Cenozoic Laramide Orogeny

A third subduction scenario is the flat subduction of the Farallon plate eastward during the Laramide orogeny (74–40 Ma). The low-angle slab may have reached as far east as the Colorado Plateau (Lee, 2005), and so the possibility of the slab reaching even farther east to Knippa should be considered. However, Ar-Ar dates of the Knippa basalts suggest that they erupted well before the apogee of flat subduction (Miggins et al., 2004). Geochemical evidence from arc basalts suggests that dehydration of the slab beneath the arc is highly efficient (>90%) (Shaw et al., 2008), so it is unlikely that much fluid would survive transport as far east as Knippa to cause the observed signature in the Knippa xenoliths.

Additional Discussion on the Origin of the Subduction-Zone Geochemical Signatures

Based on the proposed geometries for the Proterozoic and Paleozoic orogenies described here, we speculate that the observed geochemical signature is best explained by the Proterozoic subduction zone and mantle wedge. This interpretation is consistent with isotopic studies of mantle and lower-crustal xenoliths from the Prairie Creek lamproites in Arkansas, which are contemporary, and which are assumed, on the basis of structural relationships and gravity profiles, to have similar source structural terranes as the Knippa basalts. Osmium (Lambert et al., 1995) and K-Ar (Dunn, 2008) dates of the xenoliths from Prairie Creek point to a lithospheric mantle source with a Proterozoic separation age of 1.2–1.4 Ga and a lower crust with Laurentian affinities and an age of 1.4 Ga. The occurrence of Proterozoic lower-crustal xenoliths is strong evidence that the Ouachita trough is attenuated Proterozoic Laurentian basement and not a remnant ocean basin (Dunn, 2008). These observations and our interpretation of the Knippa xenoliths suggest that lithosphere formed prior to the Grenville collision has been preserved beneath the Gulf Coast despite subsequent orogenies and rifting events, including the latest opening of the Gulf of Mexico. It has been proposed that lithosphere along the Texas-Louisiana coast has been significantly thinned or attenuated (Dunbar and Sawyer, 1987). Rifting may have also removed a large area of Proterozoic basement from the midcontinent region (Thomas, 1991; Thomas and Astini, 1996), which further raises questions about the integrity of the deep lithosphere beneath the Laurentian margin through time. The preservation of Proterozoic mantle lithosphere along Laurentia's southern margin has relevance to the dynamics and process of subcontinental lithospheric mantle stabilization in an orogen, which we discuss next.

Fate of Lithospheric Mantle in an Orogenic Belt

The growth and stabilization of continents have long been thought to be related to convergent margins. For example, along convergent margins, subducting oceanic lithosphere drives juvenile arc magmatism, while relative plate motions accrete island-arc material to the continental margins (Jordan, 1978; Kelemen, 1995; Kelemen et al., 1993; Lee et al., 2007b; Ringwood, 1975; Rudnick, 1995). However, convergent margins also appear to be regions where large-scale convective removal (e.g., “delamination”) of the subcontinental lithospheric mantle is thought to occur, as is often depicted in cartoons or invoked to drive postorogenic magmatism, uplift, and recycling of MgO-rich cumulates (Bird, 1979; Houseman et al., 1981; Kay and Kay, 1993; Rudnick, 1995). Postorogenic or anorogenic 1.1–1.0 Ga “A-type” granites occur in the Grenville province (Mosher, 1998; Walker, 1992). Exposure of medium-temperature eclogites in the Llano Uplift (Wilkerson et al., 1988) and deep crustal exhumation in Quebec (Rivers et al., 1993) suggest exhumation of half the thickness of the orogen. Mosher et al. (2008) suggested that lithospheric delamination may have driven this uplift. Such destruction interferes with the arc-accretion processes that lead to stable, thick, and buoyant cratons.

Many have suggested that after continent-continent collisions, lithospheric mantle in general will be delaminated (Bird, 1979; Corrigan and Hanmer, 1997; Davies and Vonblanckenburg, 1995; Houseman et al., 1981; Turner et al., 1996). However, the preservation of subduction-zone geochemistry in the mantle beneath Grenville-age rocks in southern Laurentia suggests that parts of the original mantle root were not delaminated during the Grenville or Alleghenian orogenies and may have instead been preserved for over 1 b.y.

This conclusion seems somewhat surprising at face value because the Knippa xenoliths are not as refractory as many Archean cratonic mantle peridotites, the densities of which are consequently so low that it is difficult to destabilize them using buoyancy considerations alone. The Mg# values of the Knippa xenoliths are similar to other post-Archean lithospheric mantle, and hence their chemical buoyancies are not likely to be as high as typical cratonic mantle peridotites. Therefore, their preservation suggests that density alone is not controlling the long-term preservation of continental lithospheric mantle. One possibility is rheology. The viscosity of the Laurentian margin could have been high, either because the lithospheric mantle had been dehydrated during partial melting or the lithospheric mantle had already stabilized, cooled, and strengthened (Hirth and Kohlstedt, 1996). The Knippa xenoliths show subduction-related signatures, which imply that the nominally anhydrous minerals (olivine and pyroxenes) may have been rehydrated, and, if so, one would expect much of the lithospheric mantle to have been weakened and possibly removed. The original lithosphere thickness along the southern Laurentian margin could have been at least ~150 km thick based on the presence of diamonds in the Prairie Creek lamproite in Arkansas. The fact that the present lithosphere thickness beneath Knippa is <80 km suggests that the lower half may indeed have been progressively removed. The upper part of the lithosphere, being colder, may have escaped delamination because it was just too cold to deform. This leads us to speculate that continental lithospheric mantle, however it is formed, is relatively resistant to being recycled completely into the mantle, but the mechanism of preservation of such lithosphere still remains elusive.

Mantle xenoliths erupted through the lithosphere along the Texas coast are characteristic of continental lithosphere and show strong evidence for metasomatism deficient in high field strength elements, suggesting a relationship with arc- or subduction-related fluids. We favor an association of this signature with the Proterozoic Grenville orogeny based on structural relationships and evaluation of tectonic models. The preservation of metasomatized material implies that part of the subcontinental lithospheric mantle remained in place (stable) following orogeny and did not delaminate everywhere.

We thank Julia Morgan for discussions and help in the field. This work was partially supported by the National Science Foundation and Packard grants to Lee.