Attribution: You must attribute the work in the manner specified by the author or licensor (but no in any way that suggests that they endorse you or your use of the work).Noncommercial ‒ you may not use this work for commercial purpose.No Derivative works ‒ You may not alter, transform, or build upon this work.Sharing ‒ Individual scientists are hereby granted permission, without fees or further requests to GSA, to use a single figure, a single table, and/or a brief paragraph of text in other subsequent works and to make unlimited photocopies of items in this journal for noncommercial use in classrooms to further education and science.


Mantle xenoliths from the Pliocene Rio Puerco volcanic field, New Mexico, record chemical modification of the subcontinental lithosphere in the transition zone between the Colorado Plateau and the Rio Grande rift. The Rio Puerco necks brought a wide variety of mantle xenoliths to the surface, including abundant spinel ± garnet pyroxenites. Garnet pyroxenites equilibrated at pressures within the spinel lherzolite stability field (16–18 kbar) but at higher temperatures (1020–1100 °C) than lherzolite xenoliths (900–1000 °C). Calcite-bearing pyroxenites record temperatures as high as 1145 °C. Textural observations combined with the high equilibration temperatures are consistent with pyroxenite formation in response to melt interaction with a lherzolite precursor in the approximate ratio of 35% lherzolite to 65% melt. Bulk chemical data and the presence of both carbonatitic and basaltic glass inclusions indicate that the infiltrating melt(s) included both carbonatite and silicate components. Different mantle xenolith populations in volcanic centers of different ages in the Rio Grande rift constrain the likely age of melt infiltration and pyroxenite formation beneath the Puerco necks to Miocene–early Pliocene. Orthopyroxene + spinel + clinopyroxene + silicate glass symplectites around relict garnets record continued rift-related decompression ± heating following pyroxenite formation. Carbonatite + silicate melt generation was likely concentrated in areas affected by earlier metasomatism related to devolatilization of the subducted Farallon slab. Deep-seated fractures associated with the Jemez Lineament may have focused melts beneath the Puerco necks and resulted in wholesale conversion of lherzolite into pyroxenite.


Mantle xenoliths have long been used to document the composition, physical properties, and tectonic evolution of the subcontinental lithosphere beneath the western United States. Most of these studies have focused on various types of olivine-bearing nodules, but many have also noted the presence of pyroxenites. In general, pyroxenites appear to be most prevalent in xenolith suites from the extended lithosphere of the Basin and Range Province, eastern Sierra Nevada, Arizona Transition Zone, and southern Rio Grande rift (Wilshire et al., 1988; Ducea and Saleeby, 1996; McGuire and Mukasa, 1997). Xenolith suites from the Colorado Plateau also contain pyroxenites, but in lesser abundance (McGetchin and Silver, 1972; Wilshire et al., 1988). The association of pyroxenites with regions of extension suggests a genetic link between asthenospheric upwelling and pervasive development of mantle pyroxenites.

Mantle pyroxenites have been proposed as possible source rocks involved in generation of both alkalic basalts (Carlson and Nowell, 2001; Hirschmann et al., 2003; Kogiso et al., 2003) and mid-ocean-ridge basalts (Hirschmann and Stolper, 1996). However, the processes by which the pyroxenites themselves form are not well understood. End-member models fall into several distinct categories; pyroxenites might represent: preserved remnants of subducted oceanic crust mixed into the mantle or melts derived from such material (Xu, 2002; Pearson and Nowell, 2004), primary melt compositions that crystallized within the upper mantle (Ho et al., 2000), cumulates derived from alkali basaltic magmas (McGuire and Mukasa, 1997; Litasov et al., 2000; Dobosi et al., 2003; Lee et al., 2006), metasomatic rocks that resulted from interaction of reactive fluids and/or melts with mantle peridotites (Irving, 1980; Wilshire et al., 1988; Mukhopadhyay and Manton, 1994; Garrido and Bodinier, 1999; Liu et al., 2005), or some combination of these models (Downes, 2001; Sen et al., 2005).

Several types of pyroxenite xenoliths, along with spinel lherzolites and harzburgites, were brought to the surface in abundance by the Rio Puerco volcanic necks in northwestern New Mexico. These volcanic necks are located in the transition zone between the Colorado Plateau and the extended lithosphere of the Rio Grande rift, a region where physical and chemical modification of the uppermost mantle is likely to have occurred since the onset of rifting. We use textural and chemical data to evaluate models for the formation of the pyroxenites. The pyroxenite samples preserve a more extensive reaction history than is evident in associated lherzolites and, hence, are particularly useful for examining the thermal and chemical modification history of the upper mantle during Rio Grande rifting. Primary mantle carbonates are also present in many of the pyroxenites and provide direct evidence for rock interaction with fluid and/or melt at depth in the mantle.


The xenoliths for this study come from Cenozoic volcanic necks in the Rio Puerco drainage of west-central New Mexico (Fig. 1). The ∼50 volcanic necks of the Rio Puerco volcanic field are small-volume, mafic plugs composed of alkaline basalts and volcanic breccias that are part of the adjacent Mt. Taylor volcanic field (Kudo et al., 1972; Hallett, 1994; Hallett et al., 1997). The xenolith-bearing basalts have been dated at 4.49–2.05 Ma (Hallett et al., 1997), and the Mt. Taylor volcanics are comparable in age at 3.0–1.5 Ma (Perry et al., 1990; Hallett et al., 1997). Both the Rio Puerco and Mt. Taylor volcanic fields occur within the Jemez Lineament, a NE-trending zone of late Tertiary–Quaternary volcanic centers that extends from southwestern Arizona to northeastern New Mexico (Mayo, 1958; Aldrich and Laughlin, 1984). The lineament corresponds to the tectonophysiographic transition zone that separates the thick lithosphere of the southeastern Colorado Plateau from the extended lithosphere of the western margin of the Rio Grande rift and the southeastern Basin and Range Province.

Extension began in the southern portion of the Rio Grande rift around 30 Ma and propagated northward. Volcanism in the rift has been intermittent since the Tertiary, and basalt compositions reflect transition from a lithospheric to an asthenospheric source driven by lithospheric thinning (Perry et al., 1988; Baldridge et al., 1991). This transition in source regions occurred at ca. 10 Ma in the southern part of the rift and in the last 4 m.y. in the central rift on the basis of εNd and 87Sr/86Sr ratios (McMillan et al., 2000).

The onset of Rio Grande rift extension may have been initiated by the removal of the east-dipping Farallon slab (Dickinson and Snyder, 1978; Humphreys, 1995; Keller and Baldridge, 1999). Low-angle subduction of the Farallon slab, possibly in contact with the base of the lithosphere under the western United States, resulted in cooling and hydration of the North American lithosphere (Coney and Reynolds, 1977; Riter and Smith, 1996; Smith et al., 1999, 2004; Humphreys et al., 2003). Slab removal likely occurred between 50 and 20 Ma, based on the ages of mantle-derived volcanism thought to have resulted from slab breakoff (Humphreys, 1995).

The Rio Puerco volcanic necks obliquely span the width of the Jemez Lineament and provide an opportunity to examine the mantle lithosphere beneath this feature. Some workers have argued that the lineament is coincident with a Proterozoic province boundary at depth (Karlstrom and Humphreys, 1998; Shaw and Karlstrom, 1999), which may have been reactivated during subsequent tectonism (Dickinson and Snyder, 1978; Humphreys, 1995; Keller and Baldridge, 1999). Seismic data support a deep-rooted structure beneath the lineament (Humphreys and Dueker, 1994; Dueker et al., 2001; Gao et al., 2004), but the physical and chemical significance of the lineament at depth remains unknown.

The LA RISTRA passive seismic line places important constraints on the structure of the upper mantle in this region. West et al. (2004) documented lithospheric thicknesses ranging from 200 km in the Great Plains to 45–55 km beneath the Rio Grande rift and 120–150 km for the Colorado Plateau and argued that slow seismic velocities beneath the rift indicate the presence of partial melt. Gao et al. (2004) attributed a fast anomaly below the Colorado Plateau to the foundering Farallon slab and a slow anomaly just to the east of the plateau as upwelling asthenosphere associated with volatiles released from the slab.


We collected more than 500 xenoliths from 11 of the roughly 50 volcanic necks in the Rio Puerco area. The xenoliths range from 2 to 30 cm in diameter and are found within basaltic plugs, dikes, and tuff rings. The mantle xenolith suite is dominated by spinel lherzolites and pyroxenites, with rare harzburgites, dunites, and megacrysts of spinel and pyroxene. An unusual feature of this xenolith suite is the presence of apparently primary mantle carbonate in many of the samples (Perkins et al., 2006).

In this paper, we focus primarily on xenoliths from the volcanic neck of Cerrito Negro (CTON, Fig. 2A), but also include data from other necks within the volcanic field. Cerrito Negro is dominated by pyroclastic breccia that contains abundant lherzolite and pyroxenite xenoliths (Fig. 2B) in addition to bombs, scoria, and inclusions of the underlying sedimentary country rock (Hallett, 1994). Pristine pyroxenite xenoliths up to 30 cm in diameter weather free of the breccia and display well-rounded and polished surfaces. Although pyroxenites are common at many of the Puerco necks, Cerrito Negro is the only neck at which we found garnet pyroxenites. Several composite lherzolite-pyroxenite xenoliths are exposed in situ within the breccia (Figs. 2C and 2D), although we were unable to obtain intact samples of these xenoliths due to their weathered condition. In all cases where crosscutting relationships are apparent in the composite xenoliths in outcrop, spinel pyroxenite crosscuts lherzolite. However, the contacts between pyroxenite and lherzolite are typically gradational over a distance of millimeters to a few centimeters, suggesting considerable interaction between the two rock types and/or formation of the pyroxenites by reaction between lherzolite and melt/fluid of a different composition. There is no obvious spatial association between pyroxenite in composite xenoliths and the surrounding host basalt.

Mineral and melt inclusion compositions were determined using the JEOL 8200 electron microprobe at the Department of Earth and Planetary Sciences at the University of New Mexico. Quantitative data for this study were obtained from 16 samples. Samples were analyzed with an accelerating voltage of 15 kV, beam diameter of 1 μm, and a beam current of 20 nA. All iron was analyzed as FeO and recalculated for thermobarometric calculations using the method of Droop (1987).

Spinel Lherzolite

Most of the lherzolite samples from the pyroclastic breccia at Cerrito Negro are extensively weathered, and we thus included representative lherzolites from basalts at Cerro de Guadalupe (e.g., CG-38 in Fig. 3), Cerro de Santa Rosa, and Cerro Negro in this study. Lherzolites from all four necks show a large degree of grain-size heterogeneity. Most samples lack an obvious shear foliation, but deformation lamellae are ubiquitous in olivine, and orthopyroxene generally shows undulose extinction and local subgrain development. A foliation defined by a shape-preferred olivine and pyroxene fabric and by spinel-rich layers is present in a few samples. Several samples, such as CTON-57 (Fig. 3), contain large pyroxene megacrysts with coarse exsolution lamellae that are set in a fine-grained lherzolite matrix. Carbonate is locally present along grain boundaries and in melt pockets within lherzolite samples.

Orthopyroxene is more abundant than clinopyroxene in most lherzolite samples, suggesting that most samples experienced at least some partial melting and melt removal. Exsolution lamellae in the pyroxenes range from very fine to coarse (<10–100 μm) within the same sample. Fine lamellae are typically confined to the interiors of subhedral orthopyroxene grains and do not extend to the rims of the grains. In most samples, brown spinels are either optically homogeneous or are zoned toward opaque rims. A few samples, however, display narrow symplectitic rims composed of olivine, pyroxene, and plagioclase around spinel. Phlogopite occurs in several lherzolite samples, generally as inclusions within pyroxene. Calcite occurs along grain boundaries and associated with melt pockets in a few samples. Fluid and melt inclusions are locally present along healed fractures within the lherzolites.

Clinopyroxene is dominantly Cr-rich diop-side with 5–6 wt% Al2O3, 6–14 mol% jadeite component, Mg numbers of 88–94, and low TiO2 contents (<0.18 wt%) Figs. 4 and 5). Orthopyroxene is enstatite with 4–5 mol% Mg-Tschermak component, 0.5–0.6 wt% Cr2O3, and Mg numbers of 89–91 (Fig. 4). Megacrystic pyroxenes are similar in composition to finer-grained, matrix pyroxenes (e.g., Fig. 4). Brown, Cr-rich spinel is ubiquitous in the lherzolites. Cr/Cr + Al ratios range from 0.17 to 0.28, with little to no compositional variability within a single sample. Mg numbers in spinel range from 0.70 to 0.75 but show no variation within individual samples. Olivine compositions range from Fo89–Fo93, but do not vary by more than 0.01 within single samples. CaO contents as high as 0.32 wt% are observed in some olivine grains, but most analyses range between 0.05 and 0.10 wt%.

Carbonates are less abundant in the lherzolites than in the pyroxenites (see following), and quantitative analyses were carried out in only a single lherzolite sample (CTON-53, Fig. 6). In this sample, calcite is intergrown with and appears to be replacing olivine in response to melt-rock interaction (Fig. 6). Some of the calcite grains in this sample show oscillatory growth zones with 0.0–1.0 wt% MnO.

Glass compositions were determined along grain boundaries and in discrete melt pockets associated with clinopyroxene, orthopyroxene ± calcite in samples CTON-57 and CTON-53. Most analyses have oxide totals in the 80–84 wt% range, consistent with large amounts of H2O and/or CO2 in the glass. The glasses have SiO2 contents of 35–48 wt%, Al2O3 contents around 12%, and Mg numbers of ∼60; all are strongly enstatite normative. The presence of calcite and pyroxene crystals in the melt pockets indicates that the glasses are not primary melt compositions.


We examined >70 pyroxenite xenoliths, which we subdivide here based on the presence or absence of garnet and the types of spinel and clinopyroxene present. In all of the pyroxenites, clinopyroxene is significantly more abundant than orthopyroxene, and olivine is rare to absent. Primary carbonate is present in many of the aluminous pyroxenites and a few of the Cr-diopside pyroxenites. All of the pyroxenites are characterized by irregular grain sizes and shapes and complex intergrowths between phases. In contrast to the lherzolites, however, there is little evidence of significant strain accumulation in any of the pyroxenites.

Cr-Diopside Pyroxenite

Cr-diopside pyroxenites are bright green in hand sample, contain abundant Cr-rich spinel, lack garnet, and are texturally well equilibrated (e.g., CTON-21 and CTON-67 in Fig. 3). A few of these samples contain up to 10% olivine, but most have little to no olivine. Many Cr-diopside pyroxenites contain euhedral orthopyroxene grains with finely exsolved cores and unex-solved rims; coarse exsolution and intergrowth textures in pyroxenes are rare. No crosscutting relationships are observed between Cr-diopside pyroxenites and either spinel lherzolites or other pyroxenite types. However, there is gradational layering between Cr-diopside pyroxenite and lherzolite in several samples.

Pyroxene and spinel compositions in the Cr-diopside pyroxenites (Figs. 5 and 7) are generally similar to those in the lherzolites and distinctly different from those in the garnet-spinel and spinel-absent pyroxenites. Clinopyroxene has 8%–11% jadeite component and orthopyroxene has ∼4% Mg-Tschermak component. Spinel has Cr/Cr + Al ratios of 0.11–0.24 and Mg numbers of 70–75.

Spinel ± Garnet Pyroxenite and Spinel-Absent Pyroxenite

These pyroxenites are black in hand sample and are dominantly composed of clinopyroxene (up to 80 modal%). Many of the pyroxene grains show complex intergrowth relationships with one another. These intergrowths usually comprise two clinopyroxene grains, with less commonclinopyroxene-orthopyroxeneintergrowths. Many clinopyroxene grains contain rounded orthopyroxene inclusions (Fig. 8) that are in optical continuity with one another, suggesting replacement or overgrowth of original orthopyroxene by clinopyroxene. The garnet-bearing pyroxenites (e.g., CTON-17, CTON-77, and CTON-81 in Fig. 3) contain small (<0.5 mm) relict garnet grains and fragments that are surrounded by dark glassy to symplectitic rims that are similar to kelyphytic rims described by others in garnet pyroxenites and peridotites (e.g., Henjes-Kunst and Altherr, 1992; Stern et al., 1999; Sapienza et al., 2001). The garnet grains are always in close spatial association with large black spinel grains, although there is no clear textural evidence for a reaction relationship between the two phases (Figs. 9 and 10). Spinel is abundant in all of the garnet-bearing samples, typically as large (up to 3 mm) black grains. Small green spinel grains locally rim the black spinels, particularly in symplectites adjacent to garnet (Fig. 11). Reaction rims surrounding spinel are discussed in a subsequent section.

Fluid and glass inclusion trails are abundant throughout all pyroxenite samples; crosscutting relationships indicate that multiple generations are present and some trails contain both types of inclusions. Glass also occurs within symplectites around garnet and spinel and in melt pockets at grain triple junctions. The melt pockets typically contain carbonate in addition to silicate glass (see following). Olivine is rare to absent in the garnet pyroxenites, and it is confined to small inclusions within pyroxene. Rare phlogopite is present as inclusions in pyroxene grains.

Many pyroxenite samples that lack garnet are texturally akin to the garnet pyroxenites. These samples contain abundant large black spinels that are similar to those observed in association with garnet in the previously described samples. Pyroxenes in these samples show complex intergrowths and replacement textures that are also similar to those in the garnet pyroxenites. Based on these characteristics, it is likely that garnet was originally widespread in the pyroxenites beneath Cerrito Negro and other necks, but it reacted out of most samples prior to entrainment or during transport to the surface.

A few samples of pyroxenite (e.g., CTON-7 in Fig. 3) lack both spinel and garnet. These samples are otherwise similar to the garnet pyroxenites in that they contain large, irregular, and interlocking pyroxene grains. Optically continuous inclusions of orthopyroxene within clinopyroxene are common. Both pyroxene types are exsolved on a variety of scales.

Pyroxene compositions within the spinel ± garnet and spinel-absent pyroxenites are much more aluminous than those in the Cr-diopside pyroxenites (Figs. 5 and 8–10). Clinopyroxene is augite with 14–16 mol% jadeite component; orthopyroxene is enstatite-rich with 7–11 mol% Mg-Tschermak component. Mg-numbers for both clinopyroxene and orthopyroxene range from 74 to 83, and garnet-bearing samples are the most Feenriched. Cr2O3 decreases and Al2O3, TiO2, and Na2O increase with decreasing Mg-number for both clinopyroxene (Fig. 5) and orthopyroxene. Cr2O3 is lower and Al2O3 and TiO2 are higher in the garnet-spinel and spinel-absent pyroxenites than in the Cr-spinel peridotites or lherzolites. CaO is higher in orthopyroxene and lower in clinopyroxene in all of the pyroxenites relative to the lherzolites, suggesting equilibration at higher temperatures (see following).

Relict garnet grains are present in pyroxenites CTON-6, CTON-17, CTON-77, and CTON-81 (Figs. 9 and 10). Garnet grains large enough to analyze have average compositions of Pyr61Alm2-6Grs13Sps1 and show a minor decrease in Mg (<5 mol %) and increase in Ca (2–5 mol %) in their glassy rim regions. Black spinels in the garnet pyroxenites are spinel-hercynite solid solutions with Mg numbers of 58–60. Cr2O3 contents are <0.3 wt% (Figs. 9 and 10). Green spinel grains in symplectites are very similar in composition to black spinel grains. Clinopyroxene compositions within the symplectites are nearly identical to matrix compositions, but orthopyroxene shows markedly lower Al2O3 (∼6 versus 10 wt%) and higher SiO2 (∼52 versus 49 wt%) contents in the symplectites than in the matrix (e.g., see analyses in Fig. 10). Plagioclase was detected optically in symplectite rims around spinel and confirmed via quantitative electron microprobe analysis and X-ray maps. Due to the very small grain size in the symplectites, most analyses of plagioclase were mixed, but the few stoichiometric analyses that were obtained yielded compositions of An93. Carbonates from Cerrito Negro are calcite with a minor dolomite component. Many samples show concentric Mn zoning that is similar to that observed in lherzolite sample CTON-53. A single sample from Cerro de Guadalupe (CG-40) contains oscillatory zoned dolomite.

Carbonate-Bearing Samples

Calcite occurs in many of the aluminous pyroxenites, including two of the garnet-bearing samples from this study, and in a few of the lherzolites. In general, xenoliths that contain abundant carbonate are the least well equilibrated texturally. Calcite grains as large as 2 mm occur at triple junctions between pyroxene grains, as inclusions in pyroxene (e.g., Fig. 12), within trails and clusters of melt inclusions, a long grain boundaries, and within fractures. Many of the calcite grains contain concentric bands of glass inclusions and show oscillatory Mn zoning (Fig. 12; see also Figure 2d of Perkins et al., 2006). In almost every case, calcite is intergrown with or adjacent to orthopyroxene grains (Figs. 12 and 13). Clinopyroxene is typically strongly zoned toward lower Al, Ti, and Na contents adjacent to calcite (Fig. 13). Fluid/glass inclusion trails terminate at calcite grain boundaries and do not crosscut either the calcite or the spatially associated orthopyroxene grains (Fig. 13). Oscillatory zoned dolomite is present in melt pockets in some xenoliths from Cerro de Guadalupe (see Figure 2g of Perkins et al., 2006) but was not observed in xenoliths from Cerrito Negro. Compositions of glass inclusions and glass pockets range from carbonatitic (SiO2 = 1%–19%, oxide totals = 52%–75%) to silicate glasses similar to those in lherzolite CTON-53 described previously (e.g., Fig. 13).


Temperatures for lherzolites and pyroxenites and pressures for garnet-bearing pyroxenites were calculated using the two-pyroxene thermometer and the Al-in-orthopyroxene barometer of Brey and Köhler (1990). Representative calculated temperatures are given for each rock type in 01Table 1. In lherzolites, compositions were selected from adjacent clinopyroxene and orthopyroxene grains with well-equilibrated textures, as well as from adjacent lamellae in locally present, coarsely exsolved megacrysts (e.g., Fig. 4). The general scarcity of orthopyroxene in some of the pyroxenites resulted in selection of both adjacent and nonadjacent mineral pairs. Temperature calculations for pyroxenites based on nonadjacent pyroxenes yield temperatures ∼30 °C lower than those from adjacent pyroxene grains. Compositional data for garnet-orthopyroxene pressure calculations were chosen based on the presence of garnets large enough for microprobe analysis that were also near orthopyroxene grains. Coherent garnet grains juxtaposed immediately against adjacent orthopyroxene are rare due to the large symplectite rims around most garnet grains, but orthopyroxene is generally abundant immediately surrounding the symplectites.

Temperatures for the lherzolites range from 900 to 1000 °C at a reference pressure of 15 kbar (Fig. 14). The presence of coarsely exsolved megacrysts in some of the lherzolites hints at a higher-temperature history prior to the most recent episode of equilibration. Temperatures are generally higher for the pyroxenites than for the lherzolites, and they also show a broader range, from 950 to 1145 °C (Fig. 14A). Cr-spinel pyroxenites overlap the high-temperature end of the lherzolite range, garnet pyroxenites record temperatures from 1000 to 1145 °C, and carbonate-bearing pyroxenites (all types) record the highest temperatures.

Two of the garnet-bearing pyroxenite samples (CTON-6, CTON-77) record pressures of 16–18 kbar, whereas samples CTON-17 and CTON-81 record pressures of 11–14 kbar. The Brey and Köhler (1990) geobarometer is extremely sensitive to calculated ferric iron content in garnet. Garnets in CTON-6 and CTON-77 have calculated Fe3+ of 0.048–0.060 cations per formula unit (p.f.u.), compared to Fe3+ = 0 in CTON-17 and CTON-81. Zero values for Fe3+ yield the lowest pressure estimates and are therefore assumed to provide only minimum constraints on pressures experienced by the samples. Pressures in all samples must have been greater than 11 kbar, which corresponds to the inferred depth to the Moho beneath the Puerco necks (West et al., 2004). The samples that record higher pressures have narrower symplectites around garnet than the samples that record lower pressures. We thus take the 16–18 kbar values as more realistic for garnet growth and initial equilibration (Fig. 14) and interpret the lower pressures to represent reequilibration during decompression.


Whole-rock major-element data were obtained for 10 spinel ± garnet pyroxenites and 11 lherzolites by X-ray fluorescence (XRF) of fused disks using the Rigaku RIX 2100 at the University of New Mexico 02(Table 2). None of the Cr-diopside pyroxenites was analyzed as part of this study. Most samples were titrated to determine concentrations of FeO and Fe2O3. Whole-rock data for the host basalts were taken from Hallett (1994). None of the Cr-diopside pyroxenite xenoliths was large enough to crush for whole-rock analysis.

The CaO/Al2O3 ratios range from 0.66 to 1.44 in the lherzolites, and from 0.34 to 2.09 in the pyroxenites. Mg-numbers are 90–91 in the lherzolites, but scatter between 72 and 82 in the pyroxenites. Most of the pyroxenites are enstatite-normative and are chemically similar to Wilshire et al.'s (1988) Cr-diopside websterite analyses, despite the low Cr content of their pyroxenes; three pyroxenites are nepheline-normative and are similar to Wilshire et al.'s (1988) Al-augite clinopyroxenite samples.

Major-element data for the lherzolites, spinel ± garnet pyroxenites, and basalts, and microprobe analyses of silicate and carbonate melt inclusions are plotted against MgO in Figure 15. For all oxides except SiO2, the pyroxenites plot within a triangle defined by lherzolite, average Rio Puerco basalt, and the most Ca-rich carbonate melt inclusion composition.


Any model for the origin of the Rio Puerco pyroxenites has to take into account whole-rock chemical data, mineral compositions, temperatures of equilibration, and field and barometric evidence for formation in the spinel lherzolite facies in the mantle.

Clinopyroxenes from the Cr-diopside pyroxenites are similar to those from lherzolite xenoliths and are characterized by high Cr2O3 and Mg/Mg + Fe, low TiO2, and moderate Al2O3 and Na2O contents (Fig. 5). These characteristics are consistent with formation of the Cr-diopside pyroxenites via removal of an alkali basaltic melt component from a lherzolitic precursor (Hart and Dunn, 1993).

As shown in Figure 15, the spinel ± garnet pyroxenites differ significantly in major-element composition from the erupted alkali basalts from the Rio Puerco necks (basalt data from Hallett, 1994) and from the silicate melt inclusion compositions. The pyroxenites show some similarities to picrite compositions from the GEOROC database (, but generally have higher CaO and MgO contents and lower Na2O and Al2O3 contents than picrites. Based on these data, it is unlikely that the pyroxenites represent primary melt compositions. The complex intergrowth textures preserved in most of the spinel ± garnet pyroxenites, coupled with localized pyroxenite development in crosscutting veins/dikes, are inconsistent with a crystal settling origin for the pyroxenites, although cumulate-like compositions could result from melt crystallization on conduit walls (e.g., Irving, 1980). Fractional removal of olivine and/or clinopyroxene from an alkali basalt would produce a remaining melt fraction that would be depleted in MgO and enriched in FeO, TiO2, Al2O3, and Na2O relative to the pyroxenites shown in Figure 15, and hence it also fails as a mechanism for pyroxenite formation.

Textural and chemical data from this study support the hypothesis of infiltration and addition of melt(s) to mantle lherzolite as an origin for Rio Puerco spinel ± garnet pyroxenites. In general, infiltration of silicic melt alone will favor conversion of olivine to orthopyroxene (e.g., Rapp et al., 1999; Prouteau et al., 2001; Liu et al., 2005). The preferential development of clinopyroxene rather than orthopyroxene in the pyroxenites, however, requires infiltration of a melt with high Ca content and relatively low Si activity. The presence of both silicate and carbonatite melt inclusions within single xenolith samples implies infiltration of either a miscible carbonosilicate melt that subsequently unmixed, or lherzolite interaction with at least two infiltrating melts of widely varying composition.

As is evident from Figure 15, the pyroxenites could not have been produced via reaction of lherzolite with basaltic melt alone, but could represent mixtures that also involved a carbonatitic melt component. Calculations using basaltic melt with 8% MgO (Negro1 from Hallett, 1994), the most calcic carbonate melt inclusion from this study (from sample CG-40), and average lherzolite from this study were carried out in order to test the hypothesis that the pyroxenites formed via bulk mixing of lherzolite with silicate + carbonate melt. Al2O3, FeOT, MgO, CaO, and MnO contents of the pyroxenites can be adequately matched via bulk mixing in the approximate proportions lherzolite30–35-basalt40–50-carbonatite15–25 (e.g., Fig. 15). TiO2 and Na2O values of the pyroxenites can also be matched via mixing of ∼35% lherzolite and 65% melt, but require a greater contribution from carbonatite relative to basal-content of the pyroxtic melt. The high SiO2 enites is not easily accounted for by any simple mixture of lherzolite + basalt + carbonatite, however. The Si values are similar to SiO2 contents of clinopyroxene and orthopyroxene and likely reflect preferential consumption of olivine and formation of pyroxenes during lherzolite-melt interaction. Interestingly, the SiO2 values are also consistent with mixing in the approximate proportions lherzolite35melt65.

Clinopyroxenes in the spinel ± garnet pyroxenites have very low Cr2O3 and Mg/Mg + Fe, and high Na2O, TiO2, and Al2O3 contents. These compositions are consistent with pyroxene growth in response to addition of melt to a lherzolitic precursor. The high Fe, Al, Ti, and Na contents are generally more consistent with infiltration of a silicate than a carbonatitic melt (Rudnick et al., 1993), but zoning toward lower Fe, Al, Ti, and Na adjacent to calcite inclusions and melt pockets (e.g., Fig. 13) clearly points toward involvement of a carbonatitic melt as well.

Thermobarometric data from the Rio Puerco xenoliths also support spinel pyroxenite formation in response to lherzolite-melt interaction if the lherzolites and pyroxenites were sampled from similar depths. These pyroxenites record temperatures that are as much as 200 °C hotter than lherzolites from the same necks (Fig. 14) and plot close to the pyroxenite solidus experimentally determined by Pertermann and Hirschmann (2003). Temperatures obtained from pyroxenes adjacent to melt pockets in spinel lherzolite samples are also ∼200 °C hotter than the equilibrium lherzolite values 01(Table 1). Geobarometry is consistent with pyroxenite equilibration at depths within the spinel lherzolite stability field. The high temperatures at these depths can most readily be reconciled with pyroxenite formation at melt conduit walls that crosscut spinel lherzolites (Fig. 16). The relatively similar calculated lherzolite:melt ratios needed to produce the analyzed pyroxenites suggest that reactive melt infiltration may have ceased once all olivine was consumed near the melt conduits. The large size (up to 30 cm in diameter) of many of the pyroxenite xenoliths indicates that melt was able to penetrate significant distances into the surrounding lherzolite during its ascent.


Textural evidence for two separate reactions involving garnet, spinel, and melt is seen in the spinel ± garnet pyroxenites. These reactions record modifications to the lithosphere below the Rio Puerco necks after initial formation of the pyroxenites.

Garnet Breakdown

In the garnet-bearing pyroxenite samples, relict garnet grains are surrounded by reaction rims up to 200 μm in width. Petrographic and microprobe data indicate that the reaction zones are composed of orthopyroxene, spinel, clinopyroxene, and silicate glass. Anorthite is locally present, but it is probably a product of subsequent breakdown of the spinel. As is evident in Figure 10, orthopyroxene is by far the most abundant phase in the symplectites rimming garnet, with lesser but subequal amounts of spinel, clinopyroxene, and glass.

Using average compositions of matrix clinopyroxene, relict garnet, symplectite orthopyroxene, spinel, plagioclase, and glass from the garnet pyroxenites, along with olivine data from the lherzolite samples, the following three balanced reactions were calculated (coefficients on a molar basis, see 03Table 3 for volumetric proportions):  

These reactions are obviously simplifications in that they do not allow for continuous changes in composition of the phases. Nonetheless, they provide a starting point for evaluating the reaction histories of these samples. Both reactions 1 and 2 imply derivation of the reaction zones from an olivine-bearing precursor in which olivine is the limiting reactant. Reaction 1 predicts production of orthopyroxene in ∼3:1 volumetric proportions relative to both spinel and clinopyroxene, with no involvement of a melt phase. Reaction 2 more closely approximates the observed phase abundances shown in Figure 10, including glass preserved within the symplectites and the high modal production of orthopyroxene. There is poor agreement between the observed abundances and those predicted by reaction 3. One possible interpretation of the reaction zones thus involves closed-system, in situ partial melting of an olivine-bearing garnet pyroxenite, resulting in total consumption of olivine and production of two pyroxenes, spinel, and quenched glass.

Glass inclusions with the highest MgO contents were used with the pMELTS program (Ghiorso et al., 2002) to determine the approximate solidus temperature of the melt involved in reaction 2. Calculations at a reference pressure of 1.5 GPa yielded solidus temperatures ranging from 1050 °C (fO2 = QFM +1.0 log unit) to 1100 °C (f O2 = QFM −1.0 log unit). In all cases, calculated mineral compositions and modes were similar to those observed in the actual assemblages. These calculations support garnet breakdown via localized partial melting at the temperatures and pressures calculated for the samples.

Spinel Breakdown

Black spinel grains, which are abundant in the garnet-bearing samples, are surrounded by 50–100-μm-wide reaction rims composed of orthopyroxene, anorthite, and glass (Fig. 11). Textural analyses coupled with microprobe data give rise to the following balanced reaction (mol%; see 03Table 3 for vol%):  

Product phase abundances predicted from this reaction are in reasonable agreement with the relative abundances of orthopyroxene and plagioclase in the reaction rims; however, the amount of melt varies significantly within and between symplectites in all samples in which glass is observed.

In order to evaluate the pressure-temperature dependence of the breakdown reactions, thermodynamic data were calculated from microprobe data for all phases using the MELTS Supplemental Calculator (; Ghiorso and Sack, 1995; Ghiorso and Kress, 2004) at reference conditions of 1100 °C and 1.5 GPa 03(Table 3). Reaction 1 has a shallow slope of 5 bar/°C and thus requires decompression to proceed. Reactions 2–4 have steeper slopes (45, −108, and 70 bars/°C, respectively). Hence, garnet breakdown via reaction 2 and spinel breakdown via reaction 4 both likely occurred in response to a combination of heating and decompression (Fig. 16).

Timing of Pyroxenite Formation and Garnet-Spinel Breakdown Reactions

Pyroxenite xenoliths are also abundant in Pleistocene (Anthony and Poths, 1992) eruptive materials at Kilbourne Hole and Potrillo Maar in the southern portion of the Rio Grande rift (Irving, 1980; Wilshire et al., 1988; Padovani and Reid, 1989), although carbonate has not been reported from these pyroxenites. No pyroxenites were observed in the mantle xenolith suite from the Cieneguilla limburgite, a locality in the center of the Rio Grande rift that was emplaced early in the extensional history (ca. 25 Ma) (Baldridge, 1979; Baldridge et al., 1980). Initial generation of the silicate ± carbonatite melts responsible for converting lherzolite into pyroxenite likely occurred in response to lithospheric thinning and/or asthenospheric upwelling associated with rifting. Oxygen isotopic disequilibrium between pyroxenes and olivine in lherzolites and between pyroxenes and calcite in pyroxenites constrains the youngest carbonatite-silicate melt metasomatic events to <1 m.y. prior to eruption in the Puerco field (Perkins et al., 2006). Preservation of fine-scale zoning in carbonate grains is also consistent with metasomatism occurring just prior to sampling by the host magmas. The preservation of different xenolith populations in volcanic centers of different age thus roughly constrains melt-peridotite interaction to Miocene-Pleistocene along the Rio Grande rift, and to the Pliocene for the region beneath the Rio Puerco necks.

Garnet breakdown within the pyroxenites obviously must postdate pyroxenite formation, and must thus record mantle processes shortly before eruption of the host basalts. The symplectite reaction rims around garnet reach up to 200 μm in diameter and are overprinted by symplectites derived from spinel breakdown. Geobarometric calculations yield lower pressures for the samples with wide reaction rims compared to those with narrower rims. The reaction was most likely driven by heating related to melt passage through the mantle source region, combined with rift-related decompression. The narrower symplectite zones surrounding spinel, coupled with the development of plagioclase in these zones, suggest that spinel breakdown may have occurred during ascent in the host magma.


Mantle-derived pyroxenites occur only locally within xenolith suites from the Colorado Plateau (e.g., McGetchin and Silver, 1972), but are abundant in xenolith suites from the southern Rio Grande rift (e.g., Irving, 1980; Wilshire et al., 1988) and the Rio Puerco necks (this study). This association is not surprising if pyroxenite formation results from melt-rock interaction, since the warm mantle of the extended regions of the western United States provides greater opportunity for such interactions than does the cooler mantle beneath the Colorado Plateau. The Rio Puerco pyroxenites differ in two significant respects from other pyroxenites in the western United States; however, calcite and carbonate melt inclusions are abundant, and pyroxene compositions are more extreme than those of pyroxenites within the southern Rio Grande rift (e.g., Fig. 5). The calcite and the carbonate melt inclusions point to involvement of carbonatitic melts in the evolution of the Rio Puerco mantle, and the pyroxene compositions imply that this mantle region was more extensively metasomatized than the mantle beneath the southern Rio Grande rift. The mantle beneath the Colorado Plateau also experienced pervasive metasomatism, but largely in response to infiltration of hydrous fluids thought to be related to subduction of the Farallon slab (e.g., Smith et al., 1999, 2004; Smith and Griffin, 2005). Magnesite has been reported in a few peridotite xenoliths from Green Knobs on the Colorado Plateau (Smith, 1979) but is not prevalent. Likewise, hydrous minerals (phlogopite, amphibole, and apatite) occur within some of the Rio Puerco xenoliths (Samuels et al., 2006) but are rare. These observations suggest that the composition of the metasomatizing fluid(s), the extent of metasomatic interaction, and the thermal history all differed between the mantle beneath the Colorado Plateau, the Rio Puerco necks, and the southern Rio Grande rift.

On the basis of stable isotopic data, Perkins et al. (2006) argued that carbonatitic melts in the Rio Puerco field originated from mantle that had interacted with fluids derived from the subducted Farallon plate. Ducea et al. (2005) reached a similar conclusion from isotopic analysis of carbonate inclusions in peridotite xenoliths from the Sierra Nevada. Although the Farallon plate is thought to have detached and sunk to depth between 50 and 20 Ma (Humphreys, 1995), its chemical influence is still apparent within recently modified mantle samples from the Puerco necks (Fig. 16) as well as the hydrated xenoliths from Tertiary eruptions on the Colorado Plateau (Smith et al., 2004; Smith and Griffin, 2005). One possible explanation for the high degree of melt metasomatism in this region comes from the location of the Puerco necks at the intersection between the Rio Grande rift and the Jemez Lineament. Although the Jemez Lineament is defined as an alignment of Cenozoic volcanic centers, it likely reflects a lithospheric discontinuity of Precambrian ancestry (Karlstrom and Humphreys, 1998). It is currently underlain by seismically slow mantle, indicating anomalously high temperatures, the presence of partial melt, and/or local variations in hydration or other chemical heterogeneity (Gao et al., 2004). If the Jemez Lineament is truly a lithospheric-scale structure, variations in fracture density at depth may have focused fluids/melts released from the subducting Farallon plate into this zone. Asthenospheric upwelling following Farallon removal would then have led to enhanced melt production within the metasomatized mantle of the region and to entrainment of a wide variety of xenolith types. This hypothesis requires further testing through more detailed studies of pyroxenite xenoliths from localities that lie outside of the Jemez Lineament.

Regardless of the fluid-focusing mechanisms at depth, the pyroxenites of the Rio Puerco necks preserve a record of silicate and carbonatite melt metasomatism, heating, and decompression in the upper mantle at the junction of the Colorado Plateau and the Rio Grande rift. The LA RISTRA experiment documented pronounced seismic attenuation to depths ≥200 km in the mantle beneath this area. Gao et al. (2004) interpreted the low velocities in terms of thermal erosion of the lithosphere, but cautioned that the data could not rule out the presence of small amounts of melt and/or hydration. The textural and pressure-temperature data from the Rio Puerco garnet pyroxenites support both heating and decompression within the lithospheric mantle prior to entrainment in the host basalts. These data are in accord with a model in which lithosphere is thinned and thermally converted to asthenosphere. However, the high degree of melt-rock interaction recorded by the entire pyroxenite suite is also consistent with pervasive chemical modification of the lithosphere. Conversion of a significant amount of lherzolite to clinopyroxene-rich pyroxenite with lower bulk Mg number will result in significant reduction in seismic-wave velocities. Chemical modification of the mantle may thus be as important as temperature effects in explaining the low-velocity zone beneath the Puerco necks. If the seismic anomaly in this region does result in large part from chemical effects, we predict that the extensively refertilized lithosphere will ultimately become negatively buoyant, once rift-related heating comes to an end (e.g., Carlson et al., 2005).

*Corresponding author:; phone: +1-505-277-6528, fax: +1-505-277-8843

This work was funded by a National Science Foundation (NSF) grant EAR02-29238 to Selverstone and a Caswell-Silver Undergraduate Research (UNM) grant to Samuels. Thoughtful reviews of the manuscript by Michael Roden, Libby Anthony, and Richard Carlson are gratefully acknowledged. We thank A.-M. Ali for assistance with whole-rock analyses, J. Berlin and M. Spilde for microprobe expertise, and G. Perkins, C. Callahan, A. Brearley, M. Roy, and Z. Sharp for extensive discussions regarding the Rio Puerco xenoliths.