Absarokites from the Western Mexican Volcanic Belt: Constraints

We have investigated the near liquidus phase relations of a primitive absarokite from the Mascota region in Western Mexico. Sample M. 102 was chosen because it has high MgO contents, a high Mg# and Fo90 olivine phenocrysts, indicating it is primitive mantle melt. Highpressure experiments on a synthetic analogue of the absarokite composition with a H20 content of either -1.7 wt% or -5.1 wt% were carried out in a piston cylinder apparatus. The composition with -1.7 wt% H20 is multiply saturated with olivine and orthopyroxene as liquidus phases at 1.6 GPa and 14000 C. At the same pressure clinopyroxene appears 300 C below the liquidus. With a H20 content of -5.1 wt% composition M.102 is multiply saturated with olivine and orthopyroxene on the liquidus at 1.7 GPa and 13000 C. Assuming batch melting, we suggest that absarokite M. 102 segregated from a depleted lherzolite or harzburgite residue at depth -50 km depth in the mantle wedge. Unlike most lavas in the region, the absarokite has not ponded and fractionated at the crust mantle interface (-35-40 km), and the temperatures of multiple saturation indicate that the mantle wedge beneath the Jalisco block is hotter than previously thought. The low degree batch melting of an original metasomatised harzburgite source, can produce the observed trace element abundances. The liquidus phase relations are not consistent with the presence of non-peridotitic veins at the depth of last equilibration. Therefore, we propose that the Mascota absarokites segregated at an apparent melt fraction of less than 5% from a depleted peridotitic source. They initially formed by a small degree of melting of a metasomatised original source at greater depth. Introduction Two main processes for generating melts in subduction zones are decompression melting due to corner flow and/or extension (Luhr, 1997; Sisson and Bronto, 1998; Elkins Tanton et al., 2001) and melting due to hydrous metasomatism of the mantle wedge above the subducting slab (Gill, 1981). If decompression melting occurs during mantle flow into the mantle wedge, it will leave behind a depleted often harzburgitic mantle that will be metasomatised by a slab component. This metasomatism can either be localized, as mineralogically distinct veins (Vidal et al., 1989; Foley, 1992a; 1992b), or distributed in the form of a phlogopite and/or amphibole bearing peridotite. Potassium, sodium and silica are the major elements that are most strongly enriched in the metasomatic agent that has been interpreted either as a H20 rich fluid (Stolper and Newman, 1994; Grove et al., 2002) or a slab melt (Rapp et al., 1999; Rose et al., 2001). Calcalkaline and especially the highly potassic volcanic rocks such as lamprophyres and members of the Absarokite-Shoshonitie-Bankanite series (Iddings, 1895) have been interpreted as partial melts of metasomatised/veined peridotitic mantle (Tatsumi and Koyaguchi, 1989). Luhr (1997) concluded that the highly potassic lavas in Western Mexico represent the "the essence of subduction zone magmatism". Absarokites are amongst the most primitive of these potassic volcanic rocks, and understanding their petrogenesis is important to constraining the conditions of melt generation in the metasomatised/veined mantle wedge. We report the liquidus phase relations of the most primitive potassic lava (M102, Table 1) from the Mascota graben on the Jalisco block in Western Mexico (Carmichael et al., 1996). Sample M102 is an absarokite that is part of a Pleistocene-Holocene volcanic suite ranging from absarokites and minettes to basaltic andesites and spessartites. Carmichael et al. (1996) used a petrographic classification (Iddings, 1895; Nicholls and Carmichael, 1969) to name the rocks from the Mascota graben. Sample M. 102 comprises phenocrysts of olivine and augite in a matrix of pyroxene, plagioclase and sanidine and therefore classifies as an absarokite. Due to its unusually low K20/Na20O-ratio compared to average absarokites (Morrison, 1980) it is also referred to as sodic absarokite (Table 1). In the commonly used chemical classification diagram of high potassic volcanic rocks, K20 vs. SiO 2 (Peccerillo and Taylor, 1976) sample M. 102 is identified as a basalt (Figure 1). In this paper we will refer to it as an absarokite. In the following paragraphs we briefly summarize the regional geology and plate tectonic setting that are important for petrogenesis of the Mascota potassic lavas. For a more extensive review of the history of the Rivera plate, see DeMets and Traylen (2000) and for tectonics and basaltic volcanism of the Jalisco block, see Luhr (1997). The Rivera plate The arrival of the Pacific-Farallon seafloor spreading center at the convergent margin of North America -28Ma ago, broke the Farallon plate up into several smaller plates (Atwater, 1970). Offshore Western Mexico two such fragments, the Rivera and Cocos plate are currently subducting along the Middle American trench (Figure 2). The Rivera plate that descends underneath the Jalisco block has moved independently for at least 10 Ma (DeMets and Traylen, 2000). The Rivera plate is an oceanic microplate with a complex plate motion history and significant internal deformation. Like other oceanic microplates its present rotation pole is close to the location of the plate and the relative plate motion is governed by shear along the plate boundaries rather than slab pull (DeMets and Traylen, 2000). Figure 2 shows that the Rivera plate moved orthogonal to the trench prior to 7.9 Ma slowing from a convergence rate of -50 mm/yr at 9 Ma to -~1 mm/yr at 5 Ma. After 3.6 Ma the motion of the Rivera plate became parallel to the trench, implying a cessation of subduction from 2.6 Ma to 1.0 Ma. In the southern part of the Rivera-North American plate boundary the trench normal subduction resumed after 1 Ma (DeMets and Traylen, 2000). At the present time the Rivera plate subducts at an angle of ~45' and the Cocos plate dips ~30' in the northwest and continually flattens towards the southeast to an angle of-200 (Pardo and Suarez, 1993;1995). The age of the Rivera plate currently subducting at the trench is -10 Ma old (DeMets and Traylen, 2000). Volcanism in the Jalisco block The subduction of the Farallon plate and later its fragments along the Middle American trench formed the Mexican Volcanic Belt (MVB), which has been active at least since the Miocene (Ferrari et al., 1994; Righter, 2000). The lavas studied here occur on the Jalisco block, sometimes also called the Tepic-Colima block, situated between the northwestern-MVB and the Middle American trench. Two continental rift zones, and the Middle American trench bound the Jalisco block (Figure 2). In the northeast a series of grabens called the Tepic-Zacoalco rift zone, coincides with the large andesitic central volcanoes of the northwestern MVB (Allan, 1986; Moore et al., 1994; Righter et al., 1995; Rosas Elguera et al., 1996). Normal faulting with a strike-slip component initiated in the early Pliocene (4.2 Ma) and continues into the Holocene. The northwestern MVB that coincides with the Tepic-Zacoalco rift is dominated by calc-alkaline rocks associated with large andesitic volcanoes (Carmichael et al., 1996), but intra-plate type alkaline rocks are common throughout the Pliocene and Pleistocene (Righter and Carmichael, 1992; Righter et al., 1995; Righter and Rosas-Elguera, 2001). The Tepic-Zacoalco rift has also been interpreted as continental expression of the Tamayo Fracture zone (Righter et al., 1995). In the east the Jalisco block is bounded by the Colima rift zone, a graben that experienced 1.53.3 km of extension since the early Pliocene (Allan, 1986) and is still active (Suarez et al., 1994). The submarine Manzanillo canyon bounds the Jalisco block to the southwest and is interpreted as the offshore extension of the Colima rift zone (Khutorskoy et al., 1994). Since the late Miocene (-10 Ma) calc-alkaline volcanism has been widespread in the Colima rift. Coeval lamprophyric volcanism occurred since the early Pliocene (-4.5 Ma) and both continue into the Holocene. The interior of the Jalisco block has also experienced extension, forming the small, NNW trending Talpa and Mascota grabens, and the larger Atenguillo graben (black star in Figure 2). During the Pleistocene and Holocene potassic and calc-alkaline lavas have erupted, in the grabens and their surroundings (Carmichael et al., 1996). The Mascota volcanic suite ranges from primitive absarokites and minettes to andesites and their lamprophyric equivalent spessartite. Although the observed trends cannot be produced by fractionation and assimilation, Carmichael et al. (1996) assume that they are genetically related due to their close proximity in space and time. Starting compositions Sample M. 102 was chosen as an experimental starting composition. M. 102 shows all signs of a melt close to equilibrium with mantle peridotite including a high MgO content, together with a high Mg# (= molar MgO/[MgO+FeO]) and high concentrations of Cr and Ni (Table 1). Another important indicator of a mantle derived melt is the presence of Fo90-olivine phenocrysts (Fo = molar MgO/[MgO + FeO]), and a Fe-Mg partition coefficient KD g = 0.29 (KDFe -M g = molar [MgOmelt*FeOolivine]/ [MgOolivine*FeOmelt]) between olivine and melt, indicating that the melt was in equilibrium with Fo 90 mantle olivine. We therefore assume the absarokite M. 102 is a primary melt from a peridotite mantle that has not been modified significantly by post segregation processes. In the Mascota series the samples with the highest potassium contents have the highest Mg#'s, arguing against contamination by continental crust (Carmichael et al., 1996). The absence of a coexisting olivine and orthopyroxene on the liquidus of several experimental studies of potassic and ultra-potassic lavas has led to the conclusion that some might not have a peridotite source and therefore the usual selection criteria for primitive melts do not apply (Foley, 1992a). However, M. 102 is only mildly potassic (Figure 1) and is closely related in space and time to basaltic andesites in the Mascota series that are assumed to be derivatives of hydrous peridotite melting (Carmichael et al., 1996). Our experimental results confirm this assumption. A dry and a hydrous synthetic analogue of the composition of sample M. 102 (Table 1) were prepared by mixing high purity oxides, hydroxides, and silicates (Gaetani and Grove 1998). H20 was added to the hydrous mix as brucite Mg(OH) 2, which breaks down to MgO and H20 at run temperatures. Adding water as brucite ensures that all hydrous experiments have same initial H20 concentration. The amount of H20 in the composition can be adjusted by choosing the fraction of MgO replaced by Mg(OH)2. In our hydrous mix we have replaced all MgO by Mg(OH) 2 and therefore added the maximum amount of 5.1 wt% H20. These synthetic mixes were subsequently ground in an agate mortar under ethanol for 6 hrs. The mixes were stored in a dessicator at room temperature. Electron microprobe analyses of superliquidus glasses of both mixes showed large alkali losses during mix preparation and made it necessary to correct the sodium and potassium concentrations in both compositions. Sodium and potassium were added as 1.0 N solutions of NaOH and KOH in methanol, which are strongly hygrophylic and resulted in -1.7 %wt H20 in the initially dry mix (Table 1). The water contents of the mixes estimated from summation deficit in Table 1, are 1.5 and 4.1 wt% H20 for the less hydrous and the hydrous mix. For selected runs the H20 content was directly calculated by oxygen balance. Oxygen abundance in the glass was analyzed by electron microscopy and all oxygen left over after forming oxides is assumed to be bound in H20. H20 estimates from oxygen balance are generally higher than the summation deficits (Table 4 & 5). For some experiments the H20 estimates from oxygen balance lead to totals higher than 101.5 wt%. Since the oxygen measurement is sensitive to the thickness of the carbon coat on the samples, the high estimates of H20 may be related to the variations in carbon coat thickness. We therefore prefer the H2 0 estimates that lead to reasonable total from 98.5 wt% to 101.5 wt%. The assumed H20 content of the less hydrous runs is therefore 1.7 wt% as calculated from oxygen balance of experiment D127. For the hydrous mix the summation deficit (4.1 wt% H20) and oxygen balance techniques (5.9 wt% H20) show a discrepancy. As compromise between the two estimates we assume that the H20 content of the mix is the nominal 5.1 wt% H20 that have been added initially. Due to problems with MgO contamination from the experimental assembly and Fe loss during the experiments, the actual experimental bulk compositions differ from the composition of sample M.102. Best estimates for the less hydrous (1.7 wt% H20) and hydrous (5.1 wt% H20) compositions are given in Table 1 and will be called MEX-1.7 & MEX-5.1 below. The estimates are the means of repeated electron microprobe analyses of runs that did not contain any visual solid phases. For Mix-1.7 the 56 analyses from experiments D87, D90 and D127 (Table 2 & 4) were included in the mean. For all major and minor oxides, except FeO* in mix MEX-5.1 138 analyses from experiments B848, B854, B855 and B856 were included in reported mean. Experiment B848 has lost Fe during the run and was not included in the mean FeO* reported in Table 1. Experimental methods Two different capsule materials were used. Prefabricated carbon capsules were used for the less hydrous experiments (mix MEX-1.7) with higher liquidus temperatures (dry assemblage). The capsule material for the lower temperature, hydrous experiments (mix MEX-5.1), was AusoPd 2 presaturated with Fe to minimize Fe loss to the capsule material during the experiment (Gaetani and Grove, 1998). The AusoPd 2o capsules were pre-saturated by filling them with MEX-1.7 mix, before running them for 5 days in an 1 atm Deltech furnace at 1250 0 C and at an oxygen fugacity at the QFM buffer. After the conditioning the glass was removed from the capsule by dissolving it in hot HF with a drop of HN03.The cleaned, pre-saturated capsules were filled with MEX-5.1 starting material and welded shut. The pre-saturated Au80Pd2o inner capsule is placed into a handmade graphite outer capsule (hydrous assemblage). The hydrous and dry experimental assemblages are shown in Figure 3. In the dry assemblage a large graphite filled cavity below the graphite capsule was necessary to avoid contamination of the sample by the MgO spacers. First experiments showed that that melts produced by the starting composition are unusually mobile in graphite, and the stress field in the assemblage caused melt channels to migrate downward (Figure 4). Placing the dry capsule onto a cavity filled with powdered graphite reduced the MgO contamination significantly. All experiments were conducted in a V2" solid medium piston cylinder device (Boyd and England, 1960). The piston-in technique was used (Johannes et al., 1971), the experiments were pressurized to 1 GPa at room temperature and heated to 8650C at a rate of 100 0C/min, where the sample was held for 6 minutes. The pressure was then raised to run pressure and the temperature was raised at a rate of 500 C/min to the final run temperature. Experiments were quenched by turning off the power supply to the temperature controller. Cooling rates were fast enough to avoid growth of quench crystals in most dry runs. Hydrous runs produced excessive quench growth at -30 0 C below the liquidus, pressure quenching did not improve the results. The run temperature was measured using W97Re3-W75Re25 thermocouple with no correction for the effect of pressure on the thermocouple EMF. The temperature at the center of the sample is 20 0 C hotter than at the location of the thermocouple bead and all reported temperatures have been corrected for this effect. A Eurotherm 818 controller maintained the temperature at ±+2C of the target value and the run temperature is believed to be reproducible to ±70 C. BaCO3 has been used as pressure medium, and pressure calibration against the Ca-Tschermak breakdown reaction (Hays, 1966) showed that no friction correction is necessary. Experimental durations were kept short to minimize MgO contamination and in the experiments using graphite capsules and H20 loss in the experiments using Au80Pd 20 alloy capsules (Table 2 & 3). Analytical methods For most experiments, the sample capsule was recovered from the pressure cell, cut in half and mounted in epoxy. At each pressure an entire experimental assemblage was sliced and mounted, to check the uniformity of the distance between the thermocouple and the sample. All experiments were polished to 0.06 gtm finish, and examined in reflected light and using backscattered electron imaging. Experimental products were analyzed with the MIT JEOL JXA733 Superprobes using wavelength dispersive spectroscopy (WDS), with a 15 kV accelerating potential and a 10 nA beam. The CITZAF correction package of Armstrong (1995) was used to reduce the data. The atomic number correction of Duncumb and Reed, Heinrich's tabulation of absorption coefficients, and the fluorescence correction of Reed were used to obtain a quantitative analysis (Armstrong, 1995). For the glass phase the beam size was -10 gtm and Na was counted at the beginning of the analysis for 5 s. Other elements were measured for up to 40 s, depending on abundance level. Analytical precision can be inferred from replicate analyses of an andesite glass working standard (38b-129) from a 0.1 MPa experiment (Grove and Juster, 1989). One standard deviation of replicate glass analyses expressed as relative percent of oxides are SiO2:0.4%, A120 3:0.9%, CaO:1.5%, MgO:1.5%, FeO:1.4%, MnO:8.1%, P20 5:5%, Na20:1.9%, K20: 1.1%, based on 289 individual measurements over 28 analytical sessions. The mean sum of analyses of the anhydrous glass is 99.4%. The H20 content in the glasses was estimated by summation deficit, and therefore a maximum estimate. The phase proportions in the run products were estimated by least-squares mass balance. All oxides given in Table 1 were included in mass balance, and sums of the squared residuals are generally smaller than 1.0. Approach to Equilibrium We have not demonstrated an approach to equilibrium in our experiments, as it would require reversal experiments for all possible compositional changes in all phases. At constant temperature and pressure, equilibrium requires that experiments have constant bulk composition for the experimental duration, and that all phases have homogeneous compositions. Achieving homogeneous compositions in the mineral phases, requires experimental run times of 2-3 days, if pyroxenes are present (Baker and Stolper, 1994). In our experiments the strong loss H20 and Fe during the experiments required us to keep the run times as short as possible, to have an approximately constant bulk composition. Despite the short run times, there are many observations indicating that the experimental phases are sufficiently close to equilibrium. Most experiments are close to the liquidus and experimental products contain more than 80 % glass (Table 2 & 3, Figure 9a & 9b). The experimental glass is homogenous judging from the standard deviations given in Table 4 & 5. It has been shown that direct synthesis is sufficient to recover equilibrium phase appearance temperatures in experiments with >40% glass/melt (Grove and Bence, 1977). Olivine, orthopyroxene and clinopyroxene generally show minor element compositions and solid-liquid partitioning comparable to longer duration experiments. Figure 5 shows that the Ca and Al composition of clinopyroxenes in our experiments is comparable to those from longer duration multiple saturation experiments using basaltic compositions (Kinzler and Grove, 1992a; Gaetani and Grove, 1998). The Fe-Mg distribution coefficients (KDFeMg = molar [MgOmelt*Feoolivine]/[MgOolivine*FeOmelt]) are close to constant in experiments with the same bulk composition. The mean KD's for all experiments with composition MEX-1.7 (Table 1) and the mean KD'S for all experiments with composition MEX-5.1 (Table 1) are within one standard deviation of each other. The mean Fe-Mg partition coefficient (KDFe-Mg)ol between olivine and melt is 0.33±0.01 (la) in experiments with composition MEX-1.7, and 0.32±0.02 in experiments with bulk composition MEX-5.1. Values of (KDFe-Mg)ol from longer duration experiments range from 0.34 to 0.30 (Kinzler and Grove, 1992a; Gaetani and Grove, 1998). The (KDFeMg)opx between orthopyroxene and melt is 0.32±0.01 or ±0.02 in experiments with bulk composition MEX-1.7 and MEX-5.1 respectively. Longer duration experiments report (KDFeMg),,px from 0.27 to 0.33 (Kinzler and Grove, 1992a; Kinzler, 1997). The (KDFe-Mgcpx for clinopyroxene from experiments with bulk composition MEX-1.7 is 0.38±0.02. From experiments with composition MEX-5.1 only experiment B833 contained clinopyroxene, its KD Fe-Mgcpx is 0.34. Compared to experimental studies using longer experimental duration the (KDFe-Mg)cpx from the less hydrous compositions is higher than the mean value of 0.36 reported by Kinzler (1997). This might reflect the influence of the elevated alkali contents on the Fe-Mg