Geochemistry, Geophysics, Geosystems

An experimental investigation of the diffusive infiltration of alkalis into partially molten peridotite: Implications for mantle melting processes

Authors


Abstract

[1] When a silica-undersaturated melt is juxtaposed with partially molten peridotite, alkali elements rapidly diffuse into the peridotite in a process referred to as the diffusive infiltration of alkalis (DIA). Four types of piston cylinder experiments were performed providing constraints on the DIA process: (1) simple phase relation experiments, (2) melt-melt diffusion couples, (3) experiments examining the reaction between powders of basanite and spinel lherzolite (direct mixture experiments), and (4) diffusive infiltration-reaction couples between basanite and partially molten peridotite. Melt-melt diffusion couples constrain effective binary diffusion coefficients (EBDC) for all major elements except Na to be in the 10−6 to 10−7 cm2/s range at 1450°C and 0.9 GPa. The Na concentration profile is not binary being coupled to gradients in SiO2. EBDCs for Cl, Li, Rb, Sr, Ba, and La are in the 10−6 cm2/s range while Nb and Zr are in the 10−7 cm2/s range. A time series of direct mixture experiments shows that basanite reacts with lherzolite to reduce the orthopyroxene mode to a constant value within 15 min at 1300°C and 0.9 GPa. Quench modified melts in the direct mixture experiments have 50–51 wt.%. SiO2, independent of the basanite/lherzolite ratio. In contrast, infiltration-reaction experiments show that sodium diffuses from basanite into partially molten peridotite in 10–30 min, resulting in quench modified melt pools with up to 64 wt.% SiO2 within the peridotite. Modal analysis shows that addition of alkalis causes orthopyroxene to incongruently break down to olivine plus silica rich melt. In experiments using a basanite with enriched Cl concentrations, a distinct boundary zone appears between the basalt-peridotite interface and the peak in SiO2 concentration. Melts within this boundary zone have elevated CaO and Cl concentrations relative to both the basanite and melts within the peridotite beyond the boundary zone. The repeated formation of this boundary zone may indicate an important aspect of the DIA process, possibly responsible for the formation of anorthitic plagioclase and CaO rich melts.

1. Introduction

[2] The concentrations of the alkali elements sodium and potassium dramatically affect peridotite melting. Small increases in alkali content will substantially lower the solidus temperature of a peridotite [Hirschmann, 2000] while variations in the alkali contents of melts significantly shift the relative stability of coexisting olivine and orthopyroxene [Roeder, 1951; Schairer et al., 1954; Kushiro, 1975; Ryerson, 1985; Conceiçao and Green, 2000]. Correspondingly, alkali rich melts in equilibrium with a spinel lherzolite at 1 GPa can have >60 wt% SiO2 [Baker et al., 1995; Hirschmann et al., 1998; Robinson et al., 1998; Draper and Green, 1999].

[3] Silica undersaturated alkali basalts are ubiquitous at ocean islands, common in off-ridge axis seamounts [Niu and Batiza, 1997], and even rarely found at mid-ocean ridges [Shibata et al., 1979; Natland, 1989]. Alkali basalts typically have high middle/heavy rare earth element ratios consistent with melting in the presence of garnet and possibly result from melting more enriched mantle sources such as enriched peridotite or pyroxenite [Hirschmann et al., 2003]. Although alkali basalts are ubiquitous on the planet, the extent to which they interact with the mantle during ascent is poorly understood.

[4] Experiments have repeatedly shown that alkalis can rapidly diffuse in silicate melts with diffusion strongly dependent on gradients in melt SiO2 content [Watson, 1982; Watson and Jurewicz, 1984; Lesher, 1994; Lundstrom, 2000] The phase relations of lherzolite change with decreasing pressure such that the olivine phase volume expands at the expense of orthopyroxene and melts become more SiO2 rich [O'Hara, 1965, 1968; Stolper, 1980; Presnall and Hoover, 1984; Hirose and Kushiro, 1993]. Therefore if high pressure, silica-poor melts are channelized during ascent [e.g., Kelemen et al., 1997], gradients in silica content will exist between melt ascending in conduits and melt in surrounding peridotite. Experimental simulations show that sodium rapidly diffuses from a high Na2O, low SiO2 melt into partially molten peridotite causing orthopyroxene to incongruently melt [Lundstrom, 2000]. Here, I present experiments further evaluating the process of the diffusive infiltration of alkalis (DIA) into partially molten peridotite.

2. Experimental Methods

[5] Four types of piston cylinder experiments were performed, either at the University of Illinois Urbana Champaign (UIUC) or at Brown University (Table 1). In the first type, I determined the phase relations and compositions of melts in the peridotite and basanite starting materials at specific P-T conditions (experiments labeled “PR”). In the second type, infiltration-reaction experiments (labeled “IR”), I juxtaposed peridotite and basanite starting materials. Initial experiments of this type simply reflected layering of two powders while later experiments reflected two-steps as explained below. In the third type, gently ground mixtures of KLB-1 and EL-10 powders were run for different lengths of time to examine the kinetics of direct mineral-melt reaction (labeled “MIX”). Last, melt-melt diffusion couples determined major and trace element diffusion coefficients relevant to the infiltration-reaction process (labeled “MMD”). In all experiments juxtaposing two materials, I placed lower density materials above higher density materials to provide convective stability; similar glass compositions over a wide horizontal spacing of microprobe spots (located parallel to the interface) indicate that convection did not occur during the experiments. In several experiments, a layer of vitreous carbon spheres preserved quench-free melt (Robinson et al. [1998], following the diamond aggregate technique of Baker and Stolper [1994] and Hirose and Kushiro [1993]).

Table 1. Run Conditions of Experiments
ExperimentStarting MaterialType/CapsuleP (GPa)T (°C)t (hours)*
  1. a

    For two stage experiments, denotes time of 2nd stage.

Phase-Relation Experiment
PR-2CAM-Lsingle/Pt0.9100024
PR-6El-10single/Pt0.913006
PR-13KLB-1, vitreous carbonsingle/Pt0.9130024
 
Direct Mixture Experiments
MIX-142:1 mix KLB-1 and EL-10single/Pt0.913001
MIX-162:1 mix KLB-1 and EL-10single/Pt0.913000.25
MIX-3810:1 mix KLB-1 and EL-10single/Pt0.913002
 
Infiltration-Reaction Experiment
IR-11KLB-1 and EL-10single/Pt0.9130012
IR-33KLB-1 and EL-102 stage/Mo-Pt0.912000.5
IR-36KLB-1 and EL-10-1% Cl2 stage/Mo-Pt0.913000.25
IR-37CAM-H and EL-10-1% Cl2 stage/Mo-Pt0.513500.33
IR-39KLB-1 on top of SiO2+OPX and EL-10-1% Cl2 stage/Mo-Pt0.913002
 
Melt-Melt Diffusion Experiments
MMD-26A2384-1 and EL-102 stage/Mo-Pt0.914505 min
MMD-30A2384-1 and EL-102 stage/Mo-Pt0.914507 min
MMD-21A2384-1 and EL-102 stage/Mo-Pt0.914500 min

[6] All experiments were run in sealed metal capsules with graphite inner capsules. With graphite present, oxygen fugacity is below the graphite-CO vapor buffer, which lies within the wustite stability field at these temperatures [Ulmer and Luth, 1991]. Diffusion experiments (IR or MMD) were either single-step or two-step piston cylinder experiments. For single-step experiments, a platinum capsule with inner graphite capsule was welded prior to the piston cylinder run. Two-step experiments consisted of a synthesis step in the piston cylinder (24–48 hours for partially molten peridotite; 6–10 hours for basanite), removal of a small portion of each material for characterization, and final juxtaposition of the polished ends of two capsules in a second piston cylinder run. An advantage of the two-stage technique is that the capsule suture clearly defines the original interface of the couple. In all experiments, the location of the peridotite-basanite interface after the run was identical to that of the capsule suture indicating that a net transfer of mass across the interface was not measurable. Two-step experiments used graphite capsules inside a Pt-lined 6.3 mm OD Mo capsule with pressure sealing the Pt lid [Ayers et al., 1992].

[7] Experiments used 3/4″ salt-pyrex assemblies and crushable MgO spacers. On the basis of bracketing the melting point of ultra-pure NaCl using the falling sphere method, a 10% friction correction was applied to measured pressures. A WRe3/WRe25 thermocouple monitored temperature with no correction made for the effect of pressure on voltage. The length of the juxtaposed metal capsules in two-step experiments sometimes exceeded 4 mm. On the basis of a two-thermocouple experiment, the maximum temperature difference along any capsule assembly was <10°C. Experimental time-temperature conditions are given (Table 1). Experiments were taken to pressure without heating, ramped to final temperature, with final pressure adjustments made by increasing pressure. Ramp rates of 500°C/min minimized ramp-up times in time sensitive experiments.

[8] Peridotite starting materials consisted of two different spinel lherzolites and a harzburgite. CAM-L (lherzolite) and CAM-H (harzburgite) are recombined, handpicked mineral separates from a spinel lherzolite xenolith from Cameroon (CAM-027; collector J. Devine). These separates were rinsed gently in dilute HNO3, ground and sieved to <44 μm powders; grain sizes < 10 μm were removed by settling in water. CAM-L recombined these separates into 63% olivine, 25% orthopyroxene, 10% clinopyroxene and 2% spinel by weight while CAM-H recombined these separates into 80% olivine and 20% orthopyroxene by weight. KLB-1 is a spinel lherzolite powder having 57% olivine, 25% orthopyroxene, 15% clinopyroxene and 2% spinel with grain size of 10–50 μm [Takahashi, 1986]. The clinopyroxene, orthopyroxene and olivine from the Cameroon xenolith and KLB-1 are very similar in terms of Mg# and Al2O3 contents and melt compositions of CAM-L are similar to those of KLB-1 under identical conditions (1300°C and 0.9 GPa). The basalt starting material is a primitive basanite (EL-10) from the 1730–36 Timanfaya eruption on Lanzarote in the Canary Islands (Table 2). In three experiments, NaCl was added to the basanite (mixed by grinding prior to synthesis runs) to increase the amount of Cl from <0.1 to ∼0.9 wt%. Last, melt-melt diffusion experiments juxtaposed the basanite with a primitive MORB tholeiite glass from the Siqueiros Transform (A2384-1 [Perfit et al., 1996]), similar in composition to the equilibrium melt of KLB-1 (Table 3). Storage of all powders and loaded capsules at 140°C for 1 day prior to a run reduced the water content of the experiments.

Table 2. Starting Material Compositions
 EL-10aA2384-1bKLB-1c
SiO242.8648.8844.48
Al2O311.6017.503.59
Na2O3.722.360.30
MgO14.249.6039.22
P2O50.980.060.03
K2O1.510.000.02
CaO10.0712.223.44
TiO22.980.920.16
FeO11.827.778.10
MnO0.200.160.12
Cr2O3  0.31
NiO  0.25
Total99.9899.47100.02
Table 3. Melt Composition (Vitreous Carbon Layer) and Modal Analysis of PR-13; 0.9 GPa, 1300°C
 This StudyErrorHirose and Kushiro [1993]
SiO248.7(6)50.5
Al2O318.4(5)17.9
MgO9.6(2)10.1
CaO11.8(1)11.4
FeO7.2(6)6.7
Na2O2.45(10)2.47
K2O0.05(2)0.09
TiO20.61(3)0.65
Total99.17  
 ModeError (1σ) 
CPX + Sp9.2±1.8 
OPX24.5±3.9 
OL59.6±4.1 
MELT6.7±2.0 

[9] Major element contents were measured using either a Cameca Camebax (Brown University) or SX-50 (University of Chicago) electron microprobe. Conditions of analysis were 15 kV and 10 nA with a spot size of ∼5 microns. Standards included a combination of minerals and glasses. Analyses were normalized to daily measurements of Smithsonian standards. Ion probe analysis of one melt-melt diffusion run occurred on the Cameca IMS 3f at Woods Hole Oceanographic Institution. Energy filtering reduced molecular ion interferences [Shimizu et al., 1978].

[10] Mineral modes for the experiments were quantified using digital analysis of 512 × 512 pixel (250 μm × 250 μm) backscattered electron (BSE) images captured using a Cameca SX-50 electron microprobe at the University of Massachusetts-Amherst or a JEOL 840A SEM in the Geology Department at UIUC (Appendix A). Simultaneous collection of Mg x-ray maps allowed positive identification of phases in the BSE image. Analysis of gray scale variations using NIH image software allowed quantification of phase proportions.

3. Results and Observations

3.1. Phase Relations of Starting Materials

[11] The melting mode and melt composition of KLB-1 at 0.9 GPa and 1300°C were determined in a static melting experiment containing vitreous carbon. The melt composition agrees with previous determinations at similar pressure-temperature conditions [Hirose and Kushiro, 1993] (Table 3). Modal analysis using digital imaging agrees with the degree of melting based on mass balance. Using the measured mode of clinopyroxene and orthopyroxene and assuming Na partitioning as given in Hirschmann [2000], the calculated degree of melting (0.09) for experiment PR-1 is only slightly outside the 1σ error of the melt mode determined by imaging. A single experiment (PR-6) on EL-10 shows that the basanite crystallizes 10% olivine at 0.9 GPa and 1300°C, in excellent agreement with that predicted by the MELTS algorithm [Ghiorso and Sack, 1995].

3.2. Direct Mixture Experiments

[12] The mineral modes after two direct mixture experiments of different length (15 and 60 min) overlap at the 1σ level (Figure 1). This result indicates that the basanite reacts with the peridotite minerals in less than 15 min to approach equilibrium with some zoning of the solid phases remaining after the experiment. In further support of a fast approach to equilibrium, the major element glass concentrations in the two experiments are generally within errors of one another despite obvious problems with quench crystallization and zoning of minerals (Table 4). Although the two liquids fail a homogeneity test, only 3 elements deviate outside their combined errors; for MgO, this simply reflects different amounts of glass modification during quench, which depends on the probe analysis locations. On the basis of these experiments, the kinetics of mineral dissolution do not play a rate-limiting role in affecting the melt chemistry in the infiltration-reaction experiments.

Figure 1.

Modal proportions of minerals and melt from two direct mixture experiments of KLB-1 (2 parts) and El-10 (1 part) as a function of time. The predicted modes prior to reaction based on the phase relations and proportions of KLB-1 and EL-10 at these conditions are given on the left axis. A rapid decrease in the mode of orthopyroxene and increase in the olivine and melt modes occurs during the first fifteen minutes of reaction, consistent with orthopyroxene incongruently dissolving to form olivine and plus silica rich melt. The addition of silica results in a tholeiitic melt, as expected for equilibrium with a spinel lherzolite at these conditions. Although the modes change slightly in the longer run experiment (possibly due to inaccuracies with differentiating melt from olivine quench rims in the image analysis), little change in orthopyroxene mode or melt composition (Table 4) occurs with increased reaction time. This indicates that mineral dissolution does not play a rate-limiting role in the infiltration-reaction experiments. That melt-solid equilibrium is quickly approached is further evidenced by the constancy of melt compositions in direct mixture experiments, despite differing reaction times or peridotite-basanite ratios (Table 4).

Table 4. Melt Analyses of MIX-14, MIX-16 and MIX-38
Ratio of KLB-1:El-10MIX-16 2:1MIX-14 2:1MIX-38 10:1
(0.25 hr)error(1.0 hr)error(2 hr)error
SiO251.00.451.00.649.70.5
Al2O315.60.916.70.316.00.5
Na2O2.70.13.10.22.90.1
MgO6.50.15.50.67.31.2
P2O50.60.10.50.00.60.1
K2O1.10.31.10.10.90.1
CaO11.51.712.00.811.40.4
TiO22.00.02.30.12.10.1
FeO7.50.57.40.67.00.4
MnO0.10.10.10.10.20.1
Total98.6 99.5 98.0 

[13] Given knowledge of the modal composition of each starting material at these P-T conditions (section 3.1), the observed modes of the mixture after reaction can be compared with the modes predicted based on the 2:1 proportion of the starting materials. Within 15 min, the orthopyroxene mode decreases from an initial mode of 16.7% to an observed mode of 5.7%. The clinopyroxene mode decreases from 6% to 4%, the olivine mode increases from 43% to ∼51% and the melt mode increases from 34.6% to ∼40%. These data suggest that during reaction with the basanite, orthopyroxene incongruently dissolves to form silica rich melt and olivine as predicted by Green and Ringwood [1967], Quick [1981], Fisk [1986], and Kelemen [1986]. The SiO2 content of the melt increases to ∼51 wt.%, a value expected for a melt with moderate Na2O concentration in equilibrium with spinel lherzolite minerals under these pressure and temperature conditions. Thus high SiO2 melts are not produced during direct melt-orthopyroxene reaction at these conditions.

[14] This conclusion does not depend on the ratio of peridotite to basanite. A second mixture of 10 parts KLB-1 to 1 part El-10 was run for two hours at the same P-T conditions resulting in a melt having ∼50 wt% SiO2 (Table 4). Indeed, the 10:1 and 2:1 mixture experiments result in nearly identical melt compositions further indicating that melt-mineral equilibrium is rapidly approached. Melt compositions in experiments with higher KLB-1 to El-10 ratios will approach that of the tholeittic melt in equilibrium with KLB-1 under these conditions. Lower KLB-1 to El-10 ratios increase the proportion of silica-poor melt and therefore cannot produce >51 wt.% SiO2 melts. Direct basanite-peridotite reaction does not result in higher SiO2 melts because, although alkali addition decreases the SiO2 activity coefficient of the melt (γSiO2), TiO2 and P2O5 are also added which have the opposite effect on γSiO2 [Kushiro, 1975; Hirschmann et al., 1998].

3.3. Melt-Melt Diffusion Experiments

[15] Tholeiitic melt-basanitic melt diffusion experiments provide constraint on the diffusion rates of both major and trace elements in the IR experiments. In order to be above the liquidus of both melt compositions, experiments were run at 1450°C, 250–350° above the temperature of the IR experiments. The high temperatures and limited capsule length required that experiments be only 5 and 7 min in duration. A “zero-time” experiment (ramp-up with immediate quench; 2.9 min total; ∼1.5 min above solidus) produced a profile width for slow diffusing elements in this experiment of <100 μm and ∼1 mm for Na2O.

[16] Reported diffusion coefficients reflect linear regression techniques (Table 5) [Mungall and Dingwell, 1997]. Using a typical activation energy for melt diffusion of 150 KJ reduces diffusion coefficients measured at 1450°C by a factor of ∼3 to values appropriate to the IR experiments. Multicomponent diffusion in a one-phase system with concentration gradients can be described by Fick's first law in which the flux of an element is equal to the diffusion matrix multiplied by the concentration gradient vector [Shewmon, 1963]. However, in complex multicomponent systems like natural silicate melts, most elements produce simple binary diffusion profiles. Cooper [1968] introduced the concept of the effective binary diffusion coefficient (EBDC) applied here.

Table 5. Regressed Effective Binary Diffusion Coefficients for MMD Experiments (1450°C)
 Experiment MMD-30Experiment MMD-26
D × 106 cm2/sErrora (−)Errora (+)D × 106 cm2/sErrora* (−)Errora (+)
  • a

    Error reflects variation in last decimal place given (e.g. 0.49 (−0.08/+0.10)).

Si0.49−8101.1−22
Al0.61−9110.93−77
Mg0.54−7100.77−1115
Fe0.62−14201.0−23
Ca5.2−12208.1−40120
K3.2−333.4−56
P0.37−8110.32−57
Ti0.19−230.73−77
Cl1.0−35   
La1.4−22   
Ba1.4−12   
Nb0.9−22   
Zr0.4−12   
Sr1.7−11   
Rb1.3−11   
Li>6     

[17] All elements except Na and perhaps Ca show simple binary diffusion profiles (Figure 2). Sodium is the major exception to EBDC behavior showing a pronounced hump in concentration due to its strong coupling to gradients in SiO2 concentration [Watson, 1982; Lesher, 1994; Lundstrom, 2000]. Although the Ca profiles for the two melt-melt diffusion experiments can be regressed as EBDCs, the data are more scattered than can be attributed to analytical error with R2 of 0.81 and 0.36 (compared to >0.89 for all other elements in either experiment). Furthermore, the EBDC for Ca is significantly larger than reported in previous studies [Watson, 1979].

Figure 2.

Diffusion profiles for major elements in a basanitic melt-tholeiitic melt diffusion experiment (MMD-30; 1450°C, 0.9 GPa, 7 min). Note that all elements except Na and possibly Ca show EBDC behavior with EBDCs in the 10−6 to 10−7 cm2/s range (Table 5). Elements with higher field strength (SiO2, TiO2, P2O5) produce distinctly steeper diffusion profiles than alkali and alkaline earth elements. Na produces a humped profile consistent with coupling to gradients in SiO2 [Lundstrom, 2000].

[18] EBDCs for other major elements compare well with previous determinations at similar temperatures. Diffusion coefficients of basaltic melt cations at slightly lower temperatures (1300°C) generally fall in the 10−6 to 10−7 cm2/s range [Margaritz and Hofmann, 1978; Watson and Baker, 1991]. All trace elements form binary EBDC profiles with variations in diffusion coefficient consistent with changes in size and chemical behavior (Figure 3). The fastest diffusing cations are alkalis and alkaline earth elements, while high field strength elements are the slowest diffusing elements. EBDCs of alkali elements increase with decreasing cationic radius. Excluding Na, whose profile reflects coupling to SiO2, Li is the fastest diffusing element with a profile reaching the ends of the capsule in less than 7 min providing a minimum estimate for DLi of 6 × 10−6 cm2/s. Last, for interpreting the Cl enhanced infiltration-reaction experiments, the profile of Cl constrains DCl to be 1 × 10−6 cm2/s.

Figure 3.

Diffusion profiles for minor and trace elements for melt-melt diffusion experiment MMD-30. (a) Diffusion profiles for Na and Cl. Circles show modeled binary diffusion profile for DCl = 1 × 10−6 cm2/s. Na profile is modeled similarly to the model in Lundstrom [2000] using the following terms DNa,Na = 2 × 10−5 cm2/s, DNa,Si = −4 × 10−6 cm2/s, DSi,Na = 1 × 10−8 cm2/s and DSi,Si = 5 × 10−7 cm2/s. (b) Normalized profiles for a variety of trace elements. All trace elements are consistent with binary diffusion behavior with alkalis and alkaline earths having higher EBDCs than high field strength elements. (c) Concentration profiles for the alkali elements. Note the progressive decrease in diffusion coefficient with increase in ionic radius for the alkalis in general. As noted above, Na does not follow EBDC behavior.

3.4. Infiltration-Reaction Experiments

[19] Initial IR experiments were run as layers of peridotite and basalt powders for several hours duration. IR-11 reflected a layer of basanite separated from KLB-1 by a thin graphite disk perforated with nine ∼50 μm holes. The KLB-1 powder contained a layer of vitreous carbon far from the basanite interface to quantify the composition of the melt unmodified by quench crystallization (Figure 4a). After 12 hours, ∼7% orthopyroxene remained in the peridotite 1.2 mm from the basanite while 4% OPX remained in the section 0.6 mm from the basanite. These modes contrast with 25% orthopyroxene present under these same P-T conditions in the static melting experiment (experiment PR-13).

Figure 4.

Results from IR-11, a single stage infiltration-reaction experiment between KLB-1 and El-10 (1300°C, 0.9 GPa,12 hours). (a) the geometry of the experiment consists of a vitreous carbon layer within the KLB-1 powder, covered by a perforated graphite disk with basanite EL-10 on top. Modal results after infiltration-reaction show a decrease in the orthopyroxene mode approaching the basanite. The large errors reflect widely varying extents of reaction due to the limited pathways for melt modulated by the graphite disk; significantly greater orthopyroxene dissolution occurred near the gap between the disk and the capsule wall than near the holes in the disk. Olivine grains along these melt pathways grow significantly (>100 μm euhedral grains) during the 12 hours of this experiment. (b) Elemental concentration profiles of the quenched glass as a function of distance down the experiment (Top = 0). Left panels show measured melt compositions while right panels show data that have been corrected for quench crystallization by addition of olivine that is in equilibrium with the melt (KD = 0.29) in 1% increments until the melt reaches 10 wt.% MgO. Gray band shows location of graphite disk. The initial starting basanite composition is given on the left axis while the expected melt of KLB-1 under these P-T conditions (Table 3) is given on the right axis. The melt within the vitreous carbon layer has lower Al2O3 and CaO and higher Na2O, TiO2, K2O and P2O5 contents than the expected melt of KLB-1 at these conditions. This cannot reflect quench modification. Although enrichments in TiO2 and P2O5 are most easily attributed to diffusion from the basanite, the distance between the vitreous carbon and the basanite is greater than would be predicted based on the measured melt-melt diffusion coefficients.

[20] Melt pool composition as a function of distance shows that all elements are transported extensively over the 12 hr experiment (Figure 4b). In this as in all IR experiment figures, I have presented both measured melt compositions and melt compositions corrected for quench crystallization to 10 wt.% MgO by addition of olivine. The discussion explains the correction method and assumptions involved in this correction in more detail. The concentrations of slow diffusing elements (e.g., TiO2 and P2O5) throughout the charge are elevated relative to the melt of KLB-1 alone (Table 3). The analysis of melt in the vitreous carbon layer (located at 1.6 mm) demonstrates that these elevated concentrations do not reflect quench modification. Given the Ti diffusion coefficient (Table 5) corrected to 1300°C, the expected length scale of diffusion is ∼0.3 mm. Thus chemical transport in this experiment is more rapid than can be explained by diffusion using the measured melt-melt diffusion coefficients. The initial melting of the powder starting materials likely causes advective transport since convection is unlikely to occur given the relatively low melt porosity within the peridotite.

[21] Two-stage experiments avoided this problem by juxtaposing pre-synthesized basanitic glass and peridotite + equilibrium melt (quenched as glass during first stage). When the initial melt porosity of the lherzolite is 10%, Na diffusion extends more than 1 mm into the lherzolite in just 10 min [Lundstrom, 2000]. However, melt porosities during mantle melting are likely <10% based on theory and observation [McKenzie, 1984; Toomey et al., 1998]. In order to investigate the DIA process at low melt porosity, experiment IR-33 juxtaposed KLB-1 near its solidus [Hirose and Kushiro, 1993] with EL-10 for 30 min. Because the initial melt porosity in KLB-1 is <2%, melt pools within the lherzolite large enough to analyze commonly develop within 100 μm of the interface but are much less common beyond the interface region. Near-interface melt pools have elevated concentrations of Na2O, K2O, and SiO2 but low concentrations of TiO2 (<1 wt.%) and P2O5 (Figure 5). Potassium diffuses up gradients in concentration into the lherzolite reaching values greater than 3 wt.%. The peak in SiO2 concentration (∼57 wt%) coincides with the Na2O and K2O peaks (although there are relatively few analyses). The compositions of the few large melt pools that occur in the lherzolite beyond 100 μm from the interface depend on proximity to the capsule walls. Melt pools that exist along the wall of the graphite capsule (open symbols in Figure 5) have high concentrations of TiO2 and P2O5, consistent with melt advection along the peridotite-capsule boundary. Interior melts are low in TiO2 and reflect diffusive transport alone.

Figure 5.

Concentration profiles for melts in experiment IR-33, a two-stage experiment run at 0.9 GPa and 1200°C for 30 min. Left axis contains the initial basanite glass composition. Despite a melt fraction of <2% in the KLB-1 starting material, relatively large melt pools containing elevated SiO2, Na2O and K2O contents are found within the lherzolite near the interface. The increase of Na2O and K2O in the melt pools in the peridotite is complemented by depletions in these elements in the basanite. Open symbols represent melt pools located along the wall of the graphite capsule. High concentrations of TiO2 and P2O5 within these melts probably indicate that melt has flowed along the peridotite-graphite boundary. This effect was not observed in other experiments.

3.5. Experiments Using the Enhanced-Cl Basanite

[22] Three experiments juxtaposing the Cl-enhanced basanite with peridotite produce melt compositional profiles distinct from experiments without added Cl. The addition of Cl appears to accelerate the reaction process leading to development of a “boundary zone” within the peridotite where all pyroxene has been dissolved. This likely relates to the dramatic effect of Cl on the phase relations of mafic minerals and melt [Webster and McBirney, 2001]. The boundary zone appears uniform in lherzolite experiments but the pyroxene-free area in the harzburgite is more finger like [Daines and Kohlstedt, 1994]. Systematic enrichments in CaO and Cl provide a possible fingerprint of the occurrence of this boundary zone in nature.

[23] Melt pool concentration profiles in an enhanced-Cl basanite and lherzolite couple (IR-36, 0.9 GPa, 1300°C, 15 min) resemble that of other IR experiments having coincident Na2O, K2O and SiO2 concentration peaks ∼200 μm from the interface (Figure 6). However, rather than forming a step in concentration at the interface [e.g., Lundstrom, 2000], melt compositions steadily increase to the peak concentrations. The boundary zone can be defined as the region between the interface and the peak in SiO2 and contains glass compositions that reflect neither the basanite nor a mixture of the basanite and tholeiite. Rather, the raw data for CaO, TiO2 and Cl show pronounced concentration peaks, exceeding both the basanite and tholeiite concentrations of these elements while Al2O3, SiO2, FeO and Na2O show noticable breaks in slope at this location. Whether this systematic behavior represents bulk infiltration of basanite followed by diffusive or quench modification or diffusion alone (as in the non-Cl-enhanced experiments) is not clear. The width of the boundary zone is wider than predicted based on Ti diffusion (∼100 μm, assuming DTi for 1300°C and a melt fraction of 0.2). Either way, the reproducible formation of this boundary zone may indicate an important aspect of the DIA process.

Figure 6.

(a) Profiles of melt pool concentration in IR-36, a two-stage experiment using the enhanced-Cl basanite (1300°C, 0.9 GPa, 15 min). The initial basanite and lherzolite glass compositions are given on the left and right axes, respectively. In contrast to IR-33, a more gradual change in the silica content of melt pools is seen at the interface with an abrupt increase in silica ∼180 μm into the peridotite (see text). The area between the interface and the peak in silica reflects the boundary zone having melt CaO and Cl concentrations that are greater than either the original basanite or the tholeiitic melt within the peridotite. (b) Cl/TiO2 and Cl/P2O5 as a function of distance in IR-36. Chlorine appears to have diffused from both the basanite and the tholeiite toward the boundary zone. For instance, the melt pools within the lherzolite farthest from the interface have Cl/P2O5 that is lower than those in the melt pools in the initial KLB-1 + melt starting material.

Figure 6.

(continued)

[24] Chlorine concentrations in the boundary zone are ∼10% higher than those in the basanite. Mass balance indicates that this enrichment is not solely due to the quench. The basanite becomes depleted in Cl, changing from 0.90 ± 0.02 wt.% in the starting material to 0.81 ± 0.03 after the experiment. Although the Cl content of the glass within the lherzolite-melt starting material was 0.1 wt.%, the glass Cl content in the lherzolite far from the interface after infiltration-reaction is 0.03 wt.%. Thus Cl appears to diffuse toward the boundary zone from both ends of the experiment, shown best by Cl/P2O5 decreasing to values less than the starting material in the lherzolite far from the interface (Figure 6b).

[25] A second Cl-enhanced basanite experiment (IR-37) examined infiltration into harzburgite. Although synthesis run conditions were chosen to produce significant amounts of melt within the harzburgite (0.5 GPa, 1350°C), <2% melt existed in the synthesized starting material based on image analysis. Nevertheless, after 20 min juxtaposition with basanite, analyzable melt pools occur throughout the harzburgite. Na2O and SiO2 concentration profiles peak at higher concentrations (6.3 wt% and 64 wt%, respectively) and at greater distances from the interface (∼400 μm) than in any other experiment (Figure 7). Similar to experiment IR-36, CaO, TiO2 and Cl all show concentration peaks in the uncorrected data. Quench corrected data continue to show a peak in CaO. Modal changes clearly indicate incongruent dissolution of orthopyroxene as its mode decreases and the melt and olivine modes rise approaching the interface (Figure 8). Cl/P2O5 rises continuously through the harzburgite indicating the mobility of Cl relative to higher charged cations.

Figure 7.

Profiles of melt concentrations in IR-37, a two-stage infiltration-reaction experiment between Cl-enhanced basanite and harzburgite with less than 2% melt (1350°C, 0.5 GPa, 20 min). Like IR-36, a boundary zone with peaks in CaO and Cl and occurs next to the interface. The highest SiO2 and Na2O contents of any melt as well as the greatest distance of alkali infiltration of any IR experiment are found in this experiment.

Figure 8.

Mode determination and Cl/TiO2 and Cl/P2O5 as a function of distance in experiment IR-37. A significant decrease in the orthopyroxene mode and increase in the melt and olivine modes occurs as the interface is approached. This is fully consistent with incongruent dissolution of orthopyroxene caused by the addition of alkalis by diffusion. Cl/TiO2 and Cl/P2O5 progressively increase away from the interface indicating enrichment in Cl relative to Ti and P during the DIA process. Note that no initial determination of Cl/TiO2 and Cl/P2O5 in the harzburgite-melt starting material is possible due to the lack of large melt pools.

[26] A third Cl-enhanced basanite IR experiment tested the hypothesis that the CaO concentration peak in the boundary zone results from Ca diffusing out of the peridotite to charge balance Na infiltration. To enhance Na diffusion in the experiment, a layer of orthopyroxene + silica was placed below the KLB-1 to act as a sink for Na. In the first stage, a 1 mm thick layer of KLB-1 covered by ∼50-50 mixture of orthopyroxene and SiO2 was synthesized for 1.5 hours at 1300°C and 0.9 GPa. The KLB-1 + melt end was then juxtaposed with EL-10 + Cl for 2 hours. The melt composition profile is similar to those described above with a boundary zone defined by both melt composition (Figure 9) and the visual absence of pyroxene (Figure 10). Na2O within the basanite is depleted relative to the starting material while Na2O contents of ∼5 wt.% occur throughout the melt pools within both the KLB-1 and QTZ plus OPX portions of the experiment. Like the other Cl-enhanced experiments, a boundary zone with concentration peaks in CaO and Cl forms between the peak in SiO2 and the interface.

Figure 9.

Profiles of melt concentrations in IR-39, a two-stage infiltration-reaction experiment (1300°C, 0.9 GPa, 2 hours). In order to accentuate sodium infiltration through the peridotite, the peridotite half of this experiment consisted of a layer KLB-1 juxtaposed with a SiO2 plus orthopyroxene mixture. This was synthesized for 1.5 hours at the same P-T conditions as the final experiment resulting in a capsule with KLB-1 plus melt in one half and orthopyroxene plus quartz plus silica rich melt in the other half. The face of the KLB-1 plus melt half was juxtaposed with Cl-enhanced basanite for 2 hours. Melt pool Na2O contents are elevated throughout the peridotite half of the experiment and correspondingly depleted from the basanite. Like the other Cl-enhanced basanite experiments, a boundary zone with peaks in CaO and Cl develops adjacent to the interface. MgO diffuses out of the lherzolite and into the basanite as evidenced by the increase in MgO in the basanite as well as the olivine compositional changes (Figure 11).

Figure 10.

Backscattered electron image from the boundary zone/interface region of IR-39. Note the absence of pyroxene within the first 150 μm, the boundary zone which corresponds to the enrichments in melt CaO and Cl contents. The increase in melt CaO concentration in this region can be seen by the more bright appearance of the melt in the lower portion of the image compared with the melt in the upper portion. Note that although clinopyroxene dissolution could explain the creation of the CaO peak in the boundary zone of this experiment, this mechanism does not explain the observed CaO peak in the harzburgite experiment.

[27] Olivine compositions change over the timescale of the experiment as a function of distance from the interface indicating fast melt-solid equilibration (Figure 11). Cores as well as rims of olivines 50–70 μm in size in the boundary zone show a pronounced peak in CaO mimicking the CaO enrichment in the boundary zone melt. This indicates fast re-equilibration of olivine during the development of the boundary zone [Gaetani and Watson, 2000]. Olivine rims throughout the basanite have remarkably constant Mg# and are in exchange equilibrium with the quenched basanitic melt (KD = 0.29). In contrast, the cores of these olivines (30 μm in size) change from having Mg# less than the rims far from the interface to having Mg# greater than the rims closer to the interface. Indeed, the core of an olivine in the basanite farthest from the interface is in Mg-Fe exchange equilibrium with the starting basanite prior to the diffusion-reaction stage.

Figure 11.

Mg# (Molar Mg/(Mg + Fe+2)) and CaO for olivines within IR-39. Crosses inside points represent rim compositions (analysis within 5 μm of rim) while symbols without crosses represent core analyses (within the middle of 50–70 μm grains in the lherzolite and 30 μm grains in the basanite). Generally, points at similar distance represent core and rim analyses of the same crystal. There is an overall trend of decreasing CaO content and increasing Mg# going from the basanite to the interior portion of the KLB-1 half of the experiment. Two notable features stand out when cores and rims are examined separately. First, olivine cores within the boundary zone have elevated CaO contents, consistent with the increased CaO content of melts in the boundary zone. Thus compositions of olivine cores change over the 2 hour timescale of the experiment. The high CaO content in olivine could provide a signature of boundary zone development in the mantle. Second, the Mg# of olivine rims within the basanite remains constant throughout the basanite. In contrast, olivine cores in the basanite progressively increase in Mg# approaching the interface. The rims are in Mg-Fe exchange equilibrium with the quenched melt (KD = 0.29). The change in core composition is interpreted to result from olivine re-equilibrating with the basanite melt which is continuously increasing in Mg# as Mg diffuses from the lherzolite into the basanite. The change in melt Mg# causes Mg-Fe exchange resulting in cores getting progressively richer in Mg# approaching the interface. Notably, the olivine core farthest from the interface has a Mg# in exchange equilibrium with the original basanite melt starting material (KD = 0.29).

[28] One possible explanation for the Mg# rim-core systematics is that olivine crystals from the lherzolite detach and ascend into the basanite; this would explain why the cores of the olivines have not re-equilibrated with the basanite. However, there is no evidence for a loss of crystals from the lherzolite at the interface (the lherzolite boundary remains sharp) nor for two populations of olivine within the basanite. A more favored explanation is that the Mg# systematics reflect olivine compositional changes in response to Mg diffusion through the melt from the peridotite to the basanite. Evidence for Mg diffusion from the lherzolite to the basinite is provided by the basanite MgO content increasing from 9.5 wt% initially to 10.5 wt% over the course of the experiment. As Mg diffuses into the basanite (up a concentration gradient), olivines in the basanite close to the interface must re-equilibrate with a higher Mg# melt. Because the basanite is already saturated in olivine, re-equilibration is accommodated by Mg-Fe diffusive exchange within the olivine. This results in Fe diffusing toward the rim producing a higher Mg# core in olivines near the interface. Because this process occurs as a function of proximity to the interface, olivines far from the interface fail to see the MgO diffusive flux into the basanite. Thus olivines far from the interface maintain core compositions identical to their starting Mg#. In contrast to crystal cores, rapid Mg-Fe exchange allows rims to always maintain Mg-Fe exchange equilibrium with the surrounding melt. On the basis of measure rates of Mg-Fe exchange in olivine [Jurewicz and Watson, 1988], the length scale of diffusive adjustment should be ∼24 μm, similar to the olivine grain size in the basanite; therefore, the observed profiles are consistent with this explanation.

4. Discussion

4.1. General Observations and Interpretation of the Experiments

[29] Quench crystallization, evidenced by low MgO contents of melt pools, is an intrinsic process operating on the melt compositions in the IR and MIX experiments. However, the relatively large enrichment in alkalis and silica in glasses near the interface does not solely reflect quench modification since these melt pools are the largest, and therefore least affected by quenching. Comparison of quench-modified glass and quench-free glass (in vitreous carbon layers) from the same experiment [Lundstrom, 2000, Figure 2] indicates the following concentration increases due to quenching: SiO2, +1%, Al2O3, +8%, FeO, +7%, CaO, +4%, Na2O, +12% and TiO2, +16%. These changes are small relative to the changes in the melt pool compositions observed spatially in traverses through the IR experiments. Because quench crystallization does affect the results, melt compositions for all IR experiments are presented both in raw form and after correction back to MgO of 10 wt.% through incremental addition of olivine in equilibrium with the melt. This process will not correct all melt pools back to their original composition given that some pyroxene crystallization may also occur; however, it should adequately compensate for the enrichment in incompatible elements from quenching and reduce some of the variability in the profiles.

[30] The rapid infiltration of Na into partially molten peridotite causes incongruent dissolution of orthopyroxene. Na2O diffuses >600 μm into the peridotite in 10–30 min experiments with the peak in Na2O generally coincident with the peak in SiO2. High SiO2 contents are unlikely to reflect diffusion because Si is typically the slowest diffusing species in silicate melts [Watson and Baker, 1991]. Rather, the correspondence between the Na2O and SiO2 peaks indicates a causal relationship between the addition of Na2O and reaction to form SiO2 rich melt. Each infiltration-reaction experiment shows a decrease in the orthopyroxene mode in the area of alkali infiltration but little reduction far from the interface. Complimentary decreases in orthopyroxene mode and increases in olivine and melt modes in the harzburgite-Cl experiment (IR-37) indicate incongruent dissolution, reflecting the shift in the olivine-orthopyroxene phase boundary with addition of Na to the melt [Kushiro, 1976; Ryerson, 1985].

[31] Shaw et al. [1998] and Shaw [1999] have shown that silica rich melt forms around orthopyroxene as it dissolves in alkali basalts. However, Shaw [1999] concludes that direct melt-mineral reaction is unlikely to explain the creation of silica rich glasses in mantle xenoliths because continued reaction and diffusion quickly erase any boundary layer around orthopyroxene. The DIA mechanism is distinct from the dissolution experiments of Shaw and co-workers as it reflects addition of alkalis into the peridotite without addition of slow diffusing elements like TiO2 which have opposite effects on melt SiO2 content [Hirschmann et al., 1998]. Indeed, the direct mixture experiment results show that addition of both alkalis and TiO2 and P2O5 to KLB-1 result in melts with small to negligible increases in SiO2 content. In contrast, DIA leads to high silica melts that can be in equilibrium with peridotite over the lengthscale of a xenolith.

[32] Little is known about the actual melt species diffusing during multicomponent diffusion in silicate melts. Charge neutrality within an experiment must be maintained during diffusion but whether individual ions diffuse or ionic complexes diffuse is not known. During basalt-granite diffusion-reaction experiments, Watson and Jurewicz [1984] observed high rates of Na diffusion with no obvious mechanism for maintaining charge neutrality. These authors suggested that O−2 diffusion compensated for Na+ diffusion. In the basalt-granite case, Na will diffuse in the same direction as Ca (from basalt into granite) whereas activity gradients drive Na and Ca to diffuse in opposite directions during a basanite-tholeiite couple as in the IR experiments. Although a definitive interpretation is not possible, several observations suggest that Ca counter diffusion occurs in response to Na diffusion. First, Ca diffuses faster than all other cations except Na and Li in melt-melt diffusion couples, with poor fits to binary profiles. Second, the CaO glass concentration in IR experiments is low in areas with elevated Na2O concentrations. Although decreasing CaO content could reflect local increase in the clinopyroxene mode, Na2O infiltration should destabilize clinopyroxene resulting in increased CaO in the melt, not decreased. Increased clinopyroxene modes that could explain CaO depletions are not observed in any experiment. Finally, in experiment IR-39, there is also evidence for MgO diffusing out of the peridotite.

[33] The repeated formation of the boundary zone in the enhanced-Cl basanite experiments may indicate an important mechanism within the DIA process. All 3 experiments produce peaks in CaO and Cl within the reacted peridotite devoid of pyroxene adjacent to the interface. Similar elemental behavior (CaO, TiO2 and Al2O3 enrichment) is observed within other diffusive infiltration-reaction experiments (Z. Morgan, and Y. Liang, An experimental and numerical study of the kinetics of harzburgite reactive dissolution with applications to dunite dike formation, manuscript submitted to Earth and Planetary Science Letters, 2003). Although the peak in CaO in the two lherzolite experiments could reflect clinopyroxene dissolution, this cannot explain the CaO enrichment in the harzburgite experiment.

[34] Chlorine enrichments may fingerprint boundary zone development during DIA in the mantle. Although Cl diffusion in silicic melts has been studied previously [Bai and Koster van Groos, 1994; Watson, 1991, 1994], little work on Cl diffusion in mafic melts exists. The relatively large Cl diffusion coefficient indicates that Cl can play an active role in DIA. The repeated correlation of CaO and Cl peaks and the observation that Cl diffuses toward the boundary zone from both ends of the couple in IR-36 suggests that Cl activity gradients are primarily influenced by variations in melt CaO content. This is consistent with work by Webster et al. [1999] and Webster and McBirney [2001] that have shown that Cl solubility is highly dependent on melt Ca and Mg contents, possibly due to the formation of Ca-Cl complexes [Webster and Mathez, 2001].

4.2. Is the DIA Process Important to Mantle Melting?

[35] Both trace element depletions in abyssal peridotites [Johnson et al., 1990] and the observation that MORB is undersaturated with orthopyroxene [O'Hara, 1965; Stolper, 1980] require that some of the melt beneath a ridge ascend by channelized flow [Kelemen et al., 1997]. On the basis of trace element relationships, Kelemen et al. [1995] have shown that dunites likely represent pathways for channelized flow. If silica-poor melt from depth ascends in channels, then gradients in melt silica content will likely exist between the melt in a conduit and the melt in surrounding peridotite at shallow mantle depths, causing DIA to occur. How important this process is in affecting the major element content of melts at the surface remains debatable. Addressing this requires better answers to three questions: (1) Over what depth range and for how long will a parcel of peridotite be exposed to the DIA process? (2) How is sodium originally distributed in the mantle source region and how do Na2O contents of ascending melts vary with depth? (3) How is a melt channel system spatially organized in a melting region?

[36] The IR experiments show that DIA could occur at 0.5–1 GPa pressure or 15 to 30 km deep in the mantle. The effects of DIA should become even more pronounced at shallower depths (<15 km) because the addition of Na has a greater effect on the melt silica activity coefficient at low pressure [Hirschmann et al., 1998]. Thus estimating that DIA occurs as peridotite ascends from 30 to 15 km depth (or even to the MOHO at fast spreading ridges) is not unreasonable. If the mantle passively upwells beneath ridges [Bottinga and Allegre, 1976; Sleep, 1975; Toomey et al., 1998; Lundstrom et al., 1998] at mm to cm/yr rates, peridotite adjacent to a melt conduit could be exposed to DIA for several million years. This timescale controls the flux of Na2O that the peridotite is exposed to.

[37] The amount of melting added by DIA will depend on the gradients in Na2O between channels and surrounding peridotite as well as the overall flux of Na2O through the channels, and the advective velocity of intergranular melt in the surrounding peridotite. Initial melts of peridotite should be rich in incompatible elements including alkalis. Alkali basalts, often interpreted as early formed small degree melts, typically have garnet trace element signatures indicating pressures of origin >2.5 GPa. However, melting experiments at >3 GPa relevant to a MORB source with 0.3 wt% Na2O produce low degree melts having low Na2O contents (<2 wt% [Longhi, 2002]). Thus some uncertainty exists about the Na2O contents of ascending basalts and gradients across a melt channel-peridotite interface. However, alkali rich basalts are observed in both MORB and OIB settings, attesting to the presence of melts similar to the basanite used here and possibly indicating sources significantly enriched in Na2O relative to depleted mantle.

[38] Inferring the gradients in Na that existed in the mantle during melting from observed melts at the surface is difficult because if the DIA process occurs, it acts to reduce any gradient in melt Na2O. Indeed, Na2O concentrations of MORB are much less variable than elements of similar incompatibility [Langmuir and Hanson, 1980] supporting the idea that Na variability may have been reduced by DIA. Melt inclusion Na2O contents, independent of being hosted by olivine or plagioclase, are also much less variable than TiO2, despite the fact that the titanium partition coefficient should be 2–4 times larger than sodium's in the shallow mantle (Figure 12). If DIA impacts melting significantly, mantle sources and initial melts might be considerably more heterogeneous in Na2O than can be inferred from observed melts.

Figure 12.

Histograms of melt inclusion Na2O and TiO2 concentrations for both olivine and plagioclase hosted samples from the study of Sours-Page et al. [2002]. Despite the fact that Na2O should be more incompatible in mantle minerals, variations in TiO2 are far greater than those of Na2O, regardless of host crystal. A possible explanation for this observation reflects the efficiency of the DIA process in homogenizing gradients in Na2O during the melting process.

[39] Assuming magma generation beneath a ridge is a steady state process, the flux of Na2O through a melt channel determines the extent of DIA influence on a parcel of peridotite adjacent to the channel. In one scenario, Na2O could be stripped from a wide region of the mantle by early formed deep melts beneath a ridge, ascend up through the melt conduit and disperse back into harzburgite during DIA at shallow depths. In this way, the Na2O flux resulting from melting over a wide area is focused onto a smaller area causing the peridotite at shallow depths to melt more extensively (a geometrical magnification of the amount of Na2O available for DIA). However, even if there is no geometric effect accentuating the Na2O flux, the DIA process could still dramatically alter melting at shallow depths. Effectively, the DIA process re-fertilizes a peridotite in Na2O at shallow depths where the effect of Na2O on melting relations (relative orthopyroxene-olivine stability) is greatest [Hirschmann et al., 1998]. Although Na2O is stripped from peridotite by the time the peridotite arrives at shallow depths in a non-DIA moderated melting process, Na2O will continue to affect the amount and composition of shallow melts in a DIA moderated melting column.

[40] Possibly the most important aspect to determining the impact of DIA is the spacing between melt channels and the extent of horizontal advective flow between the channels. Although advection is much more important than diffusion during chemical transport in the vertical direction, diffusion could play a role in horizontal transport of Na2O. Pressure gradients could drive advection of melt toward conduits [Stevenson, 1989; Hall and Parmentier, 2000; Spiegelman et al., 2001], in the opposite direction of Na diffusion. If so, advective transport of Na out toward the conduit will likely overwhelm Na diffusion away from conduits. However, the rheology of partially molten peridotite remains poorly understood and the actual flow regime remains highly ambiguous. If buoyancy forces are much greater than the suction forces driving horizontal advection, then the component of horizontal velocity will be small. If so, the rate of Na diffusion can be significant relative to that of advection. If the overabundance of Na in abyssal peridotites is attributable to DIA, then this observation may provide constraint on the problem of melt suction near channels.

[41] Dunite makes up 5–15% of most ophiolites and its observed distribution can provide constraint on the spacing between melt conduits [Kelemen et al., 1997]. Size and frequency of dunites within a dunite-rich area of the Ingalls ophiolite follow a power law distribution with dunites 1–10 cm in width occurring at a frequency of ∼10/m [Kelemen et al., 1999]. Although further work is needed to quantify channel spacing at different ophiolites, centimeter width dunites occur regularly within ophiolites. Even cm-scale channels would increase the impact of DIA by shortening the lengthscale of diffusion needed to affect the peridotite located between melt channels.

4.3. Evidence for the DIA Process Occurring During Upper Mantle Melting

4.3.1. Abyssal Peridotites

[42] Some of the most critical constraints on MORB generation come from studies of abyssal peridotites [Dick et al., 1984; Dick, 1989; Johnson et al., 1990; Michael and Bonnatti, 1985; Niu, 1997]. Controversy surrounds work using “reconstructed” modes to conclude that abyssal peridotites have more olivine than predicted by simple melting models [Niu, 1997; Niu et al., 1997; Walter, 1999; Baker and Beckett, 1999]. However, regardless of inclusion of reconstructed modes, point counted modes of abyssal peridotites remain inconsistent with simple batch or fractional melting models over the probable depth range for MORB. Namely, the olivine-orthopyroxene ratios imply melting at very shallow depths (<1 GPa), whereas average MORB is undersaturated in orthopyroxene requiring derivation from >1 GPa [O'Hara, 1965, 1968].

[43] Ascending melt reacting with mantle might increase the olivine-orthopyroxene ratio [Daines and Kohlstedt, 1994; Fisk, 1986; Kelemen, 1986, 1990; Kelemen et al., 1997; Aharonov et al., 1997]. Asimow [1999] showed that equilibrium reaction between migrating melts and residual peridotite could produce olivine rich residues compared to simple batch or fractional melting. The DIA process provides another explanation as predicted trends will increasingly point toward the olivine apex with increasing importance of the DIA process. On the basis of modal variations alone, no explanation is clearly favored or eliminated.

[44] The combination of modal variations and incompatible trace element concentrations in abyssal peridotite clinopyroxenes may better discriminate processes affecting abyssal peridotites. Abyssal peridotite clinopyroxenes are extremely depleted in incompatible elements (LREE, Ti and Zr), consistent with a process of incremental melt removal [Johnson et al., 1990]. However, sodium concentrations in abyssal peridotite clinopyroxenes are greater than 30× higher than those predicted from incremental melting models [Elthon, 1992] and, in some cases, even too high for batch melting. Mineral-melt reactions do not explain the observed Na2O-MgO relationship of abyssal peridotites [Asimow, 1999]. Models which can explain the Na2O-MgO relationship require refertilizing melts to be highly depleted in trace elements, yet only moderately depleted in Na2O [Elthon, 1992]. DIA is distinct from direct mineral-melt reaction because Na2O can be continuously added by diffusion while diffusion-immobile incompatible elements are not added but incrementally stripped from the peridotite. Thus DIA provides a viable explanation for both modal variations and incompatible element systematics of abyssal peridotites.

4.3.2. High SiO2 Glasses Observed in Mantle Xenoliths

[45] Silica, alumina and alkali rich glasses occur ubiquitously in peridotite xenoliths worldwide. Explanations for the origins of these exotic melts can be divided between those centered on the glasses being (1) melts solely derived from the xenolith (breakdown of amphibole or small degree melting of the xenolith [e.g., Frey and Green, 1974]) or (2) melts created by reaction between the host basalt and xenolith minerals [e.g., Neumann and Wulf-Pedersen, 1997]. The hypothesis that the DIA process creates such glasses is distinct from either of these explanations because it allows the composition of the xenolith glass to reflect both the xenolith (for slow diffusing elements like P2O5) and the host basalt (e.g., Na, K).

[46] Many xenoliths free of a K-rich phase include glasses that are too rich in K2O to be small degree melts of the xenolith [Draper and Green, 1997] making explanation 1 unlikely for some samples. Element systematics also argue against host basalt-xenolith reaction (explanation 2). Some interior glasses in xenoliths have Zr contents as low as 2.2 and 0.9 ppm [Wulff-Pedersen et al., 1999; Vannucci et al., 1998]. Such concentrations are generally a factor of 150 less than those of the alkalic host basalts for these xenoliths. The incompatibility of Zr in mantle minerals means that Zr concentration increases, not decreases, during host basalt-xenolith reaction. Glass compositions change systematically going from host basalts to exterior glasses to interior glasses with silica positively correlated with alkalis and inversely correlated with Zr (Figure 13). These relationships are opposite those expected for host-basalt-xenolith reaction but consistent with DIA because slow diffusing elements like Zr will only reflect partial melting of the xenolith.

Figure 13.

Zr and total alkali contents versus SiO2 for host basalts and glasses within xenoliths from the study of Wulf-Pedersen et al. [1999]. Alkali contents positively correlate with SiO2 while Zr contents show an inverse correlation. These element systematics are consistent with Zr solely reflecting derivation from the xenolith, while alkali contents reflect diffusion from the host basalt into the xenolith causing incongruent dissolution of orthopyroxene boosting the silica content of the glasses.

4.3.3. Development of Boundary Zones: An Explanation for High CaO Melts?

[47] The origin of high CaO magmas and melt inclusions, found in a variety of geologic settings [Schiano et al., 2000; Sigurdsson et al., 2000], remains anomalous. Melting of Ca rich lithologies such as clinopyroxenites does not appear to explain the origin of high CaO melts [Kogiso and Hirschmann, 2001]. The observation of boundary zone enrichment in CaO may indicate a process responsible for forming both high CaO melts and anorthitic plagioclase.

[48] All Cl-enhanced basanite experiments show enrichment of Cl and CaO within the peridotite immediately adjacent to the basanite. Thus Cl enrichments may provide a fingerprint to the development of boundary zones during mantle melting. Notably, some melt inclusions in both MORB and OIB have anomalously high Cl contents and high CaO contents. For OIB, melt inclusions with up to 2.5 wt% Cl from the Austral Islands have CaO concentrations >14 wt% [Lassiter et al., 2003]. Within MORB, Cl-rich melt inclusions have ∼14 wt% CaO and are found in An80-95 plagioclase, which themselves indicate derivation from higher Ca/Na melts than the observed host melts [Nielsen et al., 2000].

[49] The trace element systematics of the Cl-rich melt inclusions from MORB further support their derivation by DIA. Although Cl is highly incompatible in spinel lherzolite [Hill et al., 2000], the Cl-rich melt inclusions have chemical systematics opposite to expectation for partial melting with the most Cl-rich melt inclusions having the highest Ti/Zr [Nielsen et al., 2000]. Notably, trace element enrichment in plagioclase melt inclusions from MORB is directly proportional to z/r (charge over ionic ratio), consistent with trace elements being fractionated by a diffusion-based process [Michael et al., 2002]. Thus one possible explanation for the creation of Cl-rich melt inclusions and their host anorthitic plagioclase is that they result from the development of boundary zones lining melt channels in the uppermost mantle.

5. Conclusions

[50] Experiments show that alkali elements rapidly diffuse from silica undersaturated magmas into partially molten peridotite at 0.5–1.0 GPa pressure due to activity gradients. The addition of alkalis, particularly sodium, results in incongruent dissolution of orthopyroxene to form silica-rich melt and olivine. Sodium diffuses due to gradients in both Na2O and SiO2 concentration, producing “uphill” diffusion profiles. At low temperatures (1200°C), potassium also diffuses up gradients in its concentration. Experiments using a Cl-enriched basanite develop a boundary zone between the basalt-peridotite interface and the peak in melt pool SiO2 content. Melts in this boundary zone are enriched in CaO and Cl. One explanation for this elemental behavior is that Ca diffuses in the opposite direction of Na providing a mechanism for maintaining charge neutrality during the DIA process.

[51] Observations including abyssal peridotite modes and compositions, silica-rich glasses in xenoliths, and anomalous CaO and Cl-rich melt inclusions may indicate the presence of the DIA process in the mantle. Discerning the overall importance of the DIA process to upper mantle melting process remains a challenge ahead. A potential implication is that estimating melt production and melting rate could depend greatly on melt-rock interaction around melt conduits, adding complications to models of mantle melting.

Appendix A:: Demonstration of Mode Determination Through Digital Image Analysis

[52] In order to quantify the mineral modes of experiments, a technique for measuring mineral modes through digital analyses was developed. To determine mineral proportions in a specific area, three or more BSE images were collected in non-overlapping locations. These images were imported into NIH Image and analyzed for gray scale values (range 1–256) using the histogram function; the results of the histogram were imported to a spreadsheet and different phases quantified by totaling all pixels within a given range of gray scale value. Because of the overlap of low gray scale values, no attempt was made to distinguish clinopyroxene from spinel in any of the analyses. However, spinel makes up at most 2% of any sample. Accordingly, this simplification is insignificant. Imperfections in the polishing (black spots) were subtracted prior to the analysis of each image. The mode results of the separate BSE images were then averaged and a 1σ standard deviation calculated. Because the mode count does not depend on the spacing of a grid as in point counting, the areal extent of a phase should provide a robust estimate of the modal abundance of a phase as long as there are no preferred orientations of minerals in the samples and the size of the mineral grains is small relative to the image size.

[53] To ground truth this method, CAM-L was run sub-solidus in the piston cylinder apparatus (experiment PR-2); a representative BSE image of this run product is shown (Appendix Figure 1). Histograms for each BSE of this experiment produced three clear peaks corresponding to olivine (medium gray) orthopyroxene (dark gray) and clinopyroxene plus spinel (lightest gray). The results for the mode of CAM-L determined by digital processing agree well with the mode based on the amounts of each mineral originally weighed in (Figure A1). For experiments containing a melt phase, differentiation using the histogram function is much more difficult because of overlap of peaks. However, as Mg x-ray maps were collected simultaneously to the BSE image, the different phases were easily identified. Using the paint tool, a particular phase could be darkened and its area extent counted by converting the image to binary format.

Figure A1.

A representative 512 × 512 pixel backscattered electron (BSE) image from PR-2. Four of these images were digitally processed using the method explained in the text to quantify the modal proportions of minerals in the experiments. The average mode and a standard deviation were then calculated for the four images. The results for modal analysis of this experiment (shown in the inset) agree well with the modes based on the weighed in proportions of the minerals.

Acknowledgments

[54] I thank Mac Rutherford, Yan Liang, Joe Devine and Jim Kirkpatrick for discussions, advice and comments and E. Takahashi, M. Perfit and K. Hoernle for providing starting materials. The manuscript benefited from reviews by Paul Asimow and an anonymous reviewer. Ion probe results at WHOI were accomplished under grant EAR-9628749. Support of the experimental work and piston cylinder lab at UIUC came from EAR-0000924.

Ancillary