Clara S. Tsang Senior Integrative Exercise March 10, 2008

Transcription

1 Do secondary magma chambers exist in Mauna Loa s Southwest Rift Zone? - A Petrologic and Geochemical Analysis of Xenoliths Clara S. Tsang Senior Integrative Exercise March 10, 2008 Advisors: Cam Davidson, Carleton College Frank Trusdell, U.S.G.S. Hawaiian Volcano Observatory Submitted in partial fulfillment of the requirements for a Bachelor of Arts Degree from Carleton College, Northfield, Minnesota

2 ABSTRACT Table of Contents INTRODUCTION. 1 GEOLOGIC SETTING 2 SAMPLE COLLECTION 2 PETROGRAPHY.. 4 The Hapaimamo flow 4 Rift zone xenoliths 6 Gabbronorites 6 Troctolites 10 Dunites 10 Periodotites 10 Websterites 13 XENOLITH VOLUME ESTIMATES...13 MINERAL CHEMISTRY...15 Hapaimamo flow phenocrysts 15 Xenolith mineral composition 17 Olivine 17 Pyroxene 17 GEOTHERMOBAROMETRY.. 21 DISCUSSION.. 23 Xenolith Transport 23 Petrology and Geochemistry 26 Pyroxene equilibration geothermobarometry estimates 28 IMPLICATIONS FOR CRYSTALLIZATION ENVIRONMENT CONCLUSION ACKNOWLEDGEMENTS. 34 REFERENCES 35 APPENDIX I 37 APPENDIX II APPENDIX III

3 Do secondary magma chambers exist in Mauna Loa s Southwest Rift Zone? - A Petrologic and Geochemical Analysis of Xenoliths Clara S. Tsang Carleton College Senior Integrative Exercise March 10, 2008 Advisor: Cam Davidson, Carleton College Frank Trusdell, U.S.G.S. Hawaiian Volcano Observatory ABSTRACT The Hapaimamo flow contains xenoliths of gabbronorites, troctolites, dunites, websterites and periodotites. Composition of cumulate olivines ranges from Fo 78-86, clinopyroxenes ranges from En Wo Fs 8-12, and orthopyroxenes ranges from En Wo 2-5 Fs Most xenoliths display an intergranular, subpoikilitic or poikilitic texture. Petrologic and compositional information point towards at least two olivine-controlled magmatic sources in order to produce the observed mineral compositions one at 1437 o C with 18.5 wt% MgO and another at 1311 o C with 15 wt% MgO. Clinopyroxene thermobarometry indicates clinopyroxene equilibrating in chambers at 13 km and <3.3 km depth during periods free of physical disturbances. The majority of the Hapaimamo flow does not contain magmas that are characteristic of the summit magma chamber, thus magma pulses from deep chambers travelled via pathways other than the primary conduit and directly contributed to the southwest rift zone secondary magma chamber. The shallow chamber is inferred to be underneath the SWRZ located at ~3.3km depth and is isolated from the primary conduit to the summit magma chamber. Cycles of magma recharge, mixing and crystallization related to periods of high magma flux occurred to produce the observed range of xenolith compositions. Keywords: Mauna Loa; Xenoliths; Picrite; Petrology; Geobarometry; Magma chambers

4 1 INTRODUCTION Studies of various xenoliths from mineralogically distinctive flows can provide a window of opportunity to studying magma storage and crystallization processes that are otherwise inaccessible (e.g. Gaffney, 2002; Shamberger, 2006). We define xenoliths as crystal cumulate fragments entrained in rising magma. When a dike forces its way toward the surface, it can break off fragments of the reservoir, incorporating them into the rising magma, and bring them to the surface. Because mineral crystallization is controlled by depth and temperature, varying mineral compositions indicate different origins with different pressure-temperature conditions, hence the possibility of one large stratified vertical column or various magma reservoirs at different depths. The Hapaimamo flow from the southwest rift zone (SWRZ) of Mauna Loa Volcano is distinguished from other Mauna Loa flows by its picritic nature and the presence of abundant yet mineralogically diverse crystal cumulates. A previous study by Mueller et al. (2004) found MgO of the Hapaimamo flow ranges from wt.%, with a compositional gap between ~12 to 14 wt.%, whereas the bulk of Mauna Loa flows contain 7-8 wt.% MgO. The flow contains crystal cumulates of varying mineral assemblages, including olivine gabbronorites, gabbros, dunites, periodotites and pyroxenites. Abundant xenoliths were also found in two other picritic flows on the SWRZ (Gaffney, 2002). Mueller et al. (2004) thus proposed this unusual flow is the result of disequilibrium processes within the magma column through olivine zoning in the groundmass and xenolith crystals. The objective of this study is to gather petrographic and geochemical data from a representative suite of xenolith samples from the Hapaimamo flow in order to evaluate the SWRZ crystallization environment and processes. This study will answer questions of flow origin posed by Mueller et al. (2004) and ascertain whether the flow originated from the central summit magma chamber, or derived from a shallowly stratified reservoir, or eved from magma that bypassed the summit reservoir. In addition, it can help to determine if the crystal cumulates are derived from a single stratified vertical column associated with the summit magma chamber or

5 2 from a different region other than the summit source. The third objective is to analyze mineral assemblages, dimensions and textures and infer depths of crystallization using clinopyroxene thermobarometry. Our study uses petrography, whole-rock and xenolith mineral compositions, and geothermobarometry to establish crystallization history and magmatic sources relative to Mauna Loa lavas. Thermodynamic parameters, e.g. temperature and pressure, are computed using clinopyroxene thermobarometry in QUILF (Lindsley, 1983) and CpxBar (Nimis, 1999). Proportion of xenolith and xenolith volume are also estimated and presented here as supporting evidence for proposed environments. GEOLOGIC SETTING Mauna Loa is one of the five volcanoes that comprise the Island of Hawai`i and is the largest active volcano on the island and in the world. Its eruptions have been characterized by periodic summit, flank and rift zone eruptions (e.g. Lockwood and Lipman, 1987; Rhodes, 1995). Mauna Loa has two prominent rift zones trending southwest and northeast. The Hapaimamo flow is located on the southwest rift zone (SWRZ) (Fig. 1), is one of the largest flows on the SWRZ, and is dated at 240 years before present or A.D (Lockwood, 1995). It originated from spatter ramparts at an elevation of 1880 m on the SWRZ and covers an estimated area of 100 km 2 (Trusdell, unpublished data, 2004). Three different flows have been identified with the Hapaimamo flow: 1) Kaikia`ana flow at upper elevation by the Hapaimamo cone, 2) `Akau flow at lower elevation along the north edge, and 3) Hema flow at lower elevation along the south edge. Samples for this study were collected randomly from a 100m x 200m strip along the Hema flow field near the coast. SAMPLE COLLECTION Two sets of samples were collected for two different objectives. The first objective is to

6 3 (a) Spatter cone Legend Hapaimamo flow Roads, highway 11, settlement areas o o o (b) 19 o Figure 1. Map showing location of study area on the SWRZ of Mauna Loa. Light green shaded area on quad map denotes the Hapaimamo flow field and blue lines denote roads and highways. Flow originated from spatter cone at elevation of 1,880 m on the SWRZ in A.D (Lockwood, 1995). Insert shows the Island of Hawai`i, location of Mauna Loa and SWRZ, and the Hapaimamo flow.

7 4 collect quality xenoliths for geochemical analysis and pyroxene thermobarometry. Samples with xenoliths (Fig. 2) were classified, measured and described before being cut and prepared into microprobe sections. Under the assumption that this region generated a representative suite of crystal cumulate samples, random sampling is representative of the source. The other purpose was to conduct a population census or calculate the proportion of xenoliths in the host flow. Lava flow blocks greater than 50 cm containing xenoliths were systematically sampled. Six blocks were randomly collected in the area and were treated as representative of the whole flow. Block and xenolith dimensions, types, mineralogy, and textures, were noted for each block. Locations were marked with a GPS and photographed. In subsequent sections, periodotite and websterite xenoliths are denoted with w, coarsegrained gabbronorite with cg, medium-grained gabbronorite with mg, fine-grained gabbronorite with fg, dunite with d, troctolites with tr, and host rock samples with R. PETROGRAPHY Mineral assemblages and modes of crystal cumulates were optically estimated and classified according to the IUGS classification system (Streckeisen, 1979). Textures were noted as open or closed with corresponding vesicle density and as having an intergranular, poikilitic, subpoikilitic or subophitic texture. Other noteworthy features like rimming of xenocrysts and microstructures were also recorded. Dimensions of blocks and xenoliths were measured using a standard measuring tape with an accuracy of ±0.1cm. The Hapaimamo flow The Hapaimamo flow has an estimated 20-30% olivine, typically appearing as euhedral rhombic phenocrysts that range in size from 3-7 mm in diameter. There are 2-3% pyroxene phenocrysts throughout the flow as euhedral rhombic phenocrysts averaging 1-2mm in size, and less than 1% plagioclase. Phenocrysts are generally euhedral and fractured. A few olivine megacrysts are also present, ranging from 5 to 15 mm. Modal olivine content differs between

8 5 L cm Gabbroic xenolith Gabbroic xenolith Troctolite xenolith Figure 2. Photo illustrating a xenolith-rich rock with labeled xenolith types. All xenoliths are >1.5cm and are used for geochemical study.

9 6 xenolith-rich and xenolith-deficient rocks. Xenolith-rich rocks contain on average 22-23% modal olivine and <1% modal plagioclase with 10-23% modal density of vesicles. On the other hand, xenolith-deficient rock yields 31 % modal olivine with 12% modal density of vesicles. The groundmass is medium gray in hand sample and is typically medium to highly feldspathic with plagioclase microlaths present. Vesicles range from rounded to subangular, averaging 3 mm and make up approximately 40% of the rock. This flow is uncommonly xenolith-rich compared to other Mauna Loa flows. Rift Zone Xenoliths Our xenolith population census yielded approximately 80% gabbronorites, 10% troctolites, 5% dunites, 5% periodotites, and <1% websterites (Fig. 3). Gabbronorites were further classified by grain size. Medium-grained gabbronorites (grain width from 0.5 mm to 1 mm) are at ~42% and are the most common, while fine-grained gabbronorites (grain width < 0.5 mm) are at ~38%, and coarse-grained gabbronorites (grain width > 1mm) are at ~20%. Xenoliths range in size from 0.5 cm up to 25 cm in the longest dimension, with most in the 1-2 cm range. The size of xenoliths generally decreases as they become more mafic (Fig. 4). Mafic xenoliths are typically less than 1 cm in their longest axis and free of vesicles, while gabbronorite xenoliths usually exceed 1.5 cm in their longest axis and ~30% are open-textured and holocrystalline. Within the main body of the flow, all xenoliths but gabbronorites are distributed randomly. Gabbronorites tend to congregate at the lower flow boundary. Gabbronorites Gabbronorites ( 10% orthopyroxene, 10% plagioclase) are comprised of plagioclase, olivine, orthopyroxene, clinopyroxene, and minor interstitial glass. Most have ~50-55% plagioclase and ~30-40% olivine. Out of 147 gabbronorite samples, 17 contain 10% olivine. The amount of pyroxene also varies among gabbronorites, and a decrease in olivine content is found to be accompanied by an increase in plagioclase and pyroxene (Fig. 5). There is a strong relationship between olivine and pyroxene, indicating olivine control in operation in the parental

10 7 (a) (b) Websterite (10) 1.0% medium-grained (69) 42% Periodotite (66) 12.1% Trocolite (39) 8.8% Gabbronorite (147) 73.3% Dunite (34) 4.9% coarse-grained (32) 20% fine-grained (46) 38% ( ) = number of xenoliths for each type Figure 3. (a) Graph showing distribution of xenoliths by volume percent. (b) Graph showing distribution of gabbronorites subdivided by textures. Number in parenthesis behind xenolith type denotes the actual number of xenoliths for each type Average longest axis length (cm) CG MG FG troctolite dunite periodotite websterite Figure 4. Bar graph showing distribution of xenolith sizes sorted by mineralogy. More mafic xenoliths are observed to have shorter axis lengths, i.e. mafic xenoliths tend to be smaller in size. CG = coarsegrained gabbronorites, MG = medium-grained gabbronorites, and FG = fine-grained gabbronorites (refer to text for grain size definitions).

11 8 Volume percent of minerals Legend plagioclase pyroxene Volume percent of olivine Figure 5. Graph showing mineral relationships in gabbronorite xenoliths. As olivine content increases in gabbronorites, volume percent of plagioclase and pyroxene decreasein a linear trend. So olivine control is evident in gabbronorites.

12 9 melt. Pyroxene content in gabbronorites decreases to 10% when olivine reaches 40% and eventually decreases to ~0% when olivine is 50%.Gabbronorites display either a holocrystalline, subophitic or subpoikilitic texture. Non-gabbroic xenoliths have ~25% vesicles and comprises ~7-10 % of the host rock. Gabbronorites are also the largest xenoliths found within the study area, averaging ~3.06 cm 2. In coarse-grained gabbronorites (Fig. 6), plagioclase is most abundant and averages 55%. Plagioclase crystals are often euhedral, lathlike and generally 1-2 mm in size. They have intergranular and sometimes subpoikilitic textures with ~20% olivine and ~25% pyroxene. Olivine is absent in a few xenoliths and its volume percentage varies among samples. It occurs as subrounded, blocky grains and blades. Olivines are ~5-7 mm at the largest, averaging 3 mm. Pyroxene, when present, is often surrounded by rhombic olivine crystals and appears as stubby blocky grains, generally ~1 mm. Two coarse-grained gabbronorites have plagioclase laths surrounding a core of pyroxene and olivine crystals respectively. Within all xenolith sizes, medium-grained gabbronorites are most abundant and largest, averaging 3.18 cm 2 and ranging up to cm 2 in area. In medium-grained gabbronorites (Fig. 7), plagioclase averages 55%, is generally 0.5 mm in size, and appears as laths. Olivine typically occurs in a subpoikilitic or intergranular fashion, as subrounded to rounded rhombs, and ranges from 30-40%. Pyroxenes in medium-grained gabbros also vary commonly from 10% to 20% on average. They are 1 mm in size and are blocky and stubby. Most pyroxenes appear to be partially resorbed or surrounded by olivine under the microscope. Within fine-grained gabbronorites (Fig. 8), plagioclase averages 50% and generally appears as microlaths when viewed under hand lens. Laths sometimes poikilitically enclose larger pyroxene or olivine grains, forming a core of mafic mineral within a plagioclase shell. Olivine occurs as subrounded minute grains and ranges consistently from 40-60% among the 29 samples, averaging ~45%. Pyroxene is at 5% for all samples and is less abundant in fine-grained gabbros

13 10 compared to other grain-sizes. It also occurs as minute blocks and is poikilitic or intergranular with plagioclase and olivine. Troctolites Troctolites (Fig. 9) are comprised of 40% plagioclase and 30% olivine, with less than 10% pyroxene. Out of the 39 samples collected, only five have ~5% pyroxene. Samples contain on average 55% plagioclase and 40% olivine. Plagioclase laths are often intergranular with olivine and, to a lesser extent, subpoikilitic. Olivine appears as euhedral grains and is distributed evenly throughout. Some troctolites are open-textured and are comparable to some of the gabbronorites, containing 10% interstitial host rock within the xenolith. Troctolites are found often in fine and medium grain sizes (grain sizes defined in above gabbronorite subsection). Dunites Our 34 dunite samples have on average 96% olivine and 4% pyroxene. Dunites are mostly fine-grained (~0.5 mm) and have a poikilitic texture (Fig. 10). Some are olivine oikocrysts, where smaller subrounded olivine grains were engulfed in a large olivine (Fig. 10b). Olivine grains and rare pyroxene grains are ~1mm in size and appear as subhedral minute rhombs. Plagioclase is often anhedral and optically appears to be surrounding olivine or pyroxene grains. One specimen in L has 10% plagioclase interacting poikilitically with olivine. One specimen in L and two in L have a rimming of plagioclase crystals between the xenolith and the host rock. Periodotites Our 38 periodotite samples (Fig.11) contain on average 55% olivine and 44% pyroxene, with three samples containing 5-10% plagioclase. Olivine and pyroxene grains are blocky and short, generally ~1-2 mm, and occur intergranularly in all samples. Periodotite dimensions average 0.6 cm and 0.9 cm and areas average 0.49 cm 2. No open textures were found in any samples.

14 11 6a. L cg 1 cm 6b. L cg cpx plag plag 7a. L mg 1 cm 7b. L mg opx opx plag opx plag cpx opx cpx plag opx plag 8a. L fg 8b. L fg cpx plag plag 1 cm plag Figure 6-8. (a) Photos and (b) micro-photographs of gabbronorites of different grain sizes; CG = coarse-grained gabbronorites, MG = medium-grained gabbronorites, and FG = fine-grained gabbronorites. Scale in photographs is in cm and in mm for micro-photographs.

15 12 9a. L cm 9b. L plag plag 10a. L b. L cm 11a. L b. L cm opx Figure (a) Photos and (b) microphotographs of troctolite, dunite, and periodotite. Trocolite (fig. 9) has plagioclase surrounding olivines subpoikilitically. Dunite (fig. 10b) is an olivine oikocryst surrounding rounded olivines and periodotite is an orthopyroxene oikocryst surrounding rounded olivines.

16 13 Websterites Websterites are defined as 90% orthopyroxene or clinopyroxene. Our samples are loosely defined as such when pyroxene is 70% and olivine is 30%. Most websterites found are true websterites, with 90% pyroxene content. One sample found in L has 100% pyroxene and three samples found in L have 90% pyroxene and 10% plagioclase. Only three out of the sixty six samples are olivine-bearing websterites that contain 70-90% pyroxene and 10-30% olivine. Pyroxene is found as euhedral to subhedral rhombs and is sometimes clustered together. Olivine grains are ~1mm in size and also appear as rhombs. Plagioclase is found as laths and averages 1mm in length. All websterites have a granular texture, with pyroxene occurring intergranularly with plagioclase. Websterites are the smallest out of all xenolith types, averaging 0.16 cm 2 with ~ 0.5 cm for both axes. XENOLITH VOLUME ESTIMATES Lava flow blocks greater than 50 cm on a side were used to determine the volume percent xenoliths in the Hapaimamo flow. For each side, the area of the side and xenoliths are computed using area equations for rectangles and ellipses respectively due to their different shapes found in the field. Using the resulting areas, proportion of xenoliths to host rock within one side is calculated using (area of xenolith)/(area of block) x 100%. When area percent of xenolith is determined for all available sides of a block, the average xenolith proportion is calculated for the whole block. By projecting the calculated average xenolith ratio onto the volume of the block, the average volume of xenoliths within a block can be evaluated. Table 1 presents the average xenolith percentages and subsequent estimated xenolith volumes calculated for each block. The average xenolith proportion within the flow is 1.21 volume percent. Out of the six blocks examined, there appears to be two distinct xenolith suites. L , -235 and -237 have a consistent average of 1.5 vol.% of xenoliths, while L and

17 14 T AB L E 1. SUMMARY T AB L E OF CAL CUL AT ED X ENOL IT H AREA% AND V OL UME Sample L L L L L L Area % xenolith Side A Side B Side C Side D Side E 0.26 n/a n/a 0.07 Side F n/a n/a 1.36 n/a n/a n/a Overall xenolith % Total rock volume (cm ) 33,888 92,453 28, ,989 42, ,453 Total xenolith volume (cm 3) Note: Area % xenolith is calculated using Areaxeno/Arearock x100 and overall xenolith % is the average of area % xenolith of all sides. Total rock volume is calculated by length x width x height of block. Total xenolith volume is computed using total rock volume x overall xenolith volume %. Measured areas of xenolith and blocks are in Appendix Percentage Periodotite Legend Average xenolith % Dunite Periodotite Websterite Dunite Websterite 0 Average xenolith % L L L L L L Figure 12. Graph showing correlation between percentage of mafic xenoliths out of a rock s xenolith population and average percentage of xenoliths present in a rock. The more mafic xenoliths there are within a rock s xenolith population, the less total xenoliths there are in the rock. This is interpreted as segregation according to density of xenolith and the ability to be entrained as the flow propagates. Mafic xenoliths are deposited earlier in the flow; most xenoliths are also able to be entrained further down the SWRZ, thus those with more mafic xenoliths are also found to have a smaller total xenolith population.

18 have 0.15 vol.% of xenoliths. Volume percent of xenolith is found to decrease with the increasing presence of mafic xenoliths (Fig. 12). L has 0.17 vol.% xenolith and 1.53% of its xenoliths are websterites, while L has only 0.15 vol.% xenolith and 48% of its xenoliths are periodotites and 3.79% are websterites. MINERAL CHEMISTRY Host rock olivine and xenolith mineral compositions were measured at the University of Hawai`i using a five-spectrometer, Cameca SX-50 electron microprobe. An accelerating voltage of 20kV was maintained and peak counting times of 30s was used for all elementts. For olivine analysis, a beam current of 20nA was maintained. For pyroxene analysis, a lower beam current of 15nA was used due to susceptible loss of sodium (Na). During analysis, Na spectra were counted first. At least three points were analyzed at mineral cores. One point was analyzed at olivine rims for host rock samples. The reported mineral analyses are an average of the three spot analyses. Calibrations were performed on natural and synthetic minerals. Relative analytical error, based on repeated analysis of the San Carlos and Springwater olivines, Verma garnet, the Smithsonian diopside and chromite standard, natural-occurring jadeite, and sphene glass, are <1% for major elements. Hapaimamo flow phenocrysts Olivine phenocrysts were analyzed to determine their composition (Table 2) and zoning patterns. Phenocrysts are euhedral and undeformed. Core compositions range from forsterite (Fo) 78% to 90% and rim compositions from forsterite 70% to 86% (Fig. 13). Over 60% of phenocryst cores have Fo and corresponding rims have Fo This is consistent with the findings reported by Wilkinson and Hensel (1988), that historically erupted Mauna Loa picrites have olivines with Fo Thirty two olivine phenocrysts displayed normal zoning patterns and three show no zoning. Two phenocrysts are reversely-zoned and show 1% increase in forsterite on the rim respectively, but there is no apparent relationship between core forsterite content and zoning

19 16 TABLE 2. GEOCHEMICAL COMPOSITION OF OLIVINE Sample 230R 230R 207w 202cg 202d 202tr Type wr core wr rim xeno core xeno core xeno core xeno core SiO FeO NiO MgO CaO Total Number of cations based on 3 cations Si Fe Mg Ca Total %Fo %Fa %La Note: Cation calculation is generated using QUILF (Lindsley, 1983) Parental magma Forsterite content % Summit-derived Rift-stored tr 202d 207w 202cg WR phenocrysts Sample number Figure 13. Plot of olivine forsterite content of samples. Tr = trocolite, d = dunite, w= websterite, cg = corase-grained gabbronorite, and WR = whole rock.

20 17 pattern of the mineral. Change in forsterite content along crystal profile varies from 3-10%. CaO content of olivine phenocrysts ranges from wt.% and increases with MgO weight percent and forsterite content of the phenocrysts. The low concentration of CaO in olivine is in agreement with crystallization at lower crustal pressure as suggested by Stormer (1973). Xenolith Composition Olivine Representative analyses of cumulate olivine grains are summarized in Table 2. Olivines are mostly subhedral and rounded, with a few kink-banded olivines showing a small degree of deformation and some with resorption edges. Forsterite content in troctolite ranges from 78-79%, in dunite from 80-83%, in periodotite from 83-85%, and in coarse-grained gabbronorite is at 86%. Olivine Fo-Ni variability corresponds with the range of compositions represented by each xenolith types (Fig. 14). Coarse-grained gabbronorite and periodotite are formed from a more primitive Ni-rich melt than troctolites. For the olivine oikocryst (L d, Fig. 10b), Forsterite content varies ~5% but NiO weight percent remains within ± 0.05%. CaO content of all olivines is moderate and ranges from 0.22 to 0.27 wt.%. This is consistent with shallow crystallization and equilibration (Stormer, 1973). Pyroxene Representative analysis of cumulate pyroxenes is summarized in Table 3. Clinopyroxenes in troctolites (En 51 Wo Fs 10 ), periodotites (En Wo Fs 8-9 ), and gabbronorites (En Wo Fs ) span similar compositional ranges (Fig. 15). All clinopyroxenes are augites and Mg-number (100*Mg/(Mg+Fe)) ranges from 66 to 87. L cg is the only analyzed sample that contains augite of a less calcic composition (En Wo 26 Fs ) and is in equilibrium with the co-existing orthopyroxene in the stability field of 1100 o C to 1200 o C at 1 kbar. Orthopyroxenes are slightly more Fe-rich and Ca-rich in gabbronorites than periodotites. Within gabbronorites,

21 Rift-stored Summit-derived Parental magma pyx + plag NiO (wt%) pyx plag Legend 202tr 202d 202cg 207w %Fo Figure 14. NiO vs. %Fo variability diagram for xenolith olivines. Olivine, pyroxene, and plagioclase are abbreviated as, pyx, and plag respectively. Note olivines of 202cg are derived from the most primitive magma out of all other samples but have pyroxene and plagioclase present, leading us to believe that olivines in 202cg are xenocrysts. In all other samples, crystallization sequence is as described in text.

22 19 T AB L E 3. GEOCHEMICAL COMPOSIT ION OF PY ROX ENE Sample 239cg.1* 239cg.2* 202cg-1.1* 202cg-1.2* 202cg-2* 202mg.1* 202mg.2* 202tr 230w 207w 232w T ype cpx opx cpx cpx opx cpx opx cpx cpx opx opx SiO Al 2 O T io FeO MnO MgO Cr 2 O CaO Na 2 O T otal Number of cations based on 4 cations based on QUILF Si Al T i Fe Fe Mn Mg Cr Ca Na T otal %E n %Fs %W o Mg# *Data used for thermometry using two-pyroxene equilibria Note: Orthopyroxene and clinopyroxene are abbreviated as opx and cpx. Cation and end-member calculation are done through QUILF (Lindsley, 1983). Mg# = 100 x (atomic MgO/FeO+MgO). Two clinopyroxene compositions are listed for 202cg; one spans similar compositional range as other samples and the other is less calcic than all samples (refer to fig. 15). Both compositions are representative of clinopyroxenes of 202cg.

23 mole% 45 Diopside 50 Hedenbergite Color code 230w 202tr 202mg 239cg 202cg 232w 202w Mg 2 Si 2 O 6 (En) 50 Fe 2 Si 2 O 6 (Fs) Figure 15. Quadrilateral plot of orthopyroxene and clinopyroxene compositions with geotherms at 1 kbar (Lindsley, 1983). Orthopyroxenes and clinopyroxenes of all samples span similar compositional ranges. Co-existing orthopyroxenes and clinopyroxenes are in disequilibrium for all samples. 20

24 21 orthopyroxenes are more ferrous in coarse-grained samples compared to other grain sizes. All orthopyroxenes are enstatites. Enstatites in coarse-grained gabbronorites are En Wo 4 Fs 21-25, medium-grained gabbronorites are En Wo 4-5 Fs 17, and periodotites are En Wo 2-4 Fs Mgnumber ranges from 74 to 87. Cr 2 O 3 variability diagram shows the trend of pyroxene crystallization and melt evolution (Fig. 16). Pyroxenes crystallized from a more primitive melt would have a high Mg# (>87) and high Cr 2 O 3. Clinopyroxenes and orthopyroxenes in periodotites and troctolites are derived from a more primitive melt than gabbronorites, contrary to the trend displayed by NiO variability of olivines. Clinopyroxene is often found surrounding euhedral plagioclase laths in its subhedral form. This provides evidence for clinopyroxene as a later crystallization phase than plagioclase. Clinopyroxene can also be found rimming olivine and orthopyroxene while being in optical continuity with orthopyroxene. This clinopyroxene rim might be a reaction from olivine and orthopyroxene re-equilibrating as the melt eves and cools (olivine + orthopyroxene + liquid clinopyroxene). GEOTHERMOBAROMETRY Representative geochemical data consisting of both clinopyroxene and orthopyroxene were considered and analyzed in QUILF (Lindsley, 1983) using the two-pyroxene equilibria. Temperatures of equilibration with an error of ± 10 o C can be estimated using Ca-Fe-Mg partitioning between coexisting phases (Lindsley, 1983). For all pyroxene compositions, QUILF failed to converge for both orthopyroxene and clinopyroxene. Precise formation temperature is therefore impossible to calculate through QUILF. However, minimum formation temperature can be estimated using different blocking temperatures of pyroxenes. Previous research showed that clinopyroxenes stop equilibrating at a higher blocking temperature than orthopyroxenes (e.g. Drueppel et al., 2001). Since T cpx > T opx,

25 22 Cr2O3 (a) (wt%) Legend 202cg 202cg_rim 202tr 230w 230w_rim 202mg 239cg 202mg Periodotite+Troctolite decrease in Mg# towards rim Gabbros Mg# (b) Periodotites Cr2O3 (wt%) Gabbros Mg# Legend 202cg 232w 207w Figure 16. (a) Mg# vs. Cr2O3 variability plot for clinopyroxenes. (b) Mg# vs. Cr2O3 variability plot for orthpyroxene. Rim data for minerals are plotted in the same color but different symbols. Periodotites and troctolites are derived from a more primitive magma (higher Mg# and Cr2O3) than gabbros. Mg# = 100 x (atomic MgO/FeO+MgO).

26 23 clinopyroxene compositions are more likely to preserve their equilibrium compositions. Thus, clinopyroxene compositions were held fixed while orthopyroxene compositions were allowed to vary in QUILF to allow convergence and to determine a minimum formation temperature (T cpx ) for the xenolith. In addition, clinopyroxene compositions were also allowed to vary in order to calculate the cooling temperature of orthopyroxene (T opx ). Temperature calculated using fixed clinopyroxene compositions from QUILF (T cpx ) were used in CpxBar (Nimis, 1999) with corresponding clinopyroxene composition to calculate crystallization pressure. Through iterating between values calculated in CpxBar and QUILF, a crystallization pressure and temperature for each sample was determined. Only pressures calibrated for tholeiitic melts in CpxBar are reported. In addition to estimated temperatures generated by QUILF, temperatures computed using glass thermometry (Montierth, 1995) from glass probe data from Mueller (2004) were also used to constrain pressure estimates in CpxBar. Calculated T cpx ranges from 1227 o C to 1100 o C in QUILF for pyroxene pairs (Table 4). Whenever calculated T cpx is lower than the minimum glass temperature calculated using glassbased geothermometer by Montierth et al. (1995) or whenever orthopyroxene is not present, minimum glass temperature of 1156 o C and 1162 o C were used to substitute for T cpx to give more realistic pressure estimates. Pressures calculated in CpxBar range from unrealistic negative pressures to 3.9 kbar (Table 4). DISCUSSION Xenolith transportation Crystal cumulates are found more abundant at the bottom of the flow compared to at the top of the flow. This is probably due to high effusive rates for the Hapaimamo flow calculated by Mueller et al. (2004). Because of higher flow rates, there is more energy to entrain xenoliths, which are denser than basaltic liquid, and prevent them from settling. Thus more xenoliths get transported further downslope. As topography becomes gentler downslope and closer to the sea,

27 24 TABLE 4. PRESSURE AND TEMPERATURE ESTIMATES Sample Glass QUILF CpxBar T ( C) Tcpx ( C) Topx ( C) P (kbar) 202mg* cg* cg* cg cg w** tr** Note: * Temperature calculated by QUILF was unrealistic and lower than glass temperature. Glass temperature was used instead of calculated QUILF temperature for estimating pressure in CpxBar. ** No orthopyroxene was present in these samples. Glass temperature from whole rock was used for CpxBar calculation to roughly estimate pressure conditions. The absolute values calculated are not used to infer crystallization enviroments.

28 25 flow velocity decreases and it becomes more difficult to suspend xenoliths. Crystal cumulates eventually sink to the bottom of the flow, resulting in a high concentration of xenoliths at the lower flow boundary. There is especially a tendency for gabbronorite cumulates to congregate at the bottom flow boundary, possibly due to their large size overcoming compositional density difference. Textures of xenoliths and their mineralogical relationships can provide clues to crystallization environment and magma chamber dynamics. The lack of vesicles and groundmass in mafic xenoliths indicate cumulus crystallization in an environment that does not experience frequent physical disruption and chemical resetting (Rhodes, 1988), whereas gabbronorites with intergranular and subpoikilitic textures represent pods of crystal mush in the magma chamber prior to eruption. Also, pyroxene content decreases as olivine content increases in gabbronorite xenoliths, which demonstrates olivine control where chemical variation of the melt can be explained by the addition or subtraction of olivine (Tilling et al., 1987) and therefore magma crystallizing gabbronorites must have >6.8 wt.% MgO. Furthermore, the presence of olivine and orthopyroxene oikocrysts is indicative of rapid undercooling. Crystal mushes from lower depths were probably captured during periods of high magma flux and transported to a shallower environment, resulting in rapid cooling of liquid in the crystal mush and formed oikocrysts. In addition, olivine typically appears to be surrounded or partially resorbed by pyroxene, which indicates mineral re-equilibration from prolonged period(s) of cooling. Also, plagioclase is found rimming mafic cumulates, which imply plagioclase is a late crystallization phase. Furthermore, pyroxene is less abundant in fine-grained gabbros than coarse- and medium-grained gabbros. This can be interpreted as the liquid starting to eve beyond pyroxene crystallization, resulting in fewer pyroxene nuclei being formed and thus less fine-grained pyroxenes present in the melt.

29 26 Petrology and Geochemistry High MgO content of host rock and relatively high abundance of olivine phenocrysts require crystallization in a high temperature environment. Previous studies on SWRZ submarine picritic lava found Fo 90 olivines, which formed from magmas with 17.5 wt.% MgO at 1415 o C (e.g. Garcia et al, 1995). In addition, temperature of the Hawaiian plume has been estimated to be >1550 o C (e.g. Watson and McKenzie, 1991). With this temperature range, it would be possible to generate magma of wt.% MgO. Most xenoliths follow the crystallization sequence of olivine ± plagioclase olivine + pyroxene ± plagioclase olivine + pyroxene + plagioclase as the melt eves. However, L cg does not appear to follow this crystallization sequence. In the NiO-Fo variability diagram (Fig. 14), olivine (Fo ) from this sample appears to have crystallized from a primitive melt before olivine in the websterites, periodotites, and troctolites. However, pyroxene from this sample apparently crystallized from a more eved melt after pyroxene in the websterites, periodotites, and troctolites in the Cr 2 O 3 -Mg# variability diagram (Fig. 16). We interpret this as olivines were crystallized at a lower depth and at a higher temperature before being entrained and transported to a shallow chamber. These high forsterite olivine xenocrysts were then mixed with pyroxenes of a lower Mg# crystallized from the more eved melt. In other words, olivines in L cg are xenocrysts and L cg has to crystallize from two different melt compositions. In addition, it is important to place constraints on the melt composition from which olivine crystallized. It is essential to assess whether olivine is in equilibrium with melts of a parental magma or whether they crystallized from a melt more comparable to summit reservoir magma in order to gain insight into its crystallization environment. Using the Fe-Mg partitioning relationship between olivine and melt where K D = 0.3 ± 0.03 (Roeder and Emslie, 1970), olivines in equilibrium with melts corresponding to whole-rock composition will plot in the equilibrium field (Fig. 17). Mg#49 from glass data was supplied by Mueller et al. (2004) and used in place for corresponding Mg-number of the host rock. We find olivine phenocrysts from host rock, dunite,

30 27 gabbro, periodotite, and troctolite all plot outside of the equilibrium field, where their forsterite contents are too high to be in equilibrium with their whole rock Mg-number. Therefore they are probably xenocrysts. Olivines with compositional continuity are interpreted to represent magma evolution and olivines with compositional gaps between them to be derived from different magmatic sources (Fig. 17). The olivine control line represents magmatic composition with 6.8 wt.% MgO and separates magmas that operate under olivine control from those that operate under fractionation (Tilling et al., 1987). At least two separate magmatic sources can be distinguished by the compositional gap and their placement relative to the olivine control line. Olivines that crystallized above the olivine control line and in compositional continuity are from host rock, coarse-grained gabbro, websterite, and dunite. They were crystallized from a melt of Mg# 72-67, 69-64, 65-60, and respectively. Olivines that are in compositional continuity are interpreted to be crystallizing from an eving melt. However, sources responsible for olivines in host rock and xenoliths often overlap. For example, the highest-forsterite olivine from L w crystallized from a melt with Mg# (source II), but overlaps with a source of Mg# 63 (source III) crystallizing medium-forsterite olivines in L d. Thus olivines that have compositional gaps between them can be inferred as crystallizing from different pulses of magma or fractionating from the melt, where olivine is the only mineral being fractionated. Forsterite content of host rock olivines indicate they were derived from a parental magma with initial ~14-15 wt.% MgO (using a ferrous iron ratio of 0.1 from Rhodes et al., 2005). Olivines below the olivine control line are crystallized from differentiated magma with less than 6 wt.% MgO, where olivines, plagioclase, and pyroxenes were being fractionated. Olivines that plot within the equilibrium field of K D = 0.3 ± 0.03 are derived from a more eved differentiated melt. Therefore, more than one magma source is needed to produce the overall observed range of forsterite content in olivine. Clinopyroxenes are plotted on a similar diagram to assess their origins (Fig. 18). The

31 28 experimental Fe-Mg partitioning relationship between clinopyroxene and melt at low pressure can be expressed as K D =0.23 (Grove and Bryan, 1983). FeO/MgO of glass is supplied by Mueller et al. (2004) and is equal to All clinopyroxenes in cumulates plot below the equilibrium line. They have lower FeO/MgO than associated clinopyroxene composition and are too magnesiumrich to be in equilibrium with their whole rock FeO/MgO ratio. They were thus crystallized from a more primitive liquid than the host rock. Clinopyroxenes in mafic cumulates crystallized from a more primitive melt with FeO/MgO of >0.65, which indicates a melt of 18.5 wt.% MgO, while clinopyroxenes in gabbroic cumulates crystallized from a more eved melt with FeO/MgO up to 1.35, indicating a melt of 8.6 wt.%. MgO characteristic of a summit composition. Contrary to olivine compositional plot, no compositional gap is observed here. Clinopyroxenes appear to be crystallizing from either one eving melt or several pulse of magma that are compositionally similar and did not drastically change the eving liquid composition enough to produce compositional gaps. The higher MgO weight percent magmatic compositions are similar to parental magma composition, which has wt.% MgO (Rhodes, 1995), and the magmatic source of lower MgO weight percent is characteristic of differentiated or reservoir lavas of Mauna Loa (Tilling et al., 1987). According to a glass-based geothermometer for Mauna Loa lavas (Montierth, et al., 1995), magmas with MgO content of 18.5 wt.%, 15 wt.%, 8.6 wt.% and 6 wt.% should have temperatures of approximately 1437 o C, 1311 o C, 1209 o C and 1154 o C respectively. Pyroxene equilibration geothermobarometry estimates Temperatures failed to converge in QUILF using the two-pyroxene equilibria with orthopyroxene and clinopyroxene chemical data. This is interpreted as a lack of chemical equilibrium in our samples from prolonged periods of cooling (e.g. Shamberger, et al., 2006), where ion exchange between minerals was allowed to continue. Two distinct sets of calculated temperatures and pressures emerged from our clinopyroxene thermobarometry estimates. One set yields unrealistic negative to <1 kbar pressure

32 FeO/MgO Cpx Forsterite (%) eving melt source I source II source III source IV Kd=0.30 (+/- 0.03) Legend Olivine control 202tr 202d 202cg 207w WR 80 source V Fractionation 85 Forsterite (%) Whole Rock Mg# Figure 17. Forsterite content of olivine cores vs. whole rock Mg# for olivines from xenoliths and host rock. 80 The equilibrium field is from Roeder and Emslie (1970) for lower pressure crystallization. Olivine control operates in magmas of >6.8 wt% MgO and fractionation occurs in magmas of <6.8 wt% MgO (Rhodes, 1995). Olivines in compositional continuity are interpreted to derive from the same magmatic source. Mg# = 100*MgO/(MgO+FeO) Legend Equilibrium field microlites 202cg source I cg_rim mg tr source II Whole Rock Mg# 207w w source III 230w_rim 239cg source IV source V KD= tr 202d 202cg 207w source VI xenocrysts FeO/MgO glass Figure 18. FeO/MgO of cumulate clinopyroxenes vs. FeO/MgO of glass (Mueller et al., 2004). The KD = 0.23 equilibrium line is from Grove and Bryan (1983) for low pressure crystallization. Clinopyroxenes that plot below the KD line are more MgO-rich and are interpreted as xenocrysts. Clinopyroxene is denoted as cpx.

33 30 at 1100 o C, which corresponds to crystallization or equilibration at <3.3 km. When taking into account the 2 kbar error bar of CpxBar calculations, the unrealistic negative pressures are interpreted as crystallization or re-equilibrium at extremely shallow conditions. The other set contains minimum T cpx ranging from 1233 o C to 1227 o C with pressure at 3.9 to 3.7 kbar, which corresponds to a depth of ~13km. Only coarse-grained gabbronorite (L cg) contains both sets of conditions; all other analyzed samples display only extremely shallow condition. Calculated temperatures and pressures of clinopyroxenes are not correlated with their locations within the coarse-grained gabbronorite xenolith. At 1 kbar, all clinopyroxenes lie within geotherms of o C while orthopyroxenes are within o C (Fig. 15). This appears to contradict our earlier assumption that T cpx > T opx. However this is assuming both pyroxenes crystallized at 1 kbar. Also, the steep gradient of geotherms for the orthopyroxene field indicates a slight change in pressure can lead to drastic changes in temperature. Thus figure 15 only serves to illustrate the disequilibrium nature of crystal cumulates. In addition, since pyroxene pairs are in disequilibrium, our estimations are at best minimum cooling temperatures recorded during magma ascent and storage. Our temperatures are in agreement with results of a previous study on a nearby picritic xenolith-rich flow (Gaffney, 2004), though our compositions indicated lower pressure environments. Clinopyroxene equilibration temperatures are also predictably lower than temperatures of parental magmas estimated using clinopyroxene Fe-Mg distribution coefficient, but are consistent with estimated temperatures of lower MgO magmas. However, pyroxene barometers and thermometers can only provide a rough estimate of equilibrating conditions because of relatively large error bars in QUILF and CpxBar and because there is evidence for chemical disequilibrium. Thus temperature and pressure calculated should be treated as approximate. Nevertheless, it provides a good indicator of the range of environments present. For this study, it successfully shows that separate sets of conditions existed and presents evidence of more than one crystallization environment.

34 31 IMPLICATIONS FOR CRYSTALLIZATION ENVIRONMENT Re-equilibration textures (e.g. resorbed grain boundaries), reversely-zoned olivines, olivine and pyroxene compositional variation, pyroxene compositional disequilibrium, and the two distinct sets of estimated thermodynamic conditions suggest magma mixing between a parental magma composition and one with lower MgO content such as a differentiated lava. There is evidence that more than one pulse of high-pressure magma rising and mixing with differentiated lava as well, because of the wide range of MgO weight percent magma estimated from olivine and clinopyroxene K D diagrams. Rhodes (1995) points out lavas erupted from the summit magma chamber typically have wt.% MgO because the summit magma chamber is large and tends to homogenize magma compositions. When compared to our magmatic sources with 18.5 wt.% MgO, 15 wt.% MgO, and 6 wt.% MgO, it is obvious that the majority of Hapaimamo flow escaped the buffering effect of the summit magma chamber. This can be achieved by bypassing the summit chamber and reaching the rift zone by conduits other than the primary pathway. However, a small fraction of olivines in host rock and xenoliths were formed from a magma source of wt.% MgO. Therefore magma was also transported from the summit magma chamber to the SWRZ during periods of high magma flux. At least one pulse of relatively primitive magma from at least 13 km rose through dikes that deviate from the primary conduit and arrived at a shallow chamber at 3.3 km underneath the SWRZ during periods of high magma flux (Fig. 19). Garcia et al. (1995) proposed a similar pathway in which a magma density filter exists to allow SWRZ picritic magma to travel via deep conduits directly to the rift zone. The deeper, more picritic magma probably entrained pyroxene, olivine, and possibly other cumulates that crystallized at deeper levels and transported them to the shallow chamber, where it then mixes with a shallow differentiated magma. Chemical disequilibrium in minerals and mineralogical textures also indicate a prolonged period of time, free of frequent physical disruption, and chemical resetting. This can be interpreted as a