1.1. Méthodes et organisation de la thèse

      La thèse s'organise en quatre chapitres distincts: le Chapitre I est une introduction et un résumé des résultats. Le Chapitre II rapporte les résultats des détails texturaux, minéralogiques, et géochimiques des xénolites. Il est basé sur des analyses de microsonde électronique de tous les minéraux ainsi que sur l'analyse sur roche totale des éléments majeurs et traces et des isotopes du Sr et Ar. Dans ce chapitre, deux grands groupes de xénolites sont distingués et leur séquence de cristallisation sera discutée, incluant les évidences de migration de liquide et de fluide. Le Chapitre III rapporte les détails texturaux ainsi que les compositions en éléments majeurs et traces des plagioclases zonés d'un sous-groupe de xénolites. Cette étude est basée sur l'imagerie "Normaski Differential Interference Contrast" et sur les analyses par microsonde électronique et ionique. L'évidence des processus de migration de liquide proposé dans le Chapitre II est développée en recalculant la composition du liquide en équilibre avec le plagioclase. Le Chapitre IV porte sur l'origine des phlogopites riches en sodium présentes dans les xénolites. Pour ceci, des analyses par microsonde électronique et des images aux rayons X, ont été effectuées. Le détail des techniques analytiques est reporté sous forme d'annexe à la fin de chaque chapitre.

2.1 Volcán San Pedro et xénolites

      L'activité volcanique du Volcán San Pedro (1.5 km³) comprend deux phases: une phase de formation du cône avec des laves andésitiques et dacitiques et une phase plus jeune qui postdate l'effondrement du flan est, accompagnée d'une éruption explosive produisant des dépôts pyroclastiques dacitiques (Singer & Dungan, 1992). Cet ensemble est suivi par une séquence de laves qui enregistrent l'éruption d'une chambre magmatique fortement zonée: (1) 0.2 km³ de dacite à hornblende-biotite contenant d'abondants xénolites mafiques (jusqu'à 45 cm de diamètre) et des inclusions mafiques trempées (IMT). (2) 0.5 km³ de dacite à 2 pyroxènes avec d'abondantes IMT, et (3) 0.1 km³ d'andésites avec de rares IMT. La dernière activité volcanique consiste en 0.2 km³ d'andésites basaltiques qui reforment le cône sommital (Fig. 2). La majorité des xénolites sont gabbroïques (22 échantillons), mais quelques granites et roches métamorphiques ont également été échantillonnés. Les xénolites de petite taille sont arrondis ou sub-arrondis, tandis que les larges fragments sont anguleux (Fig. 2). Les xénolites sont présents de façon homogène dans la coulée, et représentent un volume inférieur à 10 %. Le fait que les xénolites soient exclusivement présents dans la première coulée qui a suivi l'éruption explosive du à l'effondrement du flanc est du Volcán San Pedro, suggère qu'ils sont les fragments des conduits ou des parties supérieures de la bordure de la chambre magmatique qui ont été arrachés lors de l'éruption (mécanismes similaires à ceux de l'éruption du Mt. St. Helens, le 18 mai 1980; Hekiler, 1995).

      

Figure 1. Carte géologique simplifiée du complexe Tatara-San Pedro (TSPC).

Figure modifiée d'après Singer et al. (1997).

      Les principaux centres volcaniques peuvent être distingués: Volcán Tatara, Volcán San Pedro. Les xénolites ont été trouvés dans une des coulées dacitiques du Volcán San Pedro.

      

Figure 2 A-C. Photos du Volcán San Pedro.

      A. Regardant vers le nord, vue du flanc sud du Volcán San Pedro. La coulée de lave, dans laquelle les xénolites sont présents, est marquée d'une flèche. Nous pouvons également distinguer l'affleurement de leucogranite du pluton Huemul (blanc) sur la gauche de la photo, recouvert par des roches métamorphiques et les laves du Volcán Tatara. L'escarpement du flanc effondré du volcan, qui a induit l'éruption est également visible. B. Nombreux xénolites angulaires à sub-arrondis dans la lave dacitique du Volcán San Pedro. C. Large xénolite gabbroïque (40 cm de diamètre).

3.1 Xénolites du Groupe I: norites à hornblende et olivine

      La principale caractéristique pétrographique de ces échantillons est la présence de textures de réaction: l'olivine est typiquement résorbée, xénomorphe et entourée de hornblende, d'orthopyroxène et de phlogopite, et est plus rarement en contact avec le verre. Les rares clinopyroxènes sont également résorbés, xénomorphes et sont présents seulement à l'intérieur des hornblendes. Le spinelle chromifère est présent en inclusion dans l'olivine, l'orthopyroxène, la horblende et dans le coeur du plagioclase. Par contre, l'orthopyroxène, la hornblende, le plagioclase, la phlogopite, et l'apatite sont euhedrals. A partir des critères texturaux, je propose que les réactions entre olivine + spinelle chromifère + clinopyroxène et un liquide évolué, ont produit l'assemblage hornblende + orthopyroxène + phlogopite avec un fort "mg-number" et de hautes teneurs en Cr₂O₃ (Table 1). Plusieurs caractéristiques pétrologiques suggèrent que ces réactions ne sont pas simplement dues à une cristallisation en système fermé: (1) les réactions entre le clinopyroxène ou l'olivine et le liquide produisant la hornblende, et entre l'olivine et le liquide produisant l'orthopyroxène, ont été observées lors d'expériences effectuées avec des basaltes et andésites à faible pression (<3 kbar) (e.g., Sisson & Grove, 1993; Grove et al., 1997; Moore & Carmichael, 1998; Barclay & Carmichael, in prep.). Cependant, dans aucune de ces expériences, la co-cristallisation de la hornblende et de l'orthopyroxène (et la phlogopite) a été observée. La phlogopite n'est présente dans aucune expérimentation, même avec une cristallinité allant jusqu'à 90 % (Kawamoto, 1996). (2) Les valeurs du rapport Na/Ca des hornblendes sont élevées (pour des teneurs en Al similaires) comparées à celles des hornblendes des xénolites gabbroïques des zones de subduction et des hornblendes des expériences effectuées sur des magmas mafiques à intermédiaires à faible pression (<3 kbar). Nous proposons que ces fortes valeurs du rapport Na/Ca des hornblendes reflètent les rapports élevés Na/Ca du liquide (e.g., les dacites de San Pedro ont un rapport Na/Ca d'environ 1,2). (3) Les valeurs du rapport Na/K des phlogopites sont plus hautes que celles reportées dans les expériences à faible pression (< 3 kbar) ou que celles des autres plutons ou xénolites gabbroïques. Ceci suggère que le liquide présent lors de la cristallisation des phlogopites était riche en K₂O et H₂O. (4) Le plagioclase montre un abrupt changement de composition de An_85-70 au coeur à An_45-20 en bordure. Ceci est difficile à expliquer lors d'une cristallisation en système fermé (e.g., Brophy et al., 1996). (5) La coexistence d'olivine forstéritique et de verre rhyolitique est atypique dans les système fermé, toutefois, quelques exemples ont été décris dans des systèmes ou un mélange mécanique entre magmas felsiques et mafiques à lieu (e.g., Feeley & Dungan, 1996).

      

Table 1. Compositions représentatives des minéraux des xénolites (écart maximal)

      Afin d'expliquer toutes ces observations, je propose un scénario ou les xénolites étaient formés d'un réseau cristallin composé d'olivine, de spinelle chromifère, de clinopyroxène et de plagioclase anorthitique avec un liquide mafique remplissant les interstices. Le liquide mafique a été remplacé par un liquide plus évolué qui a réagit avec les minéraux mafiques et engendre la cristallisation de la hornblende, de l'orthopyroxène et de la phlogopite. Ceci est suivi par la cristallisation des bordures albitiques des plagioclases. Le processus de migration du liquide est probablement régit par le contraste de densité entre le liquide mafique inter-cumulus et le liquide plus évolué et plus léger qui le remplace (i.e., convection compositionnelle). Un tel processus a été reproduit en laboratoire dans des expériences de cristallisation de sels (e.g., Chen & Turner, 1980; Kerr & Tait, 1986; Tait & Jaupart, 1992).

3.2. Xénolites du Groupe IICL: leuconorites à clinopyroxène

      La majorité de ces échantillons montre une texture en mosaïque, sériée, avec parfois une orientation préférentielle des plagioclases. Les clino- et orthopyroxènes sont xénomorphes à hypidiomorphes, avec des lamelles d'exsolution et sont souvent en amas interstitiels entre des cristaux de plagioclase. La hornblende et la phlogopite sont xénomorphes à hypidiomorphes, typiquement poecilitiques et entourent les minéraux résorbés et xénomorphes de pyroxène, d'olivine et de plagioclase. La hornblende et la phlogopite sont également présentes en remplissage de microfractures sub-linéaires. Les oxydes Fe-Ti sont automorphes et présents en inclusion dans les pyroxènes, mais dans de nombreux échantillons, ils sont poecilitiques et entourent le plagioclase, les pyroxènes et occasionnellement la hornblende et la phlogopite. L'apatite, automorphe, est présente entre les limites de grain des cristaux de plagioclase ou en inclusion dans la phlogopite. Contrairement au Groupe I de xénolites, de nombreux échantillons montrent des textures de rééquilibration sub-solidus le long des limites de grain des cristaux de plagioclase et entre le plagioclase et les pyroxènes (i.e., angle dièdre des limites de grain constant; Hunter, 1987). Le plagioclase montre des macles déformées, des microfissures, des limites de grain dentelées, et occasionnellement des recristallisations dynamiques de sous-grains. De plus, des microfractures discontinues (<0.5 mm de large) remplies d'oxydes, de hornblende et de phlogopite (et fréquemment d'orthopyroxène) sont présentes dans de nombreux échantillons. De telles microfractures recoupent tous les minéraux à l'exception des hornblendes poecilitiques et des phlogopites. Lorsque les microfractures intersectent le contact pyroxène-plagioclase, les deux minéraux sont résorbés, xénomorphes et sont entourés d'une bordure de hornblende ou de phlogopite. Ainsi, je propose que les microfractures étaient remplies d'un liquide ou d'un fluide aqueux qui a réagit avec les pyroxènes et le plagioclase afin de produire les minéraux hydratés.

      

Table 2. Variations compositionnelles des xénolites et basaltes du TSPC.

      Les teneurs en éléments majeurs de ce groupe de xénolites sont similaires à celles observées dans les basaltes riches en aluminium, typiques des zones de subduction. Comparés aux basaltes du TSPC, les xénolites ont de faibles concentrations en éléments traces incompatibles (e.g. Zr, Ce, Rb) et des teneurs similaires en éléments compatibles (e.g., Sr, Ni) (Table 2). Celles-ci ne peuvent pas s'expliquer par l'accumulation des minéraux comme l'olivine et le plagioclase et suggèrent plutôt que ces xénolites ont perdu du liquide interstitiel évolué. L'étendue du rapport des éléments incompatibles (e.g., P/Zr) suggère que la perte du liquide s'est effectuée avant et après la cristallisation de l'apatite. Les lattes de plagioclase déformées et les microfractures remplies essentiellement par la hornblende et la phlogopite sont des évidences texturales et minéralogiques qui indiquent que l'expulsion du liquide interstitiel évolué s'est faite par un processus de compaction des cristaux. Le rapport des éléments incompatibles prenant en compte K et Rb montre des valeurs très disparates (e.g., Rb/Y). La mobilisation du K et Rb par rapport aux autres éléments incompatibles (e.g., Zr, Y) peut s'expliquer au moyen d'une phase fluide aqueuse. Le coefficient de partage du K et Rb entre un fluide et un liquide est beaucoup plus grand que celui de Y et Zr (e.g., Keppler, 1996). Par conséquent, les fortes teneurs en Rb et K₂O et le rapport élevé Rb/Y (ou faible P/Rb) de certains xénolites peuvent être expliqués par l'arrivée d'une phase aqueuse (sous la forme de bulles) qui est dissoute dans le liquide restant. Le fluide a pu migrer le long des limites de grain des cristaux ou à travers les microfractures (e.g., Shinohara & Kazahaya, 1995). De plus, le haut rapport Cl/F de certaines apatites appuie l'idée que certains xénolites ont été affectés par l'arrivée d'un fluide aqueux. Les analyses d'isotopes stables disponibles (d¹⁸O) suggèrent que les xénolites n'ont pas été affectés par une circulation d'eau hydrothermale. Le processus décrit ci-dessus implique donc des fluides magmatiques.

3.3. Xénolites du Groupe IIHN: norites à hornblende

      Les échantillons de ce groupe montrent des textures hétérogènes avec de larges hornblendes xénomorphes poecilitiques (> 1cm) qui entourent les cristaux xénomorphes et résorbés d'olivine, de clinopyroxène, d'orthopyroxène, de plagioclase, de spinelle chromifère et d'oxydes Fe-Ti. La phlogopite est également présente sous forme de cristaux poecilitiques qui entourent les cristaux résorbés et xénomorphes d'olivine, d'orthopyroxène et de plagioclase. De rares apatites sont présentes aux limites de grain des plagioclases. Comme pour les xénolites du Groupe I, j'interprète les cristaux xénomorphes et résorbés d'olivine, de pyroxènes et de plagioclase à l'intérieur de la hornblende et de la phlogopite, comme une évidence de réaction entre les minéraux anhydres et un liquide riche en eau. Il est important de noter que le plagioclase et l'orthopyroxène dans la hornblende poecilitique sont déformés (macles courbées du plagioclase et extinction irrégulière de l'orthopyroxène). Les cristaux poecilitiques de hornblende et de phlogopite ne montrent pas d'évidence claire de déformation texturale. Ainsi, je propose que la déformation a eu lieu avant la cristallisation de la hornblende et de la phlogopite.

      La composition de la roche totale de ces xénolites montre de faibles teneurs en éléments incompatibles (e.g., Zr, Ce) et de fortes teneurs en Ni comparées aux basaltes TSPC (Table 2). Etant donné que la cristallisation de la hornblende se produit par réaction entre l'olivine, le plagioclase et le liquide, il est difficile d'établir, à partir des abondances modales, si les faibles concentrations en éléments incompatibles de ces xénolites sont dues uniquement à une accumulation de plagioclase et d'olivine, ou si la perte du liquide est également importante. Les rapports d'éléments incompatibles comme P/Zr et P/Y sont comparables à ceux des basaltes TSPC et suggèrent ainsi que si la migration d'un liquide interstitiel s'est produite, elle a eu lieu avant la cristallisation de l'apatite. Par contre, les rapports Rb/Y et K/Y sont plus haut que ceux observés dans les basaltes du TSPC. Utilisant les même arguments présentés pour les xénolites du Groupe IICL, je suggère qu'un fluide aqueux a percolé. Je propose ainsi, que le liquide interstitiel s'est échappé du réseau cristallin, et qu'un fluide s'est dissout dans le liquide résiduel, l'enrichissant en éléments volatils, incluant K, Rb et H₂O. Ceci à produit une réaction entre les minéraux et le liquide, formant la hornblende et la phogopite avec de forts "mg-numbers" et de fortes teneurs en Cr₂O₃. De telles réactions sont probablement responsables des hauts rapports Na/Ca de la hornblende et Na/K de la phlogopite (Table 1). L'arrivée d'un fluide et les réactions proposées ici, peuvent être analogues aux processus décris par Boudreau (1999) pour "olivine-bearing Zone I" du Complexe Stillwater (Montana). De plus, le haut rapport Cl/F de certaines apatites supporte l'idée que les xénolites ont été affectés par un fluide. Les analyses d'isotopes stables disponibles (d¹⁸O et dD) suggèrent que le fluide aqueux est magmatique et non météoritique.

Conclusions

      L'étude pétrographique et géochimique des deux groupes de xénolites gabbroïques du complexe de Tatara-San Pedro a montré que ces xénolites ont enregistré plusieurs épisodes de cristallisation liés à la migration de liquide. Les réactions des minéraux réfractaires (spinelle chromifère, olivine, pyroxènes et plagioclase) avec les liquides évolués et les fluides percolants, ont produit la cristallisation de la hornblende et de la phlogopite. Nous suggérons que la grande abondance de ces minéraux hydratés dans les gabbros calco-alcalins par rapport aux basaltes ou andésites basaltiques peut s'expliquer par une interaction entre des liquides évolués et les minéraux précocement cristallisés. Le résultat des calculs de dynamique des fluides simples suggère que la migration de ces liquides par convection compositionnelle à l'intérieur des zones partiellement cristallisées des chambres magmatiques calco-alcalines est un mécanisme de différentiation réalisable.

4.1. Texture et zonation en éléments majeurs des plagioclases

      Des traverses de 42 cristaux de plagioclase allant du centre à la bordure (échantillons Hx14n and Hx14b) ont été effectuées avec la microsonde électronique (environ 2500 analyses). Dans la discussion suivante, le coeur des plagioclases correspond à des compositions comprises entre An₈₆ et An₇₀ mol%, la zone de transition entre le coeur et la bordure correspond à des compositions comprises entre An₇₀ et An₄₅, et la bordure correspond à des compositions < An₄₅.

      Les cristaux de plagioclase des deux échantillons montrent des zonations en éléments majeurs et des textures comparables, avec l'exception que les plagioclases de l'échantillon Hx14b montrent plus de surfaces de dissolution et une zonation oscillatoire plus développée que dans l'échantillon Hx14n. Les cristaux de plagioclase qui ne sont pas inclus dans l'orthopyroxène, la hornblende, ou la phlogopite montrent un coeur An₈₆ à An₇₀ avec une zonation normale. Le coeur des plagioclases est entouré d'une bordure à zonation normale de An₄₅ à An₂₀. Le passage entre le coeur et la bordure se fait par un abrupt (en moins de 100 µm) changement de composition de 30 à 35 mol% d'anorthite (Fig. 3). Les cristaux de plagioclase en inclusion dans l'orthopyroxène, la hornblende, ou la phlogopite ont les mêmes compositions et présentent une zonation normale de An₈₆ à An₄₀, ce qui indique que ces trois minéraux mafiques ont co-cristallisé.

4.2. Zonation en éléments traces des plagioclases

      Des traverses du centre au bord des plagioclases par microsonde ionique ont été effectuées sur neuf cristaux de plagioclase représentatifs de l'ensemble des échantillons. Au total, 90 analyses ont été effectuées et les éléments suivants ont été analysés: Ca, Mg, Fe, Ti, K, Rb, Sr, Ba, La, Ce, Y, et Li.

      Les concentrations en Ti, Sr, Ba, La, Ce, et Li des coeurs des plagioclases augmentent en allant vers la zone de transition. Les teneurs en Fe restent plus ou moins constantes, tandis que les teneurs en K augmentent dans certains cristaux et diminuent dans d'autres. Les concentrations en Mg diminuent toujours. Les concentrations en Rb et Y ne définissent pas réellement un enrichissement ou un appauvrissement, probablement du fait de leurs larges erreurs analytiques (25-30 %). De grandes et abruptes variations de concentration en Fe, Ti, Sr, Ba, K, La, Ce, et Y sont présentes à la transition coeur-bordure. Les bordures ont des concentrations plus élevées en Sr, Ba, K, La, Ce et plus faibles en Fe, Ti, et Y que les coeurs (Fig. 3). Les concentrations en Mg, Fe, Ti, La, et Ce restent plus ou moins contantes de l'intérieur vers l'extérieur de la bordure de tous les cristaux. Les concentrations en Sr, Ba et K ont tendance à augmenter vers la partie extérieure de la bordure.

4.3. Liquides en équilibre avec les plagioclases

      Dans cette partie, je calcule la composition des liquides en équilibre avec les plagioclases en utilisant les équations de Blundy & Wood (1991; équations #18 and #19) pour Sr et Ba, et les équations de Bindeman et al., (1998) pour Mg, Ti, Fe, K, La, et Ce. On observe que:

1. Les liquides en équilibre avec les coeurs des plagioclases montrent un enrichissement en TiO₂, Sr, Ba, K₂O, La, et Ce et un appauvrissement en MgO, qui sont comparables à l'évolution des laves basaltiques à basaltiques-andésitiques du Volcán San Pedro, et qui s'expliquent par la cristallisation de olivine + clinopyroxène + plagioclase anorthitique.
2. Les liquides en équilibre avec les zones de transition coeur-bordure des plagioclases montrent de grandes variations de concentration, comparables à l'évolution des laves basaltiques à andésitiques du Volcán San Pedro.
3. Les liquides en équilibre avec les bordures des plagioclases ont de plus faibles concentrations en TiO₂, FeO* et Sr, et de plus fortes concentrations en K, Ba, et La que le liquide en équilibre avec le coeur des plagioclases. De telles différences de composition peuvent être expliquées par la cristallisation de la hornblende (± orthopyroxène et phlogopite).
4. Les liquides en équilibre avec les bordures des plagioclases montrent une diminution en Sr et Ba et une augmentation en K/Ba et K₂O, tandis que les concentrations en TiO₂, MgO, La, et Ce restent plus ou moins constantes. Cette tendance n'est pas représentée dans les laves de San Pedro et peut s'expliquer par la cristallisation d'un assemblage minéralogique dominé par la phlogopite et le plagioclase albitique (e.g., <An₄₀) et montre ainsi l'évolution d'un liquide de composition dacitique à rhyolitique. Le fait que les concentrations en La et Ce restent plus ou moins constantes, pourrai s'expliquer par la cristallisation de l'apatite (présente dans le verre) ou par les hauts coefficients de partage de La et Ce reportés pour les biotites (e.g., Nash & Crecraft, 1985).

      Bien que les liquides en équilibre avec les bordures des plagioclases montrent une évolution en accord avec le fractionnement d'un liquide dacitique à rhyolitique, les concentrations absolues calculées pour Ba et K₂O sont environ 50 % plus basses et celles de La et Ce sont 50 % plus hautes, comparées aux compositions du verre interstitiel rhyolitique et aux laves et plutons dacitiques et rhyolitiques du TSPC. Après avoir considéré les différents facteurs (incluant la cristallisation en système fermé et la cinétique de cristallisation) pouvant être responsables de la différence entre les compositions calculées et celles des liquides évolués du TSPC, deux semblent possibles: (1) les coefficients de partage obtenus à partir des équations empiriques de Blundy & Wood (1991) et Bindeman et al. (1998) ne sont peut être pas valables pour les compositions évoluées des liquides et plagioclases considérées ici, et (2) étant donné que le K, Rb et Ba sont fortement compatibles dans la phlogopite, il semble possible que les faibles concentrations en ces éléments calculées pour le liquide percolant résultent de la cristallisation de la phlogopite. L'arrêt de la cristallisation de la phlogopite et la poursuite du processus de migration du liquide pourrai expliquer que la composition du verre interstitiel analysé ne reflète pas une cristallisation prolongée mais la composition du liquide migrant.

      

Figure 3. Profils de composition en anorthite et en éléments traces des cristaux de plagioclase de l'échantillon Hx14n.

      Les bandes grises indiquent la zone de transition entre le coeur et la bordure. Notez que les éléments traces suivent l'abrupt changement en éléments majeurs suggérant un changement majeur dans la composition du liquide interstitiel. La taille des symboles est égale ou plus grande à la précision de 2-s de tous les éléments.

Conclusions

      La zonation texturale et de composition en éléments majeurs et traces des plagioclases, combinée avec la composition du verre interstitiel d'une suite de xénolites gabbroïques, a permis de retracer les complexes processus de différentiation qui peuvent avoir lieu dans les zones partiellement cristallisées des chambres magmatiques. Les compositions des liquides recalculées en équilibre avec les plagioclases indiquent qu'après une première étape de différentiation d'un liquide basaltique à basaltique-andésitique, le liquide interstitiel a évolué brusquement vers des compositions dacitiques. Ceci peut être expliqué par la migration du liquide à travers le réseau cristallin. Le liquide mafique interstitiel a été remplacé par un liquide dacitique qui a réagit avec les minéraux préexistants (olivine, clinopyroxène et spinelle chromifère), et a formé l'orthopyroxène, la hornblende, et la phlogopite. Pendant et après ces processus, le liquide interstitiel a évolué vers des compositions rhyolitiques. Cependant, pour les compositions des plagioclases (e.g., < An₄₀) ou des liquides évolués, les compositions recalculés sont différentes de celles du verre rhyolitique interstitiel des xénolites et des laves et plutons dacitiques du TSPC. Ceci peut être du au fait de la complexité des processus de migration des liquides et des réactions qui se produisent dans les zones partiellement cristallisées des chambres magmatiques, ou au fait de l'incertitude des coefficients de partage plagioclase-liquide des magmas siliciques.

2.1. Volcán San Pedro and xenoliths

      Volcanic activity at Volcán San Pedro (1.5 km³) is divided (Singer & Dungan, 1992) into a cone-building phase comprising andesitic and dacitic lavas and a younger phase that post-dates a sector collapse of the eastern flank, which was accompanied by an explosive eruption that produced air-fall dacitic deposits. This was followed by a sequence of lavas that apparently records the downward tapping of a strongly zoned magma chamber. The eruptive sequence comprises: (1) 0.2 km³ of biotite-hornblende dacite containing abundant mafic xenoliths (up to 45 cm in diameter) and quenched mafic inclusions (QMI), (2) 0.5 km³ of two-pyroxene dacite with abundant QMI, and (3) 0.1 km³ of two-pyroxene andesite with rare QMI. The last volcanic activity at this cone consisted of 0.2 km³ of basaltic andesite magma that rebuilt the summit cone. The majority of the xenoliths are gabbroic (22 samples), although scarce granites (1 sample) and metamorphic rocks (1 sample) were also collected. Small xenoliths are rounded to subrounded, whereas larger fragments are angular (Fig. 3). The xenoliths are homogeneously distributed in the first and most evolved flow of the post-collapse eruptive sequence. They commonly are <10 vol. %. The fact that the xenoliths have exclusively been found in the first lava flow following the catastrophic structural failure of the east flank of Volcán San Pedro suggests that they are fragments of the conduits or upper parts of the margins of the San Pedro magma chamber that were shattered during the eruption (in a similar fashion to the May 18, 1980 Mount St. Helens eruption; Heliker, 1995).

      

Figure 1. Simplified geological map of Central Chile showing the location of the Tatara-San Pedro Volcanic Complex.

      Black triangles indicate main Quaternary volcanic centres. Pz-Palaeozoic rocks, Mz-Mesozoic rocks. Grey shaded areas indicate Tertiary plutons.

Figure modified from Hildreth & Moorbath (1988) and from Dungan et al. (submitted). The location of Tertiary plutons is from Mapa Geológico de Chile (1982).

      

Figure 2. Simplified geological map of the Tatara-San Pedro Volcanic Complex.

      Note the two plutons that crop out in the northern and southern part of the complex. The xenoliths were found in a dacite flow of Volcán San Pedro.

Figure modified from Singer et al. (1997).

      

Figure 3 A-D. Field and hand speciment aspect of the xenoliths.

      A. Numerous angular to subrounded xenoliths in the biotite-hornblende dacite lava flow of Volcán San Pedro. B. Large gabbroic xenolith (40 cm diameter). C. Olivine norite xenolith (sample Hx14n) bearing interstitial glass. Lava is on the left corner of the picture. Scale bar is 12 cm across. D. Hornblende leuconorite xenolith (sample Hx14y). Note the NW-SE trending microfractures. Brown patches are hornblende and phlogopite. Scale bar is 12 cm across.

3.1. Group I. Partially crystallized olivine norites

      These samples consist of olivine, orthopyroxene, hornblende, plagioclase, and phlogopite forming a medium grained (1-5 mm) crystal network with interstitial vesiculated SiO₂-rich (67-72 wt %) glass filling the interstices. Glass is distributed in isolated and occasionally interconnected pockets bounded by euhedral crystal faces suggesting that it is residual from crystallization and not due to partial melting (Fig. 5). A salient petrographic feature of these samples is the presence of reaction textures: olivine is typically resorbed, anhedral and surrounded by hornblende, orthopyroxene, and phlogopite, but is also much less commonly in contact with glass (Fig. 5). Rare clinopyroxene is also resorbed, anhedral and present only within hornblende. Cr-spinel occurs as inclusions in olivine, orthopyroxene, hornblende, and in plagioclase cores. In contrast, orthopyroxene, hornblende, plagioclase, phlogopite, and apatite are commonly euhedral (Fig. 5). From textural criteria we infer reaction relations between the early-crystallized cumulus olivine and clinopyroxene and the interstitial liquid to produce post-cumulus hornblende, orthopyroxene, and phlogopite (Fig. 5 and Table 2).

      

Figure 5 A-D. Photomicrographs of Group I xenoliths.

      A. Sample Hx14n. Anhedral, resorbed Ol surrounded by Hbl and Opx. This suggests a reaction relation between the interstitial liquid and Ol to produce Hbl and Opx. B. Sample Hx14k. Hbl enclosing anhedral, resorbed Ol, but euhedral Pl. Note also Phl surrounding anhedral Ol.. C. Sample Hx14b. Vesiculated interstitial SiO₂-rich glass (66-72 wt %) in contact with euhedral Opx, Hbl and Phl. Also note the Ol crystals (Fo₈₂) in contact with the silica-rich glass. D. Sample Hx14b. Detail of an euhedral Phl in vesiculated residual glass. Phl is surrounding anhedral, resrorbed Ol crystals suggesting a reaction relation between the interstitial liquid and Ol to produce Phl.

3.2. Group IICL. Clinopyroxene leuconorites with subsolidus textures

      The majority of these samples are characterized by a mosaic, seriate texture (Shelley, 1993), in some case with plagioclases displaying a preferred orientation. Anhedral to subhedral clino- and orthopyroxenes display exsolution lamellae in many samples and commonly occur as clusters interstitial to larger plagioclase crystals. Orthopyroxene occasionally occurs as oikocrysts including plagioclase, and in some samples is present in microfractures (see below). Olivine is commonly anhedral and contains inclusions of Cr-spinel (only present in sample Hx14e). It is commonly surrounded by rims of orthopyroxene, Fe-Ti oxide symplectites, and occasionally, by small flakes of phlogopite which we refer to as late phlogopite.

      Subhedral to anhedral hornblende and phlogopite typically occur as small (< 1 mm across) crystals surrounding resorbed, anhedral pyroxenes, olivine, and plagioclase. Hornblende and phlogopite are also found filling sub-linear microfractures (Fig. 6). Radial aggregates of late phlogopite are occasionally present as the replacement products of hornblende, as reaction rims around olivine, and in some cases after orthopyroxene. Euhedral Fe-Ti oxides occur as inclusions in pyroxenes, but in many samples these are poikilitic, anhedral and they surround anhedral plagioclase, pyroxenes and occasionally hornblende and phlogopite (Fig. 6). Anhedral apatite is the only accessory mineral and it occurs between grain boundaries of plagioclase crystals or inside phlogopite.

      In contrast to Group I xenoliths, most samples show subsolidus textural reequilibration along grain boundaries between plagioclase crystals and between plagioclase and pyroxenes (i.e., constant grain boundary dihedral angles between crystals; Hunter, 1987). These samples also have been variably deformed (see Table 1). Plagioclase commonly shows bent twins, microcracks, and serrated grain boundaries, and occasionally subgrain dynamic recrystallization (Fig. 6). Orthopyroxene and hornblende oikocrysts occasionally display weak uneven extinction may be related to subgrain recrystallization, whereas phlogopite oikocrysts show kink-bands. In addition, discontinuous microfractures (<0.5 mm in width) filled with Fe-Ti oxides, hornblende and phlogopite (and frequently also orthopyroxene) are present in many samples (Fig. 6). Such microfractures cut across all minerals except poikilitic hornblende and phlogopite. Where microfractures intersect pyroxene-plagioclase contacts, both minerals are resorbed, anhedral and they are mantled by a rim of hornblende or phlogopite (Fig. 6). From this textural relation we infer that the microfractures hosted melts or aqueous fluids that reacted with pyroxenes and plagioclase to produce the hydrous minerals (Table 2).

      Minor amounts of glass (<6.5 vol. %) are present in some of these samples. Glass occurs exclusively along resorbed grain boundaries of plagioclase crystals, plagioclase-orthopyroxene, and plagioclase-hornblende, indicating that it is the result of partial melting and not residual from crystallization as is the case of Group I xenoliths. Small euhedral crystals of orthopyroxene, clinopyroxene and plagioclase are present in the glass.

      

Figure 6 A-F. Photomicrographs of Group IICL xenoliths.

      A. Sample Hx14a. Typical leuconorite consisting of abundant Pl, Cpx, Opx and Fe-Ti oxides. B. Sample Hx14i. Poikilitic Fe-Ti oxides enclosing resorbed Pl crystals. Note Hbl crystallized on pyroxenes and the small microfracture on the right corner of the picture. C. Sample Hx14i. Note microfracture cutting across Pl and filled with Hbl and Phl. Where the microfracture intersects pyroxenes, larger Hbl crystals are present. D. Sample Hx14q. Large microfracture filled with Hbl and Phl. Note also the numerous smaller microfractures filled with Hbl and Phl. E. Sample Hx14y. Poikilitic Hbl surrounding subrounded Pl crystals. Note how Pl crystals with bent twins (black arrows) are also fractured. F. Sample Hx14y. Fractured and bent Pl crystals with microfractures filled with Hbl, Phl and in some cases Opx (center of the picture).

3.3. Group IIHN. Hornblende norites with subsolidus textures

      Samples from this group are texturally heterogeneous and are characterized by large anhedral hornblende oikocrysts (>1cm) that surround anhedral, resorbed olivine, clinopyroxene, orthopyroxene, plagioclase, Cr-spinel, and Fe-Ti oxides (Fig. 7). Where hornblende is absent, the samples consist of large orthopyroxene crystals (up to 5 mm long) and a mosaic of subhedral plagioclase. Phlogopite is also present as oikocrysts that include resorbed, anhedral olivine, orthopyroxene, and plagioclase. It occasionally occurs inside hornblende oikocrysts. Rare anhedral apatite is present along plagioclase grain boundaries.

      As for the previous groups of xenoliths, we interpret the anhedral and resorbed olivine, pyroxenes, and plagioclase inside hornblende and phlogopite as evidence of reaction relations between the anhydrous minerals and a water-rich melt (Table 2). It is worth noting that plagioclase and orthopyroxene inside hornblende oikocrysts are deformed (bent twins in plagioclase and uneven extinction in orthopyroxene, Fig. 7). Hornblende oikocrysts do not show clear textural evidence of deformation, whereas phlogopite shows occasionally kink bands. From these textural relations we infer that a main deformation event occurred prior to hornblende and phlogopite crystallization.

      Small and discontinuous microfractures (<0.5 mm) cut through plagioclase crystals and are filled either with hornblende, phlogopite or plagioclase. Hornblende oikocrysts are not cut by the microfractures. These samples are also partially melted, with minor amounts (<0.5 vol. %) of glass occurring along resorbed grain boundaries of plagioclase crystals, plagioclase-orthopyroxene and plagioclase-hornblende.

      

Table 2. Proposed reactions derived from textural observations
	Orthopyroxene	Hornblende	Phlogopite
Group I.Olivine norites	li + Ol ± Cr-Spl	li + Ol ± Cr-Splli + Cpx ± Cr-Spl	li + Ol ± Cr-Spl
Group IICL.Clinopyroxene leuconorites	li + Ol	li + Cpxli + Opxli + Ol + Pl ± Cr-Spl li + Cpx + Pl ± Cr-Splli + Opx + Pl ± Cr-Spl	li + Ol ± Pl ± Cr-Spl
Group IIHN.Hornblende norites	li + Ol	li + Ol + Pl ± Cr-Spl li + Cpx + Pl ± Cr-Splli + Opx + Pl ± Cr-Spl	li + Ol + Pl ± Cr-Spl
Note: li = liquid. Mineral symbols after Kretz (1983).

      

Figure 7 A-B. Photomicrographs of Group IIHN xenoliths.

      A. Sample Hx14z. Large poikilitic Hbl surrounding anhedral, resobed Pl and Ol, suggesting a reaction relationship between the interstitial liquid, Pl and Ol to produce Hbl. Note the bent twins of the Pl crystal in the upper left corner of the picture (black arrow). B. Sample Hx14v. Poikilitic Phl surrounding anhedral Ol and subhedral Pl.

4.1. Group I xenoliths. Partially crystallized olivine norites

---

---

---

---

---

4.2. Group II. Xenoliths with subsolidus textures

---

---

---

---

---

---

---

---

5.1. Group I. Partially crystallized xenoliths

      The major element compositions of these xenoliths are characterized by low SiO₂ (46-47 wt %) and high MgO (20-21 wt %) when compared to the basaltic composition of the TSPC (Table 11 and Fig. 15). The xenoliths have low concentrations of incompatible minor and trace elements (K₂O = 0.52-0.71 wt %; Rb =10-12 ppm; Zr = 53-56 ppm; Y = 8-9 ppm), whereas concentrations of Ni (446-643 ppm), and Cr (707-1308 ppm) are much higher than those of the TSPC basalts (Table 11). The high MgO, Ni, and Cr concentrations together with the high modal proportions of olivine suggest that the xenoliths have accumulated olivine (and Cr-spinel) (Fig. 15). Ratios of elements that are not affected by olivine accumulation, such as Ca/Na mol (2.4-3.1), K/P (8-13), P/Zr (9-10), and Rb/Y (1-1.5) fall within the range of TSPC basalts (Fig. 16). The REE abundances are low (e.g., Ce = 12.4-15.5 ppm), and the La/Yb values are within the mean of the TSPC basalts (La/Yb = 8.1-10.6) (Fig. 17).

5.2. Group IICL xenoliths

      The major element compositions of most of these xenoliths are comparable to those of high-alumina basalts typical of subduction zones (e.g., Gust & Perfit, 1987; Kersting & Arculus, 1994). The SiO₂ concentrations range from 49 to 52 wt %, MgO from 5.2 to 8.2 wt %, Al₂O₃ from 17.2 to 19 wt %, and Ca/Na mol from 2.3 to 3.8 (Table 11). Samples Hx14e, Hx14h, and Hx14y stand out from the rest by their lower concentrations of SiO₂ (46-47 wt %), generally higher MgO (7.6-10.7 wt %), higher Al₂O₃ (21.8-23 wt %), and higher Ca/Na (4.4-7.7) (Fig. 15). Trace and minor element abundances are highly variable. Most xenoliths have concentrations of incompatible elements that range from those like TSPC basalts to much lower values (Fig. 15). For example, concentrations of K₂O range from 0.2 to 0.8 wt %, Rb from 3 to 40 ppm, Zr from 7 to 84 ppm, and Y from 5.3 to 23.4 ppm (Table 11). The P₂O₅ and TiO₂ also show large concentration ranges, from lower to higher values than those of the TSPC basalts. In contrast, the Sr and Ni concentrations of most xenoliths fall within the concentration ranges of the TSPC basalts (Fig.15). Samples Hx14e, Hx14h, and Hx14y also have lower concentrations of some incompatible elements (i.e., Y and Zr), but the Ni concentrations (133-237 ppm) are higher than most other xenoliths, whereas K₂O (0.34-2.48 wt %) and Rb (10-80 ppm) are highly variable. The low concentrations of incompatible element of most xenoliths could be due to accumulation of plagioclase or olivine. However, the fact that most xenoliths have Sr and Ni concentrations within the values of the TSPC basalts suggests that loss of interstitial melt rich in incompatible elements is a more plausible explanation. Accumulation of olivine may be a contributing factor for samples Hx14e, Hx14h, and Hx14y, as they have higher Ni concentrations than the rest of the xenoliths and the TSPC basalts.

      Ratios of incompatible elements are also highly variable. Most of the xenoliths have values of K/P (5-76) that range from higher to lower than those of the TSPC basalts. The P/Zr (11-38) and Rb/Y (0.4-27) values vary from those of the TSPC basalts to higher (Fig.16). Samples Hx14e, Hx14h, and Hx14y have P/Zr values (10-11) that fall within those of the TSPC basalts, whereas Rb/Y (2-27) and K/P (8-78) range from values of the TSPC basalts to much higher (Fig. 16). Such a large range of incompatible element ratios further indicates that open-system processes (e.g., melt or fluid migration) may have modified the bulk chemistry of the xenoliths.

      The REE contents are generally low, but range from higher (e.g., Ce = 34.8 ppm) and steeper patterns (La/Yb = 11.9) than those of the TSPC basalts, to lower REE concentrations (e.g., Ce = 7 ppm) and less steeper patterns. Some xenoliths show a positive Eu anomaly (Eu/Eu* up to 2.5) when normalized to primitive mantle (McDonough et al., 1992). In general, samples with lower REE concentrations have higher positive Eu anomalies, and their REE patterns are flatter (Fig. 17). Positive Eu anomalies can be produced by plagioclase accumulation, although they could also occur if the parent liquid had initially a positive Eu anomaly or, as it will be discussed later, by loss of interstitial liquid with a negative anomaly. The REE concentrations of samples Hx14e, Hx14h, and Hx14y are low (e.g., Ce = 10-11 ppm) with the HREE (Heavy Rare Earth Elements, Gd to Lu) below determination limits (Fig. 17).

5.3. Group IIHR xenoliths

      These samples have low concentrations of SiO₂ (~ 45 wt %) and high MgO (13.3-16.5 wt %) when compared to the mean basaltic composition of the TSPC (Fig. 15). The concentrations of incompatible elements are also low (K₂O = 0.26-0.47 wt %, Zr = 28-30 ppm, and Y = 3.5-4.2 ppm). The concentration of Ni (226-335 ppm) and the Ca/Na values (5.2-6.3) are higher than those of the TSPC basalts, thus olivine and plagioclase accumulation could be partly responsible for the low incompatible element abundances. Ratios of incompatible elements such as P/Zr (6-10) are within the values of the TSPC basalts, but the K/P (7-25) and Rb/Y (2-3) values range from those of the TSPC basalts to higher (Fig. 16). This indicates that processes other than plagioclase and olivine accumulation have modified the composition of these samples (e.g., melt or fluid migration; see Discussion). They have very low REE concentrations (e.g. Ce= 6.7-7.9 ppm), with the HREE below detection limits (Fig. 17).

      

Figure 15. Major and trace element variation diagrams of the three groups of xenoliths.

      Also shown is the mean composition of ten TSPC basalts. Black arrows indicate the effects of accumulation of Pl and Ol to the TSPC composition. Ni concentration is that of Ol analysis n-V-1c. The Sr concentration of An₈₈ and An₆₀ is taken from the isotope dilution analyses of Pl of samples Hx14v and Hx14a, respectively (Table 12). See text for discussion.

      

Figure 16. Variation diagrams of element ratios of the three groups of xenoliths.

      The legend is the same as in Fig. 15. The arrow with the "a" letter indicates the geochemical effect (qualitative) of removing interstitial melt (after apatite crystallization) to a bulk composition of the TSPC. The arrow with the "b" letter indicates the geochemical effect (qualitative) of addition of a fluid phase. See text for discussion.

      

Figure 17 A-C. Rare earth element concentrations normalized to primitive mantle (McDonough, 1992).

      A. Group I xenoliths. B. Group IICL xenoliths. C. Group IIHN xenoliths. See text for discussion.

6.1.⁴⁰Ar/³⁹Ar analyses

      Incremental furnace heating ⁴⁰Ar/³⁹Ar analyses were performed at the University of Geneva following the methods described in Singer & Pringle (1996). Hornblende separates from the two samples of Group IIHN xenoliths show comparable apparent age spectra (Fig. 18), but they do not define a plateau that could be used to derive a reliable age (e.g., Singer & Pringle, 1996). Apparent ages increase with temperature from ca. 0.5-3 Ma (at 750 ^oC) to 8-9 Ma (at 1050 ^oC), and then decrease to ca. 4 Ma (at 1200 ^oC).

      The high K/Ca of the first temperature steps (up to 950 ^oC) suggest the presence of phlogopite in the mineral separate, but afterwards the K/Ca value remains low (0.04-0.05) and within the values of hornblende obtained by electron microprobe analyses (0.03-0.08). This suggests that hornblende was the main degassing mineral (Fig. 18). Determining the cause(s) of the discordant ³⁹Ar release spectra is beyond the scope of this paper, but since the samples are xenoliths, heat from the host lava may have modified their Ar distribution and partially degassed the hornblendes. In any case, we interpret these Ar data as indicating that the xenoliths are certainly older than 1 Ma and may be up to 8 Ma. The total fusion age is ca. 5 Ma, which is ~1 Ma younger than the age of the Huemul and Cerro Risco Bayo plutons (6-6.4 Ma; Nelson et al., 1999) that form the basement of the TSPC.

      

Figure 18. Results of ⁴⁰Ar/³⁹Ar analyses of two hornblende separates (samples Hx14v and Hx14z) of Group IIHN xenoliths. See text for discussion.

6.2.⁸⁷Sr/⁸⁶Sr and Rb-Sr concentrations.

      ⁸⁷Sr/⁸⁶Sr plus Rb and Sr concentrations of bulk-rock, hornblende and plagioclase mineral separates were determined by Thermal Ionization Mass Spectrometry at the University of California at Los Angeles (Table 12). Two samples of Group I xenoliths have indistinguishable ⁸⁷Sr/⁸⁶Sr of 0.703948-0.703955, which are lower than those of Group II xenoliths and among the lowest of the entire volcanic complex (J. Davidson and M. Dungan, unpublished data).

      As the ⁴⁰Ar/³⁹Ar analyses suggest that Group II xenoliths are < 9 Ma and they have low Rb concentrations, the reported Sr isotopic compositions are close to the initial values. Whole-rock and plagioclase ⁸⁷Sr/⁸⁶Sr of three samples of Group IICL xenoliths vary from 0.704009 to 0.704117, which are within the values of the majority of the lavas erupted in the TSPC. The bulk-rock, plagioclase, and hornblende ⁸⁷Sr/⁸⁶Sr of sample Hx14v from Group IIHN xenoliths are within error (0.704048-0.704060). Those of sample Hx14z are also within error (0.704037-0.704021) but slightly lower than those of sample Hx14v. The small range of ⁸⁷Sr/⁸⁶Sr (< 0.00014) of minerals and bulk-rock of all xenoliths suggests that their composition has not been modified by interaction with radiogenic liquids.

7.1. Interpretation of mineral and whole-rock compositions of Group I xenoliths

---

---

---

7.2. Interpretation of mineral and whole-rock compositions of Group IICL xenoliths

---

---

---

7.3. Interpretation of mineral and whole-rock compositions of Group IIHN xenoliths

---

---

---

7.4. Implication for the presence of hornblende and phlogopite in calc-alkaline gabbroic rocks

      The three groups of gabbroic xenoliths have significants amounts of hornblende and phlogopite, either as small crystals filling microfractures, or as large poikilitic post-cumulus crystals that can make up >50 vol. % of the rock. In this respect they are not unusual, and a survey of the literature shows that the majority of calc-alkaline gabbroic xenoliths and plutons have important amounts of hornblende (xenoliths: Aoki, 1971, Itinome-gata, Japan; Arculus & Wills, 1980, Lesser Antilles; Conrad & Kay, 1984, Adak, Aleutian arc; Grove & Donnelly-Nolan, 1986, Medicine Lake, California; Beard, 1986, compilation from different sites; Yagi & Takeshita, 1987, Japan; Fichaut et al., 1989, Martinica, Lesser Antilles; Beard & Borgia, 1989, Arenal volcano, Costa Rica; Heliker, 1995, Mt. St. Helens; Hickey-Vargas et al., 1995, Calbuco volcano, southern Chile; gabbroic plutons: Smith et al., 1983, Peninsular Ranges batholith, California; Ulmer et al., 1983, Adamello batholith, Italy; Fabriès et al., 1984, Saint Quay-Portrieux intrusion, Britanny, France; Otten, 1984, Artfjället gabbro, Swedish Caledonides; Regan, 1985, Coastal batholith of Peru; Whalen, 1985, Uasilau-Yau Yau Intrusive Complex, New Britain; Beard, 1986, compilation from different sites; Himmelberg et al., 1987, Yakobi intrusion, Alaska; Beard & Day, 1988, Smartville Complex, California; DeBari & Coleman, 1989, Tonsina Complex, Alaska; Springer, 1989, Pine Hill Complex, California; Kepezhinskas et al., 1993, Kamchatka; DeBari, 1994, Fiambalá intrusion, Argentina; Tepper, 1996, Chilliwack batholith, Washington; Sisson et al., 1996, Hornblende gabbro sill, California). This is in marked contrast with the rare occurrences of hornblende phenocrysts in basalts or basaltic andesites in arc volcanoes world wide (Sigurdsson & Shepherd, 1974, Kick'em-Jenny Volcano, Lesser Antilles; Arculus, 1976, Grenada, Lesser Antilles; Arculus et al., 1976, Bogoslof Volcano, Alaska; Luhr & Carmichael, 1985, Cerro la Pilita, Mexico; Peterson & Rose, 1985, Ayarza caldera, Guatemala; Rose, 1987, Santa María Volcano, Guatemala). The occurrence and compositions of hornblende (e.g., high mg-number and Cr₂O₃ contents) in gabbroic xenoliths has led to some authors (e.g., Conrad & Kay, 1984; Yagi & Takeshita, 1987; Beard & Borgia, 1989) to propose that it was an early crystallizing mineral, with the implication that it might be responsible for the calc-alkaline trend of subduction zone magmas (e.g., Yagi & Takeshita, 1987). Experiments on basaltic and basaltic andesitic water-bearing compositions have shown that, in some instances, hornblende can be a near liquidus mineral even at low pressures (< 3 kbar; e.g., Sisson & Grove, 1993; Barclay & Carmichael, in prep). This has led to the interpretation that the paucity of hornblende phenocrysts in mafic lavas is due to an increase in magma crystallinity once hornblende crystallizes, thus hornblende-bearing mafic magmas are unable to erupt (Barclay & Carmichael, in prep). The petrological, mineralogical and geochemical characteristics of the San Pedro xenoliths suggest that the large amount of hornblende in gabbroic rocks can be due to open-system percolation in crystal piles. Aside from the possibility of hornblende crystallizing due to protracted closed-system crystallization, we propose that migration of evolved melt (and fluid) in plutonic systems can produce reactions with early crystallized minerals (olivine, Cr-spinel, pyroxenes and plagioclase) leading to the production of large proportions of hornblende (and phlogopite) with compositions that resemble those of early crystallization from mafic magmas (e.g., high mg-numbers and Cr₂O₃ contents).

4.1. Hornblende

      Hornblende is magnesiohastingsite (classification after Leake et al., 1997), with Cr₂O₃ contents commonly <0.30 wt %, but occasionally up to 1 wt %. Center to margin compositional traverses show that TiO₂ concentration decreases (from ~ 4 wt % to ~1 wt %) whereas Al₂O₃, MgO, and Na₂O abundances increase slightly towards the crystal margin.

      Some crystals are sector zoned and important compositional differences in Ti, Al, Mg and Ca occur across sectors. The two hornblende crystals have overlapping trace element compositions, but they display rather large concentration ranges. Rb varies from 7.4 to 8.3 ppm, Sr from 171-225 ppm, Ba from 57 to 74 ppm, La from 3.5 to 6.6 ppm, and Y from 31 to 45 ppm (Table 1).

4.2. Interstitial glass

      The interstitial glass is vesiculated and is variable in composition. Glass from sample Hx14n ranges from dacitic to rhyolitic, with SiO₂ varying from 66.7 to 71.6 wt %, Al₂O₃ from 18.7 to 14.5 wt %, Na₂O from 2.6 to 6.6 wt %, and K₂O from 3.7 to 8.6 wt %. Glass from sample Hx14b is rhyolitic, with higher SiO₂ (69 to 75 wt %), similar Al₂O₃ (14-17 wt %), but lower Na₂O (3.2 to 5.3 wt %) and K₂O (3.6 to 5.5 wt %) contents (Table 1; additional glass analyses can be found in Chapter II).

      The high MgO concentration (0.8-1.4 wt %) of analyses b-3 and b-8 compared to the rest of the glasses (0.17-0.32 wt %) and to the electron microprobe analyses suggest that they are mixtures of glass and minerals, and thus they will not be further considered. The trace element composition of glass in sample Hx14b is variable, with large concentration ranges in Ba (252-464 ppm), Rb (96-139 ppm), and Sr (6-9 ppm). The concentrations of La (20-24 ppm), Ce (41-45 ppm), and Sm (1.6-2 ppm) are more homogeneous. The single glass analysis from sample Hx14n has higher Sr (46 ppm) and Ba (588 ppm) concentrations, lower Ce (34 ppm) and Sm (1.1 ppm) than glass of sample Hx14b.

6.1. Plagioclase cores (An₈₆ to An₇₀)

      Concentrations of Fe, Ti, Sr, Ba, K, Rb, La, Ce, and Y in plagioclase cores are comparable among the two samples and among different crystals, whereas Mg and Li are highly variable. Concentrations of Ti, Sr, Ba, La, Ce, and Li tend to increase towards the margins of cores (Figs. 4 and 5). Iron concentrations stay more or less constant, whereas K increases in some crystals (Pl4n, Pl7n and Pl4b) and decreases in others (Pl2n and Pl5b). The Mg concentration always decreases. The concentration profiles of Rb and Y (not shown) do not display well-defined enrichment or depletion trends, probably due to the large analytical errors (25-30%). The increasing Ti, Sr, Ba, K, La, and Ce concentrations coupled with decreasing Mg (although see Discussion) can be explained by the crystallization of a mineral assemblage consisting of olivine, clinopyroxene, and anorthitic plagioclase (An_85-70) from a basaltic magma. The concentrations of Mg, Ti, Sr, Ba, and K of plagioclase cores are similar to those measured for the anorthite-rich xenocrysts (An_82-76) in the host dacite (Fig. 6), which were also inferred to have crystallized from basaltic magmas (Singer et al., 1995).

6.2. Transition zones between cores and rims (An₇₀ to An₄₅)

      Large and abrupt changes in Fe, Ti, Sr, Ba, K, La, Ce, and Y concentrations occur at the boundaries between cores and rims, and mimic abrupt changes in major elements. Rims have higher concentrations of Sr, Ba, K, La, Ce and lower concentrations of Fe, Ti, and Y than cores (Figs. 4 and 5). Enrichment factors between rims (or transition zone) and cores are 1.5-2.1 for Sr, 2.5-4.8 for Ba, 2-4.5 for K, 2.2-7.3 for La, and 1.9-5.5 for Ce, and depletion factors are 1.2-2.6 for Fe, and 1.5-10.6 for Ti. The Li and Rb concentrations do not change significantly between cores and rims. It is worth noting that the Mg, Ti, Sr, Ba, and K concentrations in the plagioclase phenocrysts (An_64-46; Singer et al., 1995) of the host dacite lava are between those of plagioclase cores and rims, and partly overlap with compositions of the transition zone (Fig. 6). This suggests that plagioclase with compositions transitional between cores and rims crystallized from melts of a composition comparable to that of the host dacite. The large Ti and Fe concentration differences between cores and rims might reflect concomitant crystallization of hornblende and orthopyroxene, as no Fe-Ti oxides are present in the xenoliths.

6.3. Plagioclase rims (<An₄₅)

      The Mg, Fe, Ti, La, and Ce concentrations of plagioclase rims remain more or less constant from the inner to outer rim of all crystals (Fig. 4). The concentration profiles of Sr, Ba, and K are more complex. Plagioclase crystal Pl2n shows two Ba and Sr concentrations lows associated with two K concentration peaks, which in turn are related to small (<10 mol %) anorthite shifts. In general, however, Sr and Ba concentrations tend to decrease towards the outer part of the rim, whereas K concentrations increase. The decreasing rim-ward Sr and Ba concentrations can be explained, to a first approximation, by crystallization of a mineral assemblage consisting of low anorthite plagioclase (e.g., <An₄₅) and phlogopite.

      

Figure 4. Anorthite mol % and trace element profiles of selected plagioclase crystals of sample Hx14n.

      Gray band indicates the plagioclase composition of the transition zone. The size of the symbol is similar to or larger than the 2-s precision of all elements.

      

Figure 5. Anorthite mol % and trace element profiles of two plagioclase crystals of sample Hx14b.

      Gray band indicates the plagioclase composition of the transition zone. The size of the symbol is similar to or larger than the 2-s precision of all elements.

      The higher K content of the plagioclase rims compared to the plagioclase phenocrysts of the host dacite suggests that plagioclase rims crystallized from more evolved melts than the dacite (Fig. 6).

      

Figure 6. Variation diagrams of trace element composition of plagioclase.

      Also shown are the compositions of the plagioclase phenocryst and xenocryst of the dacite lava in which the xenoliths were found (excluding four outermost rim analyses of the plagioclase phenocryst, see Singer et al., 1995). The size of the symbol is similar to or larger than the 2-s precision of all elements. See text for discussion.

7.1. Origin of the plagioclase compositional gap

      The An mol % of plagioclase depends not only on magma composition but also on its temperature, pressure, and water content (e.g., Yoder, 1969; Housh & Luhr, 1991; Nelson & Montana, 1992; Sisson & Grove, 1993). Specifically, significant changes in the water pressure can induce shifts in the plagioclase of up to 20 An mol % (e.g., Panjasawatwong et al., 1995). However, the coupled major and trace element variations displayed by the plagioclases in these xenoliths suggest that the large anorthite shift is mainly due to an abrupt change in the composition of the melt rather than to changes in the water content.

      A mechanism capable of producing large and rapid compositional changes in calc-alkaline magmas was proposed by Grove & Donnelly-Nolan (1986) in a study of silicic lavas, andesite inclusions and gabbroic xenoliths of Medicine Lake Volcano. Specifically, they suggested that closed-system peritectic reactions that involve liquid, olivine and clinopyroxene as reactants, and orthopyroxene and hornblende as products could lead to large compositional changes in the melt. Due to the shallow slope in temperature-composition diagrams of such reactions, sudden increases in the crystallinity of the magma and large changes in the liquid composition can occur over small temperature intervals (Grove & Donnelly-Nolan, 1986). As the San Pedro xenoliths display the same reactions involving hornblende and orthopyroxene (but also phlogopite) to the Medicine Lake gabbroic xenoliths, it seems plausible that a brusque and large amount of hornblende crystallization could produce rapid and major compositional changes in the liquid (particularly lowering the Ca/Na). However, Brophy et al. (1996) investigated plagioclase zoning patterns in several Medicine Lake gabbroic xenoliths and found smooth, monotonic trends of decreasing An mol % (An_74-30) instead of the abrupt zoning patterns that would be expected from a sudden crystallization of hornblende and orthopyroxene. The plagioclase trace element traverses also display smooth zoning profiles, with the onset of hornblende and orthopyroxene (and Fe-Ti oxides) crystallization being marked by a change in the slope of the Fe and Ti plagioclase concentrations rather than by an abrupt change. Brophy et al. (1996) interpreted these profiles as a record of progressive closed-system crystallization from andesitic to rhyolitic compositions. We take this as evidence that simple, closed-system crystallization involving orthopyroxene and hornblende does not explain the extreme and abrupt plagioclase zoning present in the San Pedro xenoliths, and hence another interpretation is required.

      In Chapter II, we suggested that the mafic intercumulus liquid in equilibrium with olivine, clinopyroxene, Cr-spinel, and plagioclase cores, was displaced through porous flow by a dacitic liquid that reacted with the mafic cumulus minerals to produce post-cumulus orthopyroxene, hornblende, and phlogopite. In such a scenario, the abrupt and large compositional gap displayed by the plagioclase can be explained by: (1) the much lower Ca/Na values of the dacitic liquid compared to the displaced mafic liquid, and (2) the crystallization of hornblende which further decreased the Ca/Na values of the interstitial melt from which plagioclase rims crystallized. The abrupt major element shift from cores to rims, and the minimal occurrence of plagioclase of intermediate compositions (e.g., An₆₀), support the idea that melt displacement occurred as a wave or a pulse without significant mixing between the two melts, consistent with their presumably large viscosity difference.

      The trace element concentration differences between cores and rims provide additional support for the melt migration event. Enrichment factors between cores and rims range from 2.2 for K, to 7.3 for La, and depletion factors are as large as 10.6 for Ti (Fig. 6). Such large compositional changes are difficult to reconcile with closed-system evolution. To further investigate the proposed crystallization sequence and the origin of the plagioclase compositional gap, in the following section we have calculated the liquids in equilibrium with the plagioclase composition.

7.2. Some considerations prior to reconstruction the liquid composition in equilibrium with plagioclase

      Using the melt composition in equilibrium with plagioclase to model magmatic processes has been successfully applied by several authors to trace magmatic processes (Blundy & Shimizu, 1991; Singer et al., 1995; Brophy et al., 1996; Papike et al., 1997; Bindeman et al., 1999). Nevertheless, the validity of such approach relies on several assumptions: (1) diffusion has not modified the plagioclase composition after it crystallized, (2) appropriate plagioclase-melt partition coefficients are available, and (3) plagioclase crystallized under local equilibrium conditions. The first two assumptions are discussed below. The third can only be addressed by calculating the equilibrium melts.

---

---

7.3. Composition of the calculated melts

      In general, calculated melts in equilibrium with plagioclase show progressive compositional variations from the inner to outer cores and from the inner to rims (not shown). This means that the geochemical trends displayed by the calculated melts in variation diagrams can be roughly interpreted as reflecting the successive compositional changes that occurred in the liquid. The compositional variations of the calculated concentrations of MgO, TiO₂, K₂O, Sr, Ba, and La of both samples are displayed in Figures 7 and 8, together with the composition of the basaltic to dacitic San Pedro lavas, and the calculated melts in equilibrium with hornblende (partition coefficients for dacitic melts of Sisson, 1994). Melts in equilibrium with plagioclase from both samples show the same compositional trends, although those displayed by sample Hx14b (Fig. 8) are somewhat more scattered than those of sample Hx14n (Fig. 7).

      The trends displayed by the calculated melts in equilibrium with plagioclase show the same events that have been discussed up to now:

1. melts in equilibrium with plagioclase cores display increasing concentrations of TiO₂, Sr, Ba, K₂O, La, and Ce with decreasing MgO contents and mimic the trends displayed by the basaltic to basaltic andesitic lavas, which is consistent with the crystallization of olivine+clinopyroxene+anorthitic plagioclase (Figs. 7 and 8). The calculated TiO₂ concentrations also increase with Sr concentrations, although it is not clear why the trends are different from those displayed by the basaltic to basaltic andesitic lavas.
2. melts in equilibrium with the core-rim transition zones display dramatic variations, comparable to the range of basaltic to andesitic lava whole-rock compositions.
3. melts in equilibrium with plagioclase rims have lower TiO₂, FeO* (not shown) and Sr, and higher K, Ba, and La than melts in equilibrium with plagioclase cores. Such compositional differences can be explained primarily by the crystallization of hornblende (± orthopyroxene and phlogopite). The calculated melts in equilibrium with hornblende show TiO₂, Sr, La, and Ce concentrations that range from those of the basaltic andesitic to dacitic lavas, and further support the hypothesis that hornblende crystallization began just prior to crystallization of the plagioclase rims. Ideally, one could use the calculated melts to perform quantitative modeling and determine if the abrupt compositional changes between cores and rims are due to closed system crystallization of hornblende+orthopyroxene +phlogopite, or if these minerals crystallized after the melt migration and reaction hypothesis discussed above. Unfortunately, the uncertainties in the partition coefficients preclude such an approach (see discussion below).
4. melts in equilibrium with plagioclase rims display decreasing Sr and Ba concentrations and increasing K/Ba and K₂O concentrations, whereas TiO₂, MgO, La, and Ce abundances remain more or less constant (Figs. 7 and 8). These trends are not displayed by the San Pedro lavas, suggesting that rims have recorded the evolution from dacitic to rhyolitic liquid compositions through crystallization of a mineral assemblage dominated by phlogopite and anorthite-poor plagioclase (e.g., <An₄₀). The fact that La and Ce concentrations remain more or less constant could be explained by the crystallization of apatite (present in the glass) or by the high partition coefficients for La and Ce that have been reported for some biotites and may also apply to phlogopite (e.g., Mahood & Hildreth, 1983; Nash & Crecraft, 1985).

      

Figure 7. Variation diagrams of calculated compositions of the melts in equilibrium with plagioclase of sample Hx14n.

      Also shown are the calculated compositions of the melts in equilibrium with Hbl and the composition of the interstitial glass. The dashed line and arrow indicate the inferred differentiation trend of Volcán San Pedro lavas. The 2-s precision of the calculated melts is shown by horizontal and vertical bars for two points in each variation diagram. See text for discussion.

      

Figure 8. Variation diagrams of calculated compositions of the melts in equilibrium with plagioclase of sample Hx14b.

      Also shown are the calculated compositions of the melts in equilibrium with hornblende and the composition of the interstitial glass. The 2-s precision of the calculated melts is shown by horizontal and vertical bars for two points in each variation diagram. The dashed line and arrow indicate the inferred differentiation trend of Volcán San Pedro lavas. See text for discussion.

      Although melts in equilibrium with plagioclase rims display trends in accord with a fractionation from dacitic to rhyolitic melt, the absolute calculated concentrations are unlike those of the interstitial rhyolitic glass and unlike those of natural magmas found at the TSPC. To further illustrate this (Fig. 9) we have compared the calculated compositions with the dacitic to rhyolitic compositions of TSPC lavas and of two Miocene granitoid plutons from the basement of the TSPC (Nelson et al., 1999). The calculated melts in equilibrium with plagioclase rims show similar differentiation trends to those of the plutons but the absolute calculated concentrations of Ba and K₂O are shifted by about 50 % towards lower concentrations. The calculated La and Ce concentrations are shifted by a similar amount but towards higher values (not shown). Analytical problems with the ion microprobe trace element analyses of plagioclase can be discarded (Appendix I). Thus, the discrepancy between the calculated compositions and those of natural liquids at the TSPC has to be due to other factors. We will consider the following:

1. protracted closed-system crystallization may have led to individual pockets of melt that evolved independently. If this were the case, the composition recorded by every plagioclase crystal should be different and would reflect the composition of the local environment. The relatively minor differences between the major and trace element zoning profiles of plagioclase crystals Pl2n and Pl7n argues against the existence of pockets of melt that evolved in vastly different ways.
2. crystallization kinetics may result in partitioning relations quite different from the equilibrium case if the rate of crystal growth is much faster than chemical diffusion in the melt. Fast crystal growth results in apparent partition coefficients that are higher than equilibrium values for elements with D<<1, and lower than equilibrium values for elements with D>>1 (see equation 5 of Smith et al., 1955). The low La and Ce diffusivities in silicate melts (e.g., Brady, 1995) together with their small plagioclase-melt partition coefficients (D_La = 0.19 and D_Ce = 0.15, at 950^oC and An₄₀; Bindeman et al., 1998), would suggest that their high calculated concentrations can be explained at least in part by high plagioclase growth rates. If this were the case, similar effects should be observed in the behavior of K₂O, as D_K (0.13; at 950^oC and An₄₀; Bindeman et al., 1998) is smaller than D_La and D_Ce. As the calculated K₂O concentrations are low when compared to dacitic and rhyolitic compositions, high plagioclase growth rates seems not to be the explanation for the anomalous calculated compositions.
3. the partition coefficients derived from the empirical equations of Blundy & Wood (1991) and Bindeman et al. (1998) might not be appropriate for the evolved melt and plagioclase compositions considered here. To evaluate this we have plotted the calculated liquid compositions in equilibrium with the plagioclase phenocrysts of the host dacite lava (Fig. 9). The calculated concentrations of Ba and Sr in equilibrium with the dacite plagioclase phenocryst are comparable to those of the bulk-rock dacites of San Pedro, whereas the calculated K₂O concentration is ~ 50 % lower. This would suggest that D_Sr and D_Ba are appropriate for plagioclase-melt compositions studied here, whereas D_K is not. The D_K (0.13 ± 0.02) should be about 50 % smaller, and D_La (0.19 ± 0.08) and D_Ce (0.15 ± 0.04) should be about 50 % larger for the composition of the calculated melts in equilibrium with plagioclase rims to lie between those of the San Pedro dacite and the rhyolitic glass. Such partition coefficients have been reported in the literature, and it is apparent from the compilations of plagioclase-melt partition coefficients of Bindeman et al. (1998) that at ~ An₄₀, estimates of D_K vary from 0.08 to 0.24, D_La from 0.24 to 0.5, and D_Ce 0.22 to 0.3. Thus, the discrepancy between the calculated compositions in equilibrium with plagioclase rims and those of dacitic to rhyolitic natural liquids (including xenoliths interstitial glass) found in TSPC could be partly explained if the partition coefficients of Bindeman et al. (1998) were not suited for calculating plagioclase-melt partition coefficients in water-rich silicic systems.
4. As the elements that show low calculated concentrations are those that are highly compatible in phlogopite (K, Ba, and Rb), it seems plausible that during melt percolation through the crystal-network, phlogopite crystallization kept the K, Ba, and Rb concentrations low, and once phlogopite crystallization stopped (and melt migration continued) the liquids returned to more "usual" compositions. In this scenario, the composition of the interstitial glass would not be the product of protracted crystallization but would represent the composition comparable to that of the migrating liquid.

      

Figure 9. Variation diagrams of calculated compositions of the melts in equilibrium with plagioclase of both samples.

      The compositions of dacitic to rhyolitic lavas and plutons (triangles) of the TSPC display an inferred differentiation trend indicated by the dashed line and arrow. Squares are the calculated composition in equilibrium with plagioclase phenocrysts of the host dacite lava; the other symbols are the same as in Figs. 7 and 8. See text for discussion.

Appendix I. Analytical methods

      Electron microprobe analyses were done on a Cameca SX50 at the University of Lausanne, with wavelength dispersive spectrometry. Hornblende, plagioclase and glass were analyzed with an accelerating voltage of 15kV and beam currents of 15nA for hornblende and plagioclase, and 7 nA for glass. Beam diameter was ~ 2µm for plagioclase and hornblende, whereas for glass large areas were rastered when possible (~ 25µm²).

      Ion-microprobe analyses were performed at the University of Edinburgh using a Cameca IMS-4f ion microprobe with a O- primary beam of net impact energy of 15 keV (10.7 keV primary and 4.5 keV secondary) and an operating current of 5-8 nA. The secondary ion accelerating voltage was 4500 keV. Only ions with energies falling between 55 and 95 eV (75+/-20 eV) were analyzed to reduce molecular ion interferences. Beam diameter ranged between 15 to 30 mm. For plagioclase and glass, the following isotopes were measured: ⁷Li, ²⁶Mg, ³⁰Si, ⁴¹K, ⁴²Ca, ⁴⁷Ti, ⁵⁶Fe, ⁸⁵Rb, ⁸⁸Sr, ⁸⁹Y, ¹³⁸Ba, ¹³⁹La, ¹⁴⁰Ce, plus mass 60 for CaO overlap on Fe, and mass 130.5 to monitor the background. For hornblende the following isotopes were measured: ⁷Li, ³⁰Si, ⁴¹K, ⁴⁷Ti, ⁸⁵Rb, ⁸⁸Sr, ⁸⁹Y, ⁹³Nb, ¹³⁸Ba, ¹³⁹La, ¹⁴⁰Ce, ¹⁴¹Pr, ¹⁴³Nd, ¹⁴⁹Sm, ¹⁵¹Eu, ¹⁵⁶Gd, ¹⁵⁷Gd, ¹⁵⁹Tb, ¹⁶¹Dy, ¹⁶⁵Ho, ¹⁶⁷Er, ¹⁷²Yb, plus mass 130.5 to monitor the background, and mass 154 for BaO overlap on Eu. Mass 156 is a combination of CeO and minor ¹⁵⁶Gd and was used to calculate the overlap of the light REE on the heavy REE by assuming that REEO/REE follows CeO/Ce. The overlap of Si₂ on Fe could not be corrected directly but was calculated using Fe-free systems to be equal to 229 ppm on all feldspars (R. Hinton, personal communication). The concentrations were obtained by normalization of mass ³⁰Si to the silica values obtained by electron microprobe analyses. The SRM-610 glass standard was used as a monitor of day to day changes instrumental conditions (the nominal concentrations for SRM 610 are given in Hinton, 1990). The ion yields for glass and plagioclase relative to SRM-610 were determined using Lake County (LC) plagioclase, an alkali feldspar SHF1 (AF), and Corning glass (CG). Corning glass has similar ion yields to feldspars for elements such as K, Sr, and Ba and is assumed to give similar ion yields for those elements not available in the feldspar standards. The correction factors are as follows: 0.921 for Li (LC), 0.881 for Mg (LC), 0.768 for K (LC), 0.788 for Ca (LC), 0.793 for Ti (LC), 0.76 for Fe (CG), 0.768 for Rb (LC), 0.796 for Sr (LC), 0.796 for Y (LC), 0.773 for Ba (LC), 0.858 for La and Ce (AF). No ion-yield corrections were done for hornblende, because the values obtained using SRM 610 give good agreement with bulk analyses (R. Hinton, personal communication). The 2-s precision for most elements is under 5 % relative. Only La and Ce are in some cases slightly above 10 %, and Y and Rb are around 25 %.

      Comparison of the electron and ion-microprobe concentrations of MgO, K₂O and BaO agree within 10%, whereas FeO* determined by the electron-microprobe is ~ 20% higher than determined with the ion-microprobe. Sanidine analyses from O. Bachmann. (see the following figure).

      

Appendix II. Error of the plagioclase-melt partition coefficients used

      The error associated with the partition coefficients determined by using equations of Blundy & Wood (1991) and Bindeman et al. (1998) was derived through error propagation (assuming independent errors; Taylor, 1982) of the uncertainties in their regressed parameters only. We did not take into account the error in An mol % determination from the electron microprobe analyses, the error in the trace element composition of plagioclase from ion-microprobe analyses, or the error in the temperature estimations from geothermometers. The derived equation is:

      |D| = {{[[X_An exp((a X_An + b )/ RT)]/RT] |a|}^2 +

      {[[exp((a X_An + b )/ RT)]/ RT] |b|}^2}^0.5

      Where |D| is the absolute error on the partition coefficient, |a| and |b| are the absolute errors on the regressed values of a and b, and the rest of the symbols are the same as for equation (1).

3.1. Group I. Olivine-hornblende norites.

      The samples consist of Cr-spinel, olivine, minor clinopyroxene, orthopyroxene, hornblende, plagioclase, and phlogopite forming a medium grained (1-5 mm) crystal network with interstitial vesiculated SiO₂-rich (67-72 wt %) glass filling the interstices (0.5 13 vol. %). Glass is distributed in isolated and occasionally interconnected pockets bounded by euhedral crystal faces suggesting that it is residual from crystallization and not due to partial melting. Phlogopite makes up to 8 vol. % and is euhedral to subhedral. Euhedral phlogopite occurs in the interstitial glass attesting to a magmatic origin and indicating that it was among the last minerals to crystallize (Fig. 1). Phlogopite is almost invariably present surrounding resorbed olivine and Cr-spinel (Fig. 1), indicating that a reaction between these minerals and an evolved water-rich liquid occurred.

3.2. Group II. Hornblende norites and clinopyroxene leuconorites with subsolidus textures.

      Clinopyroxene leuconorites display seriate, mosaic textures with abundant plagioclase (up to 80 vol. %), olivine, minor Cr-spinel, clinopyroxene, hornblende, Fe-Ti oxides and phlogopite. Exsolution lamellae in pyroxenes, bent twins in plagioclase, and the presence of microfractures record subsolidus cooling and deformation. Phlogopite (up to 15 vol. %) occurs as small (< 1 mm) anhedral crystals surrounding resorbed plagioclase, orthopyroxene, and Fe-Ti oxides (Fig. 1). Phlogopite also occurs filling discontinuous microfractures (< 0.5 mm in width) together with hornblende, orthopyroxene, and Fe-Ti oxides (Fig. 1).

      Hornblende norites consist of large anhedral oikocrysts (> 1 cm across) of hornblende that enclose resorbed, anhedral olivine, Cr-spinel, clinopyroxene, orthopyroxene, plagioclase. Phlogopite makes up to 4 vol. % and is commonly subhedral to anhedral. It occurs inside hornblende crystals, surrounding resorbed olivine, Cr-spinel, orthopyroxene, and in contrast with the Group I xenoliths, it is also present surrounding resorbed plagioclase (Fig. 1).

4.1. Group I xenoliths. Olivine-hornblende norites.

      Phlogopites have mg-numbers (mg-number = 100*Mg/(Mg+Fe^t) in mols; Fe^t= total iron) ranging from 82 to 77, and Cr₂O₃ concentrations from <0.05 to 0.43 wt %. The Na₂O contents are between 1.5 and 3.5 wt %, and Na/K apfu (atoms per formula unit) values vary between 0.17 and 0.94 (Fig.2). The occupancy of the A-site ranges 0.9 to 1.0, which differentiates these phlogopites from the interlayer-deficient mica wonesite (Spear et al., 1981). The major element composition (e.g., MgO) of the phlogopites is comparable among different crystals, but minor elements (particularly Cr and Ba) are highly variable (Fig. 2). We interpret the presence of phlogopite with high Cr₂O₃ contents and mg-numbers as due to reactions between olivine + Cr-spinel and an evolved water-bearing liquid. Volfinger et al. (1985) suggested that high Mg/Fe^t in mica enhances substitution of Na for K. The composition of phlogopites in San Pedro xenoliths show a broadly positive correlation between the mg-number and the Na/K values (Fig. 2), and thus it seems plausible that high Na/K of the phlogopites is partly due to their high mg-numbers.

      

Figure 2. Variation diagrams of the phlogopite compositions from the xenoliths and of the biotites from the host dacite lava.

4.2. Group IICL xenoliths. Clinopyroxene leuconorites.

      Poikilitic phlogopites have mg-numbers that vary from 81-70 and Cr₂O₃ contents are typically < 0.2 wt %. Their Na₂O contents (1.1-2.1 wt %) and Na/K (0.2-0.4) are lower than those of phlogopites from Group I xenoliths (Fig. 2). Phlogopites filling microfractures have lower mg-numbers (76-70), lower Cr₂O₃ (< 0.15 wt %) and Na₂O (0.7-1.3 wt%) contents, and lower Na/K (~ 0.2) than poikilitic phlogopites. The structural formulae (Table 2) shows that some poikilitic phlogopite crystals have the tetrahedral position partly filled with Fe³⁺, whereas phlogopites filling microfractures and biotites of the dacite host lava have invariably Fe³⁺ in the tetrahedral position. Oxygen fugacity estimates of the host dacite using ilmenite and magnetite pairs are close to NNO+2 (B.S. Singer, unpublished data) which is in support with the observation of Briggati et al. (1996) that Fe³⁺ in the tetrahedral position reflects crystallization at low pressure and high fO₂ conditions. We propose that phlogopites filling microfractures crystallized from oxidized, evolved water-rich liquids comparable to the host dacite. The higher mg-number and Cr₂O₃ contents of the poikilitic phlogopites compared to those filling microfractures are interpreted to be the result of reactions between mafic minerals (e.g., olivine, Cr-spinel, and pyroxenes) and evolved melts that migrated through the microfractures.

      

Figure 3. X-ray map of Na and K distribution of the phlogopite with up to 5 wt % Na₂O.

      Na and K are distributed as bands approximately parallel to the cleavage. The mapped area is marked in Figure 1c. The dashed line marks the position of the electron microprobe traverse shown in Figure 4.

4.3. Group IIHN xenoliths. Hornblende norites.

      Phlogopites have mg-numbers ranging from 84-77 and Cr₂O₃ contents up to 0.2 wt %. Most phlogopites have Na₂O contents between 1 and 2.5 wt %, except for one poikilitic crystal (3 mm across; Fig. 1c) that has up to 5 wt % Na₂O, and Na/K mol values up to 2.2 (Fig. 2). An X-ray map of this phlogopite crystal (Fig. 3) shows that the sodium concentration is not homogeneous. The crystal consists of Na-rich bands (< 50 µm) that are parallel to the phlogopite cleavage (approximately parallel to the 001 plane) and which alternate with K-poor zones. An electron microprobe traverse across the Na-rich zone of this phlogopite crystal (Fig. 4) shows that variations of up to 2 wt % of Na₂O occur in less than 10 µm, suggesting that even the Na-rich zones are heterogeneous and probably consists of fine (< 1-2 µm) intergrowths of Na-rich phlogopite and phlogopite. Such abrupt changes in the phlogopite Na₂O and K₂O concentrations are difficult to explain by changes in the liquid composition alone, and they probably reflect the presence of a solvus.

      

Figure 4. Traverse across a Na-rich zone of the phlogopite crystal shown in Figure 3.

      The Na-rich zone is complementary with K-poor zones. Within the Na-rich zone variations of up to 2 wt % of Na₂O occurs, suggesting that it consists of mixtures of almost pure sodium phlogopite and phlogopite. Note that the spacing of the electron microprobe analyses in the Na-rich zone is 5 times smaller (2 µm) than for the rest of the traverse.

5.1. Genesis of the Na-rich phlogopites

      The presence and composition (e.g., high Cr₂O₃ and mg-number) of the phlogopites in the San Pedro gabbroic xenoliths can be explained by reactions between evolved water-rich liquids and olivine, pyroxenes, and Cr-spinel. We propose that these reactions were not due to closed-system crystallization because experiments on medium-K calc-alkaline water-rich liquids (e.g., Sisson & Grove, 1993) have not reported Na-rich phlogopite, or phlogopite even at crystallinities of up to 90 % (Kawamoto, 1996). Moreover, the microfractures filled with phlogopite in Group II xenoliths are evidence of open-system processes involving migration of evolved water-rich melts (see also Chapter II).

      To explain the high Na/K values of most phlogopites we consider below: (1) the possibility that plagioclase was involved in the phlogopite-forming reactions, (2) the composition of the reacting liquids, and (3) the potential role of a solvus.

---

---

---