Significance of Hornblende in Calc-Alkaline Andesites and Basalts

American Mineralogist, Volume 65, pages 837-851, 1980

Significance of hornblende in calc-alkaline andesites and basalts

A. T. ANDERSoN. JR.

Department af the Geophysical Sciences, The University of Chicago 5734 South Ellis Avenue Chicago, Illinois 60637

Abstract

Hornblende occursin someandesitic and basalticrocks of calc-alkalineaffini1y, where it is commonly associatedwith olivine. The texture,eruption history, and compositionare inter- pretedto indicate that hornblendecommonly forms as a product of reactionbetween olivine and basalticand andesiticliquids. The natural hornblendesprobably formed within the crust at temperaturesbetween 960o and 1080'C from liquids with lessthan about 6 weight percent HrO. The formation of hornblendefrom basalticliquid within the crust has implications for evolution of continentalcrust, as well as for the origin of andesiteand thermal conditions in subductionzones.

Introduction Definitions Boettcher (1973) and others (Mueller, 1969;Hol- In using the word liquid, I mean to refer to a single loway and Burnham, 1972;Green, 1972;Helz, 1976; state of matter (a homogeneous liquid phase). It is vi- Cawthorn and O'Hara, 1976;Allen and Boettcher, tal to distinguish between the composition of a liquid 1978)argue that amphibole plays an important role (generally a residual liquid) and the composition of in governing the compositionsof calc-alkalineande- the bulk material. Petrographically a certain horn- sites. The mechanismswhereby amphibole suppos- blende may be associated with a particular residual edly exertsits role are not clear. Is it left behind in a liquid. Experimentally a certain bulk composition crystallins residue-where is it: in the crust or in the may crystallize hornblende at or below the liquidus. mantle? Does amphibole crystallize and separate The residual liquid associated with hornblende will from a parental liquid-what are the composition generally differ in composition from the bulk. Ther- and temperatureof the parental liquid? Does amphi- modynamically the composition and temperature of bole crystallizefrom a derivative liquid which mixes a certain liquid can remain constant as the propor- with a parental liquid to yield andesite?Amphibole tions of crystals and liquid vary. To interpret a given occurs in some calc-alkaline volcanic and plutonic association of hornblende and liquid, the petrogra- rocks with basaltic and andesitic bulk compositions, pher must seek experimental conditions which yield but the significance of such occurrencesis uncertain hornblende and liquid with compositions similar to without textural evidenceof how it got there: is the those observed. If analogy is drawn with a particular amphibolea primary igneousmineral or a product of bulk composition (and liquidus hornblende) the im- alteration?If igneous,over what interval of crystalli- plied pressure (and concentration of HrO in the liq- zation did it form, and was the composition of the uid) is larger than if analogy is made with a residual liquid in that interval andesitic,dacitic, or basaltic?If liquid. The implications for temperature are complex amphibole can be shown to have actually formed and depend on the concentration of HrO in the liquid from a fiquid with a certain composition, what does as well as on compositional features such as Fe/Mg that portend for the concentrationof HrO in and the and concentrations of alkalis. temperatureof the liquid? My aim is to help point the way to answersto someof the abovequestions by Hornblende in experimental products critically sxxmining evidenceon the growth of am- Experimental studies yielding hornblende in prod- phibole from basaltic and andesiticliquids, both ex- ucts with either calc-alkaline bulk compositions or perimental and natural. with residual liquids having analogous compositions 0003-o04x/80/09I 0-0837$02.00 83'l ANDERSON:HORNBLEN DE IN CALC-A LKA LI N E A N DESITES reveal:(l) Hornblende is a liquidus or near-liquidus mineral in some basaltic bulk compositionsif P",o exceedsabout l0 kbar (Yoder and Tilley, 1962; Green and Ringwood, 1968;Allen et al., 19751'Stern et al., 1975)and if H,O dissolvedin liquid exceeds about 12 weight percent. (2) In some andesiticbulk compositionshornblende is a liquidus mineral rf Pn,o exceedsabout 2 kbar (Piwinskii, 1973).(3) As P",o 1 decreasesbelow P,o,u,,the hornblende-outcurve lies progressivelyfarther below the liquidus (Robertson T and Wyllie, l97l; Holloway and Burnham, 1972; Green, 1972;Eggler and Burnham, 1973;Allen and Boettcher, 1978).(4) The proportion and composition of residual liquid change dramatically over an interval of temperatureless than 50"C after the beginning of crystallization of hornblende (Holloway and Burnham, 1972).(5) Near the liquidus low oxygen fugacity favors the crystallization of hornblende f-t--ii:) comparedto pyroxene(Helz, 1973,1976). (6) The re- actionsbetween hornblende, liquid and other crystals 1:1a(ry22 (Bowen, 1928,p. 61, 85, lll; Helz, are complex a\*>,,/ ./'// 1976).(7) Hornblendeis stableto the highesttemper- atures and lowest concentrationsof HrO in liquid in systems(a) where olivine is present(Helz,1976), (b) where P,,o ( P,",",(Holloway, 1973),and (c) where Fig. l. Simplified hypothetical P-T-XH2o equilibrium diagram for a high-alumina basalt. The diagram is drawn in perspective. significant F is present(Holloway and Ford, 1975). The origin is not shown. The concentration of H2O refers to the The above relations are summarizedon Figure l. bulk material. The stippled and dashed regions are saturated with The coordinatesof Figure I are qualitative. Quan- vapor rich in H2O. Below the liquidus surface the pyroxene (Px)- titative values will vary with bulk composition. out and olivine (Ol)-in surfaces are assumed to coincide for Lesser MgO and Mg/(Fe+Mg) will diminish the simplicity. Fields for spinel and magnetite are omitted for simplicity. Hornblende (Hb) is a liquidus mineral in both vapor- temperature of olivine, pyroxene, and hornblende absent and vapor-present regions. The liquidus hornblende is liquidus surfaces.The field for liquidus hornblende accompanied either by liquidus Ol or by both plagioclase (Pl) and will then extendfarther into the vapor-absentportion Ol. Only one symbol is shown for pyroxene (Px). Liquidus Px is of the diagram. Lesserconcentrations of AlrO. will calcium-rich. Low-calcium pyroxen€ probably predominates if Hb shift more of the liquidus surfaceinto the olivine and is abundant. pyroxenefields and diminish the area of the liquidus field for plagioclase.Consequently, the combination simplifications,the diagram is consistentwith exist- of low Al,O, and high MgO may obliterate the liq- ing experimentaland natural data. uidus surfacehaving hornblende,olivine, and plagio- It would be helpful to know the positions of sur- clase(see Bowen, 1928, p. 1l l). For a naturalmagma facesof equal SiO, in the liquid (H,O-free basis).At with CO, as well as HrO, there will be a rangein X",o presentthe coordinatesof such surfacesin Figure I in addition to a boundary marking vapor-absentand are uncertain. Qualitatively, it is possible to give vapor-present regions. Reactions between olivine, somelimits. I considera surfaceof SiO, : 60 weight pyroxene,hornblende, and liquid add complications percent on an anhydrous basis, corresponding not shown on the diagram. Spinel and magnetiteare roughly to andesite.Because the compositionof the probable additional liquidus minerals but are not residual liquid changessteeply with temperatureaf- shown.For simplicity only one pyroxeneis depicted. ter the beginning of crystallization of hornblende Calcium-rich clinopyroxene is the liquidus pyroxene (Holloway and Burnhan, 1972),it is reasonableto but is joined atd/or replacedby orthopyroxenebe- expect the 60 percent SiO, surface to lie close to the low the liquidus at temperaturesbelow or near that hornblende-outsurface. In the region ofthe diagram of the appearanceof hornblende. Except for such where hornblendeis a liquidus mineral (hornblende- ANDERSON: HORNBLENDE IN CALC-ALKALINE ANDESITES out corresponds to liquidus), the 60 perc€nt SiO, sur- by either contrastinginitial concentrationsof HrO in face must lie below the hornblende-out surface. In parental high-alumina basaltic magma, variable the region ofthe diagram where the hornblende-out depth of crystallization,or extent of reaction during surface is far below the liquidus (for example, sub- crystallization. aerial basalts with about 0.1 weight percent HrO), re- Some natural calc-alkahneliquids associatedwith sidual liquids with more than 60 percent SiO, may hornblende are less siliceous and hotter than most develop in the absence ofhornblende. Therefore, the experimental analogs.The natural occurreneessug- 60 percent SiO, surface may intqisect the horn- gest that the field of liquidus hornblendeextends to blende-out surface, unless crystallization intervenes. lower concentrationsand partial pressuresof HrO It is probable, but uncertain, that residual liquid in than is indicated by most existing experimentalre- anhydrous high-alumina basalt attains 60 percent sults. SiO, before complete crystallization if equilibrium is maintained. Residual glassesin natural high-alumina Hornblendeassociated with natural andesiticliquid basalts have 62 (Fuego volcano, Guatemala-Rose e/ Many andesiteshave phenocrystsof hornblende al., 1978) and 67 (Hat Creek flow, California-An- (see for example,Moorhouse, 1959, p. 189-190; derson and Gottfried, 19'71)percent SiOr. However, Turner and Verhoogen,1960, p.272-282) but most equilibrium is not maintained throughout the natural do not. Many, perhapsmost, andesitescontain clots rocks. Because the minerals (An55), aug- [plagioclase of anhydrous crystalswhich may be reaction prod- ite, olivine (Fo66), and magnetitel associated with ucts of hornblende (Stewart, 1975);however, the oc- the residual glass at Fuego are compositionally simi- currence of oscillatory zoned plagioclaseswithin lar to the'normative minerals (An50), [plagioclase such clots suggestsa different origin (Garcia and Ja- pyroxene, olivine (Fo70), magnetite, and ilmenite, cobson,1979). based on the weighted average composition with l/3 Most hornblende andesiteshave more than about of the iron trivalent], I conclude that anhydrous, 60 percentSiO, (Kuno, 1950;JakEs and White, 1972; calc-alkaline, high-alumina olivine tholeiites ex- Sakuyama,1979); the groundmassprobably is dacitic emplified by Fuego probably yield equilibrium resid- to rhyolitic in composition. The hornblendescom- ual liquids with 60 percent or more SiO, before commonly are rimmed or cored by crystalsof pyroxene, plete crystallization. Therefore the equilibrium plagioclase,and other minerals.The compositionsof surface for residual liquids with 60 percent SiO, the hornblendesmay be modffied after crystalliza- probably intersects the hornblende-out surface. tion. Thus the compositionalrelations between horn- The intersection of the 60 percent SiO, and hornblende and andesiticliquid are difficult to infer from blende-out surfaces may have major petrological sig- most occurrences. nificance. In a given bulk composi$on residual liq- Below I describean associationofhornblende and uids with 60 percent SiO, will vary compositionally andesitic glass included in olivine from a lapillus of in accordance with the amounts of associated miner- porphyritic andesite. als, particularly hornblende. Many workers have pointed out that crystallization of hornblende can Asama Volcano,Japan. Back- yield calc-alkaline residual liquids (Green and Ring- The 1783eruption of products wood, 1968; Boettcher, 1973; Holloway and Burn- ground:the eruptionand its ham, 1972; Heh, 1976; Cawthorn and O'Hara, 1976). Asama volcano is a composite andesitic strato- Andesitic residual liquids not associated with horn- volcano in central Honshu, Japan. Its geology is blende may be alkalic or alkali-calcic. The residual documentedby Aramaki (1963) and briefly surnma- liquid with 60 percent SiO, may have both a smaller rued by Anderson (1979). ratio of Fel(Mg+Fe) and a lower temperature if The history of the 1783 eruption helps constrain hornblende is present than if it is absent. For a given the history of cooling and decompressionwhich af- bulk composition the amount of hornblende associ- fectedthe extrudedproducts. Although complex,this ated with a residual liquid having 60 percent SiO, eruption probably is typical of the principal way will be maximized: (l) if the initial concentration of Asama has grown. It involved a successionof Plinian HrO is large; (2) if crystallization occurs at constant explosionswhich produced about 0.17 km3 of pyro- or increasing pressure; and (3) if equilibrium (reac- clastic material (Minakami, 1942) and an equal vol- tion) is maintained. Consequently, many differences ume of lava (Aramaki, 1956). The early products between various andesitic liquids may be explained were air-fall pumice lapilli and ash; ash flows formed ANDERSON: HORNBLENDE IN CALC-ALKALINE ANDESITES on and after August 3 when the culminating blast oc- curred; an andesiticlava flow formed during the final phase of the eruption. The details are reported by Aramaki (1956, 1957),who interpreted the eruption sequencein terms of an initial vent-clearing phase betweenJune 25 and August 2, 1783followed b! extrusion from a subvolcanicbody of magma on and after August3, 1783. Petrography. The 1783 andesitic pumice lumps vary from light to medium brown; somehave streaks of different shadesof brown. Lapillus AM2-2 (An- derson, 1979) is medium brown. It was collected from a layer of air-fall lapilli which probably formed on or before August 2. The pumice contains mm-sized phenocrystsof plagioclase,augite, hypersthene,magnetite, and ilmenite. Plagioclaseis the principal phenocryst and makesup about 20 volume percentof the pumice on a void-free basis. Most of the plagioclasesare subhedral to anhedralcrystals with irregular coressieved with vermicular inclusions of glass and gas. The coresare surroundedby a normally-zonedrim which Fig. 2. Inclusion of glass (G), hornblende (H), and probable is free of inclusions and l0 to 50 pm thick. Other pyroxene (P) in olivine (O) The long dimension of the plagioclasesare poor in inclusions and oscillatory- photograph is about 200 pm. Dip and strike symbols refer to the attitude of the contact between olivine and glass referred to the zoned with skeletalprotrusions. The augitescontain polished surface as horizontal The olivine crystal is rimmed by inclusionsof glassand lamellaeof pyroxenein irreg- orthopyroxene and is from a lump of pumice extruded from ular, pale green cores. Some augites occur in clots Asama volcano, probably on or before August 2, 1783. where they are subhedraland irregular againsteach other. Commonly a zone of inclusionsof plagioclase analysisyields an estimated4 percent of HrO in the and pyroxene marks the boundary between a pale glass(Anderson, 1979). greencore and deepergreen rim. The rims are 50 to A second, smaller inclusion in the same olivine about 150 pm thick. Glassesincluded in pyroxene, crystal contains a pale brown crystal which is prob- plagioclase,and oxide phenocrystsrange from about ably hornblende (Fig. 3). It comprisesabout 2 vol- 70 to 79 weight percentSiOr; the hostglass outside of ume percent of the inclusion. Anisotropic and the phenocrystshas abont 72 percent SiO, (Ander- opaque crystals line the inside wall of the vapor son,1979). bubble in the small inclusion. Crystals of olivine (Fo79-Fo85) are rare. The oli- The otvine host crystal of the hornblende-bearing vines have 5O-pm-thickrims of hyperstheneor aug- inclusion is Fo83 (Table l) compared to Fo79 to ite. There is more olivine in medium-brown pumice Fo85 for olivines from other lumps of pumice of the than in light brown pumice. Inclusionsof glassin oli- 1783eruption (Anderson, 1979).The compositionof vine crystalshave 54 to 6l weight percentSiO, on an the preservedglass (Table l, column 2) is closeto but anhydrousbasis (Anderson, 1979). more silicic than that of other inclusions in olivine Hornblende associatedwith andesitic glass. Horn- which lack hornblende (Table l, column 4). blende occurs together with andesiticglass included Hornblende comprisesabout 30 volume percentof in an olivine crystal from the lapillus AM2-2 of me- the large inclusion (Fig. 2), some of which was re- dium-brown andesiticpumice (Fig. 2, Table 1). The moved by sectioning.Probable orthopyroxenecom- hornblende is pale brown and faceted against glass. prises about l0 volume percent. Most inclusions in A secondmineral (probably orthopyroxene)occurs Asama olivines are subsphericalto ellipsoidal. The with hornblendein the inclusion. It is coloiless,long hornblende-bearing inclusion has walls which dip and prismatic,and has a refractiveindex closeto that away from the inclusionsat high angles(Fig. 2). of the hornblende. The included glasshas a round Attached phenocrystsof magnetite and ilmenite contactwith the otvine host crystal.The microprobe occur in the same lump of pumice. Their composi- ANDERSON: HORNBLENDE IN CALC-ALKALINE ANDESITES 841

Table l. Chemical compositionsof hornblendeand andesiticliquid with comparisons

Col.I

Si0z 4?.6 61.3 57.0 56.0 60.5 55.5 63.2 63.8 60.0 A1203 2t.0 20.9 t9.7 20.0 17.3 20.3 18.6 20.4 Fe02 8.2 3.4 3.6 6.9 4.2 8.5 3.0 4.3 4.5 Mgo lq 6 2.6 3.4 3.0 1.1 3.4 0.3 0.7 4.3 Ca0 12.7 5.5 8.9 10.0 8.4 9.2 7.4 6.9 7.5

Na20 L.d 4.3 4.0 2.9 3.2 L.6 3.0 2.8 0.9 Kz0 n? 1.5 1.2 0.6 0.8 0.7 0.9 1.0 0.9 Ti02 2.4 0.3 1.2 0.8 1.3 2.2 1.2 t.2 7.2 CI 0.05 0.1s 0.I2 0.08 n.o. n.o. n.o. n.d. n.d.

Sum3 98.2 97.4 95.5 e7.3 85.7 84.0 83.7 n.r. Hr0a 3 4 3 i0 10 10 6 ?10 l5 1080-960 1080-960 1080-960 1015 1045 1050 999 1010 p6 0.1-1? 0.1-l? n 1-1? 5 5 8.0 5.2 13 Assoc . Fo83,Px? Hb,FoB3, Fo79 Hb,Cpx, Hb,Cpx, Hb,Cpx, Cpx,Hb, Hb,Cpx Px? 0l 01,0x 0l ,Mt Ol,Mt 0x

rsee key to columnsbelow.

'Al I rron as Feu.

35umof the original analysis. Reportedanalyses are recalculatedto 100percent except for roundingerrors. Original sumfor column9 is not published. Sumsare Hz0-free.

4Estimatedconcentration of Hz0. Estimatedby difference(Anderson, 1973) for columnsI,2,4; estimatedfrom pressure,9as compositionand solubility otherwise. sEstimatedtemperature in degreesCelsius. Seetext for columns1,2,4. 0therwisepublished run temperature.

6Estimatedpressure in Kbar. Seetext for 1 and 2. otherwiserun pressure(total pressurenot necessarilysame as PH"0/. .l00 = 'Co.*irtinq minerals: Fo = olivine with atomicratio of tlS/(Mg+Fe);ttU = hornblende,Px = pyroxene,Cpx calcium-richpyroxene, 0l = olivine, Mt = magnetite,0x = oxjde. Key to Co1umns:

L = Hornblendeincluded in olivine togetherwith glassof column2 (analysis9 of Table3 of Anderson,1979).

2 = Andesiticglass includedin olivine, Asama1783 (analysis 8 of Tab.le3 of Anderson,1979)'

3 = Result of adding0.4 parts hornblende(column 1), 0,7 parts glass (column2), and 0.1 part of orthopyroxene (stoichiometricEn83) and subtracting 0.2 parts olivine (FoB3).

4 = Andesiticalass includedin olivine, Asama1783 (analysis 10 of Table3 of Anderson,1979). 5 = Syntheticandesitic glass (analysis15a of Table7b of Helz, 1976).

6 = Syntheticandesitic glass (analysis16a of Table7b of Helz, 1976).

7 = Syntheticandesitic glass (fromTable 8 of Hollowayand Burnham, 1972).

8 = Syntheticandesitic glass (fromTable B of Hollowayand Burnham, 1972).

9 = Syntheticandesitic Alass (averageof analyses2l and?2 of Table 6a of Allen and Boettcher,1978). tions (Hm24, Usp34)suggest a temperatureof equili- 1070"C,according to my visual interpolation of data bration of 1080'C, accordingto the unpublishedgeo- by Buddington and Lindsley (1964) and Taylor thermometer of Lindsley and Rumble (1977). The (1964).Similar compositionsbut lower temperatures samecompositions suggest a temperatureof 10600to (990"-1020"C) were reported by Oshima (1976). 842 ANDERSON: HORNBLENDE IN CALC-ALKALINE ANDESITES

that reaction and crystallization took place after fragmentation. The faceted (not skeletal) shape and size of hornblende included in olivine suggest it to be a daughter crystal or reaction product rather than a quench crystal or product of devitrification. The round contact between the olivine and glass associated with hornblende is consistent with a reaction relation between olivine, liquid, and hornblende. Crys- tals intergrown with a residual rhyolitic liquid were broken apart and overgrown probably because in- vading, olivine-porphyritic magma caused effervescence of the interstitial liquid (due to pressure release and/or temperature increase) and contaminated the liquid with relatively refractory constituents. Liquids trapped within the olivine crystals became modified in composition as a result of growth of daughter crystals of hornblende and/or pyroxene and reaction with olivine, because the temperature decreased from that in the initial environment to near that of the invaded andesitic magma with residual rhyolitic liquid. Fig. 3. Photograph of a second, smaller inclusion in the same The initial composition of the trapped liquid be- crystal as that shown in Fig. 2. The inclusion is about 30 pm long fore crystallization of hornblende and pyroxene can and contains a pale brown crystal of probable hornblende (H) and parts a dark bubble of vapor. be estimated by adding 0.4 hornblende, 0.1 parts pyroxene and 0.7 parts andesitic glass, and subtracting 0.2 parts olivine. The result (Table l, column 3) is somewhat similar in composition to other in- Interpretation clusions of andesitic glass in olivines which lack The andesiticglass (Table 2, column 1) probably is hornblende (Table l, column 4). If other minerals are in a state of quenched equilibrium with olivine included in the calculation, the similarity can be in- (about Fo83) and hornblende, becausethe pumice creased. Some inclusions of glass in the Asama 1783 lump in which it occurswas explosivelyerupted and olivines have as little as 54 percent SiO, (Anderson, cooled in air before falling to the ground. The in- 1979). lt is not possible to relate all the glassesto a clusion probably cooled from about 1080'C to a few conmon precursor without including either minerals hundred degreesC in a few minutes,judging from not now in contact with the glasses,or residual rhyo- the Plinian explosionswhich probably erupted it in litic liquids as mixing components, or both. The ori- late July or early August of 1783(Aramaki, 1956; gins of the glassesincluded in the olivines are com- written communication,1979). Possibly the inclusion plex if considered in detail. Such detail is beyond the was containedin a froth of magma at a few hundred scope of this paper. Subtraction of olivine in the atmospheresor lessfor a few days,because it was ex- above calculation corresponds to resorption of oli- truded during the early part of the eruption when the vine concomitant with crystallization of hornblende vent probably was intermittently filled with froth. and pyroxene from a less silicic liquid precursor. I interpret the textural, compositional,and histori- Probably the andesitic liquid was in equilibrium with cal facts to suggestthat andesitic or basaltic liquid hornblende, pyroxene, and olivine at the temperature carried crystalsof olivine into a crystal-rich body of of eruptive quenching; however, the relation with oli- andesiticmagma during or before the early stagesof vine probably was one of reaction. the eruption (May 9 to August 2). The irregular an- Two kinds of experimental results can be com- hedral cores of many plagioclase crystals suggest pared with the natural association of hornblende, py- fragmentation.The abundanceof gas in the sieved roxene, olivine, and andesitic liquid. Consequently, it cores suggeststhat fragmentation was accompanied is possible to obtain two estimates of the temperature by and/or causedby effervescence.The rims on the and the concentration of HrO in the silicate liquid. In crystalsof plagioclase,pyroxene, and olivine indicate the first case comparison is made with the liquidus ANDERSON: HORNBLENDE IN CALC.ALKALINE ANDESITES 843

Table 2. Chemicalcompositions of hornblendesand associatedbasaltic materials

Col.1

si02 42.8 5i.0 42.1 5r./ 41.4 54.0 40.7 55. 5 Alz0g 13.6 18.1 19.1 14.3 18.4 16.5 17.7 13.3 17.3 Fe02 10.5 9.4 10.0 i0.8 q'1 12.4 9.2 L4.2 8.5 Mgo 1.5.J 3.5 q4 15.6 14.1 4.8 13.0 3.4 Ca0 12.2 o.J v.J 11.8 10.3 11.0 8.3 10.7 9:3

Na20 2? 3.7 3.0 2.5 2A L.O 3.8 2.0 2.8 Kz0 0.4 1.0 0.6 0.3 0.5 0.1 0.8 0.5 0.7 Ti02 2.7 1.2 1.1 2'l 1.0 1.8 4.4 2.2 cl 0.05 0.12 0.09 n.d. n.o. n.d. n.o. n.d. n.d.

5Um" 98.9 93.8 94.3 99.8 99.8 98.7 99.8 98.7 85.7 Hr0q 1 4 4 1.8 n,o. n.o. n.d. n.d. 8 T5 1030 1030 1030 n.0. n.o. 1000 1000 1045 1045 p6 1' r-a I-2 8 8 12-30 L2-30 55 Assoc.7 An92 An86 Fo79 0l,Cpx, 01,Cpx Fo83 OI,PI Fo82-84 Fo82-84 PI,Mt Hb,PI, Cr,Cpx Hb,Cr,Cpx Mt Keyto Columns(see Table 1 for footnotes):

1 = Hornblendeincluded in core of plagioclasephenocryst from Fuego(analysis reported in Iegendto Figure16, p. 20 of Roseet al., 1978).

2 = Glassincluded in rin of plagioclasephenocryst (An83) of column7 (analysisNo. 6 of Table6 of Roseet al,, 1978).

3 = Glassjnciuded in olivine (Fo79)from Fuego (analysis 4 of Table6 of Roseet al.,.l978, andclose in composition to the weightedaverage of the 1974erupted ash).

4 = Hornblendefrom gabbroic block, La SoufriEre,St. Vincent(analysis 1 of Table6 of Lewis,1973a). = S f!lg grained,interstitial materialfrom samegabbroic block as column9 (analysis1 of Table4 of Lewis, I 973b).

6 = Hornblendeinclusions in olivine phenocrysts,Little lYt.Hoffman basalt, California (column 2, Table2 of Mertzman,i978) . = 7 Groundmassof Little Mt. Hoffmanbasalt (calculatedby subtracting10 percentolivine IFo80]and 25 percent plagioclase[An70] from the bulk rock--analysis1 of Table I of l4ertzman,1978). 8 = Synthetichornblende (tielz, tgZO, Table 13a, No. 2, p. 192). 9 = Syntheticliquid associatedwith hornblende(column B) calculatedby Helz (1976,Table 7b, analysisNo. 16a, p, t5z).

phasesof an andesiticbulk composition.In the sec- its pale color suggestsa negligible concentrationof ond casecomparison is madewith the residualliquid oxycomponent;and (3) its fluorine content is un- in a basalticbulk composition. known, but probably small, becausefluorine is a mi- For the first comparisontwo assumptionsare nec- nor constitutent of most magmas. The second as- essary:(l) the hornblende is a hydroxy-hornblende; sumption may be questioned because olivine (2) the mineralsin contactwith the natural andesitic probably is in a reaction relation. For the present liquid are thermodynamicallyequivalent to the liq- purposesI assumethat the presenceof olivine may uidus phasesof a system composedof the liquid be ignored.Consequently, I seekexperimental condi- alone. tions which yield hornblende and pyroxene on the The first assumptionis justifiable because:(l) the liquidus of andesite.The compilation of Stern et al. natural hornblende contains negligible chlorine; (2) (1975)suggests that compositionallysimilar andesitic ANDERSON:HORNBLENDE IN CALC-ALKALINE ANDESITES liquids have hornblende and clinopyroxene on the as a reactantis a principal causeofthe greaterstabil- liquidus if the temperatureis below about 980oCand ity of hornblendein basalticsystems. Other composi- the liquid contains more than about l0 weight per- tional differences [greater AlrO, (Bowen, 1928, p. cent HrO. The resultsof Allen and Boettcher(1978) lll), and alkalis (Cawthorn and O'Hara, 1976)lmay on andesitic bulk compositionsare consistentwith help stabilize hornblende,but they cannot be eval- the aboveestimate: more than l0 weight percentH2O uated quantitatively with published information. In in the liquid. sum, comparisonof the natural andesiticliquid and For the secondcomparison only one assumptionis associatedminerals with experimental products sug- nec€ssary:the hornblende is a hydroxy-hornblende gests that the natural liquid contained less than 6 (ustffied above).I now seekexperimental conditions weight percent HtO and quenched from a temper- which yield hornblende, pyroxene, and andesitic liq- ature greaterthan l0l5'C. uid in a reactionrelation with olivine. Holloway and Sekins et al. ll979l studied experimentally the Burnham (1972) and Helz (1973, 1976)encountered Asama 1783andesite. They found no hornblendeon andesitic residual liquidus associatedwith horn- or near the liquidus, but did estimatethat the resid- blende, pyroxene, and olivine (and in some cases ual dacitic liquid (69-72 weight percent SiO,) of this magnetite)at temperaturesbetween about 1000and andesite contained 3.3 weight percent of HrO at 1050'C and with about 6 to l0 weight percent HrO 960'C before extrusion. in the silicate liquid (Table l, columns 5-8). Allen Could the concentrationand fugacity of H'O in and Boettcher (1978) found a comparable residual the included andesiticliquid exceedthat of the sur- liquid (Table l, column 9, but no olivine). However, rounding residual dacitic liquid? Roedder's (1965) the compositionsof residualliquids are difficult to es- work reveals that olivines are strong enough at tablish for experimentalproducts. The detailed as- 1200'C to contain excesspressures ofa few thousand sessmentby Helz (1976) suggeststhat the residual atmospheresin an inclusion. However, a significant liquids at temperaturesgreater than about l0l5'C gradient of HrO fugacity from inclusion to rim of (at P",o : 5 kbar) are less siliceous than the natural crystal would probably disappearby diffusion in a liquid. The resultsof Allen and Boettcher(1978) re- few days or weeks at temperaturesnear l000oc, veal that the maximum temperatureof hornblende judging from data on Fe, Mg, and O diffusivity in stability increasesfrom about 990oto l050oC as the olivine (Buening and Buseck, 1973;Muehlenbachs mole fraction of HrO in the gas decreasesfrom I to and Kushiro, 1975). The diffusion of HrO probably is about 0.25, consistentwith the findings of Holloway comparable to that of oxygen becauseof the similar- (19'73).At the mole fraction of 0.25 and a total pres- ity in size.The concentrationof HrO (or OH-) in oli- sureof l0 kbar, the partial pressureof HrO would be vine probably is low in comparisonto oxygen; con- about 2.5 kbar and there would be about 6 weight sequently,the concentrationgradient and diffusion percent HrO dissolvedin the liquid. The Hawaiian of HrO in olivine may be small. Laboratory heating basalt investigated by Holloway and Burnham experimentson natural olivines with inclusions of (1972), H.elz (1973, 1976) and Allen and Boettcher glasssuggest that diffusivetransfer of HrO is effective (1978) probably is neither appropriate nor optimal over tens of microns in a few hours at temperatures for the generation of natural andesitic residual liq- abovel000oC (Anderson,1974). Probably both tem- uids becauseof its low concentrationsof AlrO, and perature and the fugacity of HrO in the trapped an- alkalis. Consequently,it is likely that high-alumina desitic liquid associatedwith hornblendeare similar basalt would yield andesitic residual liquids associ- to those in the surrounding melt. Consequently, the ated with hornblende,pyroxene, and olivine at pres- most probable conditions for growth of the horn- suresless than 5 kbar and with P",o lessthan P,o.^,at blende from the trapped andesitic liquid are 1080" temperaturesgreater than that found for Hawaiian (Fe-Ti oxides)to 960oCand 100atm (plausiblepres- tholeiite at 5 kbar : Pr,o: Ptor.r(namely, greater sure of froth in vent) to about 1000atm of HrO pres- than aboutl0l5"C). sure (Sekineet al., 1979). The above comparisonssuggest that andesiticliq- In sum, the eruptive history, texture,and composi- uids and hornblende are stable to higher temper- tional relations of the hornblende-bearinginclusion atures and lesserconcentrations of HrO in basaltic suggestthat its glass representsan andesitic liquid bulk compositionsthan in andesitic bulk composi- quenchedfrom a temperatureof 1020+60"C and a tions. Probably the availability of olivine (or augite) partial pressureof HrO betweenabout 100and 1000 ANDERSON: HORNBLENDE IN CALC.ALKALINE ANDESITES atm. The liquid was in equilibrium with hornblende, of the bulk rock extruded in the October eruption. probably orthopyroxene, and olivine. The andesitic The basaltic inclusions and bulk rock probably are liquid probably contained a few percent HrO. sfunilar in composition to the parental liquid from which the earliest crystals formed. The anorthitic Hornblende associated with natural calc-alkaline cores and the included hornblendeprobably formod basaltic liquid from a lquid with a compositionnear that of the basaltic glasses(Table 2, column 3). Observational data The temperature of the parental basaltic liquid Hornblendes associated with calc-alkaline basaltic was about 1030'C; it contained about 4 weight per- material are known from La Soufridre, St. Vincent cent HrO (Rose et al., 1978).Subsequently, Harris (Lewis, l973a,b), Fuego (Rose et al.,1978), and Med- (1979) has directly measured 1.6 to 3.2 weight per- icine Lake Volcano, California (Mertzman, 1978). cent of HrO in inclusions of glassin olivine pheno- Analyses of these hornblendes and related materials crysts from Fuego. The lower concentrationof 1.6 are compared in Table 2. The compositions of the weight percent HrO applies to parental liquid in hornblendes are similar and have 13.6 to 16.5 weight Fo77 olivine. AccordinEly, a revised estimate of percent Al2O3, 2.3 to 2.6 weight percent NarO, and 1030150"Cand 1.6weight percentHrO is preferable 1.8 to 2.7 weight percent TiO, on an anhydrous basis. for the Fuego liquid which probably precipitatedthe hornblende. Interpretation It is uncertain whether any of these hornblendes La Soufriire, St. Vincent are in a state of quenched equilibrium with basaltic Lewis's (1973a,b)descriptions show that vesicular liquids, because the textural relationships are not de- and microcrystalline dark interstitial material finitive and because quenched basaltic matrix is touches hornblende as well as olivine, plagioclase, termed andesite by some workers. A case-by-case pyroxene,and magnetitein xenolirh 770. The inter- analysis and interpretation is necessary. stitial material (termed scoriaby Lewis) analyzedby him is basaltic in composition (column 5, Table 2). Fuego volcano, Guatemala Lewis (written communication, 1979) considershis The analyzed hornblende (Table 2, column l) is in separateto be a good one, little affectedby products the anorthitic core (An92) of a plagioclase pheno- of weathering and contaminatingpieces of the crys- cryst (An92 to An80). The most closely associated tals from the gabbro. His analysesshow that inter- analyzed glass (column 2, Table 2) is an inclusion in stitial scoria from different blocks differ in composi- the rim of the same crystal. The glass and the horn- tion, consistent with an indigenous origin of the blende are not in contact. The crystal is depicted in interstitial material. The most siliceous interstitial Figure 16 of Rose et al. (1978). The glass in the scoria comes from a block of gabbro lacking horn- plagioclase is transitional in composition between ba- blende but containing olivine, plagioclase,pyroxene, salt and andesite. and magnetite. I would expect progressivesolidi- Rose et a/. suggestedthat the hornblende included fication to increasethe number of minerals. s6lilize in the anorthitic core actually grew from a liquid hornblende, and enrich the residual liquid in KrO with a more basaltic composition than the inclusion and SiOr. Possiblythe gabbro with the most siliceous of glass found within the rim of the plagioclase. Their interstitial material originates from a lesser depth. reasoning was as follows: many cores of anorthite Lewis'sdata suggestthe crystallizationof hornblende from the same eruption contain inclusions of olivine. from basalticliquid. Subhedral crystals of anorthite (An92-90) are com- mon in olivine phenocrysts. The anorthitic cores of Liltle Mt. Hoffman, California the plagioclase phenocrysts have round corners, com- Olivine phenocrystsin a basaltic lava from a vent prise a low proprotion of the total plagioclase, are of Modoc Basalt (Powers,1932) contain subroundto subequant and little zoned. Most of the Fuego irregular inclusionsof hornblende(Mertzman, 1978). plagioclase is oscillatory zoned. The textures of the Olivine phenocrystsin the scoria of the cinder con€ anorthite cores suggest that it is an early mineral to (Little Mt. Hoffman) at the sourceof the flow have form in the Fuego magma. Some olivine phenocrysts round inclusions of devitrified glass, but no horn- contain inclusions of basaltic glass (51 percent SiOr) blende (A. T. Anderson, unpublished data). The which are close to the weighted average composition compositionsof the hornblendeand the groundmass ANDERSON: HORNBLENDE IN CALC-ALEA,LINE ANDESITES of the basalt are given in Table 2, columns 6 znd 7. (1973a,b)on ejecta of Fuego and Soufriire volca- Becausethe hornblendeoccurs included in olivine noes, respectively, suggest that hornblende grows phenocrystsin basaltic lava, Mertzman (1978) rea- stably from some high-alumina basalt liquids with sonablyargued that the hornblendeformed from ba- about 5l weight percent SiO, at temperaturesnear saltic liquid. The occurrenceof hornblendeincluded 1030'C. The basalticmelts are comparativelyrich in in olivine together with andesiticglass (seeAsama, Fe and poor in Mg and crystallizeanorthitic plagio- above) suggestspossible alternative interpretations: clase (commonly greater than An85), olivine (Fo84 (l) the olivines with hornblende are xenocrystsde- to 66), and magnetite,as well as hornblende. rived from an andesitic environment, (2) the horn- Experimental results (Helz, 1973, 1976) demon- blendesare daughtercrystals which crystallizedafter strate the crystallization of hornblende at crustal the trapped melt attained an andesitic composition pressuresfrom basalticandesite but not from basaltic by crystaltzing olivine. Mertzman'sfigures show that liquids. As argued above for andesitic liquids, the the hornblende is associatedwith microcrystalline crystallization of hornblende is fostered by greater material (probably formerly tiquid). Becauseof its concentrations of alumina and alkats and P,"o, > round shape and association with microcrystalline P",o probable for natural high-alumina basaltic liq- material, it is likely that the hornblende grew from uids compared to the Hawaiian basalt studied by trapped liquid after it had crystallized olivine. Helz. The Warner high-alumina basalt studied by Growth of hornblende from supercooledliquid at a Yoder and Tilley (1962)has atypically greaterMgO, pressureless than 1000atmospheres is a third possi- MgO/(MgO + FeO), and low alkalis for basaltic bility. The compositionof the liquid from which the members of calc-alkaline magma series.Rose et a/. hornblendein the Little Mt. Hoffman basalt fcrmed (1978)argued that the smallerMgO/MgO + FeO) is uncertain. Mertzman (written communication, ratio for Fuego high-alumina basalt would have the 1979)notes that the 0.1 weight percentof K,O in the effect of making its olivine liquidus temperature Little Mt. Hoffman hornblendessuggests that the liq- about 80oC lessthan that of the Warner basalt. The uid from which they formed had lessthan about 0.5 difference might permit liquidus hornblende for ba- percent KrO. Possibly the groundmassof the Little salt with a concentrationof about 5 percent HrO in Mt. Hoffman basalt is enriched in KrO by con- the liquid if PH,o: P,o,or.Experimental results are not tamination (comparestudies of East Sand Butte and consistent with crystallization of hornblende from cinder cone M39 of Anderson, 1976). Additional some natural high-alumina basaltic liquids with less work on the Little Mt. Hoffman materials is rteeded than a few weight percent HrO. to evaluatethe alternatives.I suggestthat the liquid discussion from which the hornblende grew was richeriin SiO, . General than is the bulk rock becauseof crystallizationof oli- The facts and interpretations outlined above sug- vine. Probably the liquid was similar in composition gest that hornblende crystallizesfrom andesitic and to that of the calculatedgroundmass given as column basaltic liquids at crustal pressures.Most such horn- 7 of Table 2. The compositionis that of a basaltican- blende probably is a product of a reaction between desite. liquid and olivine, as first suggestedby Bowen (1928, Mertzman (1978)extrapolated the resultsof Yoder p. 6l). The field and textural evidence of plutonic and Tilley (1962)on the Warner high-alumina basalt rocks is consistentwith crustal formation of horn- to estimate that the hornblende initially formed at blende by reaction between liquid and olivine pressuresgreater than about 12 kbar. In my judge- (Miller, 1938;Best and Mercy, 1967;Lewis, 1973b, ment Mertzman's apptcation and extrapolation of Ikeda, 1976;Walawender,1976; Mullan and Bussell, Yoder and Tilley's work is not justified, becausethe 1977).The l0o- to 106-yearlifetimes of subduction- hornblende in question probably grew from a more zone volcanoes(McBirney et al., 1974;Rose et al., siliceous liquid with a greater Fel(Fe * Mg) ratio 1977 ; Mertzman, 1977 ; McBirney, 1978), the shorter- than the Warner basalt. term variationsin compositionsof extrudedproducts (Crandell and Mullineaux, 1973; Crandell et al., Summary 1962;Newhall, 1979),and the I km' and smallervol- Evidence for the crystallization of hornblende umes of products of most individual eruptions, as from natural calc-alkalineliquids at least as basic as well as the larger volumes of rare caldera-forming basaltic andesite (about 54 weight percent SiOr) is eruptions(Smith, 1979),suggest that the roots of sub- compelling.The data of Roseet al. (1978)and Lewis duction zone volcanoeshave thicknessesless than ANDERSON: HORNBLENDE IN CALC-ALKALINE ANDESITES

about I km and solidify (and differentiate) on time tends to buffer the Fel(Mg + Fe) ratio of crystalliz- scalesappropriate for cold (crustal) environments. ing minerals and residual liquid. Helz (1976) noted Geophysicalmodels of Cenozoicisland arcs suggest that the Fel(MB * Fe) ratio of pyroxenesand resid- they consistmostly of rock denserthan andesite(for ual liquids decreaseswith crystallization in part of example,see Grow, 1973).Gabbroic rocks,including the range of solidification of hornblende, for ex- hornblende gabbro, have appropriate densitiescon- ample. Reactionwith magnesium-richminerals is an sistentwith Kuno's (1968)model of a gabbroic crust effective mechanism which minimizes increasesin in island arcs. Dominant hornblende gabbro in is- the Fel(Mg + Fe) ratio of residual liquids. land arcs is particularly appealing becauseit can The influenceon Fel(Mg + Fe) of the crystalliza- yield derivative magmaswith the calcium-rich cores tion of a hornblende-bearingassemblage in place of of plagioclasecrystals, the isotopic ratios of stron- an equivalent anhydrous assemblageof minerals is tium and lead, and the bulk compositionstypical of complex and disputed (seeRingwood, 1974and Al- many rocks in granodioritic batholiths (Piwinskii and len and Boettcher,1978 for contrastingviews). What Wyllie, 1968; Faure and Powell, 1972; Doe, 1967; is meant by "equivalent"? Considerthe hornblende- Wyllie, 1977).Thebasaltic and andesiticliquids from out surface of Figure l. The mineralogy above the which hornblende forms appear to contain 1.6i0.3 surfaceand the compositionof the liquid on the sur- (Harris, 1979)to 4t-2 wei5ht percentof HrO (Roseer face vary with P, T and X,,o. Many definitions of al.,1978;Anderson, 1979). The temperatureat which "equivalent" are possible.I adopt the following defi- hornblende forms in such liquids is probably be- nition becauseit seemsconsistent with most petrolo- tween 1080oand 960oC(see above). gical discussionsregarding the role of hornblende in Various factors probably contribute to the crystal- the formation of calc-alkalinerock series:on any sur- lization of hornblendefrom natural basalticand an- face marking identical bulk compositionsof liquids, desitic liquids with smaller concentrationsof HrO regionswith different solids are equivalent. than the minimum of about 6 weight percent sug- For a given incrementof crystallizationhow do the gestedby experinents (Yoder and Tilley, 1962;Hol- Fel(Mg * Fe) ratios of residualliquids in two equiv- loway and Burnham, 1972;Green, 1972;Helz, 1973, alent fields compare?For a given bulk composition 1976;Stern et al.,1975; Allen and Boettcher,1978): (Hro-free), the liquidus surfaceis one (and possibly (l) olivine commonly is presentas a reactantin natu- the only) surfacewith identical compositionsof liq- ral associations;(2) natural liquids have someCO, in uids. Consider the difference between the Ol + Pl addition to HrO (Harris, 1979),consequently P^,o I and the Ol + Hb fields. I assumethat the equilibrium P,"*,i(3) natural basalticliquids associatedwith calc- compositionof olivine on the liquidus is everywhere alkaline magma serieshave greaterconcentrations of the same (Roeder and Emslie, 1970).In general the alumina and alkalis than Hawaiian basaltsused in Fel(Mg + Fe) of hornblendeis greaterthan for asso- experiments;(4) natural calc-alkalinebasaltic liquids ciatedolivine (Helz, 1973;Lewis, l973a,b). However, have smaller MgO/(MgO + FeO) and greater con- the effecton Fel(MB + Fe) of the residualliquid de- centrationsof alkalis than the Warner basalt used in pendson the proportionsof the crystallizingminerals experiments.All these differencesprobably enlarge as well as on their compositions.The net iron enrich- the field of stability of hornblendebeyond that found ment causedby crystallizationof Ol + Hb possiblyis in experiments.Reaction with olivine and other fac- greaterthan that for Ol + Pl. Probably the variation tors probably foster the crystallizationof hornblende of Fel(Mg * Fe) of residualliquids with equilibrium in some natural basaltic and andesitic liquids with crystallization of equivalent hornblende-freeand less than 6 weight percent of HrO, as suggestedby hornblende-bearingassemblages is complex,but not the natural occurrencesdescribed above. as large as the variationscaused by fractional (as op- Reactionbetween olivine and liquid yielding horn- posed to equilibrium) crystallization.I agree,there- blende acts to minimize the increasein Fel(Mg + fore, with Cawthorn and O'Hara (1976) that the Fe) ratio in derivativeliquids. Two factorsare signif- omission of iron from their experimentsprobably is icant in minimizing the increasein Fel(Mg + Fe): not crucial to an evaluation of the role of horn- (l) reaction betweenliquid and olivine; (2) crystalli- blende. My emphasisis on reaction,however, rather zation of a hornblende-bearingassemblage of miner- than fractionation. als in place of an equivalent anhydrousassemblage. Minor increasein Fel(Mg + Fe) is a characteristic The effect of reaction was emphasizedby Bowen feature of calc-alkalinemagma series(Fenneg 1926; (1928,p. 6l). With reactionmagnesium-rich olivine Kuno, 1950; Miyashiro, 1974). Consequently, the ANDERSON:HORNBLENDE IN CALC-ALKALINE ANDESITES spectrumof rock seriesfound in subductionzone en- betweenthe Benioff zone and the volcanic front, as vironments(Peacock, l93l; Kuno, 1950,1959) can be has been emphasizedby Hasebeet al. (197O). related in part to variable reaction between olivine Some of the HrO subducted as oceanic crust is and liquid yielding hornblende:in generalthe more probably not returned to th€ surface by the out- hornblende. the more calcic and the less iron-en- gassing of parental basaltic liquids in subduction riched (less tholeiitic) the series.Many tholeiitic se- zones.There is an averageofabout I to 2 weight per- ries are calcic, however (for example Higashiyama, cent of mineralogically-boundHrO in the combined Isshiki, 1963)and have basalticmembers with small oceanic crustal layers 2 (Melson et al., 1968;Hum- concentrations of alkalis. The concentration of al- phris and Thompson, 1978) and 3 (Melson and kalis in the parental liquid is important. Serieslack- Thompson,l97l-Pnnz et al.,1976;lto, 1979;Ander- ing hornblendeprobably are tholeiitic or alkalic and son et al., 1919). Parental basaltic liquids of sub- tend to have groundmasspigeonite. Variable produc- duction zonesprobably contain an averageof about tion of hornblende by reaction betweenliquid and 2 to 4 weight percent H'O (Eggler, 1972;Anderson, olivine can help explain variable iron-enrichment 1973, 1979;but see also Marsh, 1976;Garcia et al., and lime-alkali index of manv subduction zone 1979; and Sekine et al., 1979,who estimate lower magma series. concentrationsof HrO). The ratio of massproduction of crust in subduction zones(continental growth) to Implications that at oceanicridges is about 0.2 (Anderson, 1974), The probable crystallizationof hornblende within basedon the ratios of cross-sectionalareas of Ceno- the crust from subduction zone basaltic liquids with zoic oceanicisland arcs(Murauchi et ul., 1968;Grow, lessthan 6 weight percent HrO has implications for: 1973)and the estimatedcross-sectional areas of sub- (l) the genesisof andesite,(2) the temperatureabove ducted oceaniccrust implied by late Cenozoiclocal the Benioff zone, (3) dehydration of subductedoce- rates of plate convergence(Chase, 1972).If all the anic crust, and (4) the origin of continental crust. HrO bound in subductedoceanic crust returns to the Genesisof andesitewithin the crust is implied by surface by first entering into basaltic parental liquids, the crustal formation of hornblendegabbro. As dis- the expected concentration of HrO in the liquids cussed above, the Fel(Mg + Fe) ratio and calc- ranges from 5 (l/0.2) to l0 (2/0.2) weight percent. alkalic index for subduction zone magma seriesare The estimatesof HrO actually present (2 to 4 per- explicable according to crystallization of hornblende cent) are uncertain,but are consistentwith the notion by reactionbetween olivine and liquid. Although the (Andersonet al.,1979) that H'O is releasedfrom de- storageand releaseof HrO in and from hornblende hydrating chlorite (the principal mineralogical resi- in subductedoceanic crust may be important in ini- denceof HrO in oceaniccrust) at too shallow a depth tiating subduction-zonemagmatism, reactions be- (Delany and Helgeson,1978, but seeJsnkins, l!l!) tween hornblende and liquids in the Benioff zone to contribute to melting in subduction zones. The probably have no direct influence on the character- HrO associatedwith the parental liquids of sub- istics of magma seriesin subductionzones (compare duction zone magmas is similar in amount to the Green and Ringwood, 1968;Green, 1972;'Boettcher, HrO supplied by amphiboles in the oceanic crust 1973,with Stern et al., 1975;Cawthorn and O'Hara, (Ito, 1979). 1976;Allen and Boettcher,1978). A complex origin of continental crust is implied by Temperaturesgreater than about 1000"C are im- parental basaltic liquid and hornblende gabbro in plied by basalticliquids with lessthan about 6 weight subductionzones. A principal idea regardingthe for- percentHrO. If subductionzone basaltic liquids orig- mation of continental crust involves the postulated inate betweenthe Benioff zone and the surface.then production and accretion of andesitic island arcs the temperaturemust be at least as high as l000oC (Wilson, 1954,p.206), becauseaverage continental somewherein the interval. Geophysical models of crust is andesiticin composition(Goldschmidt, 1933; subductionzone processes generally yield lessertem- Taylor, 1967).The chemicalcomposition of the bulk peratures, particularly close to the relatively cold and lower continental crust is uncertain and debated Benioff zone (compare Andrews and Sleep, 1974; (see Taylor and Mclennan, 1979,and Tarney and Anderson et al., 1976,with Hasebeet al., l97O; Ito, Windley, 19'79for contrastingviews). If the bulk of 1978).In order to be consistentwith volcanological Cenozoicisland arcsis composedof hornblendegab- and geologicaldata, geophysicalmodels should yield bro, then how can the andesiticcomposition of aver- temperaturesin excessof l000oC in the magma path age continental crust be explained? The implication ANDERSON: HORNBLENDE IN CALC-ALKALINE ANDESITES 849

of dominant hornblendegabbro in island arcs is that Boettcher, A. L. (1973) Volcanism and orogenic belts-the origin continental crust either was made diferently before of andesites. Tectonophysics, I 7, 223-240. Cenozoic times or is further differentiated Bowen, N. L. (1928) The Evolution of the Igneous Roc&s. Dover, bv addi- New York. tional processes. Buddington, A. F. and D. H. Lindsley (1964) Iron-titanium oxide minerals and synthetic equivalents. J. Petrol., 5,310-35'1. Note addedin proof Buening, D. K. and P. R. Buseck (1973) Fe-Mg lattice diffusion in The following important article appearedafter this olivine. "/. Geophys Res., 78, 6852-6862. paper was acceptedfor publication: Ritchey, J. L. Cawthorn, R. G- and M. J. O'Hara (1976) Amphibole fractionation in calc-alkaline magma genesis. Am. (1980) Divergent J. Sci., 276,309-329. magmas at Crater Lake, Oregon: Chase, C. G. (1972) The N plate problem of plate tectonics. products of fractional crystallization and vertical Geophys. J. Roy. Astron. Soc., 29, ll'l-122. zoning in a shallow, water-undersaturatedchamber. Crandell, D. W. and D. R. Mullineaux (1973) Pine Creek volcanic J. Volcanol.Geothermal Res., 7,373-386. assemblage at Mount St. Helens, Washington. U.S. Geol. Sum. Bull., 1383A, l-23. Acknowledgments Recent age at Mount Rainier, Washhgton. U.S. Geol. Sum. I thank P. J. Wyllie for helpful discussion and review of an early Prof. Pap., 450-D, D64-D68. version of this paper. Bacon (Menlo C. Park) provided a helpful Delany, J. M. and H. C. Helgeson (1978) Calculation of the ther- review. I have benefited from rock samples and advice contributed modynamic consequences of dehydration in subducting oceanic by Shigeo Aramaki arid Naoki Onuma (Asama) and by W. I. Rose crust to 100 kb and >800oC Am. J. Sci.,278,638-686. (Fuego). The work reported here was supported by NSF grant Doe, B. R. (1967) The bearing of lead isotopes on the source of EAR76-15016A0l. granitic magma. J. Petrol., & 5l-83. Eggler, D. H. (1972\ Water-saturat€d and undersaturated melting References relations in a Paricutin andesite and an estimate of water content in the natural magma. Contrib. Mineral. Petrol., 34,261- Allen, J C. and A. L. Boettcher (1978) Amphiboles in andesite 27t. and basalt: II. Stability as a function of P-T-1..,6-fo,. Am. - andC. W. Burnham(1973) Crystauization and fractiona- Mineral., 63, lW4-1087. tion trends in the system andesite-H2O-Co2-O2 at pressures to --, and G. Marland (1975) Amphiboles in andesite l0 Kb. Geol. Soc. Am. Bull., 84,2517-2532. and basalt: I. Stability as a function of P-T-f o,. Am. Mineral., Faure, G. and J. L. Powell (1972) Strontium Isotope Geology. 60. 1069-1085. Springer-Verlag, Berlin. Anderson, A. T. (1973) The before-eruption water content of some Fenner, C. N. (1926) The Katmai magmatic province. J. Geol., 34, high-alumina magmas. Bull. Volcanol., 37, 53V552. 673-7',12. - (1974) Chlorine, sulfur, and water in magmas and oceans. Garcia, M. O and S. S. Jacobson (1979) Crystal clots, amphibole Geol. Soc. Am. Bull., 85, 1485-1492. fractionation and the evolution of calc-alkaline magmas. Con- - (1976) Magma mixing: petrological process and volcano- trib. Mineral. Petrol., 69, 319-327. logical tool. J. Volcanol. Geotherm. Res., 1,3-33. N W. K. Liu and D. W. Muenow (1979) Volatiles from - (1979) Water in some hypersthenic magmas. J. Geot., g7, submarine volcanic rocks from the Mariana Island arc and 509-53 l. trough. Geochim. Cosmochim. Acta, 43, 305-3 12. - and D. Goufried (1971) Contrasting behavior of p, Ti and Goldschmidt, V. M. (1933) Grundlagen der quantitativen Geo- Nb in a differentiated high-alumina olivine tholeiite and a calc- chemie. Fortschr. Mineral. Kristollogr. Pelrogr., 17, 112-156. alkaline andesitic suite Geol. Soc. Am. Bull., 82,1929-1942. Green, T. H. (1972) Crystallization of calc-alkaline andesite under --, D. M. Harris and E. Ito (1979) Toward geologic cycles for controlled high-pressure conditions. Contrib. Mineral. Petrol., H2O, CO2, Cl and S. (abstr.). Int. Union Geodesyand Geophysics 34. t5Ft66. Assembly, Canberra, Interdisciplinary Abstr., 26. - and A. E. Ringwood (1968) Gehesis of the calc-alkaline ig- Anderson, R. N., S. Uyeda and A. Miyashiro (1976) Geophysical neous rock suite. Contilb. Mineral. Petrol., 18,105-162. and geochemical constraints at converging plate boundaries. Grow, J. A. (1973) Crustal and upper mantle structure of the cen- Part I. Dehydration in the downgoing slab. Geophys. J. R. As- tral Aleutian arc Geol. Soc. Am. Bull., 84,2169-2192. tron. Soc., 44, 333-357. Harris, D. M. (1979) Pre-eruption variations of H2O, S and Cl, in Andrews, D. J. and N. M. Sleep (1974) Numericat modelling of a subduction zone basalt (abstr.). EOS, 60,968. tectonic flow behind Island Arcs. Geophys.J. R. Astron. Soc., 38, Hasebe, R., N. Fujii and S. Uyeda (1970) Thermal processes un- 237-25r. der island arcs. Tectonophysics, 10,335-355. Aramaki, S. (1956) The 1783 activity of Asama volcano, part l. Helz, R. T (1973) Phase relations of basalts in their melting range Jap. J. Geol. Geogr.,27,189-229. at PH2o:5 kb as a function ofoxygen fugacity. J. Petrol., 14, - (1957) The 1783 activity ofAsama volcano, part2 Jap. J. 249-302. Geol. Geogr., 28, ll-33. - (1976) Phase relations of basalts in their melting ranges at - (t963) Geology of Asama volcano. J. Fac. Sci. (Jniv To- Pr,o : 5 kb. Part IL Melt compositions. J. Petrol., 17,139-193. kyo,14,229-443. Holloway, J. R. (1973) The system pargasite-H2O-CO2: a model Best, M. G. and E. L. P. Mercy (1967) Composirion and crystalli- for melting of a hydrous mineral with a mixed volatile fluid. I. zation of mafic minerals in the Guadalupe igneous complex, Experimental results to 8 kbar. Geochim. Cosmochim. Acta, 37, California. Am. Minbral.. 52. 41647 4. 65 l'666. 850 ANDERSON: HORNBLENDE IN CALC-ALKALINE ANDESITES

- and C. W Burnham (1972) Melting relations of basalt with 2). The distribution of Mt. Asama pumice in 1783. Bull. Earthq. equilibrium water pressure less than total pressure. J. Petrol., Res.Inst.. 20.93-106. 13. l-29. Miyashiro, A. (1974) Volcanic rock series in island arcs and active - and c E. Ford (1975) Fluid-absent melting ofthe fluoro- continental rnargins. Am. J. Sci., 274,321-355. hydroxy amphibole pargasite to 35 kilobars. Earth Planet. Sci. Moorhouse, W W. (1959) The Study of Rocks in Thin Section. Lett.. 25.4448 Harper and Row, New York. Humphris, S. E. and G. Thompson (1978) Hydrothermal altera- Muehlenbachs, K. and I. Kushiro (1975) Measurements of oxygen tion of oceanic basalts by seawater. Geochim. Cosmochim. Acta, diffusion in silicates (abstr.). EoS, 56,459. 42, t07-126. Mueller, R. F. (1969) Hydration, oxidation, and the origin of the Ikeda, Y. (1976) Petrology and mineralogy ofthe lritono basic in- calc-alkali series. NlSl TN, D-5400, l-27. trusive rocks, central Abukuma plateau. J. Fac. Sci. Univ. To- Mullan, H. S. and M. A. Bussell (1977) The basic rock series in kyo, 19,229-251 batholithic associations. Geol. M ag., 11 4, 265-280. Isshiki, N. (1963) Petrology of Hachijo-jima volcano group, Seven Murauchi, S., N. Den, S Asano, H. Hotta, T. Yoshii, T' Asanuma, Izu Islands, Japan. J. Fac. Sci. Univ. Tokyo,15,91-134. K. Hagiwara, K. Ichikawa, T. Sato, W. J. Ludwig' J. I. Ewing' Ito, E. (1979) High-temperature alteration of oceanic gabbros N. T. Edgar and R. E. Houtz (1968) Crustal structure of the frorn the mid-Cayman rise (abstr.). Geol. Soc. Am. Abstracts Philippine Sea. J. Geophys. Res., 73,3143-fl'71. with Progrums, 11,449. Newhall, C. G. (1979) Temporal variation in the lavas of Mayon Ito, K (1978) Ascending flow between the descending lithosphere volcano, Philippines. J. Volcanol. Geothermal. Res., 6, 6l-83. and the overlying asthenospherc. J. Geophys. Res.,83,262-268. Oshima, O. (1976) Fe-Ti oxide minerals of the 1973 eruption of Jakds, P and A. J. R. White (1972) Hornblendes from calc-alka-. Asama volcano. Univ. Tokyo, Coll. of Gen' Educ., Sci. Pap.' 26' line volcanic rocks of island arcs and continental margitts. Am. 39-50. Mineral.. 57.887-902. Peacock, M. A. (1931) Classification of igneous rock series' "/. Jenkins, D. M. (1979) Experimental phase relations of chlorite- Geol., 39,54-67. and amphibole-bearing peridotites (abstr.). GeoI. Soc. Am. Ab- Piwinskii, A. (1973) Experimental studies of granitoids from the stracts with Programs, 11,451. central and southern Coast Ranges, California. Tschermaks Kuno, H. (1950) Petrology of Hakone volcano and the adjacent Mineral. Petrogr. Mitt., 20, 107-130. areas,Japan. Geol. Soc. Am. Bull., 61,95'l-1020. - and P. J. Wyllie (1968) Experimental studies of igneous - (1959) Origin ofCenozoic petrographic provinces ofJapan rock series: a zoned pluton in the Wallowa batholith, Oregon. .I. and surrounding areas Bull. Volcanol., 20, j7-76. Geol., 76,205-234. - (1968) Origin of andesite and its bearing on the island arc Powers, H. A. (1932) The lavas of the Modoc lava bed quadrangle, structure. Bull. Volcanol, 32, 14l-176. California. Am. Mineral., 17, 253-293. Lewis, J F. (1973a) Mineralogy of ejected plutonic blocks of the Prinz, M., K. Keil, J. A. Green, A. M. Reid, E' Bonatti and J. Soufridre volcano, St. Vincent: olivine, pyroxene, amphibole Honnorez (1976) Ultramafic and mafic dredge samples from the and magnetite paragenesis. Contrib. Mineral. Petrol., 38, 197- equatorial mid-Atlantic Ridge and fracture zone. J. Geophys. 220. Res., 81, 408'l-4103. - (1973b) Petrology of ejected plutonic blocks of the Souf- Ringwood, A. E. (19'14) The petrological evolution of island arc ridre volcano, St. Vincent, West Indies. J. Petrol., 14,8l-112. systems. Geol. Soc. Lond. J., 130, 183-2M. Lindsley, D. H. and D. Rumble, III (1977) Therrnodynamic solu- Robertson, J. K. and P. J Wyllie (1971) Rock-water systems with tion model for coexisting Ti-magnetite plus ilmenite (abstr.). special reference to the water-deficient tegion. Am. J. Sci., 271, EOS,58,519. 252-277- Marsh, B. D. (1976) Some Aleutian andesites: their nature and Roedder, E. (1965) Liquid CO2 inclusions in olivine-bearing nod- source. ,/. Geol., 84,2746 ules and phenocrysts frombasalts. Am. Mineral., 50,1746-1782. McBirney, A. R. (1978) Volcanic evolution of the Cascade range. Roeder, P. L. and R. F. Emslie (1970) Olivine-liquid equilibrium. Am. Rev. Earth Planet. Sci.,6,437456. Contrib. Mineral. Petrol., 29, 2'l 5-289. J. F. Sutter, H. R. Naslund, K. G. Sutton and C. M. White Rose, W I., Jr., A. T. Anderson, Jr., L. G. Woodruff and S. B' (1974) Episodic volcanism in the central Oregon Cascade Bonis (1978) The October 1974 basaltic tephra from Fuego vol- Range. Geology, 2, 585-589. cano: description and history of the magma body. "/. Volcanol. Melson, W. G. and G. Thompson (1971) Petrology of a transform Geotherm. Res.,4, 3-53. zone and adjacent ridge segments. R. Soc. Lond. Phil. Trans., N. K. Grant, G. A. Hahn, I. M. Lange, J. L. Powell, J. 268.423441. Easter and J. M. Degraf (1977) The evolution of Santa Maria - and T. H. van Andel (1968) Volcanism and metamorphism volcano, Guatemala. J. Geol., 85, 63-87. in the mid-Atlantic ridge, 22oN latitude. J. Geophys. Res., 73, Sakuyama, M. (1979) Lateral variations of H2O contents in Qua- 5925-5941. ternary magmas of northeastern Japan. Earth Planet. Sci. Lett., Mertzman, S. A., Jr. (1977) The petrology and geochemistry of the 43,t03-lrl. Medicine Lake volcano, California. Contrib. Mineral. Petrol., Sekine. T.. T. Katsura and S. Aramaki (1979) Water saturated 62.221-247. phase relations of some andesites with application to the estima- - (1978) A tschermakite-bearing high-alumina olivine tho- tion of the initial temperature and water pressure at the time of leiite from the southern Cascades, California. Contrib. Mineral. eruption. Geochim. Cosmochim. Acta, 43, 1367-1f76. Petrol , 67,261-265. Smith, R L. (1979) Ash-flow magmatism. Geol. Soc. Am. Spec- Miller, F S. (1938) Hornblendes and primary structures of the San Pap., 180,5-27. Marcos gabbro. Geol. Soc.Am. Bull.,49, l2l3-L232. Stern, C. R., W. L. Huang and P. J. Wyllie (1975) Basalt-ande- Minakami, T. (1942) On the distribution of volcanic ejecta (Part site-rhyolite-H2O: crystallization intervals with excess H2O and ANDERSON: HORNBLENDE IN CALC-ALKALINE ANDESITES 851

HrO-undersaturatedliquidus surfacesto 35 kilobars, with im- Turner, F. J. and J. Verhoogen (1960) Igneous and Melamorphic plications for magma genesis.Earth Planet. Sci.Lett., 2g, lgg- Petrology. McGraw-Hill, New York. 196. Walawender, M. J. (1976) Petrology and emplacement of the Los Stewart, D. C. (1975) Crystal clots in calc-alkalineandesites as Pinos pluton, southern California. Can. f. Earth Sci, 1J, 1288- breakdownproducts of high-Al amphiboles.Contrib. Mineral. 1300. Petrol.,53,195-2A4. Wilson, J. T. (1954) The development and structure of the crust. Tarney, J. and B. F. Windley (1979)Continental growth, islatrd In G. P. Kuiper, Ed., The Earth as a Planet, p. 138-214. Univer- arc accretionand the nature of the lower crust-a reply to S. R. sity of Chicago Press, Chicago, Illinois. Taylor and S. M. Mcl-eonar..Geol. Soc.Lond. J., 136,5Ol-504. Wyllie, P. J. (19'17) Crusral anatexis: an experimental review. Tec- Taylor, R. W. (1964)Phase equilibria in the systemFeO-FerO3- tonophysics, 43, 4l-7 l. TiO2 at 1300"C.Am. Mineral.,49,1016-1030. Yoder, H. S. and C. E. Tilley (1962) Origin of basalt magmas: an Taylor, S. R. (1967)The origin and growth ofconrinents. Tectono- experimental study of natural and synthetic rock systems. ./. physics,4,17-34. Petrol., 3, 342-532. - and S. M. Mclennan (1979) Discussionon ,.Chemistry, thermal gradientsand evolution of the lower continentalcrust. In J. Tarney and B. F. Windley, Eds., Geol. Soc.Lond. J., 136, Manuscript received, October 27, 1979; 49't-500. acceptedfor publication, March 25, 1980.