American Mineralogist, Volume 67, pages 686-a99, |962

Nucleation in laboratory charges of basaltic composition

C. Ar-eN Benrpslr-e

Department of Marine GeologyCorpus Christie State University

Corpus Christie, Texas 78411

nun Enrc Dowry

Department of Mineral SciencesAmerican Museum of Natural History

New York. New York 10024

Abstract

Isothermal nucleation experiments using liquid beads suspended in 5 mm-diameterplatinum-wire loops were carried out on several compositions near the thermal minima inthe system diopside-anorthite-forsterite. Homogeneous nucleation was never observed, attemperatures from 400o C to the liquidi. Crystals of diopside, anorthite and forsterite allnucleate predominantly on the platinum wire. Nucleation on the surface of the bead occursless often and is apparently promoted by mechanical stress. Similar isothermal experimentson a natural diabase also did not show homogenous nucleation, nor did dynamiccrystallization experiments in nickel and alumina crucibles. Homogeneous nucleationprobably never occurs in the laboratory or in nature in basaltic compositions. The principalquestion concerning natural basaltic rocks is whether nucleation in each case was internal,on suspended crystals, or external, on wall rocks or other surfaces.

IntroductionDynamic crystallization experiments have re-

cently become an important tool in volcanic petrol-ogy (Lofgren, 1980). In these experiments a moltenrock is crystallized by cooling at a programmedrate, the purpose being to obtain direct informationon textures and actual crystallization sequences ofminerals. Understanding the results of such experi-ments, as well as the textures of natural rocks,requires knowledge of the kinetic processes ofcrystal nucleation and growth. The theory of crystalgrowth rates is fairly well known (Kirkpatrick,1974), and it is reasonably easy to measure growthrates in the laboratory; indeed, there is a surprisingamount of data on growth rates in basaltic systems(Dowty, 1980a). On the other hand, nucleation is amuch more difficult phenomenon to deal with theo-retically and experimentally (Dowty, 1980a; Kirk-patrick, 1981). In principle, there are two basicmechanisms, homogeneous nucleation, in whichthe new crystal forms by itself in the liquid, andheterogeneous nucleation, in which the new crystalforms on another phase, usually a solid.0003-{04)V82l0910-0886$02.00 8E6

Previous experience of many investigators hadsuggested that most nucleation in the laboratory isheterogeneous, usually on container walls. Recentdynamic crystallization experiments have alsoshown that textures of some natural basalts are bestreproduced when the starting material is not com-pletely molten, implying that suspended crystalsoften act as heterogeneous nucleation centers.However, some recent experiments (e.g., Lofgrenet al., 1974; Donaldson et al., 1975b; Donaldson,1979) in which the starting material was completelymolten were interpreted in terms of internal nucle-ation (not on the container or surface), which undersuch circumstances would presumably be homoge-neous. Many of these experiments were carried outin samples suspended in platinum-wire loops, atechnique which minimizes the area of the interfacebetween melt and solid container.

If homogeneous nucleation were predominant inany rocks, it would simplify theoretical modeling ofcrystallization. This study began as experiments onsimple synthetic compositions, for testing and cali-bration of numerical models of crvstallization

Anorthite

BERKEBILE AND DOWTY: NUCLEATION IN BASALTIC COMPOSITIONS

DioPsi6g 1387

CaMgSi "QForsterit€Mg.SiO.

Fig. l. Compositions studied in the system diopside-anorthite-forsterite (in weight percent). Numbers givetemperatures of invariant points.

(Dowty, 1980a), on the implicit assumption thatnucleation would be homogeneous if the chargewere completely molten. However, no homoge-neous nucleation was observed in the syntheticcompositions or in a natural basalt melt. Instead,nucleation took place by a variety of heterogeneousmechanisms, depending on sample treatment.

Experimental methods

Furnaces were made of 2-inch diameter aluminamumes with SiC heating elements. The furnacescool at about 0.5'lsecond and heat at 0.1 to 0.5"/second. For dynamic crystallization runs, tempera-ture was controlled by a microcomputer. Runs inPt-wire loops were carried out largely as describedby Donaldson et al. (1975a), but instead of pressingpowdered material into pellets as these authors did,it was found to be easier to make glass beads by apre-melting process. To do this, holes of about 0.6mm in diameter were drilled in a graphite block andweighed amounts of glass or rock powder, normallyabout 140 mg, were inserted. The block was thenplaced in the furnace at 1350-1400'C in a Ptcrucible for 2-5 minutes. The spherical beads thusproduced were tacked onto Pt wire loops as de-scribed by Donaldson et al. (1975a). All beads ofsynthetic compositions were annealed at 1350-1400" C in air for at least one hour prior to crystalli-zation at lower temperature. Any graphite on thebeads is rapidly burned off in air at these tempera-tures.

Synthetic compositions were weighed out in 10 g

batches from reagent-grade MgO and CaCO3, spec-trosil glass, and Al2O3 powder originally suppliedby Frank Schairer of the Geophysical Laboratory.In order to form homogeneous glass, the mixeswere cycled at least twice through a process ofmelting in a Pt crucible and grinding in a siliconcarbide ball mill. Most compositions were checkedby atomic absorption analysis.

Photographs of whole synthetic glass beadsshown in this paper were taken by immersion of thebead in a transparent cell containing liquid with arefractive index close to that of the glass. All runson synthetic compositions were examined in thiscell both under the binocular microscope in incidentlight, and under the petrographic microscope intransmitted light.

Nucleation mechanisms-synthetic compositions

The synthetic compositions chosen in the diop-side-anorthite-forsterite (Di-An-Fo) system (Fig.1) constitute simplified analogs of various types ofbasalt, with clinopyroxene, plagioclase or olivine asthe initial liquidus phase. It will be noted in thefollowing descriptions that the sequence of phasescrystallizing is sometimes not that expected if equi-librium were always maintained during cooling. Thedetermining factor in nucleation and growth ratesunder real conditions is the supercooling, AZ, withrespect to the liquidus of the phase in question,whether this liquidus is stable or only metastable;normally only stable liquidi are shown in phasediagrams such as Figure l. For example, in aeutectic system such as Di-An, the metastablediopside liquidus must be visualized as an extensionof the stable liquidus past the eutectic to lowertemperature and higher An content. If a liquid withAn content greater than that of the eutectic issuddenly brought to a temperature below the meta-stable diopside liquidus, diopside may crystallize inaddition to or instead of anorthite. More detaileddiscussions of the theory of metastable crystalliza-tion are given by Baker and Cahn (1974) and Dowty(1980a).

In the present experiments, nucleation was foundto be rather sporadic and unpredictable (the numberof crystals did not follow a Poisson distribution) atsupercoolings of less than about 60". This is similarto the findings of Fenn (1977) and Swanson (1977)on nucleation in granitic systems and Gibb (1974) onnucleation of plagioclase in basalt, but contrastswith the findings of Donaldson (1979) on nucleationof olivine in natural basalts.

888 BERKEBILE AND DOWTY: NUCLEATION IN BASALTIC COMPOSITIONS

Table l. Nucleation and morphology data for simple isothermal crystallization experiments

Composit ion # of AT range Phasesruns

Locus of Rate MorphologyNuc lea t ion (# /h r )

Di 6gAn4g

D i 6 5An35

Di 56An59

Di44An44Forz

10-60

60-120

. 20 L20-260 Di

10-55

65-285

20-33

40-140

240-640 Fo

on w i re 0 -3

o n w i r e 2 - L 0

on w i re 4+

D i

[ 1 8

li 15

0-4

0-4

fs

l,u

Fo[ "

{ 8

t,

An on wire

An,D i on w i re

(110) p r isms; ho l low (pyramiCa l reent ran ts ) . Crys ta lsbecome longrer and m6re hollow with increasing AT.

Dendrites; ini tal ly long and pointed, they become stouterwith increasing AT.

Spheru l i tes .

Plates on (010); polygonal at low AT, become rectangularhour -g lass , then dendr i t i c w i th inc reas ing AT.

An dendr i tes , 0 i spheru l i tes .

Fo on w i re 0 -0 .5 Euhedra l ; t yp ica l igneous shape.

wire, surf. l .-6 Rectangular plates; become hourglass, then dendrit ic withincreasing AT.

wire, surf. 1.0-20 Dendrites and spherul i tes.

AT ranges are approximate; rnorphologies overlap considerably in some cases.

Distinctly different modes of nucleation wereobserved in isothermal runs on synthetic composi-tions in Pt-wire loops, depending on whether thebead was (1) subjected to a simple isothermaltreatment, being taken directly from the annealingtemperature of 1350-1400" C to the crystallizationtemperature without removal from the furnace; (2)subjected to more complex thermal treatment, stillwithout removal from the furnace, or (3) removedfrom the furnace at the annealing temperature andquenched to room temperature before being rein-serted at a crystallization temperature.

(l) Synthetic compositions-isothermal runs, noquench

In simple isothermal experiments, the predomi-nant mechanism of nucleation of all three phases,and the only one for diopside and anorthite, washeterogeneous on the Pt wire. The results aresummarized in Table I and Figure 2. The morpholo-gies are somewhat similar to those observed inprevious studies (see Lofgren, 1980, for review),although the rectangular hour-glass and dendriticplates of anorthite and forsterite which we found tobe common at moderate AT do not seem to beobserved in natural basaltic compositions.

Although nucleation on the wire was alwaysdominant, forsterite also nucleated sometimes on

the surface of the bead, the plates usually lying flat.The proportion of nucleation on the wire to that onthe surface was quite variable from run to run. Insome cases the crystals on the surface can be seento have originated from a foreign particle, often a Ptfragment, but in others there is no obvious seedmaterial.

(2) Synthetic compositions-complex thermalhistory, no quench

In several previous studies of synthetic glass-forming systems (James, 197 4 ; Kalinina e t al., 1980 ;Klein er al., 1977; Strnad and Douglas, 1973; Kleinand Uhlmann, 1976), internal, presumably homoge-neous nucleation was observed, but at supercool-ings of 200'to over 700". It is known that the growthrates of diopside and anorthite maximize at about100" and decrease at larger A?'s (Leonteva, 1948,1949; Kirkpatick et al., 1979).It is possible, then,that nuclei generated at very high AI may not growto visible size in reasonable times. Therefore, thecornpositions Di65An35, Di5sAn5s and DiaaAnaaFol 2,after annealing at 1350-1400" C, were held for 20 to50 hours at temperatures of 400", 500', 600", 700o,800" and 900' C, then brought back near the liquidusfor growth of any nuclei produced at the lowertemperatures. The charges were not removed fromthe furnace until final quench and examination.

BERKEBILE AND DOWTY: NUCLEATION IN BASALTIC COMP)SITI)NS EE9

Fig.2. Simple isothermal runs. The diameter of the wire loop is always about 5 mm. The small spherical objects adhering to thewire are air bubbles. (a) Pointed needles of diopside in Di-Ana6 held for 1.5 hours at 1200'C (AZ: 65). (b) Diopside crystals,transitional from pointed needles to dendrites, in Di65An35 held for 3 hours at l160'C (AZ: ll5'). (c) Wedge-shaped diopsidedendrites in Di66Anas held for ll2ho:ur at I 140" C (AI : 125). (d) Forsterite plates in DiaoAnoaFo,2 run for 15 hours at 1250' C (Af =60).

None of these runs showed internal nucleation,although they all produced crystals of the liquidusphase on the Pt wire. These crystals were probablygenerated during heating and cooling-the beadsmust have spent the better part of an hour in the ATranges at which nucleation and growth are rapid.Runs of 20 hours or longer were thus not possible attemperatures of 1000" and higher, because rapidnucleation and growth of crystals on the wire

causes essentially complete crystallization. Manyruns were made in this temperature range forshorter times with the same result. i.e., no internalnucleation.

For a bead of 4 mm diameter, a 20 hour run timethus establishes the maximum rate of homogeneousnucleation for the temperature interval 400-900" Cat about 2 x l}-a cm-3 sec-l. For temperaturesabove 900o, rates could be as much as an order of

890

magnitude greater, since our observation timeswere shorter. These inferred maximum rates areconsiderably less than the measured maximum rateof 4 x 103 at 480'C in LizSizOs (James, 1974).

Although a complex thermal history did not pro-duce homogeneous nucleation, it did result in someadditional modes of heterogeneous nucleation. Inthe compositions Di6sAnas and Di65An35 crystals ofdiopside sometimes grew on the surface during theperiod of holding at low AI for growth. If there waslittle previous growth of diopside on the wire, suchsurface nucleation was rare and resulted in a fewisolated crystals. This type of growth may in somecases be due to surface impurities, perhaps dustparticles generated by friction between the furnacetube and the upper ceramic plug during heating andcooling. However, several observations suggestthat nucleation on the surface may be produced bymechanical stress. If a large volume of diopsidecrystals is generated on the wire, which may occurif the initial supercooling for "nucleation" is up toabout 250", the additional diopside crystals whichgrow on the surface during the second period of"growth" at lower AT usually occur in quasi-lineararrays, in which the grain size increases away fromthe larger diopside crystals. This mode of growthmay be due to the accumulation of stress near thesurface, due either to volume changes on crystalli-zation or to differential thermal expansion of crys-tals and liquid. This in turn may lead to cracks ordislocations in the glass which facilitate nucleation.Stresses are undoubtedly great in quenched beadswhich consist partly of crystals and partly of glass;such beads are usually observed to be cracked, andin some cases the cracking occurs hours or daysafter the bead is removed from the furnace.

Somewhat similar behavior was observed inDiaaAnaaFol2. Crystals of forsterite, sometimes instellate clusters, grew on the surface in the two-stage runs. These surface crystals were usuallysparse and did not form linear arrays like diopside.When the temperature of the first stage was about900-1000" C, abundant dendritic or spherulitic for-sterite grew on the wire. If beads containing suchgrowth were brought back to near the liquidus, insome cases cracks developed in the interior of thebead and additional forsterite crystals nucleated onthe cracks. Although some crystals were seen in theinterior which did not obviously nucleate on cracks,in no case did crystals grow in the interior unlessthere was already a significant volume of growth onthe wire. Again, differential volume effects arestrongly indicated.

BERKEBILE AND DOWTY: NUCLEATION IN BASALTIC COMPOSITIONS

In Di56An5s, if the bead was held at over 100' Af,diopside sometimes crystallized instead of anor-thite, usually in the form of spherulites. If thecharge was then held just below the anorthiteliquidus, anorthite plates nucleated on the diopsidespherulites (Fig. 4a). This temperature is above themetastable diopside liquidus, as shown by the factthat the diopside would dissolve in due course,leaving the anorthite plates suspended in the interi-or of the bead. No anorthite or diopside was ob-served to crystallize on the surface under theseconditions, nor did any other type of thermal his-tory ever cause anorthite to crystallize on thesurface if the bead was not removed from thefurnace.

Some experiments were made specifically to testfor crystals nucleating on one another. The possibil-ity of diopside nucleating on anorthite was presum-ably tested in the composition Di66Anas. First, totest for the nucleation of diopside and anorthite onforsterite, the composition DiaaAn+qFoD was heldat a temperature at orjust below the ternary eutec-tic (1270") for fairly long periods (over 12 hours). Alimited amount of forsterite crystallized on the wireunder these conditions, usually in the form of largeskeletal crystals or hopper plates. Then the chargewas taken to about 40' lower than the eutectic,causing the crystallization of diopside and anor-thite. Most of the diopside seemed to nucleateindependently on the wire, but in several casessmall pointed dendrites of anorthite grew on theforsterite. Two special compositions were made upin order to test the possibility of nucleation offorsterite on other phases: Dia2An51Fo7, in whichthe metastable forsterite liquidus should be justbelow the stable anorthite liquidus; andDi62An32Fo6, in which the metastable forsteriteliquidus should be just below the stable diopsideliquidus. At moderate supercoolings, forsterite nu-cleated independently on the wire in both composi-tions (in addition to the stable liquidus phase), butno clear cases offorsterite nucleating on diopside oranorthite were seen.

(3) Synthetic compositions-removed from thefurnace

Nucleation of all three phases could be inducedby removing the bead from the furnace at theannealing temperature and then reinserting it atsome small A?. If the bead was not touched, and ifit contained no crystals already, crystals formed bythis mode of nucleation were rather sparse, thereusually being less than about fifteen new crystalliza-

BERKEBILE AND DOWTY: NUCLEATION IN BASALTIC C)MP)SITI)NS

Fig. 3' Runs quenched to room temperature, without handling, then reinserted in the furnace for crystallization. All crystals in thebeads are on the surface. (a) Diopside in DiuoAna6 held for 3 hours at 1240" C (LT : 25"). (b) Anorthite in Di56An-5s held for 45 minutesat 1260" C (A? : 25') (c) Forsterite in DiaoAnaaFo,2 run for l0 hours at 1270" C (AT : 40). (d) Close-up taken in polarizingmicroscope(crossedpolars)offorsteri teplatesandsparsediopside(?)starsinDiaaAnaaFolrrunforl0minutes at1240" C(AZ:70).Height of the field of view is about 2 mm.

891

tion centers on the surface. The dominant phasewhich grew was in all cases the stable liquidusphase, although in the composition DiaaAnaaFol2,diopside sometimes grew in addition to forsterite at

moderate AT (Fig. 3d). Anorthite and diopside havea tendency to form stellate clusters (Fig. 3a,b),while forsterite is normally in the form of individualskeletal or hopper plates (Fig. 3c,d). This type of

892

nucleation is most likely caused by the adherence ofdust particles, generated by friction between theupper plug and the muffie, but mechanical stressmay also be involved.

More abundant nucleation was caused if the beadwas touched with a finger before reinsertion. Thisled to dense crystallization over the entire surface,usually consisting of individual crystals at low AT,and stellate clusters at higher A?. Possibly the saltsin perspiration cause some chemical change in thesupercooled liquid, or act as a mineralizer. Even attemperatures below the liquidus, surface diffusionof NaCl or other substances must be rapid enoughto spread them over the surface within minutes.Another possibility is that the water from perspira-tion may act as a surfactant. In experiments onalkali feldspar compositions, Fenn (1977) showedthat the presence of water in the melt greatlyreduced the Af needed for nucleation.

Dependence of nucleation rate on superheating

Much incidental information on the effect ofannealing time at 1375-1400'C on nucleation wasaccumulated. There seems to be a tendency fornucleation to become more difficult as time goes onin beads that are recycled through many crystalliza-tion runs, or otherwise annealed for long periods athigh temperature. The time necessary for this proc-ess is usually on the order of a hundred hours totalat annealing temperatures. It seems most likely thatthis behavior can be attributed to changes in thephysical state of the wire. Wires in beads whichhave been recycled many times or otherwise held atannealing temperatures for long times develop pro-nounced grooves on their surfaces, which apparent-ly represent grain boundaries. Thus it is apparentthat some annealing of the platinum is taking place.

In contrast with the findings of Donaldson (1979)on nucleation of olivine, no evidence was found forshort-term effects of annealing time or temperatureon nucleation rate. However, the number of ourexperiments pertaining to this problem is not suffi-cient to allow any firm conclusions to be drawn.

Experiments on natural basalt

To verify that the nucleation phenomena de-scribed above, in particular the absence of homoge-neous nucleation, are not a peculiarity of syntheticcompositions, experiments were carried out on anatural diabase from Rocky Hill, New Jersey, atypical Triassic-basin basaltic rock. Analyses aregiven by Dowty and Berkebile (1982). The sample

BERKEBILE AND DOWTY: NL]CLEATION IN BASALTIC COMPOSITIONS

used had original grain size of about 2 mm diameterand contained no olivine.

Experiments were carried out both in 20 mlcrucible and in Pt-wire loops. For most of the Pt-wire experiments, the starting material was basaltmelted at 1350' C and ground repeatedly to obtain ahomogeneous glass. All preparation and crystalliza-tion runs were carried out in a furnace atmosphereof 10 parts COz to one part CO, which gives anoxygen fugacity between the quartz-magnetite-fayalite and magnetite-wtstite bufers at the tem-peratures of interest. The beads are brown andopaque to incident light, so sections were cut paral-lel to the wire loop and ground until translucent-usually to a thickness of about 0.5 mm.

When beads were quenched from 1350" andtouched before reinsertion at supercoolings of up to50o, abundant pyroxene needles grew on the sur-face, and they persisted if the bead was annealed atup to 1200". This temperature is presumably theliquidus and pyroxene the primary liquidus phase.However, when the beads were annealed at 1350'and taken directly to the crystallization temperaturewithout being removed from the furnace, olivinewas at least as common as pyroxene (Fig. ab-d).Both phases nucleated exclusively on the wireunder these conditions. Morphologies are similar tothose previously observed in crystallization experi-ments on natural basalts (e.9. Donaldson, 1976;Lofgren, 1980), rather than those seen in the syn-thetic compositions (Table 1). When the startingmaterial was homogeneous glass, plagioclase wasnot observed unless the bead had essentially com-pletely crystallized. Plagioclase nucleated on thewire at the top of beads when sintered powderedbasalt was the starting material (Fig. 4d); such runsdifferentiate during melting (Dowty and Berkebile,re82).

In order to test the possibility that the high ratioof surface area (both wire and air interfaces) tovolume in the Pt-wire beads causes an artificialpredominance of heteogeneous nucleation, severalruns were made on the Rocky Hill composition in30 ml nickel crucibles. Most of these runs weredynamic cooling experiments, at cooling rates from2o to 50'per hour. Some were made in an atmo-sphere of argon and some in the 10/1 mixture ofCO11CO. The furnace atmosphere seemed to makeno difference in the phases crystallized or thetexture, and it is likely that the bulk of the liquid didnot come to redox equilibrium with the atmosphere(see Dowty and Berkebile, 1982). These runs wereactually done before the experiments in platinum-

BERKEBILE AND DOWTY: NUCLEATION IN BASALTIC COMPOSITIONS

Fig. 4. (a) Anorthite plates growing on diopside spherulites in Di5eAn5s held for Vz hour at 900' C (AZ : 385'), then % hour at1270' C (LT : 1 5). (b) Thick section of bead of Rocky Hill diabase, transmitted light, uncrossed polars. All crystals in this bead areolivine. Starting material homogenized glass, held for l6 hours at 1125" C (AZ: 75"). The long thin bright object running verticallythrough the center is a crack in the section. Air bubbles appear as bright circles. (c) Same, held for 6 hours at l 100" C (A? : 100).One large, wedge-shaped olivine dendrite is present on the lower part ofthe wire to the right ofcenter. (d) Starting material crushedrock, held for 2 hours at I 150' C (AI : 50'). Olivine crystals are all in the lower half and have elongated, chainlike shape, as in (b).Pyroxene crystals appear dark and feathery and occur in the right halfofthe bead, both top and bottom. The bright, feathery clusterjust below the top and to the left of center is plagioclase

893

wire loops described above; initial runs used largechunks of the rock, while later ones were madefrom homogenized glass. There were very impor-tant differences between these two types of runs, in

that the runs made from rock chunks differentiatedduring melting (Dowty and Berkebile, 1982), but thenucleation behavior seems to be the same.

In all these runs, olivine clearly nucleated on the

894

walls and bottom of the crucible, and grew inwardas long skeletal blades (See Dowty and Berkebile,Fig. 1). Pyroxene usually crystallized as curved,feathery dendrites and it fills in the intersticesbetween olivine crystals. The morphologies of oli-vine and pyroxene are very similar to those ob-served previously in laboratory experiments onbasalts (e.9. Donaldson, 1976; Lofgren, 1980). Inthe runs made from glass, and in the lower parts ofruns made from rock chunks, only olivine andpyroxene are present. In the runs made from rockchunks, the liquid became enriched in the plagio-clase component in the upper part (Dowty andBerkebile, 1982), and this part shows only plagio-clase and pyroxene. The plagioclase is usually inthe form of hollow prisms, as observed previouslyin natural basalts (e.g. Crawford,1973) and synthet-ic charges (Lofgren, 1974). Some plagioclase cer-tainly nucleates on the wall, but other crystalsappear, in thin section, to be coming down from thesurface (gas interface). It is not clear whether thesecrystals actually nucleated directly on the surface,or whether crystallization spread across the surfaceor near it from an origin on the walls.

One run was made from homogenized glass in analumina crucible at l0"/hour. In this run, olivine wasentirely absent, and long plagioclase and pyroxeneneedles grew inward from the walls and bottomsurface of the crucible (the appearance is somewhatsimilar to Figure lc of Dowty and Berkebile, 1982).The diference between alumina and nickel cruci-bles may be due in part to Al2O3 and NiO respec-tively being incorporated in the liquid adjacent tothe walls. However, slow diffusion rates imply thatsuch compositional differences could not have ex-tended very far from the walls, and the persistenceof the difference in phases crystallizing all the wayto the center of the charges demonstrates that eitherplagioclase or olivine is metastable.

Summary of nucleation results-implications forlaboratory studies

In detail the observations on nucleation form alarge and confusing mass; in many cases we haveinsufficient data to elucidate all the phenomenaencountered, and many of our interpretations arespeculative. Nevertheless, the most important re-sult is quite unambiguous; no clear example ofhomogeneous nucleation was observed in any runon synthetic or natural basalt compositions, despiteconcerted efforts to cause it under a wide variety ofconditions. When the charges were completely mol-

BERKEBILE AND DOWTY: NUCLEATION IN BASALTIC COMPOSITIONS

ten, and not subjected to mechanical stresses orcontamination, pyroxene, feldspar and olivine nu-cleated almost exclusively on the Pt-wire, or Ni orAl2O3 crucible, and when not nucleating on thecontainer they nucleated on the gas interface. An-nealing the charge for very long periods (ca. 100hours) seemed to depress the rate of nucleation onthe wire, but did not cause any other mechanism toappear; indeed, longer annealing seemed to elimi-nate the nucleation of forsterite on the surface,which is perhaps understandable if the latter is dueto impurity particles on the surface.

In previous dynamic crystallization experimentsin which the charge was not completely molten,nucleation probably occurred in most cases onsuspended crystals. However, if no actual remnantcrystals were present, we infer that nucleationprobably occurred on the container. Except for thestudy of synthetic diopside composition by Kirkpat-rick et al. (1981), in which forsterite was observedto nucleate on the wire, none of the recent dynamiccrystallization studies acknowledged the likelihoodof nucleation on the container. It seems that thenature of the nucleation mechanism was simply notobserved in these studies. There are two probablereasons for this. (1) The natural basalt compositionsstudied previously, being opaque, were examinedonly in thin section, which in the case of the beadsin Pt wire loops was usually cut perpendicular to thewire. Our results indicate that nucleation may origi-nate from only a few points on the wire (e.9. Fig.2b) and the probability of observing these points ina thin section is small. (2) Most runs were examinedonly after essentially complete crystallization,which further tends to obscure nucleation relations.

In some studies (e.g. Lofgren et aI., 1978) it wassuggested that there may be sub-microscopic incipi-ent nuclei which can exist stably or metastablyabove the liquidus, and which can cause internal,but not strictly homogeneous nucleation. Since ourexperiments show that there is no internal nucle-ation if the charge is completely molten, this hy-pothesis appears very improbable.

It is perhaps understandable that previous inves-tigators would in some cases have reserved judg-ment on the nature of nucleation mechanism, butthe problem is one which simply cannot be evaded;reproduction of the crystallization of natural rocksin the laboratory obviously depends on duplicationof all conditions, and if the nucleation mechanism isdifferent, inferences derived from comparison ofnatural and laboratory textures may not be valid. In

BERKEBILE AND DOWTY: NUCLEATION IN BASALTIC COMPOSITI)NS E95

particular, quantitative cooling rates of the naturalrocks derived by comparison of grain sizes cannotbe expected to be accurate.

Some previous nucleation results require specialcomment. Donaldson et al. (1975b) did suggest thatthe relatively minor differences in texture betweentheir runs from glass in Pt wire loops and those ofWalker et al. 1976) on a similar composition in ironcapsules, were due to heterogeneous nucleation oniron droplets within the liquid in the runs of thelatter, but Donaldson et al. (1975b) concluded thatnucleation in their own runs was probably ,.sponta-

neous" or internal. Nabelek et al.'197$ nucleatedplagioclase from glass which had been homoge-nized, then ground and pressed into pellets andinserted into the furnace at a temperature below theliquidus. They concluded that unmelted glass parti-cles acted as nucleation sites for plagioclase. How-ever, it seems more likely that the mechanism ofnucleation in these runs was the same as that in ourruns in which the glass bead was quenched fromabove the liquidus, i.e. crystals nucleated in theglass as a result either of mechanical stress orcontamination. The grinding of the glass by Nabe-lek et a/. could have introduced both stress andcontaminants. Bianco and Taylor (1977) likewiseinserted pressed pellets of homogenized and groundglass into the furnace below the liquidus for anneal-ing before starting dynamic crystallization runs.Despite rather small supercooling at these tempera-tures, the prior treatment of th€ glass must havecaused near-instantaneous nucleation, probablythroughout the bead. Donaldson (1979) suggestedthat olivine probably nucleated homogeneously inhis isothermal experiments in Pt-wire loops onnatural basaltic and ultramafic compositions. How-ever, he did not make thin sections, but onlycrushed the beads and examined the fragments inoil to determine if crystallization had taken place.There is no strong theoretical reason to believe thatthe systematic effects of annealing temperaturewhich he found could not be a result of heteroge-neous as well as homogeneous nucleation.

Since our own experiments did not produce ho-mogeneous nucleation, and we do not find convinc-ing evidence for it in previous experiments, wetherefore tentatively conclude that homogeneousnucleation has not been observed in isothermal ordynamic crystallization of laboratory charges ofbasaltic or ultramafic composition. There is alwayssome danger in making such generalizations inkinetic studies, but it seems clear at least that our

observations shift the burden ofproofon to anyonewho contends that he has observed homogeneousnucleation. Such observations will have to meettwo main criteria; it must be shown in three dimen-sions (not in thin sections only) that there is noconnection of the crystals with the container orsurface; and it must be shown that the liquid washomogeneous and contained no remnant crystals,or at least the liquid must have been annealed at atemperature which was definitely above the liqui-dus for several hours.

Nucleation mechanisms in natural rocks

Our results, combined with previous studies ofsynthetic and natural silicate compositions, suggestthat homogeneous nucleation is not to be expectedin natural basaltic rocks. The chief uncertainty atpresent seems to be whether nucleation in basalticrocks is primarily external, with crystallization orig-inating on wall rocks or other surfaces, or internalheterogeneous nucleation, on crystals or other ma-terial suspended in the melt. Perhaps the leastambiguous situation is that in which the texture isdominated by elongated "quench" crystals growinginward from the walls or upward from apparentcumulate crystals, as found in the "perpendicularfeldspar", crescumulate and harrisitic textures oflayered intrusions (Wager and Brown, 1968; Don-aldson, 1977), the spinifex textures of ultramaficextrusive rocks (Pyke et al., |973;Donaldson, 1974)and the comb or Willow-Lake layering of granodio-ritic rocks (Lofgren and Donaldson, 1975). Thesetextures have well-ch atacterized laboratory analogsand may be considered to arise from heteroge-neous, external nucleation.

In some cases, external nucleation may not be soobvious. The textures of lunar pyroxene-phyric orquartz-normative basalts were fairly closely repro-duced by dynamic crystallization experiments usingglass as starting material (Lofgren et al., 1974).ltseems likely from the results of our own experi-ments that the prominent pyroxene phenocrysts inthe dynamic crystallization experiments nucleatedon the Pt wire. This inference is reinforced by thefact that using iron capsules (Grove and Walker,1977) instead of Pt-wire loops in dynamic crystalli-zation experiments on similar compositions led tosignificantly different grain sizes. Some of our ex-periments show that crystallization which origi-nates on the wire or surface does not always yieldcrystal arrays which are perpendicular to the nucle-ating surface. The examples shown in Figure 5a and

896 BERKEBILE AND DOWTY: NUCLEATION IN BASALTIC COMPOSITIONS

Fig. 5. Thick section ofbead ofRocky Hill diabase, starting material homogenized glass, quenched to room temperature, touched

andheldat l l90'Cfor I hour. (b) Similar, heldat 1185" Cfor I hour. (c)Thinsectionof lunarsample 15125, crossed polars. Widthofthe field of view is 3 mm.

b were produced by growth at rather low supercool-ing in beads which had been quenched and touched.Under these circumstances, growth seems to pro-ceed inward by continued heterogeneous nucleationwith a more-or-less random orientation of the elon-gated pyroxene crystals. There seems to be consid-erable resemblance between such arrays of crystalsand the phenocrysts in the lunar basalts (Fig. 5c).This mechanism of external nucleation is certainlymore plausible than homogeneous nucleation of thephenocrysts which was suggested by Dowty et al.

(1974). However, internal heterogeneous nucle-ation, on previously existing crystals (chromite) oriron droplets (Donaldson et al., 1975) is also adistinct possibility.

Laboratory crystallization experiments on basal-tic compositions very often produce porphyritictextures, perhaps more often than observed innatural basalts (e.g. Walker et al., 1979; Lofgren,1980). It certainly seems significant that such tex-tures are more common when controlled cooling isinitiated from a temperature above the liquidus than

from one below it. Our experiments permit thespecific suggestion that in some cases, laboratory-produced porphyritic textures are due to externalnucleation on the container. It would not be war-ranted, however, to conclude that porphyritic tex-ture in synthetic or natural rocks in general is anindication of external nucleation.

Most basalts and gabbros show neither obviouselongated quench crystals at the margins, nor therather dense interlocking fabric of elongated pyrox-ene phenocrysts of the lunar pyroxene-phyric ba-salts. They tend to have fairly equant crystals rightup to the margins, the crystals simply becomingsmaller in most cases. Such observations, com-bined with the fact that basalt textures are generallymost closely reproduced in dynamic crystallizationexperiments when the melt is annealed below theliquidus, or for only a limited time above theliquidus (Usselman and Lofgren,1976; Bianco andTaylor, 19771'Lofgren, 1977; Lofgren et al., 1978;Nabelek et al., 1978; Walker et al., 1978), have ledLofgren (1980), among others, to suggest that inmost cases crystals in basaltic and intermediaterocks nucleate on pre-existing crystals within theliquid. The study by Kirkpatrick (1977) of parriallycrystalline Hawaiian lava lakes, in which crystalswere usually found in clusters and nucleation wasinferred to be continuous, is at least consistent withthis conclusion, although since these lakes appearto be crystallizing from the top, it also seems quitepossible that nucleation was originally external.

The phenomenon of "lateral accretion" proposedfor batholithic rocks (Bateman and Chappell, 1976),while not absolutely requiring external nucleation,would certainly seem to be facilitated by it. For thebulk composition of a pluton to change as it crystal-lizes inward requires extensive material exchangeover long distances. Diffusion in molten silicates istoo slow to accomplish this unaided, so that somesort of convection or other liquid motion is re-quired. This in turn seems to require that the newlyaccreted crystals be rather firmly attached to thewalls. The generally greater facility of mafic crys-tals to nucleate may assist in this type of differentia-tion (Naney and Swanson, 1981).

The conclusion that homogeneous nucleation isabsent or very rare in basaltic rocks is not necessar-ily applicable to magmas of granitic composition, inwhich laboratory experiments have produced inter-nal nucleation. For example, Swanson (1977) showsphotographs of quartz crystals, grown in syntheticgranite, which are clearly isolated plates with no

E9'l

connection to other crystals or to the container. Thereason for the difference between basaltic and gra-nitic compositions may be the presumably slowerdiffusion rates of Si and Al in the granite, whichmight have a more drastic effect on growth thannucleation, or the considerably higher water con-tents of the granitic liquids. Quartz and feldsparphenocrysts in porphyritic granitic rocks sometimesform in clusters, which has often been interpretedas evidence for synneusis (Vance, 1969), but ismore simply explained as the result of heteroge-neous nucleation (Dowty, 1980b). However, euhe-dral crystals also commonly occur in isolation. Agood example is the occurrence of orthoclase phe-nocrysts in a rhyolitic rock near Goodsprings, Ne-vada, originally described by Drugman (1938). Wecollected and examined a suite of phenocrysts fromthis locality and qualitatively verified Drugman'sconclusion that the great majority are singleuntwinned crystals of a remarkably uniform euhe-dral habit. The remainder of the crystals are twins,almost all of which consist of two individuals withreasonably symmetric development and whichprobably grew from twinned nuclei (Buerger, 1945).These facts tend to suggest homogeneous nucle-ation for this particular rock.

Conclusions

Most recent studies of nucleation seem to haveraised more questions than they have answered,and our work is no exception. However, our con-clusion that nucleation occurs exclusively on exter-nal surfaces when the charge is completely moltenis a firm one, and it is clear that some previousresults should be reinterpreted or reinvestigated.Our results also provide further confirmation thathomogeneous nucleation in natural basaltic rocks isimprobable. We find that heterogeneous nucleationmechanisms in the laboratory are quite varied, andvaried (though probably different) mechanisms evi-dently operate in nature.

AcknowledgmentsWe thank Maria Borscik for atomic absorption analyses. This

work was supported in part by NSF Grant 76-22543.

References

Baker, J. C. and Cahn, J. W. (1971) Thermodynamics of Solidifi-cation. In Solidification, American Society for Metals, MetalsPark, Ohio.

Bateman, P. C. and Chappell, B. W. (1976) Crystallization,fractionation and solidification of the Tuolumne intrusive

BERKEBILE AND DOWTY: NUCLEATION IN BASALTIC COMPOSITIONS

898 BERKEBILE AND DOI,IITY: NUCLEATION IN BASALTIC COMPOSITIONS

series, Yosemite National Park. Geological Society of Ameri-ca Bulletin. 90. 465-483.

Bianco, A. S. and Taylor, L. A. (1977) Application of dynamiccrystallization studies: Lunar olivine-normative basalts. Pro-ceedings of the Lunar Science Conference, 8th, 1593-1610.

Buerger, M. J. (1945) The genesis of twin crystals. AmericanMineralogist, 30, 469-482.

Burnett, D. G. and Douglas, R. W. (1971) Nucleation and crys-tallization in the soda-baria-silica system. Physics and Chemis-try of Glasses, 12, ll7-124.

Crawford, M.-L. (1973) Crystallization of plagioclase in marebasalts. Proceedings of the Lunar Science Conference, 4th,705-717.

Donaldson, C. H. (1974) Olivine crystal types in harrisitic rocksofthe Rhum Pluton and in archean spinifex rocks. GeologicalSociety of America Bulletin, 85,172l-1726.

Donaldson, C. H. (1976) An experimental investigation of oli-vine morphology. Contributions to Mineralogy and Petrology,57,186-213.

Donaldson, C. H. (1977) Laboratory duplication of comb layer-ing in the Rhum Pluton. Mineralogical Magazine, 41,323-336.

Donaldson, C. H. (1979) An experimental investigation of thedelay in nucleation of olivine in mafic magmas. Contributionsto Mineralogy and Petrology, 69,21-12.

Donaldson, C. H., Williams, R. J. and Lofgren, G. (1975a) Asample holding technique for study of crystal growth in silicatemelts. American Mineralogist, ffi , 324-f26.

Donaldson, C. H., Usselman. T. M.. Williams. R. J. and Lof-gren, G. E. (1975b) Experimental modeling of the coolinghistory of Apollo 12 olivine basalts. Proceedings of the LunarScience Conference. 6th. 843-869.

Dowty, E. (1980a) Crystal growth and nucleation theory and thenumerical simulation of igneous crystallization. In R. B. Har-graves, Ed., Physics of Magmatic Processes, p. 419479.Princeton University Press, Princeton, New Jersey.

Dowty, E. (1980b) Synneusis reconsidered. Contributions toMineralogy and Petrology, 24, 7-29.

Dowty, E. and Berkebile, C. A. (1982) Differentiation and diffu-sion in laboratory charges of basaltic composition duringmelting experiments. American Mineralogist, 67, 900-906.

Dowty, E., Keil, K. and Prinz, M. (1974) Lunar pyroxene-phyricbasalts; Crystallization under supercooled conditions. Journalof Petrology, 15, 419-453.

Drugman, J. (1938) On some unusual twin-laws observed in theorthoclase crystals from Goodsprings, Nevada. MineralogicalMagazine, 25, l-14.

Fenn, P. M. (1977) The nucleation and growth of alkali feldsparsfrom hydrous melts. Canadian Mineralogist, 15, 135-161.

Gibb, F. G. F. (1974) Supercooling and the crystallization ofplagioclase from a basaltic melt. Mineralogical Magazine, 39,gl-653.

Grove, T. L. and Walker, D. (1977) Cooling histories of Apollo15 quartz normative basalts. Proceedings ofthe Lunar ScienceConference. Eth. l50l-1520.

James, P. F. (1974) Kinetics of crystal nucleation in lithiumsilicate glasses. Physics and Chemistry ofGlasses, 15,95-105.

Kalinina, A. M., Filipovich, V. N. and Fokin, V. M. (1980)Stationary and non-stationary nucleation rates in a glass of2NazO.CaO.3SiO2 stoichiometric composition. Journal ofNon-Crystalline Solids, 38139, 721-128.

Kirkpatrick, R. J. (1975) Crystal growth from the melt-a re-view. American Mineralogist, 60, 798-814.

Kirkpatrick, R. J. (1977) Nucleation and growth of plagioclase,Makaopuhi and Alae lava lakes, Kilauea volcano, Hawaii.Geological Society of America Bulletin, 88, 78-83.

Kirkpatrick, R. J. (1981) Kinetics of crystallization of igneousrocks. In A. C. Lasaga and R. J. Kirkpatrick, Eds., Kineticsof Geochemical Processes. Reviews in Mineralogy, Volume 8.Mineralogical Society of America.

Kirkpatrick, R. J., Gilpin, R. R. and Hays, J. F. (1976) Kineticsof crystal growth from silicate melts: anorthite and diopside.Journal of Geophysical Research, 81, 2565J57 l.

Kirkpatrick, R. J., Kuo, L.-C. and Melchior, J. (1981) Crystalgrowth in incongruently-melting compositions: programmedcooling experiments with diopside. American Mineralogist,66, 223-241.

Klein, L. C. and Uhlmann, D. R. (1976) The kinetics of lunarglass formation, revisited. Proceedings of the Lunar ScienceConference, 7th, lll3-l l2l.

Klein, L. C., Handwerker, C. A. and Uhlmann, D. R. (1977)Nucleation kinetics of sodium disilicate. Journal of CrystalGrowth. 42.47-51.

Leonteva, A. A. (1948) Issledovaniya lineinoi skovosti kristalli-zatsii v sisteme albit-anortit-diopsid. Zhurnal FizicheskoiKhimi.22. 1205-1213.

Leonteva, A. A. (1949) Vichisleniya lineinoi skovosti kristalli-zatsii plagioklazov. Dokladi Akademiya Nauk SSSR, 68, 145-r47.

Lofgren, G.E. (1974) An experimental study of plagioclasemorphology: Isothermal crystallization. American Journal ofScience. 264. 243-273.

Lofgren, G. E. (1977) Dynamic crystallization experiments bear-ing on the origin of textures in impact-generated liquids.Proceedings of the Lunar Science Conferenc e, 8th, 207 9-209 5.

Lofgren, G. E. (1980) Experimental studies on the dynamiccrystallization of silicate melts. In R. B. Hargraves, Ed.,Physics of Magmatic Processes, p. 487. Princeton UniversityPress, Princeton, New Jersey.

Lofgren, G. E. and Donaldson, C. H. (1975) Curved branchingcrystals and differentiation in comblayered rocks. Contribu-tions to Mineralogy and Petrology, 49,309-319.

Lofgren, G. E., Donaldson, C. H., Williams, R. J. Mullins, O.and Usselman, T. M. (1974) Experimentally reproduced tex-tures and mineral chemistry of Apollo 15 quartz normativebasalts. Proceedings of the Lunar Science Conference, 5th549-567.

Lofgren, G. E., Smith, D. P. and Brown, R. W. (1978) Dynamiccrystallization and kinetic melting of the lunar soil. Proceed-ings ofthe Lunar Science Conference, gth,959-975.

Nabelek, P. I., Taylor, L. A. and Lofgren, G. E. (1978) Nucle-ation and growth of plagioclase and the development oftextures in a high-alumina basaltic melt. Proceedings of theLunar Science Conference, 9th, 725-:741.

Naney, M. T. and Swanson, S. E. (1980) The effect of Fe andMg on crystallization in granitic systems. American Mineral-ogist,65, 639-653.

Pyke, D. R., Naldrett, A. J. and Eckstand, O. R. (1973) Archae-an ultramafic flows in Munro Township Ontario. GeologicalSociety of American Bulletin, 84, 955-978.

Strnad, Z. and Douglas, R. W. (1973) Nucleation and crystalliza-tion in the soda-lime-silica system. Physics and Chemistry ofGlasses. 14.33-36.

Swanson, S. E. (1977) Relation ofnucleation and crystal-growthrate to the development of granitic textures. American Miner-

alogist,62, 966-978.Usselman, T. M. and Lofgren, G. E. (1976) The phase relations,

textures and mineral chemistries of high-titanium mare basaltsas a function of oxygen fugacity and cooling rate. proceedingsof the Lunar Science Conference, 7 th, 1345-136i.

Vance, J. A. (1969) On synneusis. Contributions to Mineralogyand Petrology, 24, 7-29.

Wager, L. R. and Brown, G. M. (196E) Layered Igneous Rocks.Oliver and Boyd, Edinburgh.

Walker, D., Kirkpatrick, R. J., Longhi, J. and Hays, J. F. (1976)

899

Crystallization history and origin of lunar picritic basalt 12002:Phase equilibria, cooling rate studies and physical propertiesof the parent magma. Geological Society of America Bulletin,87, 6/;6-656.

Walker, D., Powell, M. A., Lofgren, G. E. and Hays, J. F.(1978) Dynamic crystallization of a eucrite basalt. Proceedingsof the Lunar Science Conference. 9th. 1369-1 391.

Manuscript received, August 24, l98l;acceptedfor publication, May 10, 1982.

BERKEBILE AND DOWTY: NUCLEATION IN BASALTIC COMPOSITIONS