INTRODUCTIONEver since volcanoes were recognized as emitters of moltenrock, the subsurface structure and intimate workings ofmagmatic systems have been a subject of mystery and fas-cination. Aristotle suspected that Stromboli’s prolongedagitation might be due to enflamed coal at depth. With thegradual recognition of the meaning of dikes, sills, and plutons,all the disciplines of igneous petrology or magmatologywere born. The basic problem of understanding the physicaland chemical evolution of magma is that so very little ofany magmatic system, active or extinct, has ever been seen;or if it has been seen it has not been recognized for what itis. The science of magmatism has thus developed along thelines of those parts of magmatic systems offering easiestaccess to examination. These areas of study have developedinto mature, highly technical and quantitative fields ofresearch and thought, now so erudite that it is often noteasy to move freely through the literature of these adjoin-ing fields. And researchers in each field have a distinct viewof what magmatic systems are and how they function toachieve the perceived ends. One whole class of volcanolo-gists, for example, is mainly interested in the physicalaspects of magma delivery, degassing, vesiculation,mechanical fragmentation, and eruption. The chemicalcompositions of the volcanic products are not necessarily ofmuch interest or meaning to this puzzle. Other volcanolo-gists are primarily interested in the chemistry of the products,and some of these only in the nature of the chemical varia-tions from one lava to the next, which might reveal theprocesses leading to the observed chemical diversity. Stillothers are interested only in the most primitive composi-tions that might reveal the ultimate parental magma or per-haps shed light on the evolution of the mantle itself. Eachdiscipline, in concert with the research results gained, con-

jures up detailed conceptualizationsof the nature of the underlying sys-tem. Volcanologists, plutonists, sillpeople, dike people, and layeredintrusionists each have strict ideasof the general magmatic system towhich their rocks are attached. Inspite of this method of scientificinquiry having achieved majoradvances over the past two hundredyears, in the drive to increasingdetail, major, seemingly insur-mountable problems have arisen,forcing the need for an overarch-ing conceptualization useful to allbranches of magmatic study.

HISTORICAL PERSPECTIVEThe vast diversity of igneous rocks on Earth’s surface, thestriking compositional bimodality between ocean basinsand continents, and the gross planetary structure itself havealways driven the need to understand the origin and evolu-tion of magma. From the earliest melting experiments usinggun barrels, James Hall recognized in 1798 the temperature-dependent sequence of appearance of minerals during crys-tallization (see Dawson 1992). The discovery of the tech-nique of making thin sections by H.C. Sorby and the splendiddevelopment of petrography by Zirkel and Rosenbushallowed intimate visual access through microscopy toboundless natural experiments exemplefact in plutonic andvolcanic rocks. Coupled with analytical advances in rockchemistry, this led to the clever insight of Fouque (1879)that the physical separation or fractionation of feldsparfrom Santorini lavas might produce silica enrichment. Andby the end of the nineteenth century, the general nature ofmagma crystallization was known so well that Cross,Iddings, Pirsson, and Washington laid out a recipe (CIPWnorm), accurate to this day, for the crystallization of anyrock given its chemical composition. Little to nothing wasyet known of the actual phase equilibria involved, but thekind and modal amount of each mineral phase were wellpredicted. The endless confusion over the equivalence ofvolcanic and plutonic rocks due to differing degrees of crys-tallization was erased. The scene was now set for NormanBowen.

Magmatology at the beginning of the twentieth centurywas exceedingly well developed based on field relations andpetrography. The insights and understandings into physicalprocesses leading to the diversity of rocks were masterful.But two important conditions led to a major revolution.First, with a few notable exceptions, the existing generationof petrologists faded into retirement. Second, devoid of theconcepts of physical chemistry and physics of fluids, thefield was at an impasse. The lightning rod was Bowen, who

E L E M E N T S , V O L . 2 , P P . 2 8 7 – 2 9 2 OCTOBER 2006

Bruce D. Marsh1

1 M.K. Blaustein Department of Earth and Planetary SciencesJohns Hopkins UniversityBaltimore, Maryland 21218, USAE-mail: bmarsh@jhu.edu

Dynamics of MagmaticSystems

287

An intimate physical and chemical interplay between crystals and meltin magmatic systems gives rise to a vast diversity of igneous rocks andthe very structure of terrestrial planets. Yet the actual physical means

by which this happens is unclear. The long-standing notion of crystals nucleating,growing, and settling ad infinitum from the interior of large pools of magmato eventually form continental rocks is foundering. Processes operating at thesmallest scales within marginal solidification fronts and in mingling crystalslurries throughout highly integrated, vertically extended mush columns giverise to planetary-scale effects.

KEYWORDS: magma dynamics, planetary differentiation, solidification fronts,

crystallization, magma evolution, Sudbury melt sheet

began systematic experimental investigations on silicatephase equilibria. Prior to this time the volumetric prepon-derance and compositional contrasts of basalt and granitewere suspected to be due to eutectic-like “pole” composi-tions separated by a major thermal divide. Magmas natu-rally crystallized to one side or the other. Bowen rapidlyfound no pole compositions but, instead, a continuum ofsliding compositions for most minerals and host magmas.The most primitive MgO-rich, SiO2-poor mantle rock couldbe genetically connected to SiO2-rich continental granitethrough the progressive separation of ever-lower-tempera-ture minerals. Thus was born—in the days of Darwinianawakening and the Scopes monkey trial—the concept of a“liquid line of descent” connecting one magma to anotherthrough an endless chain of crystal fractionation. The physicalseparation of crystals from solutions to fractionate compo-sition was a widely known, commonly used technique inphysical chemistry. Madame Curie had used it extensivelyto discover polonium, actinium, and radium. H.A. Harkeralso actively used this principle to explain comagmaticsequences of rocks (Wilson 1993). But it was Bowen whoput it on firm ground, as it has remained to this day. He didthis not just by conducting experiments. He was a first-classgeologist with a keen analytical mind. He marshaled exper-iments, principles, and field relations to build an unassail-able theory of the evolution of all igneous rocks. He care-fully and quantitatively evaluated all competingmechanisms (e.g. Soret diffusion, gas fluxing, magma mix-ing, etc.) and showed them to be ineffective. Bowen (1915)is a must read, and his 1928 tract, The Evolution of theIgneous Rocks, remains an unsurpassed masterpiece.

Bowen nucleated a century of research into every possibleway that minerals can be separated on a phase diagramfrom the host melt to yield the desired chemical offspring.The physical process has always been the same: crystalsgrow in magmas and settle out under the action of gravity,leaving behind a chemically refined melt (FIG. 1). The dic-tum has become McBirney’s Adage: “When in doubt, settleit out.” It is abundantly clear from the most meager massbalance considerations that the separation of crystals and

melt gives rise to the diversity of igneous rocks. There is noother process so pervasively operative. But how this actuallyhappens is much less clear. These doubts were around evenin Bowen’s day. Indisputable field evidence showing gran-ites grading imperceptively into metasediments andgneisses gave rise to a generation of debate on the “GraniteControversy.” Bowen himself saw the difficulty of generat-ing silicic magma from basalt: the needed late-stage differ-entiate is interstitially distributed in barely connected pock-ets at high degrees (80–90%) of crystallinity. He thoughtthat perhaps tectonic action might squeeze out the melt,but relative to solidification the rate is much too slow. Mostpetrologists to this day still think that basaltic magmas sim-ply go on and on growing and fractionating crystals in thecenters of pools of magma to yield rhyolite. The mystery ofthis process, along with others, has forced igneous petrologyto rethink its foundations.

SOLIDIFICATION FRONTSOddly enough, from the earliest times an abundance offield evidence seemed to suggest strong and pervasive crystalfractionation. Many straightforward intrusions, like theShonkin Sag Laccolith in Montana and the Palisades Sill inNew Jersey, show thick layers of coarse crystals on the floorand relatively thin basaltic upper margins. A reasonable sce-nario for solidification, still in use today, was put forth in1898 by USGS chemist George Becker. He postulated thatlaccoliths crystallize like over-chilled bottles of wine: just ascrystallizing wine becomes richer in alcohol, coarse crystalsgrow on the walls of the magma chamber, and the remain-ing melt becomes progressively richer in silica. Laccolithsand magma chambers being so large, they cool much fasterat the roof, nucleating and growing dense populations ofcrystals that settle and collect on the floor. With the dis-covery of large exquisitely layered bodies like Skaergaard,Stillwater, and Bushveld, the following mantra wasadopted: a crystal-free magma, instantly injected, com-mences cooling at the roof, nucleating and growing crystalsthat are deposited on the floor by periodic cascading con-vective currents; eventually the floor cumulates reach the

288E L E M E N T S 288 OCTOBER 2006

(A) The classical concept of a magma chamber wherecrystals nucleate, grow, and settle from the interior to

chemically fractionate the residual melt. (B) The same magma chamber

encased in marginal solidification fronts within which all crystallizationoccurs. The chemically fractionated residual melt is trapped within thefront and is normally inaccessible to extraction and eruption.

FIGURE 1

A B

289E L E M E N T S OCTOBER 2006

thin roof crystallites, “sandwiching” a thin horizon of sil-ica-rich differentiate. It seemed to work. But in these bigsystems, critical parts of the geologic picture are alwaysmissing. Where are the feeders? How was the magmaemplaced? All at once? Crystal free? Did convection occur?

In smaller systems, where the full picture was better known,things were not so clear. Many sills and laccoliths show nointernal sorting or structure of any kind. These can be dis-missed as being too small for slow cooling and effectivecrystal fractionation. But the Shonkin Sag laccolith, one ofthe best examples of layering, is only 70 m thick. In con-trast, many 300–400 m thick basaltic sills are featureless,whereas others of this size show remarkable internal order-ing. On broader scales, in highly active, large-volume vol-canic systems like Hawai’i and along ocean ridges, crystalfractionation and the production of siliceous materialshould abound. In fact, little is found. At Kilauea, for example,compositionally primitive (picritic) magmas charged witholivine crystals evolve chemically by dropping out olivine.Their composition evolves from ~25% MgO to 7% MgO andto ~52% silica, and then suddenly stops changing. Crystalfractionation ceases. Nothing approaching rhyolite (~70%SiO2) or even dacite (~65% SiO2) appears. The same is trueat ocean ridges. Only in large immobile systems, like Ice-land and Galapagos, are significant volumes of rhyoliteformed. Why?

A fundamental obstacle to understanding magmaticprocesses, as mentioned already, is that large pools of activemagma are inaccessible to long-term examination. Theclosest approximation is the Hawaiian Lava Lakes. Steep-sided basins in the landscape suddenly filled with lava todepths of over 100 m have furnished isolated pools wheremagma can be examined throughout solidification. Once alava lake fills, a crust begins forming. When the crust inMakaopuhi Lava Lake became strong enough, T. L. Wrightand co-workers had the foresight to drill progressively intothe magma as crystallization proceeded (e.g. Wright andOkamura 1977). The value of this work to understandingmagmatic processes cannot be overemphasized. What hap-pens during solidification is quite straightforward. A floodof lava carrying a mass of large olivine crystals pools to forma lake. After a brief (days to weeks) period of instability, acrust forms on top and bottom and advances steadilyinward. Some of the large olivines are caught up in the ini-tially rapidly advancing top crust, but as the rate of coolingslows with crust thickening and as crust advancementlessens, the olivines escape wholesale and settle to form athick cumulate pile on the floor. The lower solidificationzone suddenly becomes much thicker than the top one.Crystallization proceeds in the upper, and lower, inwardadvancing solidification fronts, exactly along the lines pre-dicted from heat conduction calculations. The hottest partof the system is located inward from the solidificationfronts. This inner melt is devoid of crystals and remains sountil the arrival of the solidification fronts, when solidifica-tion begins. [The same sequence (i.e. loss of crystals, etc.)occurs for magmas generated in the crust through basaltunderplating.] With no cooling in the center, there can beno crystallization and certainly no crystal fractionation.The melt composition is constant. Tiny crystals falling fromthe leading part of the upper solidification front arereheated and dissolve back into the magma.

The mystery of cumulate piles of crystals, thin upper solidifica-tion zones, and featureless bodies is solved. The answer simplyrevolves around the presence or absence of large crystals(phenocrysts) in the initial magma. The Shonkin Sag Lac-colith was formed from magma laden with ~35 vol% phe-nocrysts, and the Palisades Sill arose from repeated injec-tion of phenocryst-bearing magma (Gorring and Naslund

1995). Thick featureless sills, like the massive Peneplain Sill(350 m x 100 km) in the McMurdo Dry Valleys, Antarctica,formed from phenocryst-free magma, and the underlyingBasement Sill is pervasively layered because its magma wasladen with large orthopyroxene crystals (Marsh 2004). Thecorollary of this simple phenocryst axiom is that large exot-ically layered intrusions form from prolonged trains ofdeliveries of varieties of magma, some exceedingly ladenwith phenocrysts, some devoid of phenocrysts. As a largebody requires more deliveries to build it, the probability isthen greater that some deliveries will be phenocryst rich.The rich and varied nature of the deliveries themselves isdue to the diverse magmatic environments, structural insta-bility, style of dynamic loading and unloading, and spa-tially integrated nature of the underlying magmatic mushcolumns within which the deliveries are spawned (Marsh2004). In size of magmatic bodies, it’s not slow cooling, butthe number and nature of magmatic deliveries that lead toexoticness. The style of cooling has little to do with the out-come. All sheets of magma cool in a similar manner, regard-less of depth of burial. Modest variations on this themearise from the shape of the container holding the magma—boxes with slanting walls, funnels, and slots each have dis-cernable influences on filling, layering, and cooling. Buthow then does differentiation take place?

The most obvious means of differentiation is by massivecrystal settling during emplacement. The loss of the phe-nocryst load, which overall can give the magma a primitivebulk composition, suddenly produces an abundance of rel-atively fractionated crystal-free melt. This is punctuated dif-ferentiation. It is dramatic, but limited in scope. Now thechemical evolution of the magma stalls. Further crystallizationproceeds only in the solidification fronts, encasing theentire body (FIG. 1). The evolution of the body is tracked bythe behavior of the solidification fronts (FIG. 2), which aredefined by a bundle of isotherms contained between thesolidus and liquidus, the beginning and end of melting. Thehotter the isotherm the faster it advances, thickening thesolidification fronts with time. In magmas of the upper

A basaltic solidification front depicting the change incomposition and viscosity of the interstitial melt with

position or crystallinity within the solidification front. The highly silicicmelt resides within the strongest part of the front.

FIGURE 2

290E L E M E N T S OCTOBER 2006

crust, solidification fronts sweep inward, producing feature-less bodies. The highly refined, differentiated or fractionatedmelt is locked in strong material deep in the solidificationfronts, inaccessible to extraction and eruption. This isBowen’s enigma (Bowen 1947).

Compositions plotted on a phase diagram form a differen-tiation array that progresses to a point and suddenly stops(FIG. 3). From phase equilibria alone there is no obviousexplanation for the truncation of differentiation. The key todeciphering this stop rests in the dynamics of solidificationfronts and especially the realization that temperature isalways tied to space or position in the magma. A movementin temperature and composition space on a phase diagramis a movement in position space in magma. Droppingbelow the liquidus is equivalent to entering the solidifica-tion front where differentiation is greatly restricted if notcurtailed. This also explains the conundrum at Kilauea overthe inability of lavas erupted from the summit to differen-tiate beyond ~7 wt% MgO and ~52% SiO2. Olivine-ladenmagma stalling anywhere during ascent drops its load ofcrystals and fractionates to the composition defined by theleading edge of the solidification front (i.e. the liquidus),which for these tholeiitic magmas is at the magic composi-tion of ~7 wt% MgO and ~52% SiO2. This also explains the

origin of oceanic crust. But it doesn’t explain the origin ofcontinental crust. To proceed beyond this point, specialprocesses have to operate within the solidification frontsand within the continental crust, which can be appreciatedby considering a large-scale experiment.

SUDBURY: THE ULTIMATEMAGMATIC EXPERIMENTA massive ~12 km meteorite hit Earth 1.85 billion years ago.Within two minutes, it penetrated the entire continentalcrust and formed a transient cavity ~30 km deep and ~90km in diameter. Within five minutes, the transient cavityrelaxed to form a multi-ring crater ~200 km in diameterfilled with 3 km of superheated magma. Unlike magmasgenerated deep within Earth, which are always at or belowthe liquidus, impact-generated melts can reach almost anytemperature. For the Sudbury impactor the temperaturemay have in places reached ~2500°C, well beyond the pointof vaporization of typical silicate melts, i.e. ~2000°C. Theimpactor itself vaporized, but the main body of the melt at~1700°C was much cooler, although still strongly super-heated relative to the liquidus at ~1200°C. All solids in themelt vanished. The stage was then set for the ultimate mag-matic experiment: a large sheet-like (3 km x 200 km) volume(~30,000 km3) of magma free of crystals had been emplacedinstantaneously. What happened?

Impacts are marked by massive fragmentation (brecciation)of the target rocks. At Sudbury, flash melting of these brec-cias produced a heterogeneous initial magma reflecting thenature of the local crust. Blobs, globs, and dollops of slightlydifferent compositions formed a confused magmatic emul-sion due to slight density and viscosity differences reflectingchemical composition. Emulsions operate much like sedi-mentation, except at Sudbury, parcels rapidly rose and felldepending on chemical composition. In a short time(months to years) the impact emulsion separated into twomassive layers, one silica rich (~70% SiO2) and one silicapoor (~55% SiO2). Through instant melting all the variouscompositional components of the local crust were allowedto cleanly reorganize into two massive layers. After rapidly(tens of years) dissipating the superheat by vigorous ther-mal convection, which also served to further chemicallyhomogenize each layer, solidification fronts migratedinward from the top and bottom. Two essentially uniformlayers of rock formed the final product. There is now nosign of layering and very little sign of differentiation. Thus,instantaneous injection of crystal-free magma does notform exotically layered, well-differentiated intrusions. TheSudbury testimony is clear: no phenocrysts, no layering.This is the Null Hypothesis (Zieg and Marsh 2005).

CONTINENTSContinents are ultimately differentiates of basalt (e.g.Davidson and Arculus 2005). There is no way around it.There is no indication from meteorites that granitic rockwas any part of the original solar debris forming Earth, tobe sorted and collected, like at Sudbury, through meteoriteimpacts into continents. Basalt comes from partial meltingof the underlying primitive mantle during convectionthrough a phase boundary, the mantle solidus. But themain problem is that basalt does not physically fractionatedirectly into silicic continental building blocks. There aretwo principal physical routes to the formation of conti-nents, and both involve solidification fronts.

Roof solidification fronts in basaltic magma commonlythicken to the point of becoming structurally unstable, fail-ing under the action of gravity by tearing internally (e.g.Marsh 2002). Forming well within the solidification front,

The relationship between position on a ternary phase dia-gram and spatial position within a basaltic solidification

front. During the loss of the crystal load carried by the ascendingmagma, the composition slides along the cotectic boundary in the ter-nary diagram and stops at the black dot representing the bulk compo-sition of the solidification front. Further differentiation by crystal frac-tionation to produce siliceous, continental-like melts is prohibited as theresidual melts are trapped within the crystalline network of the solidifi-cation front and can only be collected by solidification front instability.In the upper diagram, temperature and viscosity increase along theindicated curves.

FIGURE 3

291E L E M E N T S OCTOBER 2006

the tears fill with local residual interstitial melt, which isenriched in silica (~60–65% SiO2) and other continent-likecomponents. These isolated silicic segregations are scatteredat specific horizons and can only be amassed or collected bywholesale disruption of the solidification front itself. Duringthe earliest stages of Earth formation, it is likely that magmaoceans prevailed and solidified much like lava lakes.Because of the scale and steady heavy meteorite bombard-ment, the upper solidification front continually founderedand remelted in the underlying superheated magma, free-ing the silicic segregations; these steadily collected, almostlike lint, on the surface. This runaway process might wellhave rapidly produced the first continental vestiges (FIG. 4).Additional processes (e.g. erosion, sedimentation, meta-morphism, etc.) are needed to transform these differentiatesinto continental material (e.g. Rudnick 1995), but themechanism described above is perhaps the critical physicalprocess required to bridge the basalt differentiationimpasse. From a purely chemical perspective, solidificationfront instability is indistinguishable from normal fractionalcrystallization.

The key to this physical process of producing primordialcrustal material is the recycling of the solidification front,which is unavoidable in any major magmatic system sta-tionary long enough to undergo substantial reprocessing.All major oceanic volcanic systems of this nature producehighly silicic material. Iceland, Rapa Nui (Easter Island),and the western Galapagos Islands are good examples. AtIceland, lying athwart the Mid-Atlantic Ridge, about 10% ofthe surface rocks is highly silicic. Periodic jumps of theoceanic ridge into older parts of Iceland promote reprocess-ing, and the silicic lavas are full of telltale signs (e.g. clots ofleucocratic silicic rocks) of this process (e.g. Gunnarsson etal. 1998). Volcanic centers like Hawai‘i are too mobile toallow any significant reprocessing—once the system is builtit becomes extinct, moves on, and a new center is built. The

Moon is similar. Although strongly differentiated, there isno sign whatsoever of granitic rock on the Moon becausethere has been no protracted magmatism to reprocess theinitial crust.

MUSH COLUMNS, CRYSTAL CARGO, AND GRANULAR PROCESSESThe traditional approach to understanding the chemicalevolution of erupted magma is to assume a source rock,generate the magma through partial melting, and move itto a near-surface magma chamber where crystal fractiona-tion achieves the final product. The magma gains its chemicalidentity through “source” and “high-level” generic crystalfractionation and contamination processes, and the spa-tially and temporally protracted intervening path fromdepth to near surface is, per force, unneeded and ignored. Anincreasing abundance of evidence in near-surface anderupted magma reveals, however, that the most importantpart of the life cycle of magma is what happens duringascent, when a series of physical processes buttressed bychemistry operate. The magnitude of the load of large crystals,carried by the moving magma and critical to punctuateddifferentiation when the magma tarries during ascent, isproportional to the local transport flux. Playing the role ofsuspended sediment in a surging river at flood stage, thecrystal cargo also acts as a tracer to the dynamics of ascentand emplacement or eruption. In fissures and dikes thiscrystal slug forms a central tongue or swarm, which isdeposited on the floor of a sill after the magma in the dikestalls and moves horizontally into the sill (Simkin 1967).This Simkin Sequence is a key link between the dynamics ofmagma in dikes, sills, layered intrusions, and major volca-noes. The regional distribution of phenocrysts in sillsreflects the point of infilling and recharge. Local magmaponding and successive avalanching of crystal slurries pro-duces distinct layering through granular sorting. Mechanical

A schematic depiction of early proto-continental crust for-mation by massive and widespread solidification front

instability at the roof of a magma ocean. Internal tearing segregateslocal siliceous melt into lenses (arrows denote melt flow) that arereleased by foundering and thermal disintegration of the whole solidifi-cation front. The viscously immiscible silicic blobs collect to from proto-crustal material. The compositional profile (right) becomes bimodal dueto this instability, reflecting the inherent compositional differencebetween continents and ocean basins.

FIGURE 4

REFERENCESBowen NL (1915) The later stages of the

evolution of the igneous rocks. Journal ofGeology 23: 1-89

Bowen NL (1947) Magmas. Bulletin of theGeological Society of America 58: 263-280

Davidson JP, Arculus RJ (2005) Thesignificance of Phanerozoic arc magma-tism in generating continental crust. In:Brown M, Rushmer T (eds) Evolution andDifferentiation of the Continental Crust,Cambridge, pp 135-172

Davidson JP, Hora JM, Garrison JM,Dungan MA, (2005) Crustal forensics inarc magmas. Journal of Volcanology andGeothermal Research 140: 157-170

Dawson JB (1992) First thin sections ofexperimentally melted igneous rocks:Sorby’s observations on magmacrystallization. Journal of Geology 100:251-257

Dungan MA, Davidson JP (2004) Partialassimilative recycling of the maficplutonic roots of arc volcanoes: Anexample from the Chilean Andes.Geology 32: 773-776

Fouque FA (1879) Santorini and itsEruptions. McBirney AR (1998)(translator), Johns Hopkins UniversityPress, Baltimore, 560 pp

Gorring ML, Naslund HR (1995) Geochemi-cal reversals within the lower 100 m ofthe Palisades sill, New Jersey. Contribu-tions to Mineralogy and Petrology 119:263-276

Gunnarsson B, Marsh BD, Taylor HP Jr(1998) Generation of Icelandic rhyolites:silicic lavas from the Torfajökull centralvolcano. Journal of Volcanology andGeothermal Research 83: 1-45

Marsh BD (2002) On bimodal differentia-tion by solidification front instability inbasaltic magmas, part I: Basic mechanics.Geochimica et Cosmochimica Acta 66:2211-2229

Marsh BD (2004) A magmatic mushcolumn rosetta stone: The McMurdo DryValleys of Antarctica. EOS, Transactionsof the American Geophysical Union85(47): 497 and 502

Rudnick RL (1995) Making continentalcrust. Nature 378: 571-578

Simkin T (1967) Flow differentiation in thepicritic sills of North Skye. In: Wyllie PJ(ed) Ultramafic and Related Rocks, JohnWylie and Sons, New York, pp 64-69

Wilson M (1993) Magmatic differentiation.Journal of the Geological Society,London, 150: 611-624

Wright TL, Okamura RT (1977) Coolingand crystallization of tholeiitic basalt,1965 Makaopuhi lava lake, Hawaii. U.S.Geological Survey Professional Paper1004, 78 pp

Zieg MJ, Marsh BD (2005) The SudburyIgneous Complex: Viscous emulsiondifferentiation of a superheated impactmelt sheet. Geological Society of AmericaBulletin 117: 1427-1450 .

292E L E M E N T S OCTOBER 2006

processes produce the basic layering, which is then but-tressed by diffusive chemical processes. The Dais section ofthe Basement Sill in the McMurdo Dry Valleys, Antarctica(FIG. 5) exhibits this critical relationship. The cargoes ofcrystals themselves also have a rich story to tell.

The provenance or ultimate source of the individual crystalsoften reflects contributions from deep wall rock, deep oldercumulate deposits cognate to the system, and crystalsfreshly nucleated and grown in transit. At any level, themagmatic system consists of a “carrier” magma containingvarying proportions of these genetic contributions. None ofthe crystal ensembles are fully at chemical equilibrium, buteach locally reacts with the magma to produce, in detail,assemblages of older crystals armored by overgrowths, crystalsdissolving and regrowing, and fresh crystals growing inmagma of changing composition. Detailed microsamplingof individual nearby (centimeter-sized) crystals revealsstrong compositional and isotopic diversity (e.g. Dunganand Davidson 2004; Davidson et al. 2005), indicating amultidimensional magmatic mush column open andhighly integrated from source to surface and operatingunder the temporal pulse of volcanism (Marsh 2004).

CONCLUSIONThe ultimate challenge to understanding the dynamics ofplanetary magmatism is the quantitative perception of aunified process accurately combining dynamics operatingon the smallest spatial scales in solidification fronts withthe spatial structure and rhythms of volcanism operatingon the largest scales. Processes operating between crystalsand melt at the smallest scales within solidification frontsand in mingling slurries give rise to planetary-scale effects.Regardless of the system, there is every indication of a singlebroad magmatic style operating under various themes.

ACKNOWLEDGMENTSThis work is supported by NSF Grant OPP 0440718. Usefulcomments from Dougal Jerram, Grant Henderson, and BruceWatson are much appreciated. .The Dais section of the Basement Sill, McMurdo Dry Val-

leys, Antarctica. The 450 m section exhibits an extremevariation in composition, ultramafic (~mantle) at the base and dioritic(~continental) at the top. Rich modal layering on many scales is due tothe mechanical sorting of large pyroxenes and tiny plagioclases, pro-ducing layers of orthopyroxenite and anorthosite. This sorting isreflected in strong chemical variations (see CaO–MgO inset) over short(5–15 m) distances (after Marsh 2004).

FIGURE 5