skip to main content

Granulite sections:

Photomicrograph: borosilicates - werdingite and grandidierite

Rare borosilicates from Namaqualand, S. Africa: an intergrowth of blue grandidierite and dark spinel, 1mm across, invades pale yellow werdingite.

Granulites and Migmatites

Dehydration melting is believed to be an important process in the deep crust, responsible for generating water-undersaturated melts and leaving granulite facies assemblages in the residue. The melting process can be studied in situ in various rock types in well-preserved granulite terrains. Questions being addressed, using bulk chemical and microprobe analytical techniques, include the amount and geochemical signature of melt which can be extracted from different source rocks. Accessory phases are zoned and vary markedly in abundance and composition between the palaeosome and neosome of migmatites. These relationships offer the possibility of constraining the amount of melt loss, of assessing the extent of disequilibrium in melting and melt separation, and potentially also of constraining the time-scale of the migmatisation process.

High-T metamorphic belts, many of which show isobaric cooling or anticlockwise P-T-time paths and contain geochemically-distinctive granitoid and charnockite rock suites, present their own challenge in interpreting the tectonic setting of ancient orogeny, and potentially hold the key to understanding Precambrian crustal development and the character of the present-day lower continental crust. We have projects, involving international collaborations, in the Archaean of West Greenland, the Proterozoic of southern Africa, and the Grenville of the eastern USA.

Index of Abstracts

Compositional zoning in monazite: relative and absolute chronology of the metamorphic evolution of granulites

D.J. Waters and N.R. Charnley

Department of Earth Sciences, University of Oxford

Granulite facies migmatites (750 - 900°C, 5 - 6 kbar) from the Namaqualand Metamorphic Complex, South Africa, contain complex zoned monazite in both palaeosome and neosome domains. Backscatter electron imaging, compositional mapping and spot analysis using both EDS and WDS electron microprobe techniques reveal several chemically-distinct generations of growth. The relative chronology of monazite growth is readily established from concentric patterns of zoning, from internal compositional unconformities, and from chemical correlations from grain to grain. Early core zones are rich in Y and U. The dominant stage of growth, filling lobate embayments in palaeosome monazite and forming broad, weakly oscillatory growth zones in neosomes, has strongly depleted HREE + Y. One set of rim zones has high Th and depleted HREE, and a later set has lower Th and high Y + HREE.

Each monazite generation can be correlated with mineral assemblage changes in the body of the rock, the most obvious effects relating to the growth and resorption of garnet. At the prograde stage the rock was a garnet-poor biotite gneiss. Towards peak conditions abundant garnet developed by dehydration melting of biotite. After the peak, new biotite formed by limited back-reaction between melt and neosome garnet. Monazite cores developed at the prograde stage, before the appearance of garnet. The main growth stage of HREE-depleted monazite reflects the melting stage, when these elements are partitioned into garnet. Back-reaction of garnet and release of Y and HREE is recorded in the late rims. In one sample, provisional chemical dating by electron microprobe of these growth generations gives 1063 ± 24 Ma for cores, 1038 ± 11 Ma for the main growth stage, and 1013 ± 11 Ma for rim generations, broadly consistent with zircon geochronology on associated magmatic rocks.

Most of the monazite microstructures can be interpreted as the result of precipitation and dissolution reactions. There is only limited evidence for recrystallised zones comparable to those described from zircons, and no evidence for diffusional modification. The chemical uniformity of monazite generations despite spatial separation in the rock indicates effective communication during periods of reaction via interstitial fluid or melt, and the implication is that most of the fluid or melt interactions with the rock are being recorded to some degree by monazites. In this case all the growth generations appear to belong to a single tectono-metamorphic cycle.

(Presented at: Timing, Transition, Tectonics: Mineralogical Society Winter Conference, Derby, January 2002)

Conceptual models for mass balance in migmatites - where did the melt go?

D. J. Waters, J. E. Haynes and J. Burton

Dept. of Earth Sciences, University of Oxford, Parks Road, Oxford OX1 3PR

This study is concerned with combining and extending existing approaches to the major element mass balance of migmatites, with particular application to "nebulitic" granulite-facies phenomena which may result from dehydration melting. We have devised a diagrammatic representation which displays graphically the mass differences between the neosome (coarse-grained domains or "segregations") and the "palaeosome" (used non-genetically to signify the matrix hosting the neosome domains), and which assists in choosing an appropriate hypothesis to account for the differences. Many other approaches rely on the identification of a part of the system as representing either segregated melt, or unaltered precursor, but we make no such assumption.

First, a mass balance model similar to that of Bryan et al. (1969, Science 163, 926-927) and Bea (1989, J. Metamorphic Geol. 7, 619-628), in which

Palaeosome = a0 . Neosome + a1 . P1 + a2 . P2 + … + an . Pn

where P1 - Pn are a specified set of independent phase components, is solved for the unknown coefficients a0 - an by least squares. In principle any independent set of components P could be chosen, but a choice of actual phase components is most likely to aid interpretation.

The graphical tool consists of a two-dimensional space, in which area is proportional to mass, divided into quadrants as if by Cartesian axes. To positive y are plotted those parts of the system that are real and visible (i.e. the actual palaeosome and neosome), and at negative y those (P1 - Pn) which are not visible. On the left of the diagram (negative x) are plotted masses of components which have negative coefficients, and to the right, those with positive coefficients. When used in parallel with other techniques aimed at establishing a compositional reference frame, the distribution of non-visible materials in the bottom left and bottom right quadrants of the diagram helps to distinguish among different processes:

Our studies of granulite-facies pelitic and semi-pelitic migmatites from Namaqualand, South Africa, and of metabasic migmatites from the Akia Terrane, West Greenland, reveal a variety of patterns. Results from one group of samples, typically those with smaller isolated neosomes, indicate migmatisation in systems closed on the whole rock scale, but not uncommonly with local compositional heterogeneity consistent with initial melting in more Fe-Al-rich (pelitic) or Si-rich (basic) domains. A second group is interpreted as representing accumulation in the neosomes of melt which has been either transferred from the palaeosome, or derived from outside the analysed system. A third group, important in metapelitic migmatites, indicates the loss of a granitoid melt from the domains now represented by neosome, suggesting that under appropriate circumstances there can be effective melt extraction from relatively undeformed migmatites at temperatures of around 800°C.

(Presented at: Magmas to Mud - and back again; Mineralogical Society Millennium Winter Conference, Reading, December 1999)

A new time framework for Kibaran high-grade metamorphism in southern Africa: what drives granulite formation?

D.J. Waters (1), L.J. Robb (2) and R.A. Armstrong (3)

(1) Department of Earth Sciences, Oxford University, Parks Road, Oxford, OX1 3PR
(2) Department of Geology, University of the Witwatersrand, Johannesburg, South Africa
(3) Research School of Earth Sciences, Australian National University, Canberra, Australia

The Namaqua-Natal belt in southern Africa experienced high-grade metamorphism during the period 1200-1000 Ma (the Kibaran or Grenvillean orogeny). In western Namaqualand, the event involved compressional structures, voluminous early syn- and late-tectonic magmatism, and metamorphic mineral growth which is partly syn- and partly post-tectonic. Three distinctive suites of Kibaran magmatic rocks are recognised:

  1. The Little Namaqualand Suite of granitoid gneisses which experienced the main intense regional deformation D2 and show the L2 lineation. Rb/Sr whole rock data indicate an age of about 1200 Ma, which has been interpreted by some as recording the peak of granulite-facies metamorphism.
  2. The Spektakel Suite of late-tectonic granites, much in sheet-like bodies deformed by D3 folding. An age in the interval 1100-1150 Ma has been inferred from Rb/Sr data.
  3. The Koperberg Suite of minor basic to intermediate intrusions, which appear to be post-D3 in age, and were intruded under granulite-facies conditions (1040-1100 Ma ago according to older data).

Previous interpretations of the metamorphic history include (i) granulite-facies metamorphism of long duration 1200-1050 Ma, which is unlikely on thermal grounds; (ii) two peaks of granulite-facies metamorphism at 1200 and 1050 Ma (for which there is no petrographic evidence); (iii) a single, shorter-lived thermal climax at ca 1150 Ma.

New precise zircon age determinations, mainly from intrusive gneisses and granites in the Okiep Copper District, require revision of this view. The emplacement ages of Kibaran magmatic rocks fall into two well-separated clusters. The precursors of the Little Namaqualand Suite gneisses appear to record a magmatic age (not a metamorphic age) of around 1200 Ma. Both the Spektakel and Koperberg Suites appear to show magmatic zircon populations in the range 1065-1030 Ma. Rims on zircons from older units give ages of 1030 Ma, coincident with the later group of magmatic ages. Whereas the geochemical character of the earlier granitoids is consistent with convergent tectonics and collisional orogeny, that of the later magmatism is not, and rather indicates a thermal event accompanied by the melting of older deep crustal material and by the granulite-facies metamorphism.

Petrographic evidence in supracrustal rocks is consistent with amphibolite-facies syn-D2 fabric development, whereas granulite-facies minerals are syn- to post-D3. The importance of mimetic crystalliation (of granulite minerals on earlier fabrics) has probably been underestimated in some previous accounts.

In our working hypothesis, the Kibaran history of Namaqualand is resolved into discrete events separated by ca. 150 Ma. In the early Kibaran a low-P facies series metamorphic event of amphibolite grade accompanied D2 convergence and calc-alkaline magmatism. In the late Kibaran an intense thermal event caused deep crustal melting and ductile deformation (D3) but only a moderate amount of horizontal shortening. This event was not associated with extensional breakup, which took place ca 200 Ma later along a very different trend, and alternative mechanisms must be sought. Its timing and location suggest a delayed response to the change in subduction regime after an early Kibaran collision. Similar age relationships probably apply in adjacent areas of eastern Namaqualand and in Natal, and the framework should be tested by attempting to date the metamorphic climax in other areas.

(Presented at "What drives metamorphism and metamorphic reactions", Metamorphic Studies Group, Kingston University, 11-13 September 1996)

Contrasting accessory mineral behaviour during migmatisation and melt extraction

J. Burton and D.J. Waters

The Namaqualand Metamorphic Complex contains amphibolite- to granulite-facies rocks formed during a Mid-Proterozoic metamorphic event. In paragneiss belts the onset of partial melting has been observed, with the development of migmatites. These take the form of cross-cutting, garnet-bearing leucocratic regions (neosome) with an anhydrous mineral assemblage, in a schistose matrix with a metamorphic aspect (palaeosome). The melting reactions inferred for these migmatites are of the type:

Bt + Pl + Sil + Qtz = Grt + Kfs + melt

involving the consumption of biotite to produce an anhydrous assemblage and a water-undersaturated melt.

Five migmatite samples have been carefully separated into neosome and palaeosome domains and analysed for major and trace elements. Mass balance models of the form

Neosome = Palaeosome +/- Qtz, Pl, etc.

point to open-system behaviour of the migmatites, with removal of a granitoid melt from the neosomes. This conclusion is supported by the bulk REE signatures of the neosomes and palaeosomes. The neosomes display strong HREE enrichment, related to the residence of HREE in the newly-formed garnet, and prominent negative Eu anomalies resulting from the removal of feldspar components in extracted melt.

The non-opaque accessory minerals in the five migmatites show considerable variations in abundance, texture and composition between the neosome and palaeosome. The dominant accessory phases are zircon, apatite, and monazite. These are the dominant hosts for Zr (zircon), P (apatite and monazite), and the LREE (monazite).

In the palaeosomes, zircon is the most common accessory mineral, forming crystals up to 60 µm in diameter. The neosomes display considerably lower zircon abundances and smaller crystal diameters, except where zircons are included in garnet. This can be attributed to zircon breakdown during anatexis, and possibly also to the removal of small crystals entrained in melt.

In contrast, both apatite and monazite are enriched in the neosomes compared to the palaeosomes. Apatite shows strong evidence of growth during melting, with entrapment of many crystals in garnet. Monazite growth during and subsequent to melting has produced broad low-Y rims on xenocrystic high-Y core regions: the cores are very similar to monazites in the palaeosome. Monazite included in peritectic garnet rarely shows the low-Y rim zones, suggesting that most of the monazite growth took place after melt generation, but probably before melt extraction.

(Presented at "Research in Progress", Metamorphic Studies Group, Oxford University, 6 March 1996)

A record of processes in very hot middle crust, preserved in osumilite- and spinel-bearing metapelites from Namaqualand, S. Africa.

David J. Waters

Metapelites in the Proterozoic Namaqualand Metamorphic Complex display a large enough range of bulk composition to reveal an AFM facies type for the maximum grade assemblages in low-pressure granulites. Key KFMAS assemblages are Hc+ Grt + Crd + Kfs + Qtz and Osu + Opx + Crd + Kfs + Qtz. Hercynite coexisted with quartz in reduced, relatively Fe-rich metapelites. Osumilite is restricted to highly oxidised rocks with abundant haematite and magnetite, in which Osu and Crd have XMg > 0.85. Peak conditions can be placed at 860-900°C and 5 - 6 kbar by phase diagram calculations and by comparison with recent experimental studies. Somewhat surprisingly, conventional thermobarometers, including Fe-Mg exchange thermometers, give similar values.

In the osumilite-bearing area, both migmatitic and unmigmatised metasediments occur. Rocks with Qtz + Kfs + Pl + Phl/Bt have undergone partial melting, generating leucosome domains with large crystals of osumilite and orthopyroxene. Si-rich and/or K-poor lithologies have no leucosomes and preserve a regular compositional layering. Metamorphic conditions were largely buffered to the water-undersaturated solidus, defining a water activity of ca. 0.2. Associated granitoids are partly Opx-bearing.

During the very high-T event there was apparently a net gain of material at this crustal level, consisting mainly of syn-metamorphic granitoids of deeper origin. The relatively shallow level places constraints on the time span of high-T metamorphism and raises questions, potentially answerable with modern techniques, about the synchroneity of heat input across this 200 km wide granulite terrain.

(Presented at "Controls of Metamorphism 94", Liverpool, September 1994)

Melt Extraction from Granulite-Facies Migmatites

D.J. Waters
Dept of Earth Sciences, University of Cambridge, Downing Street, Cambridge CB2 3EQ, U.K.
F.J. Baars
Dept of Geology, University of Cape Town, Rondebosch, 7700 South Africa.

Migmatitic rocks in granulite facies terrains commonly consist of coarse anhydrous patches, stringers or vein networks (mobilizates) in a finer-grained, foliated matrix. We interpret these features as the result of incongruent dehydration melting reactions (Waters, 1988). The mobilizates do not represent the composition of segregated melt, but rather represent varying proportions of (i) melt, (ii) solid products of incongruent melting, and (iii) excess phases not consumed during the reaction. In metapelites and biotite gneisses, where the near-solidus melt composition is granitic and its water content can be estimated as a function of T, the melting reactions can be balanced and relative volumes of product phases predicted.

We have compared the calculated products with the observed distribution and abundance of phases in granulite facies migmatites from Namaqualand, South Africa. At the amphibolite-granulite transition (ca.750°C), the modal composition of mobilizates indicates that 10-15 vol% melt was produced, but was not extracted from its site of generation. At higher grade (>800°C) some mobilizate patches are dominated by solid products (e.g. garnet, K-feldspar) implying the loss of 10-20 vol% melt. The existence within a small outcrop area of an apparently complete gradation from garnet- or pyroxene-rich patches to leucogranitic stringers and sheets suggests that the extracted liquid has not migrated far.

A second set of predictions follows from an estimation of the physical properties of water-undersaturated silicic melts. The dihedral angle for melt in a quartzofeldspathic aggregate is probably less than, though close to, 60°. Therefore, if textural equilibrium is approached, the melt phase should be interconnected. The efficiency of melt extraction will be critically dependent on the viscosity of the melt (McKenzie, 1985; Wickham, 1987). Calculated values for the water-under saturated case are several orders of magnitude greater than those for water-saturated melts, and fall in the range 106 - 108 Pa.s. These values correspond to a length scale for compaction on the order of centimetres, which is similar to that of the mobilizates, but imply an implausibly long time scale for melt extraction of 10 - 1000 Ma. In the absence of deformation, short-range controls such as the minimization of surface energy may determine the distribution of the melt phase. If this is the case, the matrix regions are predicted to contain a few percent melt in an equilibrium distribution along grain boundaries, while the mobilizates contain pooled excess melt.

Many, but not all, of these predictions are supported by the observations on natural rocks. The solid products of incongruent melting are concentrated in the mobilizates, and are not distributed through the rock. Commonly, evidence suggests that mobilizates develop along compositional or structural inhomogeneities in the rock. For example, at the granulite transition, small mobilizates appear to have nucleated on pre-existing matrix garnets. These features indicate that melt is generated within, and is perhaps confined to, the mobilizate volumes. The matrix regions of rocks are texturally similar to their amphibolite facies equivalents, and give no indication that the melt has penetrated the grain boundary network outside the mobilizates. Possibly, textural equilibrium is not achieved on the time scale of the melting event, or else the true dihedral angle is greater than 60°. Elsewhere, some mobilizates follow earlier fabric elements, or are aligned in the hinge zones of decimetre-scale crenulations and in minor shear zones. The structural control of mobilizates, particularly those with more leucocratic compositions, suggests that melt extraction is assisted by deformation or tensional fracturing.

In summary, field evidence suggests that some mobilizates represent sites of melt generation, while others represent structurally-controlled zones where melt has locally accumulated.

Amphibolite facies migmatites, in contrast, commonly show leucosomes whose compositions are similar to minimum melts, and are markedly different to the matrix (palaeosome) composition. A melanosome is commonly also present. These features imply a much more efficient separation of solids from melt at low melt fractions, consistent with the lower viscosity predicted for melts nearly saturated in H2O.

We conclude that water content has a critical effect on the viscosity (and/or possibly the dihedral angle) of siliceous melt, so that small melt fractions are even less likely to be extracted from partially melted granulite facies rocks at about 800°C than from amphibolite facies migmatites. Large, mobile magma bodies will result only when the melt fraction becomes large enough to cause disaggregation and convective instability of the whole rock mass. A higher melt fraction may result from higher temperature, a higher initial proportion of hydrous minerals, or an external source of H2O. In the case of Namaqualand, the region of magma generation lay at hotter, deeper levels in the crust. The magmas are represented in part by late-tectonic, approximately syn-metamorphic granitoids, which intrude the migmatites, and which probably assisted in the transfer of heat to the exposed levels.

McKenzie DP (1985) Earth Planet Sci Lett 74: 81-91.
Waters DJ (1988) J Metamorphic Geol. 6: 387-404.
Wickham SM (1987) J Geol Soc London 144: 281-297.

(Presented at the 28th International Geological Congress, Washington, July 1989)

Created/modified 15 January 2002