Sei sulla pagina 1di 16

JOURNAL OF PETROLOGY

VOLUME 48

NUMBER 5

PAGES 885^900

2007

doi:10.1093/petrology/egm005

Nature of Sub-volcanic Magma Chambers, Deccan Province, India: Evidence from Quantitative T extural Analysis of Plagioclase Megacrysts in the Giant Plagioclase Basalts
MICHAEL D. HIGGINS1* AND D. CHANDRASEKHARAM2
1 2

' SCIENCES DE LA TERRE, UNIVERSITE DU QUEBEC A CHICOUTIMI, CHICOUTIMI, QUE. G7H 2B1, CANADA DEPARTMENT OF EARTH SCIENCES, INDIAN INSTITUTE OF TECHNOLOGY, MUMBAI 400076, INDIA

Downloaded from http://petrology.oxfordjournals.org/ by guest on January 4, 2013

RECEIVED SEPTEMBER 27, 2006; ACCEPTED FEBRUARY 5, 2007 ADVANCE ACCESS PUBLICATION MARCH 8, 2007

Sub-volcanic magma chambers might be a widespread component of flood basalt provinces, and their presence can be revealed in some cases by plagioclase megacrystic basalts. In at least four levels within the Deccan flood basalt sequence the generally low abundance of small plagioclase crystals increases to 5^25%, with some as large as 30 mm long. These Giant Plagioclase Basalt (GPB) flows were formed by mixing of megacryst-rich and megacryst-poor magmas. The crystal size distributions (CSD) of these megacrysts mostly plot as almost straight lines on a classic CSD diagram. For a plagioclase growth rate of 1010 mm/s steady-state magma chamber models and simple continuous growth suggest residence times of 500^1500 years. However, the lack of crystals smaller than 2 mm suggests that coarsening may have been involved and crystal shape can help define the environment where this happened. Plagioclase megacrysts are very tabular and commonly form clusters of sub-parallel crystals, characteristics that are also found in the plagioclase of anorthosites formed by flotation at the top of shallow magma chambers and crystallization in a high Peclet number regime (e.g. Skaergaard; Sept Iles) A possible history is as follows. () . 1 Plagioclase megacrysts crystallize in a convecting magma chamber just below the lava pile. (2) Currents sweep the crystals to the top of the chamber, where they accumulate as a result of their buoyancy.The crystals coarsen in response to the continuous supply of hot magma. (3) New magma sweeps through the plagioclase mush, mingles and mixes, then erupts to form the GPBs.The residence time recorded by the megacrysts in the GPBs is that of the magma chamber where the megacrysts formed, not that of the magmas that make up the megacryst-poor part of the GPBs or the other megacryst-poor lavas. Lavas with megacrysts similar to the GPBs are uncommon but widespread

(Galapagos, Surtsey, etc.) and suggest the presence of sub-volcanic , magma chambers elsewhere.
KEY WORDS: texture; microstructure; continental basalt; megacryst; plagioclase; crystal shape

I N T RO D U C T I O N
The Deccan traps are one of the most important continental flood basalt provinces (Sen, 2001; Chandrasekharam, 2003; Jerram & Widdowson, 2005). They erupted about 65 Myr ago, synchronous with the K^T boundary, and it has been suggested that they might have caused or contributed to the major mass extinction event at that time (Duncan & Pyle, 1988). However, the effect of magmatism on climate is critically dependent on eruption rate. It is not easy to determine the eruption age of the Deccan Traps, as most of the rocks are hydrothermally altered. Hofmann et al. (2000) considered that the best ages are from 40Ar/39Ar dating of mineral separates, mostly plagioclase. They determined the age of lavas at the top of the Western Ghats pile to be 654 07 Ma and that of the base as 652 04 Ma. Hence, the main eruptive period was probably less than 1 Myr. Nevertheless, magmatism may have occurred in a much shorter period or episodically, neither of which can be resolved by current isotopic techniques. It has been suggested that the duration of magmatism of the Deccan group can also be determined from rock

*Corresponding author. E-mail: mhiggins@uqac.ca

The Author 2007. Published by Oxford University Press. All rights reserved. For Permissions, please e-mail: journals.permissions@ oxfordjournals.org

JOURNAL OF PETROLOGY

VOLUME 48

NUMBER 5

MAY 2007

textures (Sen, 2001, 2002). Most formations in the lower part of the Deccan group terminate in an evolved magma that contains plagioclase megacrysts up to 50 mm long (Giant Plagioclase Basalts; GPBs). If these crystals grew continuously in the magma storage zone and their growth rate could be estimated then the duration of each magmatic cycle could be determined. If there were no major hiatuses between the magmatic cycles then the total duration could be determined, or at least a minimum value. Using this method, Sen et al. (2006) proposed that the duration of magmatism may have been as little as 22 800 years. Obviously, such a rapid rate of magmatism would have perturbed the climate considerably and could have influenced biotic evolution. In this paper we will examine quantitatively the textures ( microstructure) of plagioclase in the GPB flows and propose a model for their formation.

100 km

y-rift Camba

22

Mhow Girnar Shahada 20 Mumbai (Bombay) Area of interest Toranmal

Narm

if ada-rt

Jabalpur

Tapi-rift

Shirpur Dhul e Buldana Nasik Lonar

Mahabaleswar Ambol i

16 Arabian Sea

Bay of Bengal

Downloaded from http://petrology.oxfordjournals.org/ by guest on January 4, 2013

D EC C A N G EOLO GY A N D T H E G I A N T P L AG I O C L A S E B A S A LT S Introduction
The basalts of the Deccan province were erupted from many centres, but the most important is a shield-volcanolike structure in the Western Ghats, near Mumbai (Fig. 1) (Subbarao et al., 1994). Rifting and erosion have exposed a 17 km thick section of flows, which is the focus of this study. Aphyric or almost aphyric tholeiitic basalts dominate the section. It has been divided into sub-groups using field geology, petrography and geochemistry (Beane et al., 1986). The lowest sub-group, the Kalsubai, contains a number of horizons of lavas with abundant, large tabular plagioclase crystals termed megacrysts: these are the GPBs (Beane et al., 1986; Hooper et al., 1988). Megacrysts are crystals that are strikingly larger than other crystals in the rock, but there is no genetic implication as to whether such crystals are phenocrysts or xenocrysts. Similar GPB flows have also been reported from the northern Deccan province but are not considered here (Chandrasekharam et al., 1999). In the western Deccan, GPBs cap each formation, have the highest total Fe and Ti contents, and are considered to be the most chemically evolved.Plagioclase-phyric flows are similar to the GPBs, but with less and/or smaller plagioclase megacrysts; they are combined in this study with the GPBs. The tops of some flows are covered by red bolesa fine-grained material rich in haematite thought to have formed by weathering of the basalts and hence indicating a period of volcanic repose (Fig. 2).

70

74

78

Fig. 1. The Deccan volcanic province. The samples are from the region east of Mumbai, where the basalt section is thickest (circled).

Deccan-Western Ghats section 1.5

Tunnel Five GPB

1.0

Height (km)

Kashele GPB

0.5 Thalghat GPB Red boles

0 0 10 Phenocrysts (%)
Fig. 2. Part of the Western Ghats section of the Deccan series. The phenocrysts are dominated by plagioclase. GPB, Giant Plagioclase Basalt. Red boles are red fine-grained material rich in haematite thought to represent periods of erosion. Modified from Sen (2001).

20

Giant Plagioclase Basalt flows


Individual GPB flows are 5^10 m thick and occur at distinct horizons within the eruptive sequence (Beane et al., 1986). However, the extent of individual flows is unclear and hence it is impossible to say if the same flow occurs at

different places or if it is just another flow of the same package. GPB flows were mostly sampled in road cuts and quarries, where the flows are relatively well exposed (Table 1). At these locations it is clear that the GPBs are not homogeneous, but are composed of at least two magmatic components: one aphyric or sparsely phyric and another rich in plagioclase megacrysts (Fig. 3). The fabric defined by

886

HIGGINS & CHANDRASEKHARAM

SUB-VOLCANIC MAGMA CHAMBERS

T 1: Sample locations able


Sample no. Latitude (N) Longitude (E) Altitude (m) Location Notes

MH-04-04 MH-04-05A MH-04-05B MH-04-05C MH-04-06 MH-04-07 MH-04-10 MH-04-12 MH-04-14 MH-04-16

1984155 1985645 1985643 1985643 1985630 1985668 1985840 1985901 1981161 1982500

7382981 7384385 7384383 7384373 7384387 7384356 7382657 7382668 7385136 7381500

422 697 730 725 676 675 425 503 868 775

Nasik road Pandavlena quarry Pandavlena quarry Pandavlena quarry Near Pandavlena quarry Near Pandavlena quarry TrimbakJawhar road TrimbakJawhar road Shivaneri Fort Bote hill

Thalghat Kashele Kashele Kashele Kashele Kashele Thalghat Thalghat Manchar Manchar

Downloaded from http://petrology.oxfordjournals.org/ by guest on January 4, 2013

5 cm

Megacryst-poor component

Megacryst-rich component

Fig. 3. Variations in the plagioclase megacryst content within a single GPB flow. The field of view is 40 cm.

plagioclase crystals within the megacryst-rich component is extremely variable in direction and intensity. The spatial arrangement of these components and their fabric suggests that they originated by incomplete mixing (mingling) of two components. This may have occurred by simultaneous eruption of two magmas into the flow, or it could also have happened at depth in the feeder and been preserved during transport. This effect is seen in rivers, where water from a tributary may be clearly distinguished far from the junction of the streams. All GPB flows examined in this study had similar inhomogeneous structures.

Petrography
The GPB lavas consist of plagioclase megacrysts in a finegrained matrix. The plagioclase megacrysts are 2^50 mm long. They have compositions ranging from An61 to An64 (Hooper et al., 1988) and are generally weakly zoned, except for a narrow rim. Many crystals appear to have a

narrow hollow centre, which is filled with fine-grained material similar to the matrix. This is assumed to be trapped melt. Plagioclase in the matrix is much smaller than the smallest megacryststypically 025^1mm long. Hence, it is generally easy to distinguish megacrystic and matrix plagioclase. The texture of the plagioclase megacrysts was examined both qualitatively and quantitatively (see below). In both cases the first step was the creation of a binary image of plagioclase distribution. Rock samples 10^20 cm square were slabbed normal to the foliation, if present, and polished. In most samples the plagioclase was sufficiently altered that it stood out clearly from the matrix. However, in some very fresh samples it was not possible to distinguish clearly the plagioclase megacrysts. In these samples the polished surface was painted with hydrofluoric acid, which generally accentuated the contrast. Slabs were then scanned using a conventional document scanner and the images transferred to a vector drafting program (CorelDraw). There the crystals were outlined on the screen using a mouse. Crystals that intersect in the plane of the section, but are clearly two separate crystals were outlined and translated to separate them. The crystal outlines were then filled and exported as tiff files for further processing (Fig. 4). Large thin sections (50 mm by 75 mm) from the same samples were also cut, scanned and digitized in the same way. Most plagioclase megacrysts are not distributed homogeneously, but are clustered in touching groups (Fig. 4). An important qualitative aspect of this texture is the angular relationship of crystals within these clusters. In some of the samples with the coarsest textures plagioclase crystals occur as clusters of sub-parallel crystals (Fig. 4). In samples with smaller plagioclase crystals some of the clusters may have a radiating texture. We have not yet found a way to quantify this aspect of textural variation.

887

JOURNAL OF PETROLOGY

VOLUME 48

NUMBER 5

MAY 2007

MH-04-04

Thalghat

MH-04-06 Kashele

MH-04-14 Manchar

20 mm MH-04-05a Kashele

20 mm MH-04-07 Kashele

20 mm MH-04-16 Manchar

Downloaded from http://petrology.oxfordjournals.org/ by guest on January 4, 2013

20 mm 20 mm MH-04-05b Kashele Kashel e MH-04-10 Thalghat

20 mm

Plagioclase megacrysts in GPB

20 mm 20 mm MH-04-05c Kashele MH-04-12 Thalghat

20 mm

20 mm

Fig. 4. T extural variations of plagioclase megacrysts in GPBs. Samples MH-04-05a, 05b and 05c are from the same flow, a few metres apart. Other samples are from other GPB flows. MH-04-12 is from a plagioclase-phyric flow just above the Thalghat GPB. Images were produced by manual digitizing of crystal outlines in polished slabs.

888

HIGGINS & CHANDRASEKHARAM

SUB-VOLCANIC MAGMA CHAMBERS

T 2: Number of crystals per unit area able


Sample: Area (mm2): MH-04-04 17090 MH-04-05A 39000 MH-04-05B 27000 MH-04-05C 17200 MH-04-06 33800 MH-04-07 10900 MH-04-10 23100 MH-04-12 11860 MH-04-14 10100 MH-04-16 20130

Size bin range (mm) 398251 251158 158100 100631 631398 398251 251158 158100 1000631 06310398 03980251 5 33 39 55 63 35 32 12 9 1 0 0 0 13 25 43 51 43 25 6 4 5 0 11 35 77 83 64 20 9 2 2 0 0 0 8 34 60 59 76 38 16 5 2 1 5 18 27 41 39 51 28 24 9 1 1 7 19 32 25 33 34 29 9 2 0 0 0 23 49 52 39 29 11 1 1 0 0 4 21 62 99 119 94 90 27 11 2 1 4 35 55 77 76 61 52 16 5 1 0 17 31 49 48 59 47 38 16

Downloaded from http://petrology.oxfordjournals.org/ by guest on January 4, 2013

6 1

Q UA N T I TAT I V E T E X T U R A L MEASUREMENTS Introduction


The binary plagioclase images described above were quantified using the program ImageJ, a java version of the wellknown program NIHImage. This program calculates dimensions of a best-fit ellipse to the crystal outlines and its orientation and position. Intersection size data were converted to true crystal size distributions (CSDs) using CSDCorrections 1.3 (Higgins, 2000, 2002). All CSDs were calculated for a shape of 1:5:5 (see below) and a roundness parameter of 02 (close to parallelepiped). CSD calculations are not very sensitive to the weak fabrics observed here, especially for sections cut orthogonal to the foliation; hence a massive fabric was used. Intervals with fewer than two crystals were eliminated from the diagrams, as they are not precise.

Crystal size distribution


CSDs were determined in 10 samples from three GPB groups (Table 2). Three samples from a single flow of the Kashele GPB had CSDs that were almost straight on a classic CSD diagram [S-type CSD of Higgins (2006b)], but lacked small crystals (Fig. 5a). This was a real effect, and not a measurement artefact, as size data were combined from one or more slabs and large thin sections for each sample. Two samples, MH-04-05a and MH-04-05c, were sub-parallel, but separated by 15 ln(population density)(mm4) units. The mean size of the largest interval whose population density could be measured was 12 mm, and there were no plagioclase megacrysts in intervals smaller than 2 mm. Much smaller plagioclase crystals are present in the matrix of these samples, but were not

measured in this study. The third sample, MH-04-05b, was also straight, but extended to larger crystals, in the 20 mm interval. There were no crystals smaller than 35 mm. The slope of this sample was significantly less than that of the other samples. Samples MH-04-06 and MH-04-07 were taken from the same flow as sample MH04-05, about 1km away. They both have CSDs that are similar to those of samples MH-04-05a, MH-04-05b and MH-04-05c. However, sample MH-04-06 continues to larger crystals than samples MH-04-05a and MH-04-05c. The Thalghat GPB was sampled at two locations. Sample MH-04-04, from the Nasik road section, also has an almost straight CSD, the largest crystals of any sample studied and the shallowest slope. An attempt was made to quantify plagioclase megacrysts in the field at this location by measuring crystals on fresh surfaces with a ruler. The CSD for larger crystals followed that determined from the slabs of sample MH-04-04, but the lower cutoff size in the field was only 10 mm, compared with 2 mm in the slabs. Because of the importance of crystals smaller than 10 mm this method was not pursued further. Samples MH-04-10 and MH-04-12 were taken from the Trimbak^Jawhar section. There is a considerable difference in elevation for these two samples, but both are considered to be part of the Thalghat GPB unit. The CSD of sample MH-04-12 was straight and steep, whereas MH-04-10 was collinear with MH-04-04, but did not extend to such large crystals. The Manchar GPB was sampled at two locations. Both CSDs are approximately straight and not dissimilar to CSDs from the other GPBs. The two CSDs are parallel except for the largest size interval, which is not well determined.

889

JOURNAL OF PETROLOGY

VOLUME 48

NUMBER 5

MAY 2007

(a)

4 Kashele GPB 6

(b) 4 Thalghat GPB ln (population density)(mm4) 6 MH-04-12

MH-04-05c

ln (population density)(mm4)

10 MH-04-05a

10

MH-04-10 MH-04-04

12

MH-04-07 MH-04-05b 15 20 MH-04-06 25 30

12

14 0 (c) 4 5 10 Size (mm)

14 35 0 5 10 15 20 25 30 35 Downloaded from http://petrology.oxfordjournals.org/ by guest on January 4, 2013 Size (mm)

Manchar GPB ln (population density)(mm4) 6 Data point MH-04-14 Termination of CSD (no smaller or larger crystals)

10

MH-04-16

12

14 0 5 10 15 20 25 30 35 Size (mm)
Fig. 5. Crystal size distributions in the Kashele (a), Thalghat (b) and Manchar (c) Giant Plagioclase Basalts plotted following the conventions of Higgins (2006a). All CSDs terminate to the left; that is, there are no megacrysts smaller than 2 mm. However, there are very small plagioclase crystals in the matrix that were not measured in this study.

Because all the megacrysts have CSDs close to S-type (Higgins, 2006b) they can be reduced to a characteristic length (Cl 1/slope) and intercept. Higgins (2002) showed that closure produces a correlation between intercept and characteristic length, even if the volumetric phase proportion is variable. Hence, the easiest way to summarize the CSD data is with a graph of characteristic length against modal plagioclase (Fig. 6a). The characteristic length of a CSD says nothing about the largest crystal in the rock. This can be determined precisely only by 3D methods, but can be estimated from the largest intersection (Fig. 6b). In general, there is a good correlation between the largest intersection and the total amount of plagioclase, with the exception of sample MH-04-06.

(e.g. Kostov & Kostov, 1999; Higgins, 2006a). Shape is best examined in three dimensions, but for regular crystals with more or less uniform shapes intersection data can be used to estimate overall shape. Shape is generally expressed in terms of the ratio of short:intermediate:long (S:I:L) for a bounding parallelepiped or best-fit ellipsoid. For parallelepipeds (Higgins, 1994) and triaxial ellipsoids (Higgins, 2006a) the mode of intersection width/intersection length (2D aspect ratio) is equal to the ratio S/I. Euhedral plagioclase crystals do not fit either of these models but are close enough for this simple criterion to be used to estimate accurately this aspect of their overall shape. It is not easy to determine precisely the ratio I/L from intersection data of randomly oriented crystals (Higgins, 2006a). I/L has been estimated from the statistical parameters of the intersection length/width

Crystal shape
The relationship between the shape of crystals in rocks and their petrogenesis has been examined for a very long time

890

HIGGINS & CHANDRASEKHARAM

SUB-VOLCANIC MAGMA CHAMBERS

(a)

6 5 1600

(a) Model plagioclase tablet (101) (001) (101)

(b) Giant phenocryst

Characteristic length (mm)

04

Residence time (yrs)

4 10 3 06 2 1 0 0 5 10 15 Plagioclase (%) 20 05a 05c 12 16 05b 07 14

1200

(110)

(110)

(010)

(001)
Fig. 7. (a) Ideal shape of a plagioclase crystal. Here, the growth rate of {010} is 02 times that of the other faces. This figure was produced using the program WinXMorph (Kaminsky, 2005). (b) Broken surface of a GPB sample with a plagioclase crystal showing the (010) face. The crystal is 20 mm wide.

800

400

0 25 1000

(b) 30 04 07 25 06 14

Largest intersection (mm)

16

05b 12

20 15 10 5 0 0 5 10 15 10 05c 05a

500

ratio will be used to characterize the shape of the plagioclase crystals. Although plagioclase in all samples is strongly tabular the I/S ratio is variable, from 4 to 13 (Fig. 8). All Kashele and Manchar GPB samples have I/S of 4^7. Two of the Thalghat samples also have similar values of I/S, but sample MH-04-10 stands out with I/S 13. There is a weak correlation between the I/S ratio and the characteristic length of the CSDs. For the purposes of calculating the CSDs an aspect ratio of S:I:L 1:5:5 was used for all samples so that they could be readily compared. It should be remembered that changes in the ratio I/L will change the size scale of the CSD, but different values of S/I will change only the tailing corrections (Higgins, 2000).

Downloaded from http://petrology.oxfordjournals.org/ by guest on January 4, 2013

0 20 25

Residence time (yrs)

Crystal orientation
The orientations of crystal outlines were also measured. Orthogonal sections were not measured, hence the true 3D orientation cannot be determined. The sections were cut parallel to the foliation, therefore the only useful parameter is the dispersion of orientations, which is a measure of the quality of foliation Here, we used the normalized length . of the direction vector as the quality parameter (Higgins, 2006a): for perfectly aligned crystals it has a maximum value of one; massive rocks give a value of zero. The orientation quality of the 10 samples studied here does not appear to correlate significantly with crystal shape (Fig. 9) or any other textural parameter; however, the sample with the lowest quality foliation, MH-04-10, also has one of the lowest plagioclase contents and the most tabular crystals.

Plagioclase (%)
Fig. 6. (a) Characteristic length vs plagioclase abundance. (b) Maximum intersection size vs plagioclase abundance. Residence times were calculated for a growth rate of 1010 mm/s.

distribution (Higgins, 1994; Garrido et al., 2001) and by modelling (Morgan & Jerram, 2006). However, none of these methods give precise values, especially for tabular crystals (I/L close to one). Morgan & Jerram (2006) developed a method where the whole intersection length/ width distribution is fitted to model distributions, thus they can obtain S/I and I/L. However, the method does not take into account the much better precision with which the S/I values can be determined, compared with I/L. In this study a more pragmatic approach is used with sections of crystals with special orientations (Higgins, 2006a): plagioclase tablets are commonly flattened along {010} (Fig. 7a). A good cleavage parallel to (010) ensures that the tabular face of plagioclase crystals is sometimes exposed on broken surfaces (Fig. 7b). This can be used with the observation that crystals viewed in the (010) plane are equant to establish that I L. Hence only the S/I (or more conveniently I/S)

DISCUSSION Equilibrium crystallization of phases


The composition of the magma from which the plagioclase megacrysts of the GPB crystallized is of fundamental interest for the calculation of equilibrium crystallization. Ideally, this composition could be determined from melt inclusions in the plagioclase megacrysts; however, this is

891

JOURNAL OF PETROLOGY

VOLUME 48

NUMBER 5

MAY 2007

(a) 4 Frequency density

05b 05c 07 06

Kashele GPB

(b) 6

10 04

Thalghat GPB

Frequency density

2 12

1 05a 0 0 0.2 0.4 0.6 0.8 1 2D aspect ratio (c) 4 Frequency density 16 Manchar GPB (d) 25 3 Frequency density 14 20 15 10 5 0 0 0.2 0.4 0.6 0.8 1 2D aspect ratio (e) 14 10 12 10 8 6 05c 4 05a 1 2 3 06 12 14 07 4 5 16 05b 04 Summary 0 1:07:07 1:04:04 1:12:12 0 0 0.2 0.4

0.6

0.8

2D aspect ratio

Blocks-0.2 rounding

Downloaded from http://petrology.oxfordjournals.org/ by guest on January 4, 2013

0.2

0.4

0.6

0.8

2D aspect ratio

Shape I/S

Characteristic length (mm)


Fig. 8. Intersection shape distributions for samples from (a) Kashele GPB, (b) Thalghat GPB and (c) Manchar GPB. The frequency density is the number of crystals in an interval of 2D aspect ratios divided by the width of the interval and the total number of crystals (see Appendix). (d) Intersection shape distributions of parallelepiped tablets with a rounding factor of 02. (e) Summary of I/S shape ratio vs characteristic length. These values were determined from the mode of the intersection shape distributions. Schematic cross-sections of crystals with I/S 4 and 12 are shown at the left of the diagram.

beyond the scope of this paper. Sano et al. (2001) examined the compositions of melt inclusions in olivine phenocrysts from a regular Deccan basalt that lacked megacrysts. They were able to determine all the more abundant

elements, as well as the water content of the magma, which is essential for modelling crystallization. We used the mean composition of glass inclusions in their sample IC15 to model crystallization using the PELE

892

HIGGINS & CHANDRASEKHARAM

SUB-VOLCANIC MAGMA CHAMBERS

0.5

Orientation Dispersion (quality)

0.4

05b 6 05a 16

0.3 14 0.2 12 4

05c

Massive

0.1

7 10

0.0 4 6 8 Shape I/S


Fig. 9. Orientation dispersion (quality of foliation) vs crystal shape (I/S). Orientation dispersion varies from zero for massive rocks to 10 for parallel crystals. Schematic cross-sections of crystals with I/S 4 and 12 are shown at the bottom of the diagram.

10

12

14

program (Boudreau, 1999), which is a version of MELTS (Ghiorso & Sack, 1995). Plagioclase was the first phase to crystallize when oxygen fugacity was buffered at FMQ (fayalite^magnetite^quartz). The density of this plagioclase, 2670 kg/m3, is considerably less than that of the magma, 2730 kg/m3, confirming that plagioclase floats and hence can accumulate at the top of a magma chamber. It is likely that the original magma from which the GPB megacrysts grew was similar in composition to IC15 melt inclusions and hence plagioclase also crystallized first and floated in the GPB magma.

Crystal size
In igneous petrology we are concerned commonly with the balance between kinetic and equilibrium processes (Higgins, 2006a). This is expressed by the kinetic processes of nucleation and growth, which are controlled by the degree of undercooling (or supersaturation) of the magma. If the crystallization driving force is reduced then the texture of the magma or rock will adjust to minimize the overall energy of the system. This can be achieved by coarsening, also known as Ostwald ripening or textural equilibrium. Mechanical processes can also further modify both kinetic and equilibrium textures. Finally, increases in undercooling may rejuvenate kinetic processes. It is not always easy to distinguish kinetic and equilibrium textures, especially as perfect equilibrium is never achieved. We will start with a simple kinetic model for crystal size development. A simple kinetic model for the GPB is crystallization from a single magma in a chamber that is continuously

filled and drained. Marsh (1988) showed that in such a steady-state system the CSDs are S-type and that the mean residence time of a crystal characteristic length/ growth rate. The choice of growth rate is clearly very important. Cashmans (1993) compilation of data for basaltic systems suggested a growth rate of 109 mm/s for a cooling time of 3 yearsperhaps typical of thicker lava flows. For a cooling time of 300 years she suggested a growth rate of 1010 mm/s. This is similar to what has been estimated for the Palisades sill (Cashman, 1993). Growth rates for plagioclase in the Makaopuhi lava lake were slightly greater at (35^65) 1010 mm/s [data from Cashman & Marsh (1988), recalculated by Higgins (2006a)]. Crystallization at greater depths would have occurred at lower degrees of undercooling and hence slow growth rates: if the relationship between cooling time and growth rate continues to be linear then a cooling time of 30 000 years would give a growth rate of 1011 mm/s. A plagioclase growth rate of 1010 mm/s is chosen here as cooling must have been beneath a lava pile, at the very least, and hence slower than the Makaopuhi lava lake. This should probably be considered a maximum value, as crystallization may have occurred at greater depths. It should be noted that Sen et al. (2006) used a growth rate of 109 mm/s in their study of the Deccan GPB. If plagioclase megacrysts crystallized in a steady-state system with a growth rate of 1010 mm/s, then residence times vary from 600 to 1500 years (Fig. 6a). Another kinetic model involves simple, continuous growth of a crystal. For the same growth rate the largest crystal, as estimated from the largest intersection, would grow in 500^1000 years (Fig. 6b). The two methods correlate loosely, except for two points. The CSD method is more robust as it depends on the whole population of crystals, whereas the second method depends only on one crystal, the largest. The volume of the magma chamber where the megacrysts grew can also be loosely estimated: it equals the residence time multiplied by the refilling rate. If all Deccan magmas passed through the magma chamber at this time then the refilling rate equals the eruption rate. The eruption rate of a volcanic province equals the volume divided by the duration of volcanism. In the Deccan both these parameters are poorly known. The most recent compilation of eruption rates (White et al., 2006) proposed a value of 09 km3/year for flood basalt provinces, which would give a chamber volume of 450^1350 km3. These values would be proportionally lower if magma bypassed the megacryst-bearing magma chamber. The lower limit is only about twice that of the Skaergaard intrusion, Greenland, recently estimated at 280 km3 (Nielsen, 2004). This intrusion was emplaced beneath flood basalts and has a roof with plagioclase crystals similar to those seen in the GPB (Naslund, 1984).

Foliated

Downloaded from http://petrology.oxfordjournals.org/ by guest on January 4, 2013

893

JOURNAL OF PETROLOGY

VOLUME 48

NUMBER 5

MAY 2007

Larger volume chambers are also possible: Higgins (2005) has proposed that the Sept Iles intrusion, Canada, with a volume of 35 000 km3 was emplaced beneath flood basalts. It, too, has a roof that contains large, tabular plagioclase crystals similar to the GPB megacrysts. However, such a large intrusion would probably produce a distinctive gravity and magnetic anomaly, which is not observed in the Deccan. Sen et al. (2006) have proposed that the eruption duration was only 23 000 years, which would give an eruption rate of 87 km3/year. The chamber volume would then expand to 50 000^130 000 km3, which seems to be unrealistically large and would certainly be evident from geophysical measurements. The steady-state and simple growth models presented above produce a single CSD. Clearly, the different GPBs have different CSDs and hence the situation must be more complex. There are a number of processes that can change the CSD. Compaction of a crystal mush involves loss of intercrystal fluid, whereas mixing with an aphyric magma (dilution) increases the amount of intercrystal fluid; hence the two processes are texturally similar. In neither case is Cl changed, hence the analysis discussed above is unchanged. On a plot of Cl vs volumetric phase proportion compaction will displace the points to the right, whereas dilution will produce a vector towards the left. Within each packet of GPB flows, or even within a single GPB flow, the variation of Cl exceeds analytical error, hence dilution and compaction cannot account for all the textural variation. Clearly, another mechanism must be invoked that can change the Cl. If the overall crystal growth rate increases, then the Cl will also increase. However, we would then expect to see a correlation between Cl and the volumetric abundance of plagioclase, which is not observed. Hence, pure kinetic steady-state models do not seem to be able to account for textural observations. Before we leave pure kinetic models the approach of Sen et al. (2006) must be discussed. Crystals grow not only in chambers, but also in conduits. Sen et al. (2006) proposed that plagioclase megacrysts grew in a chicken-wire mesh across a conduit, bathed by a quasi-continuous flow of magma. They applied the Johnson^Mehl^Avrami equation and found a value of 1580 years, for a growth rate of 109 mm/s. If they had used a growth rate of 1010 mm/s, as above, then they would have found a growth time of 15 800 years. It seems unlikely that a dyke would have been continuously active for such a long period of time. Also, this model does not fit the idea of Sen (2001) that the GPBs develop during times of magmatic repose. Finally, it will be shown that the textures of the megacrysts do not seem to support such a model. The lack of small megacrysts in the GPB magmas suggests that the system is returning towards an equilibrium texture by coarsening. This is the process by which small

crystals dissolve at the same time as larger crystals are growing, so that the overall surface energy is minimized (Voorhees, 1992). It occurs when the temperature of the system is maintained close to the mineral liquidus. Under these conditions the nucleation rate is zero, but the growth rate is significant. Higgins (1998) showed that coarsening following the Communicating Neighbours model of Dehoff (1991) will lead to increases in characteristic length. If the system is allowed to be open (i.e. material is added from circulating fluids), then the volumetric phase proportion can also increase. The phase proportion can also increase if enthalpy is withdrawn from the system at low degrees of undercooling. Either process will give a diagonal vector on a diagram of characteristic length vs volumetric phase abundance. The distribution of data points in Fig. 6 suggests that we are seeing the combined effect of coarsening of a megacryst-rich magma and mixing with an aphyric magma. We will now consider the shape of the megacrysts and what it can tell us about the crystallization environment.

Downloaded from http://petrology.oxfordjournals.org/ by guest on January 4, 2013

Crystal shape
The shape of plagioclase crystals has been determined quantitatively in a number of studies, as an estimate of shape is necessary for the calculation of CSDs from intersection lengths. Higgins (2006a) has summarized these studies and concluded that tabular crystals occur in two environments: as microlites in lavas and as much larger crystals in laminated cumulate rocks such as anorthosites and troctolites. He proposed that tabular growth occurs

(001)
(010)

(001)

Face growth rate

Rate{001}

Rate{010} Chemical potential gradient


Fig. 10. Schematic illustration of the response of plagioclase crystal faces to the chemical potential gradient around the growing crystal.  It is assumed that the growth rates of the {110}, {101} and {1 10} faces are the same as that of the {001} faces. The overall size of the illustrated crystal reflects both growth rates and crystallization time.

894

(010)

HIGGINS & CHANDRASEKHARAM

SUB-VOLCANIC MAGMA CHAMBERS

in situations where there is a high chemical potential gradient around the growing crystal (Fig. 10). This is because the growth rate of the {010} faces responds less than the other faces to increases in the chemical potential gradient. It should be noted that experimental studies of plagioclase morphology vs growth have been primarily concerned with growth forms at large degrees of undercooling, such as the transitions from laths to skeletal forms to spherulites (Lofgren, 1974). Here we are concerned with variation of the shape of euhedral crystals, which form at much lower degrees of undercooling where time constraints limit experiments. The chemical potential gradient around a growing crystal is controlled by the addition of crystal nutrients and the removal of unwanted components. These occur in response to diffusion and advection (relative movement between the crystal and the growing medium). The ratio of the time scales of mass transfer by advection and diffusion is called the mass transfer Peclet number (PeMT) and is defined as PeMT vl D

PeT

vlc k

where v and l are defined as before,  is the density, c is the specific heat and k is the thermal conductivity. If the thermal Peclet number is high then latent heat will be removed by advection rather than conduction. As before, the thermal Peclet number can be estimated for a basaltic magma under typical conditions where the density is 2730 kg/m3 (from calculations using PELE; see section 41), the thermal conductivity is 14 W/m per K (Horai & Susaki, 1989) and the specific heat is 840 J/kg per K (Clarke, 1966). A crystal size of 10 mm and an advection velocity of 1mm/s (as above) would give a thermal Peclet number of 16. Hence, under these conditions heat is removed by advection and not conduction. The shape of the crystals can therefore give an idea of the environment of crystallization. The most tabular plagioclase crystals must have grown in an environment with the strongest advection, that is shearing or stirring.

Downloaded from http://petrology.oxfordjournals.org/ by guest on January 4, 2013

Crystal orientation and clustering


If the crystal mush was transported from the magma chamber by laminar flow then the relative orientation of the crystals could be preserved. We would then expect that well-foliated samples formed in a strongly sheared environment and hence would have the most tabular crystals. The absence of a significant correlation between crystal shape and quality of foliation suggests that transport was turbulent. Although quantitative measures of the overall orientation of crystals do not appear to be very helpful, qualitative observations of short-range order may help clarify the petrogenesis of the GPB. In many samples tabular plagioclase occurs as clusters of sub-parallel crystals. This structure has been termed synneusis (swimming together) and is considered to form during flow (Schwindinger, 1999). The driving force is the minimization of the surface energy of the crystal aggregate (Ikeda et al., 2002), as in the processes of coarsening, evidenced here by the shape of the CSDs. Some samples have radiating clusters of crystals, rather than sub-parallel aggregates (e.g. MH-04-12; Fig. 4). It could be significant that these samples are poorly foliated and have CSDs with the lowest slopes. This suggests that in some samples a later phase of crystal growth may have occurred rapidly in a static environment. We will now discuss the occurrence of plagioclase megacrysts in other igneous rocks, to allow the creation of a plausible model for the origin of GPB.

where v is the velocity of the growing medium with respect to the crystal, l is the length scale (size of crystal), and D is the chemical diffusivity (diffusion coefficient). If the Peclet number is high then advection will renew the supply of nutrients around the crystal and the chemical potential gradient will be greater than that which is possible by diffusion alone. In the case of microlites a high growth rate ensures a high Peclet number because the ends of the crystal outpace the growth of the depleted zone around the crystal. Mechanical movement of magma (stirring) by convection currents can also ensure a high Peclet number and has been proposed as an explanation for tabular crystals (Kouchi et al., 1986; Higgins, 1991). Such conditions may occur near the margins of magma chambers, where there is a significant velocity gradient. The mass transfer Peclet number in a magma chamber can be estimated for a basaltic magma under typical conditions. Chemical diffusivity varies with magma composition, temperature and element (Chakraborty, 1995). Order of magnitude values for a mafic magma at 13008C are 1011 m2/s for network-modifiers such as Ca and Mg, 109 m2/s for Na and K, and 1012 m2/s for network formers such as Si. If we take a typical value of D of 1011 m2/s, a length scale equal to the length of a typical crystal, 10 mm, then a convection velocity of 1mm/s would give a Peclet number of 106. Hence, advective transport is much more important than diffusive transport where crystals are exposed to convection currents. The thermal Peclet number (PeT) can also be used to determine if transport of latent heat away from a growing crystal is controlled by advection or diffusion. Here the equation is

Plagioclase megacrysts in other volcanic and plutonic rocks


Basaltic lavas with plagioclase megacrysts similar to the GPBs are uncommon but widespread. They occur in

895

JOURNAL OF PETROLOGY

VOLUME 48

NUMBER 5

MAY 2007

oceanic basalts in a wide variety of settings: spreading ridges, intraplate hotspots, aseismic ridges and fracture zones (Cullen et al., 1989). Plagioclase megacrysts occur in certain phases of the Surtsey volcano and sporadically in the Eldfel lavas (Furman et al., 1991). They are also well documented on the northern Galapagos Islands (Cullen et al., 1989). Here, the plagioclase megacrysts are heterogeneously distributed and have a texture very similar to the GPBs: plagioclase crystals are tabular and occur in subparallel glomerocrysts. The cores of the megacrysts are more calcic than the rims, which are in equilibrium with the host magma, as in the GPBs. Many mafic intrusions contain layers or zones of rocks rich in plagioclase, parts of which may be laminated. For instance, the upper border zone of the Skaergaard intrusion, Greenland, is partly composed of well-laminated leucogabbro (Naslund, 1984). The position of this unit necessitates that it formed by accumulation of plagioclase by flotation. This reflects the greater density of evolved, iron-rich, magmas with respect to plagioclase (Scoates, 2000). The upper part of the Sept Iles intrusive suite, Canada, is dominated by anorthosite, some of which is well laminated with tabular crystals (Higgins, 1991). Recently, Higgins (2005) has proposed that these rocks were also formed by flotation of plagioclase. Partial disruption of these plagioclase cumulates by granitic fractionates shows that such accumulations remained very loose and poorly cemented, even at high crystal concentrations. In both cases convection currents would have brought plagioclase primocrysts up from lower levels and their low density would have allowed them to remain in the upper border zone. These currents would have produced a local environment with high Peclet numbers and hence the plagioclase crystals would have tended to grow with a tabular habit. It is now possible to integrate field data, quantitative textural measurements and the occurrence of megacrysts in other environments in the form of an emplacement model for the GPBs.

Emplacement model of GPB flows


We propose that a cycle of magmatism started with the eruption of normal basalts, with few or no megacrysts (Fig. 11a). Such magmas were derived from the mantle, but in most cases must have been stored deep in the crust, where there was differentiation and possibly contamination by the lower crust. No large magma chambers are envisaged, but more localized swellings in the conduits. Magmas rise to the surface along faults, probably related to rifting (Hooper, 1990). Such a cycle may end when the magma is unable to penetrate the lava pile or perhaps when the production rate of magma wanes. At this point erosion of fresh lavas may produce a regolith, now preserved as the red boles. The next stage of the cycle starts with plutonism and may continue directly without a pause. Magmatism

continues, perhaps at a lower rate, but the magma now starts to accumulate at depth. The most likely place for this is along the interface between the lava pile and the basement, where there is likely to be a density contrast (Fig. 11b). The magma may wedge its way out between the base of the lava pile and the basement (right side of Fig. 11b), or tectonic forces may produce a rift (central part of Fig. 11b). The chamber will be filled gradually with hot, new magma, ensuring that vigorous convection occurs. Crystallization of plagioclase and mafic minerals produces dense, iron-rich magmas and basal cumulates. Plagioclase will float if its density is less than that of the magma, but the density difference is not usually large. Plagioclase crystals may nucleate and start to grow at depth and be wafted by convective currents up to the roof, where some crystals will tend to remain, or they may nucleate and grow at the top of the chamber. Passage of convection currents will keep the crystals bathed in hot magma, such that they remain close to their liquidus temperature and coarsen. Such currents will also ensure a high Peclet number environment conducive to the crystallization of tabular crystals. The upper border zone of the magma chamber will comprise a loose crystal mush that is easily remobilized. Mafic minerals will be denser than the magma and will continue to accumulate along the base of the magma chamber. This situation could continue until the magma has solidified completely as a pluton, or is disrupted by tectonic forces. The next stage is the transport of the megacryst-bearing magma to the surface. Reopening of a channel from the partially solidified magma chamber to the surface might result from rejuvenation of existing faults, or by the initiation of new faults (Fig. 11c). Once a conduit has formed magma will be drained from the chamber and feed flows of GPB basalts. The magma drawn from the uppermost layer will be rich in megacrysts and may mix turbulently with more evolved magmas devoid of crystals drawn from lower levels of the magma chamber. Some magma may continue to crystallize in higher-level static staging areas, producing radiating clusters of crystals. Once magma is drained from the chamber the cycle of normal basalt flows may recommence. Sen et al. (2006) have proposed two models for the origin of the GPB, the first of which somewhat resembles the model proposed above. In their first model the plagioclase megacrysts form statically on the walls of a magma chamber in the advancing solidification front (Marsh, 1996). However, we have suggested above that the plagioclase megacrysts preserve evidence of accumulation by flotation and crystallization in a dynamic regime, neither of which processes is included in the model of Sen et al. Their second model, which they developed quantitatively, proposes that the megacrysts grew in a magma conduit, where they formed a chicken-wire network. Magma flowing along the conduit

Downloaded from http://petrology.oxfordjournals.org/ by guest on January 4, 2013

896

HIGGINS & CHANDRASEKHARAM

SUB-VOLCANIC MAGMA CHAMBERS

(a) Eruption of normal flood basalts

Magma conduit Faults

(b) Surface weathering and sub-volcanic plutonism Red bole

Downloaded from http://petrology.oxfordjournals.org/ by guest on January 4, 2013

Magma chamber Plagioclase flotation cumulates Magma conduit Hot, rising magma Coarsening

(c) Draining of magma chamber and eruption of Giant Phenocryst Basalts

Magma chamber

Plagioclase flotation cumulates

Mafic cumulates

Magma conduit

Fig. 11. Model for the formation of Giant Plagioclase Basalt flows, Deccan.

fed the crystals, accounting for their great size. Subsequent flow disrupted the network and incorporated the megacrysts into a low-crystallinity magma. This model has the advantage that the crystals grew in a dynamic environment, but cannot produce the observed textures: there is no reason why plagioclase should occur as groups of aligned crystals. It also seems unlikely that a chicken-wire network could exist for the 1538 years that Sen et al. (2006) calculated. Finally, the theory does not account for the presence of similar plagioclase crystals at the top of sub-volcanic magma chambers such as Skaergaard.

CONC LUSIONS
Plagioclase megacrystic basalts are a widespread, but variable component of flood basalt provinces. They should be viewed as evidence of magma storage in the crust. Plagioclase is the most common mineral to occur as megacrysts in these rocks because it is the only mineral that can float in evolved basaltic magmas. Hence, the presence of plagioclase megacrysts indicates that sub-volcanic magma chambers were an important component of the magmatism. The Skaergaard intrusion, Greenland, is probably a good model for the magma

897

JOURNAL OF PETROLOGY

VOLUME 48

NUMBER 5

MAY 2007

chamber in which the megacrysts of the GPB grew. Larger sub-volcanic intrusions, such as Sept Iles, Canada, could produce the megacrysts, but would produce a significant geophysical anomaly, which is not observed. Quantitative studies of plagioclase megacrysts can provide information on the duration of crystallization and possibly the size of such magma chambers, but cannot give an idea of the total eruption duration, as has been proposed by Sen et al. (2006). This is because there is no evidence that all magmas have passed through the same magma chamber. It is likely that sub-volcanic magma chambers existed intermittently, just before the GPBs formed. At other times magma passed almost unchanged from the lower crust or the mantle source to the surface.

AC K N O W L E D G E M E N T S
This research was partly funded by operating grants from the Natural Science and Engineering Research Council of Canada to M.D.H. We would like to thank Hazel Jenkins, Melroy Borges, Gautam Sen and Ayaz Alam. The manuscript was improved by the insightful reviews of Dougal Jerram, James Scoates and Dick Naslund.

R EF ER ENC ES
Beane, J., Turner, C., Hooper, P., Subbarao, K. & Walsh, J. (1986). Stratigraphy, composition and form of the Deccan basalts, Western Ghats. Bulletin Volcanologique 48, 61^83. Boudreau, A. E. (1999). PELEa version of the MELTS software program for the PC platform. Computers & Geosciences 25, 201^203. Cashman, K. V. & Marsh, B. D. (1988). Crystal size distribution (CSD) in rocks and the kinetics and dynamics of crystallisation II. Makaopuhi lava lake. Contributions to Mineralogy and Petrology 99, 292-305.. Cashman, K. V. (1993). Relationship between plagioclase crystallisation and cooling rate in basaltic melts. Contributions to Mineralogy and Petrology 113, 126^142. Chakraborty, S. (1995). Diffusion in silicate melts. In: Stebbins, J. & McMillan, P. (eds) Structure, Dynamics and Properties of Silicate Melts Reviews in Mineralogy, 32, Washington, DC: Mineralogical Society of America, pp. 411^504. Chandrasekharam, D. (2003). Deccan flood basalts. Geological Society of India Memoir 53, 197^214. Chandrasekharam, D., Mahoney, J. J., Sheth, H. C. & Duncan, R. A. (1999). Elemental and Nd^Sr^Pb isotope geochemistry of flows and dikes from the Tapi rift, Deccan flood basalt province, India. Journal of Volcanology and Geothermal Research 93, 111^123. Clarke, S. (1966). Handbook of Physical Constants. Geological Society of America Memoir 97. Cullen, A., Vicenzi, E. & McBirney, A. R. (1989). Plagioclase-ultraphyric basalts of the Galapagos Archipelago. Journal of Volcanology and Geothermal Research 37, 325^337. Dehoff, R. T. (1991). A geometrically general theory of diffusion controlled coarsening. Acta Metallurgica et Materialia 39, 2349^2360. Duncan, R. A. & Pyle, D. G. (1988). Rapid eruption of the Deccan flood basalts at the Cretaceous^T ertiary boundary. Nature 333, 841^843.

Furman, T., Frey, F. A. & Park, K.-H. (1991). Chemical constraints on the petrogenesis of mildly alkaline lavas from Vestmannaeyjar, Iceland; the Eldfell (1973) and Surtsey (1963^1967) eruptions. Contributions to Mineralogy and Petrology 109, 19^37. Garrido, C. J., Kelemen, P. B. & Hirth, G. (2001). Variation of cooling rate with depth in lower crust formed at an oceanic spreading ridge; plagioclase crystal size distributions in gabbros from the Oman Ophiolite. Geochemistry, Geophysics, Geosystems 2, 2000GC000136. Ghiorso, M. S. & Sack, R. O. (1995). Chemical mass-transfer in magmatic processes. 4. A revised and internally consistent thermodynamic model for the interpolation and extrapolation of liquid^solid equilibria in magmatic systems at elevated temperatures and pressures. Contributions to Mineralogy and Petrology 119, 197^212. Higgins, M. D. (1991). The origin of laminated and massive anorthosite, Sept Iles intrusion, Quebec, Canada. Contributions to Mineralogy and Petrology 106, 340^354. Higgins, M. D. (1994). Determination of crystal morphology and size from bulk measurements on thin sections: numerical modelling. American Mineralogist 79, 113^119. Higgins, M. D. (1998). Origin of anorthosite by textural coarsening: quantitative measurements of a natural sequence of textural development. Journal of Petrology 39, 1307^1325. Higgins, M. D. (2000). Measurement of crystal size distributions. American Mineralogist 85, 1105^1116. Higgins, M. D. (2002). Closure in crystal size distributions (CSD), verification of CSD calculations, and the significance of CSD fans. American Mineralogist 87, 171^175. Higgins, M. D. (2005). A new model for the structure of the Sept Iles Intrusive suite, Canada. Lithos 83, 199^213. Higgins, M. D. (2006a). QuantitativeT extural Measurements in Igneous and Metamorphic Petrology.Cambridge: Cambridge University Press. Higgins, M. D. (2006b). Use of appropriate diagrams to determine if crystal size distributions (CSD) are dominantly semi-logarithmic, lognormal or fractal (scale invariant). Journal of Volcanology and Geothermal Research 154, 8^16. Hofmann, C., Feraud, G. & Courtillot, V. (2000). 40Ar/39Ar dating of mineral separates and whole rocks from the Western Ghats lava pile: further constraints on duration and age of the Deccan traps. Earth and Planetary Science Letters 180, 13^27. Hooper, P., Subbarao, K. & Beane, J. (1988). The giant plagioclase basalts (GPBs) of the western Ghats, Deccan Traps. In: Subbarao, K. (ed.) Deccan Flood Basalts Memoir 10, Geological Society of India, New Delhi, India, pp. 134^144. Hooper, P. R. (1990). The timing of crustal extension and the eruption of continental flood basalts. Nature 345, 246^249. Horai, K. & Susaki, J. (1989). The effect of pressure on the thermal conductivity of silicate rocks up to 12 kbar. Physics of the Earth and Planetary Interiors 55, 292^305. Ikeda, S., T oriumi, M., Y oshida, H. & Shimizu, I. (2002). Experimental study of the textural development of igneous rocks in the late stage of crystallization: the importance of interfacial energies under non-equilibrium conditions. Contributions to Mineralogy and Petrology 142, 397^415. Jerram, D. A. & Widdowson, M. (2005). The anatomy of Continental Flood Basalt Provinces: geological constraints on the processes and products of flood volcanism. Lithos 79, 385^405. Kaminsky, W. (2005). WinXMorph: a computer program to draw crystal morphology, growth sectors and cross sections with export files in VRML V2.0 utf8-virtual reality format. Journal of Applied Crystallography 38, 566^567.

Downloaded from http://petrology.oxfordjournals.org/ by guest on January 4, 2013

898

HIGGINS & CHANDRASEKHARAM

SUB-VOLCANIC MAGMA CHAMBERS

Kostov, I. & Kostov, R. I. (1999). Crystal Habits of Minerals Prof. Marin Drinov Academic Publishing House; Pensoft Publishers, Sofia, Bulgaria. Kouchi, A., Tsuchiyama, A. & Sunagawa, I. (1986). Effects of stirring on crystallization of basalt: texture and element partitioning. Contributions to Mineralogy and Petrology 93, 429^438. Lofgren, G. E. (1974). An experimental study of plagioclase crystal morphology: isothermal crystallization. American Journal of Science 274, 243^273. Marsh, B. D. (1988). Crystal size distribution (CSD) in rocks and the kinetics and dynamics of crystallization I. Theory. Contributions to Mineralogy and Petrology 99, 277^291. Marsh, B. D. (1996). Solidification fronts and magmatic evolution. Mineralogical Magazine 60, 5^40. Morgan, D. J. & Jerram, D. A. (2006). On estimating crystal shape for crystal size distribution analysis. Journal of Volcanology and Geothermal Research 154, 1^7. Naslund, H. R. (1984). Petrology of the Upper Border Series of the Skaergaard Intrusion. Journal of Petrology 25, 185^212. Nielsen, T. F. D. (2004). The shape and volume of the Skaergaard Intrusion, Greenland: implications for mass balance and bulk composition. Journal of Petrology 45, 507^530. Sano, T., Fujii, T., Deshmukh, S. S., Fukuoka, T. & Aramaki, S. (2001). Differentiation processes of Deccan Trap basalts: contribution from geochemistry and experimental petrology. Journal of Petrology 42, 2175^2195. Schwindinger, K. R. (1999). Particle dynamics and aggregation of crystals in a magma chamber with application to Kilauea Iki olivines. Journal of Volcanology and Geothermal Research 88, 209^238. Scoates, J. S. (2000). The plagioclase^magma density paradox reexamined and the crystallization of Proterozoic anorthosites. Journal of Petrology 41, 627^649. Sen, G. (2001). Generation of Deccan Trap magmas. Proceedings of the Indian Academy of SciencesEarth and Planetary Sciences 110, 409^431. Sen, G. (2002). Giant Plagioclase Basalts, eruption rate versus time response to Sheths comments and some additional thoughts. Proceedings of the Indian Academy of SciencesEarth and Planetary Sciences 111, 487^488. Sen, G., Borges, M. & Marsh, B. (2006). A case for short duration of Deccan Trap eruption. EOS Transactions, American Geophysical Union 87, 197^198. Subbarao, K. V., Chandrasekharam, D., Navaneethakrishnan, P. & Hooper, P. R. (1994). Stratigraphy and structure of parts of the central Deccan basalt province: eruptive models. In: Subbarao, K. V. & Radhakrishna, B. P. (eds) Volcanism: Radhakrishna Volume. New Delhi: Wiley Eastern, pp. 321^332. Voorhees, P. W. (1992). Ostwald ripening of two-phase mixtures. Annual Review of Materials Science 22, 197^215. White, S. M., Crisp, J. A. & Spera, F. J. (2006). Long-term volumetric eruption rates and magma budgets. Geochemistry, Geophysics, Geosystems 7, Q03010.

A P P E N D I X : C A L C U L AT I O N O F F R EQU E NC Y DE NSI T Y
The modal value of the 2D aspect ratio is not easy to determine precisely for strongly skewed distributions, as we have here. The conventional approach is to use bins of fixed width and compile a simple frequency diagram (e.g. Higgins, 1994). If too many bins are used then the frequency diagram is very irregular, whereas too few do not make it possible to define exactly the modal value. A new diagram of frequency density is proposed here, which is analogous to simple population density diagrams for CSDs (Higgins, 2006a). The 2D aspect ratios of the intersections ( values) are sorted into ascending order and divided into a number of bins, each with approximately the same number of intersections. T bins were used here, en so each bin had 20^40 intersections. The frequency density is the number of intersections in each bin divided by the difference between the upper and lower 2D aspect ratio bounds and the total number of intersections. This diagram has the advantage that the 2D aspect ratio bins are narrowest around the modal value and hence can define it precisely (Fig. A1). An example of the calculation is shown in Table A1.

Downloaded from http://petrology.oxfordjournals.org/ by guest on January 4, 2013

6 5 Fixed number bins

Frequency density

4 3 2 1 Fixed width bins 0 0 0.2 0.4 0.6 0.8 1 2D aspect ratio

Fig. A1. Comparison of 2D aspect ratio diagrams compiled with fixed width and fixed number bins. The peak is defined by two intervals for fixed width bins but six for fixed number bins.

899

JOURNAL OF PETROLOGY

VOLUME 48

NUMBER 5

MAY 2007

T A1: Calculation of frequency density for sample MH-04-04 able


values of crystal intersections mean range Number of intersections Frequency density

00371, 00427, 00566, 00594, 00631, 00653, 00664, 00676, 00687, 00732, 00736, 00745, 00745, 00761, 00765, 00779, 00794, 00795, 00801, 00822, 00831, 00842, 00844, 00862, 00872, 00877, 00880, 00883 00892, 00914, 00942, 00954, 00973, 00982, 00987, 00992, 01005, 01010, 01014, 01015, 01020, 01029, 01034, 01042, 01042, 01050, 01051, 01056, 01059, 01064, 01077, 01087, 01097, 01105, 01107, 01100 ...

00737

00883 00371 00512

28

2186

01025

01100 00892 00208

28

4568

Downloaded from http://petrology.oxfordjournals.org/ by guest on January 4, 2013

...

...

...

...

is the 2D aspect ratio of measured crystal intersections. Ten bins of 28 or 29 intersections made the total of 284 intersections.

900

Potrebbero piacerti anche