Silicic Segregations of the Ferrar Dolerite Sills, Antarctica

JOURNAL OF PETROLOGY VOLUME 52 NUMBER 10 PAGES1927^1964 2011 doi:10.1093/petrology/egr035 

Silicic Segregations of the Ferrar Dolerite Sills, 

Antarctica 

KARINA ZAVALA 1 *, ALISON M. LEITCH 2 ANDGEORGEW.FISHER 3 

1 BRUNEAU CENTRE FOR RESEARCH AND INNOVATION, MEMORIAL UNIVERSITY OF NEWFOUNDLAND, 

ST. JOHN’S, NL A1C 5S7, CANADA 

2 

DEPARTMENT OF EARTH SCIENCES, MEMORIAL UNIVERSITY OF NEWFOUNDLAND, ST. JOHN’S, NL A1B 3X5, 

CANADA 

3 

DEPARTMENT OF EARTH AND PLANETARY SCIENCES, JOHNS HOPKINS UNIVERSITY, BALTIMORE, MD 21218, USA 

RECEIVED JULY 21, 2007; ACCEPTED JUNE 28, 2011 

The upper parts of the 300 m thick Basement Sill, one of the Ferrar 

Dolerite Sills in Antarctica, contain sharply defined, coarse-grained, 

leucocratic segregations that are temporally, spatially and chemically 

related to the host dolerite. The morphologies of the segregations 

vary gradationally with distance from the upper contact of the 

Basement Sill from thin (1cm) streaks, pods and stringers close to 

the upper contact, (0·03^0·2 m) pipes and veins about 15 m below 

the contact, and (0·5^2 m) sub-horizontal, anastomosing lenses 

about 30 m beneath the contact.Their compositions range from diorite 

to granodiorite; they are enriched in P 2O 5,TiO 2,FeO T, and to a 

lesser extent in alkalis compared with the host dolerite, and they display 

linear trends in Harker variation diagrams that are compatible 

with an origin as fractionated liquids formed after simultaneous 

crystallization of pyroxene and plagioclase from the host sill 

magma. The segregations show systematic variations in mineralogy 

and texture with composition, with the lowest SiO 2 segregations 

having a similar mineralogy to the host and ophitic to sub-ophitic 

textures; the highest SiO 2 segregations are granodioritic and exhibit 

a granophyric texture. In segregations with highly siliceous compositions, 

plagioclase crystals have resorbed edges and pyroxene forms 

blebby intergrowths with plagioclase and Fe^Ti oxides. Most segregations 

are interpreted to have formed when gravitational instability 

caused the upper solidification front of the sill to tear, and mixtures 

of crystals and residual liquid entered the tears by porous flow from 

the surrounding crystal mush and by channelling from lower levels 

in the sill. Modeling using the thermodynamic program MELTS 

shows that the bulk compositions of the segregations are consistent 

with wet (0·5wt % H 2O) fractional crystallization of the host 

dolerite magma at 100 MPa, although the data are not well matched 

by a single parent magma composition. We present a physical model 

*Corresponding author: E-mail: kzavala@mun.ca 

of segregation formation involving multiple episodes of tearing triggered 

by fresh influxes of magma into the sill. 

KEY WORDS: Ferrar Dolerites; segregations; mush instability; differentiation; 

lenses 

INTRODUCTION 

The role of fractional crystallization in producing the diversity 

of igneous rocks has proven to be one of the most 

persistent problems in igneous petrology (e.g. Harker, 

1909; Bowen, 1928; Holmes, 1931; Wager & Deer, 1939; 

Yoder, 1973; McBirney, 1975). Lenses of coarse-grained, leucocratic 

material in the central and upper parts of thick 

(4100 m) basaltic sills, lava lakes, and flood basalts are 

among the richest sources of information on how basaltic 

magmas have differentiated in nature. In this study we 

refer to these leucocratic lenses as segregations, following 

the usage of Tomkeieff (1928), Richter & Moore (1966), 

Moore & Evans (1967), Wright et al. (1976), Wright & 

Okamura, 1977; Helz (1980, 1987), Helz et al. (1989),Marsh 

(1996, 2002, 2004) and Boudreau & Simon (2007). These 

features have elsewhere been described as pegmatites 

(Cornwall, 1951; Walker, 1953; Ragland & Arthur, 1985; 

Greenough & Dostal, 1992; Steiner et al., 1992; Puffer & 

Horter, 1993; Puffer & Volkert, 2001; Philpotts et al., 1996) 

and as pegmatoids (Lindsley et al., 1971; Bailey, 1989; 

Larsen & Brooks, 1994; Be¤ dard et al., 2007). 

ß The Author 2011. Published by Oxford University Press. All 

rights reserved. For Permissions, please e-mail: journals.permissions@ 

oup.com 

Downloaded from 

http://petrology.oxfordjournals.org/ by guest on April 30, 2013

JOURNAL OF PETROLOGY VOLUME 52 NUMBER 10 OCTOBER 2011 

The Jurassic age Ferrar Dolerite Sills (FDS) of the Dry 

Valleys region, Victoria Land, Antarctica, contain unusually 

well-exposed examples of such segregations 

(Marsh, 2004). Most are sub-horizontal 0·5^2 m thick 

lenses in the upper 100 m of the sills, containing crystals 

one to two orders of magnitude larger than those of the 

host dolerite, and have sharp upper and lower contacts. A 

few form smaller sub-vertical veins and amorphous pods. 

Some have gradational contacts with the host dolerite. 

The compositions of these segregations span the full range 

from the host dolerite to granite, and so offer an unusually 

detailed window on differentiation mechanisms in basaltic 

magma systems. 

In this study we develop a new petrogenetic model for 

the formation of the segregations based on detailed field 

observations made by K. Zavala and B. D. Marsh in 2001, 

together with petrographic data, mineral chemistry and 

bulk-rock major element geochemistry combined with 

MELTS modelling. 

GEOLOGICAL SETTING 

The FDS are well exposed in the McMurdo Dry Valleys 

region (Fig. 1) c. 20 km inland from the Ross Sea between 

the McKay Glacier to the north (c. 10 km north of Fig. 1b) 

and the Ferrar Glacier to the south. The basement in this 

region consists of Precambrian to Devonian metamorphic 

rocks and interbedded marbles, hornfels, and schists of 

the Koettlitz group (Gunn & Warren, 1962; McKelvey & 

Webb, 1962) and undeformed granitic plutons of the 

Granite Harbour Intrusive Complex (Gunn & Warren, 

1962). These units are overlain by gently dipping 

Devonian^Triassic sedimentary rocks known as the 

Beacon Supergroup. These two main groups are separated 

by the Kukri Peneplain unconformity, which formed after 

the emplacement of the plutons and corresponds to a 

period of uplift and erosion. Both the basement and the 

cover sequence are intruded by Jurassic ( 180 Ma) Ferrar 

dolerites, which are exposed over an area of 10 5 km 2 

and have an estimated volume of 10 5 km 3 (Kyle, 1980). 

The dolerites form part of the Ferrar Large Igneous 

Province (FLIP), which extends more than 4000 km from 

the Weddell Sea to southeastern Australasia. The FLIP 

occurs as an elongate belt along the Transantarctic 

Mountains and has been interpreted as forming in a 

failed rift tectonic setting (Elliot & Fleming, 2004) associated 

with the break-up of Gondwanaland (Brewer et al., 

1992). The FDS represent by far the largest preserved 

volume of rock in the FLIP, which also includes the Dufek 

layered mafic intrusion, pyroclastic rocks and the 

Kirkpatrick basalt flows (Elliot & Fleming, 2008). Based 

on 40 Ar/ 39 Ar geochronology, the duration of the magmatic 

activity was of the order of 1 Myr (Heimann et al., 1994; 

Fleming et al., 1997). Two sills and the Dufek intrusion 

have been dated at 183·6 1·8 Ma, based on U^Pb 

1928 

geochronology of zircon and baddeleyite, the same 

age as the Karoo igneous province in southern Africa, 

demonstrating the vast size of the igneous province 

(Encarnacion et al., 1996; Minor & Mukasa, 1997). 

Emplacement paths for the Ferrar magmas are unknown: 

dyke swarms are not seen and dykes cutting basement 

rocks are uncommon. Elliot & Fleming (2004) advocated 

long distance (thousands of kilometres) crustal transport 

of magmas from a source suggested to be in the Weddell 

triple junction region through deep basement rocks, followed 

by localized vertical transport to shallower levels 

and then lateral transport in flows and sills over scales of 

a few hundred kilometres. The Dufek Massif is located at 

the edge of the Weddell Sea about 2000 km distant from 

the Dry Valleys. Though clearly part of the FLIP, Elliot & 

Fleming (2004) argued that, based on its stratigraphic 

placement above the basement rocks, the Dufek Massif is 

unlikely to be a feeder region for the FDS. 

Fieldwork by Gunn (1962), Gunn & Warren (1962), 

Hamilton (1965) and Marsh (2004) in the Taylor and 

Wright valleys identified a stack of four massive sills over 

a 4 km vertical extent (Fig. 1b and c) that are named, 

from top to bottom, the Mt. Fleming Sill (100 m thick), 

the Asgard Sill (250 m), the Peneplain Sill (PPS) (350 m), 

and the Basement Sill (BS) (350 m). Marsh (2004) proposed 

that each sill occurs as a series of lobes fed from a 

central location near the Dais, which is the lowest topographic 

exposure of the BS (Fig. 1b), and that at the time 

of emplacement there was a stack of vertically interconnected 

sills and magma chambers within the crust forming 

a magmatic mush column. The upper three sills consist of 

homogeneous fine- to medium-grained dolerite, whereas 

the lowermost BS has a central 50^250 m thick cumulate 

layer or ‘tongue’ consisting of large (1^20 mm) orthopyroxene 

and smaller (0·5^1mm) plagioclase crystals, enclosed 

by fine- to medium-grained dolerite like that of the other 

sills. Silicic segregations are common in the upper 100 m 

of the BS and PPS. This study focuses on the petrography, 

chemical composition and formation of the segregations 

in the BS. Studies by Be¤ dard et al. (2007) and Boudreau & 

Simon (2007) suggest that Fe-rich pegmatoids (here 

referred to as segregations) were formed from an evolved 

and degassed silicate liquid that was expelled from a compacting 

crystal mush below and that vapour-saturated 

zones, along with late interstitial liquid from the lower 

and central parts of the sill, may have ascended into localized 

regions in the upper zone to form these pegmatoids. 

FIELD AND ANALYTICAL 

METHODS 

Sampling locations 

Samples of the segregations were collected from the 

Taylor Valley region in 1993 by Marsh and Wheelock 



ZAVALA et al. SILICIC SEGREGATIONS, FERRAR DOLERITE SILLS 

Fig. 1. (a) Map showing location of the Dry Valleys, Antarctica. Modified from Be¤ dard et al. (2007). (b) Geological map of the Dry 

Valleys Area. Modified from B. D. Marsh & M. J. Zieg (unpublished, with permission). White stars indicate the locations where detailed field 

observations were made. WBP, West Bull Pass; EBP, East Bull Pass; Dais, Dais Intrusion. Black stars indicate locations where analyzed samples 

were collected: SV, Simmons Valley; PV, Pearse Valley; SR, Solitary Rocks; McT, McMurdo Tip; MP, Mount Peleus. (c) Cross-section AB 

through Bull Pass. Elevation is relative to sea level. Simplified from McKelvey & Webb (1962). 

1929 




Table 1: Basement Sill (BS) segregations and adjacent dolerites 

Seg profile Seg group Location Type Sample Sub-lenses D (m) Comments 

A S5 PV lens 93-BS-78 1-A 12 lens with medium- to 

S 3 1-B 12 coarse-grained 

S 5 2-A 12 sub-lenses 

S 3 2-B 12 

A-1 S 4 SV lens 93-BS-80 60·3 coarse-grained 

S3 lens 93-BS-81 60·6 medium-grained 

S4 lens 93-BS-82 60·82 coarse-grained 

S1 dolerite 93-BS-84 60 fine-grained above 80 

B S5 SV lens 93-BS-88 52·7 homogeneous 

S 4 lens 93-BS-89 A-1 52·7 seg with sub-lenses 

S 4 A-2 52·7 

S 4 A-3 52·7 

S 4 lens 93-BS-90 A 52·4 above and adjacent to 89 

S1 dolerite B 52·4 

S1 dolerite 93-BS-91 52·7 above and adjacent to 89 

S4 lens 93-BS-92 A-1 52·68 lens cut into 

S4 A-2 52·71 alternating mafic and 

S 4 A-3 52·76 felsic sub-lenses 

S 4 B-1 52·83 

S 4 B-2 52·91 

S 5 1-A-1 52·68 

S3 1-A-2 52·74 

S2 1-B-1-A 52·86 

S1 1-B-1-B 52·96 

S4 1-B-2-A 53·09 

S 1 1-B-2-B 53·25 

C S 1 SR lens 93-BS-95 24·2 homogeneous 

S 1 dolerite 93-BS-96 A-1 24·1 between two segs 

S 1 A-2 24·1 

S1 B-1 24·1 

S1 B-2 24·1 

D S4 SR lens 93-BS-97 23 

S1 dolerite 93-BS-99 A 24 

B 24·45 dolerite 

E * SR lens 93-BS-107 A-1 20 seg lens beneath 

A-2 20·5 grey felsic rock 

B 20·9 

F-1 S5 SR lens 93-BS-141 30·3 upper part of the BS 

F S4 SV lens 94-BS-150 A 93·8 lens 

S4 lens 94-BS-151 A 93·8 adjacent to 150-A 

S1 B 93·75 

S 5 lens 94-BS-152 93·8 same as 150-A 

S 1 lens 94-BS-153 93·42 0·4 m above 150-A 

S 2 lens 94-BS-154 A 93·8 lateral end of 150 

S 1 B 93·8 

1930 

(continued) 


http://petrology.oxfordjournals.org/ 

by guest on April 30, 2013

Table 1: Continued 


Seg profile Seg group Location Type Sample Sub-lenses D (m) Comments 

G S1 SV lens 94-BS-155 60·3 top of seg lens 

S4 lens 94-BS-157 A-1 60·32 seg across all 

S3 A-2 60·36 lenses of 155 

S1 A-3 60·43 

S2 A-4 60·52 

S4 A-5 60·61 

S1 A-6 60·76 

S1 dolerite 94-BS-158 59·94 0·36 m above 155 

S1 dolerite 94-BS-159 59·98 0·32 above 155 

G-1 S1 SV lens 94-BS-160 A-1-1 60·38 spans seg 155, and 157 

S1 A-1-2 60·51 

S4 A-2-1 60·71 

S1 A-2-2 60·96 

S4 A-3-1 61·26 

S4 A-3-2 61·59 

S1 A-4-1 61·97 

S4 A-4-2 62·42 

S3 B-1-1 62·96 

S4 B-1-2 63·56 

S4 B-2-1 64·24 

S4 B-2-2 65·04 

I S1 SV lens 94-BS-210 A 38·96 shows directional growth 

S1 B 38·96 

S4 lens 94-BS-211 38·96 next to 210 

S5 lens 94-BS-212 38·96 next to 211 

J S4 SV lens 94-BS-219 A 23·7 adjacent to dolerite 

S1 B 23·7 

K S4 SV lens 94-BS-223 A 48·1 adjacent to dolerite 

S1 B 48·1 

S4 lens 94-BS-224 48·1 coarse-grained top part 

S4 lens 94-BS-225 47·9 med-grained lower part 

L S3 SV vein 94-BS-231 A 39 top of vertical 0·5 m vein 

S1 B 39·1 

S4 vein 94-BS-232 39·1 1 m vein next to 231 

S4 vein 94-BS-233 A 39·07 bottom of 232 

BS McT dolerite 96-A-26 0 chilled margin 

dolerite MP dolerite 96-A-33 120 dolerite midway in BS 

D, distance from the upper contact of BS. 

*Thin sections only. 

(Wheelock & Marsh, 1993) and in 2001 by Zavala and 

Marsh (Zavala & Marsh, 2001). Detailed sample descriptions 

are given inTable 1 and major element chemical analyses 

in Table 2. Sample locations are indicated by black 

stars in Fig. 1b. Twenty-nine of these samples were collected 

in the Simmons Valley (SV) and one was collected in the 

Pearse Valley (PV), both offshoots of the Taylor Valley. Six 

1931 

samples were collected at Solitary Rocks (SR), a promontory 

at the junction of the Taylor and Ferrar glaciers. 

During the summer field season of 2003^2004 further 

sampling and detailed field observations were made at different 

heights and locations along the Wright Valley where 

the upper and lower contacts of the BS are well exposed, 

particularly near West Bull Pass (WBP), East Bull Pass 



Table 2: Representative major element compositions of host dolerites of groups S1^S5 

Profile: SegA-1 Seg B Seg C Seg D Seg F Seg G Seg G-1 Seg G-1 Seg I Seg J Seg K Seg L av. min. max. 

Sample: 84 90-B 96-B-1 99-A 154-B 157-A-3 160-A1-1 160-A1-2 210-A 219-B 223-B 231-B 

Group S 1 

SiO 2 53·05 51·89 51·10 54·84 53·25 53·35 53·48 54·35 55·16 53·03 52·13 53·16 53·23 51·10 55·16 

TiO 2 0·53 0·38 0·36 0·69 0·49 0·62 0·62 0·75 0·71 0·46 0·47 0·48 

Al 2O 3 14·86 13·81 17·86 14·45 13·53 15·68 14·37 16·60 14·83 16·93 19·29 18·59 

FeOT 9·31 10·57 8·85 10·62 8·31 10·42 10·93 9·75 10·29 9·01 7·65 8·48 

Fe2O3 1·55 1·76 1·48 1·77 1·39 1·74 1·82 1·63 1·72 1·50 1·28 1·41 

FeO 7·91 8·98 7·52 9·03 7·06 8·86 9·29 8·29 8·75 7·66 6·50 7·21 

MnO 0·17 0·19 0·15 0·17 0·18 0·17 0·18 0·15 0·16 0·16 0·13 0·13 

MgO 8·16 9·33 6·63 5·61 9·10 5·31 6·00 4·43 4·93 5·84 4·83 4·79 

CaO 11·85 11·77 12·59 10·15 11·60 10·79 10·76 9·80 8·95 11·66 11·33 11·22 

Na 2O 1·45 1·33 1·62 1·81 1·43 2·05 1·72 2·22 2·14 2·07 2·16 2·07 

K 2O 0·57 0·42 0·44 1·01 0·57 0·81 0·88 0·99 1·00 0·72 0·72 0·73 

P2O5 0·06 0·04 0·05 0·10 0·07 0·09 0·08 0·12 0·12 0·07 0·07 0·07 

LOI 0·58 0·46 0·41 0·67 0·66 0·65 1·66 1·30 0·41 0·62 1·08 1·23 

Total 100·74 100·37 100·20 100·18 99·27 100·03 100·91 100·42 100·36 100·71 99·99 101·10 

Mg-no. 64·75 64·91 61·09 52·54 69·65 51·65 53·50 48·78 50·11 57·60 56·96 54·21 57·15 48·78 69·65 

Alkalis 2·02 1·75 2·06 2·82 2·00 2·86 2·60 3·21 3·14 2·79 2·88 2·80 2·58 1·75 3·21 

FI 1·49 1·36 1·65 1·89 1·47 2·11 1·79 2·30 2·22 2·12 2·21 2·12 1·89 1·36 2·30 

MI 0·53 0·53 0·57 0·65 0·48 0·66 0·65 0·69 0·68 0·61 0·61 0·64 0·61 0·48 0·69 

Profile: Seg B Seg F Seg G av. min. max. 

Sample: 92-1-B-1-A 154-A 157-A-4 

Group S 2 

SiO 2 54·68 52·84 54·72 54·08 52·84 54·72 

TiO 2 0·96 0·43 0·88 

Al 2O 3 12·43 7·96 10·46 

FeOT 12·69 12·06 13·50 

Fe2O3 2·12 2·01 2·25 

FeO 10·79 10·25 11·48 

MnO 0·20 0·23 0·23 

MgO 6·47 13·02 6·81 

CaO 8·85 11·71 10·08 

Na 2O 1·75 0·82 1·46 

K 2O 0·83 0·41 0·87 

P2O5 0·12 0·05 0·09 

LOI 0·73 0·60 1·45 

Total 99·80 100·27 100·64 


Mg-no. 51·66 69·35 51·39 57·47 51·39 69·35 

Alkalis 2·58 1·23 2·33 2·05 1·23 2·58 

FI 1·82 0·85 1·53 1·40 0·85 1·82 

MI 0·66 0·48 0·66 0·60 0·48 0·66 

1932 

(continued) 





Profile: Seg A Seg A SegA-1 Seg B Seg G Seg G-1 Seg L av. min. max. 

Sample: 78-1-B 78-2-B 81 92-1-A-2 157-A-2 160-B1-1 231-A 

Group S 3 

SiO 2 55·56 55·11 56·48 55·21 56·03 55·14 57·33 55·72 55·14 56·48 

TiO 2 0·84 0·84 0·94 0·89 1·00 0·87 0·79 

Al 2O 3 14·34 14·38 13·75 14·46 14·78 14·49 13·93 

FeOT 10·62 11·05 10·79 11·05 11·76 11·08 9·96 

Fe2O3 1·77 1·84 1·80 1·84 1·96 1·85 1·66 

FeO 9·03 9·39 9·17 9·39 10·00 9·42 8·47 

MnO 0·16 0·16 0·17 0·17 0·16 0·16 0·16 

MgO 4·65 4·61 4·19 4·38 3·18 4·09 4·16 

CaO 9·21 9·32 9·11 9·38 7·96 8·89 8·84 

Na 2O 1·95 1·94 2·26 1·95 2·30 2·14 2·06 

K 2O 1·21 1·13 1·18 0·97 1·26 0·95 1·10 

P2O5 0·13 0·12 0·12 0·13 0·16 0·12 0·14 

LOI 1·78 1·50 1·10 1·39 1·12 0·80 0·69 

Total 100·53 100·29 100·15 100·07 99·81 99·67 99·32 

Mg-no. 47·86 46·65 44·87 45·38 36·18 43·62 46·68 42·51 36·18 45·38 

Alkalis 3·16 3·07 3·44 2·92 3·56 3·09 3·16 3·25 2·92 3·56 

FI 2·05 2·03 2·35 2·03 2·41 2·22 2·15 2·25 2·03 2·41 

MI 0·70 0·71 0·72 0·72 0·79 0·73 0·71 0·74 0·72 0·79 

Profile: SegA-1 SegA-1 Seg B Seg D Seg F Seg G Seg G-1 Seg G-1 Seg I Seg J SegK SegL Seg L av. min. max. 

Sample: 80 82 90-A 97 151-A 157-A-5 160-B1-2 160-B2-2 211 219-A 225 232 233-A 

Group S 4 


SiO 2 58·52 58·18 62·92 57·92 60·37 56·78 60·77 54·43 57·46 58·47 57·20 55·31 61·76 58·19 54·43 62·92 

TiO 2 1·30 1·10 1·26 1·39 1·31 0·87 1·15 1·42 1·78 0·97 1·10 0·81 1·10 

Al 2O 3 12·56 13·55 12·19 12·30 12·68 12·33 11·91 11·13 13·03 14·14 14·02 10·17 12·67 

FeOT 12·01 11·01 10·31 13·20 10·82 11·47 11·35 15·14 13·66 10·99 10·86 12·39 10·07 

Fe2O3 2·00 1·84 1·72 2·20 1·80 1·91 1·89 2·52 2·28 1·83 1·81 2·07 1·68 

FeO 10·21 9·36 8·76 11·22 9·20 9·75 9·65 12·87 11·61 9·34 9·23 10·53 8·56 

MnO 0·17 0·17 0·15 0·19 0·15 0·19 0·16 0·23 0·18 0·16 0·15 0·23 0·15 

MgO 3·12 3·21 2·42 2·83 2·89 5·17 2·70 4·55 2·51 3·25 3·24 7·22 3·10 

CaO 7·18 7·68 5·76 6·90 6·30 8·90 6·19 8·92 7·31 7·66 7·44 10·09 6·02 

Na 2O 2·41 2·44 2·48 2·34 2·36 1·88 2·38 1·82 2·35 2·41 2·28 1·44 2·31 

K 2O 1·32 1·40 1·62 1·51 1·59 1·12 1·51 0·93 1·29 1·34 1·25 0·87 1·42 

P2O5 0·18 0·16 0·20 0·20 0·21 0·13 0·20 0·12 0·21 0·17 0·18 0·07 0·19 

LOI 1·32 1·13 1·12 1·16 1·13 1·55 1·41 1·39 0·71 1·02 0·98 1·44 1·97 

Total 100·29 100·21 100·60 100·15 99·99 100·58 99·92 99·81 100·72 100·76 98·88 99·24 100·92 

Mg-no. 35·25 37·93 32·98 31·00 35·89 48·58 33·27 38·65 27·81 38·27 38·47 54·98 39·22 37·76 27·81 54·98 

Alkalis 3·73 3·84 4·10 3·85 3·95 3·00 3·89 2·75 3·64 3·75 3·53 2·31 3·73 3·53 2·31 4·10 

FI 2·53 2·56 2·64 2·48 2·52 1·97 2·53 1·90 2·47 2·53 2·39 1·51 2·46 2·34 1·51 2·64 

MI 0·79 0·77 0·81 0·82 0·79 0·69 0·81 0·77 0·84 0·77 0·77 0·63 0·76 0·77 0·63 0·84 

1933 

(continued) 





Profile: Seg A Seg A Seg B Seg B Seg F-1 Seg F Seg I av. min. max. 

Sample: 78-1-A 78-2-A 88 92-1-A-1 141 152 212 

Group S 5 


SiO 2 66·95 64·52 64·07 59·00 59·28 59·09 64·78 62·53 59·00 66·95 

TiO 2 1·01 1·28 1·54 1·90 1·44 1·32 1·35 

Al 2O 3 11·60 12·09 11·51 12·55 13·08 13·49 11·55 

FeOT 6·88 8·40 10·98 12·60 10·61 10·55 10·66 

Fe2O3 1·15 1·40 1·83 2·10 1·77 1·76 1·78 

FeO 5·85 7·14 9·33 10·71 9·02 8·97 9·06 

MnO 0·11 0·13 0·14 0·18 0·16 0·16 0·13 

MgO 1·94 2·28 1·67 2·20 2·99 3·06 0·99 

CaO 3·50 4·40 4·28 5·76 6·07 6·38 3·61 

Na 2O 3·52 3·26 2·98 2·39 2·62 2·40 3·06 

K 2O 1·40 1·45 1·43 1·36 2·18 1·62 2·49 

P2O5 0·25 0·29 0·27 0·24 0·19 0·20 0·35 

LOI 1·82 2·24 1·66 1·12 1·14 1·45 0·97 

Total 99·09 100·47 100·71 99·51 99·94 99·90 100·11 

Mg-no. 37·15 36·26 24·17 26·79 37·13 37·81 16·29 30·80 16·29 37·81 

Alkalis 4·92 4·71 4·41 3·75 4·80 4·02 5·55 4·59 3·75 5·55 

FI 3·69 3·42 3·14 2·53 2·82 2·56 3·33 3·07 2·53 3·69 

MI 0·78 0·79 0·87 0·85 0·78 0·78 0·92 0·82 0·78 0·92 

FI ¼ (Na2O þ K2O)/(Na2O þ K2O þ CaO); MI ¼ [FeOT/(FeOT þ MgO)]. 

(EBP), and 12 km west of Bull Pass area at the Dais. 

These locations are indicated by white stars in Fig. 1b. 

Although there are as yet no chemical data for this set of 

rocks, the detailed field observations made during this 

field season have provided valuable new information 

about the morphology of the segregations in the sill. 

Sampling methods and nomenclature 

Sample collection was started at the upper contact of the 

BS and continued down section about 300 m (Table 1). 

Sample size varied between tens of centimetres to 2 cm, depending 

on the thickness and texture of the segregation. 

Some samples consist of several layers of coarse-grained 

felsic rock alternating with medium- to fine-grained mafic 

rock (Fig. 2a), whereas others comprise homogeneous 

coarse-grained felsic rock (Fig. 2b). 

We use the term ‘segregation profile’ for a vertical sequence 

of samples from a single locality. For example, segregation 

profile A-1 consists of segregation samples 

93-BS-80, 93-BS-81 and 93-BS-82 and the overlying dolerite 

sample 93-BS-84. Most segregations have textural variations; 

the letters A and B indicate layers with different 

textures within the profile. Trailing numbers after the 

letter refer to any additional textural variations within a 

layer. For example, segregation sample 94-BS-160 was cut 

1934 

into six layers, 94-BS-160-A-1 to A-4, and 94-BS-160-B-1 

and B-2 (Fig. 2a). The layer described by 94-BS-160-A-1 

was further subdivided into two sub-layers 94-BS-160-A- 

1-1 and 94-BS-160-A-1-2, based on observed textural 

variations. 

Analytical methods 

Samples are typically fresh and non-vesicular, although 

some have micro-fractures with minor alteration. Samples 

were cut taking into account the textural and contact relationships 

observed in the field and in hand specimen. 

Thin sections and chemical analyses were made of each 

layer in each segregation and of the adjacent dolerite. For 

chemical analyses, samples were crushed to a mesh size of 

80 mm, using a ball mill and a nickel shatter box. The powders 

were prepared for major element analyses by X-ray 

fluorescence (XRF) at Franklin and Marshall College following 

the procedure of Boyd & Mertzman (1987). The 

procedure involved melting the powders using a cadmium^nickel 

charge, then quenching the liquid into a 

solid disc. Each analysis was run three times and the average 

value was used. Relative analytical error is between 

0·11 and 0·26% for all major elements, and accuracy 

is 0·2%, estimated by analyzing accepted standards 

(Abbey, 1983; Govindaraju, 1994). FeO is reported as FeO T, 





Fig. 2. (a) Segregation sample 94-BS-160 (profile G-1) has internal 

stratification consisting of multiple felsic coarse-grained horizons 

(160: A-2, A-3, and B-1) and less felsic medium-grained (160: A-1, A-4, 

and B-2). (b) Coarse-grained segregation with no internal stratification. 

Segregations have sharp contacts with the host dolerite and are 

composed of coarse pyroxene and plagioclase crystals with minor 

opaque minerals and variable amounts of granophyric intergrowths 

of quartz and K-feldspar. 

1935 

Table 3: Segregations and dolerite modes obtained by the 

SEM^MLA method (see text for details) 

Sample: 93-BS-96-B-1 94-BS-154 93-BS-78-1-B 93-BS-88 93-BS-78-1-A 

Group: S1 S2 S3 S5 S5 

augite 21·1 24·5 16·1 1·8 2 

pigeonite 12·1 10·7 1·6 0·1 0·2 

orthopyroxene 4·0 6·4 0·2 0 0·1 

plagioclase 49·3 46·3 40·3 14·2 14·3 

K-feldspar 2·0 2·4 5 8·1 7·1 

quartz 5·0 5·3 17·1 59·7 63·1 

Ca-amphibole 6·0 4 18·1 12·5 11·2 

Fe–Ti oxides 0·4 0·3 1·4 3·5 2 

and an FeO/FeOT ratio of 0·85 is assumed (Ragland, 1989, 

p. 36). 

Mineral compositions of the major phases in segregation 

profile A-1 (samples 93-BS-80, 81, 82 and adjacent dolerite 

93-BS-84) were obtained using the JEOL 8600 electron 

microprobe at Johns Hopkins University. The microprobe 

was operated at 15 kV with a beam current of 20 nA. The 

beam size was 5 mm for augite, diopside and pigeonite, 

and 10 mm for plagioclase and quartz and K-feldspar intergrowths. 

Counting times were 60 s at the peak and 30 s 

for background. Data were reduced using the CITZAF 

program of Armstrong (1989). Analytical relative errors 

for major elements are considered to vary between 0·5 

and 1%. Pyroxene and plagioclase crystals were probed 

in transverse sections to check for compositional zoning. 

High-resolution imaging and automated quantitative estimates 

of the modal mineralogy of selected thin sections 

was carried out at Memorial University using an FEI 

Quanta 400 scanning electron microprobe (SEM)^mineral 

liberation analyzer (MLA). The data are reported in 

Table 3. 

Sample grouping 

If segregations are the result of fractional crystallization of 

the host dolerite, then mass balance requires that the composition 

of the segregations is equal to the composition of 

an initial, fractionated liquid plus or minus the crystallized 

phases entrained into or lost from this liquid. A set of 

MELTS modeled liquids and crystals fractionating between 

12008C and 9008C were linearly regressed to 

the final bulk composition of the segregations, using the 

singular value decomposition technique (Zavala, 2005). 

This technique has previously been used by Fisher (1989) 

to analyse metamorphic assemblages. The best-fit model 

to a given segregation provided the temperature at which 

the residual liquid separated from the modeled liquids 

and the proportions of minerals subsequently lost through 




Table 4: Definitions of segregation groups 

Group MML Plagioclase Augite Pigeonite 

S1 þ þ þ þ 

S 2 þ – þ þ 

S 3 þ þ þ – 

S 4 þ – þ – 

S 5 þ – – – 

MML (melts modeled liquids); þ, added; , subtracted. 

crystallization and/or gained through entrainment. When 

the segregations are grouped according to whether they 

had lost or gained the major mineral phases plagioclase, 

augite and pigeonite they fall into five categories, given in 

Table 4. Group S 1 corresponds to the host dolerite: that is, 

compositions are best fitted by a MELTS model liquid 

(MML) plus all three crystallizing phases. The segregations 

S 2^S 5 correspond to more fractionated rocks, in 

which one or more of the crystallizing phases has segregated 

from the MML. GroupS 5 is the most fractionated, 

corresponding to segregations that are best modelled as 

pure residual liquids. More detailed description of the 

technique has been given by Zavala (2005). The designated 

group is given for each sample in Table 1, and samples are 

arranged by group inTable 2. 

LOCATION AND MORPHOLOGY 

OF SEGREGATIONS IN THE 

BASEMENT SILL 

Sets of sub-horizontal, anastomosing, 0·1^2 m thick lenses 

are common in the upper 100 m of the BS and make up 

about 15% by volume of the sill. The most common types 

of segregation in both Taylor Valley and Wright Valley are 

sub-horizontal lenses (Fig. 3c and d) that extend for tens 

of metres along strike and are connected to other lenses 

by smaller sub-vertical veins. Small ( 5^10 cm) felsic 

pods (Fig. 3a) and thin ( 2^10 cm) felsic stringers 

(Fig. 3b), which occur above and adjacent to these lenses, 

are irregular features that are usually located a few 

metres ( 5^15 m) below the upper contact. Overall segregation 

textures, morphologies, and field relationships are 

similar throughout the BS; however, at the Dais, which is 

the lowest topographic exposure of the BS and consists of 

an orthopyroxene- and plagioclase-rich cumulate body, 

there is a larger variation in segregation morphology that 

varies systematically with distance from the upper contact 

(Fig. 4). 

1936 

Dais segregations 

In the interval 5^7 m below the upper contact of the BS at 

the Dais, thin (1^5 cm) sub-horizontal felsic streaks 

(4^10 cm long) are ubiquitous (Fig. 4a). These give way 

10 and 15 m below the upper contact to sparse 1^5 cm 

diameter felsic, pod-shaped, segregations (Fig. 4b) oriented 

parallel to the upper contact of the sill. Some pods are connected 

by 1^5 cm thick and 10^15 cm long undulating 

stringers. Both pods and stringers have sharp contacts 

with the host-rock. Between 15 and 20 m below the upper 

contact, segregations form 3^10 cm diameter sub-vertical 

pipes and 15^20 cm wide by 20^50 cm long sub-vertical 

veins that swell and pinch out laterally forming a 

saw-tooth pattern (Fig. 4c). Finally, about 30 m below the 

upper contact, there is a thick ( 300 cm), anastomosing, 

segregation lens, which extends laterally for more than 

20 m (Fig. 4d) and has undulating sharp upper and lower 

contacts with the host-rock. The upper contact of this massive 

layer with the host-rock is convex downward (white 

dashed line in Fig. 4d), and texturally the layer shows internal 

stratification, with a coarse-grained horizon in the 

middle that becomes progressively finer-grained towards 

the margins. 

PETROGRAPHY OF THE SILICIC 

SEGREGATIONS 

As described above, the dolerites and segregations have 

been subdivided into five groups (S 1^S 5) based on 

mass-balance relationships. Petrographic study of samples 

from each group indicates that there are distinct mineralogical 

and textural features associated with each group. 

Table 3 gives the mineral modes of representative samples 

obtained by the SEM^MLA method and Table 5 provides 

a summary of the petrography of each group. 

S 1 rocks (dolerites) are characterized by fine- to 

medium-grained ophitic textures (Fig. 5a) and contain 

prismatic (0·5^4 mm) euhedral to subhedral moderately 

embayed augite phenocrysts, smaller (0·1^3 mm) equant, 

subhedralânhedral, fractured orthopyroxene phenocrysts, 

all enclosed by a network of (0·05^1mm) plagioclase 

laths. There are minor amounts of interstitial quartz and 

alkali feldspar in granophyric intergrowths and a few 

(0·01^0·1mm) grains of Fe^Ti oxide (ilmenite and 

magnetite). A representative modal analysis of the dolerite 

(sample 93-BS- 96-B-1) is given in Table 3. Intergrowths of 

quartz and alkali feldspar are commonly hard to distinguish 

optically but they are readily identified with the 

SEM. Ca-amphibole occurs as an alteration product 

around pyroxene crystals rather than as a primary phase. 

The S2 segregations have fine- to medium-grained ophitic 

to sub-ophitic textures with sub-equal amounts of pyroxene 

(low- and high-Ca pyroxene) and plagioclase 

(Fig. 5b). Low-Ca pyroxene crystals are 1^5 mm long, 





Fig. 3. Field photographs showing that segregations can occur as small ( 5^10 cm) felsic pods (a) and thin irregular ( 5 cm) stringers subparallel 

to the upper contact of the BS (b). Between 30 and 100 m below the upper contact segregations form (0·5^2 m) thick, anastomosing, climbing 

felsic lenses (c, d) that are often interconnected to other smaller lenses above by thin sub-vertical veins. Stratigraphic section shows the 

locations of the photographs. 

equant and partly embayed, with fractures and reaction 

rims consisting of dark alteration minerals, and contain 

fine exsolution lamellae of pigeonite. Other low-Ca pyroxene 

crystals are subhedral, moderately embayed and contain 

abundant fine exsolution lamellae of augite parallel 

to (001) and inclined to (100). Augite crystals are 2^5 mm 

long, euhedral to subhedral, prismatic and less embayed 

and compositionally twinned. They commonly have exsolution 

lamellae of pigeonite parallel to (001), as confirmed 

by the different spectral patterns in the SEM^ 

MLA, and some have inverted pigeonite cores with complex 

twinning patterns. Plagioclase crystals are 0·4 

1mm long, euhedral to subhedral, commonly bladed 

and normally zoned. A representative modal analysis of 

this rock (sample 94-BS-154) is given in Table 3. This rock, 

unlike the others, has sub-rounded orthopyroxene crystals 

with reaction rims and fractures filled with alteration 

1937 

minerals; these resemble the orthopyroxene cumulate crystals 

found in the opx tongue. 

The S 3 segregations have sub-ophitic and poikilophitic 

textures consisting of approximately sub-equal proportions 

of plagioclase and augite (plus minor pigeonite), and between 

20 and 30% vermicular intergrowths of quartz and 

alkali feldspar (Fig. 5c). The rims of the plagioclase crystals 

are commonly optically continuous with the granophyric 

intergrowths. In the poikilophitic textures 

plagioclase and augite crystals have similar size: the poikilophitic 

domains consist of medium- to coarse-grained 

prismatic augite interlocking with tabular plagioclase 

crystals; granophyric intergrowths fill the interstices. 

The sub-ophitic domains consist of large ( 0·5mm) 

augite crystals embedded with smaller (0·1^1mm) plagioclase 

laths. In addition to their occurrence in granophyric 

textures, quartz and alkali feldspar form anhedral, 




Fig. 4. At the Dais in theWright Valley, segregations vary in morphology with increasing distance from the upper contact. (a) Close (5^7 m) to 

the upper contact segregations occur as (1^5 cm) thin felsic streaks; (b) about 10^15 m below the upper contact segregations are ( 5 cm) 

pod-shaped. (c) Between 15 and 20 m below the upper contact segregations form pipes and (15^20 cm) sub-vertical anastomosing lenses. 

(d) Below 30 m, an unusually ( 3·5 m) thick felsic lens with variable grain size forms a sharp upper wavy contact with the host-rock. All segregation 

types have sharp contacts with the host-rock at this location. Stratigraphic section shows the locations of the photographs. 

sub-rounded grains from 0·1to0·5 mm in size that appear 

to be recrystallized. Pyroxene crystals are commonly 

rimmed with Ca-amphibole and biotite and the size and 

amount of Fe^Ti oxide grains is higher than in S 1 and S 2. 

Figure 5c illustrates the poikilophitic texture within a 

coarser felsic segregation (94-BS-231-A) in the upper left 

corner, and the sub-ophitic texture associated with the 

dolerite (94-BS-231-B) in the lower right corner. Between 

these two textural zones there is a contact zone consisting 

predominantly of fine-grained plagioclase laths. 

Segregations in the S 3 group have higher modal amounts 

of quartz and alkali feldspar in granophyric intergrowths 

than the previous two groups (e.g. see Table 4). This group 

1938 

also has more alteration rims of Ca-amphibole and biotite 

around subhedral pyroxene crystals. 

The S 4 group is composed mostly of poikilophitic domains 

with some sub-ophitic domains; the amount of 

granophyric quartz and K-feldspar intergrowths and alteration 

is considerably greater than in all previous groups 

(Fig. 5d). The poikilophitic domains consist of interlocking, 

1^8 mm, subhedral to anhedral, altered augite and 

0·5^7 mm euhedral to subhedral, tabular, fractured plagioclase 

surrounded by medium- to coarse-grained granophyric 

intergrowths. The augite and minor inverted 

pigeonite crystals, which are subhedral to anhedral and 

partially to completely replaced by Ca-amphibole, biotite, 




Table 5: Summary of petrographic observations of the segregations and adjacent dolerites 

Group Rock type Texture Mineral size Characteristics 

S1 dolerite ophitic with minor sub-ophitic 

domains 

S2 segregation sub-ophitic with minor ophitic 

domains 

S3 segregation sub-ophitic with poikilophitic 

domains 

S4 segregation Poikilophitic with granophyric 

domains 

S5 segregation Granophyric with few crystalline 

domains 

epidote, actinolite and chlorite, commonly form bleb-like 

intergrowths with fractured and altered plagioclase crystals. 

The Fe^Ti oxides grains (mostly ilmenite and magnetite) 

are larger (0·5 mm) and are always associated with 

the altered pyroxene and with the granophyric intergrowths. 

The plagioclase crystals contain some sericite, 

minor titanite and epidote alteration, and are commonly 

fractured. Whereas some crystals are euhedral and not resorbed, 

others have inclusions of granophyric material 

through their cores and margins that are optically continuous 

with the granophyric intergrowths. Some crystals 

have kinked twins and are bent or broken (Fig. 6). The 

coarse-grained granophyric intergrowths are more pervasive, 

with fewer crystalline domains forming interstitially, 

and the K-feldspar in them is commonly altered to sericite, 

imparting a brown color, whereas the quartz remains unaltered. 

There is no modal analysis available for this 

sample. 

The S5 group is distinguished by having more than 50% 

of the mode composed of granophyre (Fig. 5e and f), consisting 

of both coarse-grained vermicular intergrowths of 

K-feldspar and quartz and large (0·5^10 mm) sub-rounded 

crystals of these phases. Typically a few, long (5^20 mm) 

fine to medium euhedral–subhedral (0·5–4 mm) augite 

subhedral, equant (0·1–3 mm) pigeonite 

euhedral bladed (0·05–1 mm) plagioclase 

minor interstitial granophyric textures and opaque minerals 

fine to medium euhedral–equant (1–5 mm) orthopyroxene and 

pigeonite 

euhedral, tabular (2–5 mm) augite 

euhedral bladed and tabular (0·4–1 mm) plagioclase 

minor opaque minerals 

fine to medium euhedral–subhedral, prismatic, (0·5–5 mm) augite 

euhedral–subhedral equant (2–5 mm) pigeonite 

euhedral bladed and tabular (0·1–5 mm) plagioclase 

interstitial alkali feldspar and quartz intergrowths 

minor opaque minerals 

medium to coarse altered, subhedral, prismatic (1–8 mm) augite 

euhedral– 

subhedral, tabular (0·5–7 mm) plagioclase 

medium- to coarse-grained granophyric textures 

minor opaque minerals (0·5–1 mm) 

medium to coarse euhedral–subhedral, resorbed (5–20 mm) plagioclase 

anhedral, altered, intergrowths of augite and opaque 

minerals 

quartz and K-feldspars ( 5 mm) 

1939 

euhedral, moderately fractured plagioclase crystals are 

present, some of which may contain inclusions and have 

rims partially replaced by granophyric intergrowths. The 

pyroxene crystals are completely replaced by intergrowths 

of Ca-amphibole, biotite, minor phlogopite and chlorite, 

and form intergrowths with the plagioclase crystals and 

the granophyric textures. Fe^Ti oxides (ilmenite and magnetite) 

are also coarse-grained (up to 5 mm) and are 

associated with the replaced pyroxenes. The measured 

modal compositions of two representative samples of this 

group (93-BS-88 and 93-BS-78-1-A) are given in Table 3. 

The S 5 group is distinguished from the S 4 group by 

having higher amounts of quartz and alkali feldspar intergrowths 

and fewer crystalline domains, and from groups 

S 1 to S 3 by having larger crystal sizes, significantly less pyroxene, 

more alteration minerals and more granophyric 

intergrowths. 

Summary 

Overall there is a gradual increase in the grain size and 

amount of granophyre and alteration minerals from the S 1 

to the S 5 group. This is particularly evident in groups S 3 

to S 5, which progressively contain more regions with large 




Fig. 5. Representative photomicrographs of petrographic groups S 1^S 5.(a)S 1 sample 93-BS-96-B-1; (b) S 2 sample 94-BS-154-A; (c) S 3 sample 

94-BS-231-A; (d) S 4 sample 93-BS-97; (e) S 5 sample 93-BS-88; (f) S 5 sample 94-BS-212. Groups S 1 and S 2 have fine- to medium-grained ophitic 

and subophitic textures with sub-equal amounts of plagioclase and pyroxenes and minor quartz and K-feldspar intergrowths (a, b). Groups S3 

and S4 have medium- to coarse-grained poikilophitic textures with minor subophitic domains and increasing amounts of quartz and alkali feldspar 

granophyric intergrowths and alteration minerals (c, d). Group S5 consists mostly of granophyric intergrowths with few crystalline domains 

(e, f) and greater amount of alteration and opaque minerals. Common phases are plagioclase (plg), augite (aug), pigeonite (pig), 

orthopyroxene (opx) and granophyric quartz^K-feldspar intergrowths (gr). Minor phases are epidote (ep), biotite (bt), and opaque minerals 

(opq). 

1940 




Fig. 6. (a, b) Representative photomicrographs of pyroxene crystals with curved twin planes: (a) sample 93-BS-107-A-1; (b) sample 93-BS-97 

(S 4). (c, d) Broken and partially resorbed plagioclase crystals: (c) sample 94-BS-157-A-2 (group S 3); (d) sample 93-BS-92-1-B-2-B (S 1). (e^h) 

Other plagioclase crystals are bent, have kinked twins, and have inclusion of granophyric intergrowths: (e) sample 93-BS-92-A-2 (S 4); 

(f) sample 94-BS-157-A-2 (S3); (g) sample 94-BS-160-B-1-1 (S3); (h) sample 94-BS-211 (S4). 

1941 





interlocking fractured plagioclase and altered pyroxene 

crystals enclosing or enclosed by quartz and alkali feldspar 

in granophyric intergrowths. The coarse-grained plagioclase 

and augite crystals in groups S3 and S4, the large 

( 2 cm) euhedral plagioclase crystals and increasing 

amount of granophyric intergrowths in group S5, and the 

increasing proportion of alteration minerals in all three 

groups S3 to S5, suggest that minor amounts of volatiles 

were present during the formation of these segregations. 

In groups S 1^S 4 some augite crystals have curved twins 

(Fig. 6a and b) and some plagioclase crystals are bent 

(Fig. 6f) or broken (Fig. 6c, d and h), and have kinked 

twins (Fig. 6c^h). The bent and deformed plagioclase crystals 

contain recrystallized pockets of granophyric material 

(see Fig. 6e and h) and most of these deformed crystals 

are enclosed by granophyric intergrowths. 

The compositional progression from dolerite to the most 

felsic segregations, together with the observed range of textural 

relationships between early (plagioclase and pyroxene) 

and late (granophyre) phases and the increasing 

amounts of Ca-amphibole and Fe^Ti oxides (magnetite 

and ilmenite) as crystallizing phases suggest that the segregations 

formed at different stages of crystallization of the 

dolerite magma. 

Table 6: Anorthite content of plagioclase crystals from 

Profile A-1 

Sample Group n Core Rim Average 

93-BS-84 S1 4 50–84 40–77 64 

93-BS-80 S4 5 54–63 38–60 53 

93-BS-81 S3 3 60–70 50–59 61 

93-BS-82 S4 5 33–63 37–61 55 

Detailed analyses are given in Supplementary Data Tables 

6–1 to 6–4. 

Table 7: Representative compositions of pyroxene crystals from Profile A-1 

MINERAL CHEMISTRY 

Chemical compositions of plagioclase and pyroxene were 

determined within Profile A-1, which consists of three consecutive 

segregations (93-BS-80, 93-BS-81 and 93-BS-82) 

situated below dolerite (93-BS-84). Plagioclase compositions 

are shown schematically in Fig. 7. Segregations 

93-BS-80 and 82 are coarse-grained and have a patchy texture, 

whereas 93-BS-81, which is situated between them, is 

medium- to coarse-grained and less patchy. The compositions 

of the minerals change systematically with bulk-rock 

composition and texture. The two most felsic and differentiated 

segregations, 93-BS-80 and 82, are similar in composition 

and texture and belong to the S4 group; the less 

felsic and differentiated segregation, 93-BS-81, belongs to 

the S 3 group, and the dolerite, 93-BS-84, which has the 

most primitive composition and an ophitic texture, belongs 

to the S 1 group. On average five plagioclase and pyroxene 

crystals per thin section in segregation profile A-1 were 

analyzed to determine compositional variations within a 

given sample and across samples (Table 6 and Table 7, 

respectively). 

Plagioclase 

Plagioclase crystals in dolerite sample 93-BS-84 are small 

(0·01^2 mm), euhedral laths commonly embedded in 

larger pyroxene oikocrysts (Fig. 7a). Most crystals are normally 

zoned, although there are a few, 0·5^1mm, subhedral, 

tabular grains that show complex zoning. The most 

felsic segregations, 93-BS-80 and 82, have 1^20 mm long, 

euhedral to subhedral tabular and blocky crystals that are 

commonly fractured and altered, and have resorbed edges 

and cores partially replaced with granophyric intergrowths 

(Fig. 7b and d). Segregation 93-BS-81, which is 

less felsic, has 1^8 mm long, euhedral to subhedral, tabular 

and bladed crystals that are less fractured and altered 

than those of the other two segregations (Fig. 7c). 

Table 6 summarizes anorthite contents of plagioclase 

crystals (cores, rims and averages) from the four samples 

of Profile A-1. Details of the electron microprobe analyses 

Sample Group n Cores Rims Average 

93-BS-84 S 1 3 En(46–68)Fs(13–21)Wo(11–36) En(34–44)Fs(23–28)Wo(30–42) En 46Fs 21Wo 34 



93-BS-82 S4 3 En(28–41)Fs(16–30)Wo(41–43) En(26–38)Fs(19–31)Wo(41–43) En32Fs25Wo42 

Detailed analyses are given in Supplementary Data Tables 7-1 to 7-4. n, number of crystals analyzed in each sample. 

Average is of (core and rim) analyses. One crystal in group S1 has a core composition of Wo11 and two crystals have a 

core composition of Wo35 and Wo36. 

1942 





Fig. 7. Photomicrographs of segregation profile A-1, which consists of (a) dolerite (93-BS-84) overlying three consecutive segregations: 

(b) 93-BS-80; (c) 93-BS-81; (d) 93-BS-82. The white circles in the photomicrographs indicate micro-probed plagioclase crystals. The compositions 

of their cores, rims and intermediate zones are shown in the feldspar ternary diagram below (e). Panels are arranged according to the 

order in which samples occur in the field, from top to bottom. 

1943 




Fig. 8. Compositional profiles of plagioclase crystals in segregation profile A-1. Left panels shows the variation of anorthite content (An wt %) 

in crystals taken from one edge of the crystal to the other, except those labeled ‘C’ where the profile terminates in the core of the crystal. 

Right panels show representative electron backscatter images indicating where the profiles were taken in the crystals. (a) Dolerite 93-BS-84 

shows irregular concave-down profiles; (b) segregation 93-BS-80 has smooth concave profiles; (c) segregation 93-BS-81 (least felsic) has little 

compositional variation from rim to core resulting in flat profiles; (d) segregation 93-BS-82 (most altered) has irregular slightly concave profiles. 

The main phases are plagioclase (plg), pyroxene (pyx), granophyre (gr), and opaque minerals (opq). Panels are arranged according to 

stratigraphic sequence in the field from top to bottom (see Fig. 7). 

are given in Electronic Appendix Tables 6.1^6.4 (all of the 

supplementary material is available for downloading at 

http://www.petrology.oxfordjournals.org). In accord with 

trends expected if the segregations are related to the dolerites 

by crystal^liquid differentiation, the dolerite 93-BS-84 

has the most Ca-rich cores but Na-rich rims similar to the 

most felsic segregations (93-BS-80 and 82), whereas the 

1944 

most mafic of the three segregation (93-BS-81) has more 

Ca-rich cores and less Na-rich rims than the other two 

felsic segregations (Fig. 7e). Plagioclase varies in size and 

composition with changes in the bulk composition of the 

host-rocks. 

Compositional profiles through plagioclase grains are 

shown in Fig. 8. Plagioclase in the dolerite (93-BS-84, 




group S 1) has cores with the highest anorthite concentrations 

and albite-rich rim compositions. The compositional 

profiles in this rock are irregular, although exhibiting 

an overall concave-downward shape from rim to core, 

and they display significant compositional variations 

both between crystals and from one edge of a crystal to the 

other (Fig. 8a). One of the smallest (0·15 mm) crystals analysed 

(plg-01) in the dolerite has a relatively uniform and 

low-anorthite core region and the smallest rim to core compositional 

variation (An40^62) when compared with the 

other crystals in this dolerite sample (Fig. 8a). 

Plagioclase in the S 4 segregations (93-BS-80 and 82) 

displays generally smooth, concave-down compositional 

profiles; however, there are compositional variations 

between crystals in these samples (Fig. 8b and d). In 

sample 93-BS-82 (Fig. 8d), plagioclase crystals 82-02 

and 82-04 exhibit irregular, jagged compositional profiles 

with a localized decrease in An (wt %) in the intermediate 

region between cores and rim, suggesting 

that these crystals may have partially resorbed intermediate 

regions. The downward dip in the compositional profile 

plg-82-01 corresponds to a fracture going across the crystal. 

Plagioclase in the S 3 segregation (93-BS-81) has the 

smallest compositional variation both between and within 

analyzed crystals, resulting in almost horizontal profiles 

(Fig. 8c). The observation that the compositional profiles 

for the small (51mm) and the larger (41·5 mm) crystals 

are almost identical suggests that the larger crystals may 

have completely recrystallized when they came in contact 

with the residual liquid, or the larger crystals may have 

grown in the same environment as the smaller ones. 

Pyroxenes 

Pyroxene crystals in segregation profile A-1 vary in size, 

shape and composition as well as modal abundance. In 

the dolerite sample 93-BS-84, the pyroxene crystals consist 

of both augite and pigeonite, whereas in the segregations 

93-BS-80, 81 and 82 the primary pyroxenes are augite in 

composition with pigeonite exsolution lamellae parallel to 

(001) and less commonly (100). Augite crystals are commonly 

twinned. In general, the pyroxene crystals show 

a regular zoning from more Mg-rich cores to more 

Fe-rich rims. Detailed mineral compositions of cores 

and rims are given in Electronic Appendix Tables 7.1^7.4. 

Representative compositions are given in Table 7; these 

compositions are plotted in Fig. 9 as not all pyroxene analyses 

yielded good totals. 

All analysed pyroxene crystals in the dolerite, 93-BS-84, 

have distinct core and rim compositions, and they are 

characterized by more enstatite-rich and less Ca-rich compositions 

than the segregations (Table 7, Fig. 9a^d). One 

out of three augite crystals analyzed (84-03) has a pigeonite 

core partially replaced by orthopyroxene lamellae as 

indicated by the anomalously En-rich data points in 

Fig. 9a. 

1945 

Pyroxenes in the more felsic, S 4 segregation samples 

show considerable compositional overlap between the 

cores and rims. Crystals in sample 93-BS-82 contain the 

most Fe-rich cores whereas those in sample 93-BS-80 contain 

the most Fe-rich and Ca-rich rims. Pyroxenes in the 

less felsic S 3 segregation sample (93-BS-81) overall have 

compositions that are intermediate between the S 1 and S 4 

samples, but much closer to the S 4 samples. 

Representative enstatite (En wt %) compositional profiles 

illustrating some of these features are shown in 

Fig. 9e. The profile corresponding to the dolerite (84-01) 

has the highest enstatite concentrations and shows the 

most rim to core variation compared with profiles from 

the segregations samples. All four profiles display asymmetric 

concave shapes with higher enstatite concentrations 

within the core of the crystal. Profile 81-03, from a crystal 

in the S 3 segregation, displays a slightly higher peak enstatite 

concentration than the profiles from the S4 segregations. 

Profile 82-04A (diamonds in Fig. 9e) shows little 

variation in composition from the center towards the rim 

on the right side: this could be because exsolution lamellae 

within that region of the crystal prevented accurate sampling 

of the average composition. 

Granophyre 

There is a systematic increase in the proportions of quartz 

and K-feldspar in the granophyric intergrowths with 

increasing silica content in the segregations that is particularly 

noticeable in the most differentiated samples represented 

by the group S 5. In the segregation profile A-1, the 

S 4 segregations 93-BS-80 and 82 have the highest amount 

of granophyric intergrowths and are also the most felsic 

and differentiated samples. 

In general, as the granophyric texture becomes more 

pervasive, plagioclase crystals with partially replaced 

cores and margins that are optically continuous with the 

granophyric intergrowths become more common. In the 

S 5 group plagioclase crystals are generally replaced by 

granophyric intergrowths and only a few large ( 1cm) euhedral 

isolated plagioclase crystals appear to be in equilibrium 

with the granophyre. 

SEM electron backscatter images were used to study 

granophyric intergrowths of alkali feldspar and quartz in 

two dolerite samples (93-BS-84 and 93-BS-96-B-1) and 

one segregation (94-BS-154-A) sample. The data are given 

in Electronic Appendix Tables 6.5 and 6.6, respectively. In 

the dolerite and segregation samples the granophyric intergrowths 

are characterized by light grey areas corresponding 

to K-feldspar and dark grey areas corresponding to 

quartz. The K-feldspar commonly surrounds euhedral to 

subhedral plagioclase crystals in small interstitial pockets, 

whereas in the segregation sample 94-BS-154-A it forms 

larger interstitial pockets. In all samples the composition 

of the K-feldspar ranges between 84 and 89 wt % orthoclase 

(Or) with an average of 85 wt % Or. In the dolerite 




Fig. 9. Clinopyroxene compositions in dolerite 93-BS-84 (a) and segregations (b) 93-BS-80, (c) 93-BS-81, and (d) 93-BS-82. (e) Representative 

enstatite (En wt %) compositional profiles of these crystals. Right panels show the electron backscatter images of where the compositional profiles 

shown in (e) were made. Panels are arranged according to stratigraphic sequence in the field from top to bottom. 

samples, the quartz forms small intergrowths but in the 

segregations the quartz intergrowths are larger and occasionally 

form isolated crystals. The quartz analyses yielded 

compositions close to 100% SiO 2 within about 0·5% measurement 

uncertainty. 

1946 

Summary 

Plagioclase crystals in segregation profile A-1 show regular 

compositional variations between the dolerite and the 

three adjacent segregations. Plagioclase in the dolerite has 

the largest compositional variation, both between crystals 




and between core and rim, indicative of continued 

fractionation. 

The segregations all have progressively more albite-rich 

core and rim compositions (Table 6) as would be expected 

if these rocks formed as fractionated liquids from the dolerite. 

In addition, the segregations have systematically 

larger crystal sizes and they display flatter core^rim profiles 

than plagioclase in the host dolerite, possibly reflecting 

faster diffusion rates enhanced by a more volatile-rich environment 

that promoted growth and re-equilibration of 

the crystals. Plagioclase in the segregations has an average 

composition between An 53 and An 61, which is within the 

range of rim to core compositional variation (An41^62) of 

crystal 84-01, which is one of the smaller crystals from the 

dolerite sample 93-BS-84. This is consistent with the segregations 

forming from the fractionated liquid of a dolerite 

parent. 

The two most felsic segregations, which belong to group 

S 4, have comparable plagioclase compositions, but the 

more altered segregation 93-BS-82 has a wider compositional 

range and more irregular plagioclase compositional 

profiles. Plagioclases within segregation 93-BS-81 (group 

S 3) have smaller ranges of compositional variation characterized 

by slightly more anorthite-rich cores and less 

albite-rich rims (Table 6) than the two S 4 segregations, as 

would be expected for this less differentiated segregation. 

This sample shows very little compositional variation between 

cores and rims within analyzed crystals. 

The pyroxenes in the segregations have more Fe-rich 

cores and rims, and higher Ca contents than the pyroxenes 

in dolerite 93-BS-84, which is consistent with the more 

evolved composition of the segregations. Segregations 

93-BS-80 and 93-82, which have similar textures and 

major element compositions, have overlapping core and 

rim compositions with overall more Fe-rich cores and 

more Fe-rich rims than the less differentiated segregation 

93-BS-81 and the adjacent dolerite 93-BS-84 (Table 7). 

MAJOR-ELEMENT 

GEOCHEMISTRY 

In the QuartzÂlkali Feldspar^Plagioclase (QAP) ternary 

diagram (Streckeisen, 1973), the dolerites and felsic segregations 

define a trend from quartz diorite to granodiorite 

passing through the quartz monzodiorite field (Fig. 10a). 

The S 1 group rocks (black diamonds) have quartz diorite 

and quartz monzodiorite compositions, except for one 

sample, 94-BS-160-A-2-2, which plots in the granodiorite 

field. This sample is part of a large composite segregation 

that contains alternating felsic and mafic sub-lenses. The 

S2 segregations (grey filled squares) have granodiorite 

compositions except for one less differentiated sample 

94-BS-154-A, which corresponds to the less felsic component 

of segregation 94-BS-154. This sample is unusual in 

1947 

other ways and will be discussed further below. The S 3, S 4 

and S 5 segregations plot in the granodiorite field with the 

quartz component generally increasing with group 

number. Only two samples (93-BS-141 and 94-BS-212) 

from the S 5 group deviate from the main trend: these samples 

have higher K 2O concentrations and are more altered 

than any other segregation. 

In an Na 2O þK 2O^FeO T^MgO (AFM) diagram 

(Fig. 10b), the S 1 and the S 2 groups plot closer to the 

Mg-rich corner, and both display a wide range of FeO T; 

the S 2 group, which contains a higher modal abundance 

of orthopyroxene, plots closest to the FeO T^MgO sideline. 

The S3 and the S4 segregations show a systematic enrichment 

in FeO T, except that the majority of the S 3 samples 

have consistently lower FeO T concentrations and form a 

tighter cluster than the majority of the S 4 samples. The S 5 

segregations exhibit two arrays of data; one that is slightly 

higher in FeOTand one that is lower in FeOT with continued 

fractionation. Although these rocks are continuously 

enriched in the alkalis, the varying FeO T trends in 

Fig. 10b could be due to the different modal amounts of 

pyroxene and the varying amounts of alteration to Fe^Ti 

oxide and biotite. 

The S 1^S 5 segregation suite shows a wide bulk-rock compositional 

range of 51^67% SiO 2, 1^13 wt % MgO, 

7^15 wt % FeO T,8^20wt%Al 2O 3, 4^13 wt % CaO, 

0·04^0·35 wt % P 2O 5, 0·4^2·3wt % TiO 2, 1^4 wt % 

Na2O, and 0·4^2·5wt % K2O. The Mg-number [MgO/ 

(FeO T þ MgO)] ranges from 16 to 70. The alkalis 

(Na 2O þK 2O) range from 1·2 to5·5wt % with the S 5 

group having the lowest Mg-number and the highest alkalis, 

and the S 1 group having the highest Mg-number 

and the second lowest alkalis. Bulk-rock compositions of 

groups S 2, S 3 and S 5 are given in Table 2; only representative 

bulk-rock compositions of groups S 1 and S 4 are reported. 

The complete dataset for Table 2 is available as an 

Electronic Appendix. The bulk-rock compositions are 

plotted on Harker variation diagrams and shown in 

Fig. 11. In this figure, two of the three samples from group 

S 2 behave like highly differentiated members of group S 1: 

the outlying samples is 94-BS-154-A, which we will argue 

is not part of the differentiation sequence. 

There is a systematic increase in the less compatible 

elements (P 2O 5,K 2O and Na 2O) with increasing SiO 2 

across the sample suite, and a less linear but regular 

decrease in the compatible elements (MgO, CaO, Al 2O 3). 

The FeO T and TiO 2 concentrations exhibit more 

scatter. Overall, as SiO2 increases FeOT concentrations 

show an increase for group S 1 and a decrease for groups 

S 3^S 5.TiO 2 concentrations increase with SiO 2 in groups 

S 1^S 3, show a large scatter in group S 4, and decrease in 

group S 5. 

Figures 10 and 11 show that there is a compositional continuum 

between the host dolerites and the most felsic 




Fig. 10. (a) Normative mineral compositions plotted in a QuartzÂlkali Feldspar^Plagioclase diagram of the dolerites described by group 

S1 (diamonds, black ellipse), and segregations described by groups S2 (grey squares, dashed grey ellipse), S3 ( , continuous-line grey ellipse), 

S4 (triangles, black ellipse) and S5 (circles, black ellipse). Numbered regions correspond to IUGS classification rock types: 3, granite; 4, granodiorite; 

5, tonalite; 9*, quartz monzodiorite^quartz monzogabbro;10*, quartz diorite; 9, monzodiorite^monzogabbro;10, diorite^gabbroânorthosite. 

(b) AFM diagram for the same segregation^dolerite suite showing a tholeiitic trend characterized by iron enrichment followed by alkali 

enrichment with progressive differentiation, in particular in groups S 4 and S 5. 

segregations and this differentiation trend can be best 

observed in the systematic decrease of CaO with increasing 

silica in groups S 1 to S 5. The Harker diagram trends 

can be explained as the result of simultaneous fractional 

1948 

crystallization of pyroxene and plagioclase from the most 

mafic (dolerite) magma. The decrease in FeO T and TiO 2 

at higher SiO 2 is explained by the late crystallization of 

Fe^Ti oxides. 




Fig. 11. Major element compositions of the segregation^dolerite suite vs wt % SiO 2. Trends form a continuum between the most primitive composition 

corresponding to the dolerite (S 1) and the most differentiated compositions correspond to the most felsic segregations (S 5). 

MELTS MODELING OF THE 

SILICIC SEGREGATIONS 

The whole-rock and mineral chemistry data suggest that 

the silicic segregations are compositionally related to the 

host dolerite. Their compositions are broadly compatible 

with a recent model for the formation of felsic segregations 

1949 

during cooling of thick sills (Marsh, 2002), in which a crystal 

mush growing downward from the top contact of the 

sill becomes gravitationally unstable when it is 50^70% 

crystalline and tears. These tears are then filled with residual 

liquid filter pressed from directly below. According 

to this model, the segregations in the BS represent residual 

liquids after 50^70% crystallization of the host dolerite. 




As such, their bulk compositions should follow the liquid 

lines of descent of their parent magma composition. 

In this study the program MELTS (Ghiorso & Sack, 

1995; Asimow & Ghiorso, 1998) was used to model the 

crystallization of the segregations from a parent dolerite 

liquid. After determining the best parent melt composition, 

we model the formation of the segregations using different 

crystallization models to obtain the best fit with the 

compositions of the segregations. 

MELTS methods 

The MELTS program combines thermodynamic principles 

and optimization procedures to compute stable heterogeneous 

equilibria between solids and liquids at 

specified temperatures and pressures in magmatic systems. 

Given a bulk composition, equilibrium states can be calculated 

at chosen temperature or pressure intervals during 

crystallization, melting, assimilation or mixing processes. 

The proportions and compositions of all phases are given 

at each interval, and the program allows forward modeling 

of the system under equilibrium crystallization (EC) 

or fractional crystallization (FC) conditions. MELTS has 

been successfully used to model magmatic processes in 

peridotite, basalt and andesite systems (Ghiorso & 

Carmichael, 1985; Hirschmann et al., 1999; Asimow et al., 

2001; Eason & Sinton, 2006). However, it has been shown 

that the MELTS model tends to become unstable at high 

degrees of differentiation of a basaltic parent magma 

(Ghiorso & Sack, 1995); consequently, in this study all 

models were run until a lower temperature of 9008C, or 

80% crystallization in FC and 90% in EC was achieved. 

Determining the crystallization conditions 

in the Basement Sill 

To model the evolution of the BS, an initial liquid composition 

is assumed and the temperature is decreased at regular 

intervals between the liquidus and the solidus at a 

constant pressure. The change in liquid composition is 

defined as the liquid line of descent (LLD), analogous to a 

differentiation trend. If segregations in the FDS are 

simple fractionated liquids from a parent basaltic magma, 

then they should lie on the LLD. 

Based on the mineralogy and geological setting of Ferrar 

tholeiites from the southern Prince Albert Mountains in 

south Victoria Land, Demarchi et al. (2001) inferred that 

crystallization occurred at a pressure between 100 and 

500 MPa in a system that was closed to oxygen exchange. 

This estimate for the depth of emplacement may be 

extended to the BS on the basis of the work of Hersum 

et al. (2007), who estimated a paleodepth of 3^4 km 

based on the thickness of the capping comagmatic 

Kirkpatrick flood basalts. For systems that are closed to 

oxygen exchange, MELTS calculates the oxygen fugacity 

(fO2) by defining the ferrous to ferric ratio according to 

the bulk composition at each temperature for a constant 

1950 

pressure, whereas for systems that are an open to oxygen 

exchange (which we also considered) MELTS requires 

that the f O2 of the system be specified along a known 

oxygen buffer, which for magmas of this composition is 

taken to be the FMQ (fayalite^magnetite^quartz) buffer. 

MELTS calculations were also done with a more oxidizing 

buffer (nickel^nickel oxide) but as these results did not 

reveal any significant difference, they are not shown. 

The mineralogy of the dolerite is predominantly anhydrous, 

arguing for a relatively dry magma. However, the 

segregations are characterized by higher concentrations of 

hydrous phases and hydrous alteration of previously crystallized 

phases, and are coarse-grained, all indicative of 

higher H 2O concentrations in the residual liquids. 

Moreover, the quartz and K-feldspar granophyric intergrowths 

characteristic of the segregations have been interpreted 

to develop as a result of sudden H 2O loss (e.g. 

Winter, 2001). We elected to run the MELTS models for anhydrous 

conditions (0 wt % H 2O in the initial liquid composition) 

and hydrous conditions (0·5 wt % of H 2O). 

Models for the solidification of thick sills (Philpotts et al., 

1996; Marsh, 2002) suggest that over a large part of the 

crystallization history there is little relative motion between 

crystals, which are connected together in permeable 

networks, and the surrounding interstitial liquid. This 

might suggest that equilibrium crystallization (EC) would 

be the appropriate crystallization mode for the BS. 

However, conductive cooling models indicate that the 

upper levels of the sill, where the segregations occur, 

would have solidified relatively quickly (within hundreds 

of years) so there would not have been time for the centers 

of early formed crystals to equilibrate with the surrounding 

liquid by diffusion; thus we infer that the earlier-formed 

crystals were chemically removed from the system, as in 

the fractional crystallization (FC) mode. As a result, we 

chose to run the MELTS models for both EC and FC 

conditions. 

Determining the best parent magma 

compositions for the segregations 

The BS (excluding the orthopyroxene cumulate tongue) is 

a relatively uniform dolerite and so representative dolerite 

liquids were sought as the parent magma composition. To 

determine the best parent composition for the segregations, 

two representative dolerite samples were chosen 

from each of three locations within the BS; however, as 

the results for the two samples from each location were 

very similar, only one of each pair is shown in Fig. 12. 

Their locations in the BS and their bulk compositions are 

given in Table 8. The three samples shown are from the 

upper chilled margin of BS (96-A-26), a dolerite adjacent 

to a segregation lens (93-BS-90-B), and a medium-grained 

dolerite located from the centre of the sill above the cumulate 

layer (96-A-33). These samples are referred to in the 

discussion below as the CM (‘chilled margin’), 




Fig. 12. Representative MgO and CaO element modeling using three initial dolerite compositions: SA, 93-BS-90-B (continuous black line); 

CM, 96-A-26 (long-dashed line); CS, 96-A-33 (short-dashed line), in EC (left panel) and FC (right panel) MELTS models that are closed to 

oxygen exchange, at 100 MPa, for anhydrous conditions (Dry) and hydrous (0·5wt%H 2O) (Wet) conditions. 

SA (‘segregation-adjacent’) and CS (‘central sill’) samples, 

respectively. 

Although these candidates for the parental liquid have 

broadly similar compositions, there are significant differences 

in their major element compositions (Table 8). 

Interestingly, the chilled margin sample has the most 

evolved composition, suggesting that it does not represent 

1951 

the parental liquid for the entire sill and that the sill was 

built up by more than one magma pulse. 

The three candidate parent liquid compositions were 

used as starting materials in the MELTS models with the 

following permutations: (1) EC or FC, and (2) anhydrous 

or hydrous (0·5wt % H 2O present in the initial liquid). 

All models were run at a constant pressure of 100 MPa 




Table 8: Ferrar dolerites initial liquid compositions 

Type D (m) Sample Location SiO2 TiO2 Al2O3 FeOT Fe2O3 MnO MgO CaO Na2O K2O P2O5 Total 

CM 0 96-A-26 McT 54·83 0·61 14·88 9·90 1·65 0·17 7·00 10·95 1·70 0·83 0·08 101·12 

SA 52 93-BS90-B Seg B 51·89 0·38 13·81 10·57 1·76 0·19 9·33 11·77 1·33 0·42 0·04 100·37 

CS 30 96-A-33 MP 53·84 0·64 13·83 9·20 1·81 0·17 7·56 10·30 1·58 0·81 0·08 98·20 

and, in the first instance, for systems closed to oxygen 

exchange. 

MELTS results 

The results from these MELT models are shown in Fig. 12 

and Electronic Appendix Fig. 12-S. Figure 12 shows LLDs 

for the major elements MgO and CaO for the three dolerite 

samples compared with the compositions of the segregations 

(black diamonds). Each symbol along the LLDs 

corresponds to a 108C decrease in temperature, with a corresponding 

increase in silica. The left and right panels in 

the figure correspond to EC and FC conditions, respectively. 

Supplementary Data Fig. 12-S shows LLDs for major 

elements FeO and Al 2O 3 for the three dolerite initial 

liquid compositions compared with the segregations. 

As expected for liquids with a dolerite composition, the 

models indicate that with increasing SiO 2 and decreasing 

temperature the concentrations of MgO, CaO and Al 2O 3 

in the residual liquids should decrease systematically 

whereas the concentrations of FeO first increase and then 

decrease when the iron oxide phases crystallize. 

It is noteworthy that the segregation compositions are 

scattered over a range at least as wide as that between the 

LLDs, suggesting that there was more than one parent 

liquid. It can also be seen that the LLDs for the CM 

(long-dashed line in Fig. 12) often fall between those of 

the SA and CS, and that for the best-fit cases many of the 

segregation compositions also fall between these two dolerite 

LLDs. Therefore, only the SA and CS limiting compositions 

are shown in the following MELTS models. 

For the present models, closed to oxygen exchange, the 

best-fit LLDs differ for different oxides. MgO and CaO 

concentrations are best matched by a wet EC model, FeO 

concentrations by FC models, and Al 2O 3 by a dry FC 

model. 

Refining the crystallization model for 

the segregations 

The roles of starting composition and oxygen diffusion are 

evaluated in Figs 13 and 14 and Electronic Appendix 

Figs 13-S and 14-S, which illustrate LLDs of selected 

oxides for the SA and CS parent liquid compositions, for 

systems closed (dark grey lines) or open (light grey lines) 

to oxygen exchange. 

1952 

These figures show that the degree of fit between LLDs 

and the segregation data is not sensitive to oxygen diffusion 

for any of the oxides, except perhaps for FeO when 

using the wet FC model. The starting composition is a 

more important variable. For the FC models, the FeO concentrations 

of the segregations are bracketed by the 

LLDs for the SA and CS dolerite starting compositions, 

with the hydrous FC model providing the best fit 

(Fig. 13d). The inflection points in these LLDs correspond 

to the crystallization of magnetite, which occurs at 

55 wt % SiO 2 and 60 wt % SiO 2 for the SA and CS 

samples, respectively. 

In the segregations the measured CaO concentrations 

decrease rapidly with increasing SiO 2 up to 59 wt % 

SiO 2 after which the decrease is less rapid, whereas in the 

MELTS models this change occurs at 55 wt % and 

60 wt % SiO 2 in the FC models using SA and CS dolerite 

compositions, respectively (Fig. 13f and h). In the EC 

models there is little change in the slope regardless of the 

initial composition used except at the lowest SiO 2 

(Fig. 13e and g). In the models, the initial rapid decrease is 

due to the crystallization of Ca-rich pyroxene and plagioclase 

but with continued crystallization the plagioclase becomes 

progressively more albite rich and thus consumes 

more sodium than calcium. Supplementary Data Fig. 13-S 

shows LLDs for MgO and Al2O3. The models that best approximate 

the compositions of the segregations are for 

MgO the EC models (Supplementary Data Fig. 13-Sa and 

c), and for Al 2O 3 the dry FC model (Supplementary Data 

Fig. 13-Sf). 

Modeling of the less compatible elements, Na2O and 

K 2O, generally shows an increase in concentration with 

increasing silica and decreasing temperature, as expected 

(Fig. 14). However, the concentrations in the segregations 

differ significantly from those predicted by the MELTS 

LLDs. Above 59 wt % SiO 2,K 2O in the segregations remains 

approximately constant at 1·5 wt %, whereas the 

K 2O concentration in the model liquids continues to increase 

with increasing SiO2. With respect to Na2O, the 

concentrations in the model liquids rise to a maximum 

before declining at high wt % SiO2, whereas concentrations 

in the segregations increase monotonically with 

increasing SiO 2. The reason for the discrepancy is not 




Fig. 13. Representative FeO and CaO element modeling using SA, 93-BS-90-B (continuous line) and CS, 96-A-33 (dashed line), in a EC model 

(left panel) and FC model (right panel) for systems that are closed to oxygen exchange (dark grey (black)), and systems that are open to 

oxygen exchange (grey) (light grey), with no H 2O present in the liquid, and 0·5wt % H 2O present in the initial liquid composition. The 

black diamonds show segregation compositions. 

clear; however, evidence for small amounts of deuteric and 

hydrothermal alteration in the segregations manifest by 

the presence of amphibole and biotite replacing pyroxene, 

minor secondary white mica replacing biotite, and sericite 

replacing plagioclase may be significant in this regard. 

Also, there is some K and Na ion exchange along the 

1953 

regions where the granophyric intergrowths surround subhedral 

or partially resorbed plagioclase crystals that 

might account for the difference in the K and Na trends 

in the segregations. Although none of the models provides 

a good fit to the alkali concentrations in the segregations, 

the wet FC model is closest (Fig. 14d and h). 




Fig. 14. Representative K 2O and Na 2O element modeling using SA, 93-BS-90-B (continuous line) and CS, 96-A-33 (dashed line), in a EC 

model (left panel) and FC model (right panel) for systems that are closed to oxygen exchange (black) (dark grey), and systems that are open 

to oxygen exchange (grey) (light grey), with no H2O present in the liquid, and 0·5wt % H2O present in the initial liquid composition. The 

black diamonds show segregation compositions. 

The P 2O 5 concentrations in the segregations are well 

approximated by the LLDs for this element, in particular 

the FC model (Electronic Appendix Fig. 14-Sb and d), 

whereas the TiO 2 concentrations show too wide a scatter 

to be properly modeled by either EC or FC models 

1954 

(Electronic Appendix Fig. 14-Se^h), probably owing to the 

coarse-grained and heterogeneous textures of the 

segregations. 

In summary, the MELTS model results show that none 

of the models provides good fits for all oxides as the 




segregation compositions do not follow the simple LLDs 

expected for fractionated residual liquids. However, overall, 

considering both compatible and less compatible elements 

(Figs 12^14 and 12-S-14-S), the wet FC closed-system 

model provides the best fit to the segregation compositions 

for most elements. For MgO and CaO EC models provide 

better fits but principally at high silica values for the most 

differentiated of the segregations (group S 5). 

The differences for single elements indicate that no 

single, simple crystallization model can adequately explain 

all the data for such a dynamically evolving system, but 

overall we conclude that the most representative model 

for the segregations is an FC model that starts from a 

range of different initial magma conditions (between SA 

and CS), which crystallize close to the FMQ buffer 

(closed to oxygen exchange) and have at least 0·5wt % 

H 2O present in the initial liquid composition. 

MELTS crystallizing phases 

The MELTS models predict the sequence and temperature 

of crystallization and the per cent crystallization (f) of 

the solid phases from the liquid. This information is summarized 

in Fig. 15 for liquids SA and CS, using FC hydrous 

models closed to oxygen exchange. The top panel displays 

the silica content of the residual liquid, and the crystals 

that form from it, as a function of crystallinity, and the 

bottom panel shows temperature as a function of crystallinity. 

The models were run from the liquidus to a 

near-solidus temperature of 9008C, beyond which the 

MELTS models became unstable. 

Figure 15a shows that the cotectic mineral assemblage 

pigeonite (pig), augite (aug) and plagioclase (plg) form at 

18% and 16% crystallization (f) for the SA and the CS 

liquids, respectively. The first mineral to crystallize is 

Mg-rich olivine in the SA liquid and Mg-rich orthopyroxene 

in the CS dolerite liquid, which has a higher initial 

SiO 2 concentration. It is important to note that the first 

mineral to crystallize in both cases is shortly replaced by 

pigeonite or augite and subsequently joined by plagioclase. 

Magnetite begins to crystallize at 10778C corresponding 

to f ¼ 59% in the SA liquid, and 10918C corresponding 

to f ¼ 42% in the CS liquid. When the temperature 

decreases to 9978C (f ¼75%) in the SA liquid and 9218C 

(f ¼72%) in the CS liquid, an Fe-rich olivine phase replaces 

pigeonite and continues to crystallize along with 

augite, plagioclase and magnetite. The composition of the 

augite becomes more Fe- and Ca-rich, the composition of 

the plagioclase becomes more Na-rich and the composition 

of the magnetite becomes more Ti-rich with decreasing 

temperature. Only in the dry FC model using dolerite 

liquid CS does ilmenite (FeTiO 3) crystallize along with 

olivine, augite and plagioclase at a temperature of 9278C. 

These crystallization sequences were also calculated for 

different models. For the two starting dolerite compositions, 

both EC and FC models have similar crystallization 

1955 

sequences involving two pyroxenes (pigeonite and augite), 

plagioclase and magnetite, and at very low crystallization 

temperatures and under varying H 2O conditions, ilmenite, 

quartz and an H2O phase form, but at different temperatures 

and degrees of crystallization. The main differences 

between the EC and FC models are that the mineral compositions 

are more evolved in the FC model at each temperature 

step, as would be expected for an FC model. 

Also, well after (1008C or so cooling) magnetite crystallizes, 

an Fe-rich orthopyroxene in the EC model and an 

Fe-rich olivine in the FC model replace pigeonite. The 

quartz phase crystallizing during the last stages of crystallization 

in the EC model liquids could react with the 

Ca-poor pyroxene to form more Fe-rich olivine. 

Another distinction between EC and FC models is that 

quartz crystallizes in the dry EC models. This occurs at a 

temperature of 9078C (f ¼92%) for the SA liquid and at 

9198C (f ¼82%) for the CS liquid. Water vapour exsolves 

in the wet EC model below 9178C (f487%) for the SA 

liquid only. These phases were not seen in the FC models 

probably because the upper limit on crystallization, 

beyond which the MELTS model became unstable, was 

only 80% for the FC models, whereas for the EC model 

it was 98%. By extrapolating from the EC models it is 

plausible that quartz would crystallize and H 2O vapour 

would exsolve if the FC models could reach similar degrees 

of crystallinity. The crystallization of quartz and exsolution 

of volatiles (H2O) at the last stages of crystallization is 

compatible with the granophyric textures observed in the 

segregations. 

The LLDs in the FC models (Figs 12 and 13) show that 

during the first several tens of degrees of cooling, the 

MgO content in the liquid decreases sharply and there is 

a less marked decrease in CaO, whereas the FeO content 

of the liquid increases. These changes can be directly tied 

to the crystallization of Mg-rich pyroxene and Ca-rich 

plagioclase, which remove MgO and CaO while allowing 

FeO and the less compatible elements to build up in the 

liquid. After magnetite joins the cotectic there is an 

abrupt decrease in the FeO and an abrupt increase in 

SiO 2 concentrations in the liquid. With continued crystallization, 

the TiO 2 concentration in the liquid also decreases 

abruptly as the magnetite becomes more Ti-rich 

and ilmenite (FeTiO3) joins the cotectic. 

The addition of H 2O depresses the liquidus temperature 

and promotes early crystallization of magnetite in both 

EC and FC models, decreasing the FeO concentrations in 

the modeled liquids. However, as mentioned above, an 

H 2O vapour phase is observed only in one of the wet EC 

models. With continued fractionation and increasing volatiles 

the MELTS program becomes unstable for these 

types of compositions, making it impossible to quantify 

processes such as the formation of the granophyric intergrowths 

and the deuteric alteration process in the 




Fig. 15. (a) SiO 2 (wt %) variation with per cent crystallization for two LLDs corresponding to SA, 93-BS-90-B (circles) and CS, 96-A-33 

(squares) dolerite liquids using a closed, hydrous, FC model. The highest silica content of the segregations (67%) is used to provide an upper 

bound on the crystallization interval over which the segregations formed. For the SA liquid the segregations crystallized up to 81% and for 

the CS liquid they crystallized up to 67%. A lower bound of 33% crystallization (which is well above the cotectic) is used for both liquids. 

It should be noted that silica becomes significantly enriched only after magnetite (mt) crystallizes. (b) Temperature vs crystallinity for the 

same two LLDs. The shaded regions represent the crystallization interval over which the segregations could have formed from the SA liquid 

(light grey) and the CS liquid (dark grey). For the SA liquid the corresponding crystallization temperatures are between 1138 and 9348C and 

for the CS liquid they are between 1113 and 9788C. 

1956 




minerals, all of which occur at the very last stages of 

crystallization. 

Crystallinity and temperature during 

felsic segregation formation 

Using the two model LLDs for the closed, hydrous FC 

model, we can use the silica content of the segregations 

(52^67 wt %) to attempt to find bounds on the crystallinity 

interval over which the segregations formed, as illustrated 

by grey shaded region in Fig. 15a. The highest silica contents 

of 67 wt % correspond to 81% crystallization of 

liquid SA and 67% crystallization of liquid CS. We should 

note, however, that the silica content of the liquid changes 

little until magnetite begins to crystallize, which occurs at 

59% crystallization for liquid SA and 42% crystallization 

for liquid CS. Thus if a segregation forms prior to these 

crystallization intervals, it would not be particularly enriched 

in silica. 

Because there is significant overlap in the SiO2 contents 

of the dolerite samples (50 54 wt %) and the low SiO 2 

segregations in group S 1 (51^55 wt %), it is not possible to 

constrain accurately the lower bounds on the crystallinity 

interval over which the segregations form. Philpotts & 

Dickson (2000) suggested that an interconnected crystalline 

mush capable of being torn would form at as low as 

33% crystallinity in a basaltic magma. From Fig. 15a we 

observe that multiple saturation of two pyroxenes and 

plagioclase is reached at 18% and 16% crystallinity for 

liquids SA and CS, respectively, so that an extensive interconnected 

network of crystals would exist at a crystallinity 

of 33%. Using SA as the initial composition, the segregations 

might have separated from a mush at between 33 

and 81% crystallization of the parent liquid (light grey 

shaded area Fig. 15b), which corresponds to a temperature 

of between 1138 and 9348C. For a CS initial composition, 

and a range of crystallinity from 33 to 67%, the temperature 

range is from 1113 to 9788C (dark grey shaded area 

in Fig. 15b). 

At crystallinity of 33% for the CS composition, the silica 

content of the residual liquid is about 56%, corresponding 

to the middle of the range for the group S 3 segregations 

and the lower end of the range for the S4 segregations 

(Fig. 11). This suggests that the low-SiO 2 segregations 

(groups S 1 and S 2) originated from a parent melt similar 

to SA. We note, however, that according the physical 

model we present below, it is possible that poorly differentiated 

or undifferentiated liquids and crystals might be 

channelled upward into a segregation lens from deeper in 

the sill rather than separating from the adjacent crystal 

mush. 

At temperatures as low as 9348C the liquid may be too 

viscous to segregate effectively, so we suggest that if the segregations 

have silica concentrations greater than, say, 

55^60 wt % they were probably segregated from a parent 

more like CS than SA. A residual liquid of CS with 67% 

1957 

SiO 2 would have separated from a mush that was 67% 

crystalline at a temperature of 9788C. Alternatively, the 

most silica-rich segregations might have formed from a 

parent liquid like SA at high crystallinity, if the increasing 

concentration of volatiles (H2O) significantly reduced the 

viscosity of the residual liquid, making it possible to segregate 

it at these low temperatures. 

Summary of the MELTS modeling 

Although none of the MELTS models carried out in this 

study provide good matches for all of the segregation 

data, the segregation compositions are best explained by 

fractional crystallization (FC) of a parent dolerite liquid 

at about 100 MPa, with up to 0·5wt % H2O. This model 

is in agreement with previous models of the FDS 

(Demarchi et al., 2001) and more recent models on the crystallization 

of the BS (Marsh, 2004; Be¤ dard et al., 2007; 

Boudreau & Simon, 2007). 

The FC model is arguably more appropriate than EC 

models owing to rapid crystallization relative to solid-state 

diffusion, such that that the cores of early formed crystals 

remain chemically isolated from the liquid. This effect 

may be enhanced as a result of multiple intrusions within 

the BS (see below). 

The segregation compositions do not follow the LLDs 

expected for residual liquids. A number of factors are expected 

to contribute to this discrepancy. First, the segregation 

compositions could represent mixtures of liquids and 

crystals entrained from the mush, as they are filter pressed 

through the mush to form the segregations (Zavala, 2005; 

Fodor, 2008). The scatter of segregation compositions also 

suggests that these segregations were not derived from a 

single parent liquid; many of the segregations fall between 

the liquid lines of descent of the two dolerites: 93-BS-90-B, 

adjacent to a segregation lens, and 96-A-33, from a deeper 

level in the BS sill. 

MODEL FOR FORMATION OF 

THE SEGREGATIONS 

The proposed physical model for formation of the segregations 

is illustrated in Fig. 16, which shows schematic snapshots 

of the sill at different times. The main solidification 

process is based on existing models for sheet-like magma 

bodies (Philpotts & Dickson, 2000; Marsh, 2002; Philpotts 

& Philpotts, 2005). In these models solidification fronts 

propagate inwards with time from the upper and lower 

boundaries. These fronts are crystal networks with lower 

crystallinity and higher temperature than the more completely 

crystallized magma nearer the contact. Near the 

‘capture front’ the crystal network becomes weak and very 

porous and most crystals are connected together on a 

short length scale; below the front a few crystals exist as 

single suspended crystals. The boundary between mush 

and magma is a feathery, irregular region with a large 



(a) 

(b) 

(c) 


roof 

surface area providing many sites for nucleation of new 

crystals. The upper solidification front is inherently gravitationally 

unstable, and we expect that in any but the thinnest 

sills, clusters of loosely bound crystals 

(glomerocrysts) periodically detach from the capture 

front and settle onto the lower solidification front; some of 

these crystals may be resorbed if they reach the hotter 

liquid below, but on their way they stir the intervening 

magma (Tait & Jaupart, 1992). This stage is illustrated 

schematically in Fig. 16a. With time, as the solidification 

front becomes broader and heavier, larger and more crystalline 

clusters can detach (Fig. 16b). 

For sills 100 m thick, detachment can occur as propagating, 

plastic tears with sharp contacts at levels in the 

front where crystallinities are as low as 30^35% 

(Philpotts & Dickson, 2000). The segregations in the BS 

usually have sharp upper and lower contacts. This suggests 

floor 

Fig. 16. Schematic model for the evolution of a thick (4100 m) sill. Not to scale. Panels indicate crystallinity and dynamics of a representative 

section at different times during solidification. Shades of grey indicate degree of crystallinity: black 70^100%, dark grey 40^70%, medium 

grey 25^40% crystals, lightest grey is crystal-poor magma, and white is segregated residual magma. Arrows indicate direction of motion, 

hatched pattern indicates settled cumulate crystals. (a^c) Evolution of a single pulse of crystal-poor magma. (d, e) Evolution after a second 

pulse of crystal-rich magma, entering from the left. The increased thickness of the section should be noted. (See text for explanation.) 

1958 

(d) 

(e) 

that when tearing occurred the mush was sufficiently crystalline 

to preserve both margins of the tears. Eventually 

the upper and lower solidification fronts will meet, 

although not necessarily at the same time in all locations 

(Fig. 16c), filling the centre of the sill with a weak, heterogeneous 

crystal network. Tearing and the formation of segregations 

may occur in the upper front before the fronts 

meetçor after the fronts meet as the weak network compacts. 

Tearing ceases as the upper front rests stably on the 

lower front. 

The weakest, central part of the crystalline network may 

subsequently be disrupted if a new pulse of magma inflates 

the sill (Fig. 16d). The melt injection may rip off and disrupt 

weak crystal clusters from the fronts and introduce 

suspended crystals. The development of a crystal network 

requires a low-energy environment. In the energetic, 

forced flow of a new magma pulse, any solid fraction will 




consists of single crystals that can move relative to one another 

and can be sorted by sedimentary processes (e.g. 

Middleton, 1970). Sedimentation of introduced crystals is 

probably responsible for the cumulate tongue of the BS 

(Fig. 16d). 

When the crystal network in the centre of the sill is 

replaced with a layer of liquid (Fig. 16c and d), the upper 

solidification front becomes newly gravitationally unstable. 

If the upper solidification front is thin, a feathery boundary 

region is re-established and solidification continues as 

shown in Fig. 16a and b. If, however, the solidification 

front is thick, then a new magma influx can instigate a 

new episode of tearing. Where earlier tears are incompletely 

crystallized and constitute weak regions in the front, 

they may be reactivated and extended, propagating sideways 

and vertically in subparallel interconnected fractures 

(Hersum et al., 2005) (Fig. 16d). Earthquakes and fluid 

shear stresses associated with fresh magma influxes would 

add to the gravitational stresses to cause reactivation and 

extension of old tears. The sets of sub-horizontal anastomosing, 

thick ( 0·5^2 m) lenses observed in the upper 

parts of the BS are thought to have formed by this process. 

Interpretation of mafic pegmatoids at the Dais suggests a 

similar formation process, in which liquids forming the 

pegmatoids were filter pressed from the surrounding gabbronorite 

during compaction of the mush (Geist et al., 

2005). 

Tearing would cease when the gravitational instability 

causing the tear was eliminated, either because the mush 

at deeper levels beneath the tear delaminated, reducing 

the weight, or because the foundering underlying mush 

became partially supported by the lower solidification 

front (Fig. 16e). 

The most felsic segregations, felsic pods and stringers, 

seen particularly well at the Dais, could have formed 

during the last stages of crystallization when an H 2O-rich 

fluid phase segregated from the residual liquid by 

crystallization-induced degassing (i.e. second boiling). It 

has been proposed that exsolution of gas leads to an increase 

in pressure, which drives late-stage residual fluid 

from the crystal-rich matrix into crystal-poor zones under 

lower pressure (Anderson et al., 1984; Goff, 1996; Sisson & 

Bacon, 1999; Caroff et al., 2000; Boudreau & Simon, 2007). 

Philpotts et al. (1996) related centimetre-scale felsic blobs 

within mafic segregation sheets to exsolution of volatiles. 

A higher partial pressure of H 2O or cooling (e.g. Puffer 

& Horter, 1993) could cause fracturing in the highly crystalline 

region near the upper contact of the BS. Late residual 

liquid could collect in these cooling and degassing 

fractures to form pods and streaks of felsic material 

(Fig. 16e). 

Multiple influxes of magma 

The present model differs from that of Marsh (2002) in 

that the tearing process caused by gravitational instability 

1959 

of the upper solidification front occurred at various times 

and places throughout the crystallization of the BS, triggered 

by multiple episodes of magma influx. Marsh 

(2002) suggested that segregation lenses ceased opening 

when they froze, whereas we propose tearing ceased when 

the gravitational stress was removed. 

The existence of multiple pulses of slightly different 

input magma (Philpotts & Philpotts, 2005) would help to 

account for the range of parent liquids suggested from the 

MELTS modelling (Figs 12^14). These pulses are inferred 

to have undergone varied degrees of fractional crystallization 

at deeper levels. The most dramatic evidence for multiple 

inputs is the opx-rich tongue midway in the BS. Its 

location indicates that the lower BS had previously crystallized 

(thereby preventing the orthopyroxene phenocrysts 

from accumulating at the lower margin of the BS). Recent 

work on the crystallization of the BS (Marsh, 2004; 

Hersum et al., 2007; Be¤ dard et al., 2007; Boudreau & 

Simon, 2007) suggests that the orthopyroxene-rich tongue 

in the BS was intruded as a crystal-rich mush after being 

texturally equilibrated at a much deeper level. 

There is not a strong correlation between the SiO 2 concentration 

in the segregations and their position in the 

BS. Although high-SiO 2 segregations tend to occur closer 

to the top contact, low-SiO 2 segregations are found at all 

levels (Fig. 17). Additionally, many segregations show internal 

layering that may be related to multiple pulses into 

a given lens (Table 1). The anomalous group S 2 segregation 

sample 94-BS-154-A occurs at a level close to the top of 

the opx tongue (square at D 100 m in Fig. 17). This location 

in combination with its unusual composition suggests 

that it is related to the opx tongue intrusion, that it may 

in fact be a part of that crystal-rich influx forced into a 

tear in the dolerite mush. 

Therefore, it is suggested that there may have been multiple 

episodes of segregation formation triggered by fresh 

influxes of magma into the sill, and that old tears could 

have been reactivated at different times. This is consistent 

with the wide range of temperatures and predicted crystallinities 

for the melt segregations. Moreover, tears in the 

mush may propagate by a system of fractures (Philpotts 

et al., 1996; Zavala, 2005) allowing liquid derived from the 

crystal mush at one level to move to a higher level, so that 

SiO 2-poor segregations need not necessarily be found adjacent 

to their parent mush. For instance, segregations in 

the Kilauea Iki lava lake include material from the adjacent 

mush and a deeper source ( 77^105 m) in the lake 

(T. R. Helz, personal communication, 2007). 

Mixed liquids and crystals 

Previous workers have considered that felsic segregations 

are filter-pressed residual liquids, whereas here we suggest 

that they are made up of mixtures of crystals and residual 

liquids that might come from deeper levels in the sill as 

well as locally. If tearing is an abrupt process triggered by 




Fig. 17. SiO 2 (wt %) content of rocks vs distance from the upper contact of the BS. S 1 dolerites (light grey diamonds), S 2 segregation (squares), 

S3 segregation ( ), S4 (triangle), S5 (circles), cumulates (þ), and Ferrar Dolerites elsewhere (black diamonds). All segregations occur within 

the upper 110 m of the BS. 

influxes of magma, rather than a very gradual process, 

then it is easy to envisage that crystals could be entrained 

from the mush, whereas others were trapped in the mush 

owing to unfavourable aspect ratios. The spread in the 

bulk chemical compositions of the segregations and the 

way in which the segregations can be grouped according 

to the minerals added to or subtracted from a residual 

liquid support this idea. The regular but varied range in 

texture, as well as the presence of deformed pyroxene 

and plagioclase crystals, are additional lines of evidence. 

Helz et al., 1989 reported a narrow compositional range 

( 1^6 wt % CaO) for filter-pressed interstitial liquids 

found in the upper parts of the Kilauea Iki lava lake. The 

fact that the BS segregations show such a wide compositional 

range (52^67 wt % SiO 2 and 4^13 wt % CaO) 

could suggest that the segregations represent mixtures of 

crystals and liquids. 

SUMMARY AND CONCLUSIONS 

Coarse-grained leucocratic segregations are a ubiquitous 

feature within the upper 100 m of the 300 m thick 

Basement Sill (BS), the lowermost of four related Ferrar 

Dolerite Sills exposed in the McMurdo DryValleys region 

1960 

of South Victoria Land, Antarctica (Fig. 1). Based on field 

observations, petrography, major element geochemistry 

and thermodynamic modelling, we suggest that the segregations 

were formed by tearing of the upper solidification 

front of the partially solidified BS as a result of gravitational 

instability, and that their range of compositions and textures 

can be explained by multiple episodes of tearing and 

infilling by liquids and crystals of various compositions 

during the crystallization of the BS. 

Segregations within the BS have bulk chemical compositions 

and mineralogy that suggest they are composed of 

varying proportions of interstitial liquid and crystals 

derived during fractional crystallization of the host dolerite. 

Their compositions form a continuum between the 

host sill (52 wt % SiO 2) and the most fractionated segregations 

(67 wt % SiO 2), following a tholeiitic differentiation 

trend (Fig. 10b). 

There are systematic changes in mineralogy and texture 

with increasing silica concentration. Dolerites have fineto 

medium-grained ophitic and subophitic textures, with 

minor amounts of interstitial quartz and alkali feldspar in 

granophyric intergrowths, opaque minerals and alteration 

minerals. As the silica content increases, the segregations 

become coarser-grained and develop sub-ophitic and 




poikilophitic textures, and the amounts of secondary alteration, 

opaque minerals and granophyric intergrowths increase. 

These features are explained by separation of 

residual liquid, plus or minus crystals, from the crystallizing 

magma over a range of crystallinities. In segregations 

with intermediate compositions and variable grain size, 

some augite crystals are bent and some plagioclase crystals 

are broken, bent and have kinked twins, suggesting these 

crystals were entrained in a partially crystalline, deforming 

medium. 

Modelling using the thermodynamic program MELTS 

demonstrates that the major element compositions of the 

segregations are largely consistent with their being residual 

liquids formed at various stages during fractional crystallization 

of a basaltic magma. The segregations could not 

have been derived from a single parent liquid; however, 

three dolerite samples from different positions in the BS 

constrain a range of fractionation paths that could account 

for most of the felsic segregation compositions. This suggests 

that the BS was built up from more than one pulse 

of slightly heterogeneous magma. However, considering 

the petrographic evidence of resorbed and deformed crystals, 

it is likely that the scatter in segregation compositions 

is at least partly due to the presence of crystals entrained 

into the melt lenses with the residual liquid. 

The above observations indicate that the felsic segregations 

were derived from the host dolerite mushes at various 

stages of their crystallization. Most segregations, in particular 

the most common, sub-horizontal anastomosing 

lenses, appear to have formed by tearing of partially crystallized 

mush within the upper solidification front and infilling 

of the tears with a mixture of interstitial liquid and 

crystals from the adjacent mush and from deeper in the 

sill. 

Based on the distribution of segregation compositions 

with location in the BS (Fig. 17), we suggest that multiple 

episodes of tearing occurred over the same range of 

elevations within the upper solidification front at different 

times. The episodes could have been triggered by new 

influxes of magma into the sill, which swept away a weak 

crystal network that was stabilizing the upper front, and 

suspended it on a layer of magma, making it newly susceptible 

to tearing. A thick, coarse-grained cumulate ‘tongue’ 

located midway in the sill indicates that there was at least 

one major late-stage magma influx event into the BS. 

Collection of residual liquid within the segregations was 

initially by porous flow through a crystal mush adjacent 

to the tears, and then by channel flow along propagating 

tears and fractures. Stratified segregations are thought to 

represent tears that were reactivated multiple times, 

whereas the uniformly coarse-grained segregations with 

gradational contacts are thought to represent residual liquids 

that were injected into a cooler part of the mush via 

fractures or channels. 

1961 

Smaller, irregular, sub-vertical veins, pods and stringers 

are located above the anastomosing lenses, where the 

mush was presumably cooler and more crystalline. 

Vertical transport at this level could have been aided by 

tearing or weaknesses in the front that formed in response 

to cooling and contraction, exsolution of volatiles at high 

degrees of crystallinities effectively reducing the viscosity 

of the liquid, and by the positive buoyancy of collections 

of interstitial liquid derived from deeper levels in the front 

similar to the segregation veins in the Kilauea Iki lava 

lake (Helz, 1980, 1987; Helz et al., 1989). 

A key objective in igneous petrology is to understand the 

range of physical processes by which rocks of diverse composition 

can be produced. Many such processes involve 

the physical separation of crystals and liquid in partially 

molten systems. The segregations at the BS have compositions 

intermediate between basaltic and rhyolitic rocks 

that are rarely found in large-scale basaltic systems (e.g. 

Hildreth, 1979; Ewart, 1982; Marsh et al., 1991), and studies 

like the present one can provide insight into the processes 

that operate at these intermediate stages of differentiation. 

Our model, involving gravitational detachment of unstable 

crystal mushes, the propagation of plastic tears, and the 

movement of interstitial liquids by porous flow, channel 

flow and diapirism, takes important elements from the 

models of previous workers (e.g. Tait & Jaupart, 1992; 

Philpotts et al., 1996; Marsh, 2002; Bachmann & Bergantz, 

2004). However, we argue for the importance of multiple 

inputs of magma into thick sills in generating and preserving 

segregation features, and point out that the material 

entering the segregations is rarely pure interstitial liquid 

(Zavala, 2005). 

Deformation and disruption of the crystal mush in the 

sill and particularly in the gravitationally unstable upper 

solidification front are key elements of our model. To fully 

understand the temporal and spatial relationships of segregations 

like those in the BS, a better physical understanding 

of the rheology of crystal mushes is needed. Further 

field and geochemical studies, and mathematical and experimental 

modeling of crystal^liquid separation will be 

essential for fully understanding the physical processes of 

segregation formation. 

ACKNOWLEDGEMENTS 

K.Z. would like to thank Bruce Marsh for the opportunity 

to participate in his Dry Valleys Antarctica project, and 

for input in the early stages of this work. K.Z. also thanks 

Maya Wheelock for collecting the samples that were analyzed 

in this study. Toby Rivers, Jean Be¤ dard, Teal Riley, 

Rosalind Helz and an anonymous reviewer are thanked 

for their helpful reviews. Careful and constructive comments 

from Executive Editor Marjorie Wilson are greatly 

appreciated. 



FUNDING 

This research was supported by NSF Research grant 

P420C652232. 

SUPPLEMENTARY DATA 

Supplementary data for this paper are available at Journal 

of Petrology online. 

REFERENCES 


Abbey, S. (1983). Studies in ‘Standard Samples’ of silicate rocks minerals 

1969^1982. Geological Survey of Canada Paper 83, 1^114. 

Anderson, T. A., Jr, Swihart, H. G., Artioli, G. & Geiger, A. C. (1984). 

Segregation vesicles, gas filter-pressing, and igneous differentiation. 

Journal of Geology 92, 55^72. 

Antonini, P., Piccirillo, M. E., Petrini, R., Civetta, L., D’Antonio, M. 

& Orsi, G. (1999). Enriched mantle-Dupal signature in the genesis 

of the Jurassic Ferrar tholeiites from Prince Albert Mountains 

(Victoria Land, Antarctica). Contributions to Mineralogy and Petrology 

136,1^19. 

Armstrong, J. T. (1989). CITZAF: Combined ZAF and Phi-rho (Z) Electron 

Beam Correction Programs. Pasadena, CA: California Institute of 

Technology. 

Asimow, P. D. & Ghiorso, M. S. (1998). Algorithmic modifications extending 

MELTS to calculate subsolidus phase relations. American 

Mineralogist 83, 1127^1132. 

Asimow, P. D., Hirschmann, M. M. & Stolper, E. M. (2001). 

Calculation of peridotite melting from thermodynamic models of 

minerals and melts, IV. Adiabatic decompression and the composition 

and mean properties of mid-ocean ridge basalts. Journal of 

Petrology 42, 963^998. 

Bachmann, O. & Bergantz, G. W. (2004). On the origin of crystal-poor 

rhyolites: extracted from batholithic mushes. Journal of Petrology 45, 

1565^1582. 

Bailey, M. M. (1989). Evidence for magma recharge and assimilation 

in the Picture Gorge Basalt Subgroup, Columbia River Basalt 

Group. In: Reidel, S. P. & Hooper, P. R. (eds) Volcanism and 

Tectonism in the Columbia River Flood-Basalt Province. Geological Society 

of America, Special Papers 239,343^355. 

Be¤ dard, J. H. J., Marsh, D. B., Hersum, G. T., Naslund, H. R. & 

Mukasa, B. S. (2007). Large-scale mechanical redistribution of 

orthopyroxene and plagioclase in the Basement Sill, Ferrar 

Dolerites, McMurdo Dry Valleys, Antractica: petrological, 

mineral-chemical and field evidence for channelized movement of 

crystals and melt. Journal of Petrology 48, 2289^2326. 

Boudreau, A. & Simon, A. (2007). Crystallization and degassing in the 

Basement Sill, McMurdo Dry Valleys, Antarctica. Journal of 

Petrology 48, 1369^1386. 

Bowen, N. L. (1928). The Evolution of the Igneous Rocks. NewYork: Dover. 

Boyd, F. R. & Mertzman, S. A. (1987). Composition and structure of 

the Kaapvaal lithosphere, Southern Africa. In: Mysen, B. O. (ed.) 

Magmatic Processes. Geochemical Society Special Publications 1,13^24. 

Brewer, T. S., Hergt, J. M., Hawkesworth, C. J., Rex, D. & Storey, B. 

C. (1992). Coats Land dolerites and the generation of Antarctic 

continental flood basalts. In: Storey, B. C., Alabaster, T. & 

Pankhurst, R. J. (eds) Magmatism and the Causes of Continental 

Break-up. Geological Society, London, Special Publications 68, 185^208. 

Caroff, M., Maury, C. M., Cotten, J. & Clement, J. P. (2000). 

Segregation structures in vapor differentiated basaltic flows. 

Bulletin of Volcanology 62, 171^187. 

1962 

Cornwall, R. H. (1951). Differentiation in lavas of the Keweenawan 

series and the origin of the copper deposits of Michigan. Geological 

Society of America Bulletin 62, 159^202. 

Demarchi, G., Antonini, P., Piccirillo, M., Orsi, G., Civetta, L. & 

D’Antonio, M (2001). Significance of orthopyroxene and major 

element constraints on the petrogenesis of Ferrar tholeiites from 

southern Prince Albert Mountains, Victoria Land, Antarctica. 

Contributions to Mineralogy and Petrology 142(2),127^146. 

Eason, D. & Sinton, J. (2006). Origin of high-Al N-MORB by fractional 

crystallization in the upper mantle beneath the Galapagos 

Spreading Center. Earth and Planetary Science Letters 252,423^436. 

Elliot, D. H. & Fleming, T. H. (2004). Occurrence and dispersal of 

magmas in the Jurassic Ferrar Large Igneous Province, 

Antarctica. Gondwana Research 7, 223^237. 

Elliot, D. H. & Fleming, T. H. (2008). Physical volcanology and geological 

relationships of the Jurassic Ferrar Large Igneous Province, 

Antarctica. Journal of Volcanology and Geothermal Research 172,20^37. 

Encarnacion, J., Fleming, H. T., Elliot, H. D. & Eales, V. H. (1996). 

Synchronous emplacement of Ferrar and Karoo dolerites and the 

early breakup of Gondwana. Geology 24, 53^58. 

Ewart, A. (1982). The mineralogy and petrology of Tertiary^Recent 

orogenic volcanic rocks with special reference to the andesite^ 

basalt compositional range. In: Thorpe, R. S. (ed.) Andesites; 

Orogenic Andesites and Related Rocks. Chichester: John Wiley, 

pp. 1^200. 

Fisher, G. W. (1989). Matrix analysis of metamorphic assemblages and 

reactions. Contributions to Mineralogy and Petrology 102,69^77. 

Fleming, T. H., Heimann, A., Foland, K. A. & Elliot, D. H. (1997). 

40 Ar/ 39 Ar geochronology of Ferrar Dolerite sills from the 

Transantarctic Mountains, Antarctica; implications for the age 

and origin of the Ferrar magmatic province. Geological Society of 

America Bulletin 109, 533^546. 

Fodor, V. R. (2008). Diorite segregations in gabbro: geochemical characteristics 

and conditions for origin assessed at diorite^gabbro contacts. 

Journal of Geology 117, 109^125. 

Geist, D., Boudreau, A., Garcia, M., Harpp, K., Mathez, E. & 

Marsh, B. (2005). Origin of magic pegmatoids in the Dais 

Intrusion, Wright Valley, Antarctica, American Geophysical 

Union Meeting, San Francisco, abstract V14C-07. 

Ghiorso, M. S. & Carmichael, I. S. E. (1985). Chemical mass transfer 

in magmatic processes. II. Applications in equilibrium crystallization, 

fractional and assimilation. Contributions to Mineralogy and 

Petrology 90,121^141. 

Ghiorso, M. S. & Sack, R. O. (1995). Chemical mass transfer in 

magmatic processes IV. 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. 

Goff, F. (1996). Vesicle cylinders in vapor-differentiated basalt flows. 

Journal of Volcanology and Geothermal Research 97,167^185. 

Govindaraju, K. (1994). Compilation of Working Values and Sample 

Description for 383 Geostandards, Special Issue. Geostandards 

Newsletter 18,1^158. 

Greenough, D. J. & Dostal, J. (1992). Cooling history and differentiation 

of a thick North Mountain Basalt Flow of Nova Scotia, 

Canada. Bulletin of Volcanology 55,63^73. 

Gunn, B. M. (1962). Differentiation in Ferrar Dolerites, Antarctica. 

New ZealandJournal of Geology and Geophysics 5, 820^863. 

Gunn, B. M. & Warren, G. (1962). Geology of Victoria Land between 

the Mawson and Mulock Glaciers, Antarctica. New Zealand 

Geological Survey Bulletin 71, 1^157. 




Hamilton, W. (1965). Diabase Sheets of the Taylor Glacier Region 

Victoria Land, Antarctica. US Geological Survey, Professional Papers 

456-B,1^71. 

Harker, A. (1909). The Natural History of Igneous Rocks. London: 

Methuen. 

Heimann, A., Fleming, T. H., Foland, K. A. & Elliot, D. H. (1994). A 

short interval of continental flood basalt volcanism in Antarctica 

as demonstrated by 40 Ar/ 39 Ar geochronology. Earth and Planetary 

Science Letters 121, 19^41. 

Helz, T. R. (1980). Crystallization history of Kilauea Iki Lava Lake as 

seen in drill core recovered in 1967^1979. Bulletin of Volcanology 43, 

675^701. 

Helz, T. R. (1987). Differentiation behavior of Kilauea Iki Lava Lake, 

Kilauea Volcano, Hawaii: An overview of past and current work. 

In: Mysen, B. O. (ed.) Magmatic Processes: Physiochemical Principles. 

Geochemical Society, Special Publications 1, 241^258. 

Helz, T. R., Kirschenbaum, H. & Marinenko, W. J. (1989). Diapiric 

transfer of melt in Kilauea Iki Lava Lake, Hawaii: A quick, efficient 

process of igneous differentiation. Geological Society of America 

Bulletin 101, 578^594. 

Hersum, G. T., Hilpert, M. & Marsh, B. D. (2005). Permeability and 

melt flow in simulated and natural partially molten basaltic 

magmas. Earth & Planetary Science Letters 237, 798^814. 

Hersum, G. T., Marsh, D. B. & Simon, C. A. (2007). Contact partial 

melting of granitic country rock, melt segregation, and re-injection 

as dikes into Ferrar Dolerite Sills, McMurdo Dry Valleys, 

Antarctica. Journal of Petrology 48, 2125^2148. 

Hildreth, W. (1971). The Bishop Tuff; Evidence for the origin of compositional 

zonation in silicic magma chambers. In: Chapin, C. E. & 

Elston, W. E. (eds) Ash-flow tuffs: Geological Society of America Special 

Paper, 180, 43^75. 

Hildreth, W. (1979). The Bishop Tuff: Evidence for the origin of compositional 

zonation in silicic magma chambers. Geological Society of 

America, Special Papers 180, 43^75. 

Hirshmann, M. M., Asimow, P. D., Ghiorso, M. S. & Stolper, E. M. 

(1999). Calculation of peridotite partial melting from thermodynamic 

models of minerals and melts. III. Controls on isobaric 

melt production and the effect of water on melt production. 

Journal of Petrology 40,831^851. 

Holmes, A. (1931). The problem of the association of acid and basic 

rocks in central complexes. Geological Magazine 68, 241^255. 

Kyle, P. (1980). Development of heterogeneities in the subcontinental 

mantle: evidence from the Ferrar Group, Antarctica. Contributions 

to Mineralogy and Petrology 73, 89^104. 

Larsen, B. R. & Brooks, C. K. (1994). Origin and evolution of gabbroic 

pegmatites in the Skaergaard intrusion, East Greenland. Journal of 

Petrology 35, 1651^1679. 

Lindsley, D. H., Smith, D. & Haggerty, S. E. (1971). Petrography and 

mineral chemistry of a differentiated flow of Picture Gorge basalt 

near Spray, Oregon: geologic setting and mechanism of differentiation. 

Carnegie Institution of Washington Yearbook 69, 264^285. 

Marsh, B. D. (1996). Solidification fronts and magmatic evolution. 

Mineralogical Magazine 60, 5^40. 

Marsh, B. D. (2002). On bimodal differentiation by solidification front 

instability in basaltic magmas, part 1: basic mechanics. Geochimica 

et Cosmochimica Acta 66,2211^2229. 

Marsh, B. D. (2004). A magmatic mush column Rosetta Stone: the 

McMurdo dry valleys of. Antractica. EOS Transactions, American 

Geophysical Union 85, 497^502. 

Marsh, B. D., Gunnarsson, B., Congdon, R. & Carmody, R. (1991). 

Hawaiian basalt and Icelandic rhyolite: indicators of differentiation 

and partial melting. Geologische Rundschau 80, 481^510. 

1963 

McBirney, A. R. (1975). Differentiation of the Skaergaard Intrusion. 

Nature 253, 691^694. 

McKelvey, B. C. & Webb, P. N. (1962). Geological investigations in 

southern Victoria Land, Antarctica. New Zealand Journal of 

Geophysical Research 5,143^162. 

Middleton, G. V. (1970). Experimental studies related to problems of 

flysch sedimentation. In: Lajoie, J. (ed.) Flysch Sedimentation in North 

America. Geological Association of Canada, Special Papers 7,253^272. 

Minor, R. D. & Mukasa, B. S. (1997). Zircon U^Pb and hornblende 

40 Ar^ 39 Ar ages for the Dufek layered mafic intrusion, Antarctica: 

Implications for the age of the Ferrar large igneous province. 

Geochimica et Cosmochimica Acta 61, 2497^2504. 

Moore, J. G. & Evans, B. W. (1967). The role of olivine in the crystallization 

of the prehistoric Makaopuhi tholeiitic Lava Lake, 

Hawaii. Contributions to Mineralogy and Petrology 15, 202^223. 

Philpotts, A. R. & Dickson, L. D. (2000). The formation of plagioclase 

chains during convective transfer in basaltic magma. Nature 406, 

59^61. 

Philpotts, A. R. & Philpotts, D. E. (2005). Crystal-mush compaction 

in the Cohassett flood-basalt flow, Hanford, Washington. Journal of 

Volcanology and Geothermal Research 145, 192^206. 

Philpotts, A. R., Carroll, M. & Hill, M. J. (1996). Crystal-mush compaction 

and the origin of pegmatitic segregation sheets in a thick 

flood-basalt flow in the Mesosozic Hartford Basin, Connecticut. 

Journal of Petrology 37, 811^836. 

Puffer, H. J. & Horter, L. D. (1993). Origin of pegmatitic segregation 

veins within flood basalts. Geological Society of America Bulletin 105, 

738^748. 

Puffer, H. J. & Volkert, A. R. (2001). Pegmatoid and gabbroid layers in 

Jurassic Preakness and Hook Mountain Basalts, Newark Basin, 

New Jersey. Journal of Geology 109, 585^601. 

Ragland, P. C. (1989). Basic Analytical Petrology. New York: Oxford 

University Press. 

Ragland, P. C. & Arthur, J. D. (1985). Petrology of the Boyds diabase 

sheet, northern Culpeper Basin, Maryland. In: Robinson, G. R., 

Jr & Froelich, A. J. (eds) Proceedings of the Second US Geological 

Survey Workshop on the Early Mesozoic Basins of the Eastern United 

States. US Geological Survey Circular 946, 91^99. 

Richter, D. H. & Moore, J. G. (1966). Petrology of the Kilauea Iki Lava 

Lake, Hawaii. US Geological Survey, Professional Papers P 0537-B, 

B1^B26. 

Sisson, W. T. & Bacon, R. C. (1999). Gas-driven filter pressing in 

magmas. Geology 27, 613^616. 

Steiner, J. C., Walker, R. J., Warner, R. D. & Olson, T. R. (1992). A cumulus^transport^deposition 

model for the differentiation of the 

Palisades sill. In: Puffer, J. H. & Ragland, P. C. (eds) Eastern North 

American Mesozoic Magmatism. Geological Society of America, Special 

Papers 268, 193^217. 

Streckeisen, L. A. (1973). Plutonic rocks classification and nomenclature 

recommended by the IUGS Subcommission on the 

Systematics of Igneous Rocks. Geotimes 18, 26^30. 

Tait, S. & Jaupart, C. (1992). Compositional convection in a reactive 

crystalline mush and melt differentiation. Journal of Geophysical 

Research 97, 6735^6756. 

Tomkeieff, S. I. (1929). A contribution to the petrology of the Whin 

Sill. Journal of the Mineralogical Society 126, 100^120. 

Wager, L. R. & Deer, W. A. (1939). Geological investigations in East 

Greenland, Part III. The petrology of the Skaergaard intrusion, 

Kangerdlugssuaq, East Greenland. Meddelelser om GrÖnland 105, 

1^352. 

Walker, F. (1953). The pegmatitic differentiates of basic sheets. 

AmericanJournal of Science 251, 41^60. 




Wheelock, M. M. & Marsh, B. D. (1993). Silicic segregations in basaltic 

sills: bimodal differentiation. EOS Transactions, American 

Geophysical Union 74, 336^337. 

Winter,D.J.(2001).An Introduction to Igneous and Metamorphic Petrology. 

Upper Saddle River, NJ: Prentice Hall. 

Wright,L.T.&Okamura,T.R.(1977).Cooling and crystallization of tholeiitic 

basalt, 1965 Makaopuhi Lava Lake, Hawaii. US Geological Survey, 

Professional Papers 1004,1^78. 

Wright,L.T.,Peck,L.D.&Shaw,R.H.(1976).KilaueaLavaLakes: 

Natural laboratories for study of cooling, crystallization, and 

1964 

differentiation of basaltic magma. In: The geophysics of the Pacific 

Ocean basin and its margin: American Geophysical Union 19, 375^390. 

Yoder, H. S. (1973). Contemporaneous basaltic and rhyolitic magmas. 

American Mineralogist 58, 153^171. 

Zavala, K. (2005). Differentiation of silicic segregations in the Ferrar 

Dolerites, Antarctica: A geochemical study, PhD thesis, Johns Hopkins 

University, Baltimore, MD. 

Zavala, K. & Marsh, B. D. (2001). On the filling process forming 

silicic segregations. American Geophysical Union Spring Meeting, 

Washington D.C, abstract V41A-05.

Silicic Segregations of the Ferrar Dolerite Sills, Antarctica

You also want an ePaper? Increase the reach of your titles

Delete template?

Save as template?