Nb-Th-La in komatiites and basalts: constraints on komatiite petrogenesis and mantle evolution

Page 1

272 Earth and Planetary Scwnce Letters, 107 (1991) 272-289 Elsevier Science Pubhshers B.V, Amsterdam

[XLeP]

Nb-Th-La in komatiites and basalts" constraints on komatiite petrogenesis and mantle evolution

K . P . J o c h u m , N . T . A r n d t * a n d A . W . H o f m a n n

Max-Planck-Instltut fur Chemw, Postfach 3060, W-6500 Matnz, Federal Repubhc of Germany

Received December 12, 1990, rewsed and accepted July 17, 1991

ABSTRACT

New spark-source mass spectrometnc analyses of Nb, Th, REE and other trace elements m Archaean to Teruary komatlites and basalts were undertaken to test the model of Hofmann et al [1] of secular variation of the composmon of the upper mantle Most 3 4 Ga komatutes and basalts have Nb/Th between 7 and 9, values that straddle the pnmltlVe mantle ratm of - 8 Several 2 7 Ga komatntes also have Nb/Th = 7-9, but most samples of tins age have higher Nb/Th, xn the range 10-15 Cretaceous-Ternary komatntes and basalts from Gorgona Island (Colombia) have Nb/Th between 11 and 24, values that approach those of modern oceanic basalts (~ 15-34).

Although these results generally support the model of Hofmann et al, there are several comphcatmg factors (1) most of the Cretaceous-Tertaary Gorgona komatntes have Nb/Th rauos httle bagher than those of 2 7 Ga komatutes and prnmtave mantle, which suggests that low Nb/Th may be a peculiarity of komatntes and not a feature of the Archaean mantle, (2) many Archaean komatntes are depleted m both Nb and Th relative to the REE, a feature that as inconsistent with their denvataon from pnrmtwe mantle We speculate that komatntes come from a source that evolved independently from normal upper mantle, and that the depletion of Nb and Th resulted from fractlonauon of an unknown phase dunng the deep melting

Certain tholentac basalts do not show unusual Nb-Th-La fracuonatmn but nonetheless show a secular ancrease in Nb/Th. Thas vanataon may md~cate a change m upper mantle composmons resulting from progressive withdrawal of contanental crust dunng the past 3 Ga

1. Introduction

H o f m a n n et al [1] d e m o n s t r a t e d tha t the ra t ios

o f c e r t a m t race e l e m e n t s in o c e a n i c basa l t s a re

r e m a r k a b l y cons t an t , w i t h ave rage va lues tha t d i f -

fer m a r k e d l y f r o m m e a s u r e d o r e s t i m a t e d va lues

o f c o n t i n e n t a l c rus t and p r i m i t i v e man t l e . T h e

ave rage N b / U ra t io of 47 + 10 o f m i d - o c e a n r idge

basa l t s ( M O R B ) a n d ocean i c i s l and basa l t s (OIB) ,

is h tgher t han va lues e s t i m a t e d for the p r i m i t i v e

m a n t l e ( - 30) and c o n t i n e n t a l c rus ta l ( - 10). T h e

e x p l a n a t i o n p r o p o s e d for these g loba l d i f f e r ences

was seg rega t ion of c o n t i n e n t a l crus t f r o m the

m a n t l e d u r i n g the A r c h a e a n and ear ly P r o t e r o z o l c

t h r o u g h a p rocess tha t e n r i c h e d the c rus t in U

m o r e s t rong ly t h a n in N b . T h e res idua l m a n t l e

was h o m o g e n i z e d d u r i n g o r a f t e r segrega t ion , a n d

s u b s e q u e n t l y d i f f e r e n t i a t e d in to the c h e m i c a l l y and

i so top l ca l l y d i s t inc t sources of p r e s e n t M O R B a n d

OIB.

I f this m o d e l is co r r ec t a n d the g loba l changes

in N b / U resul t f r o m the f o r m a t i o n o f c o n t i n e n t a l

c rus t d u r i n g the A r c h a e a n or ear ly P ro te rozo lc ,

t hen the m a n t l e tha t ex i s ted p r io r to thas p e r i o d

w o u l d h a v e h a d ra t ios of these c r i t i ca l t race ele-

m e n t s s imi la r to t hose in p r imi t ive , u n d i f f e r e n t i -

a t ed man t l e . T o tes t t tus hypo thes i s , we used

s p a r k - s o u r c e mass s p e c t r o m e t r y ( S S M S ) to ana lyse

a la rge n u m b e r o f t r ace e l e m e n t s in k o m a t i i t e s and

basa l t s f r o m f ive A r c h a e a n g r e e n s t o n e bel ts , f r o m

a P r o t e r o z o l c v o l c a n i c bel t , a n d f r o m C r e t a c e o u s -

T e r t i a r y G o r g o n a I s land .

2. Samples and analytical techniques

* Present address InStltUt de Grologle, Unlverslt6 de Rennes, 35042 Rennes, France

T a b l e s 1 a n d 2 c o n t a i n the l o c a t i o n s and the

c h e m i c a l c o m p o s i t i o n s o f the samples . T h e

0012-821X/91/$03 50 © 1991 - Elsevier Science Publishers B V All rights reserved

Page 2

K O M A T I I T E P E T R O G E N E S I S A N D M A N T L E E V O L U T I O N 273

Archaean rocks fall into two age groups: - 3 . 4 Ga old komatiites and basalts from the Onver- wacht Group in South Africa and the Pilbara Craton in Australia; and 2.7 Ga komatiites and basalts from the Abitibl, Wawa and Swayze belts in Canada, the Norseman-Wl luna belt in Australia, and the Belingwe belt in Zimbabwe. The Proterozoic samples are basaltic, - 1 . 9 Ga old, and come from the Ottawa Islands in Hudson Bay, Canada. The youngest samples are Creta- ceous-Ter t iary komatlites and basalts from Gorgona Island, Colombia.

With the exception of two olivine cumulates from the Onverwacht Group, all the komatfites are splnifex-textured and probably had composi- tions similar to those of the magmas from which they crystallized. The basalts are fine-grained with few phenocrysts, and probably also are noncumu- late. The degree of alteration varies widely. Sam- ples from Gorgona Island and the Belingwe belt are virtually unmetamorphosed and represent the freshest komatiites available; the Kambalda rocks, on the other hand, were subjected to upper greenschist facies metamorphism and carbonate metasomatism. The degree of alteration in other samples hes between these two extremes. More detailed accounts of the petrology and chenustry are found in the references hsted in Tables 1 and 2.

Major and trace element analyses are given m Tables 1 and 2. The major element data are from other publications or other sources, as indicated in the table captions. Trace elements in all samples were analysed by spark-source mass spectrometry (SSMS) using the method described by reference [2].

3. Results

Trace element data normalized using the primi- tive mantle values of [3] are presented in Fig. 1. Because all samples are altered to a greater or lesser extent, the concentrations of mobile ele- ments such as the alkalis, alkaline-earths, U and Pb, do not necessarily correspond to concentra- tions m the original magmas. Although reported an Tables 1 and 2, these elements are not shown in most diagrams, and will be treated with caution in the following discussion. Thorium is less mobile than U, and can be used in its place because the

two elements behave-similarly during magmatlc processes.

3.1 ~ 3.4 Ga komatutes and basalts from the Onverwacht Group and the Pllbara Craton

Two komatiltes (8241 and 8243) have relatively flat patterns (Figs. la and lb). The other two (5019 and 5031) and all three basalts show mod- erate enrichment of the more incompatible ele- ments and depletlon of the HREE and Y. These features are typical of Al-depleted komatlites and associated basalts [4]. An unexpected feature of most samples is the low concentrations of Th and Nb, two of the most incompatible elements, which are markedly lower than the normahzed con- centrations of La, the next most incompatible element (Fig. la). Nb-Th ratios vary within a small range that straddles the primitive mantle value of - 8. Two samples (8243 and 5038) also show relative enrichment of the high-field strength elements, Zr, Hf and Y; and one (5031) has a pronounced negative Ce anomaly.

Three M-depleted komatiit~c basalts from the Pllbara Craton (Fig. lc) have relatwely flat pat- terns w~th slight enrichment of the more incom- patible elements, again with the unusual depletion of Th and Nb (Fig. lc). In two samples Zr, Hf and Y are relatively enriched.

3.2. 2. 7 Ga komattttes and basalts from the Abltlbl, Swayze and Wawa belts, Canada

Komatiltes from Munro Township (C18) and the Alexo area (M664) belong to the Al-unde- pleted suite and have chondrltlC ratios of A1203/TiO 2 and HREE. Incompatible element patterns in Figs. ld and le are relatively smooth, without conspicuous positive or negative anoma- lies. The more incompatible elements are mod- erately depleted, as is typical for rocks of this type. Komatiltic basalts from Munro Township (C6 and C126) and Newton Township (NEW20, Fig. le) have similar patterns but with less pro- nounced depletion of the incompatible elements.

Patterns in tholefitic basalts are variable. Sam- ple C1 from Munro Township shows relattve en- richment of moderately incompatible elements, a maxamum at Pr, and slight depletion of Th, Nb and La; sample C31 from Munro Township and

Page 3

100,

,

i i

i i

i i

' I

i i

i i

i ~

i i

i i

i

Onv

erw

acht

~

--~

---

- 82

~,I

¢~

/

basa

lts

I --

--0

--

5031

I

Q

" 5°

'982

--~

~

rt

27 1

32

0

10

" 35

0

29

6

1 i

i i

i i

i f

i =

i i

i 1

i i

i i

i L

i

-o

IOnv

erw

acht

] -

-~

--

50

38

b

Nm

Lkom

atiit

es

J ~

5080

D~

-.

..

~

- 50

77

1°o

o o

n

10

i i

i i

l t

i i

i i

L i

I i

i i

i i

i i

pd92

pi 2

6a

58

15

62

18

..

..

..

..

..

.

G' .

..

..

..

.

1 T

h

Nb

L

a

Ce

P

r N

d S

m

Zr

Hf

Eu

d

Dy

Ho

Y

E

r Y

b

Lu

MgO

EN

d

100

E

o c E

E

n

i i

f I

~ i

i T

i i

i ,

i i

i ~

f ,

r

C1

d ~

¢ C3

~

C6

C12

6 ~0

-

._

,.

c.

-~

_~

~-

-=

--

~_

..

,.

•

c~a

- ~

~ 55

35

-

~ 75

2

2

73

2

0

1 27

2 25

10

I I

I I

I I

I I

i I

L I

I I

I I

1 I

I

e /~

-"-C

]-~

O--

-'C

) .~

D "

-''O

~

78

24

14

53

0

~ ~

~H

D~

~L

~m

2

96

25

1

NE

W20

•

M66

4

I t

I ]

I I

] I

I l

I l

I l

I I

I I

I ~

I

f

• 23

23

02

5

Z2

• W

A59

--

-~]-

--

ZA1

-"~

O--

Z

V85

~

WA

75

T,h

Jb

i L

, ,

I i

i i

L L

i L

I I

yI

I L

L N

L

a

Ce

P

r N

d S

m

Zr

Hf

Eu

G

d

Tb

D

y H

o

Y

Er

b L

u

MgO

EN

d

Fig

1

Tra

ce

elem

ent

dat

a,

no

rmah

zed

u

sin

g

pn

rmtl

ve

man

tle

val

ues

o

f [3

],

for

ko

mam

tes

and

b

asal

ts

,-o

C

>

Page 4

K O M A T I l T E P E T R O G E N E S I S A N D M A N T L E E V O L U T I O N 275

N

E o

E

Q_

106 F Q . . . . . . . . . . . . . . '

I \ g . K,, f \ ~ - - c - - 4783 ~_ \ J ~ • 4774

10~ t/ ~ _ . ~ ' 1 ' 4 °

I~I I I I l I I I I l I i I I I I I I i I

1 8 0

1 2 7 0 7

G,2

I I l I f l I I I I I I I [ I I I I I

1 0 -

8 5 9 7 / . - lal

~ ~ ' " 1 ~ 190 - - ~ - -- _ ~.r-,.-,~18 3

F ,,At' I - GOR117 I [K/ ~ 1----o-- Go.,67 I [~ I Island I I -----~- G OR1891

' ' I ~ GOR, E L_-so-- Go.,60j 1 1 1 I I I I I I I I 1 I I I I I [ I I [ I

Th Nb La Co Pr Nd Sm Zr HI Eu Gd Tb Dy Ho Y Er Yb Lu MgO E;N,

F i g 1 ( c o n t i n u e d )

WA75 from the Wawa belt (Fig. If) have flat patterns with small negative Nb anomalies; and sample NEW91 from Newton Township and WA59 from the Wawa belt have flat HREE, en- riched LREE and pronounced depletion of Th and Nb.

3.3. 2.7 Ga komatutes from Zwlshavane, Behngwe greenstone belt, Zimbabwe

Three samples of MgO-poor, Al-undepleted komatnte have very similar, relatively flat patterns with shght depletion of the more incompatible elements. All three show depletion of Nb and, to a lesser extent, Th (Fig. lf).

3.4. 2.7 Ga komatHtes and basalts from Kambalda, Norseman- Wiluna belt, Austraha

Two komatiites (477.4 and 478.3) have smooth patterns similar to those of komatutes from the Abltlbl belt, with moderate depletion of the more incompatible elements (Fig. lg). A third sample (Al l40) has a flatter pattern with a shght negative Nb anomaly but no corresponding depletion of Th. The tholeiltic basalt (KA1) has a smooth pattern, with very slight depletion of the more incompatible elements, and no negative Nb or Th anomalies. The pattern in sample C592 from the overlying Paringa Basalt Formation stands m marked contrast to those described before. The

Page 5

276 K P J O C H U M ET A L

< [...

~_~o~o~9 o o O o ~ _

Page 6

AI2

03

/TIO

2 11

17

11

11

24

21

20

23

20

21

20

18

18

La

/Sm

1

6 1

6 4

1 2.

9 1

3 0

8 1

1 0

8 0

9 1.

3 1.

2 1

3 0

5 0

5

Nb

/U

26

30

13

17

25

38

46

32

51

23

21

19

44

44

Nb

/Th

8

11

7 7

6 15

14

13

15

8

8 8

16

24

Nb

/La

0

7

08

0

3

04

0

8

1 1

08

0

9

1 0

08

0

6

06

0

7

07

Z

r/S

m

26

30

25

29

31

26

25

22

26

25

26

23

23

23

eNd

3 8

4 0

1.7

2.5

2 5

2 4

2 5

Sou

rce

a =

G

Gru

au,

un

pu

bh

shed

dat

a, b

= A

rnd

t an

d J

enn

er [

9],

c =

Arn

dt

and

Les

her

(m

pre

p),

d =

Arn

dt

and

Nes

blt

t [3

1],

e =

Arn

dt

[32]

, f

= N

lsb

et e

t al

[3

3],

g =

N.T

A

rnd

t, u

np

ub

hsh

ed d

ata,

h =

Ech

ever

ria

and

Att

ken

[34

,35]

TA

BL

E 2

Maj

or

and

tra

ce e

lem

ent

anal

yse

s of

bas

alts

Lo

caU

on

O

nv

erw

ach

t P

db

ara

Kam

bal

da

New

ton

M

un

ro

Sam

ple

No

50

38

5080

50

77

92

56A

2

6A

K

A1

C

59

2

NE

W9

1

NE

W2

0

C1

C1

26

C

31

C6

Sou

rce

x 1

1 j

j .1

b b

k k

1 1

1 1

SIO

2 51

53

55 9

6 53

77

51 5

8 49

94

50 9

9 50

58

54 1

7 49

.52

48 5

5 50

14

51 0

2 52

05

50.5

4

TIO

2 1

67

0 84

0

65

0.71

1

93

1 39

0.

81

0.93

0

92

0 68

1

45

0 67

0.

88

0 82

A

1203

13

07

10 1

5 7

52

7 59

14

11

13.6

3 15

64

14 0

6 16

28

10 6

4 13

04

12.2

0 14

.43

14 0

8

Fe2

03

15 1

6 10

26

12.3

9 10

97

15 9

1 15

91

12 4

3 12

06

12 6

2 12

87

15 7

1 11

60

11 1

8 13

90

Mn

O

0 26

0

17

0 17

0

22

0.24

0

25

0 18

0

19

0.25

0

20

0 26

0

19

0 17

0.

22

Mg

O

4 26

9

59

11 5

7 12

.35

5 84

6.

20

5.96

5.

53

7 82

14

46

5 52

12

20

7.51

7.

27

CaO

9.

44

9.60

10

.77

14 9

8 9

15

9 66

11

24

9 57

9

83

10.5

5 11

.71

8.52

10

54

9 51

N

a2

0

42

5

31

5

30

3

1.47

2

47

1

48

3

62

2

64

1

97

1

76

1

27

3

39

3

09

3.

22

K2

0

01

5

01

7

00

5

0.08

0

26

0

35

0

60

0

75

0

56

0

23

0

86

0

09

0

09

0.

27

P205

0

20

0 11

0.

08

0 04

0

15

0 13

0

00

0 10

0

25

0 06

0.

14

0 07

0.

09

0.07

L

OI

05

2

0

1 0

01

0

5

1 9

26

2

8

20

Page 7

TA

BL

E 2

(co

nti

nu

ed)

Lo

cati

on

O

nv

erw

ach

t

Sam

ple

No

50

38

5080

Pd

bar

a

5077

92

5

6A

2

6A

Kam

bal

da

KA

1

C5

92

N e

wto

n

Mu

nro

NE

W9

1

NE

W2

0

C1

C1

26

C

31

C6

Cs

01

0

0

00

R

b 2.

0 0

9 0

7 B

a 17

8 0

16 3

11

8

U

0 58

0 0

077

0 10

3

Pb

37

0

5

05

S

r 27

6 0

30 7

32

.8

Th

2 35

0 0.

334

0.37

9 N

b

17 1

0 2.

97

3 26

La

19.7

0 6

22

3 65

Ce

43 4

0 12

70

10 3

0

Pr

6 27

1

80

1 56

N

d

24 3

0 7.

82

7 25

Sm

5

30

2.10

2

04

Zr

163

0 57

.8

52 2

Hf

4 25

1.

47

1 51

Eu

1 43

0

66

0 66

Gd

5

38

2 68

2

55

Tb

0.79

0

43

0 40

Dy

5.

41

3 05

2

70

Ho

1

13

06

7

0.63

Y

39 2

18

4

16 7

E

r 3

33

1.83

1.

63

Tm

0

45

0.29

0

24

Yb

3

29

1.78

1

38

Lu

0 41

0.

24

0 23

0.1

3 8

0.7

0 5

1 1

24 2

23

.5

3 6

113

0 97

.6

79.8

11

4 0

279

0 68

7

47 4

21

4 0

0 07

9 0.

198

0 18

2 0

049

1 60

0 0

073

0.04

3 0

177

10

3

4

19

4

4

11

17

.3

15

63

0

17

40

1

48

0

170.

0 1

20

0

14

90

76

.5

13

70

0 31

0 0

731

0 72

3 0.

197

6 32

0 0

273

0.12

0 0.

582

2 22

6

25

4 13

2.

13

5 08

1

83

1 30

6.

11

28

6

11

00

5.

78

25

2

14

00

3

56

1

69

6.

98

7 14

22

.30

13 7

0 6

80

25 5

0 8

41

4 75

18

40

1 15

3.

00

1 95

1

18

3 36

1

28

0 79

3

01

5 50

13

90

9 00

6

00

12 3

0 5

86

3.95

13

90

1 75

4

44

2 98

1

99

2 65

1

75

1.36

3

89

40.2

14

1 0

117

0 54

0

89.3

50

8

37 5

10

2.0

1 07

4

01

3 11

1

47

2 39

1.

46

1 07

2.

88

0 67

1

66

1.08

0.

71

0.89

0

67

0 52

1

32

2 13

6

02

3.96

2

39

2 33

1

84

4 13

0 36

0

94

0.69

0

46

0 50

0

40

0.33

0

68

2 21

6

48

4 92

3

01

3 22

2

86

2 22

3

98

0 52

1

50

1 05

0

71

0 66

0

67

0 52

0

94

15 6

47

0

33 1

22

3

25 5

18

5

15.4

29

2

1 41

4

90

3 29

2

23

1 87

2

00

1.58

2

67

0 23

0

74

0 33

0

30

0 23

0.

40

1 55

4

52

3 04

2

03

1 99

2.

04

1 35

2.

31

0 19

0

72

0 45

0.

33

0 29

0.

31

0 22

0

34

03

0.

4 0

2

15

2

3

62

107.

0 68

3

53.8

0

04

2

01

15

0

05

9

07

0

8

06

73 1

98

1

101

0 0.

132

0.40

7 0

155

1.55

3.

12

2 16

1 58

4

57

2 12

4 77

11

70

5.70

0.

70

1 63

0.

94

3.57

7.

48

4 88

1 31

2

39

1 61

37 6

61

5

42 3

0.88

1

45

1 20

0.50

0

81

0.63

1 61

2

86

2.52

0 27

0

49

0.47

2

00

3

33

3

07

0 48

0

77

0 69

16 7

22

.2

19 4

1.

49

2.14

1

90

0.22

0

34

0 28

1.53

2.

20

2.04

0

20

0 31

0.

28

A1

20

3/T

IO2

8

12

12

11

7 10

19

15

L

a/S

m

3 7

3 0

1 8

1 6

2 5

1 9

1 3

5 3

Nb

/U

29

38

32

28

32

23

44

3

Nb

/Th

7

9 9

7 9

6 11

0

8 N

b/L

a

0 9

0.5

0 9

0.8

0 6

0 7

0 8

0 4

Zr/

Sm

31

28

26

23

32

39

27

34

end

--3

5

--0

.2

02

0.

0 1

5 1.

8 2.

7 --

1 4

18

16

9 2

0

12

1

8

25

31

35

7 11

10

0.5

08

0

9

29

28

26

2.4

30

3

5

18

16

17

12

1

9

13

37

27

37

12

8 14

10

0

7

1.0

29

26

26

22

2

0

Page 8

KOMATIITE PETROGENESIS AND MANTLE EVOLUTION

TABLE 2 (continued)

279

LocaUon Wawa Ottawa Islands Gorgona

Sample No WA59 WA75 G6 G37B G12 GOR167 GOR117 GOR54

Source m m n n n h h S l O 2 51 40 46.20 50 60 46.90 50 40 T102 0 86 0 72 0.80 0.67 0 77 AI203 12 20 9 70 12 00 13 90 13 81 Fe203 12 78 12.30 13 09 12 68 12 62 MnO 0.20 0.23 MgO 11.30 18 00 12.70 13 20 8 65 CaO 11 40 11 40 11 10 12 00 12 39 Na20 1 51 0 98 1 40 1 16 2 43 K20 0.08 0 08 0 20 0 03 0 02

P205 0 05 0 06 LOI 0 6 3 1 1 4

Cs 01 19 01 0.1 02 0.0 00 00 Rb 73 262 12 13 43 10 03 09 Ba 81 6 77 9 27 3 10 5 244 5 2 21 0 32 9 U 0 056 0.091 0.159 0 048 0.272 0 012 0.034 0 241 Pb 1 1 09 1.0 06 0.8 02 121 06 Sr 133 0 124 0 77 6 116 0 71 1 65 6 107 0 148.0 Th 0 214 0 365 0 731 0 149 0 640 0 031 0 082 0.849 Nb 1.59 1 90 2.58 1 89 2 30 0 45 0 87 10 70 La 2 92 3 03 3 99 2 07 3 64 0 73 1 02 6 49 Ce 8 53 8 82 9 33 5.58 10 20 2.46 2 99 17 90 Pr 1 07 1 07 1 36 0.87 1 12 0 48 0 60 2 42 Nd 4.79 4 93 6 13 4 31 4 82 2 90 3 53 9 51 Sm 1 63 1.55 1 69 1 35 1 51 1 44 1 74 2 57 Zr 45 2 48 2 44.6 32 6 44.5 41 0 39 4 91 9 Hf 1 22 1 14 1 30 1 02 1 12 1 09 1 05 2.34 Eu 0 60 0 57 0 61 0 60 0 48 0.60 0 63 0.95 Gd 2 33 2.19 2.12 1 80 1 54 1 82 2 07 3 38 Tb 0.40 0 36 0 36 0 31 0 27 0 34 0 41 0 50 Dy 2.89 2 73 2 54 2 19 1 91 2 50 3 13 3 42 Ho 0 72 0 69 0 56 0 50 0 46 0.56 0 75 0 74 Y 22 1 18 9 15 7 13.6 17 1 17.1 22 8 24 3 Er 2 18 2 13 1 58 1 34 1 31 1 68 2 18 2 10 Tm 0 26 0 22 0 26 0 37 Yb 2 24 2 02 1 51 1 29 1.33 1 49 2.20 1 82 Lu 0 33 0.29 0.23 0 21 0.16 0 23 0 32 0 23

A1203/TIO 2 14 14 15 18 16 La /Sm 1 8 20 24 1 5 24 05 06 25 N b / U 28 21 16 40 8 38 25 44 N b / T h 7 5 4 13 4 15 11 13 N b / L a 0 5 0 6 0 6 0.9 0 6 0.6 0 9 1 6 Z r / S m 28 31 26 24 29 28 23 36

end 2 5 2.3 0.6 0 7 9 7

Source b = Arndt and Jenner [9], h = Echeverria and Attken [34,35], l = Jahn et al [36], j = Gruau et al [37,38]; k = Cattell and Arndt [39,40], 1 = Arndt and Nesbltt [41], m = E Hegner, unpubhshed data, n = Arndt et al [11]

i n c o m p a t i b l e e l e m e n t s , i n c l u d i n g T h , a r e s t r o n g l y

e n r i c h e d , t h e r e is a l a r g e n e g a t i v e N b a n o m a l y ,

a n d s m a l l p o s i t i v e a n o m a h e s f o r t h e o t h e r H F S E .

3 5. - 1.9 Ga basalts f rom Ottawa Islands, Hudson Bay, Canada

T h r e e b a s a l t s , o n e a p i l l o w e d m a g n e s i a n t ho l e l -

i te, t h e o t h e r t w o s p i n i f e x - t e x t u r e d k o m a t f i t l C

Page 9

2 8 0 K P J O C H U M E T A L

100

D

Z

10

~2

Z

o

Prtmltlve [ ] manlle • • A []

O

÷ • • 3 4 G a 7- • 2 7 Ga K o m a t l t t e s

[ ~ [ ] GorgonaJ

[3 3 4 G a 7- b A 2 7 Ga B a s a l t s

O 1 9 G a j [ ] Gorgona ~ \ ~ ~ ....

34Ga ~ G 6 G12 samples [ ~

1 ÷ • C592

i i i i i i 1 1 1 i i i i i i i i i i i L i i i i

1 1 10 100 N b ( p p m )

Fig 2 N b / U and N b / T h plotted agmnst Nb concentration

The primitive mantle value ( N b / U = 30, N b / T h = 8, Nb = 0 6 ppm) is from [3], the field of modern oceanic basalts from [1]

and the composluon of continental crust (CC, N b / U =12,

N b / T h = 3, Nb = 11 ppm) is from [5] The dashed field enclo-

ses the compositions of 3 4 Ga old komatlltes and basalts

basalts, have patterns analogous to those of the enriched and depleted rocks at Kambalda (Fig. lh). The magnesian tholente (G37b) has a rela- tively flat pattern with slight depletion of Th, Nb and La; the spinifex basalts (G6 and G12) show enrichment of all the Incompatible elements, in- cluding Th, and have negative Nb anomahes.

3 6 Cretaceous-Tertmry komatlttes from Gorgona Island, Colombm

Three komatlltes and a komatilUC basalt from Gorgona Island show moderate to extreme deple- tion of the more Incompatible elements (Fig li). The patterns are smooth, with no obvious posiUve or negative anomalies. A tholentic basalt (GOR54) has moderate ennchment of the incompatible ele- ments with a distinct positive Nb anomaly.

4. Nb / U and Nb / Th relationships

In Fig. 2a we plot our data in terms of N b / U vs. Nb, the type of diagram used by Hofmann et

al. [1], and compare them with measured composi- tions of modern oceanic basalts and with esti- mated composluons of prlmitwe mantle and con- tmental crust. The N b / U data for komatntes and related basalts scatter widely, from about 10 to 50, but broadly coincide with the estimated composi- tion of primitive mantle. With few exceptions the data lie below the horizontal trend defined by the modern oceanic basalts. This behavlour might be taken to support the imtlal hypothesis that the Precambrian lavas were derived from undifferenti- ated mantle, but in view of the well-known mobil- Ity of U and the large scatter m the data from Precambrian rocks, the result is inconclusive at best.

In Fig. 2b, Th is substituted for U. The di- agram is analogous to the N b / U vs. U diagram, but because Th is more mcompauble than Nb [3,5], Nb-Th ratios in the oceanic basalts decrease with increasing Nb content: enriched OIB have lower N b / T h than depleted N-MORB. Nonethe- less these basalts define a linear trend that lies distinctly above the compositions of primitive mantle and continental crust, and the relative positions of the three fields are comparable to those in the N b / U vs. Nb diagram.

The Nb-Th rauos in the Precambrian komatl- ites and basalts are distinctly lower, at a given Nb concentration, than the Nb-Th ratios in modern oceamc basalts In a majority of Precambrlan samples, Nb-Th ratios range between 5 and 15 This range partially overlaps the range measured in modern oceanic basalts ( - 5 to 40), but because the lowest values in the modern basalts are in enriched OIB, which have far higher Nb contents than the Precambrlan rocks, the fields are almost completely separate. The range of N b / T h values in the komatutes and associated basalts overlaps the estimate for primative mantle. Three basalts, G6 and G12 from the Ottawa Islands, and C592 from Kambalda, plot well below the other data. The N b / T h of the Kambalda basalt (0.8) is less than that in many continental crustal esumates e.g [5].

Komatiites and basalts from Gorgona Island have very low Nb contents and Nb-Th ratios between 11 and 24. Three samples fall within the field of Precambrlan rocks; a fourth plots between the fields of Precambrmn and modern rocks; and the fifth, a tholentlc basalt, plots within the OIB field.

Page 10

K O M A T I I T E P E T R O G E N E S I S A N D M A N T L E E V O L U T I O N 2 8 1

Closer mspecUon of the data reveals further regularities. 3.4 Ga old komatiltes and basalts from Barberton and Pdbara have a relatwely nar- row range of N b / T h , from about 6 to 11. Unhke in modern rocks, the Nb-Th ratio does not de- crease with increasing Nb content. Most 2 7 Ga lavas have slightly higher N b / T h , from - 10 to 15, but several others have values similar to those of the 3.4 Ga rocks. Again there is no correlation between N b / T h and Nb content.

An iniual assessment of these results m~ght support the concept of an essentxally undifferenU- ated mantle m the late Archaean and Proterozoic, as proposed by Hofmann et al [1]. ParUcularly xmpresswe is the clustering of Nb-Th ratios in the Archaean rocks about the prlrmtive mantle value, and the overall contrast between the ratios m Precambrlan and modern rocks. However, several aspects of the data are inconsistent with thas inter- pretauon. Firstly, the Nb-Th ratios of three out of four Cretaceous-Tertmry komatiltxC lavas from Gorgona Island fall wtthin the field of Archaean komatntes and plot distinctly below the trend defined by the modern oceanic basalts The Gorgona lavas are strongly depleted in incompati- ble trace elements and have overall geochetmcal and ~sotop~c characterlsUcs similar to those of modern MORB. The fact that they plot together with Archaean komatfites raises the possibility that differences in Nb-Th ratios between Pre- cambrian and modern volcanic rocks are not due to secular variation m the composition of their mantle sources, but may be related to the petro- logical differences between komatntes and basalts. The relatively low N b / T h of both Precambrlan and Cretaceous-Tertmry komatntes, rather than being a function of the age of the rocks, might arise from pecuharltles in the processes that gener- ate ultramafic magmas.

A second problem with the concept that the Precambnan lavas come from undifferentiated mantle becomes apparent when the overall distri- bution of data in Fig. 2 is compared with trace element patterns of individual samples, as il- lustrated in Fig. 1, and with their Nd isotopic compositions. Although many komatiites have Nb-Th ratios similar to the prirmtive mantle value, both elements are depleted relative to the REE. For example, the Barberton komatfite 5031 has N b / T h of 7.2, only slightly lower than the prima-

l o

z

Pnm,t,ve N - M O R B mantle /

i ,-, • ~,~ ~4o so[ 5019 A • 3 4 G a

5031 • 2 7 Ga ~ ' K o m a t l l t e s [ ] Gorgon~]

c] 3 4 G a ] 0 1 A 2 7 Ga r- Basal ts

01 1 10 0 19Ga / N b ( p p m ) [] Gorgona.J

Fig 3 Nb/La vs Nb Sources of data as m Fig 2 Mlxtng hne connects a magma from prlmmve mantle with continental crust [5] Certain komatntes plot well below this hne, indicat- ing that their low Nb contents are not due to crustal con-

tarmnaUon

txve mantle value of 8, but its N b / L a is 0.30 (Table 1, Figs. l a and 3), considerably lower than the primltwe value of - 1 . In this sample the LREE are relatively enriched and the anomalous concentrauons of Nb and Th are particularly con- spicuous: normahzed concentraUons show a steady increase from Lu to La (except for the Ce anomaly, which is probably secondary), then a dramauc drop at Nb and Th (Fig. la). This marked depar- ture from primitive element raUos is a clear sign that the apparently primitwe N b / T h must be interpreted with caution.

5. Processes that could have influenced N b / T h in komatiites and basalts

The foregoing discussion provides ample indi- cation that interpretation of the trace element data from the komatlite and basalts is not straightfor- ward. The simple story of secular variaUon in mantle composlUons is neither supported nor con- tradicted; rather it seems that several different processes may have influenced the relative con- centraUons of incompatible elements. Posslblhues include: (a) element mobility during alteration; (b) crustal ,contamination; (c) the presence of sources with different composit ions in the Archaean mantle, a n d / o r variations in the condi- tions of melting.

5.1. Element mobthty Although the trace elements we have used in

the discussion are normally considered immobile

Page 11

282 K P J O C H U M E T A L

durmg hydrothermal alteration and low-grade metamorphism, the types of fractionation we ob- serve are highly unusual, and secondary effects cannot be ruled out. However, we believe such effects are not ~mportant in the present situation for the following reasons.

(a) The relative abundances of REE, HFSE and Th do not correlate with degree of alteration. Negatwe Nb and Th anomalies are as large in the very fresh Zwishavane komatntes as in the more strongly metamorphosed (upper greenschlst facxes) Barberton and Pdbara lavas: smooth patterns without anomalies characterize both the relatwely fresh Alexo komatiites and the highly altered Kambalda samples.

(b) There is httle evidence of systematic ennch- ment or depletion of groups of elements that should behave coherently during alteration. En- richment of REE relative to the HFS elements and Th, a possible cause of low N b / L a and T h / L a raUos, can be ruled out by the relatwely high concentrations of Zr and Hf xn many of the sam- ples that display negatwe Nb and Th anomalies. For example, sample PIL26a, which is plotted in Fig. lc, has a large negative Nb anomaly ( N b / L a ) ~ = 0.7, but moderate posture Zr and Hf anomalies (Zr/Sm)N = 1 6.

5 2 Contamlnatton wtth continental crust

Komatiltes are prone to contamination by crustal rocks durmg ascent and eruption [6,7]. Such contarmnatlon increases the abundances of incompatible elements such as the LREE and Th but has a smaller effect on Nb, Ta and TI The result is a magma relatively enriched In LREE and Th, with negative Nb, Ta and T1 anomalies. These are the characteristics of three of the basalts we analysed: sample C592 from Kambalda (Fig. lg), and G6 and G12 from the Ottawa Islands (Fig. lh). In both regions there is independent evidence of contarmnatlon with older crust. Chauvel et al [8] showed that the eyd(X ) value of the Kambalda basalt was -1 .4 , lower than that of associated komatmes (eNd(T) = -t-2 to + 5); Arndt and Jenner [9] demonstrated that the overall major and trace element compositions of Pannga basalts was con- sistent with their having formed by contamination of komatnte, and Compston et al. [10] found zircon xenocrysts from a crustal contaminant in

100

x:

z

Ohvme crystalhzatton or accumulabon

100

1 10 N b ( p p m )

b

CC

1 20

. . . . . . . . [0592T~Nd = ¼1 4] . . . . . .

1 10 [Lo/Sm]N

Fig 4 Nb/Th vs Nb and Nb/Th vs [La/Sm]N diagrams showing the compositions of komatntes and basalts from Kambalda and the Ottawa Islands, and mixing lines between komatntlC magmas and crustal contaminants The end value of sample G37b Ls taken as that of sample G47 from the same flow [42] For the Ottawa Island basalts (G6, G12, G37b), the contarmnant is Taylor and McLennan's [5] upper continental crust, for the Kambalda samples, the contaminant has lower Nb and Nb/Th Numbers by tick marks give percentage

contaminant

samples in the same unit. Arndt et al. [11] pre- sented analogous geochemical data and arguments for basalts from the Ottawa Islands.

The effects of crustal contamination on N b / T h relationships are illustrated in Fig. 4. In the N b / T h vs. Nb dtagram (Fig. 4a), a mixang line between komatiite and continental crustal material has a steep upper limb Because of the large differences m Th concentrattons and N b / T h ratios of komati- lte and crustal material, additton of a small amount of contaminant has a great effect on the N b / T h of the resultant magma. Basalts from both regions lie on plausible mixing lines; the Ottawa Islands basalt (G6) on one between the Mg-basalt G37b and crustal material wtth a composltton like that estimated by Taylor and McLennan [5], and the Kambalda basalt on a line between Kambalda komatiite and a contaminant wxth simalar overall composition but significantly lower N b / T h . The material that contaminated the Kambalda lavas

Page 12

KOMATIITE PETROGENES1S AND MANTLE EVOLUTION 283

apparently had N b / T h lower than Taylor and McLennan's [5] crustal estimate, but similar to that of many granitoids and granulitic rocks.

Further inspection of the data in Fig. 4 suggests that contamination rmght be an explanaUon for not only the extremely low N b / T h in the two basalts discussed above, but also for the more moderate N b / T h in certain other Precambnan komatiites and basalts. For example, the Kam- balda komatilte Al140 has lower N b / T h than the other two komatiites (477.4 and 478.3) and plots m Fig. 4 close to mixing lines connecting a hypo- thetxcal parental komatiite and contmental crust As can be seen In Fig. lg, the chemical character- istics of this sample are generally consistent with the contamination interpretaUon: it has a shght negative Nb anomaly, and LREE and Th are slightly less depleted than in the other two komatl- ites. The generally lower level of incompatible element abundances of sample Al140, although inconsistent with simple contaminatxon, can be explained by differing levels of post-eruptxon ohvlne fractxonation or accumulation in the vari- ous komatute samples. When (La/Sm)N, which is insensitive to ohwne accumulation, is plotted against N b / T h (F~g. 4b), the position of the sam- ple is closer to the mixing line.

Returning to Fig. 4a, it can be seen that the amount of contaminant calculated for the komati- lte A l l40 is less than 1%. We point this out to emphasize that N b / T h is an extremely sensitive indicator of crustal contamination which, when used m conjunction with Nd isotopic data, is a powerful d i scnmmant between contaminated komatiltes and basalts and those that retain their original trace element abundances.

It must then be asked whether the overall dif- ferences in N b / T h of the Precambrlan and mod- ern volcamc rocks could not also be explained by crustal contarmnatlon. Perhaps all the komatlites and basalts that we analyzed, the Gorgona lavas included, simply are variably contarmnated de- rlvauves of parental magmas with higher N b / T h ratios. This possibility can be disrmssed for the following reasons. First, the Nd and Pb isotopic composmons of the Gorgona komatiites are so s~milar to those of MORB that they preclude the addition of even the 0.5% contaminant required by the mixing model (B. Duprr , pers. commun., and [12]). Second, the N b / L a relationships il-

• 34Ga ] • 27Ga }'Korea iites

(Nb)N I k'*'A Gorgona J / ~ J ~f34 Ga "] / \ L27Ga / _ Its

(Th) N (La) N

Fig 5 La-Nb-Th diagram All data are normahzed to pnrmtwe mantle values [3] Insets show schemauc Th-Nb-La spectra for OIB (oceamc island basalt), MORB (mld-oceamc ridge basalt), a basalt contarmnated w~th continental crust, and a komatnte

depleted in both Th and Nb

lustrated in Fig. 3 are unlike those predicted from the mixing model: the compositions of most komatiltes plot well to the left of a mixing hne between MORB or primitive mantle and continen- tal crust, and some komatntes (e.g. 5031, 5019) have N b / L a lower than either end-member. Fi- nally, and most significantly, the outstanding fea- ture of the trace element patterns of many of the komatutes with low N b / T h is their relatively low abundances of both Nb and Th. Whereas crustal contalmnatlon produces a negauve Nb anomaly but enriches Th, the komatlites, with few excep- tions (e.g. sample Al140) either have relatively smooth incompatible element spectra (e.g. sample M664), or are depleted in both Nb and Th relative to La (e.g. samples PIL56a, 5031, 5019, and most other 3.4 Ga komatiites and basalts).

5 3 Nb and Th depletlon in komatt t tes and associ- ated basalts

The contrasting trace element patterns of crust-contanunated basalts and uncontaminated komatntes are well illustrated in Fig. 5, in which abundances of Nb, Th and La, normalized to primitive mantle values [3], are plotted. (In intro- ducing this diagram, it ~s not our intention that it becomes a tectonic discrimination diagram of the type developed by Pearce [13] and subsequently

Page 13

284 K P J O C H U M E T A L

used or abused by numerous authors; rather we look upon it as merely a convenient way to il- lustrate chemical differences between different types of mafic-ul t ramafic lavas.) In the diagram, the Precambnan rocks are readily divided into three groups. The first comprises typical 2.7 Ga old komatlites and basalts, which are displaced slightly away from the Th and Nb apexes as a result of depletion of the more incompatible ele- ments. This characteristic ~s also seen in the Gorgona komatxites and in modern N-MORB. The second group includes crust-contatmnated lavas such as C592, G6 and Al140, winch are displaced away from the Nb and La apexes. The third group is made up predominantly of 3.4 Ga old komatutes and basalts, which are displaced towards the La apex and away from Nb and Th. This trend, which results from the depletion of Nb and Th relative to La (and other LREE) and is manifested in about one third of the komatntlC samples, is distinct from those of the modern basalts and the Precambrlan crust-contaminated lavas.

The O I B - M O R B data occupy a field parallel and superficially swmlar to the komatdte trend, but different because the displacement towards the La apex coincides with increasing depletion of incompatible elements. Samples at the lower-right end of the modern basalt trend are strongly de- pleted N-MORB whereas samples at the lower- right end of the komatnte trend are LREE-en- nched komatutes. The dmtlnCtlOn ~s seen more clearly in Fig. 6 in which N b / L a and N b / T h are plotted against [La/Sm]N. In Fig. 6a, Pre- cambrian komatutes show a general decrease in N b / L a with increasing L a / S m , contrary to the trend of the modern oceanic basalts In which the two ratios correlate positively Also somewhat am- biguous in the La-Nb-Th diagram (Fig. 5) are the Gorgona komatiltes, winch plot at the margin of, or within, the field of modern oceanic basalts Reference to F~g. 6b shows the Gorgona lavas at the end of komatute trends Normal depletion of incompatible elements has lowered both L a / S m and N b / T h , but along a trend largely separate from that of the modern basalts, and N b / T h in Gorgona komatiites is lower than that of the ex- trapolated O I B - M O R B trend.

The relaUonshaps of Th, Nb and La in Figs 1-5 highlight the differences between the trace

z

01

~ 1 0

z

, , , , , , , , , , , , , , , ,

5031 • 3 4 Ga 7 N - M O R B • 2 7 Ga I-KomatHtes

I [ ] G° rg°nd L , , , , , , , , ~ c3 34Ga ]

| ~ 2 7 Ga I - Basalts

[ ] L [ ] GorgonaJ

A

i i i i i J J i i

1 10 [Lo/Srn] N

Fig 6 Nb/La vs. [La/Sm]y and Nb/Th vs [La/Smly di- agrams dlustratmg differences between early Precambnan komatntes, Gorgona komatntes and basalts, and modem oc-

eanic basalts

element characteristics of komatntes and modern basalts. The depletion of Nb and Th relative to the LREE in the komatiltes is not due to alter- ation, nor contamination with continental crust, nor the processes that control the compositions of modern basalts. To understand what might have caused this characteristic it is useful to review current thinking on komatnte petrogenesls.

6. Formation of komatiites

Two features of komatntes suggest that their origin was markedly different from that of normal oceanic basalts. The first is the unusual chemistry of Al-depleted komatu tes - - the l r low A1/TI, high Ca/A1 and depletion of H R E E - - w h i c h is best interpreted as resulting from garnet (majorlte) fractionatlon. Although the problem is not en- Urely resolved, several authors [e.g. 14-16] have argued that garnet (or majortte, the high-pressure garne t -pyroxene sohd solution) was a residual phase during the melting that produced the paren- tal komatnte magma. These arguments are based largely on experimental studies which show garnet to be a hquidus phase in komatnte compositions at pressures greater than about 12 GPa. If the dlstinctwe chemistry of Al-depleted komatnte is to be explained by majonte fractlonaUon concurrent

Page 14

K O M A T I I T E P E T R O G E N E S I S A N D M A N T L E E V O L U T I O N 285

Temperature (°C)

1500 2000 2500

OI ± Px + I)q

~ , , ~~ 200 lO

Ol \ 40o

Q- 20 I + IIq

600

Solid'us & IKluk:Jus

\ Mantle ad~oat

3O

Fig 7 Pressure-temperature diagram showing the hquldus and solidus of mantle pendotlte [14], and the melting paths of basalt, plcnte and komamte, from [19] Ol = ohvlne, P x =

pyroxene, hq = hqmd, M j = majorate, Pv = perovskate, M w =

magneslowtisute, fl, 3' = bagh-P polymorphs of ohwne

with melting, the melting must been at depths greater than 360 km.

The second petrogenetacally important feature is the extreme eruption temperature inferred from the high MgO contents of many noncumulate komatntes. Jarvis and Campbell [17] and others [18,19] have argued that the eruption temperature of komatlite ( - 1600 o C) is so much greater than that of typical basalts ( < 1200°C) that the two types of magma form by different melting mecha- nisms a n d / o r come from different parts of the mantle. The essence of these arguments is il- lustrated in Fig. 7 in which are shown the solidus and liquidus of mantle peridotite and the melting paths of the sources of komatilte, picrite and MORB. It is seen from the diagram that the melting path for komatlite intersects the solidus only at a pressure greater than 20 GPa, which corresponds to a depth of greater than 600 kin. This suggests that komatiite does not form by the normal hthospheric-extenslon processes that are thought to produce modern mid-oceanic basalts [20]: rather, a komatiite erupting at 1600 o C must have already been in a molten state when it re- ached the upper mantle. Campbell and co-workers [17,18] proposed that the high temperatures of komatiite magmas are best explained by a model in which komataxte magma forms in plumes that

ascend from a thermal boundary layer deep within the mantle, perhaps at the 670 km discontinuity, but more likely at the core-mant le boundary.

If these arguments are valid (and we see no alternative mechanism to explain the l'ngh erup- tion temperatures of komatnte), then the sources of komatiite and MORB are located at entirely different levels of the mantle and may have quite different compositions. There may, however, be similarities between the origin of komatilte and modern magmas such as OIB in that both form in plumes ascending from deeper in the mantle. A modern plume probably ascends in a solid state until it intersects the sohdus at relatively shallow levels, the komatnte source starts at higher tem- peratures and is partially molten for much of its ascent.

The parallel between komatlite and modern plume-related magmas does not end with the physical aspects but may extend to the chemical character of the source. Just as the material that melts to produce modern OIB has a composition conspicuously different from that of the modern upper mantle (i.e. the MORB source), the com- position of the komatlite source may also differ from that of ambient Archaean upper mantle. Consider, for example, the Nd I s o t o p i c composi- tions of Archaean komatiltes [21,22]. The eNd val- ues of komatlites typically are distinctly positive (e.g. + 2 to +5 at 2.7 Ga) and are as high as, or higher than, those of tholelitic basalt ( + 1 to + 3), the predominant Archaean volcanic rock. If the source of komatiite is deep in the mantle, either this source is unrepresentative of the material that overlies it, or vast regions of the Archaean mantle had positive eNd values. The latter alternative is problematic, because the presence of a large volume of depleted mantle requires the existence of a commensurately large reservoir of enriched material. Various authors [23,24] have discussed the problems inherent to any explanation of the long-term depletion of even a relatively shallow layer in the upper mantle: to account for extensive or whole-mantle depletion seems impossible. It seems more reasonable to propose that the source of komatiite, like that of many types of OIB, represents material that was reworked at upper levels in the mantle, then transported to greater depths where it was stored for some time before remelting to produce the magmas that erupt on

Page 15

286 K P J O C H U M E T A L

the surface. The concept that Gorgona komatntes are plume-derived magmas from old recycled material has recently been promoted by Walker et al. [25] and Storey et al. [26].

This line of reasoning casts some light on a particular feature of the chemical and isotopic characteristics of komatntes and the Archaean mantle. For example, DePaolo [27] pointed out that material with the Sm-Nd ratio of typical 2.7 Ga komatiites Q47Sm/144Nd = 0.22-0.25) would evolve to a present-day end value of around + 40, far higher than that inferred for the upper mantle on the basis of the compositions of MORB or any other mantle-derived magmas. He assumed komatlltes to be representative of the Archaean mantle, and proposed that the discrepancy be- tween their isotopic evolution path and the path apparently followed by the upper mantle is evi- dence of recycling into the mantle of continental crust, a concept also discussed by Patchett and Chauvel [28]. If the source of komatiite was iso- lated and evolved separately from the bulk of the upper mantle, it is not valid to use its composition as a basis for models of the chermcal or isotopic evolution of the upper mantle.

It is also in the framework of a very deep origin of komatute melts that an explanation for their unusual trace element compositions can be sought. The fracuonation of Th and Nb from the LREE, which is Illustrated in Figs. 1, 2, 5 and 6, could have taken place either dunng the early reworklng of the source material, or during the high-pressure melting that produced the ultramafic magma. For example, residual phases during melting at pres- sures greater than 15 GPa would include Ca- and Mg-sIlicate perovskltes. Thorium is highly compat- ible with Ca sihcate perovsklte, Nb with Mg sdl- cate perovskite [29], and the fractlonatlon of both phases could, in theory, produce negative Th and Nb anomalies. However, Zr and Hf are also highly compatible with both perovskites, and their frac- taonatlon should also produce strong depletion of these elements, a feature not seen in the komatl- ites (Figs. la and lb). We are left to speculate that the fractionatlon of other high-pressure phases (e.g. [30]) might have led to this fractlonation, and await more results from high-pressure experimen- tal petrology. In any case, we are led to the conclusion that for the purpose of detecting the presence or absence of a Nb anomaly in the

Archaean mantle, komatiites may not be the best samples to study.

7. Basalts associated with komatiites

Even though komatlltes have been affected by processes that biased their compositions and com- promise their use as samples of typical Archaean mantle, many tholentic basalts appear to have escaped these processes. Our sample suite includes 22 basalts, many of which have petrological and chemical characteristics typical of greenstone-belt tholentes. Some have the trace element signature of crustal contamination (C592, G6, G12 and per- haps PIL26a); others display the depletion of Nb and Th that characterlses many komatutes. These rocks probably formed by fractlonation (with or without contamination) of parental komatiite magmas, and their Nb-Th ratios tell us little about ambient mantle compositions. On the other hand, other samples have smooth trace element patterns unblemished by Nb or Th anomalies. Their trace element patterns and isotopic compositions rule out derivation by crystallization of parental komatnte, and they are best interpreted as the products of melting in the upper parts of the mantle. The samples In question--5077 from Barberton, C1 and perhaps C126 from Munro Township, KA1 from Kambalda, and G37b from the Ottawa Is lands--have N b / T h between 9 and 13, slightly above the primitive mantle value, but distinctly below the values of modern MORB (Fig. 2). Especially within this group, but also within the entire suite of basalts (Fig. 8), there is a correlation between N b / T h and age. The 3.4 Ga

30

~_ 2O

Z

10

i

[ N-MORe

0 G37b 0 0~26~

Prlmlttve mantle

©

I~77

0

i i i ~ i I t

0 I 2 3 Age (Ga)

Fig 8 Diagram of Nb/Th of basalts vs age.

Page 16

KOMATIITE PETROGENESIS AND MANTLE EVOLUTION 287

old Barberton sample (5077) has the lowest value ( N b / T h = 9) and the 1.9 Ga Ottawa Islands basalt (G37b) the baghest ( N b / T h = 13)

How do the lavas from the modern komatiite locality fit into this picture? Sample G O R l l 7 , the komatfit~c basalt from Gorgona Island, has N b / T h = 11, lower than the values in most of the Precambrlan basalts and similar to values in the associated komatiltes. This sample was probably derived from parental komatnte, and its low N b / T h may be related to processes peculiar to komatlite generation. On the other hand, the basalt G O R 54 has N b / T h that is relatively high for its Nb content. This sample plots well within the field of modern oceanic basalts (Figs. 2, 3 and 6). It thus seems that only the komatiltes and related basalts on Gorgona Island show abnormally low N b / T h , supporting the concept that there is a relationship between low N b / T h and komatiite formation, whereas the tholefitic basalt has a higher N b / T h like that of other modern oceanic basalts.

From these results we infer that basalts formed by partial melting of the upper mant le- -green- stone-belt tholelites in Precambraan time, and MORB at the modern m a n t l e - - d o provxde some evidence of secular increase in N b / T h . This varia- tion may be related to continuing extraction of continental material from the mantle. The conclu- sion must be seen as preliminary, however, be- cause the data base for Archaean tholeiltes pro- vided by the present study (which at the outset had been aimed primarily at komatiites) is small.

8. Conclusions

(1) Komatiites and associated basalts have Nb- Th rauos that cluster around the estimated primi- tive mantle value and are distinctly lower than those of modern oceanic basalts with similar Nb contents. The lowest values are found m two basalts that had been contaminated wxth continen- tal crust. The Nb-Th ratio is shown to be a very sensmve indicator of crustal contamination which can be used to discriminate between contaminated lavas and those that preserve pramary trace ele- ment ratios. Nb-Th ratios in uncontaminated komatiites vary with the age of the rocks. The lowest values (7-11) were measured m - 3.4 Ga komathtes; shghtly higher values (10-15) come

from - 2.7 Ga samples, and Cretaceous-Tert iary komatiites from Gorgona Island have still higher values (16 to 24).

(2) Although the Nb-Th ratio in komattites ts close to the primitive mantle value, both elements are depleted relative to the LREE. This observa- tion, together with similar Nb-Th ranos measured in Precambrlan and Cretaceous-Tert iary komati- ~tes, is at odds wtth the expectation that the corn- positrons of Precambrlan komatute would reflect that of essentially undifferentiated mantle. The source of komatlite was not typtcal Archaean mantle but material that had been stored m a deep location where it was isolated from the main con- vecting mantle. The cause of the Nb-Th-La frac- tionatlon in komatlites is not understood, but may have taken place during the earlier history of the source material of komatiites, or through the frac- tlonatlon of high-pressure phases during the par- tial melting that produced the komatiite magmas.

(3) Certain Archaean tholentic basalts have smooth trace element spectra unblenushed by the types of Nb-Th anomalies that result from crustal contamination or the pecuhar process that has affected the komatlltes. Nb-Th ratios in > 3 Ga basalts are generally lower than in 1.9-2.7 Ga basalts, which in turn are lower than in modern basalts with s~rmlar Nb contents. This observation provides some evidence of secular variation in the composition of the upper mantle, and suggests that continental material continued to be ex- tracted from the mantle throughout the past 3 Ga.

Acknowledgements

We are grateful to the following people for supplying samples a n d / o r analytical data: Gerard Gruau, Mike Lesher (and Western Mining Corpo- ratxon Ltd.), Mike Bickle, Alan Cattell, Ernst Hegner, Llna Echeverria and Bernard Dupr& Sigrid Mtdinet-Best assisted dunng SSMS analyses and Bob Nesbitt and two anonymous reviewers provided useful comments on the manuscript. The investigation was undertaken with the support of the Deutsche Forschungsgemelnschaft (projects Kr 590/11.1 and Kr 590/11.2).

References

1 A W Hofmann, K P Jochum, M Seufert and W M Wh|te, Nb and Pb m oceanic basalts: new constrmnts on mantle evolution, Earth Planet Scl Lett 79, 33-45, 1986

Page 17

288 K P JOCHUM ET AL

2 K P Jochum, H M Seufert, S Mldinet-Best, E Ret tmann, K Schonberger and M Zlmmer, Multi-element analysis by isotope di lu t ion-spark source mass spectrometry (ID- SSMS), Fresemus Z Anal Chem. 331, 104-110, 1988

3 A W Hofrnann, Chemical differentiation of the Earth the relationship between mantle, continental crust, and oceanic crust, Earth Planet Scl Lett 90, 297-314, 1988

4 R W . Nesbxtt, S S Sun and A C Purvls, Komatntes geo- chemistry and genesis, Can Mineral 17, 165-186, 1979

5 S R Taylor and S M McLennan, The Continental Crust its Composit ion and Evolution, 312 p p , Blackwell Scien- tific Pubhcatlons, Oxford, 1985

6 H E Huppert, R S J. Sparks, J S Turner and N T Arndt, Emplacement and coohng of komatnte lavas, Nature 309, 19-22, 1984

7 H E Huppert and R S J Sparks, Kornatntes, I Eruption and flow, J Petrol 26, 694-725, 1985

8 C Chauvel, B Dupr6 and G A Jenner, The Srn-Nd age of Kambalda volcamcs is 500 Ma too oldT Earth Planet Scl Lett 74, 315-324, 1985

9 N T Arndt and G A Jenner, Crustally contaminated komatntes and basalts from Kambalda, Western Australia, Chem Geol 56, 229-255, 1986

10 W Cornpston, I S Williams, I H Campbell and J J Gresham, Zircon xenocrysts from the Kambalda volcanics age constraints and direct evidence for older continental crust below the Kambalda -Norse rnan greenstones, Earth Planet Scl Lett 76, 299-311, 1986

l l N T Arndt, G E Brugrnann, K Lehnert, C Chauvel and B W Chappell, Geochemistry, petrogenesls and tectomc envlronrnent of Clrcum-Superlor Belt basalts, Canada, in Geochemistry and Mlnerahzatlon of Proterozolc Volcamc States, T C Pharaoh, R D Becklnsale and D Rlckard, eds , pp 133-145, Geol Soc Spec Publ 33, 1987

12 B Dupr6 and L M Echeverria, Pb isotopes of Gorgona Island (Colombia) isotopic variations correlated with magma type, Earth Planet Scl Lett 67, 186-190, 1984

13 J A. Pearce and J R Cann, Tectonic setting of basic volcanic rocks determined using trace elements analyses, Earth Planet. Scl Lett 19, 290-300, 1973

14 E Takahastu, Melting of a dry perldotite KLB1 up to 14 GPa lmphcations on the ongm of pendotl te upper mantle, J Geophys Res 91, 9367-9382, 1986

15 E Ohtam, I Kaeabe, J Morlyama and Y Nagata, Parti- tioning of elements between majonte garnet and melt and implications for genesis of komatute, Con tnb Mineral Petrol. 103, 263-269, 1989

16 G Gruau, C Chauvel, N T Arndt and J Cornlchet, Al-de- pletlon in komatntes and garnet fractlonatlon In the early Archean mantle Hf isotopic constraints, Geochlrn Cosmo- chim Acta 54, 3095-3101, 1990

17 G T Jarvxs and I H Campbell, Archean komatntes and geotherrns" solution to an apparent contradiction, Geophys Res Lett 10, 1133-1136, 1983

18 I H Campbell, Heads its basalts, tails its komatutes, Na- ture, in press, 1991

19 G H Miller, E M. Stolper and T J Ahrens, The equation of state of molten kornatnte, kornatute petrogenesis, and the evolution of the Hadean mantle, J Geophys Res , in press, 1991

20 D McKenzae and M J Backle, The volume and composl- non of melt generated by extension of the lithosphere, J. Petrol 29, 625 679, 1988

21 S B Sharey and G N Hanson, Mantle heterogeneity and crustal recycling m Archean granl te-greenstone belts evi- dence from Nd isotopes and trace elements m the Rainy Lake area, Superior Province, Ontario, Canada, Geochlm Cosmocham Acta 50, 2631-2651, 1986

22 B M Jahn, B Auvray, Z Zhang, J Cornlchet, Y L Bai, Q H Shen and D Y klu, 3.5 Ga old amphlbohtes from eastern Hebel province, China field occurrence, petrogra- phy, Sm-Nd tsochron age and REE geochemistry, Pre- cambrian Res 34, 311-346, 1987

23 C G Chase and P J Patchett, Stored mafic/ul trarnafic crust and early Archean mantle depletion, Earth Planet Sci Lett 91, 66-72, 1988

24 S J G Galer and S L Goldstem, Early mantle differentia- tion and thermal consequences, Geochlm Cosmochlm Acta, m press, 1991

25 R J. Walker, L M Echeverria, S B Sturey and M F Horan, Re-Os isotopic constraints on the ongm of volcanic rocks, Gorgona Island, Colombia Os isotopic ewdence for an- cient heterogeneitles m the mantle, Con tnb Mineral Pet- rol 107, 150-162, 1991

26 M. Storey, J J Mahoney, L W Kroenke and A D Saunders, Are oceanic plateaus sites of komatnte formatlon9 Geology 19, 376-379, 1991

27 D J DePaolo, The mean life of continents estimates of continent recychng rates from Nd and Hf isotopic data and lrnphcatlons for mantle structure, Geophys Res Lett 10, 705-708, 1983

28 P J Patchett and C Chauvel, The mean hfe of the conti- nents is not constrained by Nd and Hf isotopes, Geophys Res Lett. 11 ,151-153, 1984

29 T Kato, A.E. Rmgwood and T. Infune, Expenrnental determination of element part l t lomng between sihcate per- ovskltes, garnets and liquids constraints on early differ- entlation of the mantle, Earth Planet Sc~ Lett 89, 123-145, 1988

30 M. Madon, J Castex and J Peyronnean, A new alumlno- calclc high-pressure phase as a possible host of calcium and a lummlum in the lower mantle, Nature 342, 422-425, 1989

31 N T Arndt and R W Nesbitt, Magma rmxlng In komatutlc lavas from Munro Townstup, Ontario, in Archaean Geo- chemistry, A Kroner, G N Hanson and A M Goodwln, eds , pp 99-115, Springer, Berhn, 1984

32 N T. Arndt, Differentiation of komatute flows, J Petrol 27, 279-301, 1986

33 E G Nlsbet, N T Arndt, M J Blckle, W.E Cameron, C Chauvel, M Cheadle, E. Hegner, T K Kyser, A Martin, R Renner and E Roedder, Uniquely fresh 2.7 Ga komatmes from the Behngwe greenstone belt, Zimbabwe, Geology 15, 1147-1150, 1987

34 L M. Echeverria, Kornatutes from Gorgona Island, Col- ombia, in Komatutes , N.T Arndt and E.G Nlsbet, eds , pp 199-210, Allen and Unwln, London, 1982

35 Aitken, B G and L M Echeverria, Petrology and geochem- istry of komatntes and tholentes from Gorgona Island, Colombia, Con tnb Mineral Petrol 86, 94-105, 1984

36 B.M Jahn, G Gruau and A Y Ghkson, Komatll tes of the

Page 18

KOMATI1TE PETROGENESIS AND MANTLE EVOLUTION 289

Onverwacht Group, South Africa REE chermstry, Sm-Nd age and mantle evolution, Contnb Mineral Petrol 80, 25 40, 1982.

37 G Gruau, B.M Jahn, A Y Ghkson, R Davy, A H Hick- man and C Chauvel, Age of the Talga-Talga Subgroup, Pdbara Block, Western Austraha, and early evolutxon of the mantle new Sm-Nd lsotoplc evidence, Earth Planet So Lett 85, 105-116, 1987

38 G Gruau, C Chauvel and B M Jahn, Anomalous Sm-Nd ages for the early Archean Onverwacht Group volcanlcs slgmficance and petrogenetic ~mphcatlons, Contr lb Mmeral. Petrol 104, 27-34, 1990

39 A Cattell, Enriched komatntlc basalts from Newton Town- ship, Ontario their genesis by crustal contamination of depleted komatnte magma, Geol Mag. 124, 303-309, 1987

40 A Cattell and N T Arndt, Low- and high-alumina komatl- ltes from a Late Archean sequence, Newton Townstup, Ontario, Contrib Mineral Petrol 97, 218-227, 1987

41 N T Arndt and R W Nesbltt, Geochemistry of Munro Township basalts, m Komatutes, N T Arudt and E.G Nlsbet, eds , pp. 309-330, Allen and Unwm, London, 1982

42 C Chauvel, N T Arndt, S Kaehnzcuk and A Thom, For- mation of Canadian 1 9 Ga old continental crust, I Nd isotopic data, Can J Earth So 24, 396-406, 1987.