Diffusive Isotopic Contamination of Mafic Magma by Coexisting Silicic Liquid in the Muskox Intrusion

Abstract

Shifts in 87Sr/86Sr and 143Nd/144Nd ratios measured in cumulates from the upper levels of the Muskox mafic intrusion indicate that isotopic and bulk chemical
exchange were decoupled across a mafic-silicic liquid interface during crystallization of the intrusion. Modeling of diffusive
exchange between liquid layers demonstrates that isotopic compositions of silicate liquids in layered magma chambers may be
strongly affected by this process on time scales of 103 to 104 years. Diffusive contamination can be used to place constraints on the physical processes and time scales of magmatic systems.

Full text

Diffusive Isotopic Contamination
of
Mafic Magma by
Coexisting Silicic Liquid in the Muskox Intrusion
BRIAN
W.
STEWART*
AND
DONALD
J.
DEPAOLO
Shifts
in
87Sr/86Sr
and
143Ndfl44Nd ratios measured
in
cumulates
from
the
upper
levels
of
the
Muskox mafic
intrusion
indicate
that
isotopic
and
bulk
chemical exchange
were decoupled across a mafic-silicic
liquid
interface
during
crystallization
of
the
intrusion. Modeling
of
diffusive exchange between
liquid
layers demonstrates
that
isotopic compositions
of
silicate liquids
in
layered
magma
chambers may
be
strongly
affected
by
this process
on
time scales
of
10
3
to
10
4 years. Diffusive contanlination can
be
used
to
place constraints
on
the
physical processes
and
time scales
of
magmatic
systems.
C
HEMICAL
DIFFUSION
IS
GENERAL-
ly thought
not
to be important in
magma evolution, because the
effec-
tive diffusivities
of
most elements are small
in comparison
to
the length scales
of
mag-
ma chambers. However, recent studies
of
magma chamber dynamics (1-4) show that
low-density, silica-rich magma can be
maintained for a long time
as
a stable layer
over denser mafic magma with a sharp
compositional interface between the two
and little mechanical mixing.
In
this situa-
tion, diffusion need operate only over the
short distances
of
the compositional
boundary layer between the two magma
layers. Even with a thin boundary layer,
chemical diffusion
is
minimal and the con-
centrations
of
elements within each layer
are negligibly affected by the presence
of
the adjacent layer (5-7). Isotopic ratios,
however, can be more strongly affected by
diffusion, because diffusion
of
isotopic ra-
tios occurs independently
of
diffusion
of
concentrations
(8)
9), and effective diffu-
sivities
of
isotopic ratios are
not
ultimately
controlled by the especially slow diffusion
of
the major components
Si0
2 and Al203
(10) 11).
Our
isotopic data from the
Muskox intrusion demonstrate that mea-
surable diffusive isotopic exchange occurs
between mafic and silicic magma layers
in
magma chambers.
The Muskox intrusion
is
a prism-form
layered mafic intrusion exposed along the
Arctic Circle
in
Northwest Territories,
Canada. The ultramafic and mafic cumu-
lates that form most
of
the intrusion reach
a thickness
of
-2000
m and may thicken
northward where the intrusion dips be-
neath its sedimentary cover. Muskox intru-
sion cumulates crystallized from a series
of
Berkeley Center for Isotope Geochemistry, Departtnent
of
Geology and Geophysics, University
of
California,
and Earth Science Division, Lawrence Berkeley Labora-
tory, Berkeley, CA 94720.
*To whom correspondence should be addressed. Present
address: Division
of
Geological and Planetary Sciences
170-25, California Institute
of
Technology, Pasadena,
CA 91125.
708
25 injections
of
magma into a shallow
chamber
(12)
13). Each successive injection
pushed some
or
all
of
the previously exist-
ing magma
out
of
the chamber and formed
a new cyclic
unit
characterized by repeti-
tions in the cumulate stratigraphy and
shifts
in
crystallization parameters such
as
the Mg/Fe ratio and the
Ni
concentration
in
olivine. Complete crystallization
of
the
final magma injection yielded a cyclic
unit
(C.U. 25) 150 to
200
m thick. During
deposition
of
the cumulates, melting
of
wall and
roof
rock formed a layer
of
low-
density,
Si0
2-rich liquid at the
roof
of
the
chamber that
is
preserved
as
a xenolith-rich
granophyric
unit
0 to 70 m thick directly
overlying C.U. 25
(13)
14); thus mafic and
silicic liquids coexisted
in
a gravitationally
stable configuration during crystallization
of
the mafic magma.
We measured isotopic ratios
ofN
d (
143
N d/
144Nd)
and Sr
(87Sr/86Sr)
and concentrations
of
Rb, Sr, Sm, and
Nd
in samples spanning
the uppermost
cyclic
unit
(C.u.
25), the
granophyric roof
wne
(GRZ), and the sur-
rounding wall rocks (Table 1) (15). InitialNd
and Sr isotopic ratios
(16)
of
cumulates are
assumed to record the isotopic ratio
of
the
magma from which they crystallized. Minor
postcrystallization alteration had little
ef-
fect
on
the Sm-Nd systematics
but
some
effect
on
the Rb-Sr system,
as
demonstrat-
ed by the agreement
of
mineral and whole-
rock initial
ENd
values for sample 83DM-3
from C.U. 22 (Table 1) and the discor-
dance
of
its respective 87Sr/86Sr values.
However, because the relatively large
whole-rock samples used in this study
should average
out
millimeter-scale isoto-
pic disturbances, we believe that the age-
corrected 87Sr/86Sr ratios reflect magmatic
values,
as
has been shown for other layered
mafic complexes
(3)
4}
17).
The
ENd
and 87Sr/86Sr values
of
the
GRZ
fall intermediate between wall rock end-
members and probably represent an aver-
age
of
the melted wall rock components
(Fig. 1). Values
of
87Sr/86Sr and
ENd
for
C.U. 25 samples lie within a much more
restricted range (Fig. 1 inset and Fig. 2).
The cumulates
all
have
ENd
values
of
-0.75
(dashed line, Fig. 2), within analytical un-
certainties, except the uppermost sample
(83DM-U).
In
contrast, 87Sr/86Sr values
increase with stratigraphic height (toward
GRZ
values) well outside
of
analytical
error, from 0.7051
to
0.7097; only one
sample (83DM-12) deviates from this
trend. These data show that the C.U. 25
magma was progressively enriched in ra-
diogenic Sr while its
ENd
remained essen-
tially constant.
To
test whether bulk assimilation-frac-
tional crystallization (AFC) models could
account for the discrepancy between the
Nd
and Sr isotopic data, we constructed
two models
(3)
4}
18), one with a value for
a (ratio
of
mass rate
of
assimilation to mass
rate
of
crystallization) equal
to
zero, the
other for a = 0.12.
In
the calculations we
used the
Nd
and Sr concentrations and
isotopic compositions
of
the
GRZ
for the
assimilant, and a bulk (crystals + intercu-
mulus liquid) distribution coefficient
of
Sr
(Ksr) equal
to
1. We let
KNd
increase from
0.2 at the base
ofC.U.
25
to
0.8 at the
top
to reflect increasing amounts
of
trapped
intercumulus liquid
(19)
20). The AFC
models could
not
account for both the N d
and Sr isotopic variations (Fig. 3A). The
87Sr/86Sr data require a
~
0.10, while the
143Nd/144Nd data require a 5 0.05.
Because
of
the well-established differ-
ence in diffusivities between
Nd
and Sr in
silicate melts
(5)
6), we next investigated
the possibility that the isotopic variations
were caused by diffusive contamination
rather than bulk mixing. Baker (7) found
that the elemental and isotopic effects
of
diffusive exchange between coexisting an-
hydrous rhyolite and dacite layers are small
for time scales less than 105
to
106 years.
However, recent diffusion experiments
(21) show
that
Sr tracer diffusivities in wet
rhyolites are large enough
to
allow signif-
icant isotopic exchange in geologically rea-
sonable time scales.
We use the example
of
Sr isotopes
to
deScribe
our
model
of
isotopic exchange
between two silicate liquid layers (22). The
total concentration
of
Sr
in
each layer
is
assumed to evolve independently
of
the
isotopic exchange across the interface. The
concentration
of
Sr in the silicic layer
(C
a)
is
taken to be constant; the concentration
in the mafic layer
(C
m)
is
controlled by
fractional crystallization
of
the mafic mag-
ma. The compositional interface between
the two layers
is
assumed
to
be infinitely
sharp with respect
to
all components
ex-
cept
87Sr
(23). Diffusion
of
87Sr
from the
silicic magma
to
the mafic magma takes
place through a boundary layer
of
thickness
SCIENCE,
VOL. 255
" .
.1
A

'C
Z
w
0.72
0.74
0.76 0.78 0.80
87Sr
/86S
r
x,
which lies entirely within the silicic
magma layer (24).
The
flux
of
87Sr
into
the
mafic liquid can
then
be written:
DSrPa
87
87
J87
=
--
([
Sr]a -
[Sr]'i)
(1)
x
where [
87
Sr];
is
the
concentration
in
moles
per
unit
mass
of
87
Sr
in
the
silicic magma
that
would
give it an 87Sr/86Sr ratio equal
to
that
of
the
mafic layer,
Dsr
is
the diffu-
sivity
of
Sr, and
Pa
is
the density
of
the
silicic magma. This equation can
then
be
rewritten:
DSrPa 86
J87
=
--
[ SrJa(Ea -
Em)
(2)
X
where we have used E
as
shorthand
for
87Sr/86Sr.
From
this formulation, ex-
pressions for
the
rate
of
change
of
87
Sr/
86
Sr
in
the mafic and silicic liquids can be
derived:
dEm
DsrpaACa
dt
=
xFMoC
m
(Ea
-
Em)
(3)
Fig. 1. Variations in initial (1257 Ma)
87Srj86Sr
and
ENd
values for the Muskox intrusion and
surrounding wall rocks. Sample M-156 from the
granophyric roof zone (closed square)
lies
within
the field (shaded region) defined by the various
wall rocks around the intrusion (open squares).
Cumulates from
c.u.
25 (closed circles) show a
significant spread in
87Srj86Sr
ratios but
not
in
ENd
values (inset).
where A
is
the interface area, and F
is
the
fraction
of
mafic liquid remaiuing relative
to
the initial liquid mass Mo (25).
For
the case
of
fractional crystallization with variable
K,
the rate
of
change
of
an element concentra-
tion
in
the mafic magma
is
given
by
dCm ldF
at
=
(K
sr
-
I)Cmr
at
(5)
In
order
to
solve Eqs. 3
and
5, the time
dependence
of
crystallization within the
mafic magma
must
be specified.
In
the
modeling
that
follows, we assume
that
the
rate
of
crystallization
is
proportional to
the square
root
of
time (26, 27);
that
is, F
= 1 - (t/te
)1/2,
where t
is
the
time elapsed
since initial injection
of
magma
into
the
chamber and
te
is
the
total time
of
crystal-
lization.
The
results are
not
strongly de-
pendent
on
the crystallization law used.
Equations 1
through
5 apply equally
to
Nd
isotopes.
Table
1.
Sample locations, trace element concentrations, and isotopic data.
Height* Sr
Nd
87Rbj
Sample Location (m) 86Sd
(ppm)
M-156
GRZ
98.7 50.87 5.55
83DM·ll
c.u.
25 163 155.3 33.58 3.08
83DM-12
C.U.25
155 186.9 54.16 1.227
83DM-23
c.u.
25 133 293.9 25.67 0.598
83DM-8
C.U.25
III
332.8 26.92 0.382
83DM-7
C.u.
25 89 310.1 23.03 0.336
83DM-16
C.U.25
57 312.7 22.29 0.286
83DM-5
c.u.
25 42 229.6 8.05 0.1329
83DM-4
c.u.
25 29 110.8 5.059 0.1311
83DM-3
C.u.
22
342.3 10.04 0.0899
Clinopyroxene 25.79 14.76 0.0963
Orthopyroxene 1 2.24 0.905 0.313
Orthopyroxene 2 2.6 1.118
Plagioclase 566.8 2.430 0.0608
83DM-42a Granite wall rock 58.54 41.07 15.23
83DM-42b Granite wall rock 84.4 119.9 7.49
83DM-14 Hornfels wall rock 225.7 12.56 0.316
83DM-35 Orthogneiss wall rock 193.7 20.21 0.605
To
model
the
effects
of
diffusional ex-
change combined
with
fractional crystalli-
zation (DFC)
in
the Muskox intrusion,
we
estimated values for
the
physical parame-
ters and
then
used
the
variations
of
the
model curves
with
te
to
constrain
the
crys-
tallization time
of
the
intrusion.
We
esti-
mate
that
the
thickness
of
the
silicic por-
tion
of
the
chemical
boundary
layer
is
0.1
m (28)
and
that
the
silicic magma con-
tained 5% H20 by weight.
The
silicic
boundary layer was
most
likely heated
to
supersolidus temperatures by
the
underly-
ing convecting mafic magma. Interpolating
between Baker's (21) estimates
of
Dsr
in
hydrous rhyolite
at
IOOO°C
ofl.5
X
10-
12
m2
S-l
at 3.5% H20
and
2.3 X
10-
11
m2
S-l
at
6% H20
and
assuming
that
log
Dis
proportional
to
the
H20
content
for
>2%
H20 (21, 29), we calculated a value
of
8 X
10-
12
m2
S-l
for D
Sr
' Lesher (30)
showed
that
DNd
in
rhyolite
is
~
1/8
that
of
Ds
",
so
we used DNd = 1 X
10-
12
m2
S-l.
The
same values
of
Ksr
and
KNd
were used
as
in
the
AFC modeling.
Nd
and
Sr isotopic
ratios
of
the silicic liquid
at
t = 0 were
chosen so
that
they
would
evolve
to
the
values
of
GRZ
by
the
end
of
the
crystalli-
zation
of
the mafic liquid (t = te).
The
magnitudes
of
the
isotopic shifts
in
the
silicic magma are
dependent
on
the
mass
of
the silicic liquid layer.
In
the
case
of
the
Muskox intrusion,
the
ratio
of
silicic liquid
mass
to
mafic liquid mass (Ma/Mrn)
must
be greater
than
~0.35
to
obtain
geologi-
cally reasonable silicic liquid isotopic com-
positions
at
t = 0 (31).
We
set
MjMrn
=
0.35
in
the
modeling described below.
For
the
chosen parameters, a crystalliza-
tion
time
of
8000
years for
the
uppermost
Initial:!:
147Smj
Initial:!:
87Srl6Sr
±§
144Ndll
ENd
±§
0.73764 53 0.1133
-11.2
0.5
0.70974 30 0.1291
-6.2
0.4
0.70583 13 0.1363
-0.4
0.6
0.70833 7 0.1432
-1.0
0.6
0.70639 5 0.1462
-0.7
0.6
0.70698 7 0.1458
-0.8
0.4
0.70562 6 0.1377
-1.0
0.5
0.70567 3 0.1635
-0.7
0.5
0.70507 3 0.1745
-0.5
0.6
0.70535 3 0.1424
-0.3
0.4
0.70533 4 0.1861
-0.8
0.5
0.70502 15 0.2587
-0.9
0.7
0.2409
-0.5
0.5
0.70543 3 0.0882
-1.0
0.6
0.79160 139 0.1078
-20.3
0.8
0.73982 70 0.0855
-22.0
0.5
0.70594 5 0.1657 0.4 0.4
0.70933 7 0.1308
-5.8
0.5
*Stratigraphic
height
above
base
of
C.U.
25.
uncertainty
in
87Rb/86Sr
or
147Sm/144Nd.
tUncertainty :50.5%
IIUncertainty
:50.3%.
:j:Corrected
to
crystallization
age
of 1257
Ma
(16).
§U
ncertainty
includes
± 2
SEM
in
ratio
and
7
FEBRUARY
1992
REPORTS
709

0.0
'"*"'
j 0.2
150
4-<
g 1
1
•
>-+t'
.:c
m 1
"ii 100
.....,.....
0.4
.l:
1
u •
:c
Ht"' "
Q.
0.6
as
1
..
m 50
....,..
~
........
0.8
liS
• 1
'I*""'
Fig. 2. Plots
of
87Sr/86Sr
ratios and
ENd
values for
C.
U.
25 cumulates against
stratigraphic height above
the base
of
C.V. 25 and F
(mass fraction
of
liquid
re-
maining relative
to
initial
liquid mass); F
is
assumed
to
be proportional
to
strati-
graphic height in C.V. 25.
All
ENd
values
fall
within er-
ror
of
-0.75
(dashed line)
except the uppermost sam-
ple
(83DM-ll);
in contrast,
117
Sr/
86
Sr values increase up-
ward in the section.
o 1 I I I 1 1 I I I I I I I 11.0
0.704 0.706 0.708 0.710 -7 -6 -5 -4
-3
:2
-1
0 1
87Sr/86Sr eNd
cyclic
unit
provides a reasonable fit
to
both
the Sr and the
Nd
isotopic data (Fig. 3B).
For
comparison, neither 800-year
nor
30,000-year crystallization times fit the Sr
isotopic data, although any time less than
30,000 years makes an acceptable fit
to
the
Nd
data. A crystallization time
of
30,000
years requires an initial silicic liquid 87Srj
86Sr
value
of
0.780, which
is
difficult
to
reconcile with the wall rock data.
For
tc
=
8000 years, the 87Srj86Sr ratio
of
the silicic
liquid shifts downward by 0.01 from an
initial value
of
0.748
at
t = 0, well within
the field
of
wall rock values (Fig. 1); the
shift
of
ENd
in the silicic liquid
is
negligible.
Because the underlying
24
cyclic units were
formed by partial crystallization
of
earlier
magma injections, the total lifetime
of
the
Muskox magmatic system (including the
o.O'A
0.2
~
'i:
0.4
LI..
0.6
0.8
0.2
0.4
LI..
0.6
0.8
..
o,,'l-
e,.r.
.....
1.0 I ( . ! I
!,
'"
I !
0.704 0.706 0.708 0.710 -7 -6 -5 -4 -3 -2
-1
a
87S
r
/86S
r
eNd
Fig. 3.
(A)
Isotopic evolution curves for AFC
models compared
to
C.u.
25 data. Curves are
shown for the
cases
of
a equal to 0 and a = 0.12.
F
is
the fraction
of
mafic liquid remaining in
C.
V.
25. (B) Isotopic evolution curves for
DFC
models
compared to C.V. 25 data. The curves represent
the isotopic evolution
of
the
mafic
magma
as
it
diffusively exchanges with the overlying silicic
magma. The numbers indicate the total crystalli-
zation time
of
C.
V.
25 for each model;
see
text for
other parameters.
710
surface flows from venting
of
the chamber)
would then be
on
the order
of
50,000 to
100,000 years. Additional modeling indi-
cates that the results are
not
sensitive to the
chosen partition coefficients,
nor
are they
sensitive
to
the initial concentrations
of
Sr
and
Nd
in the magma and assimilant, so
long
as
realistic estimates are used. How-
ever, the magnitudes
of
the isotopic shifts
caused by
DFC
are strongly dependent
on
the chosen values for diffusivity and
boundary layer thickness (Eqs. 3 and 4).
Improved understanding
of
boundary layer
conditions should lead
to
more accurate
estimates
of
crystallization times in
cases
where
DFC
has operated.
At deeper crustal levels where crystalliza-
tion times are likely
to
be much greater, the
effects
of
diffusional isotopic exchange may
be even larger than those observed in the
Muskox intrusion.
In
Fig. 4, we construct-
ed a model using hypothetical end-mem-
bers to demonstrate the possible isotopic
effects
of
DFC
compared
to
those
of
APC.
The path for the APC process with a = 0.1
is
compared
to
the effects
ofDFC
with
tc
=
20,000 years for a sill
of
basaltic magma
200 m thick with isotopic ratios character-
istic
of
mantle plumes
(ENd
= 0, 87Srj86Sr
= 0.7045) interacting with typical Pro-
terozoic granitic crust
(ENd
= - 20,
87
Sri
86Sr
= 0.720). Given a sufficiently lengthy
crystallization time, it
is
clear that
DFC
can
produce Sr isotopic effects comparable to
APC, while maintaining almost constant
Nd
isotopic ratios. Interpretation
of
a
DFC
trend
as
an APC trend would lead
to
gross
overestimates
of
the 87Srj86Sr ratio
or
ENd
of
the assimilant end-member.
In
some
instances, APC and
DFC
may operate si-
multaneously; in these cases, the path
would be intermediate
to
the curves shown
in Fig. 4. The magnitude
of
the
DFC
effects are directly proportional to
tc
and
inversely proportional
to
the thickness
of
the mafic magma chamber. The 87Srj86Sr
ratio
of
the silicic magma can be signifi-
cantly modified
as
well, again with little
change in its major-
or
trace-element chem-
J
~
""'"
't-
-1
-2
..,
z
w -3
-4
-5
-6L'
__
~~
__
L-
__
~
__
-L
__
~
__
~
0.704 0.706 0.708 0.710
87Sr
/86S
r
Fig. 4. Model curves for
mafic
magma
(ENd
= 0,
87Sr/86Sr
= 0.7045) interacting with Proterowic
granitic crust
(ENd
=
-20,
87SrjB6Sr
= 0.720) for
the processes
of
DFC
and AFC.
In
the
DFC
model shown, the total crystallization time
of
the
mafic
magma
is
20,000 years; the
ticks
along this
curve indicate time passed since the initial intru-
sion
of
the magma. Tie lines represent equivalent
values
ofF
(mass fraction
of
mafic
magma remain-
ing in magma chamber).
istry. This effect
is
enhanced
if
the under-
lying mafic magma
is
repeatedly replen-
ished by injections
of
fresh magma.
The extent
to
which we can assume
DFC
operates
in
natural systems depends pri-
marily
on
our
estimates
of
the crystalliza-
tion times
of
mafic magma chambers,
which
is
contentious (32). However, the
physical situation
of
silicic magma stably
overlying mafic magma
is
likely to develop
any time a basaltic magma ponds beneath
less dense, low-melting granitic material,
and
in
such cases physical mixing
is
thought
to
be inefficient. Thus,
DFC
should be considered in any interpretation
of
isotopic variations in basalts erupted
through continental crust.
In
particular,
shifts in 87Srj86Sr ratios without corre-
sponding shifts
in
143Ndjl44Nd, common-
ly
attributed
to
interaction with altered
mafic crust, need to be reexamined.
Al-
though few experiments have been carried
out
on
Pb
diffusion
in
magmas, the rela-
tively high diffusivity
ofPb
in minerals
(33,
34) leads us
to
speculate that the effects
of
DFC
on
Pb
isotopes in mafic magmas may
be even larger than those exhibited by Sr.
This
is
an additional factor to be consid-
ered in cases where
Pb
isotopic variations
in igneous mafic rocks are large compared
to those
of
Sr and N d.
REFERENCES
AND
NOTES
1.
1.
H. Campbell and
J.
S.
Turner,
J.
CeDI.
95,
155
(1987).
SCIENCE,
VOL. 255

2. C. M. Oldenburg, F.
J.
Spera, D.
A.
Yuen, Earth-
Sci. Rev.
29,
331 (1990).
3.
D.
J.
DePaolo,
J.
Petrol.
26,925
(1985).
4.
B.
W. Stewart and D. J. DePaolo, Contrib. Mineral.
Petrol.
104,
125 (1990).
5. M. Magaritz and
A.
W. Hofmann, Ceochim. Cosmo-
chim.
Acta
42,
595 (1978).
6.
__
, ibid.,
p.
847.
7. D. R. Baker, Contrib. Mineral. Petrol.
104,
407
(1990).
8.
__
, Ceochim. Cosmochim. Acta
53,
3015
(1989).
9. C. E. Lesher, Nature
344,
235 (1990).
10. E.
B.
Watson and
S.
R. Jurewicz,
J.
Ceol.
92,121
(1984).
11.
A.
F.
Trial and F.
J.
Spera,Int.J.
Heat Mass Transftr
31,941
(1988).
12. T. N. Irvine and C.
H.
Smitb, in Ultramqfic and
Related Rocks, P.
J.
Wyllie, Ed. (Wiley,
New
York,
1967), pp.
38-49.
13. T. N. Irvine, Ceol.
Soc.
S.
Aft.
Spec. Publ.
1,
441
(1970).
14. Thickness
of
GRZ
estimated from map and cross
sections
ofC.
H. Smitb, T.
N.lrvine,
D. C. Findlay,
Ceol. Surv. Can. Maps 1213-A and 1214-A (1966).
15. Sample M-156, collected from a xenolitb-rich region
just below tbe roof, was provided by T. N. Irvine;
all
otber samples were collected by D.J.D. in a 1983
expedition. Stratigraphic positions
of
samples witb-
in C. U. 25 (Table 1) are estimated from tbeir field
relations and map locations and are given relative
to
tbe base
of
C.U. 25, which
is
assumed
to
have an
average thickness
of
170 m. Whole-rock samples
of
100
to
500 g were crushed and split, and 100-
to
500-mg aliquots were dissolved witb
HF
+
HCl0
4.
Mer
spiking witb tracer solutions
of
Rb, Sr, Sm,
and
Nd,
tbe cations
of
interest were separated by
standard ion-exchange techniques. Concentrations
and isotopic ratios were determined by tbermal
ionization mass spectrometry, and
all
elements were
run
as
metal species except N d, which was
run
as
NdO+.
Most
measurements were carried
out
on
VG
single collector machines
at
tbe University
of
Cali-
fornia, Los Angeles (UCLA) and tbe Berkeley Cen-
ter for Isotope Geochemistry (BCIG), witb some Sr
isotopic analyses
run
on
tbe BCIG
VG
multicollec-
tor. Botb single-collector machines give tbe same
values for '43Nd/'44Nd witbin measurement uncer-
tainties, and
all
87Sr/86Sr
ratios are corrected
to
tbe
pre-1989
UCLA
value for NBS987
ofO.7l031.
16. All values
of
87
Sr/
86
Sr and
ENd
qnoted in tbe text are
initial ratios (corrected for decay
of
87Rb and
147Sm
since tbe time
of
crystallization).
In
this study we
assume a crystallization age
of
1257
Ma
(million
years ago), based
on
an internal Sm-Nd isochron
from a two-pyroxene gabbro collected from C.
U.
22
(sample 83DM-3, Table 1); this age
of
1257
± 40
Ma
(2<1)
is
within error
oftbe
U-Pb
age
ofl270
±
4 Ma
of
A. N. LeCheminant and L. M. Heaman
[Earth
Planet. Sci. Lett.
96,
38 (1989)]. Age-correct-
ed N d isotopic ratios are reported
as
ENd,
witb
4(
143Nd/144NdSAMP(T)
)
ENd(T)
= 10
143Nd/144NdCHUR(T)
- 1
where T
is
tbe crystallization age
of
tbe Muskox
intrusion and
CHUR
is
tbe chondritic reservoir. The
chondritic isotopic composition at time T
is
given
by: '43Nd/'44NdcHUR(T) = 0.511836 -
0.1967[exp(AsmT) -1]. Decay constants used are
0.0142 per 109 years for
ARb
and 0.00654 per 109
years for
ASm'
17. R. E. Harmer and M. R. Sharpe,
Econ.
Ceol.
80,
813 (1985).
18. D.
J.
DePaolo, Earth Planet. Sci. Lett.
53,
189
(1981).
19. T.
N.
Irvine, in Physics
<if
Magmatic
Processes,
R.
B.
Hargraves, Ed. (Princeton Univ. Press, Princeton,
NT,
1980), pp. 325-383.
20. D. N. Shirley,
J.
Ceol.
94,
795 (1986).
21. D. R. Baker, Con/rib. Mineral.
Petrol.
106,
462
(1991).
22.
In
our model, we assume tbat tbe coexisting magmas
achieve a steady-state condition
of
two-layer convec-
tion
on
a time scale tbat
is
short relative
to
tbe time
scale
of
crystallization. This
is
a particularly good
assumption for tbe Muskox intrusion, because tbe
GRZ
was already in a convective state when tbe final
7 FEBRUARY 1992
pulse
of
mafic magma was emplaced. For further
discussion
of
tbe relevant magma chamber dynam-
ics,
see C. M. Oldenburg,
F.
J.
Spera, D.
A.
Yuen,
and
G.
Sewell
[].
Ceophys. Res.
94,
9215 (1989)]
and
B.
D. Marsh
[].
Petrol.
30,
479 (1989)]; see
also discussion
of
Marsh's paper by
H.
E.
Huppert
and J.
S.
Turner
[ibid.
32,851
(1991)].
23.
To
maintain constant Sr concentration in tbe silicic
liquid, tbere must be a net flow
of
tbe otber Sr
isotopes
(88S
r,
86Sr,
and
84Sr)
into tbe silicic liquid
to
balance tbe loss
of
87Sr.
This can be ignored for
our
purposes because tbe total change
in
[
86
Sr
]a
is
only about 0.1% (see Eq. 2).
24. The boundary layer would in fact consist
of
botb
mafic and silicic components; however, diffusivities
of
Sr and N d in basalt are faster by up to an order
of
magnitude tban diffusivities in rhyolite
(5,
6), mak-
ing diffusion in tbe silicic portion
of
tbe boundary
layer tbe rate-limiting process.
25.
In
Eqs. 3 and 4, we use tbe approximation tbat tbe
concentration
of
Sr
is
independent
of
tbe
87Sr/86Sr
ratio
([
86
Sr
]/[Sr] = constant), which
is
generally
accurate to better tban 0.5%.
26. J.
c.
Jaeger,
Am.
J.
Sci.
255,
306
(1957).
27. T. N. Irvine, Can.
J.
Earth Sci. 7, 1031 (1970).
28.
S.
Clark, F.
J.
Spera, D.
A.
Yuen, in Magmatic
Processes:
Physicochemical Principles,
B.
O.
Mysen,
Ed. (Geochemical Society, University Park, P
A,
1987), pp. 289-305.
29. T. M. Harrison and E.
B.
Watson, Contrib. Mineral.
Petrol.
84,
66
(1983).
30. C. E. Lesher, Trans.
Am.
Ceophys. Union
72,
309
(1991).
31.
In
order
to
supply enough Sr
to
produce tbe increase
in
87Sr/86Sr
observed in C.U. 25, tbe
roof
zone
of
molten wall rock overlying tbe Muskox cumulates
must have been more extensive during crystallization
tban tbe current exposure would suggest. Two
possible explanations are tbat (i) tbe
GRZ
thickens
to
tbe north
of
tbe current Muskox exposure where
tbe intrusion dips beneatb sedimentary cover,
or
(ii)
much
of
tbe
GRZ
was vented during and after
crystallization
of
tbe mafic magma. The lack
of
a
downward-crystallizing
roof
sequence in tbe
Muskox intrusion, even in regions where wall rock
breccia
is
not
present, suggests tbat such a liquid
layer was present throughout tbe entire
roof
zone
during crystallization
of
tbe cumulates.
32. For contrasting views, see
H.
E.
Huppert
and R.
S.
J.
Sparks
[].
Petrol.
29,
599 (1988)] and
A.
N.
Halliday et
al.
[Earth
Planet. Sci. Lett.
94,
274
(1989)],
as
well
as
discussion by R.
S.
J.
Sparks,
H.
E. Huppert, and C.
J.
N.
Wilson
[ibid.
99,
387
(1990)].
33. J. M. Martinson, Contrib. Mineral. Petrol.
67,
233
(1978).
34. E.
B.
Watson, T. M. Harrison, F.
J.
Ryerson,
Ceochim. Cosmochim. Acta
49,1813
(1985).
35. We tbank T.
N.
Irvine for samples and field assist-
ance. The comments
of
R. C. Capo, R.
B.
Hanson,
F. M. Richter, F.
J.
Spera, and an anonymous
reviewer improved tbe manuscript. This work was
supported by NSF grants EAR84-15143 and
EAR88-04609.
11 October 1991; accepted 2 December 1991
COVER
Granophyre from the
roof
zone
of
the Muskox intrusion, Northwest
Territories, Canada, with a skeletal quartz crystal (violet,
-0.17
millimeter long)
surrounded by optically continuous vermicular quartz intergrown with alkali
feldspar (blue). The intergrowths reflect the
final
crystallization
of
a magma rich in
silica and alkalis that coexisted with an underlying silica-poor magma.
See
page 708.
[Photograph by Brian
W.
Stewart, California Instimte
of
Technology, with
cross-polarized light and a 575-nanometer retardation plate]
REPORTS
711