Rare-earth elements in the Skaergaard intrusion

Rare-earth elements in the Skaergaard intrusion

&ochimica etCosmochimica Acta,1068,Vol.32,pp.4%to447.Pergamon Press.Printed inNorthern Ireland Rare-earth elements in the Skaergaard intrusion LARRY ...

1MB Sizes 8 Downloads 55 Views

&ochimica etCosmochimica Acta,1068,Vol.32,pp.4%to447.Pergamon Press.Printed inNorthern Ireland

Rare-earth elements in the Skaergaard intrusion LARRY A. HASKIN and MARY A. HASKIN Department of Chemistry, University of Wisconsin, Madison (Received 3 Augu.st

1967; accepted in revked fmn ‘7 November 1967)

Ah&&-The chilled marginal gabbro of the Skaergaard intrusion, East Greenland, is very similar in absolute and relative rare-earth abundancesto other gabbros that have been studied. The rooks from the layered seriesthat were analyzed all had rare-earth contents similar to ‘that of the chilled gabbro, except for the specimenfrom the very top of the series,which was enriched more than ten times. The rare earths appear to have become increasinglyconcentrated in the magma as solidification progressed, but there was little change in relative abundances. No quantitative sympathetic variation betweenthe rare earths and any major element was observed. Some minor fluctuations in rare-earth abundances appear to be correlated with variations in feldspar oontents of the intrusion. It does not appear to be possible to explain the partition of the rare earths by the logarithmic theory.

UNDOUBTEDLY, the most thoroughly studied example of a layered series of igneous rocks produced by the slow cooling and fractionsl crystallisation of a basaltic magma is the Skaergssrd intrusion of East Greenland. A detailed petrologic study was fist made by WAGERand DEER (1939), who were able to estimate the compositions of the successive residual liquids during the course of fractional crystallisation. An improved determination of the changes in magma, composition during solidification has since been made by WAGER (1960). WAGERand MITCHELL(1951) studied in detail seventeen of the trace constituents of the magmrt snd their partition between the liquid and the solid, as evidenced by their concentrations in the rocks which were produced st various stages of crystal fractionation and accumulation. MCINTIRE(1963) compared the results of the analyses for the trace elements with theories of trace element partition. Of the rare-earth elements (REE), La and Y were determined by WAGER and MITCHELL(1951), but their data were scanty due to the inadequate sensitivity of their spectrographic techniques. The purposes of the present investigation are to determine the contents of the REE in rocks from the Skeergaard intrusion; to correlate, if possible, the changes in RE content and RE relative abundances of the rocks with changes in composition of the magma and with the gross composition and mineralogies of the rocks ; and to examine fractional crystallisation as a, contributing mechsnism for the production of the relative abundances of the REE found in the earth’s crust, where the light lanthenides are more predominsnt than they are in chondritic meteorites, in which the RE relative abundances characteristic of the solar system are possibly preserved (HASKINand FREY, 1966). Samples

For a detailed description of the Skaergaard intrusion, the reader is referred to the works of WAQER,referenced above, and WAGER and BROWN(1967). Briefly 433



the Skaergaard appears to have formed from a subsurface pool of basic magma (500 km3) whose solidification products can be classified as belonging to a layered series, a marginal border group, or an upper border group. The layered series, which comprises the bulk of the intrusion, has been divided into zones according to the mineral components of the rocks found in the various layers (WAGER, 1960). The lowest part of the series, the hidden zone (HZ), is inaccessible for study, and has been estimated to constitute the first 70 per cent of the layered series to crystallise. Estimates of the HZ composition are based on comparison with an earlycrystallised sequence of rocks in the marginal border group (e.g. @rites), and on extrapolations of chemical variation data. The lowest exposed rocks are in the lower zone (LZ), and comprise the crystallised fraction from 70 per cent to about 82 per cent; this zone is divided into subzones (LZa, LZb, and LZo, from bottom to top, according to the appearance of certain minerals). The middle zone (MZ), free from olivine, contains the rocks from about 82 per cent eryst~sation to about 94 per cent. The upper zone contains most of the final crystallisation products, and is divided mineralogically into subzones, with UZa containing the fraction from about 94 to 97 per cent crystallisation, UZb the fraction from about 97 to 99 per cent, and UZc approximately the final i per cent. The quantity of residual magmatic liquid which found its way into the formations of the do~~nward-crystallised upper border group, which rests over the layered series, is uncertain, but must not exceed a fraction of a per cent of the original magma. The marginal border group includes fine-grained olivine gabbros that are believed to be rapidly cooled, original magma, and whose composition is taken to represent that of the original magma. The rock samples examined in this work were supplied by Dr. G. M. Brown from the Wager Co~e~tion, Dep~ment of Geology and M~eralogy, Oxford University. A list and brief description of each, including, for those from the layered series, the estimated percentage of magma that had crystallised when the rock formed, is given as an appendix. k1NAZYSIS

The rock samples were received as powders of original analyzed material and were weighted directly into ~lyethylene tubes for neutron irradiation. The analyses were done by neutron activation, according to our usual procedures (submitted for publication). The average precision for duplicate analyses is usually & 5 per cent or better, and the average absolute accuracy is believed to be equivalent. RESULTS AXD DISCUSSION

The results of our determinations are given in Table 1. WAGER and MIrrcrr~nr., (1951) determined Y and La in Skaergaard rocks by a spectrographic technique. The sensitivity of their method was only 30 ppm for both elements, so they did not observe them in any whole rock samples from the layered series below UZc. They found 300 ppm La and 500 ppm Y in pyroxene separated from ferrogabbro’ * Referred to in this wlay in the earlier papers. Recently, a revised nomenclature (WAQE& 1960; WAUER and BROW, 1967) uses the term “ferrodiorite” to take account of the andegine pl&gioclaae.


Rare-earth elements in the Skaergaard intrusion

Table 1. Rare-earth element concentrations of Skaergaard rocks, chondritcs and sediments(ppm)

Y LS C0 PlNd Sm EU Cd Tb HO ElTm Yb &&+

margin81 gabbro

#4626 gabbro picrite

#6086 aumlllate from LZb

from MZ

from UZe

12.7 E+6 12.1 1.88 9.2 2.51 1.04 2.88 0.44 0.54 1.46 0.200 1.20 0190 65

10-2 5.7 11.5 l-81 8.6 2.18 0.71 2.37 0.39 0.41 1.17 0.182 0.93 0.156 49

9.8 7.6 14.1 1.96 8.4 2.09 o-91 2.16 0.31 0.44 1.26 0.191 1.06 0.173 63

8.3 2.80 6.5 0.95 4.5 1.37 073 l-74 o-30 0.39 1.01 0.168 0.98 0.138 31.

9.6 2.80 7.2 l-10 6.0 1.88 1.59 2:27 0.32 0.40 1.10 0,173 0.96 0.159 38

#4489 Composite #4330 trensgrsaive Composite of North oumulate bedenbergita of9 Americen from UZc grenophyre ohondrites* shales* 118 76-5 171 26 113 35 14.0 36.5 b-8 6-2 16.4 2.4 13.7 2.7 702

124 82.4 168 28.5 126 31 9.15 33.6 6.2 6.0 17.2 2.5 13.7 2.6 690

I.96 o-330 0.88 0.112 0.60 0.181 0.069 o-249 0.047 O-070 0.200 o-030 0.200 0.034 5.3

27 32 70 7.9 31 6.7 1.24 6.2 0.85 I.04 3.4 O-60 3-l 0.48 193

* HAWKINet al. (1967). t Sum of all mm-earths, inoluding estimate for Dy.

(upper UZc) and 150 ppm They found EG3649 (lower UZb) and 4000 ppm (upper UZc). Their analyses agree within about f 60 per cent. EG1881 EG414.5

(upper UZb).


La but no Y in pyroxene from ferrogabbro 1000 ppm Y and 300 ppm La in apatite from Y snd 1200 ppm La in apatite from EG4142 with ours for rocks from the same layers to

RE pattern

In order that the relative RE distribution pattern in the chilled gabbro (EG4507) can be simultaneously compared with the RE distribution for chondrites and that for a composite of North American shales, which is a good average for weathered crustal material (HASKINet al., 19668, 1967) the following procedure was used. The concentrations of the individual REE in ppm in the shale composite and in the chilled gabbro were divided, element by element, by the concentrations in ppm of the corresponding elements in chondrites. The resulting ratios were plotted on a logarithmic scale against RE atomic number as the abscissa (Fig. 1). Ratios for Y were plotted in the space belonging to Dy, which is not determined in our analytical procedures. Error bars represent standard deviation uncert&nties. These are based on duplicate analysis of the chondrites. For the shales and chilled gabbro, average errors for each element obtained from results on a number of other duplicate samples have been used (HASIEIN et al., 1967). From Fig. 1 it can be seen that the chilled marginal gabbro (No. 4507) and, therefore, presumably the Skeergaard intrusion as a whole, has a relative RE distribution intermediate between that of the chondrites and that of average weathered crustal material. In absolute abundance, the element Lu is about 6 times more concentrated in the Skaergeard chilled gabbro than in chondrites, and about 0.4 times as concentrated as in sediments. The La concentration of the gabbro is 17 times that of the chondrites, and only about O-17 times that of the shales. The Skaergaard gabbro is very similar to the analyzed Duluth gabbro





(FREY and HASKIN,to be published) and the San Marcos gabbro (TOWELL et cd.. 1965), both in absolute RE content and in RE pattern. It is anomalously enriched in Eu to about 1.2 times the value interpolated between the contents of the neighboring elements Sm and Gd, and depleted in Ce to O-77times or less the extrapolated value. The absolute RE content of the Skaergaard gabbro is 56 ppm, compared to 59 for the San Marcos gabbro and 50 for the Duluth gabbro. This is in significant contrast to the probable chilled gabbro from the Stillwater layered complex and a

Fig. 1. The RE patterns of the Skaergaard chilled marginal gabbro (EG4507) (separatepoints) and of a compositeof shales (upperdashedline) are comparedwith the pattern in chondritic meteorites.

possible chilled gabbro from the Bushveld complex, which have RE patterns that are relatively less rich in the light REE than are chondrites, respectively only 16 ppm and 6 ppm total REE, and anomalous Eu enriohments of 4-5 times the interpolated values (FREY and HASKIN,to be published). Y appears to be depleted in the Skaergaard chilled gabbro, relative to its abundance in chondritic meteorites. This apparent depletion may be real. This is uncertain, however, because the ratios shown in Fig. 1 for Yb and Lu are not lower than that for Y, and Y appears to follow these elements closely in some geochemical processes (HASKINet al., 1966b). The RE patterns for the rocks from the Skaergaard series, plus those for the picrite and the granophyre, are compared with that of the chilled gabbro in Fig. 2. To obtain the comparisons in Fig. 2, the concentration of each REE in each rock was divided directly by the concentration of the corresponding REE in the chilled marginal gabbro, rather than by the concentration in chondrites. Error bars of about f 7 per cent have been assigned to these ratios. The & 7 per oent is a representative average standard deviation uncertainty, and in the absence of dnplioate analyses, care must be exercised in interpreting any small analytical deviations as indicating genuinely anomalous behavior for an element. The RE patterns of the rocks of the layered series are all somewhat different from that of the chilled marginal gabbro, but perhaps their most striking feature is their considerable similarity to that presumed for the parent magma, and the absence

Rare-earth elementain the Skaergaardintrusion


of any strong systematic trend of relative depletion or enrichment of light or heavy REE with increasing extent of magma solitication. There may be a slight trend towards relative enrichment of the light REE in the melt, as inferred from the pattern of gabbro EG4330 from the top of the layered series, but nothing as marked as the trends found for specimens from the Southern California Batholith (TOWELL I










































Fig. 2. The RE patterns of the Skaergard rocks are compared with that of the chilled mazginal gabbro (EG4507).

et al., 1965) or in Hawaiian basalts (SC~ILLIN~and WINCHESTER, 1966). These latter systems are, of course, not strictly analogous to the Skaergaard intrusion, but both could be plausibly explained, in part at least, by the fractional crystallisation of basic magma. Among the Skaergaard samples analyzed, the pattern most different from that of the presumed average is found in EC5086 (LZb), in which the Tb content is O-7times the Tb content of the chilled gabbro, while the La content at the other extreme is l-4 times the La content of the chilled gabbro. In the latest geologically obvious member of the layered series to have crystallised (EG4330, UZc), all elements except Y have been enriched to between 11 and 16 times their concentrations in the chilled gabbro, indicating that the relative RE distribution of the last liquid had not changed much from the original.


Lannr X. HASKINand MARYA. H~skrr

There appears to be some trend of enrichment of Eu relative to Sm and Gel in the sequence from LZb (EG5OS6} to MZ (EG4427) to UZa (EG~lSl}, but it is not apparent in UZc (EG4330). Certainly, Eu is varying in relative abundance. presumably due to its reduction to the +2 oxidation state. Anomalous behavior for Eu is frequently observed (TOWELL et al., 1965; SCHILLING and WZNCHEST~R, 1966; HASKINet al., 1966b). The element Y is anomalously depleted to 0.85 times the interpolated value in the hedenbergite granophyre (EG4489) and to about O-80 times the interpolate~l value in EG4330 (UZc); it may be enriched to about I.15 times the interpolated value in EG4427 (MZ). We have not previously observed such strongly anomalous behavior for Y in igneous rocks, and are unable to offer a good explanation. As previously discussed, Y usually behaves like a typical heavy lanthanide, but not always like any one specific heavy lanthanide, and it is not a lanthanide, lacking the p~tially-pled 4f electron subshell (HASKINet at., i966b). RARE-EARTHCONTENTS OFTHELIQUID The RE contents and patterns of the liquid at various stages during the solidification of the magma can be estimated if the rocks examined for the layered series are presumed to be representati~~eof the zones from which they came, if the composition of the chilled gabbro is presumed to represent accurately the composition of the original magma, and if relative volumes for the various zones are assumed. We have used essentially the same assumptions regarding the significance of the chilled gabbro and the relative volumes as made by WAGER(1960). The number of samples analyzed is certainly minimal in terms of obtaining an adequate representation of each zone, and this is a limitation of the accuracy of our estimates. However, we do not believe this to be a serious limitation, since the particular specimens analyzed were carefully selected as being of average character in terms of crystal accumulation and gross composition ; since the absolute RE contents of the rocks from LZb to UZc vary by less than a factor of 3 ; and since there is no strong variation in relative RE abundances among all the rocks examined. In estimating the absolute RE content of the upper zone, which was derived from the last 6 per cent of the liquid, an interpolation was necessary. Rock EG5181, taken as representative of UZa, the first half of the upper zone to crystallise, is similar in RE content to the rocks of the middle and lower zones. Rock EG4330, from the top of UZc. is nearly 20 times richer in REE than EG5181, aud was taken ai representative of the last 1 per cent of the layered series. We did not have for analysis any sample from UZb, the approximately 2 per cent of the layered series preceding UZc. UZb is characterised by high P,O, contents and the first appearance of considerable apatite in the rocks. Apatite often has high RE contents, but has not been found to contain more than a few per cent of all the REE in basic rocks (HASP et al., 1966b). The onset of apatite erystallisation in the Skaergaard intrusion was not accompanied by a discontinuous increase in the RE contents of the rocks. The data of WAGERand MITCHELL (1951) show clearly that the RE contents of UZb rocks lie between those of UZa and UZc. ‘Ihey found Y and La in 3 whole-rock samples from UZc, but none in 2 whole-rock samples from UZb or in samples from


elements in the Skaergaard intrusion


lower horizons of the layered series. They found 4 times as much Y and La in a sample of apatite from UZc as in a sample from UZb. They found 300 ppm La and 500 ppm Y in pyroxene from UZc, 150 ppm La and no Y in pyroxene from a rock at the top of UZb, and no La or Y in pyroxene from UZa. For the estimates of residual liquid concentrations, we have taken the RE concentrations of UZb as equal to a mixture of 2 parts of EG6181 (UZa) plus 1 part of EG4330 (UZc). These concentrations are in line with the observations of WAQER and MITCHELL (1951) on UZb rocks. They are intermediate between those used for UZa and UZc, but because the absolute RE content of EC4330 (UZc) is about 20 times greater than that of EGSlSl (UZa) the RE pattern of UZb is dominated by that of EG4330. I















Fig. 3. The RE patternsestimated for successive residue1 liquids (curves above dashed horizontal line) and for the average of the Skaergaard hidden zone (HZ) (curvebelow dashed line), are compared with the RE patternof the chilledmarginal gabbro. Percentagesrefer to extent of solidificationof the originrtl magma. The last 1 per cent of the liquid was taken to be identical with EG4330 in RE content and pattern. Two parts of the mixture described above for UZb were added to one part of EG4330, and then divided by 3 to obtain the RE content of the last 3 per cent of the liquid. Three further parts with the RE concentration of EG6181 (UZa) were added to the previous sum and divided by 6 to obtain the RE content of the last 6 per cent of the liquid. Twelve parts with the RE concentration of EG4427 (MZ) were next added in ; finally, 12 parts with the RE concentration of EG6080 were added to the previous sum to give 30 parts representing the last 30 per cent of the magma to crystallise. The last sum was subtracted from 100 parts with the RE concentrations of the chilled gabbro (EG4607), and then divided by 70 to give an average RE content for the unexposed, presumed 70 per cent of the intrusion (HZ). The RE patterns and contents of the residual liquids are compared with that of the chilled gabbro in Fig. 3. To obtain this figure, the RE contents calculated for each stage of solidification were divided by the absolute contents of the corresponding REE in the chilled gabbro (EG4607), just as was done to produce the comparisons in Fig. 2. The lowest curve in each figure is that for the calculated average for the HZ. The RE distributions of the liquids are similar to each other and to the distribution for chilled gabbro. Fluctuations in the liquid RE contents and patterns,






A. Hasmx

caused by the particular patterns of the rocks analyzed from the lower zones, are seen. The magma at 70 per cent ~rystallised is enriched in the REE only about f-5 times over its original content, and the average RE content for HZ rocks is about 80 per cent that of the chilled gabbro and quite similar to it in RE distribution, very much in line with the RE contents and patterns actually observed for rocks of the intrusion. The magma becomes increasingly enriched in the REE right up to the final stage. The element Y appears slightly enriched in the HZ rocks and anomalously depleted from the liquid phase. Similarly, Lu appears to be anomalously depleted in the HZ, and enriched in the liquid. Both of these features are merely reflection of the RE distribution of EG4330, which was used for UZc and which dominates UZb, and are local phenomena. There is no evidence for a significant enrichment in Lu or a depletion of Y in the other rocks of the layered series which we have examined. Such strongly anomalous enrichments or depletions in the liquid would surely appear in at least some of the intermediate rocks. The apparent enrichment of Lu in EG4330 may be experimental error, since Lu is sometimes difficult to determine. Because there are no anomalies for these elements in the samples from LZ, MZ and UZa, the anomalies in UZb and UZc must be compensated in the calculation by ~orrespon~ng, opposite anomalies in the HZ. The apparent Eu depletion in the HZ may be real, since there is some tendency toward Eu enrichment in the rocks from LZ, MZ and UZa. The entire amount of REE in the average HZ rocks cannot be accounted for by trapped liquid, because the RE contents of HZ rocks are too close to that of the liquid from which the rooks are forming. The picrite (EG4526) from the marginal border group has m~eralogieal and gross compositional characteristics that are similar to those inferred for the earliest rocks (HZ) of the layered series (WAGER and MITCHELL, 1951). The actual RE contents and distribution of the picrite are similar to our estimates for the average of the rocks of the HZ. PARTITION OF THE REE


In order to illustrate better the partition of the REE between the solid and liquid during the course of crystallisation, the concentrations in ppm for several of the elements in the rocks of the layered series have been plotted against percentage magma solidified in Fig. 4. The average calculated concentrations for HZ have been plotted also, as lines extending from 25 to 50 per cent solidified; presumably, somewhere along that line the plotted value is equal to the true value for the rocks. For La, the HZ has 0.76 times the average value for the layered series, the LZ has 1.6times the average, and the MZ and UZc values are 0.5 times the average. There is an abrupt rise in La concentration for UZc to over 13 times the average value for the series. Approximately the same trend of relative concentrations is found for all the elements, but the changes in concentration are not as great for the heavier REE as for La. Yb, for example, varies only from 0.80 times the series average for the HZ and UZa to 0.94 in the LZ. MCINTYRE (1963), using the data of WAGER and MXTCRELL(1951), attempted to use the logarithmic law of trace element partition to explain the observed liquid

Rare-earth elements in the Skaergaard intrusion


and solid concentrations for trace elements in the Skaergaard intrusion. The logarithmic theory presumes that the attainment of true thermodynamic phase equilibrium between a residual melt and its solid products is prohibited by kinetics for a body the size of the Skaergaard intrusion. Rather, equilibrium is expected to obtain only between the residual liquid and the surface layers of any crystals in contact with it. This is equivalent to assuming that the concentration of a trace element entering the solid phase at any instant is proportional to the concentration





Fig. 4. The concentrations (in ppm) for several of the REE are plotted against percentage solidification of the Skaergaard magma. Only the concentrations of La (left ordinate) and of Yb (right ordinate) can be read directly from the plot, but the ordinate scale is the same for all the elements. Similarly, the absoissa is not marked, but for each element there is a point at 35, 74, 89,995 and 99 ‘A solid& oation of the magma. The purpose of this figure is to show the trends of concentration change. The approximate percentages of plagioclase in the different levels of the layered series are shown superimposed on the curve for La.

of that trace element in the liquid at the same instant. The proportionality constant is called the partition coefficient, designed by A in the equations below, which describe this type of crystallisation behavior: c= = P/n = C”(1 - x)“-1.


Equations ( 1) are taken from equations (45) of MCIN!rmE( 1963), and CLrepresents the concentration of the trace element in the liquid, Cs the concentration in the solid, CA the average concentration for all the solid plus liquid, and z the fraction of the original liquid that has solidified. MCINTIRE(1963) found that, for the trace elements Co, Ni, Mn, Ga and Ba, the agreement between the logarithmic theory and the observed trace element partition was good. The theory did not yield the observed values for Li, Zr and SC unless relatively small changes were made in their concentrations in the residual liquids, as estimated by WAGJER and MITCHELL (1951). To obtain agreement for Sr and V, the values of the partition coefficients had to be considerably altered with increasing



solidification of the magma. (All the values actuahy required slight changes, but that is to be expected as the T, P and composition of the liquid and crystals change.) The logarithmic theory does not fit the observations on REE in the Skaergaard rocks and calculated liquids. Yb is given here as an example. In Fig. 5 we have plotted the observed Yb concentrations (square points) for the layered series rocks as in Fig. 4, and with them the calculated Yb concentrations in the residual liquid (round points) at various stages of soli~fication. The upper solid curve represents the best fit of equation (l), using 1 = O-55, to the estimated liquid concentrations. The agreement is poor. The estimate for the concentrations in the solid (lower solid curve) is even worse. The logarithmic law, for values of Iz less than 1.0, is


Fig. 5. The coneentrat~ons of Yb iu ppm for the varioutr XO& of the kyered series (square points) and the estimated Yh contents in ppm for the residual magma (ciroubx points) are plotted againat peroentage soliditiaetion of the magma. Predictions of the logarithmic law of trace element partition for the Yb content of the liquida (upper solid curve) and for the solids (lower solid aurve) are shown. Frediotiom based on a model of constant Yb content for the solid (horizontal dashed line) and ctomponding incrersse of yb oon~ntr~tion in the liquid (dashed curve) are also shown.

doomed to predict a steady increase in trace element concentration as solidification progresses, and such a steady increase was simply not observed for the REE. The equation can be made to match the observations on Yb quite well if the values of 2 are approF~ately altered with increasing amount soli~fied, but the required changes are not small and are entirely arbitrary. Use of the logarithmic theory to explain the partition of the REE between the crystallising solids and residual liquids of the Skaergaard layered series is simply not supported by the data. A better approximation for all but the last dregs of liquid is to arsSumethat the RE eonoentration in the solid phase is constant thronghout t&o the magma, [email protected] of the liquid ~n~nt~tion, as if the only a certain, fixed number of sites available for the [email protected], and thme were always being filled. The appropriate equation for this behavior is: CL = cs + (CA - @)/(I

- X),


Rare-earth elements in the Skaergeard intrusion


where the symbols have the same significance as in equations (1). Equation (2) is a completely general equation describing the mass balance of the trace element in the entire intrusion. A constant RE concentration for the solid phases corresponds to holding the value of @ constant for all values of x. This approximation is, of course, destined to break down at a very high fraction of solidification, since it predicts an infinite concentration for the liquid. It is not a bad first approximation, 80A




ii :I i ‘_

IO2 E ::

/’ 1

,i’ LIauID_\,i’

0.5 ___;_tL&&+ -------




Fig. 6. The concentrations of La and Ho in the solid and liquids (caloulated) and the predictions of the model based on constant RE content in the solid are shown as in Fig. 5.

however, to the observed solid and estimated liquid concentrations for the heavier REE in the Skaergaard layered series. The upper dashed curve and dashed horizontal line of Fig. 6 show the predictions of equation (2) for Yb. (The dashed line at 1.0 does not represent a normalisation ; the average concentration of Yb just happens to be 1-Oppm.) In Fig. 6 the results of equation (2) for a second heavy REE (Ho) and for a light REE (La) are shown. The agreement for Ho is again good. La is the least favorable case that could be chosen, and the observed RE concentrations for the solid can hardly be said to be predicted accurately by equation (2), but the agreement with the estimates for the liquid is still fair. CORRELATIONSBETWEEN THE REE


The fluctuations of RE concentration among the rocks of the intrusion, which are considerable for the lightest REE but small for the heavy REE, can be compared with behavior of the major element constituents, using the data of WAGER (1960).



Only two elements, Si and K, behave qualitatively like any of the REE ; their concentrations in the solid phase remain essentially constant with increasing degree of solidification until the last rocks are formed, then they rise sharply, much as do those of the heavy REE. The similarity in behavior between the REE and Si is strictly qualitative. The RE contents change by a factor of 15 while the Si content increases only from 22 to 27 per cent, a factor of about 1.25. The relatively large RE ions would not be expected to substitute for the much smaller Si ions, anyway. The rise in K content is much greater, but there is no exact correspondence between K and the REE. For example, the ratios ppm La/percentage K and ppm Y/percentage K in EG5181 (UZa) are 12 and 43, but in EG4330 (UZc) they are 92 and 143, hardly reassuring in terms of a potential K-REE variation diagram. The semiquantitative agreement, at best, between K and the REE might be construed to imply that the REE were substituting for K in the solid phases forming. It might also be construed to imply that the REE were somehow being held in the liquid as complexes with K. The most plausible explanation seems to us merely that neither the REE nor K are welcome constituents in the crystallising phases, so that both become strongly concentrated in the liquid as crystallisation progresses. There is no relationship between the RE contents and the P,O, contents of the rocks. For example, the ratio La/P,O, in rock EG4330 (UZc) is 3-O x 10-2, but for EG5181 (UZa) is only 3.1 x 10-3; the average for MZ and LZ rocks is 6.5 x 10-3. If the same La/P%O, ratio were present in UZb as is present in EG4330 (UZc), then the amount of La in UZb would be nearly twice the amount believed present in the entire layered series. The REE are not being induced in any obvious way to substitute for any of the major constituents in the solids. There is no simple variation between any of the REE and any of the single, major element constituents. INFLUENCE OF MINERAL PHASES ON RE CONTENTS The kinds of minerals crystallised from the Skaergaard liquid do not change greatly from the lowest accessible rocks (LZ) to near the top of the intrusion although their compositions change considerably. That the appearance of large quantities of apatite in the rocks of UZb, followed by lesser quantities in UZc, does not account even qualitatively for the observed behavior of the REE has already been demonstrated, but can be further elaborated here. WAGEXXand MITCHELL (1951) gave no partition coefficients between the liquid phase and apatite. Using their data for Y and La in apatites from EG3649 (UZb) and EG4142 (UZc), with our estimates for the liquid concentrations of these elements at the same per cent of solidification, we get partition coefficients of 17 for Y and 7 for La (EG3649) and of 41 for Y and 17 for La (EG4142). The only new mineral phases whose appearance accompanies the suddenly high RE concentrations in rocks at at the top of the layered series are those of micropegmatite (WAQER and MITCHELL, 1951). It is difficult to imagine that this material could be responsible for the high concentrations of the REE in these rocks, since it consists of quartz and feldspar, neither of which has previously been observed to contain the required amounts of REE (HASKIN et al., 1966b).

Rare-earth elements in the Skaergaard intrusion


There is one possible qualitative correlation between RE concentrations in the layered series rocks and their mineral contents. Except in the last few per cent of solid to form, the proportion of plagioclase in the rock varies roughly in the same sense as the contents of the lighter REE. This sympathetic variation can be seen in Fig. 4, where the average percentage of plagioclase in the rocks at various levels of the layered series (WAGERand MITCHELL,1951) is plotted on the same abscissa as the La contents. The accuracy of these average plagioclase contents is subject to some reservation; estimates are difficult to make because of mafic-felsic layering (M. BROWN,private communication). Previous results for basic rocks have shown that plagioclase is not a strong concentrator of the REE (TOWELLet al., 1966; H&KIN et al., 1966b). WAGERand MITCHELL found no La or Y in separated plagioclase from their Skaergaard samples. Plagioclase is a somewhat selective mineral for the REE, however, in that it tends to favor admission of the lightest lanthanides over the rest of the group, and it often contains a high proportion of all the Eu in the rock, presumably as Eu2+. Thus, if variations in plagioclase were to affect significantly the RE contents of the rocks, the strongest sympathetic variation with plagioclase contents should be with the lightest REE, La-Nd and Eu. That is exactly what is found in the rocks of the Skaergaard intrusion. If plagioclase content is responsible for the observed fluctuations in RE content, however, a decrease from about 60 per cent in LZ rocks to about 48 per cent in MZ rocks causes a change in La content from 7-6 ppm (EG5086, LZb) to 2.8 ppm (EG4427, MZ). Quantitatively, the changes in the concentrations of some of the heavy REE are more similar in magnitude to the fluctuations in the plagioclase contents. Hedmbergite


The RE contents and pattern of the granophyre (EG4489) are nearly identical with that of EG4330 (UZc), which is the last geologically obvious differentiate of the layered series, and which lies only a few meters below the level from which the granophyre was taken. Thus, it is tempting to conclude that the REE in the granophyre came from the final liquids of the layered series. There is one major difference between the RE patterns of the granophyre and the gabbro from UZc; relative to the chilled gabbro (EG4507) the granophyre is deficient in Eu, down to O-73times the interpolated value, while in the gabbro from UZc the Eu concentration is not anomalous. This is not necessarily significant, except that the apparent deficiency of Eu in the granophyre just happens to match exactly in magnitude the relative Eu enrichment of the chilled gabbro and the gabbro from UZc, as compared to sedimentary materials. Thus, the relative abundance of Eu in the granophyre is the same as that for our best estimate of the average relative RE distribution for the earth’s crust. This may be strictly fortuitous. In all other aspects, including a small relative depletion in Y, it seems entirely natural to consider that the REE in the granophyre came from liquids of the layered series. Assimilation of more than a few per cent of typical sedimentary or igneous material during formation of the granophyre would be expected to have altered its RE distribution noticeably away from that of the chilled gabbro or the gabbro from UZc, especially for the lighter REE. HAMILTON(1963) found that the Sf17/Sr86ratio of O-7104 for this granophyre (EG4489) was too high to be compatible with its formation by simple differentiation 6




from the Skaergaard magma (e.g. Sr87/Sr86of 07063, marginal border group, and O-7069, layered series). Conta~nation by a material relatively rich in SF seems to be required. The nature of the contaminant was apparently such, however, that even extensive (50 per cent) assimilation of country rock was probably insufficient to produce the high SrS7/Srs6ratio. Our data suggest that any assimilated material had either a very low RE content or relative RE abundances remarkably close to those of the bulk of the Skaergaard rocks. SUMMARY

The principal conclusions that we draw from this study are the following: (1) The Skaergaard successive liquids became steadily enriched in the REE during solidification of the layered series. (2) The fractional c~s~~sation of the original Skaergaard magma did not give rise to any strong trend of RE pattern change as so~~~tion of the magma progressed. (3) There is no sympathetic variation between the contents of the REE and that of any of the major element constituents of the Skaergaard rocks. (4) The logarithmic law of trace element partition does not give a good estimate of the behavior of the REE in the Skaergaard intrusion. (5) There is nothing in the RE contents or patterns of the Skaergaard rocks that refutes the estimates and assumptions of WAC+B~~~R (1960) regarding the size of the various zones and the average composition of the intrusion. Achmowlledgmenta-Weare very grateful to Dr. G. M. BROWN for furnishing the samples of Skaergaardmaterial and to him and Dr. F. &SEX for suggestingimprovementsin the manuscript. We thank the crew of the University of Wi~oo~~ nuclear reaotor for the neutron patios. We acknowledgegratefully the partial support of this work by the National SoienceFoundation under grant no. GA-498 REB~:RENCES Bnowx G. M. and VINCENTE. A. (1963) Pyroxenes from the late stagea of fractionation of the Skaergaardintrusion, East Greenland, J. Petrol. 4, 176-97. cORyEr& C, D., C-E J. W. and W~~~ST~R J. W. (1903) A procedure for geochemicai interpretationof terrestrialrare earth abundanuepatterns. J. Ue~~~~. [email protected], 559-566. HAMXLTON E. I. (1963) The isotopic composition of strontium in the Skaergaard intrusion, East Greenland. J. Petvvl. 4, 383-91. &SKIN L. A., Hasxx~ M. A., FREY F. A. and WILDEXANT. R. (1968) Relative and absolute terrestrialabundanoesof the rare earths. Proo. &ymp. Origin,a~& AbundancRe of the Elemeat8, Paris, May S-11, 1967 (editor L. II. Ahrens), in press. IF88xrx L. A. and FREY F. A. (1966) Dispersed and not-so-rearths. S&em% l&&299-314. HASKINL. A., WILDEMAXT. R., Foxy F. A., COLLINSK. A., KEEDY C. R. and EASJCJX M. A. (1966a) Rare earths in sediments. J. [email protected] Re8. 71, 6091-6105. 5x1~ L. A., FREY F. A., SCHMITTR. A. and SMXTHR. H. (196613)Meteoritic, solar and terrestrialrare-earth distributions. Phgs. Ohem. Earth 7, 167-321. MC&TIRE W. L. (1963) Traoe elementpartition ooeffloients-A review of theory and applications to geology. #eochim. Gbmochina. Acta 27,1209-1264. ~CHILLTNU J. G. and WINGED J. W. (1966) Rare earths in Hawaiian baaal&. &$ence 1% 867-Q. TAYLOR H. P., JR. and EPS~ZXNS. (1963) 01*/016 ratios in rookaand coexisting mineraleof the SkaergaardIntrusion, East Greenland,J. PetroE, 4, 61-74. TOWELL D. G., WINCHESTIER J. W. and VOLROVSKY R. (1965) Rare-earth distribution in some rocks and assooiated minerals of the batholith of Southern Cdif~mia. J. Ge0phy.s. Res. 70, 3485-96.

Rare-easth elements in the Skaergaard intrusion


WAGERL. R. (1960) The major element vari8tion of the layered series of the Skaergeard intrusion snd 8 reestim8tion of the average oomposition of the hidden layered series and of the successive residual nmgm8s. J. Petrol. 1, 364-93. WAUERL. R. and BROWN G. M. (1968) Layered IBccke. Oliver 8nd Boyd. In press. WAGERL. R. and DEER W. A. (1939) Geological investigations in E8st Greenland. Pert III. The petrology of the SkaergElard intrusion, Kangerdlugssuek. Medd. G%nknd 106,l-362. WAQERL. R. and MITCHELLR. L. (1951) The distribution of tr8ce elements during strong fraction8tion of basic magma-A further study of the Skaergaard intrusion, East Greenland. Beochim. Cosmochim. Acta 1, 129-208.



EG4607: Chilled olivine gabbro from the mctrginctlborder group. This rook is described by WAGER (1960), who classified it 8s the best 8v8ilable representation of the composition of the Skaergeerd megma, at the time of intrusion. EG4626: Picrite from the marginal border group. This rock, from its minemlogy, is presumably similar in comparison to the earliest-formed rocks of the hidden layered series. Similar picrites were discussed by WAUERand DEER (1939) and WAQER 8nd MITCYEELL(1951). EG6086: G8bbro from the lower zone (LZB) of the layered series, approximately equivalent to EG4077 (WAGIER and MITCHELL, 1961; WAUER, 1960). Cumulate of pl8gioclase, olivine, augite end inverted pigeonite, at about 74% solidification of the m8gm8. EG4427: Olivine-free g8bbro, from the middle zone (MZ) of the layered series, 8pproximately equivalent to EC3661 or 3662 (WAQER and MITCHELL, 1951; WAUER, 1960). Cumulate of plagioclaae, augite, inverted pigeonite 8nd mctgnetite, at &bout 89 % solidification. EGBlBl: Ferrodiorite, from the upper zone (UZa) of the lsyered series, described by WAGER (1960). Cumul8te of plrtgioclase (8ndesine), ferroaugite, iron-olivine and magnetite, at about 97 % solidification. EG4330: Ferrodiorite, from the upper zone (UZc) of the layered series, and the last geologiaslly obvious differentiate of the intrusion (WAQER, 1960; BROWN and VINCENT, 1963). Cumulate of Bndesine plagioclsse, ferrohedenbergite, fayalite, msgnetite and apatite, at about 99 ‘A solidificcttion. EG4489: Transgressive hedenbergite granophyre, believed to have been expelled from between the uppermost layered series and the lowermost upper border group. Presumably derived from the last liquids of the magme, but possibly with consider8ble assimilated material (TAYLOR and EPSTEIN, 1963; BAMILTON, 1963). Note added irt proof. Dr. KENT BROOKS has kindly provided us 8 ertmple of rock EC4272 from the lower half of UZb. This rock contains nectrly 3 per cent P,O,. Preliminary analysis by an instrumental activation method yielded the following approximate values (~60 per cent): L8, 16 ppm; Sm, 11 ppm; Eu, 4.5 ppm. The values we used for the 8verage contents of UZb in the calculations in the paper were: La, 27.5 ppm; Sm, 12.9 ppm; Eu, 6.7 ppm. These estimates, besed on the date of WAQER and MITCHELL (1951), are thus consistent with the values obt8ined from our preliminary analysis and con&m our conclusion that the sudden onset of 8patite crystallisation ~8s not accompanied by 8 similctr, discontinuous rise in the RE contents of the rocks.