Notes 237

etcoemochidca Acta,1970,
Qe~~hlmica Vol.34,pp.237to243. Pergamon
Press.
Printedin
Northern Ireland

Trace element fractionation during anatexis

DENIS M. SHAW
McMaster University, Hamilton, Ontario, Canada

(Received 14 April 1969; accepted in revisedform 24 September 1969)

Ab&&--The theoretical basis of trace element fractionation presented in a paper by P. W.

GAST (1968) can be developed in a more flexible algebraic form, which is here described and
illustrated.

THE RECENT article by P. W. GAST (1968) is most stimulating and thought-provoking,

in its interpretation of mantle-crust megmatic trace-element trends on a more quanti-
tative basis. Attempts by my colleagues and myself to analyse this paper in detail
have indicated some possible improvements to the algebraic analysis. The Crst
point at issue is his derivation of an expression for the bulk distribution coefficient
for fractional melting of a multiphase solid.
Using GAST'S symbols (p. 1062), but omitting the suffix which refers to element a,
the total concentration of an element in an initial assemblage consisting of mass
fractions X”, Xs etc. is COT

coT = coaXOa+ coax,@ + c,“X,~ + . . . where IZX’ = 1

which may be rewritten as

This is GAST'S equation 5.

Now let a small weight fraction Ax* be formed, by melting the small amounts xa,
2s etc. of each phase. Let the concentration of the element in the liquid be cl and the
concentration of the element in the (equilibrated) remaining solids be cT.
Then
(1 - Axz)cT = coT - Ax” . cz
and also
cr = c”(X,Oc- 2”) + cP(X,B - x@) + . . .

= c= (X0”- %“I+ x,s - xg x,y - y

p,p + K"'Y +*..I
z x0=-x” xp - f X0? - %’
=c + KZ/aK=l@ + +
KU” K”/“K=/Y ’ ’ ’
1

Detmfug the bulk Nernst distribution coefficient as DT” = cT/czthen

238 Notes

which differs from GAST’S equation (7). His expression defines DTll as the ratio of
the element concentration in the whole system to that in any given liquid, i.e. cT,,/c’.
It is not quite clear from the context how GAST’S equation (7) was used in
subsequent machine computations. However the numerical examples used to
illustrate equation (7) seem to be erroneous.
For example, using the example given (GAST, 1968; p. 1063) where X” = Xfl =
0.5, xa = xa = 0.003, K z/e= 10 and KU’S = 1, then calculation of D1jT from equation
(7) gives 9.5, not the value of 12.3 given by CAST. Use of my equation (1) gives 10.1.
GAST’S second example (same paragraph) should calculate by his equation (7) to the
value DziT = 5.4, although the result given is 3.6; my equation (1) gives 5.6.
The main point of this Note is, however, the derivation of satisfactory expressions
for trace element behaviour arising from various fractional melting models. The
expressions used by SCHILLIN~ and WINCHESTER (1967) and by GAST (1968) in his
equations (8)-( 13) are unnecessarily complicated, because restricted to discontinuous
melting, and can be replaced by continuous variables advantageously.
Let us assume an initial solid of mass W,,g, comprising masses WO”g of each
phase i:
IV, = was + w,p + . . .

the weight fractions are Xi where

1 = xoa + x0@ + . . .

The element in question comprises w0 g of the initial solid, with initia,l concen-
tration c0 = w,/ W,. During fractional melting suppose each phase contributes a
fraction pi of the liquid which amounts to L g, where

1 =pa -l-p@ +. ..

The proportion of liquid is F = L/W,,, and if the concentration of the element in

the liquid at this point is cz then
cl = c”/DaJt (2)
where c8 is the concentration of the element in the residual solid and DSfl is the bulk
distribution coefficient, referred to simply as D in the following.
Then

c’ = w8’Iwb - w.-
w, -
wz
L

where w’ g is the aggregate weight of the element in the L g liquid formed up to this
point
Thus
1 w()-wl
cl=-____
(4)
DW,-L

and since the mass of liquid forming instantaneously is dL g, containing dw” g of the
element, then
dw’
cl = FL
 Notes 239

and
dwl 1 w()--wl
- =-_.p. (5)
d.L D W,-L
Since we also know that

XS
= cn X” + ml? + . . . where Ku/S= ca/detc.
1

= & C X” + g&+ ...1

then

(6)

Combining (5) and (6) we obtain

dwl wg - wl Xb
-= ____ +... (7)
d.L W, - L K'/"K"/P

Equation (7) is general, and requires additional assumptions before it can be

integrated.
The first case is where the phases melt in the proportions in which they occur in
the rock, i.e. p” = Xsa etc. Then D is a constant (Do), and the variation of con-
centration in the liquid cz is given by

CZ
-=- ; (1 - p)w (8)
CO 0
Similarly

$ = (I - p,&A
(9)

Equation 8 assumes that each fraction of liquid is separated continuously from

preceding liquids. The aggregate liquid composition is Cz and is given by

EZ 1
-=- F 1 - (1 - Pp .
[ 1
(10)
CO
If by contrast the liquid remains at all times in equilibrium with the residual solid we
can obtain the different expression cL for its concentration:

CL 1
-_=
(11)
CO Do + F(1 -Do)'

Expressions 9, 10 and 11 are shown in Fig. 1 for the hypothetical case where Do = 4.
If, as mentioned earlier, different phases melt in proportions pa etc., then D is a
240 Notes

1.

Relative

Concentration

1.

0~4

Fig. 1. Relative concentration of a trace element in the liquids produced by

varying degrees of fractional melting (P) of a multiphase solid, where the phase
proportions in the solid do not change throughout melting. c1 is the concentration
in a liquid momentarily in equilibrium with the residual solid but thereafter
isolated: Ezis the aggregate concentration of fractions described by cz: cL isithe
concentration when the liquid remains at all times in equilibrium with the residual
solid. D is the bulk distribution coefficient.

function of F and equations (6) and (7) need modifying. Since the initial value of D
is D,, where X” = X,a etc., then it can readily be shown that

Ds,Z = D = WIJ~O - Lp Do - pp
= l--F
(12)
w, - L
where
cl
PS
’ = j$ + KZ/“K”@ + *-*

Substituting equation (12) in equation (5) and integrating, we obtain the following
expression for ~1, analogous to equation (8)) where the liquid is continuously separated

(13)
 Notes 242

The expressions analogous to Ezand cL follow:

(14)
and
CL 1
-_=
CO Do + P(1 - P) l

Similarly
c8
-= D .-cz-2&l -gyp
CO CO

To illustrate these principles, two of GAST’S(1968) examples have been recalcu-

lated. The f?rst is concerned with the trends for K, Rb and K/Rb in the melting of
ampholite, using the data given in GA~T’S Fig. 10 and Table 1 (assuming that
Khb’*rinis infinitely large for both K and Rb, in the absence of any other data). The
results are shown in Fig. 2, in which the symbols I, I, L, s, refer to relative concen-
trations given by equations (13)-(16).
The liquid trends given by equations (13), (la), (15) are qualitatively similar,
particularly where D, is not greatly different from unity, as in the case of K. With
Rb, however, the degree of fractionation is much more dependent on the model
chosen, and this is consequently also the case for K/Rb.
Ampholite non-modal melting

30 30

20 20

Rb'

10 RbL 10
6'
Rb
KL
K'
kk L
0.1 0.1
F F

0.1 0.1
F

Fig. 2. K, Rb and K/Rb trends in various hypothetical melting models of ampho-

lite describedin the text (cf. CAST, 1908, Fig. 10).
7
 Phlogopite Lherzolite
Phlogopite Lherzolite -_ non-modal melting
-- modal melting
5
4 KI
Relative Relative

Concentration Concentration

0
0 0.05 0.1
F

Fig. 3. K, Rb and K/Rb trends in various hypothetical melting models of phlogopite lherzolite described
in the text (cf. GAST, 1968, Fig. 10). Figure 3a shows modal melting (i.e. D constant), whereas
Figure 3b shows non-modal melting.
 Notes 243

A point worthy of mention is that the trend of increasing K/Rb, as the quantity
of melt increases, applies not only to the liquids produced but also to the residual
solids. This paradox is in part a function of the numbers used in this example, but is
real and could well apply in nature.
The second example uses the phlogopite lherzolite also illustrated in GAST’S
(1968) Fig. 10, with the additional parameters given in his text. The calculations
used for Fig. 3a assume the lherzolite minerals to have melted in their original
proportions (equations S-11), whereas in Fig. 3b the melting proportions chosen by
GAST have been used. It is confirmed that the relative melting proportion of phlogo-
pite determines whether K/Rb increases or decreases.
In conclusion it should be noted that GAST’S (1968) major conclusions do not
appear to require modification, assuming that the calculations illustrated by his
diagrams did not embody the numerical errors mentioned above.

REFERENCES
GAST P. W. (1968) Trace element fractionation and the origin of tholeiitic and alkaline magma
types. Qeochim. Cosmochim. dcta 32, 105’7-1086.
SCRILLINUJ. G. and WINCHESTERJ. W. (1967) Rare-earth fractionation and magmatic processes.
In Mantles of Earth and Terrestrial Planets, (editors S. K. Runcorn F.R.S.), pp. 267-283.
Interscience.

Qeochimiea
et Cosmochimica
Acta, 1970,Vol. 34,pp. 243to 247. Pergamon
Press. Printedin Northern
Ireland

Rapid rock analysis by electron probe

J. C. RUCKLID~E, F. G. F. GIBB, J. J. FAWCETT and E. L. GASPARRINI
Department of Geology, University of Toronto Toronto 5, Ontario, Canada

(Received 17 June 1969; accepted in revisedform 17 September 1969)

Abstract-Glasses produced by direct fusion of five U.S.G.S. standard rock powders have been
successfully analyzed for Si, Al, Ti, total Fe, Mg, Ca, Na, K and P. A versatile probe data
reduction programme allows a wide range of standards to be used and data can be collected and
processed rapidly. Results suggest that the technique may find wide application as a method of
analysis of rocks, of suitable types.

INTRODUCTION
DURING an experimental study of the melting relations of some igneous rocks, a
remarkable agreement between electron probe analyses of glasses, produced from
rock powders at or above the liquidi, and the chemical analyses of the rocks,
suggested that whole rock analyses might be obtained by a similar technique.
The small quantities of material required, and the speed with which analyses can
be performed on the electron probe, make this an attractive possibility. To
evaluate the feasibility of this method of rock analysis, glasses were prepared by
direct fusion of five U.S.G.S. standard rocks and were analysed for Si, Al, Ti,
total Fe, Mn, Mg, Ca, Na, K and P using an ARL-EMX electron microprobe. The