1

The Use of Trace Elements in Igneous Petrology

Introduction:
Trace elements are those which occur in very low concentrations in common rocks
(usually < 0.1 % by weight). Their concentrations are therefore commonly expressed in
parts per million (ppm; 1 ppm = 10-4 weight%). Unlike major elements, trace elements
tend to concentrate in fewer minerals, and are therefore more useful in formulating models
for magmatic differentiation, and in some cases, in predicting the source of a particular
magma. Trace elements most commonly used for the interpretation of the petrogenesis of
igneous rocks include: Ni, Cr, Sc, V, Rb, Ba, Sr, Zr, Y, Nb and the rare earth elements (La
to Lu). Keep in mind that the concentration of trace elements will vary with the rock type;
whereas Ni and Cr show higher concentrations in mafic and ultramafic rocks, Zr and Rb
are more concentrated in acidic rocks. Accordingly, some major or minor elements as K
and P, which occur in very low concentrations in basalts (approaching trace levels), are
just as useful in petrogenetic interpretations as some trace elements (in the case of basalts
only, .. of course!).

The incorporation of a trace element in the crystal structure of one or more minerals
depends largely on its charge and radius, but also on the electronegativity of this element,
and crystal field effects. Accordingly, a trace element will either substitute for a major
element in the structure of a crystallizing mineral, or remain in the liquid.

The distribution coefficient "K d":

The distribution coefficient of a particular element is the ratio of the concentration of this
element in a mineral crystallizing from a melt to its concentration in the liquid, i.e.
Kd = Cs/C
where Cs and C are the concentrations of the trace element in the solid and liquid,
respectively. Accordingly, if a magma is crystallizing two phases (e.g. Ol and Cpx), and the
trace element that we are interested in can enter the structure of these two phases, it is
clear that it will have two different Kd values, one for Ol and the other for Cpx. In
practice, Kd values for different elements and different phases are determined either from
well controlled experiments, or from analysis of crystal - glass pairs in equilibrium in
volcanic rocks. Kd values for some commonly used trace elements are listed for common
igneous minerals in Table 1.

Factors affecting Kd:

i) Temperature: As with any other equilibrium constant, Kd depends on T. Proper
determination of the Kd value requires knowledge of T of equilibration of the
crystal and liquid.
ii) Composition of the liquid: Kd values will differ from one magma type to another
(e.g. basic vs. acidic), as such properties as viscosity and composition of the melt
will affect the diffusion rates of particular elements.
iii) Pressure: has a very limited effect on Kd
 2

iv) Oxygen fugacity: Some trace elements (e.g. Ce and Eu) have more than one
oxidation state. Differences in charge and radius between such ions (e.g. Eu 2+ and
Eu3+) will result in differences in their Kd values for particular minerals.
v) Type of mineral crystallizing from the magma: As stated above, each element will
have a unique Kd value for each mineral crystallizing from a melt of constant
composition at constant P, T and fO2.

The bulk distribution coefficient "D":

Because each element will have as many Kd values as there are phases crystallizing from
the melt at constant T, P, and fO2 (one value for each phase), it is often more convenient
to calculate a bulk distribution coefficient in order to determine whether this trace element
will be preferentially incorporated in the solid phases or in the coexisting melt. The bulk
distribution coefficient is therefore expressed as the sum of Kd values for all crystallizing
phases, taking into account the weight proportions of these phases.
D = S wi . Kd
i
where w is the weight % of the crystallizing mineral "i", and K di is the corresponding Kd
value for this phase. A trace element with a D value << 1 will therefore be concentrated in
the liquid and is considered "incompatible", whereas an element with D > 1 will be
preferentially incorporated in the solid phases, and is therefore termed "compatible".

Example:
Consider a lherzolitic assemblage consisting of 40% Ol, 30% Cpx and 30% Opx, from
Table 1, we can calculate DNi and DSr as:

DNi = 0.4 . 10 + 0.3 . 4 + 0.3 . 2 = 4.72

DSr = 0.4 . 0.001 + 0.3 . 0.01 + 0.3 . 0.07 = 0.0244

Therefore, for this composition, Ni is considered a compatible element, whereas Sr is

incompatible.

Types of trace elements:

In addition to this simple classification of trace elements into compatibles and
incompatibles, trace elements are perhaps better classified on the basis of their
geochemical characteristics (which will naturally influence whether the element becomes
compatible or incompatible).
1) Large ion lithophile elements (LILE): These elements are characterized by large
ionic radii, and low charges, and will therefore preferentially concentrate in the
liquid until a particular phase with large enough sites to accommodate them begins
to crystallize. These elements will therefore be largely "incompatible" particularly
with respect to mantle phases (Ol, Opx, Cpx, Gt, .. etc). Examples include: K, Rb,
Sr and Ba.
2) High field strength elements (HFSE): These are elements which have large cations,
but also large charges, and are also excluded from mantle phases and more
concentrated in residual liquids (i.e. they will be more incompatible). These
 3

elements are concentrated in accessory phases as sphene, zircon, and apatite.

Examples include Zr, Hf, Nb, Ta, Th and U.
3) Transition elements: Trace elements which are also transition elements are
characterized by relatively small ionic radii, and are either bi- or tri-valent. These
elements are strongly partitioned in the solid phases that crystallize during the early
stages of magmatic evolution, and are therefore "compatible" with mantle phases.
Examples include Ni, Co, Cr, and Sc.
4) Rare earth elements (REE): This is a group of elements with atomic numbers
between 57 (La) and 72 (Lu) characterized by relatively large ionic radii, and
valences of either +2 or +3. They have proven to be very important for petrogenetic
interpretations. However, these elements occur in very low concentrations in
igneous rocks, and are difficult to analyze for.

Applications of trace elements to Igneous Petrogenesis

1- Testing models of magmatic differentiation using trace elements:
Because some trace elements are compatible while others are incompatible (for a specific
magma composition at a given T, P and fO2), concentrations of some trace elements will
be very sensitive to the processes of magmatic differentiation, and are often very useful in
testing certain models of partial melting or crystal fractionation, ... etc. For example, if we
consider a rock that was formed from a magma differentiating by fractional crystallization
of significant amounts of olivine and pyroxene, this rock will be depleted in compatible
elements as Ni, Co and Cr, and enriched in the incompatible elements Rb, Sr and Ba. The
degree of enrichment or depletion depends on the amount of fractionated Ol and
pyroxene. Knowledge of the distribution coefficients for these elements will therefore
allow us to "quantify" the process of fractional crystallization (i.e. figure out how much of
these phases has fractionated from a particular source magma to produce this rock).

On the other hand, partial melting of a particular mantle rock will result in very different
trace element patterns and concentrations. Very small amounts of partial melting will result
in magmas that are highly enriched in incompatible elements, but strongly depleted in
compatible ones. Magmas produced by larger amounts of partial melting will be less
enriched in these incompatible elements, and so on. Accordingly, trace elements can be
used to infer the mechanisms of differentiation, and the degrees of partial melting,
assimilation or fractional crystallization necessary to produce a rock with particular trace
element patterns and concentrations. Although there are many models that can be tested
using trace elements, we will focus on the three most commonly used models:
i- Equilibrium partial melting:
Is the case of producing a single batch of magma by partially melting a rock, so that this
magma is in equilibrium with the source rock until it is mechanically separated from it.
Under these circumstances, the concentration of a trace element in the liquid (C) relative
to its original concentration in the unmelted rock (Co) is given by:
C /Co = 1/(F + D - FD) (1)
where F is the weight proportion of the melt produced, and D is the bulk distribution
coefficient for this element. It can be seen that for strongly incompatible elements (where
D << 1), C /Co ~ 1/F, whereas for very small fractions of melt, C  /Co ~ 1/D. These
 4

relations are shown in Fig. 1. Note that with increasing degrees of partial melting, different
minerals may be progressively consumed, causing discontinuous changes in the value of D.

ii- Incremental batch melting:

Incremental batch melting means that the rock undergoes partial melting in stages, and
that at each stage, the melt is removed. This mechanism can be modeled successfully using
the expression used for equilibrium fractional melting provided that the composition of the
source rock is modified after each batch of melt is produced to reflect its actual
composition at this stage (i.e. at the point where a new batch of magma will be produced).
Therefore, for a rock that has undergone two stages of partial melting, producing melt
fractions F1 and F2, respectively, the following expressions apply:
after the first increment of melting:
C1/Co = 1/(F1+D1-F1D1), (2)
and the concentration of the trace element in the residue (CR) will be given by:
CR = D1C1 (3)
and the concentration of the same trace element in the second batch of liquid will be given
by:
C 2/CR = 1/(F2+D2-F2D2) (4),
from (2) and (3), and substituting in (4):
C 2/Co = D1/(F1+D1-F1D1)(F2+D2-F2D2) (5)

iii- Fractional crystallization (Rayleigh fractionation)

If all the minerals crystallizing from a melt were to remain in equilibrium with the melt
from which they formed, then fractional crystallization can be modeled as the exact
reverse of equilibrium melting using the same equation (equation (1), where F is now the
fraction of the liquid remaining after crystallization). However, if the crystals are separated
(which is more realistic) then we apply the Rayleigh equation:
C /Co = F(D-1) (6)
For highly incompatible elements (D << 1), C 1/Co ~ 1/F (Fig. 2). However, because of
the differences between equations (1) and (6), it is clear that closed system crystal
fractionation is more effective than partial melting in concentrating one incompatible
element in the liquid relative to another. These relations are clear from Figs. 1 and 2.
 5

2- Determination of the depth of generation of a primary magma:

Because different trace elements have different chemical characteristics, their
concentrations in a magma produced by partial melting of a source rock will depend on
the phases occurring in this source rock (which control the value of Kd) and the
abundance of these phases (which control the values of D). For example, elements such as
Sr are strongly fractionated in plagioclase feldspars. In the absence of Plag from the
source rock, Sr will be incompatible, and will be highly enriched in a magma formed by
small degrees of partial melting. If the source rock is Plag - bearing, the magma will be
less enriched in Sr (for the same % of partial melting), until Plag melts completely.

As you are aware, primary basalts are produced by partial melting at different depths in the
mantle. At shallow depths, the mantle consists of Plag lherzolite, at depths between 40 and
80 km, it consists of spinel lherzolite, whereas at depths > 80 km, it consists of garnet
lherzolite (Fig. 3). Because Plag, Spinel and garnet concentrate Sr, V + Cr, and heavy rare
earth elements (HREE), respectively, magmas produced by small degrees of partial
melting at shallow depths will be depleted in Sr, those from intermediate depths will be
depleted in V and Cr, whereas those from depths > 80 km will be depleted in HREE.
Trace element contents of different types of primary magmas indicate that these can be
produced by different degrees of partial melting at different depths in the mantle, as listed
in the following Table:

Type of primary magma Depth (km) % melt

Tholeiite 25 - 50 10 - 30

AOB & Basanite 50 - 100 2 - 10

Nephelinite, Kimberlite, and > 75 1-2

Carbonatite

3- Prediction of the phases fractionating from a magma:

The same rationale presented in (2) above applies for the identification of the phases which
have fractionated from a magma undergoing fractional crystallization. Separation of: (a)
Plag depletes the remaining melt in Sr and Eu, (b) Ol depletes it in Ni and Co, (c) spinels
deplete it in V, Cr and possibly Zn, (d) K-spar in Ba and Rb, ... etc. The effects of
fractionation of some of these phases on some trace element patterns are shown in Fig. 4b.

4- REE and REE diagrams:

REE are very useful for petrogenetic interpretations. Their concentrations in an igneous
rock are usually divided by (i.e. normalized against) their concentrations in standard
chondrites or N-MORB (normal mid - oceanic ridge basalt) in order to smooth out large
differences in concentration between one REE and the other. These normalized values are
then plotted on diagrams similar to that shown on Fig. 4a, where the REE are arranged on
the X-axis from the lightest element to the heaviest. In the case of an igneous rock series
(a group of co-magmatic rocks), the REE concentrations will increase systematically with
 6

progressive differentiation (Fig. 4a) as they are largely incompatible (unless we have
garnet in the source rock from which the magma formed).

REE diagrams are also useful in identifying which phase or phases fractionate from a
magma, following the same rationale presented in (2) and (3) above. In order to identify
such phases, it is necessary to know which REE are preferentially incorporated in which
phases. Figure 5 shows the REE concentrations for some of the most common igneous
minerals (occurring as a phenocryst) compared to their concentrations in the coexisting
liquid (represented by the matrix in a volcanic rock). One of the most striking features of
this diagram is the enrichment of Gt and zircon in the HREE relative to the LREE, and the
strong fractionation of Eu by the three feldspars.

REE diagrams are also used to determine the type of basalt. Once more, this is a function
of the depth of generation of the basaltic magma, as well as the degree of partial melting
necessary to generate it. Figure 6 shows the REE patterns expected for nephelinites,
Ocean island tholeiites, MORB, and alkali olivine basalts (AOB). Note how MORBs,
which are products of significant degrees of partial melting at shallow depths are LREE -
depleted, whereas AOBs and Nephelinites, which form by low degrees of partial melting
of garnet peridotites are HREE - depleted, and LREE - enriched (Fig. 6). Applying the
same principles and equations (1), (5) or (6) outlined above, REE concentrations can be
used to test the various models of differentiation of a particular igneous rock suite, and the
degrees of partial melting and/or fractional crystallization necessary to produce a
particular "group of REE concentrations and patterns". An example of this is illustrated in
Fig. 7, which shows that although all patterns for these rocks collected from Iceland are
consistent with their derivation from one parental magma, the ratios of Ce/Yb are
inconsistent with such a model and require more than one parental magma.

5- Significance of Eu and prediction of oxidation states:

Eu has two valence states Eu 2+ and Eu3+. The valence of Eu in a magma will be a
function of its oxidation state; high oxidation states favor Eu 3+. Plagioclase can only
incorporate Eu2+ in its structure, not Eu3+. Therefore, if a magma is undergoing fractional
crystallization under relatively reducing conditions (low fO2), the onset of Plag
fractionation will result in the rapid depletion of Eu from the magma, particularly when
compared to the other REE (all of which are trivalent). This results in what is known as a
"negative Eu anomaly" where the concentration of Eu is much lower than that of the
neighboring REE on a normalized REE plot (e.g. the REE plot for rock # 8 on Fig. 4a).
The magnitude of this negative Eu anomaly will be a reflection of the amount of feldspar
fractionating as well as the oxidation state of the magma at this time. On the other hand,
analysis of a rock containing cumulus plagioclase will result in a "positive Eu anomaly".
Accordingly, REE plots can help identify whether plagioclase is fractionating or not, and
give us a rough idea of the fO2 conditions prevailing at the time of its fractionation. Figure
8 shows the effect of fO2 on the distribution coefficient of Eu in Plag.
 7

6- Spider diagrams:
Chondrite normalized concentrations of some (mostly incompatible) trace elements may be
used in plots similar to those presented for REE, and which are known as spider diagrams.
The order in which these elements are arranged on spider diagrams is to some extent
arbitrary, although the most incompatible elements are always placed on the left hand side
of the plot, with increasing compatibility to the right (Fig. 9). Because of the sensitivity of
incompatible trace element concentrations to the degree of partial melting, patterns on
spider diagrams tend to be diagnostic of the conditions under which the primary magma
was generated, and of its type. MORBs, which result from significant degrees of partial
melting, tend to be depleted in the most incompatible trace elements (Fig. 9). Ocean island
basalts (EMORB and AOB) are enriched in these incompatibles. On the other hand, calc-
alkali basalts forming in subduction zones show numerous spikes and troughs that are
suggestive of contamination with (continental?) crust, or metasomatism (by fluids derived
from the dewatering of the subduction slab?). Troughs at Th and Rb as well as Nb and Ta,
are quite characteristic of calc-alkaline rocks and are interpreted as being due to
contamination of the magma with lower crustal rocks.

7- Other applications:
In addition to the applications listed above, trace elements can also be used to model such
processes as contamination or magma mixing. They can also be used to detect whether
rocks were affected by hydrothermal alteration, since some trace elements are mobilized
by meteoric water (e.g. Sr, Rb, Ba), and hence partly removed from the igneous rock,
whereas others are relatively immobile (e.g. Y, Zr, Nb). Trace elements can also be used to
identify the paleotectonic setting of some volcanic rocks (i.e. to determine where they
were erupted). In this case, rather than use the absolute concentrations of trace elements
(which may have been affected by such post-magmatic processes as weathering, alteration
or metamorphism), ratios of relatively immobile trtace elements (as these are least affected
by post magmatic processes). Figure 10 shows some commonly used discriminant ternary
plots for rocks of basaltic composition.

Perhaps one of the most important conclusions derived from the application of trace
element systematics is that the mantle is not homogeneous. Careful study of patterns of
incompatible trace elements from different volcanic rocks has shown that the mantle is
heterogeneous. Several models have since been proposed to account for such
heterogeneity, the most popular of which suggests that parts of the mantle have been
metasomatised by fluids derived from subduction zones.