Lecture 3 - Trace elements & Geochemistry


I.  Chemical Variation diagrams

·      Simple display of chemical differences and trends shown by a related suite of rocks.  Line is drawn through lots of points

·      Most common is Harker diagram

Oxide vs. % SiO2.  Why is SiO2 useful? 

Liquid line of descent.  Ideally correlated but usually not (open systems)


Segmented trends:  onset of crystallization of new minerals.


Oxide vs. % MgO.  Better for basalts, cause plagioclase crystallization decreases SiO2!


Compositional variation is due to either partial melting or fractional crystallization (crystal settling, crystal zoning) or mixing


Can use major or trace elements, or both


Assumption:  Liquid compositions!!!!.  No mineral accumulations, plutonics are dangerous

More a bit later after talk about trace elements



II.  Trace elements

Any element with concentration <0.1 wt.% (less than 1000 ppm).  Very powerful petrogenetic indicators of magmatic processes (note mistake in text p. 37)

·      REE ­ (lanthanides, atomic numbers 5-71)

·      transition metals ­ atomic numbers 21-30

·      PGE ­ platinum group elements

·      HFS ­ small, highly charged cations.  High ionic potential (aka field strength).  Sc, Y, Th, U, Pb, Zr, Hf, Ti, Nb, Ta.  Not easily leached by H2O-rich fluids (immobile).  Incompatible

·      LILE ­ large cations of small charge, low field strength.  Cs, Rb, K, B, Sr, Eu.  Very easily leached by H2O rich fluids (highly mobile).  Incompatible


Defined as being either compatible (prefer to be in mineral) or incompatible (prefer to be in liquid).


Obviously will depend on which minerals are around and what the composition of the magma is


Kd = C mineral/element / C liquid/element = Partition coefficient


Kd <<<<1.  Incompatible.  ³Happy in liquid².

Kd = 1 Neutral

Kd > 1 Compatible.  Element is ³happy in solid².


Kd for Sr in plag is very high.  But Kd for Sr in olivine is very low.


Thus need to look at the whole rock or crystallizing assemblage, not just individual minerals. 

Di = SKdixi = Distribution coefficient


Di = x1Kd1 + x2Kd2 + x3Kd3 + Š..


DSr in mantle is low (relatively incompatible, will go into melt

DSr in crust is high (compatible)

DNi > 1 in mantle, same for Cr.  Preferentialy retained during melting, is first to be extracted during crystallization


Partition coefficients increase with increasing T of melt.  More silicic melts are more tightly structured, causing trace elements to be rejected and forced into coexisting crystals. 

Pressure has opposite


Can use Harker diagrams to show behavior

















Rare earth elements

La, Ce, Pr, Nd, Pm, sm, Eu, Gd, Tb, dy, Ho, Er, Tm, Yb, Lu

Atomic numbers are 57 (La) to 71 (Lu)


Light REE ­ lower atomic number but larger radius

Heavy REE ­ higher atomic number but smaller radius


Geochemically similar ­ incompatible

                                         Trivalent (+3) ­ except Eu+2, Ce (sometimes +4)

                                         Eu2+/Eu3+ depends on oxygen fugacity

                                         Eu2+ looks a lot like Ca, so substitutes in minerals with lots of calcium.

Generally the REE behave exactly the same.  They are incompatible.

However, small differences in chemical behavior is due to the decreasing size and increasing atomic number.  Thus REE can be fractionated from each other.


Oddo-Harkins effect ­ REE with even atomic numbers are more abundant than REE with odd atomic numbers

Yb70 is 6x more abundant than Lu71

Causes scaling effect in graphical representation. 



Divide abundance by concentration in some standard rock. 

Standard is chondritic meteorites (inner planets of solar sys).  Representative of primitive Earth. 

Log scale










Fractionation and partial melting will have effect on shape of REE pattern, depending on D.


If all REE had same D, would always have flat patterns.  But variability.  Incredibly useful for gleaning information on sources and processes in the origin of magmas. 

Examples:  lunar basalts with negative Eu anomaly, MORB with sloping pattern, adakites (Fig. 2.23)



Garnet ­ low Ds for light REE, larger Ds for heavy












Magma derived from source with garnet

Magma that has fractionated garnet













Spidergrams (normalized trace element diagrams)

Same idea as REE

1.  Abundance of a range of incompatible elements normalized to estimates of their abundance in primitive mantle (estimated mantle before continental crust formed, so = bulk earth) (=Wood 1979).  Same as chondrite, just multiplied by a factor of 2.9.  Adjustments for K, Rb because they are volatiles and may not have chondritic abundance in Earth.  Adjustments for Phosphorus because it may have gone to core.  

Arranged in order of increasing compatibility during melting of mantle. 


2.  Abundance normalized to chondrite ­ Thomposon (1982).  Thought to be better because is a measured proxy for primitive undifferentiated Earth


3.  Abundance normalized to MORB (Pearce-style).  Better for evolved basalts, or andesites, etc.  whose parents might be MORB rather than primitive earth.  Order of elements is different.  Uses ionic potential (charge/radius) to measure mobility in a fluid.  Most mobile elements (Sr, K, Rb, Ba) at left in order of increasing incompatibility, and immobile elements on right increasing incompatibility from right to left. (see Fig 4.25 of Rollinson)


Useful because basalts from different tectonic settings have different patterns on these normalization diagrams. 


MORB ­ depleted in the most incompatible elements.  Why?

MORB basalts are excellent mirror of the asthenospheric source that the come from.  The depletion in incompatible elements must be a characteristic of their source.  The mantle must have had previous partial melting events that extracted incompatible elements.  When mantle melts, will want to ³get rid of² those elements on the left.  Over time source becomes depleted.

Called a depleted source.


IAB - enriched in the most incompatible elements, especially Ba, Rb, K but depleted in incompatible elements with high field strengths (Nb and Ta) relative to adjacent elements on the diagram.  ³Spiked² pattern. 

Positive spikes ­ components added to mantle source by slab fluidsThought to reflect magma generation involving hydrous fluids from slab. 

Negative troughs ­ Nb and Ta.  Why?

Retention of Nb&Ta in source during melting.  Stablization of some weird mineral.  Or maybe not really a trough, just an extra depleted mantle?


OIB ­ enriched in incompatible elements.  Other enriched source OR small degrees of partial melting of non-depleted source.


Which to use?

Very chaotic in that people don¹t often let readers know which normalization they are using.  Rollinson recommends chondrite-normalized using normalized values of Thompson (1982) (=column 6 of table).  Less subjectivity in normalizing values.


Inherent worries:

1.  Compatibility vs mobility.  All elements are incompatible, but have varing degrees of mobility in a fluid phase.  HFS(Y, Hf, Zr, Ti, Nb, Ta) are immobile, LIL (Cs, Rb, K, Ba, Sr, Eu) are mobile.  So LIL behavior may be function of a fluid, HFS elements function of chemistry of source and crystal/melt processes. 

2.  Zr controlled by zircon, P by apatite, Sr by plagioclase, Ti, Nb, and Ta by ilmenite, rutile or sphene.

3.  Negative Nb anomalies characteristic of continental crust and may be an indicator of crustal involvement in magma processes

4.  LIL may also be indicator of crustal contamination, as these elements are concentrated in the continental crust. 


Bivariate trace element plots

Incompatible elements are extremely sensitive to partial melting processes.  The more highly incompatible, the more sensitive to degrees of partial melting.  Also vary during fractional crystallization, but more important during AFC.



1.  Identification of source characteristics from incompatible element plots:  Ratio of a pair of highly incompatible elements whose bulk partition coefficients are similar will not vary in the course of fractional crystallization and will vary little during batch partial melting.  Thus, the slope of a correlation line on a bivariate plot of two such highly incompatibel elements gives the ratio of the concentration of the elements in the source.


For mantle melting, these groups have almost identical bulk partition coefficents

Cs-Rb-Ba                    very common

U-Nb-Ta-K                 very common

Ce-Pb                          must have very good analyses, so not common

Pr-Sr                           not common

P-Nd                           not common

Zr-Hf-Sm                    very common

Eu-Ti                           not common

Ho-Y                           not common

Bivariate plots of element pairs taken from within these groups can show the ratio of the elements in the source. 


Also use La-Ta, La-Nb, Ta-Th, Ti-Zr and Ti-Y. 


Any variation in the ratio reflects heterogeneity in the source resulting from source mixng or contamination.



a.  Relationship does not hold for very small degrees of melting.

b.  Difficult to apply to granitic rocks because few trace elements are highly incompatible in granitic melts.  Isotopes are better.


2.  Identification of source characteristics  from incompatible element ratio-ratio plots:

Minimize effect of fractionation.  Different mantle sources plot on different correlation lines (Fig 4.31 Rollinson). Saunders et al. (1988).





Each ratioed to Nb, in order to explore the ratios Th/Ce, K/Ce, and U/Pb in MORB

Where the three elements have identical bulk partition coefficients, the ratio will not change during partial melting or fractional crystallization.


3.  Compatible element plots.  Because they change dramatically during fractional crystallization, can be used when plotted against an index of fractionation to test for fractional crystallization.  Can be used to look for AFC and in situ crystallization. 

In partial melting, these compatible elements are buffered by the solid.  So the concentration in the source after partial melting remains unchanged.  


Modelling trace element processes in igneous rocks

·      Trace element concentrations must be known with great accuracy

·      Partition coefficients must be known accurately for the conditions under which the process is being modeled

·      The starting composition must be known. 


Calculated compositions are plotted either

a.  on a bivariate graph and compared with an observed trend of rock compositions

b.  on a multivariate diagram such as a REE digram and calculated compostion compared with a measured composition


Vector diagram modeling

Changes in trace element concentrations modeled using vectors to show amount and direction of change.

Mineral vectors  show trend of fractionating mineral phase or mineral assemblage. (Fig 4.33 Rollinson)

Partial melting vectors:  show changing melt and source compositions during partial melting.  (Fig 4.34 Rollinson).  Can compare model types. For example, different degree of enrichment. 


Shows only limited number of elements, but can plot tons of samples


Multivariate diagram modeling

Can only show a few samples clearly on a single diagram.  (†)