You are on page 1of 38

ESCI 430 C-T Lee

Chapter 7
Trace-element geochemistry: Silicate Systems

7.1 Introduction
Major elements are defined as chemical constituents that reside in approximately
the 1 wt. % level or greater within any given system. The state of this system, that is its
physical properties, are dominantly controlled by its major-element composition. Major
elements largely dictate what mineral phases form in a rock, what chemical species are
present in a liquid, etc. Such properties as the density, viscosity, and melting temperature
of a rock, or the density, freezing point, and boiling point of a liquid are ultimately
controlled by the major-element composition of the system. At any given temperature or
pressure, the state of a system is essentially defined thermodynamically by its bulk
composition
A trace element is defined as any element present below ~0.1 wt. %. Units such
as ppm, ppb, ppt, and ppq (parts per million, billion, trillion, and quintillion) are some of
the units used to describe the concentrations of trace elements in a system. Because
trace-elements are present in such small quantities, they generally exert minimal control
on the state of a system. For example, in silicate systems, trace-elements rarely form
mineral phases, and in aqueous systems, they do not form major species. Instead, the
distribution of trace-elements within a system is controlled by how it is partitioned
among various phases. The partitioning between various phases (e.g., mineral and melt
phases) occurs in a predictable manner, and for this reason, trace-elements can be used as
passive tracers to study how mass is redistributed within or between systems. In
addition, a number of trace-elements decay radioactively to another trace-element,
allowing the isotopic composition of such daughter elements to be used to constrain the
timescales of petrologic processes that fractionate parent/daughter elemental ratios.

7.2 Partition coefficient
The ratio of the concentration of a trace-element in one phase relative to another
phase is called the partition coefficient. In general, the partition coefficient of an
element i for a given mineral j is defined to be the ratio of its concentration in the mineral
to that in the melt at equilibrium:
i
Melt
i
j
i
j
C
C
D = (7.1)
When D<1, the element is said to be incompatible in the mineral, that is, the element
prefers to be in the melt. When D>1, the element is said to be compatible and the
element prefers to stay in the solid. For a rock, consisting of several phases j, the bulk
partition coefficient is given by the weighted average of the individual partition
coefficients for each phase:
=
j
i
Melt
i
j
j
i
B
C
C
X D (7.2)
where X
j
represents the weight fraction of the jth phase in the solids.

ESCI 430 C-T Lee
The partition coefficient can be put into thermodynamic context using the
following example. Consider the following reaction
JMgSi
2
O
6
(cpx) = JMgSi
2
O
6
(melt) (7.3)
where the cation J is exchanged between a clinopyroxene crystal and a melt. The
equilibrium constant for this reaction is
J
Cpx
J
melt
O JMgSi
Cpx
O JMgSi
melt
O JMgSi
Cpx
O JMgSi
melt
RT
o
G
X
X
a
a
e K
6 2
6 2
6 2
6 2
/
3 . 7
) 3 . 7 (


= = =

(7.4)
The equilibrium constant can be expressed in terms of the partition coefficient as
J
Cpx
O JMgSi
Cpx
O JMgSi
melt
D
K
1
6 2
6 2
) 3 . 7 (

= (7.5)
where the measurable partition coefficient is
J
melt
J
Cpx
J
Cpx
X
X
D = (7.6)
It can be seen from Eq. 7.5 that the partition coefficient can be easily expressed in terms
of the equilibrium constant and the respective activity coefficients. If the behavior of the
trace-element is ideal, then the interchange between the equilibrium constant and the
partition coefficient is trivial. Because the equilibrium constant depends on temperature
and pressure, the partition coefficient also depends on pressure and temperature.
For non-ideal mixing, the activity coefficients deviate from unity. Two scenarios
arise. If both activity coefficients are constant over a wide compositional range (but near
infinite dilution of the trace-element), the trace-elements behavior in the mineral and
melt is said to follow Henrys Law. In the Henrys Law region, the partition coefficient
for a particular trace-element does not depend on its concentration in the mineral or melt.
If the activity coefficient varies as a function of concentration of the element of interest,
then non-Henrian behavior is displayed. This means that the partition coefficient cannot
be constant. It is critically important when applying measured partition coefficients to
evaluate whether Henrys Law is obeyed.
Before going on, it is important for us to recognize that the chemical reaction of
Eq. 7.3 assumes that the melt behaves thermodynamically as if it contained simple
ordered structural units similar to that of clinopyroxene. In reality, the speciation of Al
and Si in melts is not likely to be so simple, and therefore, reactions similar to Eq. 7.3
may only be valid for a narrow compositional interval.
More generally, we can describe trace-element partitioning as an exchange
reaction between the melt and the coexisting mineral. For example, we can describe the
partitioning of Ni
2+
between olivine (ol) and melt as an exchange reaction without the
need to specify a melt structure
0.5Mg
2
SiO
4
(ol) + Ni
2+
(melt) = 0.5Ni
2
SiO
4
(ol) + Mg
2+
(melt)


(7.7)
For the ideal case, the equilibrium constant for reaction 7.3 can be expressed as
Ni
melt
Mg
Ol
Mg
melt
Ni
Ol
Ni
melt
Mg
Ol
Mg
melt
Ni
Ol RT
o
G
X X
X X
a a
a a
e K = = =

5 . 0
5 . 0
/
) (
) (

(7.8)
or equivalently as

ESCI 430 C-T Lee
Mg
Ol
Mg
melt Ni
ol
Ni
melt
Mg
Ol
Mg
melt
Ni
Ol
D
C
C
D
C C
C C
K = = (7.9)
where, the subscripted term K
D
represents the equilibrium coefficient expressed in terms
of concentrations. In Eq. 7.9, is the partition coefficient (often referred in this
context as the Nernst partition coefficient). It can be seen from Eqs. 7.8 and 7.9 that K
Ni
ol
D
D

depends primarily on T and P (G
o
is calculated for a given P and T) and not so much on
bulk composition (for the ideal case). However, from Eq. 7.9, we can see that the
partition coefficient of Ni between olivine and melt, , depends on composition (e.g.,
Ni
ol
D
Mg
Ol
Mg
melt
C
C
) since the exchange of Ni between olivine and melt depends also on the exchange
of Mg.
In practice, geochemists use partition coefficients of the form in Eq. 7.1 to model
the geochemical processes that fractionate trace-elements. These partition coefficients are
determined by measuring the concentration of the element of interest in a mineral and the
concentration of the same element in a coexisting melt. This can be accomplished from
experimental studies or from natural phenocryst-lava pairs. Partition coefficients derived
by these approaches are typically compiled into convenient tables for quick reference.
The foregoing discussion, however, shows that partition coefficients in general depend on
temperature, pressure, the major-element composition of the system, and also the
concentration of the element of interest. The partitioning of elements with variable
valence states will also depend on the oxygen fugacity of the system, which dictates an
elements dominant valence state. Oxygen fugacity is a measure of the O
2
activity in a
system (e.g., the amount of O
2
). Oxygen has a strong oxidizing power, that is, it has a
tendency to attract electrons. For example, V can exist in a number of oxidation states,
dictated by the amount of O
2
in the environment. The following reaction shows how O
2

buffers the ratio between V
3+
and V
4+
: V
2
O
3
+ 0.5 O
2
= 2 VO
2
.
The successful application of partition coefficients towards modeling geochemical
processes requires that the partition coefficients appropriate for the temperature, pressure,
and compositional range of interest be chosen.

7.3 Controls on partitioning
7.3.1 Goldschmidt Rules
Minerals have a wide variety of crystallographic sites in which trace-element
cations can enter. For example, pyroxenes are characterized by two cationic octahedral
sites, M1 and M2. The pyroxene formula can be expressed in terms of its structure, e.g.
M1M2Si
2
O
6
. Although both the M1 and M2 sites are in 6-fold coordination, the M2 site
differs from the M1 site in that it is distorted. For example, for an orthopyroxene having
the composition Mg
0.93
Fe
1.07
Si
2
O
6
, the average cation to oxygen bond distance in the M1
site is 2.10 A
o
, whereas in the M2 site it is 2.23 A
o
.
Our intuition tells us that the ability for a different cation to enter one of the
structural sites of orthopyroxene rests on the requirement that its ionic radius is similar to
that of the replaced atom or ion (e.g., the M1 and M2 cations). Intuition also tells us that
ionic charge also exerts a control on the ability of a cation to enter a particular

ESCI 430 C-T Lee
BONDING REVIEW
When two atoms approach, they exert a force on each other. At large distances, the
force is attractive and at very close distances, the force is repulsive. Attractive forces
range from being very weak to being strong enough to form stable molecules.
Chemical bonds form because the attraction of atoms results in a decrease in
potential energy. Several types of chemical bonds exist: Ionic, Covalent, Polar, Van
Der Waals, and Metallic bonds. Ionic Bonds are common in inorganic geologic
materials. Ionic bonds form when valence electrons are exchanged between two
atoms such that each atom achieves a closed-shell configuration. The attraction of
the closed shell ions results in the formation of a chemical bond. In a covalent bond,
valence electrons are shared between two atoms in an overlapping electron cloud.
Thus, in contrast to an ionic bond where two closed-shell atoms interact, a covalent
bond involves the interaction of two open-shell atoms. In reality, most bonds are a
mixture of ionic and covalent character. For an ionic bond to form, an electron/s must
be stripped off one atom and transferred to the other atom. The energy required to
strip off an electron is called the ionization potential (IP) and the energy released
when an atom accepts an electron is called the electron affinity (EA). The tendency
for an atom A to attract an electron from atom B depends on relative stabilities of
bonds A
+
B
-
and A
-
B
+
. The differences in energy (per electron) between these two
bonds is a measure of the tendency to form one of these bonds, that is
0.5 [E(A
+
B
-
) E(A
-
B
+
)]= 0.5 [(IP
A
-EA
A
) (IP
B
EA
A
)]
or
X
A
-X
B
= 0.5 [(IP
A
+ EA
A
) (IP
B
+ EA
B
)]
where X
A
and X
B
are termed the electronegativities of atoms A and B, respectively,
X
A
= (IP
A
+ EA
A
)
X
B
= (IP
B
+ EA
B
)
When the electronegativity difference X
A
X
B
is large, the bond has an ionic
character. When this difference is small, the bond is largely covalent.
Polar bonds form by the long-range interaction of oppositely charged atoms
bonded to two different molecules. Water is the most abundant molecule that forms
polar bonds with other atoms or molecules. Because of the asymmetric structure of
the H
2
O molecule, the charge distribution on the water molecule is slightly polar (in
water, this is also known as hydrogen bonding). Polar bonds are relatively weak.
Another weak type of bond is the Van der Waals bond. These bonds are extremely
weak and result from short term interactions between closed shell atoms or molecules
owing to the fact that at any given moment in time, the valence electrons are not
homogeneously distributed. The last common type of bond is the metallic bond,
where the valence electrons of many nuclei are shared in such a way that electrons
are free to move from one atom to another. The attraction between the electrons and
the positively charged nuclei bind the metal atoms together.
crystallographic site. Two elements with identical charge are more readily exchangeable
than two elements with different ionic charge because exchange of disparately charged
cations requires that vacancies or a second element of compensating charge be created in
order to maintain electrical neutrality. Thus, two elements that have similar ionic radii
and identical charge can mix nearly randomly, such that the mixing can be said to be
nearly ideal. In such a case, these two elements can form a solid solution. Fe and Mg,
two cations that have very similar 6-fold cationic radii, thus form solid solutions in
various minerals (e.g., (Mg,Fe)
2
Si
2
O
6
and (Mg,Fe)
2
SiO
4
).
Table 1 (Shannon, 1976) shows the ionic radii of various cations for given
coordination numbers and valence states. To use this table, some assumption of the

Table 1. Ionic Radii from Shannon (1976)
ION
Oxidation
State
Elec.
Config.
Coord.
#
Spin
State
Crystal
Radius (A)
Ionic
Radius
(A)
Z/IR
Ac +3 3 6p 6 6 1.26 1.12 2.68
Ag +1 1 4d 10 2 0.81 0.67 1.49
Ag +1 1 4d 10 4 1.14 1 1.00
Ag +1 1 4d 10 4 1.16 1.02 0.98
Ag +1 1 4d 10 5 1.23 1.09 0.92
Ag +1 1 4d 10 6 1.29 1.15 0.87
Ag +1 1 4d 10 7 1.36 1.22 0.82
Ag +1 1 4d 10 8 1.42 1.28 0.78
Ag +2 2 4d 9 4 0.93 0.79 2.53
Ag +2 2 4d 9 6 1.08 0.94 2.13
Ag +3 3 4d 8 4 0.81 0.67 4.48
Ag +3 3 4d 8 6 0.89 0.75 4.00
Al +3 3 2p 6 4 0.53 0.39 7.69
Al +3 3 2p 6 5 0.62 0.48 6.25
Al +3 3 2p 6 6 0.675 0.535 5.61
Am +2 2 5f 7 7 1.35 1.21 1.65
Am +2 2 5f 7 8 1.4 1.26 1.59
Am +2 2 5f 7 9 1.45 1.31 1.53
Am +3 3 5f 6 6 1.115 0.975 3.08
Am +3 3 5f 6 8 1.23 1.09 2.75
Am +4 4 5f 5 6 0.99 0.85 4.71
Am +4 4 5f 5 8 1.09 0.95 4.21
As +3 3 4s 2 6 0.72 0.58 5.17
As +5 5 3d 10 4 0.475 0.335 14.93
As +5 5 3d 10 6 0.6 0.46 10.87
At +7 7 5d 10 6 0.76 0.62 11.29
Au +1 1 5d 10 6 1.51 1.37 0.73
Au +3 3 5d 8 4 0.82 0.68 4.41
Au +3 3 5d 8 6 0.99 0.85 3.53
Au +5 5 5d 6 6 0.71 0.57 8.77
B +3 3 1s 2 3 0.15 0.01 300.00
B +3 3 1s 2 4 0.25 0.11 27.27
B +3 3 1s 2 6 0.41 0.27 11.11
Ba +2 2 5p 6 6 1.49 1.35 1.48
Ba +2 2 5p 6 7 1.52 1.38 1.45
Ba +2 2 5p 6 8 1.56 1.42 1.41
Ba +2 2 5p 6 9 1.61 1.47 1.36
Ba +2 2 5p 6 10 1.66 1.52 1.32
Ba +2 2 5p 6 11 1.71 1.57 1.27
Ba +2 2 5p 6 12 1.75 1.61 1.24
Be +2 2 1s 2 3 0.3 0.16 12.50
Be +2 2 1s 2 4 0.41 0.27 7.41
Be +2 2 1s 2 6 0.59 0.45 4.44
Bi +3 3 6s 2 5 1.1 0.96 3.13
Bi +3 3 6s 2 6 1.17 1.03 2.91
Bi +3 3 6s 2 8 1.31 1.17 2.56
Bi +5 5 5d 10 6 0.9 0.76 6.58
Bk +3 3 5f 8 6 1.1 0.96 3.13
Bk +4 4 5f 7 6 0.97 0.83 4.82
Bk +4 4 5f 7 8 1.07 0.93 4.30
Br -1 -1 4p 6 6 1.82 1.96 -0.51
Br +3 3 4p 2 4 0.73 0.59 5.08
Br +5 5 4s 2 3 0.45 0.31 16.13
Br +7 7 3d 10 4 0.39 0.25 28.00
Table 1. Ionic Radii from Shannon (1976)
ION
Oxidation
State
Elec.
Config.
Coord.
#
Spin
State
Crystal
Radius (A)
Ionic
Radius
(A)
Z/IR
Br +7 7 3d 10 6 0.53 0.39 17.95
C +4 4 1s 2 3 0.06 -0.08 -50.00
C +4 4 1s 2 4 0.29 0.15 26.67
C +4 4 1s 2 6 0.3 0.16 25.00
Ca +2 2 3p 6 6 1.14 1 2.00
Ca +2 2 3p 6 7 1.2 1.06 1.89
Ca +2 2 3p 6 8 1.26 1.12 1.79
Ca +2 2 3p 6 9 1.32 1.18 1.69
Ca +2 2 3p 6 10 1.37 1.23 1.63
Ca +2 2 3p 6 12 1.4 1.34 1.49
Cd +2 2 4d 10 4 0.92 0.78 2.56
Cd +2 2 4d 10 5 1.01 0.87 2.30
Cd +2 2 4d 10 6 1.09 0.95 2.11
Cd +2 2 4d 10 7 1.17 1.03 1.94
Cd +2 2 4d 10 8 1.24 1.1 1.82
Cd +2 2 4d 10 12 1.45 1.31 1.53
Ce +3 3 6s 1 6 1.15 1.01 2.97
Ce +3 3 6s 1 7 1.21 1.07 2.80
Ce +3 3 6s 1 8 1.283 1.143 2.62
Ce +3 3 6s 1 9 1.336 1.196 2.51
Ce +3 3 6s 1 10 1.39 1.25 2.40
Ce +3 3 6s 1 12 1.48 1.34 2.24
Ce +4 4 5p 6 6 1.01 0.87 4.60
Ce +4 4 5p 6 8 1.11 0.97 4.12
Ce +4 4 5p 6 10 1.21 1.07 3.74
Ce +4 4 5p 6 12 1.28 1.14 3.51
Cf +3 3 6d 1 6 1.09 0.95 3.16
Cf +4 4 5f 8 6 0.961 0.821 4.87
Cf +4 4 5f 8 8 1.06 0.92 4.35
Cl-1 -1 3p 6 6 1.67 1.81 -0.55
Cl+5 5 3s 2 3 0.26 0.12 41.67
Cl+7 7 2p 6 4 0.22 0.08 87.50
Cl+7 7 2p 6 6 0.41 0.27 25.93
Cm+3 3 5f 7 4 1.11 0.97 3.09
Cm+4 4 5f 6 4 0.99 0.85 4.71
Cm+4 4 5f 6 8 1.09 0.95 4.21
Co+2 2 3d 7 4 HS 0.72 0.58 3.45
Co+2 2 3d 7 5 0.81 0.67 2.99
Co+2 2 3d 7 6 LS 0.79 0.65 3.08
Co+2 2 3d 7 6 HS 0.885 0.745 2.68
Co+2 2 3d 7 8 1.04 0.9 2.22
Co+3 3 3d 6 6 LS 0.685 0.545 5.50
Co+3 3 3d 6 6 HS 0.75 0.61 4.92
Co+4 4 3d 5 4 0.54 0.4 10.00
Co+4 4 3d 5 6 HS 0.67 0.53 7.55
Cr+2 2 3d 4 6 LS 0.87 0.73 2.74
Cr+2 2 3d 4 6 HS 0.94 0.8 2.50
Cr+3 3 3d 3 6 0.755 0.615 4.88
Cr+4 4 3d 2 4 0.55 0.41 9.76
Cr+4 4 3d 2 6 0.69 0.55 7.27
Cr+5 5 3d 1 4 0.485 0.345 14.49
Cr+5 5 3d 1 6 0.63 0.49 10.20
Cr+5 5 3d 1 8 0.71 0.57 8.77
Cr+6 6 3p 6 4 0.4 0.26 23.08
Table 1. Ionic Radii from Shannon (1976)
ION
Oxidation
State
Elec.
Config.
Coord.
#
Spin
State
Crystal
Radius (A)
Ionic
Radius
(A)
Z/IR
Cr+6 6 3p 6 6 0.58 0.44 13.64
Cs+1 1 5p 6 6 1.81 1.67 0.60
Cs+1 1 5p 6 8 1.88 1.74 0.57
Cs+1 1 5p 6 9 1.92 1.78 0.56
Cs+1 1 5p 6 10 1.95 1.81 0.55
Cs+1 1 5p 6 11 1.99 1.85 0.54
Cs+1 1 5p 6 12 2.02 1.88 0.53
Cu+1 1 3d 10 2 0.6 0.46 2.17
Cu+1 1 3d 10 4 0.74 0.6 1.67
Cu+1 1 3d 10 6 0.91 0.77 1.30
Cu+2 2 3d 9 4 0.71 0.57 3.51
Cu+2 2 3d 9 4 0.71 0.57 3.51
Cu+2 2 3d 9 5 0.79 0.65 3.08
Cu+2 2 3d 9 6 0.87 0.73 2.74
Cu+3 3 3d 8 6 LS 0.68 0.54 5.56
D+1 1 1s 0 2 0.04 -0.1 -10.00
Dy+2 2 4f 10 6 1.21 1.07 1.87
Dy+2 2 4f 10 7 1.27 1.13 1.77
Dy+2 2 4f 10 8 1.33 1.19 1.68
Dy+3 3 4f 9 6 1.052 0.912 3.29
Dy+3 3 4f 9 7 1.11 0.97 3.09
Dy+3 3 4f 9 8 1.167 1.027 2.92
Dy+3 3 4f 9 9 1.223 1.083 2.77
Er+3 3 4f 11 6 1.03 0.89 3.37
Er+3 3 4f 11 7 1.085 0.945 3.17
Er+3 3 4f 11 8 1.144 1.004 2.99
Er+3 3 4f 11 9 1.202 1.062 2.82
Eu+2 2 4f 7 6 1.31 1.17 1.71
Eu+2 2 4f 7 7 1.34 1.2 1.67
Eu+2 2 4f 7 8 1.39 1.25 1.60
Eu+2 2 4f 7 9 1.44 1.3 1.54
Eu+2 2 4f 7 10 1.49 1.35 1.48
Eu+3 3 4f 6 6 1.087 0.947 3.17
Eu+3 3 4f 6 7 1.15 1.01 2.97
Eu+3 3 4f 6 8 1.206 1.066 2.81
Eu+3 3 4f 6 9 1.26 1.12 2.68
F-1 -1 2p 6 2 1.145 1.285 -0.78
F-1 -1 2p 6 3 1.16 1.3 -0.77
F-1 -1 2p 6 4 1.17 1.31 -0.76
F-1 -1 2p 6 6 1.19 1.33 -0.75
F+7 7 1s 2 6 0.22 0.08 87.50
Fe+2 2 3d 6 4 HS 0.77 0.63 3.17
Fe+2 2 3d 6 4 HS 0.78 0.64 3.13
Fe+2 2 3d 6 6 LS 0.75 0.61 3.28
Fe+2 2 3d 6 6 HS 0.92 0.78 2.56
Fe+2 2 3d 6 8 HS 1.06 0.92 2.17
Fe+3 3 3d 5 4 HS 0.63 0.49 6.12
Fe+3 3 3d 5 5 0.72 0.58 5.17
Fe+3 3 3d 5 6 LS 0.69 0.55 5.45
Fe+3 3 3d 5 6 HS 0.785 0.645 4.65
Fe+3 3 3d 5 8 HS 0.92 0.78 3.85
Fe+4 4 3d 4 6 0.725 0.585 6.84
Fe+6 6 3d2 4 0.39 0.25 24.00
Fr+1 1 6p 6 6 1.94 1.8 0.56
Table 1. Ionic Radii from Shannon (1976)
ION
Oxidation
State
Elec.
Config.
Coord.
#
Spin
State
Crystal
Radius (A)
Ionic
Radius
(A)
Z/IR
Ga+3 3 3d 10 4 0.61 0.47 6.38
Ga+3 3 3d 10 5 0.69 0.55 5.45
Ga+3 3 3d 10 6 0.76 0.62 4.84
Gd+3 3 4f 7 6 1.078 0.938 3.20
Gd+3 3 4f 7 7 1.14 1 3.00
Gd+3 3 4f 7 8 1.193 1.053 2.85
Gd+3 3 4f 7 9 1.247 1.107 2.71
Ge+2 2 4s 2 6 0.87 0.73 2.74
Ge+4 4 3d 10 4 0.53 0.39 10.26
Ge+4 4 3d 10 6 0.67 0.53 7.55
H+1 1 1s 0 1 -0.24 -0.38 -2.63
H+1 1 1s 0 2 -0.04 -0.18 -5.56
Hf+4 4 4f 14 4 0.72 0.58 6.90
Hf+4 4 4f 14 6 0.85 0.71 5.63
Hf+4 4 4f 14 7 0.9 0.76 5.26
Hf+4 4 4f 14 8 0.97 0.83 4.82
Hg+1 1 6s 1 3 1.11 0.97 1.03
Hg+1 1 6s 1 6 1.33 1.19 0.84
Hg+2 2 5d 10 2 0.83 0.69 2.90
Hg+2 2 5d 10 4 1.1 0.96 2.08
Hg+2 2 5d 10 6 1.16 1.02 1.96
Hg+2 2 5d 10 8 1.28 1.14 1.75
Ho+3 3 4f 10 6 1.041 0.901 3.33
Ho+3 3 4f 10 8 1.155 1.015 2.96
Ho+3 3 4f 10 9 1.212 1.072 2.80
Ho+3 3 4f 10 10 1.26 1.12 2.68
I-1 -1 5p 6 6 2.06 2.2 -0.45
I+5 5 5s 2 3 0.58 0.44 11.36
I+5 5 5s 2 6 1.09 0.95 5.26
I+7 7 4d 10 4 0.56 0.42 16.67
I+7 7 4d 10 6 0.67 0.53 13.21
In+3 3 4d 10 4 0.76 0.62 4.84
In+3 3 4d 10 6 0.94 0.8 3.75
In+3 3 4d 10 8 1.06 0.92 3.26
Ir+3 3 5d 6 6 0.82 0.68 4.41
Ir+4 4 5d 5 6 0.765 0.625 6.40
Ir+5 5 5d 4 6 0.71 0.57 8.77
K+1 1 3p 6 4 1.51 1.37 0.73
K+1 1 3p 6 6 1.52 1.38 0.72
K+1 1 3p 6 7 1.6 1.46 0.68
K+1 1 3p 6 8 1.65 1.51 0.66
K+1 1 3p 6 9 1.69 1.55 0.65
K+1 1 3p 6 10 1.73 1.59 0.63
K+1 1 3p 6 12 1.78 1.64 0.61
La+3 3 4d 10 6 1.172 1.032 2.91
La+3 3 4d 10 7 1.24 1.1 2.73
La+3 3 4d 10 8 1.3 1.16 2.59
La+3 3 4d 10 9 1.356 1.216 2.47
La+3 3 4d 10 10 1.41 1.27 2.36
La+3 3 4d 10 12 1.5 1.36 2.21
Li+1 1 1s 2 4 0.73 0.59 1.69
Li+1 1 1s 2 6 0.9 0.76 1.32
Li+1 1 1s 2 8 1.06 0.92 1.09
Lu+3 3 4f 14 6 1.001 0.861 3.48
Table 1. Ionic Radii from Shannon (1976)
ION
Oxidation
State
Elec.
Config.
Coord.
#
Spin
State
Crystal
Radius (A)
Ionic
Radius
(A)
Z/IR
Lu+3 3 4f 14 8 1.117 0.977 3.07
Lu+3 3 4f 14 9 1.172 1.032 2.91
Mg+2 2 2p 6 4 0.71 0.57 3.51
Mg+2 2 2p 6 5 0.8 0.66 3.03
Mg+2 2 2p 6 6 0.86 0.72 2.78
Mg+2 2 2p 6 8 1.03 0.89 2.25
Mn+2 2 3d 5 4 HS 0.8 0.66 3.03
Mn+2 2 3d 5 5 HS 0.89 0.75 2.67
Mn+2 2 3d 5 6 LS 0.81 0.67 2.99
Mn+2 2 3d 5 6 HS 0.97 0.83 2.41
Mn+2 2 3d 5 7 HS 1.04 0.9 2.22
Mn+2 2 3d 5 8 1.1 0.96 2.08
Mn+3 3 3d 4 5 0.72 0.58 5.17
Mn+3 3 3d 4 6 LS 0.72 0.58 5.17
Mn+3 3 3d 4 6 HS 0.785 0.645 4.65
Mn+4 4 3d 3 4 0.53 0.39 10.26
Mn+4 4 3d 3 6 0.67 0.53 7.55
Mn+5 5 3d 2 4 0.47 0.33 15.15
Mn+6 6 3d 1 4 0.395 0.255 23.53
Mn+7 7 3p 6 4 0.39 0.25 28.00
Mn+7 7 3p 6 6 0.6 0.46 15.22
Mo+3 3 4d 3 6 0.83 0.69 4.35
Mo+4 4 4d 2 6 0.79 0.65 6.15
Mo+5 5 4d 1 4 0.6 0.46 10.87
Mo+5 5 4d 1 6 0.75 0.61 8.20
Mo+6 6 4p 6 4 0.55 0.41 14.63
Mo+6 6 4p 6 5 0.64 0.5 12.00
Mo+6 6 4p 6 6 0.73 0.59 10.17
Mo+6 6 4p 6 7 0.87 0.73 8.22
N-3 -3 2p 6 4 1.32 1.46 -2.05
N+3 3 2s 2 6 0.3 0.16 18.75
N+5 5 1s 2 3 0.044 -0.104 -48.08
N+5 5 1s 2 4 0.27 0.13 38.46
Na+1 1 2p 6 4 1.13 0.99 1.01
Na+1 1 2p 6 5 1.14 1 1.00
Na+1 1 2p 6 6 1.16 1.02 0.98
Na+1 1 2p 6 7 1.26 1.12 0.89
Na+1 1 2p 6 8 1.32 1.18 0.85
Na+1 1 2p 6 9 1.38 1.24 0.81
Na+1 1 2p 6 12 1.53 1.39 0.72
Nb+3 3 4d 2 6 0.86 0.72 4.17
Nb+4 4 4d 1 6 0.82 0.68 5.88
Nb+4 4 4d 1 8 0.93 0.79 5.06
Nb+5 5 4p 6 4 0.62 0.48 10.42
Nb+5 5 4p 6 6 0.78 0.64 7.81
Nb+5 5 4p 6 7 0.83 0.69 7.25
Nb+5 5 4p 6 8 0.88 0.74 6.76
Nd+2 2 4f 4 8 1.43 1.29 1.55
Nd+2 2 4f 4 9 1.49 1.35 1.48
Nd+3 3 4f 3 6 1.123 0.983 3.05
Nd+3 3 4f 3 8 1.249 1.109 2.71
Nd+3 3 4f 3 9 1.303 1.163 2.58
Nd+3 3 4f 3 12 1.41 1.27 2.36
Ni+2 2 3d 8 4 0.69 0.55 3.64
Table 1. Ionic Radii from Shannon (1976)
ION
Oxidation
State
Elec.
Config.
Coord.
#
Spin
State
Crystal
Radius (A)
Ionic
Radius
(A)
Z/IR
Ni+2 2 3d 8 4 0.63 0.49 4.08
Ni+2 2 3d 8 5 0.77 0.63 3.17
Ni+2 2 3d 8 6 0.83 0.69 2.90
Ni+3 3 3d 7 6 LS 0.7 0.56 5.36
Ni+3 3 3d 7 6 HS 0.74 0.6 5.00
Ni+4 4 3d 6 6 LS 0.62 0.48 8.33
No+2 2 5f 14 6 1.24 1.1 1.82
Np+2 2 5f 5 6 1.24 1.1 1.82
Np+3 3 5f 4 6 1.15 1.01 2.97
Np+4 4 5f 3 6 1.01 0.87 4.60
Np+4 4 5f 3 8 1.12 0.98 4.08
Np+5 5 5f 2 6 0.89 0.75 6.67
Np+6 6 5f 1 6 0.86 0.72 8.33
Np+7 7 6p 6 6 0.85 0.71 9.86
O-2 -2 2p 6 2 1.21 1.35 -1.48
O-2 -2 2p 6 3 1.22 1.36 -1.47
O-2 -2 2p 6 4 1.24 1.38 -1.45
O-2 -2 2p 6 6 1.26 1.4 -1.43
O-2 -2 2p 6 8 1.28 1.42 -1.41
OH-1 -1 2 1.18 1.32 -0.76
OH-1 -1 3 1.2 1.34 -0.75
OH-1 -1 4 1.21 1.35 -0.74
OH-1 -1 6 1.23 1.37 -0.73
Os+4 4 5d 4 6 0.77 0.63 6.35
Os+5 5 5d 3 6 0.715 0.575 8.70
Os+6 6 5d 2 5 0.63 0.49 12.24
Os+6 6 5d 2 6 0.685 0.545 11.01
Os+7 7 5d 1 6 0.665 0.525 13.33
Os+8 8 5p 6 4 0.53 0.39 20.51
P+3 3 3s 2 6 0.58 0.44 6.82
P+5 5 2p 6 4 0.31 0.17 29.41
P+5 5 2p 6 5 0.43 0.29 17.24
P+5 5 2p 6 6 0.52 0.38 13.16
Pa+3 3 5f 2 6 1.18 1.04 2.88
Pa+4 4 6d 1 6 1.04 0.9 4.44
Pa+4 4 6d 1 8 1.15 1.01 3.96
Pa+5 5 6p 6 6 0.92 0.78 6.41
Pa+5 5 6p 6 8 1.05 0.91 5.49
Pa+5 5 6p 6 9 1.09 0.95 5.26
Pb+2 2 6s 2 4 1.12 0.98 2.04
Pb+2 2 6s 2 6 1.33 1.19 1.68
Pb+2 2 6s 2 7 1.37 1.23 1.63
Pb+2 2 6s 2 8 1.43 1.29 1.55
Pb+2 2 6s 2 9 1.49 1.35 1.48
Pb+2 2 6s 2 10 1.54 1.4 1.43
Pb+2 2 6s 2 11 1.59 1.45 1.38
Pb+2 2 6s 2 12 1.63 1.49 1.34
Pb+4 4 5d 10 4 0.79 0.65 6.15
Pb+4 4 5d 10 5 0.87 0.73 5.48
Pb+4 4 5d 10 6 0.915 0.775 5.16
Pb+4 4 5d 10 8 1.08 0.94 4.26
Pd+1 1 4d 9 2 0.73 0.59 1.69
Pd+2 2 4d 8 4 0.78 0.64 3.13
Pd+2 2 4d 8 6 1 0.86 2.33
Table 1. Ionic Radii from Shannon (1976)
ION
Oxidation
State
Elec.
Config.
Coord.
#
Spin
State
Crystal
Radius (A)
Ionic
Radius
(A)
Z/IR
Pd+3 3 4d 7 6 0.9 0.76 3.95
Pd+4 4 4d 6 6 0.755 0.615 6.50
Pm+3 3 4f 4 6 1.11 0.97 3.09
Pm+3 3 4f 4 8 1.233 1.093 2.74
Pm+3 3 4f 4 9 1.284 1.144 2.62
Po+4 4 6s 2 6 1.08 0.94 4.26
Po+4 4 6s 2 8 1.22 1.08 3.70
Po+6 6 5d 10 6 0.81 0.67 8.96
Pr+3 3 4f 2 6 1.13 0.99 3.03
Pr+3 3 4f 2 8 1.266 1.126 2.66
Pr+3 3 4f 2 9 1.319 1.179 2.54
Pr+4 4 4f 1 6 0.99 0.85 4.71
Pr+4 4 4f 1 8 1.1 0.96 4.17
Pt+2 2 5d 8 4 0.74 0.6 3.33
Pt+2 2 5d 8 6 0.94 0.8 2.50
Pt+4 4 5d 6 6 0.765 0.625 6.40
Pt+5 5 5d 5 6 0.71 0.57 8.77
Pu+3 3 5f 5 6 1.14 1 3.00
Pu+4 4 5f 4 6 1 0.86 4.65
Pu+4 4 5f 4 8 1.1 0.96 4.17
Pu+5 5 5f 3 6 0.88 0.74 6.76
Pu+6 6 5f 2 6 0.85 0.71 8.45
Ra+2 2 6p 6 8 1.62 1.48 1.35
Ra+2 2 6p 6 12 1.84 1.7 1.18
Rb+1 1 4p 6 6 1.66 1.52 0.66
Rb+1 1 4p 6 7 1.7 1.56 0.64
Rb+1 1 4p 6 8 1.75 1.61 0.62
Rb+1 1 4p 6 9 1.77 1.63 0.61
Rb+1 1 4p 6 10 1.8 1.66 0.60
Rb+1 1 4p 6 11 1.83 1.69 0.59
Rb+1 1 4p 6 12 1.86 1.72 0.58
Rb+1 1 4p 6 14 1.97 1.83 0.55
Re+4 4 5d 3 6 0.77 0.63 6.35
Re+5 5 5d 2 6 0.72 0.58 8.62
Re+6 6 5d 1 6 0.69 0.55 10.91
Re+7 7 5p 6 4 0.52 0.38 18.42
Re+7 7 5p 6 6 0.67 0.53 13.21
Rh+3 3 4d 6 6 0.805 0.665 4.51
Rh+4 4 4d 5 6 0.74 0.6 6.67
Rh+5 5 4d 4 6 0.69 0.55 9.09
Ru+3 3 4d 5 6 0.82 0.68 4.41
Ru+4 4 4d 4 6 0.76 0.62 6.45
Ru+5 5 4d 3 6 0.705 0.565 8.85
Ru+7 7 4d 1 4 0.52 0.38 18.42
Ru+8 8 4p 6 4 0.5 0.36 22.22
S-2 -2 3p 6 6 1.7 1.84 -1.09
S+4 4 3s 2 6 0.51 0.37 10.81
S+6 6 2p 6 4 0.26 0.12 50.00
S+6 6 2p 6 6 0.43 0.29 20.69
Sb+3 3 5s 2 4 0.9 0.74 4.05
Sb+3 3 5s 2 5 0.94 0.8 3.75
Sb+3 3 5s 2 6 0.9 0.76 3.95
Sb+5 5 4d 10 6 0.74 0.6 8.33
Sc+3 3 3p 6 6 0.885 0.745 4.03
Table 1. Ionic Radii from Shannon (1976)
ION
Oxidation
State
Elec.
Config.
Coord.
#
Spin
State
Crystal
Radius (A)
Ionic
Radius
(A)
Z/IR
Sc+3 3 3p 6 8 1.01 0.87 3.45
Se-2 -2 4p 6 6 1.84 1.98 -1.01
Se+4 4 4s 2 6 0.64 0.5 8.00
Se+6 6 3d 10 4 0.42 0.28 21.43
Se+6 6 3d 10 6 0.56 0.42 14.29
Si+4 4 2p 6 4 0.4 0.26 15.38
Si+4 4 2p 6 6 0.54 0.4 10.00
Sm+2 2 4f 6 7 1.36 1.22 1.64
Sm+2 2 4f 6 8 1.41 1.27 1.57
Sm+2 2 4f 6 9 1.46 1.32 1.52
Sm+3 3 4f 5 6 1.098 0.958 3.13
Sm+3 3 4f 5 7 1.16 1.02 2.94
Sm+3 3 4f 5 8 1.219 1.079 2.78
Sm+3 3 4f 5 9 1.272 1.132 2.65
Sm+3 3 4f 5 12 1.38 1.24 2.42
Sn+4 4 4d 10 4 0.69 0.55 7.27
Sn+4 4 4d 10 5 0.76 0.62 6.45
Sn+4 4 4d 10 6 0.83 0.69 5.80
Sn+4 4 4d 10 7 0.89 0.75 5.33
Sn+4 4 4d 10 8 0.95 0.81 4.94
Sr+2 2 4p 6 6 1.32 1.18 1.69
Sr+2 2 4p 6 7 1.35 1.21 1.65
Sr+2 2 4p 6 8 1.4 1.26 1.59
Sr+2 2 4p 6 9 1.45 1.31 1.53
Sr+2 2 4p 6 10 1.5 1.36 1.47
Sr+2 2 4p 6 12 1.58 1.44 1.39
Ta+3 3 5d 2 6 0.86 0.72 4.17
Ta+4 4 5d 1 6 0.82 0.68 5.88
Ta+5 5 5p 6 6 0.78 0.64 7.81
Ta+5 5 5p 6 7 0.83 0.69 7.25
Ta+5 5 5p 6 8 0.88 0.74 6.76
Tb+3 3 4f 8 6 1.063 0.923 3.25
Tb+3 3 4f 8 7 1.12 0.98 3.06
Tb+3 3 4f 8 8 1.18 1.04 2.88
Tb+3 3 4f 8 9 1.235 1.095 2.74
Tb+4 4 4f 7 6 0.9 0.76 5.26
Tb+4 4 4f 7 8 1.02 0.88 4.55
Tc+4 4 4d 3 6 0.785 0.645 6.20
Tc+5 5 4d 2 6 0.74 0.6 8.33
Tc+7 7 4p 6 4 0.51 0.37 18.92
Tc+7 7 4p 6 6 0.7 0.56 12.50
Te(-2) -2 5p 6 6 2.07 2.21 -0.90
Te+4 4 5s 2 3 0.66 0.52 7.69
Te+4 4 5s 2 4 0.8 0.66 6.06
Te+4 4 5s 2 6 1.11 0.97 4.12
Te+6 6 4d 10 4 0.57 0.43 13.95
Te+6 6 4d 10 6 0.7 0.56 10.71
Th+4 4 6p 6 6 1.08 0.94 4.26
Th+4 4 6p 6 8 1.19 1.05 3.81
Th+4 4 6p 6 9 1.23 1.09 3.67
Th+4 4 6p 6 10 1.27 1.13 3.54
Th+4 4 6p 6 11 1.32 1.18 3.39
Th+4 4 6p 6 12 1.35 1.21 3.31
Ti+2 2 3d 2 6 1 0.86 2.33
Table 1. Ionic Radii from Shannon (1976)
ION
Oxidation
State
Elec.
Config.
Coord.
#
Spin
State
Crystal
Radius (A)
Ionic
Radius
(A)
Z/IR
Ti+3 3 3d 1 6 0.81 0.67 4.48
Ti+4 4 3p 6 4 0.56 0.42 9.52
Ti+4 4 3p 6 5 0.65 0.51 7.84
Ti+4 4 3p 6 6 0.745 0.605 6.61
Ti+4 4 3p 6 8 0.88 0.74 5.41
Tl+1 1 6s 2 6 1.64 1.5 0.67
Tl+1 1 6s 2 8 1.73 1.59 0.63
Tl+1 1 6s 2 12 1.84 1.7 0.59
Tl+3 3 5d 10 4 0.89 0.75 4.00
Tl+3 3 5d 10 6 1.025 0.885 3.39
Tl+3 3 5d 10 8 1.12 0.98 3.06
Tm+2 2 4f 13 6 1.17 1.03 1.94
Tm+2 2 4f 13 7 1.23 1.09 1.83
Tm+3 3 4f 12 6 1.02 0.88 3.41
Tm+3 3 4f 12 8 1.134 0.994 3.02
Tm+3 3 4f 12 9 1.192 1.052 2.85
U+3 3 5f 3 6 1.165 1.025 2.93
U+4 4 5f 2 6 1.03 0.89 4.49
U+4 4 5f 2 7 1.09 0.95 4.21
U+4 4 5f 2 8 1.14 1 4.00
U+4 4 5f 2 9 1.19 1.05 3.81
U+4 4 5f 2 12 1.31 1.17 3.42
U+5 5 5f 1 6 0.9 0.76 6.58
U+5 5 5f 1 7 0.98 0.84 5.95
U+6 6 6p 6 2 0.59 0.45 13.33
U+6 6 6p 6 4 0.66 0.52 11.54
U+6 6 6p 6 6 0.87 0.73 8.22
U+6 6 6p 6 7 0.95 0.81 7.41
U+6 6 6p 6 8 1 0.86 6.98
V+2 2 3d 3 6 0.93 0.79 2.53
V+3 3 3d 2 6 0.78 0.64 4.69
V+4 4 3d 1 5 0.67 0.53 7.55
V+4 4 3d 1 6 0.72 0.58 6.90
V+4 4 3d 1 8 0.86 0.72 5.56
V+5 5 3p 6 4 0.495 0.355 14.08
V+5 5 3p 6 5 0.6 0.46 10.87
V+5 5 3p 6 6 0.68 0.54 9.26
W+4 4 5d 2 6 0.8 0.66 6.06
W+5 5 5d 1 6 0.76 0.62 8.06
W+6 6 5p 6 4 0.56 0.42 14.29
W+6 6 5p 6 5 0.65 0.51 11.76
W+6 6 5p 6 6 0.74 0.6 10.00
Xe+8 8 4d 10 4 0.54 0.4 20.00
Xe+8 8 4d 10 6 0.62 0.48 16.67
Y+3 3 4p 6 6 1.04 0.9 3.33
Y+3 3 4p 6 7 1.1 0.96 3.13
Y+3 3 4p 6 8 1.159 1.019 2.94
Y+3 3 4p 6 9 1.215 1.075 2.79
Yb+2 2 4f 14 6 1.16 1.02 1.96
Yb+2 2 4f 14 7 1.22 1.08 1.85
Yb+2 2 4f 14 8 1.28 1.14 1.75
Yb+3 3 4f 13 6 1.008 0.868 3.46
Yb+3 3 4f 13 7 1.065 0.925 3.24
Yb+3 3 4f 13 8 1.125 0.985 3.05
Table 1. Ionic Radii from Shannon (1976)
ION
Oxidation
State
Elec.
Config.
Coord.
#
Spin
State
Crystal
Radius (A)
Ionic
Radius
(A)
Z/IR
Yb+3 3 4f 13 9 1.182 1.042 2.88
Zn+2 2 3d 10 4 0.74 0.6 3.33
Zn+2 2 3d 10 5 0.82 0.68 2.94
Zn+2 2 3d 10 6 0.88 0.74 2.70
Zn+2 2 3d 10 8 1.04 0.9 2.22
Zr+4 4 4p 6 4 0.73 0.59 6.78
Zr+4 4 4p 6 5 0.8 0.66 6.06
Zr+4 4 4p 6 6 0.86 0.72 5.56
Zr+4 4 4p 6 7 0.92 0.78 5.13
Zr+4 4 4p 6 8 0.98 0.84 4.76
Zr+4 4 4p 6 9 1.03 0.89 4.49
ESCI 430 C-T Lee
elements dominant valence state and coordination number is required. For many
elements, only one or two valence states occur under the oxygen fugacities of most
natural conditions. The coordination number of an atom is defined as the number of
surrounding ligands that are bonded to the central atom. The coordination number can be
estimated from the geometry of the crystallographic site in which a central cation is
located. If this is not known, an approximate coordination number can be estimated
using a rough estimate of the anion-cation radius ratio, R
A
:R
X
. As a general rule of
thumb, the minimum radius ratios corresponding to particular coordination numbers (CN)
are as follows: <0.155 (CN =1), 0.155 (CN=3), 0.225 (CN=4), 0.414 (CN=6), 0.732
(CN=8), 1.0 (CN=12). Once the coordination number and valence state is known, the
ionic radius can be determined.
In 1937, Victor Goldschmidt outlined these generalizations, in what has come to
be known as the Goldschmidt Rules. They are as follows:
Ions of similar radii and the same charge will enter into a mineral in amounts
proportional to their concentration in the liquid.
An ion of smaller radius but with the same charge as another will be incorporated
preferentially into a growing crystal. In other words, for a given charge, the ion with
the higher field strength or ionic potential (Z/IR, where Z is the charge and IR is the
ionic radius) will be preferentially incorporated into a crystal.
An ion of the same radius but with a higher charge than another will be incorporated
preferentially into a growing crystal.
The Goldschmidt Rules, however, should only be taken as rough generalizations. A more
quantitative approach has been provided by Wood and Blundy (1997), which we describe
below.

7.3.2 A physical model for trace-element partitioning
Wood and Blundy (1997) presented a quantitative model to describe the partitioning
of trace elements between clinopyroxene and silicate melt as a function of pressure,
temperature, bulk composition, and the elastic properties of the cationic site in which
trace-elements enter. Qualitatively, their model rests on the property that substituting a
cation with a different ionic radius than that of the major cation generates a strain energy
that is proportional to a measure of the stiffness (Youngs Modulus, E) of the cation-
oxygen bonds in that site. If a substituting cation is larger than the major cation, it will
exert a positive strain on the site. If the substituting cation is smaller, it will exert a
negative strain on the site. The strain energy (Brice, 1975) for a perfectly spherical site is
given by
|
.
|

\
|
+ =
3 2
) (
3
1
) (
2
4
o i o i
o
A strain
r r r r
r
EN G 7.10
where E is the Youngs Modulus, N
A
is Avogadros number, r
o
is the radius of the site,
and r
i
is the radius of the substituting cation. It can be seen that the strain energy has a
largely parabolic dependence on cationic radius, exhibiting a minimum strain when the
substituting cation has an identical radius to the site radius r
o
. Wood and Blundy (1997)
showed that the partition coefficient of an element substituting into a given site is
expressed as

ESCI 430 C-T Lee
|
|
|
|
|
.
|

\
|
|
.
|

\
|
+
=
RT
r r r r
r
EN
D D
o i o i
o
A
o i
3 2
) (
3
1
) (
2
4
exp

7.11
where R represents the gas constant, T is temperature, and D
o
represents the partition
coefficient of the major cation in that site, that is, the case in which the strain in that site
is zero. The value of D
o
follows from Eq. 7.6 and therefore depends on temperature,
pressure, and bulk composition of the mineral. Examples of Eq. 7.11 for clinopyrxoene
are shown in Figure 1 along with experimental partitioning data (Figure taken directly
from Wood and Blundy, 1997). For a given charge, Eq. 7.11 represents a parabola,
whose maximum is represented by D
o
and whose curvature (or tightness) depends on
Youngs Modulus, E. Youngs Modulus is proportional to the bulk modulus K (E =
3K/(1-2) where is Poissons ratio), which itself is a linear function of the cation
valences
o o
c a
V d
n e Z AZ
K
9
) 1 (
2

= 7.12
where A is the Madelung constant, Z
a
and Z
c
are the anion and cation valences, V
o
is the
molecular volume, e is the electron charge, n is the Born constant, and d
o
is the
interatomic separation. Because K (hence E) increases with increasing cationic charge, it
can be seen from Eq. 7.11 and Figure 1 that the higher the charge, the tighter the
parabola. Qualitatively, this results from the fact that the higher the charge, the higher
the Coulombic anion-cation attraction and hence the stronger the bond. The actual height
FIGURE 7.1. The Wood and Blundy (1997) model for trace-element partitioning between
clinopyroxene and silicate melt using Eq. 7.11. This figure is reproduced exactly from Figure
4a,b in Wood and Blundy. Data in a are experimental data from Blundy and Dalton
(unpublished). Data in b are experimental data from Hauri et al. (1994).


ESCI 430 C-T Lee
of each parabola depends on the equilibrium partition coefficient of the strain-free cation,
which can be determined by thermodynamic considerations.
The values of Wood and Blundys formulations are two-fold. First, they provide
a simple physical explanation for the controls of trace-element partitioning. Second, they
provide a means of predicting partition coefficients when experimental data are not
available. For example, one can use the partition coefficient of Ca between
clinopyroxene and silicate melt to predict the partition coefficient of a rare-earth element
between clinopyroxene and melt given knowledge of the cationic radii of Ca and the rare-
earth element.

7.3.3 Partition coefficients

Table 2. Partition Coefficients for basaltic systems (adapted from compilations in Rollinson (1993)
Olivine OPX CPX Hornblende Phlogopite Plagioclase Garnet
Rb 0.0098 0.022 0.031 0.29 3.06 0.071 0.042
Sr 0.014 0.04 0.06 0.46 0.081 1.83 0.012
Ba 0.0099 0.013 0.026 0.42 1.09 0.23 0.023
K 0.0068 0.014 0.038 0.96 0.17 0.015
Y 0.01 0.18 0.9 1 0.03 0.03 9
Ti 0.02 0.1 0.4 1.5 0.9 0.04 0.3
Zr 0.012 0.18 0.1 0.5 0.6 0.048 0.65
Hf 0.013 0.263 0.5 0.051 0.45
Nb 0.01 0.15 0.005 0.8 1 0.01 0.02
Ta 0.013 0.06
Th 0.03 0.5 0.01
U 0.002 0.04 0.1 0.01
La 0.056
Ce 0.0069 0.02 0.092 0.2 0.034 0.12 0.03
Pr
Nd 0.0066 0.03 0.23 0.33 0.032 0.081 0.07
Sm 0.0066 0.05 0.445 0.52 0.031 0.067 0.29
Eu 0.0068 0.05 0.474 0.4 0.03 0.34 0.49
Gd 0.0077 0.09 0.556 0.63 0.03 0.063 0.97
Tb 0.57
Dy 0.0096 0.15 0.582 0.64 0.03 0.055 3.17
Ho
Er 0.011 0.23 0.583 0.55 0.034 0.063 6.56
Tm
Yb 0.014 0.34 0.542 0.49 0.042 0.067 11.5
Lu 0.016 0.42 0.506 0.43 0.046 0.06 11.9
Ni 5.9-29 5 1.5-14 6.8
Co 6.6 2-4 0.5-2.0 2
V 0.06 0.6 1.35 3.4
Cr 0.7 10 34 12.5
Sc 0.17 1.2 1.7-3.2 2.2-4.2
Mn 1.45 1.4 .3-1.2

One of the problems faced by geochemists is choosing which element or elements
to use as a tool in unraveling geochemical processes. There are ~87 naturally occurring
elements, making it a formidable task of choosing which element to investigate. It is
particularly frustrating to non-geochemists as the choice of elements to study often seems
random. Elements are chosen for study according to their partitioning behaviors. As the
partitioning behaviors of trace-elements are systematically different, each trace element
may provide different information about the evolution of a system. Table 2 shows a
selection of partition coefficients for various minerals in basaltic systems. Below, we
provide a brief overview of how various trace-elements are used to unravel geochemical
processes.

ESCI 430 C-T Lee
Alkali Earths (Li, Na, K, Rb, and Cs)
The alkali Earth metals all have one unpaired electron in their outer s-valence shell. For
this reason, alkali Earths tend to pair up very easily with the halogens, which need one
electron to fill up their valence shell. For a given series, the alkali Earths have the largest
atomic radii, and therefore, their charge to ionic radii are small. These elements are also
known as the large ion lithophile elements (LILEs). These properties make the alkali
Earths very soluble in aqueous solutions. Their large ionic radii and their monovalent
states make them generally incompatible in most silicate minerals. However, some
minerals, such as micas, are highly enriched in these elements because micas have a
monovalent site in their structure. Seawater and the continental crust are highly enriched
in these elements because aqueous processes occurring during subduction zone
magmatism and during weathering effectively mobilize these elements. A particularly
good way of detecting the role of aqueous processes in various petrologic processes is to
monitor the ratio between a LILE and an element that is not soluble in aqueous fluids,
such as a rare-earth element or high field strength element (e.g, Zr, Nb, Ti, Ta). Because
of their incompatible nature and fluid-mobile properties, the alkali Earths are highly
depleted in the mantle. Fifty percent of the Earths K, Rb, Na, and Cs may be in the
continental crust and oceans (some recent high pressure experiments suggest that K might
be in the core). The exact bulk Earth budget of K, Rb, Cs, and Na in the Earth is not
known as these elements were moderately volatile during the condensation of the solar
nebula and accretion of the Earth. High concentrations of alkali Earths occur in
evaporate deposits (Na, K) or in hydrothermal solutions, such as pegmatitic fluids (Li,
Rb, Cs). K and Rb have radioactive isotopes (
40
K and
87
Rb). These decay systems have
been used to date a variety of rocks, particularly granitic rocks which are highly enriched
in Rb and K.

Group 2 large ion lithophile trace elements (Be, Mg, Ca, Sr, Ba)
The group 2 elements have two electrons in their valence shell. These two
elements are readily given up two atoms with high electron affinities (e.g., halogens,
metalloids) and therefore these elements often form ionic bonds (halides and oxides). Mg
and Ca are abundant elements in most terrestrial systems. Be, Sr, and Ba are trace
lithophile elements. Due to their similar ionic radii, Sr and Ca often substitute for each
other. Sr is often enriched in plagioclase (CaAl
2
Si
2
O
8
) and calcite (CaCO
3
).

The Rare Earth Elements (La-Lu, Sc, Y)
The rare-earth elements include the Lanthanides (La-Lu), Sc and Y. These
elements have valence electrons in the d (La, Sc, Y) or f orbitals (Ce to Lu). The
dominant valence state of the REEs is 3 under the oxygen fugacity conditions on Earth.
The charge to ionic radius of the REEs is ~ 3, much higher than that of the alkali Earth
metals. A close look at the Periodic Table will reveal that in going to the right in the
Lanthanide series, the number of protons increases while the number of electrons
increases to gradually fill the f-orbital. Because the filling of electrons does not involve
an increase in energy level, the end result is that the atomic radii of the Lanthanides
contracts with increasing Z due to the increased Coulombic attraction caused by adding
more protons into the nucleus. In most silicate minerals, the REEs are moderately
incompatible (0.1<D<1) because in these minerals the cation sites are divalent.
Substituting REE
+3
into a divalent site requires simultaneous substitution of another

ESCI 430 C-T Lee
element to achieve charge balance or a charge vacancy must be generated. Both of these
processes are energetically less favorable than substituting in a divalent cation into a
divalent site. We know from Goldschmidts Rules that the cation with the largest radius
will be more compatible. Therefore, we expect La to be more incompatible than Lu. For
most silicate minerals (see Table 2), partition coefficients of the light rare earth elements
(LREEs = La to Eu) are lower than that of the heavy rare earths (HREEs = Gd to Lu). It
therefore follows that partial melting should lead to melts that are enriched in LREEs
over HREEs. For very silicic melt compositions (rhyolites), the REEs may actually
become compatible in pyroxene because such melts are so highly polymerized that there
are few ways to incorporate the REEs into the melt. One of the most important minerals
that fractionate the REEs is garnet. The LREEs are highly incompatible in garnet but the
HREEs are compatible. Thus, a partial melt from a garnet-bearing source will exhibit
extreme HREE depletion relative to the LREEs. Only a few accessory minerals
preferentially incorporate the LREEs over the HREEs. Because of similar ionic radii, Sc
is found to behave like Yb, and Y behaves like Ho. Eu is an unusual REE in that it has
two oxidation states (+2 and +3) occurring over the range of oxygen fugacity conditions
on Earth. This allows Eu
+2
to be more easily incorporated into divalent cation sites. The
effect is most pronounced for the mineral plagioclase (CaAl
2
Si
2
O
8
) where it substitutes
for Ca. Magmas that have experienced plagioclase crystallization invariably show
depletion in Eu relative to the REEs. The systematic variation in REE partition
coefficients makes the REEs a powerful tool in deciphering the sources and styles of
partial melting. The continental crust is enriched in LREEs over HREEs, while mid-
ocean ridge basalts tend to be LREE depleted relative to the HREEs. The LREE-depleted
character of mid-ocean ridge basalts is an inherited feature of its mantle source, which is
itself LREE-depleted due to previous extraction of continental crust. The alpha decay of
147
Sm to
143
Nd has been one of the most successful radiogenic isotopic tools used to
decipher the differentiation history of the silicate parts of the Earth.
The REEs tend to be insoluble in aqueous fluids. Their concentrations are
therefore very low in freshwater and seawater, and they have very limited fluid mobility
in hydrothermal systems. Because of their low solubilities, the average residence time of
a REE in seawater is only a few thousand years compared to the ~2 Ma residence time of
Sr, which is very soluble. REEs precipitate out of seawater by adsorbing onto particles,
particularly Fe-Mn oxyhydroxides that form by oxidation of dissolved Fe
2+
and Mn
2+
in
surface waters. In natural waters, the element Ce also occurs appreciably in the +4 state.
Ce
+4
is even more insoluble than Ce
+3
, precipitating as CeO
2
along with Fe-Mn
oxyhydroxides. For this reason, seawater has an appreciable negative Ce anomaly while
Fe-Mn nodules and crusts have a pronounced positive Ce anomaly. Ce anomalies are
also evident in groundwaters that have passed through oxidation fronts. Because of the
short residence time of the REEs in seawater, they can be used to trace movements and
sources of water masses in the oceans. The Nd isotopic system has been extensively used
for this purpose.
REE deposits are generally rare. The richest REE deposits are typically
associated with carbonatite magmatism. Carbonatite magmas are unusual magmas
composed primarily of CaCO
3
, Na
2
CO
3
, and/or MgCO
3
. They also tend to be highly
enriched in the incompatible elements, including the REEs. Because of their odd major-
element chemistry, they precipitate unusual minerals, many of which are REE oxide

ESCI 430 C-T Lee
minerals. The cumulate deposits of carbonatite magma systems are therefore highly
enriched in the REEs.

High field strength elements (Nb, Ta, Ti, Zr, Hf)
The elements Nb, Ta, Ti, Zr, and Hf are located on the left-hand side of the
Transition series. In most geologic environments, they are characterized by very high
charges (+4). The cationic sites in most silicate minerals do not accommodate tetravalent
cations and therefore these elements tend to be incompatible. Because of their high
charge to radius ratio (Ionic Potential), they are often termed the high field strength
elements. They are particularly insoluble in aqueous fluids. They are generally more
compatible in oxide minerals. For example, Nb and Ta are compatible in rutile (TiO
2
) or
ilmenite (FeTiO
3
). If Zr concentrations are high enough, zircon can form (ZrSiO
4
).

Noble Metals (Ru, Rh, Pd, Os, Ir, Pt, Au)
The Noble metals are transition elements in periods 5 and 6. Except for Au, the other
six elements are often referred to collectively as the platinum group elements. The noble
metals have 6 to 9 electrons in their valence (d-shell) and are characterized by numerous
oxidation states. In nature, these elements often occur as metals (zero oxidation state) or
covalently bonded as sulfides. At high oxidation states these metals can combine with
oxygen or complex with halogens in solution. In metal form, these elements are not very
soluble, hence the term noble metals. Most of the Earths noble metal budget is in the
core. In a cosmochemical sense, these elements are referred to as being siderophile,
that is, iron-loving. After the core, the mantle is the main repository for noble metals as it
appears that these elements are largely compatible during partial melting of the mantle
and stay in the mantle residue as sulfides or metal alloys. The noble metals are
occasionally found concentrated in layered mafic intrusions, where they are associated
with cumulate layers of sulfides and/or metal alloys. Under certain conditions, some of
the metals are highly soluble. Pt, Pd, and Au (and possibly Os) may form soluble
chloride complexes in hydrothermal solutions, allowing them to be transported and
concentrated in the form of ores.

7.4 Partial melting models
We now make use of partition coefficients to study the evolution of trace-element
compositions during melting. In Section 7.6, we will use what we learn in this section to
understand processes of Earth differentiation. Qualitatively, we already know that if an
element is incompatible, that is D<1, it will be enriched in the melt and depleted in the
coexisting solid residue. Conversely, for D>1, the melt will be depleted in that element
and the solid will be enriched. In a closed system, the concentration of an element in the
melt is related to that in the solid by
C
melt
=C
solid
/D
B
Eq. 7.13
where D
B
is the bulk partition coefficient.

ESCI 430 C-T Lee
We can also envision a scenario by which we incrementally extract batches of melt from
the system, such that the mass of the system decreases with each increment of melt
extracted. Before each increment of melt is extracted, we let the concentration of an
element in the melt be dictated by Eq. 7.13 assuming a closed system. As in the first
scenario, an incompatible element will prefer to be in the melt. Therefore, the first
increment of melt extracted will be highly enriched in the incompatible element and the
solid residue will be depleted. If we remove the first increment of melt from the system,
and then produce another increment of melt from the remaining solid residue, we will
find that although the second increment of melt is enriched relative to the solid residue,
the concentration of the incompatible element will be lower than that of the first melt
increment simply because the concentration of the element is progressively decreasing in
the solid residue with progressive melting.

Batch melt
Solid
Residue
Aggregate
Melt
Fractional
Melt
Solid
Residue
A
B
Batch melt
Solid
Residue
Batch melt
Solid
Residue
Aggregate
Melt
Fractional
Melt
Solid
Residue
A
B


Figure 7.2a. Physical difference between a) batch (equilibrium) and b) fractional
melting.
The two endmember processes of melting are called equilibrium (batch) and
fractional melting. Physically, equilibrium melting describes the case in which melting
takes place in a system closed to mass exchange and where the melt and solid residue
always remain in equilibrium (Fig. 7.2a). Fractional melting describes the case in which
melt is instantaneously removed from the system as soon as it is produced (Fig. 7.2b).
Fractional melting can be described as incremental equilibrium melting in the limit that
each melt increment goes to zero. In general, the melting process in the mantle is closer
to fractional melting than equilibrium melting. This is because melts have much lower
densities than the solid residues, and therefore, under lithostatic conditions, the melt tends
to escape soon after it is created. In detail, a critical fraction of melt must be generated

ESCI 430 C-T Lee
before the melt can separate from the solid residue. The critical melt fraction depends on
the melt fraction and wetting angle of the melt. If the wetting angle of a melt is low (low
cohesion), it is allowed to form interconnected pathways at very low melt fractions,
allowing it to readily escape. If the wetting angle is high (high cohesion), then the melt
pools at grain boundary triple points. Only when enough melt is created will the melt
pools become large enough to form interconnected pathways for melt to migrate. These
two processes are quantitatively modeled below.
0.1
1
10
100
0 0.2 0.4 0.6 0.8 1
C
melt
/C
o
C
solid
/C
o
D=0.01
0.1
0.5
F
0.01
0.1
1
10
100
0 0.2 0.4 0.6 0.8 1
F
0.01
0.1
1
10
100
0 0.2 0.4 0.6 0.8 1
1
5
10
10
5
1
0.5
0.1
0.01
Equilibrium Melting
Melts
Solids


FIGURE 7.3. Equilibrium Melting.

7.4.1 Equilibrium melting
In any closed system, the following mass balance for an element must hold
melt melt sol sol
o
sol
M C M C m + = 7.14
where represents the total mass of the element in the system (equivalent to the mass
of the element in the unmelted solid), M
o
sol
m
sol
is the mass of the solid at any given point in
time, M
melt
is the mass of the melt at any point in time, and C
sol
and C
melt
is the
concentration of an element at any point in time. The following mass balance condition is
also required
melt sol
o
M M M + = 7.15

ESCI 430 C-T Lee
where M
o
is the total mass of the system. Using the relationships in Eqs. 7.13 and 7.15,
and defining F to be the weight fraction of melt in the system (F=M
melt
/M
o
), Eq. 7.14
becomes
) ) 1 ( ( F F D C C
B melt
o
sol
+ = 7.16
where is the initial concentration of the element in the system, e.g., in the unmelted
solid. Eq. 7.16 can be rearranged to express the concentration of an element in the melt
and solid as a function of melt fraction relative to the initial concentration of the element.
o
sol
C
) 1 (
) 1 (
1
B B
B
o
sol
sol
B B
o
sol
melt
D F D
D
C
C
D F D
C
C
+
=
+
=
7.17
Eqs. 7.17a and b are plotted in Figure 7.3. Incompatible elements are enriched
and depleted in the melt and solid, respectively, as melt is extracted. Compatible
elements are depleted and enriched in the melt and solid, respectively, as melt is
extracted. For highly incompatible elements (D~0), the maximum enrichment of the
element in the melt is 1/F.

7.4.2 Fractional melting
Let us model fractional melting as an incremental equilibrium melting process in
the limit that the mass of each melt increment, dM
melt
, goes to zero. Denoting the mass of
an element in each melt increment as dm
melt
, at any point in time, the concentration of an
element in the solid is m
sol
/M
sol
and in the coexisting melt fraction is dm
melt
/dM
melt
, where
m
sol
and m
melt
are the element masses in the solid and melt, and M
sol
and M
melt
are the
masses of the solid and coexisting melt. Thus, at any instant
( )
( )
melt melt
sol sol
B
dM dm
M m
D = 7.18
or
sol
melt
sol
melt
B
M
dM
m
dm
D = 7.19
Making use of the fact that dm
sol
=-dm
melt
and dM
sol
=-dM
melt
, Eq. 7.19 can equivalently be
expressed as
sol
sol
sol
sol
B
M
dM
m
dm
D = 7.20
We can integrate Eq. 7.20 to obtain an expression of the mass of an element in the solid
as a function of the fraction of solid mass remaining in the system

sol
M
o
sol
M
sol
sol
sol
m
o
sol
m
sol
sol
B
M
dM
m
dm
D 7.21

ESCI 430 C-T Lee

0.001
0.01
0.1
1
10
100
0 0.2 0.4 0.6 0.8 1
0.001
0.01
0.1
1
10
100
0 0.2 0.4 0.6 0.8 1
0.001
0.01
0.1
1
10
100
0 0.2 0.4 0.6 0.8 1
0.001
0.01
0.1
1
10
100
0 0.2 0.4 0.6 0.8 1
0.1
1
10
100
0 0.2 0.4 0.6 0.8 1
0.1
1
10
100
0 0.2 0.4 0.6 0.8 1
F
C
melt
/C
o
C
solid
/C
o
o
melt C C /
Fractional Melting
Instantaneous
Melt Fraction
Solid
Aggregate Melt
D=0.01
0.1
0.5
1
5
10
0.01
0.1
0.5
5
10
1
10
5
1
0.5
0.1
0.01


FIGURE 7.4. Fractional Melting
ESCI 430 C-T Lee
which yields
o
sol
sol
o
sol
sol
B
M
M
m
m
D ln ln = 7.22
or
B
D
o
sol
sol
o
sol
sol
M
M
m
m
/ 1
|
|
.
|

\
|
= 7.23
Multiplying both sides of Eq. 7.23 by
sol
o
sol
M
M
and defining the total fraction of melt
extracted as
o
sol
melt
M
M
F , Eq. 7.23 becomes

( )
1 ) / 1 (
1

=
B
D
o
sol
sol
F
C
C
7.24
It follows that the concentration in each instantaneous melt fraction is simply
( )
1 ) / 1 (
1
1

=
B
D
B
o
sol
melt
F
D
C
C
7.25
In nature, we are able to sample melts after they have reached the surface of the Earth. It
is unlikely that we can sample each instantaneous melt fraction at its source. Instead,
these instantaneous melt fractions are likely to aggregate in a magma chamber near the
surface of the Earth, and it is this aggregate reservoir that we are more likely to sample.
The average concentration of an element in the aggregate melt is determined by
integrating the following equation

=
melt
M
melt melt
melt
melt
dM C
M
C
0
1
7.26
or expressed in terms of F,

=
F
B
D
o
sol
melt
melt
dF F
D
C
F
C
0
1 ) / 1 (
) 1 (
1
7.27
Integration of Eq. 7.27 then yields
F
F
C
C
B
D
o
sol
melt
/ 1
) 1 ( 1
= 7.28
Figure 7.4 shows the effect of fractional melting on the solid residue and the melt.
It can be seen that the depletion of an incompatible element in the solid residue is more
rapid during fractional melting than during equilibrium melting. This is because with
each increment of melt extracted, the concentration of an incompatible element in the
system decreases during fractional melting, whereas during equilibrium melting, the
system concentration remains constant. It follows that although each instantaneous melt
fraction is enriched in an incompatible element relative to the solid residue, the
concentration of an incompatible element in each instantaneous melt fraction must

ESCI 430 C-T Lee
decrease rapidly. The concentration of the aggregate melt, however, is nearly identical to
that predicted by equilibrium melting. Thus, if fractional melts pool together in a magma
chamber near the surface of the Earth, it is not possible to determine from magma
compositions alone whether the melting process was fractional or equilibrium. In
contrast, the composition of the solid residue contrasts between fractional and
equilibrium melting.


7.4.3 Fractionation of trace-element relative abundances during partial melting
We can see from Eq. 7.17, that the enrichment of a trace-element relative to its
source depends on its bulk partition coefficient. If we plot Eq. 7.17 as a function of D,
we obtain Figure 7.5. In practice, we can generate plots similar to Figure 7.5 in which
the D values are replaced by elements corresponding to given D values. If we plot the
relative abundances of trace-elements in some systematic order, we will find that the
curvature of the trace-element abundance pattern depends on the relative differences in D
values of the plotted elements.
For example, the partition coefficients for rare-earth elements in various minerals
in equilibrium with basaltic melt are plotted in order of increasing atomic number in
Figure 7.6 It can be seen that the heavy rare-earth elements (HREEs) are compatible in
garnet whereas the light rare-earth elements (LREEs) are incompatible in garnet. A
garnet source will therefore generate a melt that is highly enriched in LREEs but depleted
in HREEs. A clinopyroxene- or orthopyroxene-rich source will also generate LREE
enrichment relative to the HREEs, but the relative enrichment will be suppressed
D
0.1
1
10
0.001 0.01 0.1 1 10
C
L
/C
So
incompatible compatible


FIGURE 7.5. Enrichment/depletion of elements relative to starting composition during partial
melting as a function of bulk partition coefficient. A melt fraction, F, of 10% is assumed.

ESCI 430 C-T Lee
compared to a melt from a garnet-rich source. Thus, trace-elements can be used to obtain
information about the source region if one has some understanding of the partitioning
behaviors of trace-elements in different minerals.

FIGURE 7.6. Rare-earth element partition coefficients for minerals in equilibrium with
basaltic melts.
To quantify this further, we can express the equilibrium and partial melting
equations in terms of the ratio between two different elements. For example, the
equilibrium melting equation of Eq. 7.17 can be expressed for two elements, A and B, as
follows

ESCI 430 C-T Lee
) 1 (
) 1 (
A A
B B
Bo
sol
Ao
sol
B
melt
A
melt
D F D
D F D
C
C
C
C
+
+
= 7.29
We can do the same for the aggregate of fractional melts
B
D
A
D
Bo
sol
Ao
sol
B
melt
A
melt
F
F
C
C
C
C
/ 1
/ 1
) 1 ( 1
) 1 ( 1


= 7.30
Equations 7.29 and 7.30 in fact give nearly identical results. The important features of
the above two equations are that elemental ratios of highly incompatible elements (e.g.,
D
A
<<1 and D
B
<<1) in a melt remain largely constant over a wide range in degree of
melting, F. When D
A
~D
B
~0, it can be seen from Eq. 7.29, that the second term on the
right is approximately unity for most F. Only when the degree of fractional melting is
very small can partial melting generate melts with incompatible trace-element ratios
significantly different from their source regions.

7.4.4 Non-modal melting
In modeling the distribution of trace-elements during melting, we assumed that
the modal abundance of minerals in the source region controlled the bulk partition
coefficient. In so doing, we assumed that the relative proportions of the phases within the
solid residue does not change during melting. Such a process is called modal melting.
We know from phase petrology, however, that during partial melting, the solid
does not necessarily melt in exactly the same proportions as the phases present in the
solid, that is, the solid phase proportions are continually changing. This melting process
is called non-modal melting. Eutectic melting is a case in point.
Let us investigate non-modal melting for a closed system, e.g. equilibrium
melting. If we define X
j
as the weight fraction of a phase relative to the entire system,
where X
melt
= F, the mass balance relation of a particular element i is simply
+ =
j
j
i
j melt
i
o
X D F C C ) ( 7.31
where is the original concentration of the unmelted source. We can add and subtract
the quantity, , to the right hand side of Eq. 7.31, yielding
i
o
C

j
o
j
i
j melt
X D C ) (
+ + =
j j
o
j
i
j
j
o
j
i
j j
i
j melt
i
o
X D X D X D F C C ) ( 7.32
where represents the original phase proportions in the unmelted solid. If we make
the following definitions
o
j
X

j
o
j
i
j
o
B
X D D 7.33

j
j
i
j
B
X D D 7.34
F
X X D
P
j
o
j
i
j
) (
7.35

ESCI 430 C-T Lee
where represents the bulk partition coefficient for the unmelted solid
and represents the bulk partition coefficient of the solid for at any given degree of
melting, F. Substituting Eq. 7.33 into 7.35 yields the general bulk partition coefficient as
a function of P
o
B
D
B
D
F
FP D
D
o
B
B

=
1
7.36
Substituting Eq. 7.35 into 7.32 yields
) ( FP D F C C
o
B melt
i
o
+ = 7.37
from which we can obtain the non-modal equilibrium melting relation for the melt
) 1 (
1
P F D C
C
o
B
i
o
i
melt
+
= 7.38
We can also derive the non-modal fractional melting relation by plugging in Eq.
7.36 into 7.20 and integrating. The composition of the solid during non-modal fractional
melting becomes
F
D
FP
C
C
P
o
B
i
o
i
sol

|
|
.
|

\
|

=
1
1
/ 1
7.39
The average concentration of the aggregate of fractional melting is given by
(
(
(

|
|
.
|

\
|
=
P
o
B
i
o
i
melt
D
FP
F
C
C
/ 1
1 1
1
7.40

7.4.5 Trace-element spidergrams
It is often useful to view a lot of trace-elements on the same plot in order to
maximize the information content. To do so, geochemists make use of normalized trace-
element diagrams, called spidergrams. In such diagrams, the measured trace-element
concentration of a sample is normalized by some reference composition. The
normalization serves two purposes. First, it turns out that the abundances of trace-
element masses is such that even masses are much higher than odd masses. This is a
primordial feature of the solar system (or galaxy for that matter), which results from the
nucleosynthetic processes that create the elements. Thus, if one simply plots
concentrations of various trace-elements on the same diagram, a very jagged diagram
would appear. Normalization to any other rock reference eliminates this jaggedness.

ESCI 430 C-T Lee


Table 7.1. Composition of chondrites (CI) and Primitive Mantle
CI chondrite
1
Primitive
Mantle
2
CI chondrite
1
Primitive
Mantle
2
Li ppm 1.5 1.6 Pd ppb 560 3.9
Be ppm 0.025 0.068 Ag ppb 199 8
B ppm 0.87 0.3 Cd ppb 686 40
C % 3.45 0.012 In ppb 80 11
N ppm 3180 2 Sn ppb 1720 130
F ppm 60.7 25 Sb ppb 142 5.5
Na ppm 5000 2670 Te ppb 2320 12
Mg % 9.89 22.8 I ppb 433 10
Al % 0.868 2.35 Cs ppb 187 21
Si % 10.64 21 Ba ppb 2340 6600
P ppm 1220 90 La ppb 234.7 648
S ppm 62500 250 Ce ppb 603.2 1675
Cl ppm 704 17 Pr ppb 89.1 254
K ppm 558 240 Nd ppb 452.4 1250
Ca % 0.928 2.53 Sm ppb 147.1 406
Sc ppm 5.82 16.2 Eu ppb 56 154
Ti ppm 436 1205 Gd ppb 196.6 544
V ppm 56.5 82 Tb ppb 36.3 99
Cr ppm 2660 2625 Dy ppb 242.7 674
Mn ppm 1990 1045 Ho ppb 55.6 149
Fe % 19.04 6.26 Er ppb 158.9 438
Co ppm 502 105 Tm ppb 24.2 68
Ni ppm 11000 1960 Yb ppb 162.5 441
Cu ppm 126 30 Lu ppb 24.3 67.5
Zn ppm 312 55 Hf ppb 104 283
Ga ppm 10.0 4 Ta ppb 14.2 37
Ge ppm 32.7 1.1 W ppb 92.6 29
As ppm 1.86 0.05 Re ppb 36.5 0.28
Se ppm 18.6 0.075 Os ppb 486 3.4
Br ppm 3.57 0.05 Ir ppb 481 3.2
Rb ppm 2.3 0.6 Pt ppb 990 7.1
Sr ppm 7.8 19.9 Au ppb 140 1
Y ppm 1.56 4.3 Hg ppb 258 10
Zr ppm 3.94 10.5 Tl ppb 142 3.5
Nb ppb 246 658 Pb ppb 2470 150
Mo ppb 928 50 Bi ppb 114 2.5
Ru ppb 712 5 Th ppb 29.4 79.5
Rh ppb 134 0.9 U ppb 8.1 20.3
1
Anders and Grevesse [1989]
2
McDonough and Sun [1995]

ESCI 430 C-T Lee

The second reason for normalization is that one can learn something about
geologic processes that fractionate (i.e. change the relative proportions of trace-elements)
trace-element relative abundances from their original relative abundances. For example,
if we are interested in the process of partial melting, we would normalize the composition
of the melt (or residue) to our original starting composition. Similarly, if we are
interested in how sedimentary processes fractionate trace-element abundances, we might
normalize to average continental crust. For high temperature geochemical applications, it
is common to normalize to a hypothetical primitive mantle composition or a chondritic
meteoritic composition (Table 1). For lithophile elements, both references are believed to
approximate the primordial relative abundances of elements in the silicate part of the
Earth. In the case of primitive mantle, the absolute abundances are believed to represent
the primordial silicate Earth, that is, an estimate of the starting composition.
0.01
0.1
1
10
100
1000
Cs Rb Ba Th U Nb La Ce Pr Sr Nd Zr Hf Sm Gd Tb Dy Ho Y Er Yb Lu
MORB
Plume
Cont.
Crust
Melts and crust
normalized to
primitive mantle
Primitive mantle- normalized
0.01
0.1
1
10
100
1000
Cs Rb Ba Th U Nb La Ce Pr Sr Nd Zr Hf Sm Gd Tb Dy Ho Y Er Yb Lu
MORB
Plume
Cont.
Crust
Melts and crust
normalized to
primitive mantle
Primitive mantle- normalized


FIGURE 7.7. Trace-element abundances of continental crust, mid-ocean ridge basalt
(MORB) and a plume basalt normalized to primitive mantle.
To enhance readability of a trace-element spidergram, it is necessary to agree
upon the order of plotting elements. To the non-geochemist, the choice of element
sequence seems arbitrary. However, there is some method behind all this madness! For
example, in many cases, particularly in high temperature geochemistry, trace-element
spidergrams are plotted in order of increasing compatibility. For example, Figure 7.5
shows the composition of a partial melt normalized to its starting composition and plotted
in order of increasing partition coefficient, D. As different elements have different
partition coefficients, we can replace the Ds in Figure 7.5 with elements having the
appropriate D value.
Figure 7.7 is an example of a trace-element diagram, on which is plotted the
average compositions of continental crust, mid-ocean ridge basalt and a plume magma.
The elements are plotted approximately in the order of increasing compatibility.
Compatibilities are either determined from experiments or inferred from the compositions

ESCI 430 C-T Lee
of melts. At the time when experimental data was limited, the order of compatibilities
was estimated by assuming, as a working hypothesis, that the continental crust was
extracted from the mantle by partial melting
1
. It follows from the discussions in the
previous sections that the more incompatible an element the more enriched it would be in
the continental crust. By convention, it was agreed that the relative order of
compatibilities is roughly defined such that the trace-element spidergram of continental
crust normalized to primitive mantle (or chondrite) yields a smooth negative slope.
Broadly speaking, this inferred compatibility sequence is consistent with experimental
data.
An important feature of Figure 7.7 is that the primitive-mantle-normalized pattern
of continental crust is smooth except for a few elements, such as Nb, Ta, Zr, and Hf. The
positions of these elements in Figure 7.7 obviously do not correspond to their positions in
the order of enrichment seen in continental crust! Instead, Nb is plotted next to U while
Zr and Hf are plotted next to Sm. In this particular case, the reason for plotting Nb next
to U and Zr and Hf next to Sm is based on the observation that Nb has a very similar
compatibility to U (and Zr and Hf to Sm) during formation of mid-ocean ridge basalts.
This plotting sequence may not be as arbitrary as it sounds. By plotting Nb next to U, we
see that continental crust has a negative Nb anomaly. This implies that the petrogenetic
processes controlling Nb partitioning during continental crust formation are not the same
as in mid-ocean ridge basalt generation
2
.

7.4.6 Trace-element systematics of a differentiated Earth
The important features to note in Figure 7.7 are that the continental crust appears
to be highly enriched in the incompatible elements, whereas mid-ocean ridge basalts are
depleted in the highly incompatible elements (note the difference in slope). It is
generally believed that a large part of the continental crust was formed early on in Earths
history. Extracting continental crust out of the primitive mantle by partial melting
resulted in the enrichment of incompatible elements in the continental crust. Because a
large part of the continental crust so formed was not recycled and rehomogenized back
into the mantle, the residual mantle became depleted in the highly incompatible elements.
This residual mantle is often referred to in the literature as Depleted Mantle (DM).
Progressive extraction of continental crust from the mantle will result in progressive
depletion of incompatible elements in the residual mantle - the more incompatible an
element the more it is depleted from the mantle. Magmas that subsequently derive from
this residual mantle, such as mid-ocean ridge basalts, must therefore inherit the depleted
signature of the residual mantle. For this reason, the trace-element abundance pattern of
mid-ocean ridge basalts exhibits a relative depletion in the highly incompatible elements.


1
In detail, the continental crust was probably not extracted from the mantle by a single-stage partial
melting process as its silica content (SiO
2
~59 wt. %) is too high to be in equilibrium with the mantle. The
true petrogenetic origin of continental crust is much more complicated than the formation of mid-ocean
ridge basalts, and this is a topic of continued debate.
2
The formation of continental crust may involve the melting of subducted oceanic crust or sediments. The
presence of amphibole or rutile in the melting residue of continental crust formation may be one means of
creating a negative Nb anomaly as Nb can be compatible in these minerals. By contrast, mid-ocean ridge
basalt melting does not involve either of these minerals in the residue and therefore Nb behaves as a very
incompatible element like U during the formation of mid-ocean ridge basalt.

ESCI 430 C-T Lee
7.5 Fractional crystallization
7.5.1 Major element effects of fractional crystallization
Here, we investigate the effects of crystallization on the trace-element
composition of a magma. There are two endmember styles of crystallization: equilibrium
and fractional crystallization. In the former, there is no transfer of mass into or out of the
system, such that for any amount of crystallization (e.g, 1-F, where F is the melt
fraction), all crystals remain in equilibrium with the melt. In reality, it is highly unlikely
that crystals remain in complete equilibrium for two reasons. First, crystals tend to be
denser than melts, and therefore, once a crystal is formed, it has a tendency to sink and
fall out of the magma chamber, hence exiting the magmatic system (Fig. 7.8a). The
second reason is that the diffusion timescale of chemical species in crystals is slow
compared to that of mechanical mixing and diffusion in the melt. Therefore, even though
the rims of phenocrysts may be in equilibrium with the host magma, the cores of the
phenocrysts may not be (Fig. 7.8b). Both the physical removal of crystals and their

A B
C
A B
C


FIGURE 7.8. a) Fractional crystallization whereby phenocrysts (gray rectangles) settle out of
the magmatic system (white region); b) Fractional crystallization whereby rapid growth of
phenocrysts results in the diffusive isolation of phenocryst cores from the magma; inset shows
the zoned nature of such phenocrysts, which is a result of disequilibrium; c) Assimilated
fractional crystallization in which melting/stoping of wallrock is concomitant with crystallization.

ESCI 430 C-T Lee
growth result in the incremental removal of material from the magmatic system and
therefore the composition of the magma must evolve. In the limit that each mass
increment of crystallization goes to zero, that is, continuous removal of mass, the
crystallization process is referred to as fractional crystallization.
We can use major element variation diagrams to estimate the degree and
composition of crystallizing phases. A variation diagram is a chemical plot on which the
compositions of a series of cogenetic magmas are plotted. For example, a magma
originally in equilibrium with olivine, will first crystallize olivine. Upon hitting the
olivine-plagioclase cotectic, olivine and plagioclase will precipitate. This crystallization
series will cause the melt composition to evolve. Figure 7.9 shows the schematic
evolution of magma going through this crystallization sequence. Because of the high
MgO and low SiO
2
content of olivine, crystallization of olivine drives the magma
towards MgO-poor and SiO
2
-rich compositions. When plagioclase co-crystallizes with
olivine, the magma composition is driven in the direction opposite the composition of the
olivine-plagioclase crystallization mixture. Simple mass balance allows one to estimate
the degree of crystallization.
The schematized example shown in Figure 7.9, however, assumes that the major-
element composition of the minerals remains constant. This is certainly not the case
because the major-element composition of the phenocryst will change as the composition
of the magma changes, simply because the composition of the magma and phenocrysts
are thermodynamically linked.



SiO
2
wt. %
MgO
Wt. %
L
o
Fo
Olivine precipitation
Plagioclase precipitation
SiO
2
wt. %
Al
2
O
3
Wt. %
L
o
Fo
Olivine
precipitation
Plagioclase
precipitation
An
An
SiO
2
wt. %
MgO
Wt. %
L
o
Fo
Olivine precipitation
Plagioclase precipitation
SiO
2
wt. %
Al
2
O
3
Wt. %
L
o
Fo
Olivine
precipitation
Plagioclase
precipitation
An
An


FIGURE 7.9. Schematic variation diagram showing how the composition of a melt with an
primary composition of L
o
changes with progressive crystallization of olivine and plagioclase
as shown in Figure 1.

ESCI 430 C-T Lee
7.5.2 Trace-element effects during fractional crystallization
We now model the effect of fractional crystallization on the trace-element
composition of a magma. We keep the same symbols that we used in deriving the partial
melting relationships. We model fractional crystallization as infinitesimal increments of
closed systems. Thus, at any given time, the following relationship must hold
melt
melt i
B
i
melt
i
melt
M
dM
D
m
dm
= 7.41
We can integrate Eq. 7.41,

melt
M
o
melt
M
melt
i
B
i
melt
m
oi
melt
m
i
melt
M d D m d ) (ln ) (ln 7.42
which yields
|
|
.
|

\
|
=
|
|
.
|

\
|
o
melt
melt i
B
oi
melt
i
melt
M
M
D
m
m
ln ln 7.43
Rearranging Eq. 7.43, we have
i
B
D
o
melt
melt
oi
melt
i
melt
M
M
m
m
|
|
.
|

\
|
= 7.44
Changing Eq. 7.44 to concentrations, we obtain the fractional crystallization relationship
for the melt
1
=
i
B
D
o
melt
i
melt
F
C
C
7.45
Figure 7.10 shows how the concentration of a trace-element evolves in a melt for
different partition coefficients.
It can be seen that if an element is incompatible (D<1) then the melt is enriched in
that element. If an element has a distribution coefficient of 1, then the concentration of
that element in the melt does not change. If an element is compatible in the crystallizing
phase (D>1), then the element is rapidly depleted from the melt. In Eq. 7.45, we can see
that if an element is perfectly incompatible (D=0), then the maximum relative enrichment
of such an element in the melt is given by 1/F.
We can also calculate how a trace-element ratio evolves in a fractionally
crystallizing melt by combining the fractional crystallization equation (Eq. 7.45) for each
trace-element (denoted by element A and B):
B
D
A
D
oB
melt
oA
melt
B
melt
A
melt
F
C
C
C
C

|
|
.
|

\
|
=
|
|
.
|

\
|
7.46
Eq. 7.46 expresses how the trace-element ratio of a melt evolves from its starting
composition as a function of the fraction of melt remaining F and the difference in
distribution coefficients, D
A
-D
B
. We can see that if two elements have the same
distribution coefficient, e.g., D
A
-D
B
=0, then the ratio of element A to B in the melt does
not change. Similarly, if D
A
and D
B
are both very small quantities, but different, that is,

ESCI 430 C-T Lee
both A and B are highly incompatible, then the difference, D
A
-D
B
, is a very small
number, i.e. nearly equal to zero. Thus, the ratios of highly incompatible elements in a
melt also do not change during fractional crystallization. This means that certain trace-
element ratios in a crystallizing magma retain the original signature of the primitive
magma, allowing us to obtain information about the magmas origin.





10
-3
10
-2
10
-1
1
10
10
2
0 0.2 0.4 0.6 0.8 1
Fractional Crystallization
D=0
D=
0.1
D=
0.5
D=1
D=2
D=5 D=10
C
L
/C
Lo
F
10
-3
10
-2
10
-1
1
10
10
2
10
10
2
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
Fractional Crystallization
D=0
D=
0.1
D=
0.5
D=1
D=2
D=5 D=10
C
L
/C
Lo
F

FIGURE 7.10. The evolution of trace-element concentration in a melt during fractional
crystallization. F represents the weight fraction of melt remaining.



References
"Revised effective ionic radii and systematic studies of interatomic distances in halides
and chalcogenides." R. D. Shannon 1976 Acta Cryst. A32, 751-767.


EXCERCISES
PART A
1. Unlike most of the other rare-earth elements, Europium occurs in both Eu
+2
and
Eu
+3
in nature. Using Table 2, plot the partition coefficients of the rare-earth
elements for plagioclase. Why is Eu much more compatible in plagioclase than
the other rare-earth elements? If a magma undergoes crystallization of
plagioclase and if the accumulated plagioclase is removed from the magma
system, what would be the most distinctive feature of the residual magmas rare-
earth element pattern?

ESCI 430 C-T Lee
2. Using Goldschmidts Rules as a rough guide, which element is more
incompatible: Sm or Nd, Lu or Hf, and Rb or Sr? If this relative behavior
partitioning behavior reflects most geochemical processes, where on the following
diagrams would the compositions of various rocks on Earth plot. These diagrams
show the Sm/Nd, Lu/Hf and Rb/Sr ratios normalized to primitive mantle. First
plot where primitive mantle compositions lie. Then plot schematically where
partial melts of primitive mantle and the residual solid mantle would lie. If
continental crust represents the first melt extracted from the primitive mantle,
where would it lie? If mid-ocean ridge basalts represent melts extracted from the
primitive mantle after continental crust was extracted, where would they lie? If
seawater represents a mixture between dissolved continental detritus and
hydrothermal circulation at mid-ocean ridges, can you make an educated guess
where on these diagrams seawater might lie?

S
m
/
N
d
Lu/Hf
1
1
S
m
/
N
d
Rb/Sr
1
1
X- and y-axes represent the ratios normalized to the Earths primitive mantle
S
m
/
N
d
Lu/Hf
1
1
S
m
/
N
d
Lu/Hf
1
1
S
m
/
N
d
Rb/Sr
1
1
X- and y-axes represent the ratios normalized to the Earths primitive mantle


3. Two-component mixing. Assume that two reservoirs A and B (magmas, waters,
etc.) have different concentrations of Sr and Rb denoted by Sr
A
, Sr
B
, Rb
A
, and
Rb
B
.
Write out a mass balance expression for the concentration of Sr and Rb
when the two reservoirs are mixed with the following proportions, X
A
and
X
B
, where X
A
is the weight fraction of A in the mixture and X
B
= 1 X
A
.
Show that the Rb/Sr of the mixture is given by the following equation
B B A A
B B A A
MIX
MIX
Rb Rb Rb X
Sr Sr Sr X
Rb
Sr
+
+
=
) (
) (

Show that a plot of Sr/Rb versus 1/Rb is a straight line that is independent
of the weight fractions of A and B. The answer is
(

+
(

=
B A
B A
B B
MIX B A
B A
MIX
MIX
Rb Rb
Sr Sr
Rb Sr
Rb Rb Rb
Sr Sr
Rb
Sr 1

What is the slope of this line? Can you think of how this equation may be
used to obtain information about the composition of the mixing
endmembers, A and B?

ESCI 430 C-T Lee


4. A geologist has discovered an exposed section of the lower crust of an ancient
island arc. The lower crust appears to be eclogite, a rock composed of garnet and
clinopyroxene. The question is whether the garnet and clinopyroxene in this
eclogite represent a primary liquidus phase, that is, the parent magma crystallized
garnet and clinopyroxene directly as it cooled. Alternatively, the eclogite can
represent a subsolidus transition from a plagioclase-clinopyroxene rock (gabbro)
that underwent an increase in pressure. Solving this dilemma is important
because the crystallization of garnet occurs only at high pressures (>1.5 GPa or
~45 km), whereas plagioclase is a lower pressure mineral. Thus, if garnet was on
the liquidus, a deep magma chamber is implied. One tool we can use is to model
the trace-element composition of the cumulates according to the following steps.
Assuming a closed system, write down a mass balance equation that
equates the total concentration of an element in the system to the
concentration in the solid cumulate and residual liquid. Assume that all of
the crystallized products are in equilibrium with the liquid. Verify that the
concentration in the solid cumulate C
S
is the following:
B S S
o
S
D X X
C
C
/ ) 1 ( +
=
where C
o
is the concentration in the original melt (e.g., prior to any
crystallization), D
B
is the bulk partition coefficient, and X
S
is the weight
fraction of the solid in the system.
On a spreadsheet (e.g., EXCEL), fill up the following the table. Assuming
the initial composition of the magma is that of a typical mid-ocean ridge
type basalt (see Table), calculate the rare-earth element composition of an
eclogitic and gabbroic cumulate. Assume that the cumulate makes up
30% by mass of the entire system. Assume also that the ratios of garnet to
clinopyroxene and plagioclase to clinopyroxene are equal.
Normalize your results to primitive mantle and plot the normalized
eclogite, gabbro and original basaltic melt compositions on a rare-earth
element spidergram. Plot in order of increasing atomic mass. Comment.

Primitive
Mantle
basalt D gt D cpx D plag
D
eclogite
D gabbro
Solid
Proportion
predicted predicted
ppm ppm
gt:cpx =
1:1
plag:cpx
= 1:1
XS eclogite gabbro eclogite gabbro basalt
La 0.648 2.5 0.01 0.056 0.1
Ce 1.675 7.5 0.03 0.092 0.12
Nd 1.25 7.3 0.07 0.23 0.081
Sm 0.406 2.63 0.29 0.445 0.067
Eu 0.154 1.02 0.49 0.474 0.34
Gd 0.544 3.68 0.97 0.556 0.063
Dy 0.674 4.55 3.17 0.582 0.055
Er 0.438 2.97 6.56 0.583 0.063
Yb 0.441 3.05 11.5 0.542 0.067
Lu 0.0675 0.455 11.9 0.506 0.06
Primitive Mantle Normalized