8 Free energy and phase equilibria

these stabilities must be given relative to one another and not

as absolute stabilities. For example, although the o. phase has
a lower free energy than the J phase at low temperature and is
therefore the more stable ofthe two under these conditions, a
third phase )' might exist that could have a lower free energy
than either (f. or J. We would still be able to say that (f. was
more stable than jJ under these conditions, but both phases
would be unstable relative to the )' phase.
Let us detennine the variance along the line G" = Gil
(t:,G = O) in Figure 8.2. The system has three variables G, T,
and P, which are related through Eq. (8.2) for phase (l.. A
similar but independent equation can be written for phase J.
Thus, the three variables are related through two independent
equations, and the variance must be 3 - 2 = l. The assem-
blage of o. -1- J is therefore said to be univariant. Any one of
the three variables can be changed, but the values ofthe other
two are then fixed by the equations. If only one of the phases
is present, the three variables are related by only (me equa-
tion, and the variance of the system would be 2. This divar-
iant surface would be the surface for the (1. or the jJ phase in
the G-T-P plot.
The three-dimensional plot of G-T-P space is commonly
reduced to two dimensions by taking sections through it at
constant P or constant T, or projecting the data parallel to the
G axis onto the P-Tplane (Fig. 8.2). In a section at constant P
or T, the two surfaces for phases (1. and/3 appear as lines that
intersect at a point where the univariant line G" = (;1
1
pierces
the plane ofthe section. The variance ofthe assemblage (1. -1- fJ
therefore becomes zero (invariant) under these conditions.
This follows from the fact that the number of variables is
reduced to two by holding one of them constant while two
equations still relate the variables. With only two variables
and two equations, there can be no variance; that is, only one
set of conditions will pennit phases (1. and fJ to coexist. Should
a rock be found containing these phases, we would know
preciseIy the conditions undcr which it formed.
Despite the simplicity of the sections at constant Tand P,
the most useful diagram is the P-Tprojection. In this plot, the
variance of the system is not decreased by the projection
(nothing is heId eonstant). Thus, the univariant line
G" = Gil remains a univariant lineo Even the divariant surfa-
ces for phases (1. and J can be contoured with lines of constant
G, but in most diagrams only the univariant line is shown. [t is
important when examining P-T diagrams illustrating phase
equilibria to realize that they are projections from G-T--P
space.
To plot univariant lines in P- Tprojections, we must know
how to determine their slopes. The difference in free energy
(t:,G) between the two surfaces for the (f. and /) phases in G-T-
P space is a function only of T and P in a closed system.
According to the eycIie rule for total differentials (Eq. (7.39)),
we can therefore write
(
Dt:,G) (ap) (DT) =-1
DT l' Dt:,G r DP!'.G
8.2 FREE ENERGY SURFACE IN G-T-P SPACE 165
from which it folIows that
(8.2) G-" -G-" S-"(T-7')-t-V-"(P-P)
r,1' - 7".1'" - 1 () ()
~ ~ ) P = - S
Fig.8.1 Free-energy surlace 01 phase (; plotted in G-T-P space. The
surface is ruled with lines of constant Iree energy.
Divariant surface
G"T.P= G'\ P - S"(T- To)+ \J"(p-Po)
, o' o
G
t
This relation is the equation of a surface in G-T-P space that
has an intercept on the G axis of G'h, ,1'" (Fig. 8. 1). If Pis held
constant, the slope on the surface is -S"; ifTis heId constant,
the slope is V". In reality, S and Vwould be functions of Tand
P, so that slopes would change with Tand P, and the surface
would be curved. For simplicity of illustration, however, S
and Vhave been taken as constants.
If thc system contains a second phase jJ, a polymorph of
phase 0., for example, another surface in G-T-P space can be
constructed (Fig. 8.2). Phase jJ will most likely have different
valucs ofS and Vfrom those of phase (l.. Thus, the free-energy
surfaces of the two phases will not be parallel and will
intersect along a line where G" = GfJ. But if G" = Gil and
phasc o. changes into phase jJ, the free energy would not
change; that is, t:,G,. = O. According to Eq. (7.25), this con-
dition describes equilibrium. The line of intersection of the
two surfaces therefore gives the values of T and P under
which phases o. and jJ are in equilibrium with each other;
that is, they can coexist.
We know fi'om Eq. (7.25) that a spontaneous reaction is
one in which dG is negative; that is, the reaction proceeds in a
direction that minimizes G. On the low-temperature/low-
pressure side of the line of intersection of the two surfaces
in the G-T-P plot ofFigure 8.2, the free energy ofthe system
can be lowered if phase jJ changes into phase (1. 1n this region,
then, phase (1. is stable relative to phase J. At pressures ancl
temperatures aboye the line of intersection of the two surfa-
ces, the jJ phase has the lowerfi'ee energy and is thcrefore
stable relative to the (1. phase. It is impOliant to cmphasize that
know that (aG" /aT)!' is -S" and (aG" /ap)T is V",
respective!y. If we assume that S and V remain constant
over the range of conditions considered, integration of Eq.
(8.1) gives
(8.1 )
J
'(..;'1' J.r (D.. G") . j'l' (a..G")
dG= -,- dT+ -.,- dP
. G;: . 7\, aT l' . 1'" c)p r
In Section 7.8 we saw that in a closed system, that is, one in
which no matter is transferred in or out, the change in the
Gibbs free energy, dG, is a function only of T and P; it can
therefore be cxpressed as a total differential in terms of Tand
P (Eq. (7.30)). If we consider a system consisting only of a
phase a, the free energy ofthe system at any Tand P could be
obtained by integrating Eq. (7.30); that is,
8.2 FREE ENERGY SURFACE IN G-T-P SPACE
must have the same value throughout a system; that is, no
gradients in the potentials of mechanical (P), thermal (n,
chemical (p), and electrical (E) energy can exist. Although
perhaps self-evident, this principIe is so important that it is
commonly refelTed to as the zerotJ law ofthermodynamics.
In determining the equilibrium between minerals, we con-
stantly have to evaluate how variations in the intensive and
extensive variables affect equilibrium. One particularly
important question is how many of these variables can be
changed simultaneously without disturbing equilibrium. If a
system can be described in terms of n variables, and r equa-
tions relate these variables, then (n - r) variables can be
changed independently while maintaining equilibrium. The
number (n - r) is called the variance or number of degrees ol
f'eedom of the system.
We can illustrate the significance ofvariance by examin-
ing a simple system consisting of l mol of ideal gas. Because
only 1 mol is considered, the number of moles is not a
variable. The system, then, can be described in terms of the
two intensive variables P and Tand the one extensive varia-
ble, V. But these three variables are related by the ideal gas
law (PV = RT), so that the variance of the system must be 2
(i.e. 3 - 1); that is, any two variables can be changed inde-
pendently, but the value of the third is detennined by the
equation relating the variables. We will have occasion to refer
ofien to the variance of systems. The numbers of variables
and relations between variables may be considerably greater
than those considered in this simple example, but the same
principIe will be involved (see Section 10.5).
where To = 298 K, Po = 105 Pa, and G(; = free energy per
mole at 298 K and lOS Pa. From Eqs. (7.3 1) and (7.32), we
Electrical
[dQ
Chcmical
j/ dn
In Chapter 7, our experiences of the behavior of energy in
nature were formalized into laws dealing with the conserva-
tion of energy, the natural direction of processes, and the
absolute zero of temperature. Once given mathematical
expression, these laws were used to relate important thermo-
dynamic functions to easily measured properties. The ther-
modynamic functions, in tum, were used to determine the
direction of reactions and the conditions necessmy for equi-
librium. In this chapter, we deal with the practical aspects of
applying these thermodynamic principIes to dctermining the
equilibrium relations between minerals.
We have seen that energy can take various forms - mechan-
ical, thermal, chemical, electrical. In each case, the energy can
be expressed as the product oftwo terms, one intensive and the
other extensive. Intensive ones can be measured at any point in
a system and are independent ofthe extent ofthe system; they
can be thought of as environmental lctors. They include
pressure, temperature, and chemical and electrical potentials.
The extensive ones, on the other hand, must be integrated over
the entire system. These incIude volume, entropy, number of
moles, and e!ectrical charge. Changes in energy can then be
expressed as follows:
Mechanical Thermal
PdV TdS
8.1
where I/ is the chemical potential of substance i (Chapter 9),
n the number of moles of i, [ the electrical potential, the
electrical charge, and all other terms are as defined in
Chapter 7.
Writing energies as the products of intensive and exten-
sive variables provides insight into the meaning of the more
abstract thermodynamic terms. Just as electrical charge is the
capacity of a system for electrical energy, so entropy is the
capacity of a system for thennal energy. Similarly, while
electrical potential determines the leve! to which the electrical
capacity is exploited, temperature determines how much the
thennal capacity will be exploited. SimilarIy, volume is the
capacity of a system for mechanical energy, and pressure is
the mechanical potential that exploits that volume. The
capacity for chemical energy is the number of moles of
substance present in the system, and the chemical potential
is the environmental factor that exploits that capacity.
Another important rcason for distinguishing these two
types ofvariable is that at equilibrium, any intensive variable
164
(8.8)
(8.7)
(8.6)
(8.5)
(8.4)
= O} + RT InI
= + RT In P bars
- l' - o J'E'-
GT=G
T
+ VdP
.0
The free energy of the gas at temperaturc Tand 10
5
Pa (OJ)
can be detennined directly from the data in Table 7.1.
lntegration of the VdP term requires that we know the func-
tional relation between Vand P.
Ifthe gas behaves ideally (V= RTIP), Eq. (8.4) reduces to
For an ideal gas y = and I = P. Tables of fugacities and
fugacity coefficients for high pressures and temperatures are
given for H
20 by Bumham et al. (1969) and for CO2 by
Mel'nik (1972).
and 10
5
Pa. According to Eq. (7.36), (aSI8P)r= -o: V. But we
are dealing with the difference in volumes (11V,,) between
that have very small and similar values of 0:. We
can therefore use the 11173 for the high-pressure value
(8.96 J mol-
I
K-
I
). The resulting slope of the univariant
line at 800 oC and 0.75 OPa is
(
8P) = 8.96 = 1.2 x 106 Pa K
I
8T I>.G 7.39 x 10-
6
+
which again is in good agreement with the experimentally
determined slope. The reason that the thermodynamic calcu-
lation agrees so well with the experimental results is that the
initial thermochemical data used for Table 7.1 were adjusted
to conform with as many high-quality experimental studies as
possible (Holland and Powell, 1998).
Univariant curves for reactions involving a gas are not as
simply calculated as those involving condensed phases (solid
or liquid), because 11V,. and I1S,. are strongly dependent on
pressure and temperature. The free energy of a gas at high
pressures can, however, be calculated easily from Eq. (7.32),
which upon integration gives
Thus, pressure is a measure of the free energy of an ideal gas.
Under the high pressures encountered in the Earth, most
gases are not likely to behave ideally, so they will not obey
Eq. (8.6). But the functional form of this equation is so
convenient that it is preserved for real gases by inventing a
function,f; known as the fitgacity, such that
where pressure is measured in pascal. If the pressure IS
measured in bars (1 bar = 10
5
Pa), Eq. (8.5) becomes
f
'= PlI
. I
The fugacity therefore measures the free energy of a real gas
in the same way that pressure measures the free energy of an
ideal gas. The fllgacity of a gas is related to its pressure by the
fitgacity coefficient, y, where
8.3 PLOTTING UNIVARIANT UNES IN P-TDIAGRAMS 167
From Table 7.1 we obtain 11 Molar volumes of the
minerals will increase slightly with increasing temperature,
but because the coefficients of them1al expansion of most
minerals are small and similar, the value of 11V,. remains
almost constan!. The effect of the coefficient of compressi-
bility is even less important. The value of 11 can there-
fore be used at high pressures and temperatures. For the
kyanite lo andalusite reaction, it is +7.39 x 10'6 m
3
mol-l.
Ifwe are to reestablish equilibrium by raising the pressure
at 800 oC, I1G{Ill73 wil! have to become zero (the condition
for equilibrium). In that case,
Because I1G,. is negative, the reaction kyanite --> andalusite
proceeds spontaneously at 800 oC and 10
5
Pa. But the value
of I1G,. can be changed by increasing the pressure. If it is
increased by 5571 J mol-
I
, I1G,. will become zero and the
reaction will be at equilibrium. The change in I1G,. with
pressure is given by Eq. (7.34), which in its integrated fonn
gives
'G
O
'G
And
'-" /',1073 = '-" /,1073
temperature (and pressure) range of interes!. The known
starting point might also be derived from experimental
work can'ied out at high pressures and temperatures. In either
case, it is necessary to be able to extend the standard thermo-
dynamic data in Table 7.1 to conditions ofhigh pressure and
temperature.
We will illustrate the manipulation of thermodynamic
data by considering a much-studied reaction, that of the
polymorphic transfonmation of kyanite to andalusite. At
low pressures and temperatures this reaction is too slow to
be studied in the laboratory. Even at high pressures and
temperatures the reaction is extremely sluggish, but careful
experiments have shown that the two polymorphs are in
equilibrium at 800 oC and 0.72 OPa (Richardson et al.,
1969). Is this result in agreement with the thermodynamic
data in Table 7.1, and if so, what slope should the reaction
have in a P-T diagram?
Using Eqs. ( 7.14), (7.21), and (7.27) and data from
Table 7.1, the free energies offormation (1107,1073) ofkyanite
and andalusite at 800 oC (1073 K) and 10
5
Pa are determined
(see Problems 7.4, 7.9, and 7.10). With these values, we
calculate the I1G,. ofthe transfonnation ofkyanite to andalu-
site as follows:
from which it follows that P must be 0.75 OPa, a pressure in
good agreement with the experimental data.
The slope ofthe univariant line at 800 oC and 0.75 OPa is
given by 11S;/11 V/, according to the Clapeyron equation
(Eq. (8.3)). As stated aboye, the volume change of the reac-
tion at 298 K and 10
5
Pa can be used at high pressures and
temperatures without introducing serious error. The value of
can be calculated from the data in Table 7.1 for 800 oC
T
Projection 11 G
,6 stable
relative
tOa
a stable
relative
to (J
P
T
13
,6stable
relative
to a
P
P =, Pequilibrium
PLOTTlNG UNIVARIANT UNES IN P-TDlAGRAMS
\-vIfenaforceof I newton causes a displacement of I meter
(l] = lNm). Thus the units of aPlaTbecome N m-2 K-
I
or
PaK-
1
.
Forpurposes ofillustration, the phases o: and fJ have been
taken to be polymorphs, such as o: and fJ quartz. But the
principIes outlined here are equally applicable when
o: and fJ represent assemblages of phases constituting the
reactants and products of a reaction. They can be clystalline
minerals, magmas, or gases. The entropy and volume changes
in these cases are the total changes in these properties deter-
mined for the balanced chemical reactions.
Although the Clapeyron equation (Eq. (8.3)) gives the slope
of a univariant line in a P-T diagram, it does not locate the
line with respect to absolute values of pressure and temper-
ature. To detenmine this, we must know the equilibrium
temperature of the reaction at sorne specified pressure. We
might detelmine it at 10
5
Pa, for example, using the standard
data in Table 7.1 and Eq. (7.28). Such a calculated temper-
ature, however, might be far removed from the particular
astable
relative
tO,6
G
(8.3)
, - ,
,
,
: /:/
,
: ,6 stable
, relative
: to
T
T =, Tequilibrium
Constant P
astable
relative
to ,6
Divariant surface, phase f3
Divariant surface, phase a
G
(8I1G/ 8T) p
(8I1G/8Ph
Fig. 8.2 Intersecting free-energy
surfaces of phase IX and phase fJ
(shaded). Both surfaces are ruled
with lines of constant G. The line of
intersection of the two surfaces,
where the free energies of the two
phases are equal, indicates the
pressure and temperature
conditions under which the two
phases are in equilibrium with each
other. Sections through the G-T-P
plot at constant Tand constant Pare
shown below, along with a
projection of the line of intersection
ofthe two surfaces on the P-Tplane.
But according to Eqs. (7.33) and (7.34), (811G!8T)P==----:4S
and (8I1G/8Ph = I1V. Thus,
This is known as the Clapeyron equation. Itshowsthatthe
slope ofa univariant reaction (o: --> fJ) in a P-Tprojectionis
given by the entropy change (ff- 8'") divided by thevolume
change U"J - V") of the reaction. The univariantlines can
have positive or negative slopes depending on the signsofI1S
and 11 V. A metamorphic reaction in which a gas is .liberated
will invariably have positive values of I1S and 11 T, andthus.its
slope in a P-T diagram will be positive. Similarly, most
melting is accompanied by positive values of I1S and 11V
(ice is an exception), and thus most melting reactions also
have positive slopes in P-T diagrams.
Standard units for I1S and 11V in Eq. (8.3) give units of
J m-
3
K-
1
for aPlar. But ajoule is the amount ofwork done
166 Free energy and phase equilibria
Unstable
T
\
\
\
\
\
\
\
\
\
\
(\
B
p
p
el
T
8.4 SCHREINEMAKERS RULES AND G-T-P SPACE 169
Fig. 8.6 Impossible arrangement of univariant lines, because both
reactions make the phase (! unstable in the other reaction. The
sector in which pilase (! is stable must subtend an angle of less than
180
0
at the invariallt point.
Metastable
and )' both subtend anglcs of less than 180
0
arouncl the
invariant point.
The set of univariant lines around the invariant point is
now complete. Each line, while divicling the P-T plane into
fields of clifferent rclativc stabilities ofthe phascs involvecl, is
itself divicled into stablc and metastable parts by thc other
reactions. The various parts of the P-T cliagram can now be
T

< 180
(\
Stable
, O'q.
, "
',0;0
,
A
p
Fig. 8.4 Intersection of two
univariant lines (f. =ji and (1. =y.
Because the (f. =)' reaction involves (1.,
it must occur in a region where (1. is
stable; that is, to the left of the (1. = ji
line. The configuration in (B) is
therefore impossible. In (A), the
univariant line (f. = j' is dashed to the
right of the (f. = ji line to indicate that
here the reaction is metastable. The
upper part of the (f. = ji line is similarly
dashed.
Fig.8.5 Mechanical examples of
stable, metastable, and unstable
configurations of a block. The
activation energy is the energy that
has to be expended to make the
metastable block become unstable.
also be metastable. lt follows, thcn, that the stble end or a
univariant hne must he in the/iefd in vvhich thef"ee energies
oj" both the reactants and products are lmver than the /i'ee
energies orany other relatal phases in the .system.
At thc pressurc and tcmperaturc representcd by thc point
of intersection of thc two univariant curves in Figure 8.4, thc
systcm is invariant; [hat is, phases (x, fJ, and )' can coexist only
at this specific set of P-Tconditions. Along the lines extend-
ing from this point 01' intersection, the system is univariant,
with phases (X and IJ or (X and)' coexisting. Under conditions of
pressure and temperature represented by the ficld between the
univariant lines, the systcm is divariant, ancl only one phase
exists (o. in Fig. 8.4(A)). Because the stablc cncl of a univar-
iant line must be in a fielcl where its rcactants ancl proclucts are
more stable than any other phasc (or phases), the angle
between univariant lines bounding a divariant /iefd at an
invariant point must be less than 18()o. The arrangement of
univariant lines in Figure 8.6, for example, woulcl be impos-
sible because the (l. phase in both reactions would be made
metastable by the other reaction.
Because the two univariant lines that intersect to give the
invariant point represent cocxisting (J.-J ancl (1.--)', a thircl
univariant linc representing thc coexistence of)l ancl fJ must
also pass through the invariant point. Because this reaction
involves )1, it must lie in the felcl where )1 is stable; that is,
within 180 in a clockwisc clirection from the stable end ofthe
line (X = )l. But it al so involves IJ and must thereforc occur
within 180 in a counterclockwise direction from the stable
end of the line (1. = IJ. These conclitions arc met in only one
sector (Fig. 8.7). Note that the resulting divariant fclds for fJ
/3
fJ is more
stable
than iY
(\ is more
stable
than fJ
p
T
Fig. 8.3 Relative stabilities of phases (f. and fJ on either side of the line
along which the two phases are in equilibrium because their free
energies are the same.
However, for the reaction 0.->)1 to be stable, the free energies
of (1. ancl y must both be lower than that offJ. This is satisfied
only by the relations shown in Figure 8.4(A) . To the right of
the (J. = fJ univariant line, fJ is more stab1e than (l.. So a reaction
involving (J. on the right side of this line is said to be meta-
stable. That is, the reaction can take place, but it does not
involvc the most stable phases under these conditions.
The distinction between metastable, stable, ancl unstable
can be illustrated by hkening a chemical reaction to the
simple mechanical situation of a rectangular block resting
on a surface (Fig. 8.5). When the block has its long axis
parallcl to the horizontal surface, the block's centcr of gravity
is as low as it can possibly be. The potential energy of the
block is therefore at a minimum, and this conclition is
described as stable. If, on the other hand, the block were
resting with its long axis in a vertical position, the center of
gravity would be higher than in the stable configuration. The
potential energy would therefore be higher. By knocking the
block over, the potential energy woulcl be reduced. But
energy is required to topplc the block, because as the block
tilts sideways, its center of gravity is raisecl slightly because
of the pivoting action about the lower eclge. The amount of
cnergy necessary to raise the block during the pivoting is
known as the activatioll energy; this amount of energy must
be cxpended to topple the block. Once thc center of gravity
passes over the pivot point, both the activation energy intro-
cluced during the tilting and thc initial excess potcntial energy
arc transformed to kinetic energy as the block falls. In this
state the block is unstable. Had the activation energy not been
introducecl, however, the block could have rcmained in its
vertical position indefinitely, even though in this position the
block is not in its lowest-energy state. Under these conclitions,
the block is said to be metastable.
The univariant line (/. = )' in Figure 8.4 is metastable to the
right of the univariant line (/. = fJ. This part of the line is
therefore dashecl. 1t will be noted that the line for the o. = fJ
reaction is dashed aboye thc line for the (/. = y reaction
because, under these conclitions, the (l. = IJ reaction must
(8.9)
(8.10) = -f, - 1) RT lnf
8.4 SCHREINEMAKERS RULES FOR INTERSECTING
SURFACES IN G-T-P SPACE
168 Free energy and phase equilibra
For a reaction involving condensed phases and a gas, the
free energy change can be obtained by integrating Eq. (8.4):
which can be solved for P by ttial and etTOr or by using, for
example, the Solver function in Microsoft Excel (Problem 8.8).
Note thatP and llave units ofbars (l bar= 10
5
Pa= 10
5
J m-3).
where f,V;'.T,c refers only to the volume change in the con-
densed phases; normally, the value of f, V)98 can be used. The
free-energy change resulting from the volume change of the
gas is accounted for in the (RT lnf) term, where the fugacity
is given for the specified P and T. If we are interested in
knowingatwhat pressure this reaction is at equilibrium, we
setf,G:
T
tozero, and Eq. (8.9) becomes
So far, we have considered the location and orientation of
only a single univariant reaction ((X--->fJ) in G-T-P space and
its projection onto the P-T planeo The mineral s involved in
this reaction, however, might, under certain conditions of
pressure and temperature, be involved in other reactions.
More than one univariant line may, therefore, occur in the
pressure and temperature range ofinterest. Because each line
would most likely have a different slope, the lines will inter-
sect at a point in the P-T diagram, and under these conditions
of pressure and temperature, the assemblage of minerals
represented by the two lines will have zero degrees of free-
domo The possible ways in which univariant lines can inter-
sect at an invariant point follow certain simple geometrical
rules. These were first worked out by the Dutch physical
chemist Schreinemakers and are thus known as the
Schreinemakers rules (Zen, 1966).
A univariant line describing a reaction, such as the one
illustrated in Figure 8.2, divides the P-T plane in [wo
(Fig. 8.3). To the left of the line, phase (1. is more stable than
phase J, because phase (X's frec energy is lower than that
of phase J. To thc right of the line, phase fJ is more stable
than phasc 0., because under these conditions phase fJ has the
lower free energy. This condition would be equally truc if (1.
and J represented assemblages of minerals rather than single
minerals. AIso, note the deliberate use of thc phrase "more
stable than... " to indicate that we are discussing relative, not
absolute, stabihties.
Ifa second univariant hne were present, as would happen
if (x, for example, wcre able to transfonn into a third poly-
morph )', the P-T p1ane would again be divided by this hne,
with o. being more stable on one side of the line and )' bcing
more stable on the other. Let us assume that this second
reaction has a positive slope on thc P-T diagram. The uni-
variant line for the (X->)' reaction will intcrsect the univariant
line for thc o.->fJ rcaction as shown in Figure 8.4(A) and (B).
P
P P1P
T= Tp D
B
G
G
P
end of the (y) reaction to be metastable. The univariant lines
are now al! in their correct sectors, and the infom1ation from
the third reaction, that is, that the (fJ) and (y) univariant lines
must be on opposite sides ofthe (0.) reaction, serves simply to
confinll that no mistakes were made in orienting the lines
with the information from the first two equations.
This mnemonic scheme leads to two possible arrange-
ments of univariant lines that are reverse images of each
other. Soth are thermodynamical!y consistent, and without
additional infom1ation, selection of the COlTect one might be
difficult. Of course, ifthermodynamic data for the phases are
available, the decision is simple. But even when this is lack-
ing, knowledge of the tleld occurrence of the phases can
provide the answer. rt: for example, the mineral fJ were
known to occur only in high-temperature rocks, the scheme
in Figure 8.1 O(C) would be the correct one. Similarly, if one
ofthe phases is a liquid or a gas, it must appear on the high-
temperature side of a reaction with a solido
The actual slopes of the univariant lines cannot be deter-
mined from the mnemonic scheme; this requires information
on the 6.S and 6. V of the reactions or experimental data.
Commonly, however, the slope ofat least one ofthe reactions
is known, and then the scheme provides some restrictions on
the possible slopes of the remaining univariant lines. Two
reactions must, of course, be known to determine the position
ofan invariant point in P-Tspace.
,
,
, ,
/1
,
,
8.5 SCHREINEMAKERS RULES IN MULTICOMPONENT SYSTEMS 171
"
,',
1 '
'1
CW
11
T< Tp Tp T> Tp
T
e
A
G
P
fJ=y
(fJ)I(o.)l(y)
o.=fJ o.=y
(o.)l(y)l(fJ) (o.)l(fJ)l(y)
The phases in parentheses (phase absent) now refer to reac-
tions. From the mnemonic for the first equation, we read that
the stable paIi ofthe (0.) reaction (/3 = y) lies on one side ofthe
univariant line (y) and the (jJ) reaction lies on the other. Thus,
either of the arrangements of lines in Figure 8.1 OCA) or (S)
would satisfy this requirement. According to the mnemonic
from the second equation, the (0.) reaction must also lie on the
opposite side ofthe (/3) line from the (y) reaction. Thus, ifwe
choose the option in Figure 8.10(A), this condition can be
satisfied only by making the high-pressure end of the (y)
reaction metastable (Fig. 8.1 O(C)). If the option in Figure
8.10(S) is chosen, the only way that reactions (0.) and (y)
can be on opposite sides ofreaction (/3) is for the low-pressure
One reason for using this method of labeling reactions is
that it leads to a simple mnemonic scheme for working out the
sequence of univaIiant lines around an invariant point with-
out having to resort to basic principies each time. We start by
writing each reaction as a simple equation. The equation is
then rewritten, placing each phase in parentheses and replac-
ing the equals sign with the phase absent from that reaction.
Thus, for the one-component system with phases 0., (J, and y,
we write the three reactions as
Fig. 8.9 Isothermal sections
through G-T-P space showing the
relative positions of the (/, = J, a = y,
and J = y univariant lines at a
temperature below the invariant
point (B), above the invariant point
(e), and at the invariant point
(D). The relative positions of these
sections are shown in the P-T
plot (A).
(8.11)
(8.12)
where it is understood that the 2 refers to temperature and
pressure. This relation, which is known as Gihbs phase rule,
is developed more fully in Section lOS
If a system consists of (e +2) phases, the variance will be
zero. In a P-T diagram, (e+2) univariant lines radiate from
the invariant point, each line involving (e + 1) phases. These
univariant lines divide P-Tspace into (e+ 2) divariant fields.
One or more assemblages ofe phases occur in each field; one
of these assemblages is unique to each field. Thus in the
example worked out in Figure 8.7, the fommlas ofthe phases
0., jJ, and y (polymorphs of silica, for example) can all be
expressed in terms ofone component (Si02). At the invariant
point, three (e +2) phases (o., jJ, and y) coexist. If any one of
the phases is absent, the system is univariant. For example, if
y is absent, the system consists of o. and jJ, and the P-T
conditions are given by the univariant line o. = jJ. It is con-
venient to describe each of the univariant lines by the phase
that is absent. According to the symbolism used by
Schreinemakers, this absent phase is placed between paren-
theses. Thus, the o. = jJ reaction is designated by (y). Similarly,
the univariant lines for the o. = y and jJ = y reactions are
designated by (/3) and (0.), respectively.
where n is the number of environmentally controlled (inten-
sive) variables in addition to temperature (n and pressure
(P); the fugacity of water might, for example, play such a
role. If temperature and pressure are the only intensive vari-
ables, Eq. (8.11) is commonly written as
f=e-d,J+T+P+n
In the example aboye, we considered a system consisting of
three phases, 0., jJ, and y. The principies detelmining the
arrangement of univariant lines about an invariant point,
however, apply equally to systems containing more phases,
the only difference being that as extra phases are added, the
number of univariant lines increases. This number depends
simply on how many independent reactions can be written
between the minerals involved.
If we designate the number ofphases (minerals, magmas,
fluids)present by 1;, and the minimum number of eomponents
necessary to describe the formulas ofthe phases by e, then the
number of degrees of freedom (variance) of the system is
given by
8.5 SCHREINEMAKERS RULES APPLlED
TO MULTICOMPONENT SYSTEMS
o.-->y reaction metastable at these temperatures. The intersec-
tions of the o. and jJ surfaces and the jJ and y surfaces both
become stable under these conditions. At the temperature of
the invariant point (Fig. 8.9(D)), the free-energy surfaces for
the three phases intersect at a point. Only under these con-
ditions do all three phases have the same free energies; that is,
they are equally stable (see Problem 8.1).
f=c+2 cj)
.8
,
,
P
P
T
Fig.8.7 Thermodynamically permissible arrangement of univariant
lines separating the divariant fields of the phases a, J. and y.
T
Fig.8.8 Relative free energies of phases a, J, and y in various
parts of the P-T projection.
labeled in terms of the relative free energies of each of the
phases (Fig. 8.8).
The relations between the free energies and the stability
of the phases are most ciearly seen in isothelmal G-P plots
(Fig. 8.9). At temperatures below the invariant temperature
(Tp), the trace of the free-energy surface of the jJ phase is
aboye the point of intersection of the free-energy surfaces of
the o. and y phases (Fig. 8.9(B)). Reactions with the jJ phase
(1 and 2 in Fig. 8.9) are therefore metastable. At temperatures
aboye the invariant point, the trace ofthe free-energy surface
of the jJ phase lies below the intersection of the free-energy
surfaces ofthe o. and y phases (5 in Fig. 8.9(C)), making the
170 Free energy and phase equilibria
1
1
1
1
/0/
1
11 1
/0,'
;>..,11,
/(::;,
1
1
(O) (Pe)
<;;;..\\ O
o"
(0)<::0 (Pe)
x

fa ",,"O
...e-,.-'---- (En)
Pe+ O
T
T
(Fa)
(Fa)
D
B
P
Fa = Pe + O Pe + O = En
(O) (Pe)
Continued heating oI' the assemblage Fo + En would eventu-
al!y result in the breakdown of En to Fo + Owhen it crosses
the (Pe) reaction (Fig. 8.11 (B)). The assemblage En Fo is
theretore restricted to the sector between the (O) ancl (Pe)
univariant lines.
Divariant fielcls marking the stability limits of a single
phase are also given by this diagram. Enstatite, 10r example
(Fig. 8.11 (D), is stablc only between the univariant lines (Fo)
and (Pe). Here again, the seetor in which a single phase is
stable cannot exceed 18()O
The reaetions given by the univariant lines radiating 'om
the invariant point ean be represented graphieally as shown in
Figure 8. 11 (E). We wil! start with the assemblage in the upper
left of the diagram and proeeecl in a eloekwise direetion
around the invariant point. The phases stable in the seetor
boundecl by the (O) and (Fo) univariant lines are Pe + En and
En + O, We see that the latter assemblage is possiblc by
eonsiclering what the reaetion produets would be if a mixture
of Pe + O, eontaining an exeess oI' O, was raised aboye the
(O)
O Pe O Pe En O
<l---<>----<l"(En<>-: ...
'Fa (En) P
Pe+O
8.5 SCHREINEMAKERS RULES IN MULTICOMPONENT SYSTEMS 173
T
/
/
T
/ /
(Fo)
A
(Fa)
e
P
Pe En O
(O)
(Pe)
SI e
" "'- Fo

<;;;..\\ O Pe O
Pe En O "
,o
"

" <:(0
o" El El
"
Fo
P
Fo (En)
Pe+O
"
"/
S'
,
O
Pe O
\le"""
El
" (Fo)
F T
P
Pe + En = Fa En = Fa + O
Pe En O Pe En O Pe
El El El El S e El El

Fa Fa
E (Pe)
Figure 8.11 (B). The third reaction requires that the (Pe) and
(O) lines lie on one side ofthe (Fo) line and (En) on thc other.
This can be satistled only by placing the (En) linc as shown in
Figure 8.11(C). The diagram is now complete but can be
checked by examining thc fourth rcaction. This indicates
that the (Pe) and (O) lincs should lic on the oppositc side of
the (En) reaction f!-om thc (Fo) line, which indced they do.
The tour univariant lines in Figure 8.11 (D) separate four
divariant tlelds, each characterized by assemblages oftwo (e)
phases. These fields each subtend angles of less than 180
0
at
the invariant point. Some of these assemblages, such as that
of Pe + Oin the lower sector, can be read directly from the
phases involved in the reactions. Others, such as that of
Pe+Fo, must be deduced. Consider the asscmblagc Pc+En
being heated until it crosses the (O) reaction to form Fo (Fig.
8.11 (A)). fthe initial assemblage contained an excess of En
over that which was needed tor the reaction, thc reaction
products would be Fo plus the excess En. Thus, En is stable
to the right of the (O) reaction as long as no Pe is present.
Fig. 8.11 Sequence of steps
involved in arranging the univariant
lines around the invariant point
involving periclase (Pe), forsterite
(Fo), enstatite (En), and quartz (O)
in the two-component system
MgO-Si02 . See text for explanation.
(3)
(3)
(-)
I
I
I
I
T
T
n
I
I
I
I
0=3
(n)
(ct)
D
B
P
P
2. Mg2Si04 + Si02 2MgSi03
Forsterite Ouartz Enstatite
(Fo) (O) pel (En)
3. MgO + Si02 MgSi03
Periclase Ouartz Enstatitc
(Pe) (O) IFol (En)
4.2MgO + Si02 Mg2Si04
Periclase Ouartz Forsterite
(Pe) (O) IEnl (Fo)
Thermodynamic data indicate that al! of these reactions
have positive slopes in P-T space and that quartz and peri-
clase are stable on the high-temperature side of the (Fo)
reaction (Fig. 8.11 (A)). According to the mnemonic scheme
10r the first equation, the (Fo) reaction lies on the opposite
side ofthe (O) reaction from the univariant lines for (Pe) and
(En). Starting with the (Fo) line as given, the (O) line is drawn
with a positive slope; it intersects the (Fo) line to create the
invariant point (Fig. 8.11 (A). Because the (O) reaction
involves enstatite, the reaction is metastable below the (Fo)
reaction. Similarly, because the (Fo) reaction involves ensta-
tite and periclase, it must be metastable to the right ofthe (O)
reaction. The univariant lines 10r (Pe) and (En) can, according
to the tlrst reaction, be placed anywhere on the right side of
the (O) reaction. However, according to the second reaction,
the (Fo) and (O) lines must lie on one side oI'the (Pe) line and
(En) on the other. Also, because the (Pe) reaction involves
enstatite, only that part aboye the (Fo) line can be stable. The
(Pe) line can theretore be placed only in the sector shown in
b)
(n)
T
T
I
I
(l
,
,
,
ct= 3
(ct)lul(3)
Mg2Si04
Forsterite
(Fo) 101
MgSi03
Enstatite
(En)
+
P
(3)
(;3)
e
172 Free energy and phase equilibria
A
Fig.8.10 Sequence of steps in
arranging univariant lines around an
invariant point in a one-component
system. The line fA = (1, designated by
the phase absent ()'), is taken as
given. The remaining two lines must
plot on opposite sides of ()') as P
shown in (A) or (B) Part of the ()')
reaction then becomes metastable
as shown in (e) or (D).
1. MgO
Periclase
(Pe)
Although the example worked out aboye is tor a one-
component system, the mnemonic scheme works equally
well for systems containing more components. Consider, tor
example, the two-component system MgO__Si02. Common
mineral s in this system inelude perielase (MgO), forsterite
(Mg2Si04), enstatite (MgSi03), and quartz (Si02). Because
the system has two components, it would be invariant if al!
four minerals were present together (l = O = e + 2 - (j;).
Ractiatingfrom this invariant pointwouId be tour (e+ 2) uni-
variant lines,each representing. the. coexistence of three min-
eraIs (e+ 1).
Ihe compositions. f the . minerals and the reactions
between them can be graphical!y represented by a simple
line that at one end is pure MgO and aUhe other is pure
Si02 (Fig..8.11 (E. Enstatite (UVIgO+1Si02), for example,
would plot halfway along this line, whereas forsterite
(2MgO + 1Si02) would plot only one-third ofthe way toward
Si02. Reactions between the minerals can be determined
simply by inspecting the relative positions of the minerals
on this compositional lineo A reaction product must lie
between the reactants. For example, periclase and enstatite
could react to Jorm forsterite. The four possible univariant
reactions between perielase, forsterite, enstatite, and qumiz
and their mnemonic schemes are as follows:
T
univariant lines are adjusted until they are consistent with the
mnemonic for each reaction. This leads to the unique solution
shown in Figure 8.12.
Once the univariant lines have been positioned in their
correct sectors the slopes ofthe lines can be adjusted to agree
with the slopes determined from the thel1110dynamic data.
This should not move a univariant line into a new sector
unless a mistake has been made in the positioning of the
univariant lines or if there is an error in the thermodynamic
data. Each of the lines can now be labeled with the
(An) (En) (Co) (Cd) (Sp)
A
(Sp) orAn)
CJ
x ~
,- o0\x
" CJ CJ
~
0.<:: "
P
&
Table 8.2
I1V I1S dPldT
Phase absent Rcaction Mncmonic (m' x 10
3
) (J mol
-1 K 1) (Pa K-
1
x lOS)
(An) SEn + SCo = Cd + 3Sp (En)(Co)1(An)I(Cd)(Sp) 0.068 55.42 8.1
(En) Cd + SCo = SAn + 2Sp (Cd)(Co)1(En)I(An)(Sp) -0.024 -36.37 15.2
(En) 2Cd+ Sp = SEn+ SAn (Cd)(Sp)1(Co)1(En)(An) -0.092 -91.79 10.0
(Cd) An+Sp= En+2Co (An)(Sp)1(Cd)I(En)(Co) -0.009 -3.81 4.2
(Sp) 2En+3An = Co+Cd (En)(An)I(Sp)1(Co)(Cd) 0.042 43.99 10.5
En Cd Cd An Cd En An En En Cd
++ ++ ++ ++ ++
A
CO
Sp O S,&:PA"ti, C0ti" Co
11 III IV V
8.5 SCHREINEMAKERS RULES IN MULTICOMPONENT SYSTEMS 175
Fig.8.12 Construclion of a P-T
diagram for the univariant lines at
the invariant point involving
enstatite (En), cordierite (Cd),
andalusite (An), corundum (Ca),
and spinel (Sp) in the three-
component system
MgO-AI20 r Si02. See text
for explanation.
ofthc (An) linc. A short scgmcnt ofthe metastable cxtension
ofcach line should be plottcd beyond the invariant point; this
will help in positioning thc remainder of thc univariant lines.
Ncxt, consider the (En) rcaction. According to thc mnc-
monic, the (Cd) and (Co) reactions must lie on one side ofthe
(En) linc and thc (An) and (Sp) reactions on the othcr.
Depending on how these lines wcrc drawn on either sidc of
the (An) linc in thc first step, they may need to be switched or
adjusted to bc consistent with the mnemonic for the second
reaction. Continuing in this manner, then, thc positions ofthc
As the number of components increases, so does the
number of univariant lines radiating from an invariant point.
The simple mnemonic scheme outlined aboye, however, can
still be used to determine the sequence ofunivariant lines. Let
us consider the system MgO-AI20 r Si02. Because it con-
tains three components, an assemblage offive minerals (c+ 2)
in this system will be invariant, and radiating from an invariant
point will be five univariant lines. This system contains many
minerals, any group offive ofwhich will produce an invariant
assemblage. We will consider the invariant point associated
with the five minerals andalusite (An), corundum (Co), spinel
(Sp), enstatite (En), and cordierite (Cd).
Because the system contains three components, the com-
positions ofthe minerals can be expressed conveniently in a
triangular plot where each apex ofthe triangle represents one
of the three components (Fig. 8.12). From such a plot, the
reactions between the minerals can easily be detem1ined.
First, draw a set of lines joining the compositions of each of
the minerals, but no two lines may intersect. The set of lines
shown in triangle 1 in Figure 8.12 is one possible way of
drawing these lines, but any of the arrangements shown in
triangles 1 to V would be acceptable. These lines divide a
triangle into a number ofsmaller triangles. A rock with a bulk
composition plotting within any one of these subtriangles
consists of the minerals at the apexes of the subtriangle. For
example, a rock composed of40% Si02, 30% MgO, and 30%
AI20 3 (x in triangle I) would consist of Co + En + Cd.
Next, remove any one line and replace it with another line
joining a different pair of minerals. For example, the line join-
ing En and Co can be replaced with one joining Cd and Sp (H in
Fig. 8.12). These switching lines constitute a reaction, which in
this case is En + Co = Cd + Sp. Because this reaction does not
involve andalusite, it is refelred to as the andalusite absent
reaction or (An). Note that when this one line is switched, the
remainder of the lines do not change; that is, only one line is
changed at a time. Continuing in this manner, the remainder of
the reactions are easily detel1nined (Fig. 8.12 (H) to (V.
We now write out the balanced reactions and the mne-
monic scheme for each of the five univariant lines radiating
from the invariant point (Table 8.2). From the /J.S and /J.Vof
each reaction, the slopes of the univariant lines are deter-
mined. The /J.S of the reaction also indicates which assem-
blage is favored at high temperature. For example, with
increasing temperature, reactions (An) and (Sp) proceed to
the right, but the remainder proceed to the left.
Stm1 by placing the (An) reaction on the P-Tdiagram with
its appropriate slope (it is instmctive to draw this yourself and
then compare it with Fig. 8.12). From the mnemonic, we
know that the (En) and (Co) lines must plot on one side and
the (Cd) and (Sp) lines on the other side ofthe (An) reaction.
Two complementary diagrams could be drawn ifwe did not
know that the (An) reaction proceeds to the right with increas-
ing temperature; that is, Cd and Sp are stable together on the
high-temperature side of the (An) lineo Thus the (Cd) and
(Sp) lines must be placed on the low-temperature side ofthe
(An) line, and (En) and (Co) on the high-temperature side.
These lines can be plotted temporarily anywhere on each side
(Fo) reaction. Clearly En + Q would remain. The first com-
positional line in Figure 8.11 (E) therefore has the minerals
Pe, En, and Q marked on it. If a rock has a bulk composition
that lies between En and Q, it would consist of the minera1s
En and Q. A rock with a composition between En and Pe
would, on the other hand, consist of En and Pe.
On crossing the (Q) univariant line, Pe reacts with En to
fOl111 Fo. Thus, on the compositional line, the mineral Fo is
added. Now rocks having bulk compositions between En and
Pe will consist of Pe + Fo or Fo + En, depending on whether
they fall to the left or right ofthe Fo composition.
Continuing in a clockwise direction, the (Pe) reaction is
encountered next, and En breaks down to form Fo+Q. The
disappearance of En from the compositional line means that
rocks of appropriate composition will contain Fo + Q. Note
that the assemblage Pe + Fo is still stable in MgO-rich rocks.
Next, Fo breaks down to fOl111 Pe + Q on crossing the (En)
reaction, leaving only Pe and Q as stable minerals on the
composition lineo Any composition in this system will there-
fore consist of some mixture of Pe and Q.
Finally, on crossing the (Fo) reaction, Pe and Q react to
form En, and we then have the assemblage with which we
started. It is important to note that each one of these steps
involves changing only one mineral at a time. Ifyou find that
two minerals have been changed in going from one sector to
the next, a mistake has been made in positioning one or more
of the univariant lines.
We are now in a position to label all ofthe univariant lines
and divariant fields with the appropriate mineral assemblages
(Fig. 8.11(F to complete the P-T diagram. Such diagrams
provide a convenient means of analyzing the significance of
mineral assemblages found in rocks. Most rocks will fall in
divariant fields in P-Tdiagrams, because they typically have
existed over a range of pressures and temperatures. As a
result, most rocks will contain e phases, which in this
example is two. Rocks containing more than e phases must
lie on univariant lines or at invariant points, or be out of
equilibrium.
Univariant andinvariant assemblages are not as uncom-
mon as might at first be expected. If a change in intensive
variables causes a divariant assemblage to reach a univariant
reaction, the enthalpy change of the reaction may be large
enough to buffer and control the intensive variables. For exam-
pie, by analogy,itisnot anunusual OCClllTenCe to find water
boiling in a potn a kitchen,yetthe univariant reaction water->
steam is actually invariant under the ambient pressure of 105
Pa; that is, you know that the temperature of the water is
100 oC if it is boiling. A stove may transmit considerable
energy to the water, but this heat is used in vaporization rather
than in raising temperature. Thus the boiling buffers the
reaction at 100 oC as long as there is water to boil. In a similar
way, reactions, such as the partialmelting ofthe upper mantle
or the growth of a metamorphic mineral in a contact aureole
of an igneous intrusion, can produce univariant or invariant
assemblages that will, through their enthalpies of reaction,
maintain the intensive variables at the special values ofuni-
variant or invariant equilibria, as long as reactants remain.
174 Free energy and phase equilibria
(O)
(Cd)
Pe Sp
(Cd)
(Sp)
O Cd
T
Sp Co
present in that tield. In nature, rocks may form under con-
ditions eorresponding to those of clivariant fields, univariant
lines, or, under unusual conditions, invariant points. P-T
diagrams therefore provide an invaluable tool10r the analysis
of the mineral assemblages found in rocks, ami they will be
used extensively throughout the remainder of this book.
8.8 PROBlEMS
8.1 For the system illustrated in Figure 8.9, draw three iso-
baric (constant pressure) G versus T diagrams tor con-
ditions where P < P IP, P> PIp, and P = PIp, where PIp is
the pressure at the invariant point. Label all isobaric
invariant points and indicate whieh parts ofeach diagram
are stable and which are metastable.
oSIJ
e
o-.p
0
x
o
(Sp)
Cd O
(Co) (O)
(Pe) (Cd)
(Co)

(Pe)
(O)
p
+ + + +
Sp Pe Co Co

Fig.8.13 Construction of a P-T
diagram for the univariant lines at
the invariant point involving quartz
(Q). cordierite (Cd), corundum (Co),
spinel (Sp), and periclase (Pe) in
the three-component system
MgO-AI20 r Si02. See text for
explanation.
conclitions under which the reactants amI products are at
equilibrium (i1G = O).
The line of intersection of two free-energy surfaees is said
to be univariant, beeause only (me of the two intensive
variables, T or P, can be independently varied. If a second
univariant line is present involving the same components, the
system is invariant at the point ofinterseetion ofthe two lines.
Only under this one set are the assemblages
of phases from both univariant lines stable together.
The position ancl slope ofunivariant lines in P-Tspaee can
be calculated from thc thermodynamic data for the phases. In
acldition, Schreinemakers rules provicle a simple geometrical
means of correctiy orienting univariant lines about an invar-
iant point. The univariant lines clivide P--T space into divar-
iant fields; that is, regions in which both P and T can vary
independently without modifying the mineral assemblage
8.8 PROBlEMS 177
(Q)(Sp)
(Co)(Pc)
(Cd)
I(Co)(Pe)1
I(Q)(Cd)1
I(Sp)1
Cd = 5Q + 2Sp
Sp = Co + Pe
5Q + 2Pe + 2Co = Cd
the mnemonies for whieh are
8.7 SUMMARY AND CONCLUSIONS
We have seen that the Gibbs free energy of a phase (or group
oLphases) .varies as a function of temperature (-S) and
pressure (V). Because different phases are not likely to have
identical molar entropies and volumes, free energy surlces
wil! not be parallel in G-T-P space. Where these surfaces
intersect, the free energies ofthe two will be identieal. Thus, if
the phases are related by a reaction, the line of interseetion of
the free energy surfaces defines the pressure and temperature
(Cd)
(Sp)
(Q)(Pe)(Co)
The first reaetion does not involve Co or Pe; so it can be
referred to as the (Co) or (Pe) lineo Similarly the seconcl
reaetion is both the (Q) and (Cd) reaction.
We will start with the (Sp) reaction on the P-T diagram.
According to the mnemonie, the (Q), (Pe), ancl (Co) reactions
lie on one side and the (Cd) reaction on the other side of the
(Sp) lineo Beeause Cd is stable at lower temperatures than
Q+Pe +Co, the (Cd) reaetion must lie on the high-temperature
side of the (Sp) line and (Q), (Pe), and (Co) on the low-
temperature side. But (Q) amI (Cd) are the same reaction.
Therefore, the reaction of Sp = Co +Pe passes straight through
the invariant point with no ehange in slope. On the high-
temperat1ll'e side it is labeled (Ccl) and on the low-temperature
side, (Q).
The mnemonie for the seeond reaction indicates that the
(Sp) reactionlies on one side ofthe (Q) or (Cd) reaetion and
the (Co) and (Pe) reactions on the other. But the (Co) and (Pe)
reactions are the same one; thus they must pIot as a single
lineo AIso, from the third reaction we see that the (Co) and
(Pe) reactions must plot on the same side 01' the (Sp) reaction
as the (Q) reaetion. The univariant lines for (Co) and (Pe)
therefore pIot as a single line as shown in Figure. 8.13.
We see, then, that ifthe indifferent phases lie on the same
side 01' the eollinear or singular phases, a single univariant
line passes straight through the invariant point ((Q) and (Cd)
in Fig. 8.13). This clegeneracy does not lead to a deCl'ease in
the number ofunivariant lines, but two ofthe lines will have
identieal slopes. Ifthe indifferent phases lie on opposite si des
ofthe singular phases, two univariant lines become one, This
therefore leads to a decrease in the number ofunivariant lines
about the invariant point.
(Co)(Pe)
(Q)(Cd)
(Sp)
As before, reactions between phases can be cletermined
by inspection of the compositional triangle (Fig. 8.13).
However, unlike the previous example, where five different
reactions were obtained, only three occur at this invariant
point. They are
176 Free energy and phase equilibria
appropriate reaction. Care must be taken to get the correct
assemb1ages on each side of each lineo In doing this, it is
usefu1 to recal! that the sector subtended at the invariant point
by a phase assemb1age cannot exceed 180
0

A final check on the correctness of the scheme of uni-

variant lines can be obtained by constructing the composi-
tional triangles in each divariant field based on the labeling
of the univariant lines bounding the fields. Lines can be
drawn between any pair ofphases that a reaction indicates is
stable in the particular sector. In the lower right sector of
Figure 8.12, for example, Cd +Sp are stable together and
can therefore be joined with a lineo But this assemblage
could also inelude Co, if Co had been present in excess
when the (An) reaction formed the Cd +Sp. Consequently,
Cd, Sp, and Co can al! be joined with lines. Had En been in
excess instead of Co, then Cd +Sp +En would al so be stable.
These phases, therefore, can also be joined with lines. The
resulting compositional diagrams in each sector differ by
only one line from the diagram in the adjoining sector. 1f
they do not, univariant lines have been incorrectly positioned
or labeled.
8.6 DEGENERATE SYSTEMS
In general, as many univariant lines radiate from an invariant
point as there are phases in the system; this wil! be two more
than the number of components. In sorne systems, however,
reactions between certain phases may be described in terms
of a smal!er number of components than are needed to
describe the other reactions being considered. In such cases,
the number of univariant lines decreases and the system is
said to be degenerate.
A system becomes degenerate in two ways. First, two or
more phases may have the same composition (polymorphs,
for example). Second, in systems consisting ofthree or more
components, three or more phases may be col!inear or four or
more phases may be coplanar in ternary or quaternary com-
positionaL diagrams, respectively. The result is that some
reactions will involve fewer than e + I phases. Those phases
taking part in reactions involving fewer than c+ 1 phases are
termed singular phases, and the remainderare cal!ed
ferent phases. Two types of degeneracy occur depending on
whether the indifferent phases plot on opposite or the same
sides of the line joining the singular phases. This can be
illustrated by considering a specifc example.
The invariant point involving corundum, periclase, spinel,
cordierite, and quartz in the system MgO-AI20 r Si02 is
degenerate for two reasons. First quartz, cordierite, and spinel
plot on a straight line in the compositional diagram; second,
perielase, spinel, and corundum also plot on a straight line
(Fig. 8.13). In thefirst case Q, Cd, and Sp are considered the
singular phases, whereas Pe and Co, which plot on opposite
sides of the collinear group, are the indifferent phases. In the
second case, Pe, Sp, and Co are the singular phases, and the
inditlerent ones are Qand Cd, which both plot on the same
sicle of the collinear group.