If the existence of the new, proposed phases,
and 
, can be confirmed, iron
would consist of five possible phases
from which the phase diagram would be a nightmare to untangle
without a fine-scale experimental mapping.
The new phases, plus the
disagreement between researchers from
several different laboratories
on the location of the melting
curves, have allowed much speculation
on the detailed iron phase
diagram. It is not
surprising that several
quite different phase diagrams
have been proposed recently.
The turning point of many
discussions about the melting
temperature profile of
the core is the
shock wave results
of Brown and McQueen [1986],
who found a solid-solid (s-
)
transition at 200 GPa and
4400 K (also see Boness
and Brown [1990]) and a
solid-liquid (s-
)
transition at 243 GPa
and 5720 K. Their
values were
determined from the measurement
of the shock 
Hugoniot curve by a traditional
thermodynamic formula. The presence of the
s-
transition at 200 GPa and 4388 K is
a stumbling block to some who want to extrapolate
their melting curves to 330 GPa.
For example, using his new data above 100 GPa,
Boehler [1994] proposed that the B&M s-
transition
at 200 GPa and the s-
transition at 243 GPa should have
been reported at lower temperatures. By assuming the
lowering of the B&M 
at 243 GPa and the B&M s-
transition at 200 GPa by 1100 K, he cleared out this
region of the phase diagram, so that he could
extrapolate his 
curve measured up to 190 GPa
to beyond 190 GPa unimpeded,
obtaining 
,
the lowest value reported (see Figure 2d).
But the new shock wave results of Gallagher
and Ahrens support the position of the B&M
s-
transition as they placed it. So
if Gallagher and Ahrens are correct, the
Boehler extrapolation, 
,
is untenable. Yoo et al. [1994a],
extrapolated their shock wave determination
of 
, finding 
at 330 GPa. But they had to ignore the B&M melting data
at 243 GPa and the Boehler DAC data.
The phase diagram proposed by Williams et al. [1991]
is such that their 
-
boundary
is displaced from B&M's solid-solid point at
200 GPa, and the trace of their 
-liquid
boundary is nowhere near the solid-liquid B&M
point at 243 GPa (see Figure 2a).
The 1987 Williams et al. extrapolation
yields 
,
the highest value on record
(Figure 2a). The
Williams et al. extrapolation, however,
relied on the Bass
et al. [1987] shock
wave points. Since
Gallagher and Ahrens
have moved these points
down by about 
,
the Williams et al.\
extrapolation,
,
appears to be untenable.
Saxena et al. [1994] published a paper using their newly discovered
-
boundary and proposing that it coincides with
the s-
boundary of B&M at 200 GPa and 4000 K. This satisfies the B&M
s-
datum and eliminates the need for a 
phase above 200 GPa.
The Saxena et al. 
-
phase boundary intercepts the
-liquid boundary, establishing a triple point at 216 GPa and 4500 K (Figure 2c).
Their analysis gives 
near 6000 K. Their 
curve above 216 GPa also
satisfies the B&M s-
point at 243 GPa and 5720 K.
There are, however, imperfections in this phase diagram, as
discussed by Anderson [1994]. These are: 1) the Saxena et al.
solid-solid 
-
phase boundary
shows curvature (see Figure 2c). There are thermodynamic
reasons why a solid-solid phase boundary of a transition metal,
such as iron, should be a straight line [ Young,1991].
A deflection from a straight line of a phase boundary in
the phase diagram of a metal is caused by a triple point
or a sudden change in configurational entropy [ Weathers
and Bassett, 1990]. This means that the change of the 
-
slope at about 120 GPa would probably be caused by a t.p., which
would require an s-
branch in addition to the two shown in Figure 2c. The
extra branch may cut off the 
branch from the deep core.
2) the Saxena solution ignores the 100 GPa t.p. well mapped by Boehler [1993] and shifts
the t.p. to 60 GPa, where the three branches are not well worked
out. Until these imperfections are worked out, Saxena's phase diagram will not be preferable.
Anderson's [1994] phase diagram (Figure 2b) shows a triple
point at 190 GPa that joins the Gallagher and Ahrens data,
the Boehler data, and the Brown and McQueen datum (200 GPa).
Anderson's proposed 
exiting the top of the t.p. is less
than that indicated by the Gallagher and Ahrens data
(Figure 1), but it is qualitatively the same. Using the Lindemann
formula, Anderson found that the 
of the upper branch
intersected the 330 GPa isobar at 
.
The Anderson solution shows a negative slope of the
branch of the 
-
boundary. This arises because
the virtual coincidence of an 
-
transition and
an s-
transition required the density of the 
phase
to be less than that of the 
phase in order to
satisfy the Brown and McQueen velocity-pressure behavior.
Thus the 
-
phase boundary may resemble the
better known 
-
phase boundary at low pressure.
Isaak et al. [1994] find that bcc iron appears to have
a slightly lower density than hcp iron at deep core conditions
and would satisfy the negative 
of the 
-
boundary.
Eliminating the 
estimates of Williams et al.\
and Boehler because of Gallagher and Ahrens's work
and summarizing the current experimental estimates
for 
, we have Yoo et al. [1994a],
6830 K; Gallagher and Ahrens [1994] about 6400 K (author's extrapolation)
and Brown and McQueen [1986], refined by
Boness and Brown [1990], 5800 K. All extrapolations have overlapping
error bars. For comparison, the thermodynamic estimate
of Anderson [1994]
is 
. There is a relatively small
difference between the largest and smallest estimate
of 
, 1000 K.
There has been progress
in the theoretical calculation of 
for iron. Poirier and Shankland
calculated 
from dislocation melting
theory and found that for iron, the 
at 330 GPa
is 5600--6160 K, depending on crystallographic structure.
It is close to Bukowinski's [1977] calculation, 
.
Kerley [1994], modeling iron as a fluid of
hard spheres, found 
.
In his recent review paper, Stacey [1995] adopted Poirier
and Shankland's value of 
,
6000 K, but extended its limits.
In summary, there is overall convergence
to the value of 
.
While there does appear to be convergence on the value of 
,
there is still room for disagreement on the structure of pure iron
at inner core conditions. The conventional wisdom is that hcp 
iron dominates at inner core conditions. But the
new Gallagher and Ahrens data, joining with Boehler's DAC data and the Brown and
McQueen datum to form a t.p. at 200 GPa,
cast doubt on this assumption. Nevertheless,
the hcp phase 
has been proposed for the
inner core by Saxena et al. [1994] (Figure 2c). Saxena's phase
diagram honors the data near the 200 GPa t.p. But, as mentioned
above, details of his phase diagram are questionable.
Anderson's [1994] proposed phase diagram (Figure 2b) has 
as the basis
of the inner core, and he suggested the fcc structure for
. Bukowinski's [1977] theory
of fcc iron at inner core conditions gives a theoretical
foundation for Anderson's proposal.
The final choice between the 
phase
and the 
phase for the core may depend
on the outcome of the current work on proving the existence
of the 
phase by identifying its
crystallographic structure. In the meantime,
the uncertainties of the Saxena phase diagram
make 
(fcc) the best choice for
the inner core.