CFMASH +An,Qtz,HzO
3.5.3 The effect of uncertainties in the data and activity models
As discussed
previously,low
variance assemblages from natural amphibolites suggest that the phase diagramfor
amphibolites islikely
to involve the three invariant pointsrepresented
in Fig.
3.4 (i.e.[chl, Hbl], [chl,crd],
and[chl, Grt])
and possibly also [Als,chll
or [St,Chl] which contrastswith
the calculations summarised in Fig.3.7.
Since thediscrepancy appears unlikely to be due to the effects of additional minor components or
reduced agr6,
it
seemslikely
that some of the thermodynamic data may be inerror'
Note that a similar conclusion was reached byHowell
(1991) and also byXu
(1994) who madeadjustments to the enthalpy value
for
end-membetgedritelntheFMAsH
system until thecalculated phase relations matched those known from petrological observation.
In the case of the chlorite-bearing, orthoamphibole absent CFMASH model system, the least
well
consüained thermodynamic data are those for the end-members, Fe-staurolite' Mg- staurolite , tremolite,Fe+remolite
and hornblende (see AppendixA2).
Another factor whichChaPter3-CFMASH- 46
tr
x il-¡
I¡
"x x- Þ
a
Xx
x
a0.9 0.8 0.7 0.6
s
0.5X
0.3 0.2 0.1 0
rH o
Selv.L O RAJ
^ S&R À F&H X
SPearxs
o 0.05 o.1o 0.15 0'20 0.25 o'30 o'35 o'40 0'45
0'50XFeg*,Mz
Chaprcr3-Ct;MASII-
47may cause problems in the calculations is departure in the real minerals from the a-X relations assumed in formulating the data and the permanent datafiles. The most
likely
sources of such discrepancies arein
the mixing of end-members in which some of the cations areof
significantly different size. For example, garnet
in
theCFMASH
system has been considered to be an ideal solid solution of pyrope, almandine andgrossular.
However, given thedifferences in the size of calcium, iron and magnesium ions, mixing beween these end- members is
likely
to be non-ideal.In
the case of the aluminous chlorite endmember, arnesite, the experiments used to derive the thermodynamic data were based on solid-solution chlorite which did not closely approach the amesite endmember. Thus the thermodynamic data are strongly dependent on the (possibly suspect) activity model used to derive them (R. Powell, pers. comm.,1993).
The effect of the non-idealityof
a-X relationsin
garnet, andof
adjusting the thermodynamic datawithin
error for the lesswell
constrained end-members, are discussed below.Garnet
Non-ideality of mixing may be modelled using activity coefficients which relate the activity of the ideal end-member to that of its non-ideal, more realistic counterpart (Powell, 1971;Wood
&
Fraser,1984).
(See Appendix44,
section A4.3.2for
the formulationof
activity-composition (a-X) relations for TI{ERMOCALC permanent datafrles.) Activity
coefficients
(ï)
are a function of the composition of the phase (RT/n Y= wAij/(X)
where wA¡i are interaction parametersfor
the mixing of the end-membersi
andj
in the phaseA).
Powell (pers. comm., 1gg3) has suggested two models for non-idealityin
the mixing relationsof
pyrope, almand.íne and grossularin
gamet (R. Powell, pers. comm.,1993); one for Ca-richand the other for low-Ca garnets. The garnets of interest in hornblende-staurolite amphibolites typically
fall in
the low-Ca region where the interaction parameters have determined as:wGrtF.Mg
=
190R=
1.6 kJK-l
wGtrcaFe - wGrrcaMg = -1100R = -9.1 kJ
K-l
(R. Powell, pers. comm.,
1993). A
range of valuesfor
wGttcaFe ând wGrtaaMg were used to formulate activiry models of garnet and the effects of these are illustrated in Fig. 3-12.Fot
significant non-idealityin
garnet (i.e. large interaction pammeters e.g. wcrtçaFe = 5 kJ K-1 and wG.tcuMs= l4.I
kJ K-1), theCFMASH
invariants[Hbl,
Oam] and [Chl, Oam] move a signif,rcant distance toward [Grt, Oam],implying
that non-idealityin
garnet activity-composition (a-X) relations may be part of the reason for the difficulties encountered in the chlorite-bearing
CFMASH grid.
However, the effectof
the non-idealityin
garnet is insufficient to allow the predicted phase relations to corespond with those observed petrologically.Chapter3-CFMASH- 48
[Crd]
[Als]
I I
. original I
i
adjusted staurolite
tstl tGrtl
/
tchll
o /
lHbll
10
I
I
7
L(ú
-ol¿
o-
6
5
550 600
650
700T
("C)Chapter3-CFMASII- 49
Chlorite
Although the ideal mixing model assumed in the derivation of the thermodynamic data of amesite is probably an over-simplif,rcation, the resulting data are strongly dependent on the
activity
model used to derive them (R. Powell, pers. comm',1993)'
Thus, the uncertainry associatedwith
the enthalpy of amesite is probably larger than the listedsd(fl)
(see AppendixA3).
To determine whether the assumption of the ideal mixing allows sufficiently uncertainty to cause an inversionof
the phase diagram topology, the enthalpy of amesite (AII¡ttss) was variedwithin
the furl range of its 2o error bar (+6 kJmol-l).
This caused all of theCFMASH
invariant pointsinvolving
chlorite to move toward the[Chl]
inva¡iant (which asit
does notinvolve
the end-me mber arnesire, obviously is not affected by changes in the data for amesite),(Fig.
3.12).
The degree of movement experienced by each of the invariant points ÀFla¡¡ss wâs slight and certainly insufficient to cause an inversion of the topology'Hornblende
The enthalpy of the hornblende end-members hornblende, nemolite and Fe-tremoiite are associate
with
considerable uncertainties (sd(II)
= 4'55,5'08
anð' 6'25 kJmol-l
respectively, see Appendix
A2)
which could be partly responsible for the inadequacies of the calculated phase diagram. The effect of varying the enthalpy of either harnblende or bothtremolite
and,Fe-trennlite
tothefull
extent of thei¡ uncertainties is to cause the stable and metastableCFMASH
invariant points to converge, however, the shifts are insufficient to cause an inversionof
topology (Fig. 3.13).Staurolite
The targe uncertainties in the enthalpy values
for
the stauolite end-members suggest that staurolite may be a potential source for the inadequacies of the calculated chlorite-bearing grid (see AppendixA2). If
the leastwell
defined enthalpy, ÀH¡5¡, is adjusted to thefull
extentof
its uncertainty (by aHrst = -17 kJmol-l)
the GFMASH invariant points draw very close together (Fig.3.12).
However, in doing so, the predicted compositions of the phases became unrealistic,with
staurolite calculated as significantly more Fe-rich than garnet'A
moresatisfactory result was obtained by adjusting AFI¡5¡ and AH¡¡3¡ by equal' smaller amounts (e'g' aH¡5¡, aH¡¡15¡ =
-
15 kJmol-l).
This adjustment allows the topologyof
the calculated phase diagram to invert so that the [Hbl, Oam] and [St, Oam] CFMASH invariant points become metastablewith
respect to the remaining orthoamphibole-absent points, [Chl,Oam]'
[Crd'Oam], [Als, oam],
[Grt,oam]
(Fig.3.1a).
This topology would then allow theorthoamphibole-bea¡ing invariant points [Crd,
Chl],
[Grt,Chl]
and[Hbl' Chl]
to be stable' in agreementwith
the phase relations determined from natural rocks.Chapter3-CFMASII- 50
lAlsl
. original
r
AHrru= -10 kJ
mol-1o
AHtr-
ÂH1,= -10 kJ
mol'1tsrl lGrtl
tchrl
\ -\
a lHbrl
10I
I
7
L(d
-ol¿
È
6
5
550
600 650 700
T ("C)
Fisure 3.13. The effects of the uncertainty in the enthalpy o-f hornblende ÂH¡5 = -19
¡¡
ilå,Ïi";ä'od:Ïü* i-rol Ü
*"r_r õn the rocationïof the orthoamphibole-absent CFMASH univariant Points.Chapter3-CFMASII- 51
10
I
(Ú
-o l¿
fL
8
7
6
600 650 700 750
T ("C)
(s0 CFMASH
+An,Qtz,HzO
a
o?
,ã (st)
z q t
C,fl gtÀn lCrdl
sill
(Hbr)
%
7c oé
lchll
Hzo lGrtl
o4
[Als]
õ
o' oIJ
c Eo
g)
oô.¡
I o ('
F
Hbl,
An)(Hbl)
(S0 (Hbr)Chapter3-CFMASII- 52