American Mineralogist, Volume 77, pages 30-33, 1992
.C):
The fluxing effect of fluorine at magmatictemperatures(600-800
A scanningcalorimetric study
D. B. DrNcwELL, S. L. Wrsn
BayerischesGeoinstitut, Universit?it Bayreuth, Postfach l0 12 51, W-8580 Bayreuth, Germany
Ansrn-lcr
The effect ofF on the glasstransition behavior ofalbite, diopside, and four other silicate
melts has been investigated using scanning calorimetry. The addition of F to all silicate
melts investigatedresultsin a strong,nonlinear decreaseof the glasstransition temperature
(Z' as recorded by the peak temperaturesof heat capacity). The decreasesobserved extrapolate consistently to published fluoride glass transition temperatures.The largest Z,
decreaseis observed for albite-FrO-, melts (AT = 250 "C at 6 wto/oF). The effect of F is
similar to that previously observedfor HrO (Taniguchi, 1981).
Physical properties of low-temperature silicate liquids are a valuable constraint on lowtemperature petrogeneticprocessesin granite and pegmatite petrogenesis.Low-temperature wiscositiescan be estimated from the glasstransition data. These data are combined
with previously published high-temperature,concentric-cylinder viscosity data to obtain
a much more complete description of the temperature dependenceof viscosity for these
melts.
The present data, obtained on supercooledliquids close to the glasstransition, are of
special significancebecauseit is at the glass transition that silicate glass structures are
frozen. A separatemultinuclear NMR study of glassesquenchedfrom these experiments
has shown that the predominant coordination of F in albite glassis octahedralto Al. The
coordination state of F does not appear to be concentration dependent, and thus the
structural origin of the nonlinear Z, decreasedoes not arise from such a mechanism.
silicate melt. Generally this is done in one of the two ways
schematically in Figure l. A time-temperaThe formation of igneousrocks involves processesthat illustrated
(TTT) curve is drawn for the crysture-transformation
can only be fully understood through a complete descripsystem.Superimposedon the
of
a
single-phase
tallization
tion of the rheology of silicate melts. It has recently been
description of the glass
is
the
time-temperature
suggestedthat silicate melts behavepredominantly as re- diagram
composition yields
A
favorable
liquidus.
the
and
transition
laxed liquid phases during such petrogenetic processes
experimental time scales,lie outside
under
regions
that,
(Dingwell and Webb, 1989, 1990).Some exceptionsdo
the TTT crystallization curve, both below the liquidus
occur, and the extremely high rate of melt deformation
and above the glass transition. Experiments just below
during volcanic ash flow eruptions is an example of unliquidus temperature can take advantage of the finite
relaxedbehavior of the melt (Webb and Dingwell ,1990a, the
required for significant crystallization to provide data
time
1990b). A second possible case is that of rapid crystal
propertiesof supercooledliquids. Experimentsjust
growth in extremely low-temperature melts such as peg- on the
the glasstransition interval contribute to the unabove
matitic systems.The viscosity of a silicate melt is a neardata by avoiding significant crystallizadercooled-liquid
linear measure of the relaxation time, and it is at high
from below.
tion
viscosities and the corresponding high relaxation times
This study is of the latter type. We have measuredthe
that low-temperature processesof rhyolite, granite, and
heat
capacity of F-bearing silicate melts at temperatures
pegmatite petrogenesisoccur.
just above the glasstransition but below the TTT enveProviding a complete description of the rheology and
lope for our experimentaltime scale.The resulting calorithermochemistry of silicate melts requires the study of
metric
traces of the unrelaxed and relaxed silicate melts
undercooled liquids. An increasingnumber of studies of
provide information on the low-temperthis nature are being performed as a result (Neuville and are also used to
viscosity
of
these melts in a region that is difficult
Richet, l99l; Taniguchi, 1989; Webb and Dingwell, ature
with
viscosity measurementsbecauseof
directly
access
1990a, 1990b; Webb et al., 1992). Experiments on the to
during the longer excursiontimes
crystallization
incipient
behavior of supercooledliquids must avoid significant
required.
are
properties
7",
that
above
in
the
of
the
order to observe
crystallization
Ixrnonucrrox
00
0003-004x/92/0l 02-0030$02.
30
3l
DINGWELL AND WEBB: F IN SILICATE LIQUIDS
i% o i% <--
% crystallized
1.8
significantcrystallization
TTT envelope
u
F
t..
!
i
o
a
liq.iid
-Iqu-rd =
liquid
1.4
glass
;
o
(J
glass
RECIPROCAL
TEMPERATURE
600
Fig. l. Schematicrepresentationof the time scalesof structural relaxation, crystallization, and experiments on crystal-free
supercooledliquids such as the presentstudy.
(K)
TEMPERATURE
Fig. 2. Scanning calorimetric traces of the heat capacity of
glassyand liquid rnelts on the join diopside-F O ' cooled and
subsequently reheated at 5 K,/min.
Mnrrrots
The compositions investigated in this study represent
the substitution of FrO-, into base compositions of diopside (CaMgSi,Ou)and albite (NaAlSi3O8)as well as reconnaissanceof four other compositions in the system
Na-Al-Si-O-F. The glassesusedin the study originate from
the viscosity investigations of Dingwell et al. (1985) and
Dingwell (1989). Additionally, NaAlSi3O8 glassescontaining 1.6 and 3.40loF were producedby mixing powders
of the requisite proportions of albitic and albite + 5.80/o
F glassesand fusing at 1200 "C for 2 h in a Pt crucible.
The scanning calorimetry was performed using commercial instruments (Setaram DSCI I I and HTC). The
calibration ofboth instruments was carried out with AlrOr.
The heat capacity data are accurateto 2.5o/o(DSC instrument) and 50/o(HTC instrument), and the glasstransition
peak temperaturesare reproducible to +3 "C.
Rnsur,rs
The scanningcalorimetric traces of the diopside FrO-,
glassesare presentedin Figure 2. The effect ofthis extent
ofF substitution on the liquid and glassyheat capacities
is within the precision of these data (Fig. 2). The general
appearanceofthe transition is unchangedexcept for the
shift to lower temperatureswith F addition. The ratios of
peak height to liquid heat capacity and ofglass to liquid
heat capacity are not measurablyinfluencedby the F contents.
The effect of F content on the glass transition of diopside and albite melts is parameterizedin Figure 3 and
Table I in the form of peak temperature vs. F content.
The peak temperature is a strongly nonlinear function of
F content for both the diopside and albite melts.
ature of fluoride glassesat =600 K (Gan andZhang, 1990;
Koide et al., 1990). The trends of peak temperatures extrapolate consistently to the values obtained for pure fluoride melts.
A preliminary comparison of the effect of F and HrO
on the glass transition and low-temperature viscosity of
silicic melts can be made using the data of Taniguchi
(198 l) on wet rhyolites. The effects of HrO and F are
compared in Figure 4 on the basis of equivalent F and
H2O/2. The comparison indicates similar effects of F and
HrO on Z-. The slope of the decrease in Z, with HrO
860
e
820
LrJ
E
l
F
780
E.
]U
(L
I.JJ
F
I
ALETE
.
DIOPSIDE
740
700
:<
LtJ
(L
660
620
12
DrscussroN
In Figure 3 the composition dependenceof the peak
temperature for F-bearing albite and diopside are compared. The decreaseof ?", observed is consistent with
existing information placing the glasstransition temper-
16
FLUORINE(wt%)
Fig. 3. The efect of F on the glass transition of diopside and
albite melts as indicated by the peak temperature of glassescooled
and subsequently reheated at 5 K./min.
DINGWELL AND WEBB: F IN SILICATE LIQUIDS
JL
TABLE1.
Peak temperatures from scanning calorimetry
Composition
DI
DrF0.25
DIFl
AB
ABF1.6
ABF3.4
ABF6
JDF
NEF
AKF
ALUMF
Peaktemperature
(K)
1014
978
933
1131
1000
893
880
812
833
823
870
a
(E
g
F
6
o
C)
a
X " " .- U
.^
- I
o)
o
"t_oQ1i9P+^1!""
:' +
6 e o -t t r -
- o
o
o
tr
E
+
JDF
o
l€F
a
AKF
X
ALff
11
added is within error of that for F. This comparison indicates that HrO and F are equally effective in reducing
relaxation times in low-temperature silicic melts. The inset of Figure 4 is a comparison of the effects of F and
HrO on the viscosity of albite melt (Dingwell, 1987).
Similar trends are shown for the effectsof F and HrO in
melt viscosity at 1200"C, consistentwith the calorimetric
data from 600 to 800 "C.
The low-temperature viscosities of silicic melts containing F are ofdirect relevanceto rock-forming processes in granitic and pegmatitic environments. The effects
of F are likely to be much increasedat petrogenetictemperaturescompared with the superliquidusregime,where
data are available. Attempts to extrapolate high-temperature viscosity data to lower temperatures relevant to
granite and pegmatite petrogenesisare difficult because
the high-temperature dala are too restricted to account
for any potential non-Arrhenian behavior of the temperature dependenceof viscosity. Low-temperature data are
required.
Viscositiesof undercooledF-bearingmelts are however
difficult to measureusing dilatometric means becauseof
the mineralizing action of F in these melts. Attempts to
obtain such data using the frber elongation method indicated that the strength ofsuch fibers is strongly reduced
with addition of F (Dingwell and Webb, unpublished
1 0 4/ T K )
concentric-cylinder
of high-temperature,
Fig. 5. Comparison
viscositydataof Dingwellet al. (1985)andDingwell(1989)with
low-temperatureviscositydata derivedfrom the peaktempera(this study).
turesof scanningcalorimetricexperiments
data). The causeof such a reduction is probably imperfections causedby crystallization.
The time scales of shear and enthalpy relaxation in
silicate melts are closely related (Dingwell and Webb,
1989, 1990).Recentcomparisonsofshear viscositiesand
calorimetric glasstransition temperatures(Knoche et al.,
1992; Webb eL al., 1992) have confirmed that peak temperaturesfrom enthalpy relaxation representisokoms (i.e.,
constant vicosities) for silicate melts (e.g.,Richet, 1984).
Thus the peak temperaturesof this study yield low-temperature viscosity data for the investigatedcompositions.
Calibration of the peak temperaturesof melts along the
diopside-anorthitejoin by Knoche et al. (1992) confirms
the common assumption of constant viscosity at a consistently defined glass transition temperature for glasses
with the same thermal history. Knoche et al. (1992) established that glassescooled at 5 K./min and subsequently
reheatedat 5 K/min yielded calorimetric peak temperatures at a viscosityof I1.0 a 0.3 log'oPa s.
Low-temperatureviscositiesderived from the scanning
calorimetry are plotted together with the high-temperature data of Dingwell et al. (1985) and Dingwell (1989)
in Figure 5. The low-temperatureviscositiesobtained for
o
the alkali aluminosilicate melts indicate an even stronger
-78o
viscosity decreasewith the addition of F at low temperU
(E
ature than is the caseat superliquidus temperatwes. The
74o
P
peralkaline melt is strongly non-Arrhenian, and the perE
aluminous melt is Arrhenian. The albite, jadeite, and
H
7oo
nepheline F-bearing melts exhibit only slightly non-ArUJ
F
rhenian temperaturedependenceswith the result that the
o1
02
03
X/X+O, (X=FOH.NBO)
strong lowering of activation energy observed at high
temperatures(Dingwell et al., 1985) continues to lower
temperatures.The effect of F in low-temperature silicic
o246
melts is indeed much greaterthan that observedat higher
FLUORINE (vvt%)
temperatures.
The F-bearing diopsidic melts exhibit viscosity temFig. 4. Comparison of the effectsof HrO and F on the glass
transition in rhyolite (marked H'O) and albite melts, respec- perature relations that are subparallelto that ofdiopside
itself. The activation energyofviscous flow is not strongtively.
DINGWELL AND WEBB: F IN SILICATE LIQUIDS
ly affectedby the addition ofF to diopside at high or at
low temperatures.The effect of F on melt viscosity does
however increasewith decreasingtemperature.
The frozen structures of the samples investigated in
this study have been investigatedby a variety ofnuclear
magneticresonancemethods (Schalleret al., 1992).Those
results indicate that the predominant speciation of F is
to octahedral Al atoms (Manning et al., 1980) for the
2eSi,and
F-bearing albite samplesof this study. The 27,4.1,
'eF spectraof albite + 1.6,3.4, and 5.80/oF indicate no
concentration dependence of F speciation. Thus the
strongly nonlinear nature ofthe effectsofF on the glass
transition temperature (this study) and on viscosity
(Dingwell, 1987) do not appear to result from varying
proportions of different fluoride species.
CoNcr-usroN
The effectsof F on the glasstransition of silicate melts
is characterizedby a strong, nonlinear decreaseof peak
temperature in scanning calorimetric measurementsof
heat capacity. Low-temperature viscosity data estimated
from the peak temperaturesprovide a more complete description of the temperature dependenceof viscosity in
these melts. The observed effects of F and HrO on the
glass transition temperature are similar. The effects on
albite and diopside are qualitatively similar but quantitatively different. The much larger decreasein the activation energy of viscous flow in the albite melts compared with diopside melts may result from the contrasting
solution mechanisms of F in these melts, inferred from
NMR measurementson the quenchedglasses.
AcxNowleocMnNTS
We wish to thank Detlef KrauBe for microprobe analyses and Kurt
Klasinski for software development. Calorimerry has been supported by
a lribniz Preis to F Seifert. We thank D. Neuville and P. Richet for a
prepnnt.
RnrrnnNcns crrED
Dingwell, D.B. (1987) Melt viscosities in the systern NaAlSijOs-H,OFrO ,. In B.O. Mysen, Ed., Magmatic processes:Physicochemicalprinciples, p. 423-433. The GeochemicalSociety, University Park, Pennsylvania.
-
JJ
(1989) The effect of fluorine on the viscosity of diopside melt.
American Mineralogist, 74, 333-338.
Dingwell, D.B., and Webb, S.L. (1989) Structural relaxation in silicate
melts and non-Newtonian melt rheology in geologic processes.Physics
and Chemistry of Minerals, 16, 508-516.
(1990) Relaxation in silicate melts. EuropeanJournal of Mineralogy,2,427-449.
Dingwell, D.B., Scarfe,C.M., and Cronin, D.J. (1985) The efect of fluorine on the viscosities of melts in the system NarO-AlrOySiOt: Implications for phonolites, trachytes and rhyolites. American Mineralogist,
70, 80-87.
Gan, F., andZhang, H. (1990) Viscous behavior ofoxide and non-oxide
glasses.Glastechnische Berichte, 63K, 439-446.
Knoche, R., Dingwell, D.B., and Webb, S.L. (1992) Temperature-dependent thermal expansivities of silicate melts: The system anorthite-diopside. Geochimica et Cosmochimica Aaa, in press.
Koide, M., Matusita, IC, and Komatsu, T. (1990) Viscosiry of fluoride
glassesat glass transition temperature. Journal ofNon-Crystalline Solids.125.93-97.
Manning, D.A.C., Hamilton, D.L., Henderson, C.M.B., and Dempsey,
M.J. (1980) The probable occumenceofinterstitial Al in hydrous fluorine-bearing and fluorine-free aluminosilicate melts. Contributions to
Mineralogy and Petrology,75, 257-262.
Neuville, D., and Richet, P. (1991) Viscosity and mixing in molten (C4Mg)
plroxenes and garnets.Geochimica et Cosmochimica Aaa,55, l0l l1019.
Richet, P. (1984) Viscosity and configurationalentropy of silicate melts.
Geochimica et Cosmochimica Acta. 4E. 471-483.
Schaller,T., Dingwell, D.B., Keppler, H., Kniiller, W., Merwin, L., and
Sebald, A. (1992) Fluorine in silicate glasses:A multinuclear NMR
study. Geochimica et CosmochimicaActa, in press.
Taniguchi, H. (1981) Effects of water on the glass transformation temperature ofrhyolitic rock melt. Joumal ofthe JapaneseAssociation of
Mineralogy, Petrologyand Econornic Geology, 76, 49-57.
(1989) Densities of melts in the systemCaMgSi'Ou-CaAl,SirO,at
low and high pressues, and their structural significance. Contributions
to Mineralogy and Petrology, lO3,325-3t4.
Webb, S.L., and Dingwell, D.B. (1990a) The onset of non-Newtonian
rheology in silicate melts: A fiber elongation study. Physics and Chemistry of Minerals, 17,125-132.
(1990b) Non-Newtonian rheologyofigneous melts at high stresses
and strain rates: Experimental.rcsults for rhyolite, andesite, basalt and
nephelinite.Joumal of GeophysicalResearch,95, I 5695- I 570 l.
Webb, S.L-, Knoche, R., and Dingwell, D.B. (1992) Determination of
silicate liquid thermal expansivity using dilatomery and calorimetry.
European Journal of Mineralogy, in press.
Meuuscn-rrr REcETvED
Mw 2, l99l
MeNuscnrrr AccEprEDSrsrervrnen I 1. I99 I