Metastable crystallization products and metastable phase diagram of the glassy and supercooled aqueous ionic solutions of LiCl

Journal of Non-Crystalline Solids 104 (1988) 203-210 North-Holland, Amsterdam

203

M E T A S T A B L E CRYSTALLIZATION P R O D U C T S AND M E T A S T A B L E P H A S E DIAGRAM O F T H E GLASSY AND S U P E R C O O L E D A Q U E O U S IONIC S O L U T I O N S O F LiC! A. ELARBY-AOUIZERAT 1, J-F. JAL l, p. C H I E U X 2, J.M. LETOFFI~ 3 p. CLAUDY 3 and J. D U P U Y a I D~partement de Physique des Mat~riaux, Universit~ Claude-Bernard, 69622 Lyon Villeurbanne, France 2 Institut Laue-Langevin, 156X Avenue des Martyrs, 38042 Grenoble Cedex, France 3 Laboratoire de Thermochimie, I N S A , 69622 Lyon Villeurbanne, France

Received 8 March 1988

The metastability towards crystallization of glassy and supercooled aqueous ionic solutions of LiCk has been explored as a function of temperature and concentration by a combination of differential scanning calorimetry and neutron scattering experiments. A metastable phase diagram obtained under well specified sample thermal treatment has allowed us to define two narrow concentration ranges (or deep wells) for good glass formation. A good description of the metastable phase diagram is given by the superposition of two binary diagrams, H20/LiCI 3H20 and LiCI 5H20/LiC1 2H20. The metastable crystallization end products such as LiCI 2H20 have been identified by their diffraction pattern which has been measured for the first time. In particular the metastable form of ice, cubic ice I¢, could be produced in a controlled manner and its transformation to ice lh accurately followed.

1. Introduction

The concentrated aqueous ionic solutions easily form glasses on cooling to liquid nitrogen and their glass forming ability has been extensively investigated in comparison to other glass formers and in particular to high temperature ionic glasses [1]. Indeed, depending on the cationic charge they offer a large concentration range (e.g. from about the solubility limit to R - 10, 20, 30 for monovalent, divalent or trivalent chorides) (with R = (mol H 2 0 / m o l salt) where the glassy state can be obtained. The correlation between the glass transition temperature, Tg, and the formal charge of the cation or the basicity of the anion has been pointed out and a thermodynamic description for the concentration dependence of Tg has been given [2]. More recently several studies have been made to characterize the dynamical properties of these solutions at the approach of Tg or in the glassy state [3,4]. However, many fundamental questions remain to be solved. The ease of formation of the glassy state has a counterpart. It is often difficult to 0022-3093/88/$03.50 © Elsevier Science Publishers B.V. (North-Holland Physics Publishing Division)

reach thermodynamic equilibrium and there are regions of the phase diagram where the properties are dominated by kinetic problems on the time scale of hours or days. Therefore, it is helpful to introduce new descriptions such as non-equilibrium or metastable phase diagrams drawn at an appropriate constant cooling and warming rate

[5]. Moreover, depending on the time spent in the metastable regions, the nucleation and growth of various stable or unstable precipitates or compounds might be controlled. It becomes necessary to always complement the macroscopic studies with a microscopic structural characterization. And it should be noted that although these systems abound in hydrates, very few structural determinations have been performed up to now. This is a pity since they would be very helpful in order to build up models of the local order and could act as reference structures to understand the disordering process. Let us take the example of the LiCl H 2 0 solutions. This system has been well investigated and a considerable effort has been made to accurately

204

A. Elarby-Aouizerat et al. / Metastable crystallization products

T(*C) 1oo !

93.6

T4

6O T3 -

2O T2

-20

19.1 2. Experimental conditions

-2o.5

-60

20975

3

lisation at the most dilute concentrations (6 < R < 12) of the supercooled liquid. Then we will identify the structure of the hydrates and comment on the sequence of hydrate formation from the metastable supercooled liquid at high salt concentration.

2~

1~"

Fig. 1. LiCl RH20 phase diagram. (E = eutectic point; T i = transition temperature and corresponding peritectic point).

determine its equilibrium phase diagram, a critical study being represented [6] in fig. 1. The eutectic point (E), the vertical lines representing the various hydrates, the isothermal horizontal lines where transformation between hydrates occur, and the peritectic points (Ti) are clearly noticeable. Interestingly, except for the dihydrate, only odd hydrates (1, 3, 5) are formed and the monohydrate structure is the only one to have been studied. [7]. Of course, for some concentrations, on cooling the liquid (to T < 140 K) a glass is formed and once made, it is interesting to explore its stability towards crystallisation. We shall concentrate our effort on so exploring the glass or supercooled liquid metastability regions of the phase diagram and analysing the crystallisation mechanism and its end-products. A new picture of this system will then emerge which we believe is of significance for the understanding of glass formation and should help in the optimisation of further experimental work. This study has been performed using an interactive combination of differential scanning calorimetry (DCS) for the detection of the thermal effects and the optimisation of the sample thermal treatment and neutron scattering at large and small angles for the identification of the sample microstructure under controlled thermal history. After a brief description of the experimental techniques we shall present some results on ice crystal-

Quantitative differential scanning calorimetry experiments were performed with a M E T T L E R TA 2000 B heat flow (Ar gas) apparatus controlled by a H.P.85 microcomputer, with complete automation of the data analysis using a simplified electrical representation of the calorimeter which had been calibrated independently [8]. The temperature range covered is 100 K < T < RT or above, the temperature being measured with a precision of 0.2 K. A calorimetric signal is obtained with a sensitivity of about 3 #W at room temperature, i.e. for a typical sample an accuracy of 2% for the enthalpy change AH, and 5% for the heat capacity Ce. Weighted samples of about 40 mg were inserted in aluminium cups and sealed. They were then quenched to liquid N 2 at an average cooling rate of 2 K / s and introduced into the calorimeter precooled to 120 K. The same cooling rate and procedure were maintained for all concentrations. It allowed us to fully vitrify a LiC1 12H20 concentration corresponding to the complete hydration of Li ÷ and CI- ions [9], but did not prevent the partial crystallisation of more dilure samples. The warming rate was normally kept at 3.3 × 10 -2 K / s , the apparent optimum value to observe non-equilibrium effects, but sometimes also was chosen at a slower rate of 3 × 10 -3 K / s . Neutron scattering experiments were performed at the ILL (Grenoble) on various diffractometers D2, DIA, D1B, with wavelengths 1.3 ~<~ ~< 2.5 ,/k, i.e. allowing us to vary the resolution and the incoming beam flux as needed. The small angle neutron scattering (SANS) experiments were performed on D17 and D l l at various momentum transfer (k) ranges (k = (4~r/~) sin0; 20 is the scattering angle). The cylindrical samples for diffraction were contained in 8 mm diameter, 50 mm

A. Elarby-Aouizerat et al. / Metastable crystallization products

high, tight vanadium cells of 0.1 m m wall thickness; for SANS, 1 to 2 m m thick flat samples were used and contained in sapphire cells with 1 m m thick, 20 m m diameter windows. Temperature was controlled with various I L L orange cryostats with a precision and stability of 0.1 K. The samples were cooled in liquid N 2 inside the cryostat tail with a cooling rate of - 2 K / s . The excess of liquid N 2 was generally boiled off when the sample temperature became less than 100 K, in order to prevent cracks developing in the sample, although these could be healed by subsequent warming. Instead of using a constant warming rate as in the DSC experiments most of the neutron scattering measurements were performed by step warming, the temperature being kept fixed for the time necessary for at least one spectrum acquisition (i.e. from 20 min to an hour depending on the type of scan and machine). The M E R C K (suprapur) LiC1 salt was dried out at 130 ° C for 24 h before weighing and dissolved in weighed amounts of triply distilled and deionized water. In some occasions the samples were also prepared with the monohydrate, or by a salt extraction technique with organic solvents [10]; sample filtration was performed before SANS measurements. No difference induced by the preparative method could ever be detected in the diffraction or DSC measurements. All neutron scattering experiments were made with 99.7% D20 solvent in order to prevent the strong incoherent scattering from the hydrogen atom.

3. Experimental results 3.1. Nucleation and growth o f ice f o r concentrations R >~ 6. Comments on the ice I c structure

The cooling rate chosen for our experiments is such that for the concentration R = 12, the LiC1 solution can be fully vitrified. Following Samo'flov [11], this concentration corresponds to full ionic hydration of the salt at RT, with 4 H 2 0 around Li ÷ and 8 H 2 0 around C1-. With more dilute solutions (R > 12), ice precipitates while cooling and we obtain a mixture of ice crystals and a glass

205

whose composition is given by the intersection of the extrapolated ice branch of the liquidus and the T~ line ( - 142 K), i.e. a glass of composition LiC1 6 H 2 0 [12]. In the range 12 ~< R ~< 6, well vitrified samples are prepared which on warming above T~ become unstable towards ice nucleation and growth. A careful combination of DSC and neutron scattering experiments has allowed us to define the appropriate concentrations and thermal treatments for which nucleation and growth processes could be investigated at the time scale of our measurements. It led us to discover that ice I c, which is the metastable cubic form of ice, could be produced a few degrees above T~ after an appropriate annealing of the aqueous ionic glass at temperatures around Tg. At higher temperature the cubic diffraction pattern continuously transforms into the stable hexagonal pattern of ice (150 K < T < 160 K). The concentrations LiC1 9D20 and LiC1 8.5D20 are the most appropriate for a detailed investigation of this unstable form of ice at a time scale of about 30 min per diffraction spectrum. At these concentrations, annealing of the sample near Tg is not mandatory and slow warming is sufficient [13] to detect ice Ic, although, of course, the stability of this phase and the kinetics of its transformation to I h depend on sample thermal history. We present in fig. 2 one example of a series of diffraction spectra obtained on a LiC1 9D20 glass at different stages of the thermal treatment. Except for a very slight shift of the main peak position (effect of temperature) the spectrum of fig. 2a is nearly identical to that of liquid LiC1 9D20 [14]; we have a well disordered system. At 151 K (figs. 2b, c) a crystalline pattern appears, which is characteristic of ice I~, a diamond structure of space group (fd3m). At the same time, the disordered structure is modified: (i) its intensity drops as the glassy matrix is depleted due to ice precipitation; (ii) there is a strong shift of its main peak position towards larger m o m e n t u m transfer values; (iii) the disordered pattern progressively tends to the LiC1 6D20 structure. As a matter o f fact, the I~ structure which we obtain superimposed on the glassy pattern is never a perfect cubic structure, If we look in more detail at the main (111) ice I~ peak, we see that it is strongly

A. Elarby-Aouizerat et aL / Metastable crystallization products

206

I 1.0r-

~_~ I~ ~

Z ~@

LiCl 9 I~O t (~,~2a ~)

J 1

0.5i i~, ,

a) '~'~i

1.0~

LiCl 9 D20 "

i _

0

2

g) 2o3K

'

(15

0~

i f) 160K(3h)I

02 ,i •

,~,~,

~'! k (h,-;)

0

!I

i

2

e) ~8OK

4 k(A -~)

6

i 8

Fig. 2. Neutron diffraction investigation of the precipitation of ice from a quenched LiCI 9D20 solution (A =1.22 ,~, instrument D2) (a) glassy sample as quenched; (b) and (c) precipitation of ice I¢; (d), (e) and (f) transformation of ice I c in ice Ih; (g) high temperature I h pattern.

asymmetric. On warming, this asymmetry will develop and progressively produce a deformed triplet pattern (fig. 2d) as if a fraction of the sample was crystallizing into ice Ih, the stable hexagonal form of ice. However, it is impossible to describe the deformation of ice Ic by a combination of the two pure forms of ice. Another approach is necessary. We observed that the deformation of the 111 peak of I~ ice could be described [15] by the appearance of a strongly asymmetric (tail towards the high angles) 100 peak of ice I h. On subsequent warming the 101 and 102 peaks appear as a modulation of the 100 tail, and finally the asymmetric character of the whole 101 series disappears (see figs. 2c, d, e and fig. 3). This type of behaviour is similar to the progressive correlations between bidimensional structures [16] as if sheets of I h ice were forming in I c and progressively developing their interactions as they multiplied. Although other defects might be necessary to fully describe the "cubic" ice structure [17] a quantitative description of the deformed ice Ic patterns has been

made on this basis and has shown the existence of a supplementary disordered ice phase which exists until completion of the I c --* I h transformation [18]. In any case, although the first stages of the Ic ice formation could be fully controlled by varying the thermal treatment and have been produced for several concentrations (as well as several salts, e.g. BeC12) we could never obtain a pure I c pattern free of an asymmetric deformation of the (111) peak. Given the deformation of the I¢ ice structure, it is not easy to characterise the beginning of the I~ ~ I h transformation. The best criterion is probably the emergence of a sharp (102) I h peak; a sharp (100) (002) (101) triplet I h pattern is also significant. Of course, good experimental resolution is necessary to follow the phenomenon (see fig. 3). We see in figs. 2d, e, f, g that the process is quite slow and extends over several degrees which allows examination of the intermediate steps. At the end of the process, the I h pattern obtained after 3 h at 160 K (fig. 2f) is still faulty. The correlation between the volumic fraction of I~ ~ I h transformation and the average size of the crystallites has been determined from SANS studies [13] and we should keep in mind that cubic crystallites have sizes around 200 ~,. When larger sizes are observed they are correlated to the cubic to hexagonal transformation. In summary, the glassy and supercooled aqueous ionic solutions allow us to produce, on a

I O.5

.go4

, i

LiCI 9 D20 A • 2.52

I

~o3 t

J

50

60

70 (2 e)

Fig. 3. High resolution spectra (k = 2.52 ,~, instrument D1B) of the first stages of the modification of ice I~ into ice I h.

A. Elarby-A ouizerat et al. / Metastable crystallization products

convenient time scale, all the intermediate steps for nucleation and growth of ice in its metastable I~ form as well as its progressive transformation to ice I h" To be more specific, by varying temperature, concentration or salt, it is possible to define and study three different states: (i) the supercooled liquid above Tg before ice crystals are detected, (ii) the faulty cubic ice structure, (iii) the cubic to hexagonal ice transformation. The dependence of the kinetics of ice nucleation or transformation on various parameters such as annealing or ionic strength could therefore be investigated at great length. And we are already asking ourselves several questions, such as (i) how to interpret the scattering intensity changes observed in the disordered patterns at the approach of ice precipitation [19], (ii) how to reconcile the observation of an easily identifiable diffraction pattern for I¢ ice subsequently transforming to I h with the DSC consistent finding of a single exothermic crystallization signal [20] ?

3.2. Crystallization of the hydrates under equilibrium conditions Several hydrates have been found in the LiC1. H 2 0 system. They were obtained under the equilibrium conditions summarized in the phase diagram given in fig. 1. Their characteristic diffraction patterns are given in figs. 4, 5 and 6. These patterns were obtained from samples of respective compositions LiC1 5D20 and LiCI 3D20 quenched to liquid N 2 and warmed up to the quoted temper-

207

7

LiCI 5D20

[,i 1 T=190K

6

4

21-

I

,

i

,

1

i

,

2

i

,

3

i

4

K(,&q)

Fig. 5. Neutron diffractionpattern of LiC15D20 a t 180 K. ature. The LiC1 2D20 structure was obtained at 191 K as a first crystallization of the quenched LiC1 3D20 solution and at 251 K on melting the trihydrate, in both cases the excess liquid structure is very apparent. Details of the peak positions are given in table 1. These patterns are sufficiently characteristic to permit an easy identification of the compounds.

3.3 Identification of the non equilibrium disordered and crystalline phases In the general case, DSC experiments on glassy aqueous ionic solutions prepared, as usual, by

i

'

i LiCI 3 D20 T=225K

il,,

3 3

7"

T=191 K

1-

~

,! T,201K

i

>.U.I

t i

I

r

i

K(R')

Fig. 4. Neutron diffraction pattern of LiCI 3D20 at 221 K.

L~ 2 3 k(,~"1) Fig. 6. Neutron diffraction pattern of LiC1 2D20 obtained (a) in a metastable form at 191 K from the partial crystallization of a quenched LiC1 3D20 solution; (b) in a stable form at 261 K from the decomposition of the LiCI 3D20 crystals obtained at 221 K (see fig. 4) from the same solution. 1

A. Elarby-Aouizerat et al. / Metastable crystallization products

208

Table 1 The intensities are given in the same relative units for LiC1 5 D 2 0 and LiCI 3D20. The intensities of LiCI 2D20 should be increased by a factor 1.3 to correct for the fraction of the sample which is in the glassy state LiCI 5 D 2 0

LiCI 3 D 2 0

k ( k -1)

1 k (T = 190 K)

0.92 1.12 1.30 1.47 1.53 1.81 1.87 2.06 2.21 2.39 2.48 2.68 2.77 2.9l 2.97 3.05 3.17 3.28 3.39 3.57 3.67 3.83 4.06 4.09 4.15 4.32 4.40 4.54 4.66 4.81 5.04 5.10

100 100 120 100 220 100 190 230 6300 920 760 740 170 80 80 85 240 100 40 320 150 210 280 190 180 210 70

1.06

LiCI 2 D 2 0

I k (T = 225 K) 650

2.06 100 2.13 2030 2.23 2120 2.38 2430 2.53 220 2.66 380 2.82 170

3.07 3.18 3.31 3.48

650 95 85 560

3.90

240

4.13 4.28 4.43 4.57

850 180 185 250

4.83 4.92

250 160

I (T = 191 K)

1.13 1.32

70 180

1.83

300

2.10 20 2.24 1350 2.34 960 2.46 160 2.55 60 2.73 420 2.97

40

3.35 3.50

240 60

3.70 3.80 3.92

40 40 140

4.15 4.39

160 50

4.52

50

210 K) are detected on the warming curves. In order to characterize these effects, the diffraction spectra of a LiCI 3D20 sample were collected at several temperatures and we found that for T = 191 K the material crystallizes as LiCI 2D20 with an excess disordered structure. Then on warming the dihydrate and the disordered phase react to form the trihydrate shown in fig. 5 which was obtainded above 221 K. Finally, at the peritectic temperature ( T - 2 5 3 K) the trihydrate decomposes again into dihydrate and liquid according to the equilibrium phase diagram. We give in fig. 6 the comparison between the diffraction spectra obtained at 191 K and 261 K. They are, qualitatively speaking, identical. We can therefore conclude that, at the composition LiC1 3H20, there is on warming from Tg, an initial metastable crystallization of the spectrum as LiC1 2H20. In this metastable state, the liquidus branch of LiC1 2 H 2 0 should be extended to low temperature as shown by a dashed line in fig. 7, suppressing the LiC1 3H20 peritectic. A similar metastable extension of the ice branch of the liquidus line to low temperature has already been quoted above for the concentrations R < 6. In that case a detailed description of the metastable fiquidus line shape was obtained with an equilibrium between ice and a L i C 1 . 6 H 2 0 solution.

T'K'I LiCI R

70

373 333

253

liquid quench, give a characteristic glass transtion temperature, Tg, and an exothermic effect due to sample crystallization (at To). Both Tg and T¢ measured at a chosen standard warming rate (2 K / m n ) can be drawn on the equilibrium phase diagram (see fig. 7), although the states below T~ are non-equilibrium metastable ones. A study of fig. 7 reveals several interesting points which we shall develop. At concentrations around LiCI 3H20, two exothermic effects (for T - 190K and

213 173 133

12

lb

5

3

2'0 3'0

2

R

4'o

"

mole '~ Li CI

Fig. 7. Metastable phase diagram of LiCl H 2 0 (see text). Values for Tg(O), crystallization of ice (D), of LiC! 5 H 2 0 (O), of LiCl 3 H 2 0 ( 0 ) , and of LiCI 2 H 2 0 ( × ) are quoted. The dashed lines were obtained as described in the paper.

A. Elarby-Aouizerat et al, / Metastable crystallization products

k'iI 2.0

/

,.*"

/ / /

1.9 1.8

-lb

1~-

2'0

2'5 "

mole ~o Li CI

Fig. 8. Concentration dependence of the main peak position of the structure factor in the liquid and in the glassy state.

This was confirmed by the diffraction studies since the determination of the composition dependence of the main peak position of the structure factor in the liquid and glassy state (see fig. 8) allowed us to estimate the composition of the disordered phase at equilibrium with the ice crystals. These estimates are reported in fig. 7 (open circles) and agree with the extrapolation of the ice branch of the liquidus line to low temperature made on the basis of thermodynamics. Going back to LiC1 3D20, we observe very little difference between the main peak position of the liquid structure factor in equilibrium with LiC1 2D20 at 261 K and that of the disordered phase structure factor at 191 K. The metastable extrapolation of the LiC1 2D20 liquidus line to low temperature should be very steep indeed. If we observe the composition dependence of the crystallization temperature Tc (fig. 7) we see two maxima around the values LiC1 6 H 2 0 and LiCI 4H20. As a matter of fact, these maxima are not measurable at the chosen warming rate of 2 Kmin 1, the only detectable thermal effect in the DSC experiments made at those two concentrations being Tg, the samples remaining always in the disordered state. Therefore we conclude the existence, in the phase diagram, of two narrow wells or windows for glass formation which are both at the composition of a missing even hydrate of LiC1 H 2 0 (R = 6 or 4). These windows are defined by the extrapolations of the ice branch and LiC1 2 H 2 0 branch of the liquidus which we have already introduced, and also by two other

209

dashed lines which we have drawn on either side of the LiC1 5 H 2 0 hydrate such as to contain the observed values of Tc around R = 5 which have a well-defined minimum at the pentahydrate composition. Diffraction studies at concentrations around R = 5, to confirm those non-equilibrium lines from the structure of the disordered phases coexisting with the pentahydrate have not yet been made but one of these lines seems to be the extrapolation to low temperature of the LiC1 3H20 branch of the liquidus. Fig. 7 with its two windows for "good glass" formation, is therefore a convenient summary of our observations on the metastable phases. However we should keep in mind that the Tg and Tc values have been obtained within a well-defined procedure (quench to hquid N:, followed by warming at 2 Kmin-1). A series of non-equilibrium phase diagrams at different warming rates should be produced to cover every experimental situation. It is sufficient for our purpose to know that the compositions of LiCI 6 H 2 0 and 4 H 2 0 are the best for preparing L i C 1 - H 2 0 glasses which can be kept for hours in the supercooled state without crystallization. We should add, that, generally speaking, the crystallization of metastable structures is favoured by slow warming rates or, better, by annealing at low temperature [19] ( T < Tg). The concentration dependence of Tg and in particular the range 5 > R > 3 for which there are two apparent Tg values have been analyzed in terms of competing hydrate-like structures. Finally, the effect of deuteration on Tg and T~ for R > 6 is about the same as for the ice branch of the liquidus which is raised by 2-3 degrees by deuteration. It has not been fully investigated for R<6.

4. Conclusion

Most of the results which have been presented up to now do not require further discussion. A few qualitative comments can however be made about the non-equilibrium phase diagram (fig. 7). The metastable LiC1 RI-I20 system can be described as the superposition of two binary diagrams, (i) H 2 0 / L i C 1 3 H 2 0 with a deep eutectic at the c o m -

210

A. Elarby-Aouizerat et al. / Metastable crystallization products

position LiCI 6 H 2 0 , (ii) LiC1 5H20/LiC1 2 H 2 0 with a deep eutectic at the composition LiC14H20. Although such deep eutectics are probably due to a competition between the various hydrates, the reason why they are at the composition of missing even hydrates remains to be understood at a structural level. In particular a detailed comparison of the structure and properties of the two glasses of composition LiC1 6 H 2 0 and LiC1 4 H 2 0 should be quite interesting. As far as the hydration is concerned, we have seen that for R > 12, ice crystallizes on quenching, which has been interpreted by the completion of the hydration shell of both ions for R = 12. For 12 > R > 6, the value of Tg remains constant and nucleation of ice is obtained under appropriate thermal treatment *. We therefore believe that for these concentrations, the water in excess of the value LiCl 6 H 2 0 is weakly bound. At R = 6 all the water molecules are probably shared by the hydration shells of both ions, while for R = 5 some LiC1 ion pairs could already be formed. Obviously there is a real need for a complete atomic structure determination of the hydrates (and this for various salts) in order to obtain models of the local order. Of course, an isotopic substitution technique to obtain the coordination number and structure of the hydration shell of Li + and C1- could be performed directly in the supercooled [21] or glassy state but this is not an easy experiment. Systematic studies of non-equilibrium phase diagrams of various aqueous ionic solutions should help us to understand where to expect the "good glass" windows. It should also help to focus ourselves on the key structural questions related to the deep eutectic formation and will allow us to find out what are the most appropriate systems to be investigated. Finally at the concentrations where ice can be nucleated it is worthwhile studying the effect of ionic strength on the nucleation and phase transformation of ice. * We note that on some occasions, the precipitation of ice can be complemented at T - 170 K by a precipitation of LiC1 5H20 such as to fulfil the thermal equilibrium concitions.

References [1l C.A. Angell, J. Phys. Chem. 69 (1965) 2137; C.T. Moynihun, J. Phys. Chem. 70 (1966) 3399; C.A. Angell and L.M. Torell, J. Chem. Phys. 78 (2) (1983) 937. [2] C.A. Angell and E.J. Sure, J. Chem. Phys 52 (1970) 1058. [3] I.A. Baianu, N. Boden, D. Lightowlers and M. Mortimer, Chem. Phys. Lett. 54 (1978) 169; N. Boden and M. Mortimer, J. Chem. Soc. Faraday Trans. II, 74 (1978) 353; E.W. Lung and H.D. Liidemann, Ber. Bunsenges Phys. Chem. 89 (1985) 508. [4] P. Carmona, A. Elarby-Aouizerat, J.F. Jal, J. Dupuy, J.-A. Serughetti, J.M. Letoffr, P. Claudy, M.C. Bellissent-Funel and P. Chieux, Proc. 6th Int. Conf. on the Physics of Non-Cryst. Solids, Kyoto, Japan 1987, ed. S. Sakka, J. Non-Cryst. Solids 95&96 (1987) 1009; P. Carmona, A. Elarby-Aouizerat, J.F. Jal, J. Dupuy, P. Chieux and J. Dianoux, to be published. [5] J. Dupuy, A. Elarby-Aouizerat, J.F. Jal, P. Chieux, P. Claudy and J.M. Letoffr, Rivista Della Stazione Sperimentale del Vitro 5 (1984) 63; P. Chieux, in: Physics and Chemistry of Aqueous Ionic Solutions, NATO ASI series C205. eds. M.C. Bellissent-Funel and G.W. Neilson, (Reidel, Dordrecht, (1987) p. 359. [6] R. Cohen-Adad and J. Lorimer, in: Solubility data (Pergamon, New York) to be published. [7] A.S.T.M. file no. 22-1142, Weiss et al., Chem. Bet., 102 (1969) 632. [8] P. Claudy, J.C. Commer~on and J.M. Letoffr, Thermochim. Acta 68 (1983) 305; 68 (1983) 317. [9] A.H. Narten, F. Vaslow and H.A. Levy, J. Chem. Phys. 58 (11) (1973) 5017. [10] C.C. Lynch, J. Phys. Chem 46 (1942) 366. [11] O. Ya Sami'/lov, Dokl. Akad. Nauk. SSSR 81 (1951) 641; Izvest. Akad. Nauk. SSSR Khim. 3 (1952) 398. [12] P. Claudy, J.M. Letoffr, J.J. Counioux and R. CohenAdad, J. Thermal Anal. 29 (1984) 423. [13] A. Elarby-Aouizerat, J.F. Jal, C. Ferradou, J. Dupuy, P. Chieux and A. Wright, J. Phys. Chem. 87 (1983) 4170. [14] A. Elarby-Aouizerat, Thrse no. 1171 Universit6 Claude Bernard, Lyon I (1982). [15] A. Elarby-Aouizerat, J.F. Jal, J. Dupuy, H. Schildberg and P. Chieux, J. de Phys. Coll. Suppl. 3 (48) (1987) C1-465. [16] B. Crozet, Thrse Universit6 Aix-Marseille II (1983). [17] W.F. Kuhs, D.V. Bliss and J.L. Finley, J. Phys. Coll. Suppl. 3 (48) (1987). [18] J. Dupuy, A. Elarby-Aouizerat, P. Claudy, J.F. Jal, J.M. Letoff~ and P. Chieux, in: The Physics and Chemistry of Aqueous Ionic Solutions, eds. M.C. Bellissent-Funel and G.W. Neilson (Reidel, Dordrecht, 1987) p. 447. [19] A. Elarby-Aouizerat, J.F. Jal, P. Chieux, A. Wright and R. Parreins, J. de Phys. 43 (1982) L355. [20] P. Carmona, A. Elarby-Aouizerat, J.F. Jal, P. Chieux, P. Claudy and J. Dupuy, to be published. [21] S. Cummings, J.E. Enderby, G.W. Neilson, J.R. Newsome, R.A. Howe, W.S. Howells and A.K. Soper, Nature 287 (1980) 714.