STRUCTURES OF RARE-EARTH HALIDE MELTS - Handbook on the Physics and Chemistry of Rare Earths

the existence of complex ions as well as for estimating their stabilities. The final decision concerning the formation of complicated species such as complex ions, clustered ions, molecules, and so on in the melts should be made according to direct observations of near-neighbor structures, namely, diffraction works and Raman spectroscopic measurements as will be described later.

As mentioned inSections 5 and 6, some differences which could not be dis- regarded are really found even in the crystal structures of rare-earth halides and in their densities in solid and molten states. Accordingly, local or short- range structures of rare-earth halides have to be clarified in the first place.

Up to now, numerous works have been published which report the structure of aqueous solutions containing rare-earth ions, but the number of works on the melt structures is quite limited to halides. As for the techniques to investigate the local structure of melt, potential tools can be divided broadly into three categories. The first is comprised of diffraction methods such as XRD

TABLE 17 Density of Molten NaCl–CeCl3System Versus Composition and Temperature (Sato and Yamamura, 1992)

CeCl3(mol%) rm(g cm³) Trange (K)

12.0 2.564–6.04710⁴T 1023–1134

24.5 2.945–6.82610⁴T 910–1137

37.0 3.276–7.62510⁴T 986–1138

50.0 3.508–7.76410⁴T 989–1143

61.5 3.715–8.06710⁴T 1010–1131

73.0 3.875–8.23710⁴T 1019–1136

85.0 3.972–8.00110⁴T 1042–1135

100.0 4.012–7.18310⁴T 1046–1145

TABLE 16 Density of Molten CsCl–LaCl3System Versus Composition and Temperature (Kim et al., 1987)

LaCl3(mol%) rm(g cm³) Trange (K)

12.0 3.708–9.72610⁴T 908–1213

25.0 3.714–9.37710⁴T 1072–1219

40.0 3.753–9.06410⁴T 1015–1217

51.5 3.849–9.19810⁴T 875–1208

64.0 3.918–8.91110⁴T 1148–1215

77.5 4.053–9.10710⁴T 1090–1112

90.5 4.127–8.82610⁴T 1113–1217

and ND, the second is Raman spectroscopy, and the third gathers X-ray absorption fine structure (XAFS) and computer-aided calculations. Hereafter, the structures of rare-earth halide melts obtained by the three categories of methods are described in due order.

7.1. Structures from Diffraction Experiments

Structural analyses have been continuously carried out since the 1960s so as to comprehend physicochemical properties of molten salts. At the beginning, laboratory-scale X-ray diffractometers and less stable neutron source-detector systems were, however, compelled to be used at the sacrifice of data accuracy.

Since electronics, mechanical devices, and the detector systems have gradually and continuously been developed, this has resulted in substantial improvements

TABLE 19 Density of Molten NaCl–SmCl3System Versus Composition and Temperature (Sato and Yamamura, 1992)

SmCl3(mol%) rm(g cm³) Trange (K)

11.5 2.597–6.22710⁴T 1002–1147

23.0 2.959–6.79010⁴T 915–1148

34.5 3.273–7.37910⁴T 915–1157

51.5 3.639–7.92910⁴T 919–1155

58.5 3.773–8.06210⁴T 905–1152

73.0 3.983–8.18810⁴T 935–1146

90.5 4.154–7.85410⁴T 917–1155

100.0 4.205–7.47210⁴T 925–1146

TABLE 18 Density of Molten NaCl–NdCl3System Versus Composition and Temperature (Janz, 1988)

NdCl3(mol%) rm(g cm³) Trange (K)

12.5 2.522–5.3710⁴T 1131–1284

26.2 3.032–6.9210⁴T 1093–1283

39.3 3.351–7.2010⁴T 1016–1283

55.3 3.703–7.8310⁴T 1023–1282

68.5 3.930–8.5710⁴T 1005–1281

85.0 4.120–8.4310⁴T 1021–1273

Handbook on the Physics and Chemistry of Rare Earths 120

in the apparatuses. Examples of new and performing setups are the high-energy monochromatic X-ray beam line with a diffractometer at SPring-8 (synchrotron radiation facility “Super Photon ring-8 GeV”) and the pulsed neutron sources with high-precision detectors at KEK (High Energy Accelerator Research Orga- nization) or J-Parc (Japan Proton Accelerator Research Complex) in Japan.

Therefore, accurate measurements of the melt structures are now within feasibility. Similar apparatuses are also operating at ESRF and ILL (Institut Laue–Langevin) in France, APS (advanced photon source) and SNS (spallation neutron source) in the United States, ISIS (centre for research at the Rutherford Appleton Laboratory; pulsed neutron and muon source) in the United Kingdom, and in several other countries.

Pioneering works on the structures of molten rare-earth chlorides have been done bySaboungi et al. (1991)andMochinaga et al. (1991)in the early 1990s. The former researchers have reported a ND study of the liquid structure of YCl₃ and combined the structural data with macroscopic melting and transport data to compare the behavior of this molten salt with those of

TABLE 20 Density of Molten LiF–YF3System Versus Composition and Temperature (Janz, 1988)

YF3(mol%) rm(g cm³) Trange (K)

0.0 2.074–3.32110⁴T 1140–1340

19.0 3.254–5.59510⁴T 980–1340

30.0 3.695–6.67210⁴T 1080–1340

40.0 3.902–6.15410⁴T 1080–1340

50.0 4.287–7.21610⁴T 1120–1340

60.0 4.174–5.06510⁴T 1220–1340

TABLE 21 Density of Molten LiF–LaF3System Versus Composition and Temperature (Janz, 1988)

LaF3(mol%) rm(g cm³) Trange (K)

5.0 2.491–3.64810⁴T 1140–1350

10.0 2.799–3.11810⁴T 1180–1350

15.0 3.500–6.65810⁴T 1110–1350

20.0 3.737–6.48610⁴T 1130–1350

25.0 3.799–5.50710⁴T 1130–1350

SrCl₂, ZnCl₂, and AlCl₃ as prototypes of different melting mechanisms for ionic materials. DyCl₃, HoCl₃, and ErCl₃ melts have very similar structure to YCl₃, thus confirming the earlier suggestion of a similarity of structural behavior for these systems in the liquid phase. A novel melting mechanism for trivalent metal chlorides, leading to a loose disordered network of edge- sharing octahedral units in the liquid phase, has been thereby established.

The various melting behaviors were related to bonding character with the help of Pettifor’s phenomenological chemical scale (Pettifor, 1986). These findings were in good agreement with Papatheodorou’s suggestion obtained from an extensive Raman scattering study of YCl₃that the structure of the melt may be a network of distorted chlorine-sharing octahedra (Papatheodorou, 1977).

In the first place, this suggestion emerged from a comparison of the Raman spectra of pure liquid and solid YCl₃as well as from a series of measurements on mixtures of YCl₃ and alkali chlorides, in which Raman modes of YCl₆ octahedra were observed in stoichiometric mixed crystals at room temperature and monitored first up to and across melting and then as a function of the YCl₃content in the liquid mixture up to the pureYCl₃melt.

In the latter report (Mochinaga et al., 1991), the structures of a series of LaCl₃, CeCl₃, PrCl₃, NdCl₃, GdCl₃, DyC1₃, and SmCl₃pure melts were systematically studied with a laboratory-scale X-ray diffractometer. In these rare-earth chloride melts, the existence of six chloride ions as the nearest neighbors surrounding a rare-earth cation was deduced from the interatomic distance ratio ofr(Cl–Cl)/r (R–Cl); interatomic distances as well as the coordination numbers of the rare- earth ions are listed inTable 9. Interpretation of Raman spectra also showed in the same report that the chloride ions and the rare-earth cations form octahedral complex anions RCl63

(R¼La, Ce, Pr, Nd, Gd, Dy, and Sm). Taking account of the low equivalent electrical conductivities and the separation between the nearest cations R³^þ–R³^þin the results of X-ray analysis of these melts, the existence of dimers or more polymeric complex anions may be alleged. As an example, parameters obtained from XRD data of pure DyCl₃ melt are shown in Figs. 9–11. The calculation procedure is the same as described before. The three nearest-neighbor ion pairs R^3þ–Cl, R^3þ–R^3þ, and Cl–Clwere assumed to follow a Gaussian distribution with a mean square displacement 2bijfrom the peak positions rij, and the quantitiesnij,rij, andbijcould be obtained by least- squares fitting of Debye equation for the reduced intensity functionSi(S).

We also briefly describe results obtained by other experimental techniques. The Raman spectra of pure molten YCl₃, DyCl₃, GdCl₃, PrCl₃, and LaCl₃are illustrated inFig. 12. The normal modes of vibration of octahedral XY₆-type chemical species are schematically presented inFig. 13, where only the vibrational modesn1,n2, andn5are Raman active (Nakamoto, 1978). The typical Raman peaks on the Rayleigh wings for the respective melts are observed in the wavenumber range of about 140 to 300 cm¹, corresponding to the stretching vibration mode, n1, of octahedral XY6 species. The weak bending modes are barely detectable at lower wavenumbers.

Handbook on the Physics and Chemistry of Rare Earths 122

0 2 4 6

8 0

3 6 9

1 2 3 4 5

10 12 14

1 2 3 4 5

r (Å) r (Å)

D(r) (Å-1) r-1.D(r).10-3 Å-2

6 7 8 9 10

440 360 280

LaCl₃ PrCl₃ GdCl₃ YCl₃

DyCl₃

Wavenumber (cm^-1)

Intensity (Arbitrary unit)

200 120 40

Handbook on the Physics and Chemistry of Rare Earths 124

Conventional molecular dynamics simulation was also carried out to investigate the local structure of molten DyCl₃, using the simple Born–

Mayer–Huggins-type potentials in which parameters were determined so as to reproduce the interference functionSi(S) of the melt. The octahedral coordination of the cations was ascertained from the angular distribution of—Cl– Dy^3þ–Cl, in which strong correlations were obtained at about y¼90 and 180 as given inFig. 14. The distances between the nearest-neighbor cations, r(R^3þ–R^3þ), inTable 9are commonly less than twice the distance ofr(R^3þ– Cl). Under the assumption that the octahedral species are linked through at least one or two Cl ions, that is corner- or edge-sharing, with each other, the above facts are consistent with this geometry and the —R³^þ–Cl–R³^þ angles listed inTable 22indicate bent, not linear, configurations of the octahedra. These data are in line with the possibility that clusters consisting of dimeric or more polymeric complex anions RCl₆³ exist in pure RCl₃melts.

The most probable model is illustrated inFig. 15.

As a consequence, the formation of clusters determines the physicochemical properties of molten rare-earth halides. As a typical example, the electrical conduction of molten chlorides is briefly discussed. The molar conductivities, formerly called equivalent electrical conductivities, of KCl, NaCl, CaCl₂, ZnCl₂, and RCl₃ melts so far measured are shown in Fig. 16 (Janz et al., 1968). The molar conductivities of RCl₃ melts are much smaller than that

Y Y Y

Y n(XY) n1(A1g)

d(YXY) n4(F1u)

d(YXY) n5(F2g)

d(YXY) n6(F2u) n(YXY)

n2(E_g)

n(XY) n3(F1u) X

of the CaCl₂melt in which similar octahedral complex anions CaCl64exist.

ZnCl₂has a highly developed network structure in the molten state, and thus, its electrical conductivity is known to be very low. The electrical conductivities of RCl₃melts are in the middle range between those of low-conductive ZnCl₂and high-conductive KCl and NaCl melts.

FIGURE 14 Angular distribution of the Cl–Dy^3þ–Clangle simulated by molecular dynamics.

TABLE 22 Average Angles R^3þ–Cl–R^3þestimated from XRD data (Mochinaga et al., 1991)

RCl3melt —R^3þ–Cl–R^3þ()

LaCl3 135

CeCl3 130

PrCl3 128

NdCl3 133

GdCl3 134

DyCl3 131

SmCl3 132

Handbook on the Physics and Chemistry of Rare Earths 126

After confirmation from Raman scattering experiments on molten CeCl₃ that octahedral CeCl63exists, a somewhat detailed discussion has been made on the medium range structure of this melt (Iwadate et al., 1992). The density of molten CeCl₃ dilatometrically measured at 1143 K amounts to 3.193 g cm³, corresponding to a molar volume of 77.19 cm³mol¹. Assum- ing that the Ce³^þions are homogeneously distributed in the melt, the mean interionic Ce^3þ–Ce^3þdistance is calculated to be 5.04 A˚ . But the most

FIGURE 15 Most probable model for dimeric ions R2Cl115. Black circles: R; empty circles: Cl.

600 ZnCl₂ 300

LaCl₃ CaCl₂

NaCl

KCl

ZnCl₂ PrCl₃

GdCl₃ DyCl₃ 50

L (10-4Sm2mol-1) 100 150

500 t (⬚C)

700

800 1000

probable interionic distance for Ce^3þ–Ce^3þ pairs reported in Table 23 is 5.21 A˚ , that is about 3.4% larger. This fact indicates that there is some special interaction between CeCl63complex ions.

When two discrete CeCl63ions approach each other as closely as possible, the shortest Ce^3þ–Ce^3þdistance is estimated to 6.04 A˚ by considering the steric hindrance of Cl ions and the equality of intraionic and interionic Cl–Cl distances, both of which are undistinguishable in the modeling of the Si(S) function (seeFig. 17). Therefore, the real medium range melt structure is prob- ably not composed of discrete CeCl63 octahedra.

Another model was therefore considered, named edge-sharing model and in which two octahedra share two Cl ions and form, for example, Ce₂Cl₁₀⁴ dinuclear decahalides. From short-range XRD analysis and Raman spectra, the Cl–Ce³^þ–Clangles and the Ce³^þ–Cl and Cl–Cldistances were calculated to be 90, 2.84 A˚ , and 4.05 A˚, respectively. The Ce^3þ–Ce^3þ

TABLE 23 Coordination Numbersnik, Interionic Distancesrik, and Root-Mean Square Displacementsh△rik2i^1/2for Molten CeCl3

(Iwadate et al., 1992)

i k nik rik(A˚) h△rik2i^1/2(A˚)

Ce^3þ Cl 5.6 2.84 0.249

Cl Cl 11.3 4.05 0.573

Ce^3þ Ce^3þ 6.5 5.21 0.660

Handbook on the Physics and Chemistry of Rare Earths 128

distance should be, thus, close to 4.0 A˚ , but the observed value is 5.21 A˚, indicating that the edge-sharing model is also not a good approximation for the CeCl₃melt.

In the third, corner-sharing model, two octahedra share one Cl ion and form a Ce2Cl115 ion. When the Ce^3þ–Cl–Ce^3þ configuration is linear, the interionic Ce^3þ–Ce^3þ distance is the same as the sum of the respective ionic radii, 5.64 A˚ according to Shannon (1976). This value is indeed greater than 5.21 A˚ , but if the Ce^3þ–Cl–Ce^3þ angle is set to 133.1, the corner-sharing model becomes valid as shown inFig. 15. In the same manner, trimer, tetramer, and larger polymeric ions, denoted CenClð5nþ1Þð2nþ1Þ, may exist in the melt.

As can be seen from the general form of the polymer, a linear configuration does not satisfy the CeCl3stoichiometry and values of the Cl/Ce ratio approach 5 when linear polymerization occurs. Therefore, clustering of octahedra as pre- dicted by the corner-sharing model and the existence of polymeric ions seem to be needed in order to reduce the Cl/Ce ratio down to the stoichiometric ratio 3 and to reconcile the structural model with the whole set of experimental data.

Additionally, this type of ionic bond has been thought to be not necessarily rigid, that is, to break and form again in a kinetic process. It may be concluded from the above discussion and in view of other physicochemical data that polymeric ions and their clustering exist in the CeCl₃melt.

Tosi et al. have published a series of excellent papers titledOrdering in Metal Halide Melts(Tosi et al., 1993),Melting and Liquid Structure of Poly- valent Metal Halides (Tosi, 1994a), Structure of Covalent Liquids (Tosi, 1994b), andThe Molten State of Trivalent Metal Halides and Oxides: Recent Progress(Akdeniz et al., 1998). In particular, the ordering of trihalide melts and the network formation are mentioned in these reviews. As stated before, several research groups interested in molten salts have carried out determina- tions of liquid structures for several trivalent metal halides based on measurements of total X-ray and ND patterns. They focus on two main themes: (i) the trend from cation-dominated Coulomb ordering to loose network structures across the series of sixfold-coordinated rare-earth chlorides, including YCl₃ and (ii) the competition between network formation and stabilization of molecular-type liquid structures with strong intermolecular correlations, as in associated molecular liquids. The overall structural evolution as the trivalent metal ion is changed is governed by the increasing weight of cova- lency versus ionicity. The macroscopic properties reported for trichlorides reflect melting mechanisms that are consistent with the observed liquid structures. Progressive network formation in the melt from LaCl₃to YCl₃is signaled by decreasing values ofDS, DV/Vl ands, whereDS is the entropy change, DV/Vl gives the relative difference between the volume Vl of the liquid atTmand that of the solid at room temperature, and s is the electric conductivity.

Regarding network formation, the following conclusion has been drawn on the basis of liquid structure data available from XRD for LaCl₃, CeCl₃, PrCl₃,

NdCl₃, SmCl₃, GdCl₃, and DyCl₃ (Mochinaga et al., 1991) as well as from ND for molten NdCl₃ (Saboungi et al., 1990). These compounds have the UCl₃ structure in their high-temperature crystal phase, except for DyCl₃, which transforms into the AlCl₃structure before melting. The UCl₃structure is hexagonal, with each metal cation (M) surrounded by six chlorides on the corners of a trigonal prism and further coordinated by three coplanar chlorides at somewhat larger distance (Wyckoff, 1964). The structure can be built by stacking into a chain MCl₃units shaped as trigonal pyramids with M at the apices and the three chlorides in the base of each pyramid being equally shared between two M to form the trigonal prism around each M, and then packing these chains so as to give three coplanar interchain M–Cl bonds for each M. Evidently, a molecular-type crystal structure results if each M is brought closer to one of the two triplets of chlorides forming each prism.

The AlCl₃structure can also be built from MCl₃ pyramidal units, arranged in layers so as to yield a slightly distorted cubic close packing of chlorides inside which planes of octahedral sites are alternately either occupied by M or empty (Templeton and Carter, 1954). Each layer of the AlCl₃ structure can almost be viewed as a hexagonal planar lattice of M sandwiched between two triangular lattices of chlorides having the phase relationship of adjacent (111) planes in the face-centered cubic lattice, as shown inFig. 18. In this figure, the lower cluster shows the octahedral coordination of the metal ion in the crystal, which is basically preserved in molten YCl₃. The upper cluster shows an Al₂C1₆unit, which, through the displacement of the two metal ions as indicated by arrows, yields an Al₂C1₆molecule on melting of AlCl₃.

FIGURE 18 Schematic illustration of a layer in the AlCl3, crystal structure and of melting in YCl3and AlCl3. The black spheres represent a plane of trivalent metal ions, and the gray and white spheres represent planes of chlorines above and below the plane of metal ions, respectively.

Handbook on the Physics and Chemistry of Rare Earths 130

All rare-earth trichlorides listed above have similar structural characteris- tics in the melt. The first-neighbor coordination number of the metal ions is 5.6–5.7 and the R–Cl bond length lies in the range 2.69–2.87 A˚ . The second-neighbor bond lengths are r(R–R)¼4.9–5.2 A˚ and r(Cl–Cl)¼3.8–

4.1 A˚ , indicating a Coulomb ordering primarily determined by the repulsion between the polyvalent cations, as discussed earlier for molten SrCl₂. Yet the ionic conductivity in the melt decreases steadily through the series of compounds from LaCl₃ to DyCl₃, and Raman-scattering data show that a high-frequency spectral shoulder in LaCl₃rises to become a very well-defined peak in DyCl₃(Mochinaga et al., 1991). This suggests that the essentially sixfold coordination of the metal ions becomes progressively more stable through the series, leading to a liquid structure that resembles a loose network of distorted Cl-sharing octahedra. The ND pattern of molten NdCl₃shows a relatively weak and broad structure in the region of the first sharp diffraction peak, FSDP (Saboungi et al., 1990).

DyCl₃ and YCl₃ are structurally isomorphous before melting and melt with similarly low values of DS and DV/Vl. In a ND experiment on molten YCl₃(Saboungi et al., 1991), the Faber–Ziman structure factor exhibited a well-defined FSDP at k¼0.95 A˚¹, giving unambiguous evidence of intermediate-range order. The average coordination number of the metal ions is 5.9, which confirms the Raman-scattering finding (Papatheodorou, 1977) of rather long-lived octahedral coordination of the metal ions, though in detail the octahedra in the liquid are somewhat expanded and distorted. The octahedral network must be relatively loose on a time scale longer than the period of the breathing mode of the octahedron atn¼260 cm¹, to be compatible with the value of the ionic conductivity of molten YCl₃(Akdeniz and Tosi, 1992;

Janz et al., 1968). The second-neighbor C1–C1 coordination is approximately 8.2, somewhat lower than the value of 9 for intralayer Cl–Cl correlations in the crystal.

The structural trend represented by increasing connectivity of highly coordinated local structures from LaCl₃to YCl₃can be reproduced within a purely ionic model (Tatlipinar et al., 1992). This model relates the appearance and growth of intermediate-range order in a 3:1 liquid to the decrease in the radius RMof the metal ion.Figure 19shows the evolution of the theoretical partial structure factors from LaCl₃through YCl₃and AlCl₃at constant density. As RMdecreases, Coulomb ordering becomes stronger, as indicated by the align- ment of peaks in the R–R and Cl–Cl structure factors with the valley in the R–Cl structure factor. At the same time, the main peak associated with cation ordering in LaCl₃ evolves into the FSDP in AlCl₃. When the fluid with RM¼0.82 A˚ is allowed to expand to densities appropriate to real AlCl3 at standard pressure, the model suggests stabilization of dimeric bound states.

As mentioned above, the main features of the local structures adopted by molten rare-earth halides have been elucidated during the 1990s by Tosi and coworkers (Akdeniz et al., 1998; Tosi, 1994a; Tosi, 1994b; Tosi et al., 1993).

Dalam dokumen Handbook on the Physics and Chemistry of Rare Earths (Halaman 155-195)