The Phase Diagram GdF3-LuF3
aa r X i v : . [ c ond - m a t . m t r l - s c i ] F e b I. M. Ranieri and S. L. Baldochi
Center for Lasers and Applications, Inst. Pesquisas Energeticas & Nucl.,CP 11049, Butant˜a 05422-970, S˜ao Paulo, SP, Brazil
D. Klimm ∗ Institute for Crystal Growth, Max-Born-Str. 2, 12489 Berlin, Germany
The Phase Diagram GdF –LuF Abstract
The phase diagram gadolinium fluoride – lutetium fluoride was determined by differential scanning calorimetry (DSC)and X-ray powder diffraction analysis. Both pure components undergo a reversible first order transformation to a hightemperature phase. The mutual solubility of both components is unlimited in the orthorhombic room temperaturephase. The maximum solubility of Lu in the high temperature phase of GdF (tysonite type) is about 20% and themaximum solubility of Gd in LuF ( α -YF type) is about 40%. Intermediate compositions of the low temperaturephase decompose upon heating in a peritectoid reaction to a mixture of both high temperature phases. Key words: phase diagram, solid solution, phase transformation
1. Introduction
Rare earth trifluorides (REF , RE = La, Ce,. . . Lu), including yttrium fluoride, were extensivelystudied in the past, instead of this many questionsabout the high-temperature polymorphic transfor-mations and its structure so far remain. Oftedal[1,2] and Schlyter [3], determined the LaF struc-ture as hexagonal with the P /mcm space groupand two formula units per elementary cell, usingsynthesized powder and the lanthanum mineraltysonite (= fluocerite). Mansmann [4,5] and atsame time Zalkin et al. [6] established the currentaccepted structure of the LaF . Using smaller X-ray wavelength (Mo K α ) it was possible to observeweaker additional reflections on a single crystal, ∗ Corresponding author: Tel.: +49 30 6392 3024; fax: +4930 6392 3003.
Email addresses: [email protected] (I. M. Ranieri), [email protected] (S. L. Baldochi), [email protected] (D. Klimm). and the structure of the LaF tysonite structurewas described as trigonal with space group P ¯3 c Z = 6. The light rare earths fluorides from Lato Nd crystallize in this structure. All other rareearth fluorides crystallize at room temperature inthe orthorhombic structure determined by Zalkinand Templeton [7] for YF , also referred as β -YF ,space group P nma and Z = 4.Another characteristic pointed out by Mansmannfor the change of the structure from trigonal to or-thorhombic was due to a relation between the sizeof the metal and its coordination number. When theionic radius of the rare earth r RE decreases, the flu-orine ions tend to touch each other, resulting in arepulsive energy. A critical ratio r RE /r F = 0 .
94 wasproposed for the change from the LaF to the α -YF structure, taking into account the ionic radius calcu-lated by Ahrens [8]. Furthermore, the trigonal struc-ture could be stabilized if the REF -orthorhombicstructure becomes deficient in fluorine ions or at hightemperatures [5]. Preprint submitted to J. Solid State Chem. 3 December 2007 homa et al. [9] studied the behavior of theREF from SmF to LuF taking into accountX-ray diffraction at high temperature and differ-ential thermal analysis (DTA) experiments. It wasobserved that the REF from Sm to Ho have thetrigonal P ¯3 c T structure of LaF “tysonite”type mentioned above. For the small rare earthsfrom Er to Lu upon heating, the β -YF structurechanges to a hexagonal not well identified α -YF structure. This structure was considered tentativelyisostructural with the trigonal α -UO structure bySobolev et al. [10,11], belonging to P ¯3 m D d )space group, Z = 1 [12]. Nevertheless, the α -UO structure is not yet well understood too. Eithersuperstructures or a lower (e.g. orthorhombic)symmetry were discussed, as single crystal show abiaxial optical interference figure [13,14].Jones and Shand [15] proved that it was possi-ble to grow crystals of the four tysonites, LaF ,CeF , PrF and NdF , but only the orthorhom-bic DyF and HoF using CdF as scavenger. AfterGarton [16] and Pastor [17] it was established thatGdF , TbF , DyF and HoF crystals can be grownfrom oxygen free compounds and in a reactive atmo-sphere, confirming the thermodynamic studies bySpedding et al [18].For the intermediate SmF , EuF , and GdF twosubsequent transformations P nma (1) ←→ P ¯3 c (2) ←→ P /mmc were discussed by Greis [19] (the numberin brackets indicates the order of the phase trans-formation [PT]). These PT were inferred from elec-tron diffraction experiments with LaF , where smallsynthesized crystals presented also sub-cells reflec-tions as observed by Schlyter [3] and Maximov [20].Stankus [21] claimed that at high T the β -YF type( P nma ) and the LaF type (tysonite, P ¯3 c
1) be-come practically identical. Sobolev et al. [22] con-structed the phase diagrams of the systems GdF –LnF (Ln = Tb, Ho, Er, Yb), solid solutions regionswere proposed without phase transitions in all sys-tems, when the cation mean ionic radius was be-tween that of the Tb and Er .Recently, the present authors have published aphase diagram study of the system GdF –YF [23].It was established that both components undergoa solid-state phase transformation of first order be-fore melting. Additionally, a λ -shaped maximum of c p ( T ) being characteristic for a second order trans-formation was found for GdF . Both low- T and high- T phases exhibit unlimited mutual solubility. Thisobservation raises the question, whether the high- T structures of YF (reported as P ¯3 m
1) and GdF (reported as P /mmc ) may really be different [10].In the current paper, the phase diagram of thesystem GdF –LuF is reported for the first time.The interest in this phase diagram derived fromDTA studies regarding the phase diagram of thesystem LiF–Gd − x Lu x F , which is interesting todevelop new solid-solution crystals of the typeLiGd − x Lu x F to be used as laser host.
2. Experimental
Mixtures of Gd − x Lu x F with x = 0 . , . , . .
8, respectively, were prepared using commer-cial LuF (AC Materials, 6N purity) and GdF synthesized from commercial Gd O powder (Alfa,5N purity) by hydrofluorination. The oxide wasplaced in a platinum boat inside a platinum tube,and slowly heated in a stream of argon gas (WhiteMartins, purity 99.995%) and HF gas (MathesonProducts, purity 99.99%) up to 850 ◦ C. This processis described in detail elsewhere [24,25]. Conversionrates > .
9% of the theoretical value calculated forthe reaction Gd O + 6 HF −→ + 3 H Owere measured by comparing the masses prior toand after the hydrofluorination process. These sam-ples were used to the DSC measurements.Thermoanalytic measurements were performedwith a NETZSCH STA 409CD with DSC/TGsample carrier (thermocouples type S). The sam-ple carrier was calibrated for T and sensitivity atthe phase transformation points of BaCO and atthe melting points of Zn and Au. Sample powders(50 −
70 mg) were placed in graphite DSC crucibleswith lid. As graphite is not wetted by the moltenfluorides, the melt forms one single almost sphericaldrop (diameter d ≈ or LuF , respectively, wasfound to be almost identical. Thus it can be as-sumed that the partial evaporation does not lead toa considerable concentration shift. The inhomoge-neous powder samples were homogenized in a firstheating/cooling cycle with ±
40 K/min. Here theheating was performed to that T max where the DSCmelting peak was just finished and the molten sam-ple could homogenize. Depending on x , this was thecase for 1145 ◦ C ≤ T max ≤ ◦ C and due to thelarge heating rate the mass loss did never exceed2% in these preliminary mixing cycles. Withoutintermediate opening of the apparatus, the mixingcycle was followed by a measuring run with a heat-ing rate of 10 K/min. Although the crucicles werecovered by lids, the evaporation of ≈ − ( x = 0) to pure LuF ( x = 1) weremeasurered.Other samples were melted under a flux of hy-drogen fluoride gas, then pulverized to be analyzedby powder X-ray diffraction, using a Bruker AXSdiffractometer, model D8 Advance, operated at40 kV and 30 mA, in the 2 θ range of 22 . − . ◦ .The diffraction patterns where treated with theRietveld Method [26] using the GSAS program tocalculate the lattice parameters [27].
3. Results and Discussion
It turned out that GdF as well as LuF showedsimilar DSC heating curves: A first endothermalpeak due to the first order PT ( T GdF PT = 902 ◦ Cor T LuF PT = 946 ◦ C, respectively) is followed by asecond endothermal peak due to melting ( T GdF f =1252 ◦ C or T LuF f = 1182 ◦ C, respectively). The val-ues for GdF were measured and compared with lit-erature data recently [23], and for lutetium fluorideone finds values 943 ≤ T LuF PT ≤
963 and 1180 ≤ T LuF f ≤ T in ◦ C) in the literature [9,19,21].Only Jones & Sand [15] reported the very low value T LuF PT ≈ ◦ C. Like in the recent study [23] weak λ shaped bends were found in the DSC curves of GdF rich mixtures with x < .
2. This could be an indica-tion for a second order transformation between T PT and T f but will not be discussed here.Fig. 1 shows DSC heating curves for five com-positions starting from pure LuF ( x = 1 .
0) toGd . Lu . F . The PT peak for x = 1 . T side that is assumedto result from some contamination of the sample,as no additional peak that could be related to thepresence of oxyfluorides was detected on cooling.With increasing GdF content the melting peakshifts to lower T until it reaches for x ≈ . T eut = 1092 ◦ C. On the contrary,the PT peak shifts with increasing GdF content tohigher T and reaches for x ≈ . T PT = 1051 ◦ C. The same T PT and T eut are reached
900 950 1000 1050 1100 1150 1200
Temperature (°C) -2.0-1.00.0 D S C ( W / g ) x = 1.0 x = 0.6 x = 0.8 x = 0.4 x = 0.5 exo Fig. 1. DSC curves (2nd heating with 10 K/min) ofLuF ( x = 1 .
0) and of four mixtures Gd − x Lu x F with0 . ≤ x ≤ . T PT = 1051 ◦ C and T eut = 1092 ◦ C are thetwo isothermal phase boundaries in Fig. 4. if one starts with pure GdF and adds LuF .The lattice constants of nine samples with initialLuF concentrations 0 ≤ x ≤ P nma space group, no parasitic peaks that wouldindicate the presence of other phases were observedat room temperature (Fig. 2). The experimental er-rors { ∆ a , ∆ b , ∆ c } were always ≤ × − ˚A butwere found to be maximum in the region 0 . ≤ x ≤ .
4. Fig. 3 shows the total error σ = ∆ a +∆ b +∆ c versus the LuF concentration x of the sample, to-gether with a , b , c . θ (deg)
25 30 35 40 45 50 55 60 65 i n t e n s it y ( c p s ) GdF LuF Gd Lu F Gd Lu F Gd Lu F Fig. 2. Diffractions patterns of some Gd − x Lu x F samplesutilized to determine the lattice parameters at room tem-perature. Table 1 reports the results for calculated values of a , b and c , the volume of the unit cell V , the massdensity ̺ and the mean rare-earth ionic radius in3he solid solution ¯ r = (1 − x ) r [8]Gd + x r [8]Lu . r [8]Gd and r [8]Lu are the ionic radii of Gd and Lu , respec-tively, with coordination number 8 as determinatedby Shannon et al. [28]. Table 1Lattice parameters for the
P nma phase of Gd − x Lu x F (0 ≤ x ≤ Lu content ¯ r a b c V ̺ ( x , molar) (˚A) (˚A) (˚A) (˚A) (˚A ) (g/cm ) a a a a a b b b b ba PDF 49-1804; b PDF 32-0612
The lattice parameters a and b drop linearlywith x and c is almost constant for x ≤ . (Fig. 3). The smooth variation of a , b and c over the whole concentration range is obviously theresult of unlimited mutual solubility of GdF andLuF in their low- T phases. The variation of c inthis system is very similar to the c axis variation ofthe pure rare earth fluorides. In the orthorhombicrare-earth fluorides due to the lanthanides contrac-tion, there is a linear decreasing in the c parametervalue from 4.40 ˚A (SmF ) down to 4.376 ˚A (DyF )and then a nearly exponential increase up to 4.467 ˚A(LuF ) [7]. Taking into account the ionic radius ofHo , namely 1.015 ˚A (with a coordination numberof 8) [28], and the mean ionic radius of the compo-sition Gd . Lu . F , one has the same value.The DSC and XRD results are summarized inFig. 4 where full circles represent experimental DSCpoints that could be well determined by extrapo-lated onsets of sharp peaks. Such points representthe lower boundary of PT regions or, in the caseof melting, the solidus line. The higher boundary ofPT regions or the liquidus line, respectively, are lessremarkable as here the PT process or melting justfinishes and the DSC curve returns to the basis line. x × σ ( Å ) l a tti ce c on s t a n t ( Å ) b a c Fig. 3. Top: lattice constants for the
P nma phase ( β -YF type) of Gd − x Lu x F solid solutions at room temperature.Bottom: experimental error σ = ∆ a + ∆ b + ∆ c . Such more vague determined events are marked byhollow circles in Fig. 4. T ( ◦ C ) GdF LuF mol-% b b b b b b b b b b b b b bb b b b bb b b b b b b b b bbC bC bC bCbC bC β -(Gd,Lu)F tysonite α -Gd:LuF t y s o n i t e + β - ( G d , L u ) F α - G d : L u F + β - ( G d , L u ) F tysonite + α -Gd:LuF t y s o n i t e + l i q α - G d : L u F + l i q liquid Fig. 4. Phase diagram GdF –LuF . ˆ – sharp peaks. ž –small peaks or from offsets. α, β mean high- T phase or low- T phase, respectively. Two horizontal lines at 1051 ± ◦ C and 1092 ± ◦ Care the most prominent feature of the experimen-tal phase diagram Fig. 4. Obviously, two differentprocesses with strong thermal effects start indepen-dent on composition x for a broad range 0 . . x . . . . . . T PT = 1051 ◦ C or T eut = 1092 ◦ C, respectively. The components GdF and LuF show complete miscibility in the liquidphase. If the melt is cooled, either Gd-saturatedLuF ( α -YF type, P ¯3 m
1) or Lu-saturated GdF (tysonite type, P /mmc or P ¯3 c
1) crystallize fromthe melt. Both high- T phases belong to different4pace groups and unlimited miscibility is thereforeprohibited — instead a eutectic is formed below1092 ◦ C. In the eutectic mixture between T eut and T PT both constituents are mutually saturated. Inthe β -phase both components exhibit unlimited mu-tual solubility. The phase transformation occurringat T PT for medium concentrations x tysonite + α − YF ⇄ β − YF (1)is a peritectoid reaction.Near the left or right rims of the phase diagram,single phase Lu-doped GdF (tysonite structure) aswell as single phase Gd-doped LuF (= α -Gd:LuF )transform upon cooling to β -(Gd,Lu)F , crossinga 2-phase field where α - and β -phase are mixedand the upper and lower limit of this 2-phase fielddepend on x . Both 2-phase fields “tysonite + liq.”and “tysonite + β -(Gd,Lu)F ” are broader than thecorresponding fields on the LuF rich side wheretysonite is replaced by α -Gd:LuF . The largerwidths of the 2-phase fields on the GdF rich sidemay result in stronger phase segregation of suchcompositions, resulting after cooling to room tem-perature in microscopic grains with a wider varia-tion of compositions. As different compositions ofthe Gd − x Lu x F grains result in different latticeconstants, the larger experimental error σ that isobserved for Gd-rich mixtures (but not for pureGdF ) can be explained by variations of x resultingfrom segregation.
4. Conclusion
Solid solutions Gd − x Lu x F are formed over thewhole concentration range 0 ≤ x ≤ β -YF type structure with space group P nma like the pure components GdF and LuF .Upon heating GdF undergoes a first order phasetransformation to a tysonite type structure whereasLuF undergoes a first order transformation to a α -YF type structure that is not isomorphous totysonite. As a result, a miscibility gap separates thehigh- T phases of GdF and LuF . For intermediatecompositions 0 . . x . . − x Lu x F to a mixture ofthe high- T phases of GdF (saturated with Lu) andLuF (saturated with Gd) is a peritectoid decompo-sition.The behavior of the system GdF –LuF is some-what uncommon: Usually the mutual solubility ofcomponents rises with T , but in the present caseboth components show unlimited miscibility in the low- T phase only. It should be noted that Nafzigeret al. [29] obtained similar results (constant eutec-tic and phase transformation temperatures T eut =1073 ◦ C and T PT = 1045 ◦ C, respectively, over anextended composition range & –YF system. Nafziger et al.speculated that this is an indication for a low- T in-termediate phase, but the present authors think thata peritectoid decomposition (1) as shown in Fig. 4 ofthis work would be a more simple explanation andthat the postulation of an (otherwise not proven)intermediate phase is not necessary.AcknowledgementsThe authors acknowledge financial support fromCAPES (grant number 246/06) and DAAD (grantnumber D/05/30364) in the framework of the PRO-BRAL program.References [1] I. Oftedal, Z. Physik. Chem. B 5 (1929) 272–291.[2] I. Oftedal, Z. Physik. Chem. 13 (1931) 190–200.[3] K. Schlyter, Arkiv f¨or kemi 5 (1953) 73–82.[4] M. Mansmann, Z. anorg. allg. Chem. 331 (1964) 98–101,doi:10.1002/zaac.19643310115.[5] M. Mansmann, Z. Kristallogr. 122 (1965) 375–398.[6] A. Zalkin, D. H. Templeton, T. E. Hopkins, Inorg. Chem.5 (1966) 1466–1468, doi:10.1021/ic50042a047.[7] A. Zalkin, D. H. Templeton, J. Amer. Chem. Soc. 75(1953) 2453–2458, doi:10.1021/ja01106a052.[8] L. H. Ahrens, Geochim. Cosmochim. Acta 2 (1952) 155–169, doi:10.1016/0016-7037(52)90004-5.[9] R. E. Thoma, G. D. Brunton, Inorg. Chem. 5 (1966)1937–1939, doi:10.1021/ic50045a022.[10] B. P. Sobolev, P. P. Fedorov, Sov. Phys. Crystallogr. 18(1973) 392.[11] B. P. Sobolev, P. P. Fedorov, D. B. Shteynberg, B. V.Sinitsyn, G. S. Shakhkalamian, J. Solid State Chem. 17(1976) 191–199, doi:10.1016/0022-4596(76)90220-6.[12] W. H. Zachariasen, Acta Cryst. 1 (1948) 256–268,doi:10.1107/S0365110X48000703.[13] C. Greaves, B. E. F. Fender, Acta Cryst. B 28 (1972)3609–3614, doi:10.1107/S056774087200843X.[14] S. Siegel, H. R. Hoekstra, Inorg. Nucl. Chem. Lett. 7(1971) 497–504, doi:10.1016/0020-1650(71)80238-6.[15] D. A. Jones, W. A. Shand, J. Crystal Growth 2 (1968)361–368, doi:10.1016/0022-0248(68)90029-8.[16] G. Garton, P. J. Walker, Mat. Res. Bull. 13 (1978) 129–133, doi:10.1016/0025-5408(78)90077-6.[17] R. C. Pastor, M. Robinson, Mat. Res. Bull. 9 (1974)569–578, doi:10.1016/0025-5408(74)90126-3.
18] F. H. Spedding, B. J. Beaudry, D. C. Henderson,J. Moorman, J. Chem. Phys. 60 (1974) 1578–1588,doi:10.1063/1.1681233.[19] O. Greis, M. S. R. Cader, thermochimica acta 87 (1985)145–150, doi:10.1016/0040-6031(85)85329-6.[20] B. Maximov, H. Schulz, Acta Cryst. B 41 (1985) 88–91,doi:10.1107/S0108768185001677.[21] S. V. Stankus, R. A. Khairulin, K. M. Lyapunov,High Temperatures – High Pressures 32 (2000) 467–472,doi:10.1068/htwu216.[22] D. F. Sobolev, V. S. Sidorov, P. P. Fedorov, D. D. Ikrami,Sov. Phys. Crystallogr. 22 (1977) 574–577.[23] D. Klimm, I. M. Ranieri, R. Bertram, S. L. Baldochi,Mat. Res. Bull. 42,doi:10.1016/j.materresbull.2007.04.004.[24] H. Guggenheim, J. Appl. Phys. 34 (1963) 2482–2485, doi:10.1063/1.1702768, see also: R. Bougon, J.Ehretsmann, J. Portier, A. Tressaud, in P. Hagenmuller(Ed.): Preparative Methods in Solid State Chemistry,Academic Press, New York, 1972, p. 401.[25] I. M. Ranieri, Growth of LiY − x TR x F :Nd (TR = Luor Gd) crystals for optical applications, Ph.D. thesis,IPEN/University of S˜ao Paulo (2001).[26] H. M. Rietveld, J. Appl. Cryst. 2 (1969) 65–71,doi:10.1107/S0021889869006558.[27] A. C. Larson, R. B. V. Dreele, General structure analysissystem (GSAS), Tech. Rep. LAUR 86-748, Los AlamosNational Laboratory (2007).[28] R. D. Shannon, Acta Cryst. A 32 (1976) 751–767,doi:10.1107/S0567739476001551.[29] R. H. Nafziger, R. L. Lincoln, N. Riazance, J. inorg.nucl. Chem. 35 (1973) 421–426, doi:10.1016/0022-1902(73)80553-6.:Nd (TR = Luor Gd) crystals for optical applications, Ph.D. thesis,IPEN/University of S˜ao Paulo (2001).[26] H. M. Rietveld, J. Appl. Cryst. 2 (1969) 65–71,doi:10.1107/S0021889869006558.[27] A. C. Larson, R. B. V. Dreele, General structure analysissystem (GSAS), Tech. Rep. LAUR 86-748, Los AlamosNational Laboratory (2007).[28] R. D. Shannon, Acta Cryst. A 32 (1976) 751–767,doi:10.1107/S0567739476001551.[29] R. H. Nafziger, R. L. Lincoln, N. Riazance, J. inorg.nucl. Chem. 35 (1973) 421–426, doi:10.1016/0022-1902(73)80553-6.