Thermal annealing of radiation damaged titanite J. CHROSCH,1 M. COLOMBO,1 T. MALCHEREK,1 E.K.H. SALJE,1,* L.A. GROAT,2 AND U. BISMAYER3

In short Radiogenic impurities (1) of U and Th in titanite, CaTiSiO5, lead to moderate radiation damage (ø a-decay events/g) (2) and therefore to partial amorphization (ø30%) (3). Powder X-ray diffraction on such damaged titanite from the Cardiff locality in Canada shows that two modifications of the crystalline material coexist. Both modifications are structurally b phase but differ systematically in their lattice parameters and also in their chemical composition.

One modification exhibits strong broadening in X-ray diffraction patterns, whereas it is almost unstrained with respect to fully annealed titanite. The other modification shows large strain broadening and increased specific volume (about 3%) due to a high concentration of defects. The unstrained modification consists of small nucleation centers in the damaged material, and it grows when the sample is annealed. In short

At annealing temperatures above 823 K, this modification dominates rapidly and replaces the strained titanite. The analysis of volume strain and of structural strain resulting from the peak profiles suggests a temperature- dependent activation energy for the recrystallization process. In short

Introduction Titanite, CaTiSiO 5, is an accessory mineral in a wide variety of igneous and metamorphic rocks. Authors were motivated to investigate the damage annealing of titanite for two reasons. First, titanite shows both structural damage by radiation and also structural phase transitions. This coincidence makes titanite an ideal model compound for the systematic investigation of the interplay between radiation damage and subtle structural transformations. Second, the storage of nuclear waste may involve the radiation damage of titanite-type materials.

The structure consists of chains of corner-sharing TiO6-octahedra cross- linked by edge-sharing chains of CaO7 polyhedra.Isolated SiO4 tetrahedra share corners with both structural units (Speer and Gibbs1976; Taylor and Brown 1976). Synthetic and natural titanites undergo a phase transition between the a phase and the b phase at 496 K. This phase transition has been thoroughly investigated and it is mainly related to a change in the position of the Ti atoms within the octahedron.

The b phase is observed at room temperature for samples with Ti partly substituted by Al or Fe (Higgins and Ribbe 1976; Oberti et al. 1991). The structural formula of titanite may be written symbolically as CaTi(SiO4)O, with substitution possible at both cation and anion sites within the structure. Substitution of Ti by lesser charged cations is charge balanced with (OH) or F replacing O. The Ca position is sevenfold coordinated and can accommodate minor amounts of Na, U, and Th (Hughes et al. 1997).

As a result of the two latter radiogenic substitutions, titanites may suffer substantial radiation damage (Hawthorne et al. 1991; Vance and Metson 1985). The extent of this damage depends on the radiation dose (Ewing 1987; Ewing et al. 1987) due to the progressive overlap of a-recoil collision cascades. Naturally occurring titanite has been found either as an essentially crystalline phase or as a mixture of amorphous and crystalline regions (Hawthorne et al. 1991; Vance and Metson 1985).

The titanite used is from the Cardiff locality in Ontario, Canada. and contains substantial amounts of Al2O3 and Fe2O3, replacing Ti on the octahedral sites. Therefore, on average, at least 20% of the octahedral sites are occupied by Fe and Al, rather than Ti. The U and Th concentrations reported for these samples are equally variable (between 400 and 800 ppm). Based on an equivalent uranium (EU) content of 500 ppm and an estimated age of 1000 Ma, Vance and Metson (1985) conclude a dose of a-decay events/g, corresponding to about 30% of the dose needed to render the structure completely X-ray amorphous.

The starting material used in this study was a dark red color, with well-developed crystal faces and a glassy appearance. Some impurities and alteration products could be identified by X-ray diffraction, The principal impurity phases are fluorite and calcite Other possible phases, which also crystallize in the fluorite structure, include thorianite, ThO 2, and uraninite, UO 2. Due to the presence of heavy elements, these phases might contribute to the X-ray scattering background, even though their actual volume fraction is well below 1%.

Guinier powder diffraction on quenched samples and Rietveld refinement Small amounts (35 mg) of the initial titanite crystal were wrapped in platinum foil and annealed at temperatures between 473 and 1123 K. At temperatures above 700 K, samples were kept under a nitrogen atmosphere during annealing. The average weight loss was between 0.08% at 473 K and 0.7% at 1123 K. Following Hawthorne et al. (1991) this weight loss can be attributed to the combined presence of large F concentrations and H 2 O in Cardiff titanite. The principal evolving gas species under these conditions is H 2 for samples with F at moderate temperatures, whereas at higher temperatures, 1073 K, a mixture of SiF4, O2, F, HF, and H2 escapes. Than the quenched titanite samples were crushed in an agate mortar and ground to a fine powder.

ROOM-TEMPERATURE POWDER DIFFRACTION Recrystallized titanite A full structure refinement was performed on the sample annealed at 1123 K, which prior TEM observations showed to be fully recrystallized.

Single-crystal structure refinements of metamict Cardiff titanites (Hawthorne et al. 1991) show only minor deviations of the average structure from the structure of recrystallized material. Therefore the assumption of an undisturbed average structure must be valid. However, in terms of Al content, the refinements with fixed structural parameters can only be expected to give a qualitative estimate of the true impurity concentration. ROOM-TEMPERATURE POWDER DIFFRACTION

Metamict titanite Bragg peaks in this natural Cardiff titanite are split as clearly seen at ø2.6, 3, and 3.2 A° in Figure 2. This can be attributed to contributions from two different microstructural fractions of titanite. One fraction exhibits predominantly strain broadening, corresponding to Bragg peaks shifted toward larger d- spacings. This strained fraction will be called titanite II.

The other fraction, assigned to smaller d- spacings, has approximately the same cell parameters as the final, annealed titanite. Hence, this fraction can be regarded as almost unstrained, and we will refer to it as titanite I. Broadening of the corresponding Bragg peaks is predominantly caused by small particle size. ROOM-TEMPERATURE POWDER DIFFRACTION Metamict titanite

Rietveld refinement of natural Cardiff titanite and of partially annealed samples was completed using two distinct titanite modifications. Structurally, these two modifications were taken to be identical, apart from the variation of the Al concentration on the octahedral position and their unit-cell dimensions. The diffraction signals of both titanites merge rapidly at temperatures above 873 K, so that values for only one structure are reported at this temperature. Various strain parameters can be calculated for both microstructural titanite fractions. The structural strain of the titanite II component is plotted in Figure 3 as a function of annealing time. The observed volume strain in Cardiff titanite is approximately 3%,

The logarithm ofthe volume strain during heating at 773, 873, and 973 K is shown in Figure 4 as function of logarithmic time.

The powder diffraction pattern of the starting material (cf. Fig. 5, top, t 5 0) shows a high background diffraction signal of the amorphous matrix with superimposed powder diffraction lines of the crystalline parts of the sample. The Bragg peaks are only weakly broadened.. As in the case of the peak (Fig. 2), the fine structure of the powder lines clearly shows two peaks related to the two different modifications of titanite

the splitting can be seen in the pattern at 573K

Annealing the sample at 573 K leads to the time evolution of the powder spectrum shown in Figure 5, The intensity of the (211)- reflection increases simultaneously for both phases, titanite I and II. The quantitative increase is similar for both peaks, with no indication that significant changes of the relative intensities of any powder peak occur, i.e., the time evolution is characterized by a uniform increase of all crystalline diffraction signals. This clearly shows that the amounts of both titanites in the sample increase without major structural changes.

Further annealing at 823 K follows a very different pattern The Bragg peak of titanite I swiftly gains intensity, whereas the integrated diffraction signal of titanite II decreases. However, the height of the broad titanite II peak increases with time whereas the concomitant shift toward higher diffraction angles indicates a decrease of the lattice volume for this component.

Annealing at 973 K is similar to that at 823 K. The remaining diffuse scattering signal of titanite II disappears gradually and merges with the diffraction signal of titanite I within an annealing time of approximately 30 h. The final annealing stage is now identical to the powder diffraction pattern of a standard titanite sample in its b phase (Chrosch et al. 1997).

The recrystallization is seen as the general increase of the powder peaks (Fig. 8) and a simultaneous decrease of the diffuse background diffraction, arising from theamorphous regions of the sample.

In order to quantify this effect the background scattering was integrated between 2u and 2u and plotted as a function of time (Fig. 9). The decrease of the diffraction signal of the amorphous matrix is clearly visible.

A roughly linear relationship between the decrease of the background scattering and the increase of the Bragg intensity of the (211) peak of titanite I is shown in Figure 10. The position of the (211) peak changes very little with annealing, and the effective particle size of the final stage of annealing is about 800 A°.

All of the above observations were made with respect to b titanite as the stable crystalline state. in radiation damaged titanite. This correlates with the tendency of substitutional defects to destabilize the a phase in favor of the b phase (Hughes et al. 1997; Oberti et al. 1991). The Cardiff sample contains substantial amounts of Fe and Al, and one might expect at least a partial suppression of the a phase, e.g., similar to titanite from the Rauris locality (Zhang et al. 1995). Annealing the sample does not seem to produce the equilibrium phase (i.e., the a phase) of titanite, but rather the high-temperature b phase. This observation might well be explained by the high degree of Al content of the sample before and after annealing.

A possible interpretation of the observed peak splitting in metamict titanite is that the strain-broadened fraction stems from a titanite structure, which is exposed to stress from the surrounding amorphous parts of the material. Moreover, high defect concentrations in these crystalline areas cause the large volume strain (ø3%) observed for this fraction. This volume strain might define a limiting value for the crystalline state above which long- range order ceases to exist. Such a limit to the defect concentration taken up by the crystalline structure also has been found in zircon (Holland and Gottfried 1955).

Titanite I can be attributed to nuclei of recrystallized titanite inside the amorphous regions or alternatively to remnants of titanite that had never been subject to radiation damage. The strain analysis of annealed samples can be used to obtain insight into the activation energy of the recrystallization process.

Whereas the structural strain seen in titanite II, or in the merged peaks at higher temperatures, appears to decay exponentially at temperatures above 873 K, it remains unchanged or it even increases slightly at lower temperatures.

The kinetic behavior of the Bragg intensities at 823 K (Fig. 8) and the position of the diffuse scattering (Fig. 7) show the same characteristic time evolution, a rapid change at times shorter than 2 h and a much weaker time dependence for the later stages. Such a multistage behavior is commonly observed for other radiation damaged material (Weber 1990). The short time regime is then attributed to the annealing of Frenkel defects in weakly damaged regions whereas the late time regime is dominated by the propagation of the stable, crystalline phase into the unstable, amorphous matrix.

Our results can be analyzed within the same theoretical model. The diffuse scattering is then associated with regions of high defect densities but remaining weak crystallinity, i.e., small grain sizes. At high temperatures this diffuse scattering is observed toward the low angle side of the Bragg peak, as the particle size of the defect rich, but crystalline material decreases. Annealing leads to a shift of the diffraction maximum that corresponds to a strain release of some 3%. The recombination of Frenkel defects usually dominates the low-temperature annealing behavior in radiation damaged zircon (Weber 1990; Vance and Metson 1985) whereas growth phenomena are most relevant at higher temperatures.

The combination of continuous radiation damage and partial recrystallization of metamict titanite can drive this material into a state of microstructural and possibly chemical inhomogeneity. We distinguish between three fractions of titanite: (1) The original titanite structure expands by 3% with respect to the undamaged material. Any part of the structure that is subjected to larger volume strains is so heavily damaged that it does not contribute to the long-range order seen by the X-ray diffraction experiments.

These heavily damaged areas of titanite form the amorphous matrix. The total volume fraction of amorphous material with respect to the crystalline part depends on the radiation dose, whereas the structural state of both fractions appears to be more or less constant.

The presence of small particles of largely undamaged titanite in the untreated material might be due to regions in the metamict crystals that are never affected by a-decay events. This can be caused either by inhomogeneous distribution of U and Th in the sample (Murakami et al. 1991) or by channeling effects of the metamict structure itself (Diaz de la Rubia 1996).