Crystallization of BaF2 from droplets of phase separated glass – evidence of a core–shell structure by ASAXS

Glasses with the mol% compositions 1.88 Na2O·15.04 K2O·7.52 Al2O3·69.56 SiO2·6.00 BaF2 and 1.88 Na2O·15.03 K2O·7.52 Al2O3·69.52 SiO2·6.00 BaF2·0.05 SmF3 were studied using X-ray diffraction, transmission electron microscopy, and anomalous small-angle X-ray scattering.

barrier is very different. The doped samples show long luminescence lifetimes of nearly 5 ms at a wavelength of 600 nm which is nearly as large as in fluoride phosphate glasses.

1.
Introduction: In the past few years, the formation of nano crystals from glasses was intensively studied [1][2][3][4]. It has been shown that a prerequisite for the formation of nano crystals is a core shell structure [5,6]. This can be achieved if during the course of nucleation and subsequent crystal growth the core gets successively enriched in network modifying oxides or other less mobile components. That leading to increased viscosity in the depleted residual glassy phase [7][8][9].
This results in a drastic decrease in the diffusion coefficients in the shell and to a deceleration of crystal growth. In some cases the crystals do not grow any further within the time scale of the experiments of some ten hours [10][11][12][13]. A very narrow crystal size distribution is observed, which is even narrower than the Lifshitz-Slyozov-Wagner (LSW) distributions obtained from Ostwald ripening [14,15] predicts. Previously, these core shell structures were predicted based on the chemical composition of the glass and the formed crystalline phases as well as the measured changes in the glass transition temperatures of partially crystalline samples. In the past decade new aberration-corrected transmission electron microscopes (TEM) have been developed that are able to detect the shell directly using electron energy loss spectroscopy (EELS) [16] or energy-dispersive X-ray spectrometry (EDXS) [17][18][19].
Moreover, the formed core shell structures were also confirmed with the data from anomalous small-angle X-ray scattering (ASAXS) [20].
Up to now, core shell structures were predominantly observed during homogeneous nucleation of rare earth fluorides [18] and of earth alkaline fluorides [8][9][10][11][12][13][14][15][16][20][21][22][23]. These fluoride nano crystals are good hosts for rare earth elements with interesting luminescence and up-conversion properties. The narrow size distribution is an important prerequisite for transparency in the visible light spectrum, which is required for applications in optics and photonics.
In lithium aluminosilicates, ZrTiO4 acts as nucleation agent. The first step of the nucleation process is the formation of an amorphous droplet-like phase, subsequently ZrTiO4 precipitates. This also leads to the formation of a core shell structure [7,24]. The formation of liquid/liquid phase separation with a droplet structure is observed in the case of some oxyfluoride glasses [21,22]. Already minor quantities of rare earth elements may lead to phase separation. It has been shown for an oxyfluoride glass from which BaF2 can be precipitated that an addition of rare earth oxides in a concentration as small as 0.05 mol% leads already to the formation of a droplet phase separation [23].
In the present paper, the effect of 0.05 mol% SmF3 on the crystallization of a glass with the composition of 1.88Na2O-15.04K2O-7.52Al2O3-69.56SiO2-6BaF2 is studied predominantly using the methods XRD, ASAXS and TEM.

Experimental Procedure
Glasses with the compositions (in mol%) In order to prepare electron transparent samples, the mechanical wedge-polishing approach was applied. First, a dedicated grinding and polishing tool (Multiprep, Allied company) was used. After finishing this preparation, low-energy Ar + ion broad-beam milling (precision ion polishing system PIPS, Gatan company) was performed followed by a selective carbon coating for mitigation of electrostatic charging [26].
The as prepared and annealed glass samples were studied by SAXS and ASAXS. The thicknesses of the on both sides polished sample sheets varied between 30 and 50 µm, which is required to achieve a sufficiently high X-ray transmission. The SAXS measurements were conducted at 4900 eV, an energy far below the Barium L3 X-ray absorption edge (Ba L3

Results
The prepared glasses were optically transparent without visible striae. The densities of both glasses were 2.603 g/cm 3 and equal within the limits of error. The glass transition temperatures determined by dilatometry were 473 °C and 485 °C for the glasses A and B, respectively. Unfortunately, exothermic peaks were not detected by DSC, supposedly due to the small volume concentrations of formed crystals.

XRD and Electron Microscopy
In a previous preliminary report [32], it has already been shown that the undoped glass is homogenous, while the Sm-doped glass shows liquid/liquid phase separation. Figure 1 shows TEM replica micrographs of a SmF3 doped (sample B) and an undoped sample A. In the sample with SmF3, heterogeneities are observed of sizes around 100 nm. Since the nonthermally treated sample is X-ray amorphous (Fig. 2), these heterogeneities are liquid/liquid phase separation with droplet structure. By contrast, the sample without SmF3 did not show any signs of heterogeneities and hence phase separation [20].     In Fig. 4  Further information from the SAXS curves was extracted by fitting the whole shapes of the curves with the spherical core shell model in combination with the Gaussian size distribution of particles, using the software SASfit. The evaluated fitting parameters on the particle core size and shell thickness are given in Table II. The distribution of core (BaF2) particles in both samples is shown in Fig. 5 (Right). The volume weighted mean diameters are fairly similar and centred at 11.2 nm for the crystallized Sample A and at 11.0 nm for Sample B (table 2).

Anomalous Small-Angle X-Ray Scattering
The thicknesses of the shells are 2.7 and 2.5 nm, respectively. The contrast variation values from the ASAXS curves were evaluated by fitting them with a spherical core shell model using the SASfit program and the procedure published previously [20]. Fitting the experimentally evaluated contrast values with the theoretically calculated values reveals the chemical composition and densities of core, shell and the remaining glass matrix. The evaluated parameters are provided in Table II.

Fig. 6 ASAXS scattering curves recorded from sample A annealed at 600°C for 20 h (right) and sample B annealed at 580 °C for 20 h (left) using X-ray energies close the absorption edge of Ba. X-ray energies used are 4900, 5177, 5234 and 5244 eV.
Since the scattering curves strongly depend on the energy near the Ba-absorption edge, the particles that give rise to X-ray scattering, are strongly enriched in Ba. According to the XRDpatterns they consist of BaF2. Therefore, the scattering curves were fitted assuming a BaF2 core and an SiO2 shell.  is not of importance. For example as light converters for blue LEDs to produce white light, slight scattering is not disadvantageous, and also for the use as active laser material in the infrared, scattering of visible light has not a negative effect.   Table 3. The lifetimes of the investigated samples are somewhat smaller than for the fluoride phosphate sample, an almost pure fluoride glass.

Discussion
Although glass B contains only 0.05 mol% SmF3, the structure was very different from that of the sample without samarium (glass A). While the glass doped with samarium showed phase separation with droplets with sizes of around 100 nm, the glass without samarium did not.
Thermal treatment leads to the crystallization of cubic BaF2 in both glasses. The XRD patterns show strongly broadened lines attributed to a mean crystallite sizes of around 11 nm, as calculated from the XRD-line broadening. This crystallite sizes are comparable with the size of the crystals observed in the TEM micrographs. The effect of time and temperature on the size of the crystals is comparably small as already discussed in Ref. [12]. However, the effect is notably more pronounced than in previously studied glasses, in which during annealing, CaF2 is precipitated [11,27]. For sample A, previous studies using advanced TEM measurements gave evidence of the formation of core shell structures [20]. Using SAXS and ASAXS, the existence of such a shell is confirmed. In the SmF3-doped sample annealed at 580 °C for 20 h, SAXS and ASAXS show also a core shell structure. In both cases, the shell is BaF2 is formed inside the droplet and SiO2 as well as possibly Al2O3 is expelled. It should be noted that stresses appear simultaneously with the formation of the core shell structure. These stresses can hardly relax because the composition of the shell is attributed to a higher glass transition temperature which is above the crystallization temperature applied.
It is remarkable that an addition of a SmF3 concentration as small as 0.05 mol% gives rise to drastic changes in the crystallization mechanism. It should further be mentioned that similar structures, in our case BaF2 nano crystals with SiO2 enriched shells do not allow to conclude on similar crystallization mechanisms.
It is further surprising that the luminescence lifetimes for a Sm 3+ emission at 600 nm are in the range of 5 ms. This is much longer than in silicate glasses, which usually show luminescence lifetimes of about 2-3 ms depending on their chemical composition [28,32,33].
For alumosilicate glasses, fluorescence lifetimes are in the range of 2.2 to 3.8 ms [28,29], It should be noted that luminescence of transition or rare earth cations is an effect also observed in liquids and hence not necessarily connected to the solid state. By contrast to luminescence caused by nano size semiconductors, nano size effects of rare earth containing compounds have never been reported.

Conclusions
Using X-ray diffraction, transmission electron microscopy and anomalous small-angle X-ray scattering a glass with the mol% composition 1.88 Na2Oꞏ15.04 K2Oꞏ7.52 Al2O3ꞏ69.56 SiO2ꞏ6.00 BaF2 as well as a glass doped with onlyꞏ0.05 mol% SmF3 were studied. While the undoped glass was homogeneous, the samarium doped glass showed droplet phase separation with a droplet size of around 100 nm as proved by TEM. Thermal treatment of the samples led to the crystallization of cubic BaF2. ASAXS gave evidence of a core shell structure. The spherical particles of the undoped and the doped sample had diameters of around 11 nm, while the shell had a thickness of 2.6 nm. Surprisingly, both the cores and shells possess hence similar dimensions if crystallized from the homogeneous glass and the phase separated glass. It should be noted, however, that the origin of the barrier is very different. In both cases, it is assumed that the shell acts as diffusion barrier that hinders crystal growth. The core sizes determined by ASAXS were in excellent agreement with the crystallite sizes calculated from the strongly broadened XRD-lines using Scherrer's equation. The samarium doped sample shows strong luminescence. The luminescence lifetimes were nearly 5 ms at 600 nm and hence unusually long.