Introduction

More precise heat treatment than in case of oxide glasses is required to obtain transparent, low-phonon glass-ceramics based on oxyfluoride glasses [1]. It is due to the fact that introducing fluorine atoms into glass structure causes its weakening and higher ability of crystallization of oxide framework. On the other hand, the material has to maintain its transparency and luminescent properties, so that the glass-ceramics could be used as an active element in optoelectronic systems. The luminescence properties are achieved by introducing into the structure of such glass optically active lanthanides. For receiving higher luminescence efficiency and greater lifetime, the lanthanides atoms should be located in low-phonon phase, in this case the fluoride phase with higher symmetry. For this purpose, the silicate glasses doped with LaF3 [2, 3], CaF2 [4, 5] and BaF2 [6, 7] were studied. The results of the research confirmed the improvement in luminescent properties of the materials. It is the issue of the material, in which the fluoride phase would readily host the optically active dopant into its structure. In this study, influence of Gd2O3 replacing Al2O3 in the glass composition on thermal glass stability and process of crystallization was studied in the BaF2–NaF–Na2O–Gd2O3–Al2O3–SiO2 system. The crystallization process was further studied by heat treatment, followed by X-ray diffraction experiments and FTIR analysis.

Experimental

Four glasses with varying content of Gd2O3 were prepared using chemically pure raw materials (Table 1). The glass was melted in a platinum crucible in an electric furnace in air atmosphere holding it at 1400–1470 °C for 60 min to obtain 0.1 mol of glass (ca. 10 g). During melting, the crucible was covered with a platinum plate to reduce vaporization losses. The losses of fluorides were compensated by extra amounts of components calculated from XRF analysis. The melt was poured onto a brass plate forming a layer of 2–4 mm thickness and then annealed at a temperature near the transformation temperature. The nominal glasses compositions are listed in Table 1.

Table 1 Nominal glass composition in mol%
Full size table

Thermal stability analysis was performed with DSC method, using NETZSCH STA 449 F3 equipment and a platinum crucible in argon atmosphere. Approximately 50 mg of grounded glass below 0.06 mm were used for DSC measurement from 50 to 1050 °C. Crystallization kinetics analysis was performed for four heating rates: 5, 10, 20 and 30 K min−1 after appropriate equipment calibration for each rate. The transformation temperature (T g) established on inflection point and peak temperature of crystallization (T p) were determined from the DSC traces of the respective glasses. The activation energy (E a) and Avrami exponent (n) were calculated according to the DSC data.

The heat treatment process was carried out in Nabertherm electric furnace. Phase identification was performed with X-ray diffraction analysis, using Philips X’Pert equipment at room temperature with CuKα radiation over the range of 10°–70° 2θ.

The morphologies of the crystalline phases developed upon the heat treatment of bulk glass samples were examined with a scanning electron microscope (SEM, NovaNano SEM 200, FEI Company). SEM tests were performed with an attachment for the chemical analysis of specimens in micro-areas with energy dispersive X-ray spectroscopy (EDAX). The observations were carried out in the secondary electron mode in low vacuum conditions (60 Pa) and at an accelerating voltage of 18 kV. The samples were covered with a carbon layer. The chemical composition was determined using a WD-XRF Axios Max spectrometer with Rh 4 kW PANalytical lamp.

FTIR spectra were recorded with a Bruker Company Vertex 70v spectrometer. Spectra were collected in the MIR regions (2500–400 cm−1) after 128 scans at 4 cm−1 resolution. Samples were prepared by the standard KBr pellet method.

Results and discussion

DSC analysis

DSC analysis shows that the increasing content of Gd2O3 (replacing Al2O3) influences the thermal stability (Fig. 1). A monotonic increase in transformation temperature is observed, together with higher Gd2O3 content (Table 2). Difference of T g between samples FGA_0 (with no Gd2O3 content) and FGA_10 (with no Al2O3 content) was found to be 56 °C. Jump-like change of heat capacity (Δc p) corresponding to the transformation temperature also shows a significant modification. Addition of 5% mol Gd2O3 reduces it, and then, Δc p increases along with complete replacing aluminum atoms with gadolinium. Glasses (FGA_5 and FGA_7.5) that contain both Gd and Al in their networks have the lowest Δc p value. It is due to the fact that those glasses demonstrate lower thermodynamic fragility and are not prone to strong relaxation effects [8].

Fig. 1
figure 1

DSC traces of glass: a FGA_0, b FGA_5, c FGA_7.5 and d FGA_7.5

Full size image
Table 2 Thermal characteristics of glasses from DSC curve of glass samples for α = 20 K min−1 heating rate
Full size table

Introducing of Gd2O3 causes crystallization changes, such as increase in the peak crystallization temperature (T p) with simultaneous occurrence of two exothermal effects (Table 2). Thermal stability, determined as a difference between onset crystallization temperature (T x) and T g [9–11], has higher values for glasses containing only one trivalent component, and this parameter has slightly lower value for glasses containing a mixture of Al2O3 and Gd2O3. Gadolinium addition also influences the crystallization enthalpy (ΔH) value. Comparing these values, it can be assumed that ΔH significantly decreases along with higher Gd2O3 content (Table 2). It indicates a change in mechanism of the crystallization process, associated with formation of various crystalline phases in the series of glasses, as shown in the XRD analysis.

In order to determine changes in the glasses crystallization kinetics, a non-isothermal DSC analysis was carried out with four heating rates (Fig. 1).

Crystallization kinetics analysis can be based on one of the reaction models, developed by Kissinger [12], Augis–Bennett [13] and Ozawa [14]. In the Kissinger method, the crystallization peak temperature is determined as a function of the heating rate α as following:

$$\ln (\alpha /T_{\text{p}}^{2} ) \, = \, - E_{\text{a}} /RT_{\text{p}} + {\text{const}}.$$
(1)

Ozawa and Kissinger plots are commonly used to calculate the Avrami constant, n, and the crystallization activation energy, E a, for glass [12, 15]. In the non-isothermal method, the values of the Avrami parameter are determined using the Ozawa equation:

$$\ln \left[ { - \ln \left( {1 - x} \right)} \right] \, = - n\ln \alpha + {\text{const}} .$$
(2)

where x is the crystallized volume fraction at T for the heating rate α. It should be taken into account that Kissinger’s approach does not consider the concept of nucleation and growth. In this case, transformation under non-isothermal condition is described as a first-order reaction.

Figure 2 represents the relation of ln(T 2p /α n) to 1000/T p according to the Kissinger–Matusita model [12, 16, 17] which is described as:

$$\ln \left( {T_{\text{p}}^{2} /\alpha^{{\text{n}}} } \right) = - mE_{\text{a}} /RT_{\text{p}} + {\text{const}} .$$
(3)
Fig. 2
figure 2

Plots of ln(α/T 2p ) versus 1000/T p for the oxyfluoride glasses: a FGA_0, b FGA_5, c FGA_7.5 and d FGA_10

Full size image

In our calculation, we assume the variable amount of nucleuses due to the process of formation fluorides as the first stage of the network rearrangement according to the XRD results. They can act as a new center of nucleation of silicates. In this case, m = n − 1. The obtained values have good linear correlation. The effect of crystallization was not observed for FGA_10 at heating rate of 5 and 10 °C min−1 and could not have been taken to our calculation. Thus, the values of E a and Avrami parameter (n) cannot be estimated exactly for the glass and may deviate from the proper value. It is presented as the dotted line in Figs. 2 and 3.

Fig. 3
figure 3

Plot of ln[−ln(1 − x)] versus ln(α) for the glasses: a FGA_0, b FGA_5, c FGA_7.5 and d FGA_10

Full size image

The crystallization activation energy, determined on the basis of the plots, is presented in Table 3. Analysis of crystallization kinetics from curves according to Eq. (3) indicates that the activation energy for aluminosilicate glass (FGA_0) is slightly higher than Si–O bond dissociation energy (360–430 kJ mol−1) in silicate glasses [18]. It may be connected with Al2O3 as a next component of the framework. There are enough modifier ions to compensate electric charge to form tetrahedral [AlO4]. The significant higher values are obtained when Gd2O3 was substituted Al2O3. The activation energy for the precipitation of BaF2 and BaGdF5 is much lower (250–350 kJ mol−1) [19] than those obtained in our analysis. As fluoride phases form during crystallization, Si–O bond dissociation is not actually necessary, as fluoride is not bound to silicon [20]. That is why it can be assumed that obtained in the study values apply to aluminosilica oxide network crystallization. On the DSC curves, a separate peak corresponding to fluoride phase forming was not observed. This statement is based on the XRD analysis which confirmed the fluoride crystallize below 800 °C and there is no deflection on the derivative DSC curves in the temperature region between T g and 800 °C. However, crystallization of fluoride phases can impact the increase in crystallization activation energy of silica network by shortage of cations necessary for silicates crystallization and/or by reinforcement of the oxide framework.

Table 3 Activation energy of crystallization and Avrami parameter ‘n’
Full size table

The values of n determined from ln[−ln(1 − x)] versus ln α plot slope (Fig. 3) are listed in Table 3. It is evident that it ranges from 1.7 to 1.4, which suggests a change in the crystallization mechanism [21]. Moreover, the calculated values of n are not integers, which indicates a multistage mechanism of crystallization and a complex crystallization behavior (Fig. 4). Silicate crystallization is preceded by fluoride phase crystallization as XRD analysis showed. It can be a result of higher chemical affinity of modifiers ions toward fluorine and lower strength of fluorine bonds than oxygen. Size and amount of formed fluorides can influence the silicate crystallization mechanism. The results suggest that with increase in Gd2O3 content in the glass, the crystallization mechanism changes from two-dimensional to one-dimensional growth of silicates (n parameter is lowering).

Fig. 4
figure 4

XRD patterns of FGA_0 sample after heat treatment

Full size image

Because of the crystallization effect disappearance in case of FGA_10 sample heated with 5 and 10 °C min−1 rate, it was impossible to carry out a precise crystallization kinetics analysis of this glass; thus, the E a and n values estimated based only on two heating rates can be inaccurate. However, a lower value of crystallization activation energy and Avrami parameter, n, can be observed. These changes, as well as Δc p, ΔT and n values, are similar to those corresponding to FGA_0. It indicates a similar role of Gd2O3 and Al2O3 in the glass structure. Their mutual substitution in the glass affects the change in thermal stability of silicate network, which must carry modification of crystallization process.

XRD analysis

XRD analysis revealed FGA_0 glass crystallization as a two-stage process. Heat treatment of the glass at 750 °C for 2 h leads to BaF2 formation as a first phase (Fig. 4). This effect is not visible on DSC curves, where the crystallization effect is situated above 800 °C. The sample analysis after 12 h of heat treatment at the same temperature shows the oxide network crystallization and formation of NaAlSi3O8 (Fig. 4). Intensity of the peaks corresponding to this phase is low, which indicates the initial stage of crystallization process of this phase. Heat treatment at 900 °C for 12 h leads to BaF2 phase disappearance (Fig. 4). It could be the effect of fluoride phase dissolution in the glass that becomes a melt in this temperature.

In the case of FGA_5 glass, at the first stage of crystallization (heat treatment at 770 °C for 2 h), BaGdF5 is formed (Fig. 5). Unlike BaF2, its thermal stability is higher and it is not disappearing after heat treatment at 900 °C for 12 h (Fig. 5). At the same time, similarly to FGA_0 glass, the silicate network glass crystallization in the form of albite is observed (Fig. 5). To determine the effect of heat treatment procedure on both phases crystallization ability, an experiment was conducted, in which a glass sample was placed in the preheated to 980 °C furnace and crystallized for 30 min. XRD analysis showed that under these thermal conditions, only BaGdF5 was formed (Fig. 5). Such behavior of BaGdF5 allows to modify the treatment process in order to obtain a transparent glass-ceramics. It is very important, since it gives a better opportunity to control shape and size of BaGdF5 grains (Figs. 12, 13). Similar effects to FGA_5 are observed for FGA_7.5 glass (Fig. 6), but the effect of BaGdF5 crystallization is lower and the additional crystallization of Gd9.33(SiO4)6O2 occurs at higher temperature. Complete substitution of Al2O3 with Gd2O3 (FGA_10 glass) leads to change of the course of crystallization. Despite the 12 h heat treatment, intensity of corresponding BaGdF5 peaks is very low (Fig. 7). Heat treatment at 880 °C for 12 h leads to small amounts of GdF3 formation. At the same time, gadolinium silicate Gd9.33(SiO4)6O2 is formed. However, its ability to crystallization is lower than NaAlSi3O8.

Fig. 5
figure 5

XRD patterns of FGA_5 sample after heat treatment (H index indicates the sample was put into a hot furnace directly)

Full size image
Fig. 6
figure 6

XRD patterns of FGA_7.5 sample after heat treatment

Full size image
Fig. 7
figure 7

XRD patterns of FGA_10 sample after heat treatment

Full size image

XRD results show that introducing gadolinium oxide instead of Al2O3 affects the type of the fluorides to be formed and benefits the crystallization process. BaGdF5 shows higher thermal stability than BaF2. This observation may be important for obtaining of glass-ceramics containing low-phonon fluoride phase with high transparency.

SEM/EDS study

Figures 8 and 9 show the microstructure of FGA_0 sample after heat treatment at 750 °C for 2 h and 900 °C for 30 min. BaF2 is visible in both samples. The crystallites are larger with much more spherical shape for higher temperature treatment.

Fig. 8
figure 8

SEM image (a) and EDS spectrum of FGA_0 after heat treatment at 750 °C for 2 h representing the point #1—crystallites of BaF2 (b) and point #2—glassy matrix (c)

Full size image
Fig. 9
figure 9

SEM image (a) and EDS spectrum of FGA_0 after heat treatment at 900 °C for 30 min representing the point #1—crystallites of BaF2 (b) and point #2—glassy matrix (c)

Full size image

The sample becomes mostly amorphous with increasing duration of heat treatment up to 12 h (Fig. 10). It is confirmed by small thermal effects on DSC curves and lower intensity of peak on XRD pattern. EDS of the visible crystallites indicates the silicate phase formation. There is no sign of spherical BaF2 crystallites that undergo dissolution.

Fig. 10
figure 10

SEM image (a) and EDS spectrum of FGA_0 after heat treatment at 900 °C for 12 h representing the point #1—crystallites of silicates (b)

Full size image

Figure 11 shows glass fracture near the sample surface. A difference in the cross section between the surface and the inside is visible. The distinct area of 10 μm in width with no visible BaF2 crystallites is visible near the surface, unlike the inner part of the sample (upper part of the photograph). It may indicate a different chemical composition of the surface and the inside of the sample. Lack of fluoride phase formation in the surface layer may be a result of vaporization of fluorine during heat treatment.

Fig. 11
figure 11

SEM image of FGA_0 cross section after heat treatment at 900 °C for 30 min

Full size image

A microstructure of FGA_5 sample after heat treatment at 770 °C for 2 h is presented in Fig. 12. Just comparing both heat-treated FGA_0 and FGA_5 samples, the difference in the size of crystallites can be observed. FGA_5 sample heated at even 20 °C higher temperature that FGA_0 sample shows significantly smaller crystallites, below 150 nm. However, the crystallites are well dispersed through the glass network and are not forming clusters in both cases.

Fig. 12
figure 12

SEM image (a) and EDS spectrum of FGA_5 after heat treatment at 770 °C for 2 h representing the point #1—crystallites of BaGdF5 (b) and point #2—glassy matrix (c)

Full size image

SEM microphotograph reveals the formation of crystallites in the prism shape after 12 h of heating (Fig. 13). EDS analysis showed higher concentration of Gd, Ba and F atoms in this area, which may suggest a fluoride phase. XRD analysis confirmed formation of BaGdF5 (Fig. 5) besides Na(AlSi3O8) phase which is also visible as small crystallites.

Fig. 13
figure 13

SEM image (a) and EDS spectrum of FGA_5 after heat treatment at 900 °C for 12 h representing the point #1—crystallites of BaGdF5 (b) and point #2—glassy matrix (c)

Full size image

Figure 14 presents microstructure of FGA_7.5 sample after treatment at the temperature 1070 °C for 2 h. The EDS analysis confirms the formation of three kinds of areas which may correspond to the formation of BaGdF5/GdF3, Gd9.33(SiO4)6O2 and Na(AlSi3O8).

Fig. 14
figure 14

SEM image (a) and EDS spectrum of FGA_7.5 after heat treatment at 1070 °C for 2 h representing the point #1—phase with higher concentration of Ga (b), point #2—phase with higher concentration of Ba (c), point #3—phase with composition corresponding to Na(AlSi3O8) (d)

Full size image

Figure 15 shows a microphotograph of FGA_10 sample after heat treatment at 880 °C for 12 h. Irregular areas with higher concentration of Gd precipitated in glassy matrix are observed (EDS point #1). Comparing these results with XRD analysis (Fig. 7), it can be concluded that they are mostly amorphous. The results correspond to the DSC analysis that shows lower values of ΔH.

Fig. 15
figure 15

SEM image (a) and EDS spectrum of FGA_10 after heat treatment at 880 °C for 12 h representing the point #1—crystallites of BaGdF5 (b) and point #2—glassy matrix (c)

Full size image

FTIR study

FTIR spectra for as-made glasses are presented in Fig. 16a. Broad absorption lines confirm the amorphous nature of the samples. Increasing the fraction of Gd2O3 results in minor spectral changes. Replacing Al2O3 with Gd2O3 leads to a shift of the major band maximum toward higher wavenumbers. This band corresponds to stretching vibrations of tetrahedrons [SiO4] with various polymerization degree [22, 23]. Introducing Gd2O3 to the glass does not shift the bands at 1187 cm−1, which is connected with asymmetric Si–O–Si Q4 and Si=O bond vibrations [24]. It can be an evidence that substitution of Al atoms with Gd does not cause silicate network depolymerization. On the other hands, the minor shift of 1000 cm−1 band associated with asymmetric vibrations of Si–O–Si Q2 [25] is observed. The change in Si–O bond in the glasses doped with different content Gd2O3 can be assessed by the relationship between Si–O bond strength and effective frequency as reported in the literatures [26, 27]. The shift of the peak at about 1000 cm−1 may result the Si–O bond strength increase as the doping content of Gd2O3 increase. This effect may be explained by lower Gd–O bond strength, which leads to higher strength of Si–O(Gd) bond than Si–O(Al). To confirm the assumption a deconvolution analysis is needed. The band at 711 cm−1 is observed for the FGA_0 glass as a result of Si–O–Si bending vibration. The band shifts significantly and monotonously toward higher value of wavenumber when Gd content increases.

Fig. 16
figure 16

FTIR spectra of as-made glasses (a) and glass-ceramic after 12 h of thermal treatment (b)

Full size image

Figure 16b shows FTIR spectra for samples heated for 12 h at 900 °C (FGA_0, FGA_5, FGA_10) and at 1070 °C (FGA_7.5). For sample with 7.5 and 10 mol% Gd2O3 alone (spectra c and d), the spectra differ not much to the as-made glass spectrum. It indicates that the crystallization ability of Gd9.33(SiO4)6O2 is much lower than NaAlSi3O8. It should be noted that fluoride phases vibrations, due to the chemical nature of the bonds, are not active in IR spectroscopy [28, 29].

Conclusions

Replacing Al2O3 with Gd2O3 in the glass composition significantly influenced the crystallization behavior of fluoro-silicate glasses, both in terms of fluoride phase and silicate network crystallization. The influence of the mixture of both oxides on crystallization was different than when we introduced them separately into the glass. Addition of Gd2O3 to the aluminosilicate glass increases the transformation temperature, at the same time decreasing Δc p and thermal stability ΔT. It also results in decrease of ΔH. The study indicates that Gd2O3 admixture significantly increases the activation energy of NaAlSi3O8 crystallization. Complete replacement of Al2O3 with Gd2O3 leads to Gd9.33(SiO4)6O2 formation, and Na-silicates crystallization is not observed. Due to Gd3+ ions participation in fluoride phases (BaGdF5, GdF3) formation, crystallization of Gd9.33(SiO4)6O2 is impended, which was confirmed by our study. Thus, Gd2O3 can be used as a facilitating component for fabricating the low-phonon fluoride phases in the transparent glass-ceramics.