The effect of melt-homogenization and heat-treatment on the optical properties of the rare earth doped oxyfluoride glass-ceramics

The effect of melt-homogenization and heat-treatment on the optical properties of the rare earth doped oxyfluoride glass-ceramics


Introduction
An oxyfluoride glass-ceramic (GC) is composed of both glass phase and fluoride crystals, and is produced by heat-treating its parent or precursor glass (PG) [1].The PG is obtained by quenching its liquid state at a sufficient rate to bypass crystallization [2].It was reported that the phase separation, e.g., the separation between the fluoride-rich and the aluminosilicate-rich phases in oxyfluoride glasses, could induce the formation of nanocrystals [3,4].In other words, the fluoride crystals can be precipitated from the fluoride-rich phases in the PG via appropriate heat-treatment (HT).However, the growth of fluoride crystals in some oxyfluoride glasses is limited by the energy barrier for ionic diffusion in the liquid around crystals, and hence large crystals cannot be obtained [5][6][7].It was reported that the small-sized fluoride crystals are difficult to accommodate more rare earth (RE) ions for optimizing optical properties for applications [1].To achieve large-sized fluoride crystals in oxyfluoride glasses, some of the present authors recently designed numerous compositions, from which large fluoride crystals can be generated even during melt-quenching [8].The thus-derived oxyfluoride glasses contained different kinds of fluoride crystals, such as NaLuF 4 , KTb 2 F 7 , and K 3 YF 6 [9][10][11].
Compared with traditional oxyfluoride glass-ceramics (GCs), the melt-quenching derived oxyfluoride GCs have the following two advantages.First, the latter ones contain larger fluoride crystals with low phonon energy, which can accommodate more RE ions, and this gives superior optical performances [12].Second, the preparation of such new types of GCs is simpler and more energy-saving since the HT procedure for traditional GCs can be skipped [13].Thus, the melt-quenching derived GCs have potential to be used as high-performance optical materials.However, the origin of the formation of the fluoride crystals in the oxyfluoride glasses still needs to be further investigated.In particular, revealing the relations between structural heterogeneity, crystal formation and functionalities in glass is a key for designing new types of functional glass-ceramics [14][15][16].In this work, we investigated the impact of structural heterogeneity and phase separation on fluoride crystal formation and optical properties by varying the melt-homogenization temperature (T homo ).We prepared the Er 3+ -Yb 3+ co-doped Ba 2 LaF 7 melt-quenching derived oxyfluoride PG and GC samples by quenching the glass melt from four different T homo .To study the structural and morphological evolution of the PG and GC samples during crystallization, we utilized X-ray diffraction, Raman spectroscopy and scanning electron microscopy on both the melt-quenched and the heat-treated samples.In addition, we detected the phase transition of the studied samples using differential scanning calorimetry.We determined the optical properties of the studied materials and identified the optimum T homo for fabricating oxyfluoride GCs with high optical performances.

Experimental
The oxyfluoride PG and GCs with the molar composition of 45SiO 2- -15Al 2 O 3 -12Na 2 O-21BaF 2 -7LaF 3 -0.5ErF 3 -1.0YbF 3 were prepared by the melt-quenching method.10 g raw materials, each component of which has a purity of 99.99%, were mixed and put into an alumina crucible with a lid, and then melted and homogenized in a box-type furnace in air atmosphere at 1450, 1500, 1550, and 1590 • C, respectively, for 45 min.Each melt was cast quickly onto a stainless-steel plate preheated at 300 • C. Subsequently, the obtained samples were annealed at 500 • C for 8 hours in a muffle furnace to release thermal stress and then cooled down to room temperature in the furnace.The abovederived samples were denominated as PG-1, PG-2, PG-3 and PG-4, according to the melt-homogenization temperatures of 1450, 1500, 1550, and 1590 • C, respectively.The four PG samples (PG-1, PG-2, PG-3, and PG-4) were subjected to heat-treatment (HT) at around 1.07T c1 for 2 hrs.where T c1 is the onset temperature of the first crystallization (Fig. 1), and accordingly the heat-treated samples were named HT-PG-1, HT-PG-2, HT-PG-3 and HT-PG-4.
The studied samples were upscanned using differential scanning calorimetry (DSC, NETZSCH STA 449F3 Jupiter) at 10 • C/min in nitrogen atmosphere.X-ray diffraction (XRD) measurements were performed to identify the crystalline phase by a powder diffractometer operated at 45 KV and 40 mA, using Cu-Kα as the radiation.The Raman spectra in the range of 175-1200 cm − 1 were obtained using a micro-Raman spectrometer (inVia, Renishaw) with a 532 nm green HeNe laser at room temperature.The microstructure of the samples was characterized by field-emission scanning electron microscopy (SEM, QUANTA 200) using a voltage of 30 kV.The chemical compositions of the samples were determined by the special aberration-corrected transmission electron microscope (ACTEM, FEI Titan Cubed Themis G2 300) equipped with High-Angle Annular Dark Field transmission electron microscope (HAADF-STEM) and selected area electron diffraction (SAED) at the voltage of 200 KV.The optical absorption spectra in the wavelength range of 275 to 1000 nm were acquired using a Varian Cary 50 spectrophotometer.The local structure in PG sample was characterized by performing the solid-state nuclear magnetic resonance (NMR) experiments (Bruker Avance III HD 500 MHz spectrometer (11.7 T)) at 25 • C. 27 Al magic angle spinning (MAS) NMR spectra were obtained at the resonance frequency of 130.2 MHz, operating with a 4 mm MAS probe at a spinning rate of 12 kHz.The typical pulse length was 0.83 μs (10 • liquid flip angle).The relaxation delay was 0.5 s for all the samples.The chemical shifts of 27 Al were referenced to Al(NO 3 ) 3 (1 M) aqueous solution.The UC luminescence spectra were recorded in the wavelength range from 500 to 700 nm using a HITACHI F-7000 fluorescence spectrophotometer under the 980 nm laser excitation.

Calorimetric analysis
Fig. 1a and b show the DSC curves of both PG (PG-1, PG-2, PG-3, and PG-4) and the heat-treated PG (HT-PG-1, HT-PG-2, HT-PG-3, and HT-PG-4) samples, respectively, from which the glass transition temperature (T g ), the first and second onset temperatures of crystallization peaks (T c1 and T c2 ) and the first and second melting peaks (T p1 and T p2 ) were determined (Table 1).It is seen that both T g and T c1 slightly decrease with increasing melt-homogenization temperature (T homo ) for the PG samples prepared with temperatures up to 1550 • C. By further raising T homo to 1590 • C, an increase of T g and T c1 is observed.The T g decrease could be attributed to the depolymerization of the glass network since the fluoride crystallites break down at T homo = 1550 • C and hence, the fluorine ions as modifiers dissociate the network.In contrast, the T g increase could be ascribed to the increase of the network

Table 1
The onset temperatures (in • C) of the glass transition (T g ), the first and the second crystallization peaks (T c1 and T c2 ), and the first and the second melting peaks (T p1 and T p2 ) of both PG and the heat-treated PG samples, respectively.polymerization degree as fluorides (SiF 4 and NaF 4 ) escape from the melts at higher T homo (1590 • C) and thus the number of bridging oxygens increases [17].In Fig. 1a, it is observed that T c2 of PG samples increases with increasing T homo , indicating that the secondary crystallization needs higher temperature (i.e., higher kinetic energy) to be initiated.Increasing T homo can help destroy some relatively ordered domains, that is, reduce the dynamical and structural heterogeneity in the melt, and thus increasing the energy barriers for nucleation to occur.This indicated that varying T homo is an effective way to alter the crystallization behavior of glasses.It should be noted that the first melting peak for PG-2, PG-3, and PG-4 samples might be hidden in the second crystallization peak (see T p1 ), whereas the melting peak is observable in the curve of PG-1 sample.In contrast, T p2 exhibits only a slight change with varying T homo , meaning that the melting event is almost independent of the melt-homogenization history.
As shown in Fig. 1(b), both T g and T c1 of the heat-treated PG samples are higher than those of the corresponding PG samples.This is attributed to the fact that the fraction of the connected [SiO 4 ] and [AlO 4 ] tetrahedral units in the glass matrix increases upon heat-treatment (HT) owing to the precipitation of the Ba 2 LaF 7 crystals.Thus, the network connectivity of the glass phase increases, thereby making both T g and T c1 higher [18].The heat-treated PG samples exhibit lower T c2 than the PG samples, implying that the second crystallization occurs more easily.Comparing the DSC curves of both PG-1 (Fig. 1a) and HT-PG-1 (Fig. 1b), it is seen that the first melting peak in PG-1 becomes stronger due to the further formation of Ba 2 LaF 7 crystals upon HT.In addition, the first melting peaks of HT-PG-2, -3 and -4 samples appear during DSC upscans, indicating that Ba 2 LaF 7 crystals are precipitated from the glass matrix upon HT.The crystal precipitation is also confirmed by the XRD results (Fig. 3b).Thus, T p1 of the studied samples is the onset temperature of the melting of Ba 2 LaF 7 crystals.Furthermore, each T p2 value of the heat-treated samples is lower than that of the PG samples [19].
To study the first crystallization behavior of both PG and the heattreated PG samples, the crystallization enthalpies (ΔH) are obtained by calculating the area of the first exothermic peak as shown in Fig. 1(a  and b).The ΔH of PG samples decreases with increasing T homo (Fig. 1a), implying that the glass stability preventing the first crystallization event increases.This suggests that the degree of structural heterogeneity decreases with increasing T homo [20].It is seen in Fig. 1b that the HT makes the first crystallization peaks much smaller for samples HT-PG-1 and -2, and even disappeared for the samples HT-PG-3 and -4.This means that the formation of Ba 2 LaF 7 crystals is already completed in the samples with higher T homo (i.e., in HT-PG-3 and -4) during isothermal HT, and hence, the first crystallization peak does not exist during the subsequent DSC upscans.In contrast, the formation of Ba 2 LaF 7 crystals is not completed in the samples with lower T homo (i.e., in HT-PG-1 and -2) during HT, and thus, the first crystallization peaks appear during the DSC upscans despite they are tiny.In other words, the heat-treated PG samples exhibit higher crystallinity than the original PG samples.
To observe the crystallization event in PG samples, each sample was upscanned to 740, 700, 720, and 690 • C (above T c1 ), respectively, three times at 10 • C/min equal to the subsequent cooling rate.The determination of T g is shown in Fig. 2a-d, and the T g values obtained from the first upscans are given in Table 1.It is seen that the first DSC upscan gives a lower T g value than the second and third upscans, whereas the latter two upscans exhibit almost the same T g .In addition, the area (i.e., enthalpy) of the first crystallization peak decreases with the number of DSC cycles and then remains nearly constant.This could be explained by the increased stability of the glass matrix after the crystal precipitation during the first upscan to T max .In other words, during the dynamic heating process, a large portion of F − , La 3+ , and Ba 2+ ions are depleted from the glass matrix and then migrate to the existing Ba 2 LaF 7 crystals for further growth and also to the structural domains where nucleation can easily occur [21].However, during the 2nd and 3rd upscans, the remaining glass matrix will not crystallize since both thermodynamic and kinetic barriers for nucleation are quite large owing to the high network connectivity.

Structural analysis
Fig. 3a shows the XRD patterns of the four PG samples, where diffraction peaks are detected in PG-1, -2 and -3 samples, and become weaker with increasing T homo .These peaks are attributed to the facecentered cubic (FCC) Ba 2 LaF 7 crystal (PDF#48-0099) [9].However, two weak broad humps appear for PG-4 sample, implying its glassy nature [3,22].This suggests that the formation of Ba 2 LaF 7 crystals can be suppressed since the structural heterogeneity in the glass melt is drastically reduced by increasing T homo .Upon isothermal HT, the diffraction peaks of samples HT-PG-1, -2 and -3 become stronger.Interestingly, several weak peaks are observed for sample HT-PG-4, and they are ascribed to Ba 2 LaF 7 crystals, indicating that HT leads to an increase of Ba 2 LaF 7 content in PG samples [23].
Fig. 4a and b show the Raman spectra of both PG and the heattreated PG samples, respectively, which are normalized by the intensity of the peak at around 549 cm − 1 .It is seen in Fig. 4a that the Raman peak at 269 cm − 1 becomes weaker and even disappears with increasing T homo from 1450 to 1590 • C, i.e., for samples PG-1, -2, -3 and -4.The peak at 269 cm − 1 is ascribed to the stretching mode of Ba-F and La-F bonds in Ba 2 LaF 7 crystals, confirming the existence of Ba 2 LaF 7 crystals in PG-1, -2 and -3 samples.The Raman peaks in the wavelength range of 388-691 cm − 1 are assigned to symmetric stretching vibration modes of the interconnected [Si(Al)O 4 ] tetrahedra [24,25].Specifically, the peaks at 425, 550, and 691 cm − 1 are associated with the symmetric stretch of 5-, 4-and 3-fold rings, respectively [26,27].Interestingly, only two broad peaks at 474 and 720 cm − 1 appear in the spectrum of the sample homogenized at 1590 • C (i.e., sample PG-4), indicating the amorphous nature of PG-4 [28].This implies that the structural heterogeneity in PG samples increases with decreasing T homo , hence promoting formation of Ba 2 LaF 7 crystals.In addition, the symmetric stretching vibrational peaks are not detected in PG-4, indicating that the [Si(Al)O 4 ] tetrahedra are not well interconnected [28].
To investigate the effect of HT on both crystallization and local structure of PG samples, the Raman spectroscopy measurements were conducted on the heat-treated PG samples (Fig. 4b).Compared to the Raman spectra of PG-1, -2 and -3 in Fig. 4a, the Raman peak at 269 cm − 1 becomes stronger upon HT, indicating an increase of Ba 2 LaF 7 crystal content, and this agrees with the XRD evidence (Fig. 3).Moreover, the weak Raman peak at 269 cm − 1 appears in HT-PG-4 sample, meaning that Ba 2 LaF 7 crystals form upon HT.In addition, the Raman peaks related to the symmetric stretching vibration modes of [Si(Al)O 4 ] tetrahedra also occur in HT-PG-4 sample, which are consistent with those of HT-PG-1, -2 and -3 samples.This indicates that the dissociated glass network containing [Si(Al)O 4 ] tetrahedra is interconnected again in PG-4 upon HT, being confirmed by the increase of T g .Fig. 5 shows the TEM images of samples PG-1, PG-4, HT-PG-1, and HT-PG-4.It is seen in Fig. 5a that flower-like Ba 2 LaF 7 crystals are present in the glass matrix of PG-1, and their size increases upon HT.The distance between the adjacent fringes is determined to be 0.35 nm, which matches with the spacing of the (111) planes of Ba 2 LaF 7 crystals.Furthermore, new spherical Ba 2 LaF 7 nanocrystals emerge in HT-PG-1 sample (Fig. 5c).The TEM image in Fig. 5b confirms the glassy nature of PG-4 sample since there are no distinct lattice fringes.Through HT,  some ordered domains of about 10 nm (see the dashed circles) occur in glass matrix and exhibit the spacing of 0.35 nm between the (111) planes (Fig. 5d), indicating the formation of Ba 2 LaF 7 crystals (Inset of Fig. 5d).
To study the impact of both T homo and HT on the heterogeneity and crystallization, SEM analyses were performed on samples PG-1, PG-4, HT-PG-1 and HT-PG-4 as shown in Fig. 6.Some flower-like Ba 2 LaF 7 crystals of about 200 nm (white domains) are uniformly distributed in PG-1 (Fig. 6a).In contrast, no crystals are detectable in PG-4 (Fig. 6b).This clearly indicates that increasing the homogeneity by elevating T homo can suppress the formation of Ba 2 LaF 7 crystals.In Fig. 6c and d, we can observe the influence of HT on the crystallization behavior of PG samples.For instance, when sample PG-1 is subjected to HT at 643 • C (1.07T c1 ) for 2 hrs, flower-like Ba 2 LaF 7 crystals grow up to 300 nm, and at the same time, tiny spherical Ba 2 LaF 7 nanocrystals (about 15 nm in size) are precipitated from the glass matrix (see the image of HT-PG-1 in Fig. 6c).When the amorphous sample PG-4 is heat-treated at 642 • C (1.07T c1 ) for 2 hrs, some spherical Ba 2 LaF 7 nanocrystals of 10 nm in size are precipitated from the glass matrix of PG-4 sample (Fig. 6d) [29].This crystallization behavior is also confirmed by the TEM results in Fig. 5.
To reveal the effect of T homo on the microstructures of PG samples, we performed the 27 Al magic angle spinning (MAS) nuclear magnetic resonance (NMR) measurements (Fig. 7).It is seen that 27 Al MAS NMR spectra of all the samples feature the main resonance at about 60 ppm, which is attributed to the four-coordinated aluminum (Al(IV)) [30].In addition, the Al(IV) signal in PG-2, -3 and − -4 samples has the same chemical shift as PG-1, suggesting that the coordination environment of Al does not undergo detectable change with increasing T homo .It was previously reported that Al 3+ ions can be distributed both in the interface regions and in the structural network to form [Al(O,F) 4 ] and [Al (O) 4 ] tetrahedra, respectively [31].However, owing to the volatilization of fluorides (SiF 4 and NaF) and the decrease of the size of fluoride-rich domains with increasing T homo , some Al-F linkages in the interface regions might be replaced by the Al-O linkages.This means that the coordination environment of Al could be slightly altered though this change is not detectable in 27 Al NMR spectra.where A is the absorbance and THK is the sample thickness (in mm).Four main absorption peaks of Er 3+ ions can be observed at around 378, 486, 521, and 652 nm, which are attributed to the transitions from the ground state 4 I 15/2 to the excited states 4 G 11/2 , 4 F 7/2 , 2 H 11/2 , and 4 F 9/2 , respectively [17].In addition to the above-mentioned peaks, an absorption peak at around 978 nm is seen in each transmittance spectrum,   and it is associated with the transition of Yb 3+ ions: 2 F 7/2 → 2 F 5/2 .The light transmittance of PG samples increases with increasing T homo (Fig. 8a), reaching the maximum value of around 95% (per mm) in PG-4 sample.This is mainly because raising T homo decreases the size of the Ba 2 LaF 7 crystals and alters the chemical composition of the glass matrix.This reduces the differences in the refractive index (n) between the Ba 2 LaF 7 crystal (1.54@588 nm) and the remaining glass matrix, thereby suppressing the light scattering and hence leading to the highest light transmittance [32].

Optical properties
Interestingly, the light transmittance of HT-PG-1 and HT-PG-2 samples is even higher than that of samples PG-1 and PG-2 although both the fraction and the size of Ba 2 LaF 7 crystals increase upon HT.This anomalous phenomenon is in strong contrast to what is observed in traditional GCs [1].This means that HT is an effective way to enable the n of glass matrix to match that of Ba 2 LaF 7 crystals, and hence to reduce light scattering of the samples [33].Contrarily, the light transmittance of samples PG-3 and PG-4 is lowered by HT, and therefore the heat-treated PG-3 and PG-4 samples are less transparent than the untreated ones.It was reported that n is proportional to density (ρ) for inorganic glasses [28,34,35].The ρ values of the 8 studied samples are given in Table 2, demonstrating that ρ is lowered by HT.Thus, it can be inferred that the n value of glass matrix in the heat-treated samples decreases upon HT.Thus, the n differences between Ba 2 LaF 7 crystals and the glass matrix for PG-1 and PG-2 become smaller after HT, but PG-3 and PG-4 exhibit the opposite trend.
Fig. 9 shows the UC luminescence spectra in the wavelength range from 500 to 700 nm for the PG and the heat-treated PG samples, which are obtained by excitation of the 980 nm laser.The characteristic UC luminescence peaks at 521, 541, and 652 nm are attributed to the transitions from the excited states: 2 H 11/2 , 4 S 3/2 , and 4 F 9/2 to the ground state: 4 I 15/2 of Er 3+ ions, respectively.It is seen that the UC luminescence intensity of PG samples decreases with the increasing T homo (Fig. 9a and its inset) and the heat-treated PG samples exhibit higher UC luminescence intensity than PG samples (Fig. 9b and its inset).In addition, these energy levels in the heat-treated PG samples are more strongly split than PG samples [36].This behavior in heat-treated samples can be ascribed to the incorporation of more Er 3+ ions into the low phonon energy environment in Ba 2 LaF 7 crystals, reducing the probability of nonradiative relaxation and thus enhancing UC luminescence [37].
By comparing the data in Fig. 8 with those in Fig. 3, we can find an effective way for designing glass-ceramics (GCs) with optimum luminescent behaviors, i.e., regulating both T homo and HT conditions.As described above, the light transparency of PG samples increases with raising T homo , and however, the UC luminescence becomes weaker.Based on these two opposite trends, we can tailor the optical and photonic properties of the GCs, depending on their application fields.If we want to achieve the highest UC luminescence effect (i.e., sample PG-1 in Fig. 9a) in an Er-Yb doped oxyfluoride glass-ceramic, we should fabricate the precursor sample at the lowest T homo (i.e., melting at 1450 • C),   but this sample exhibits the lowest light transmittance (Fig. 8a).However, if we desire to obtain the glass with both high light transmittance (Fig. 8b) and strong UC luminescent effect (Fig. 9b), we can subject it to a proper HT (see sample HT-PG-1).In other words, a proper HT can enable enhancing both light transmittance and UC luminescence.Depending on the application field of such GCs materials, an optimum combination between light transmittance and UC luminescence should be realized by controlling both the T homo and the HT conditions.

Conclusion
Four Er 3+ -Yb 3+ ions doped oxyfluoride precursor glass (PG) samples were prepared from the sample composition (45SiO 2 -15Al 2 O 3 -12Na 2 O-21BaF 2 -7LaF 3 -0.5ErF 3 -1.0YbF 3 ) at four melt-homogenization temperatures (T homo ) (1450, 1500, 1550 and 1590 • C), respectively.The effect of T homo on crystallization behavior, structural evolution and optical properties of the studied samples were studied by performing DSC, SEM, XRD, TEM and NMR characterizations and by conducting optical analyses.We found that the crystallization enthalpy (ΔH) corresponding to the first crystallization peak decreases with increasing T homo , indicating a decrease of crystallization tendency in PG samples.This phenomenon could arise from a decrease of the fluoride-rich domains in glass melt, i. e., a decrease in structural heterogeneity.The light transmittance of PG samples increases with increasing T homo .This is attributed to the decrease in both the fraction and the size of Ba 2 LaF 7 crystals, and hence, to the lowered difference in the refractive index (n) between the Ba 2 LaF 7 crystals and the glass matrix.In addition, the Up-conversion (UC) luminescence of PG samples becomes weaker with increasing T homo .Interestingly, the light transmittance of samples PG-1 and PG-2 samples increased upon heat-treatment (HT), whereas that of PG-3 and PG-4 decreased.Moreover, the heat-treated PG samples exhibit higher crystallinity and stronger UC luminescence than PG samples.Thus, this work provided an effective approach, i.e., the approach optimizing both melthomogenization temperature and HT conditions, to develop the glassceramics with high optical and photonic performances for various applications.

Declaration of Competing Interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Fig. 4 .
Fig. 4. Normalized Raman spectra of PG (a) and the heat-treated PG (b) samples, respectively.Note: Raman spectra of PG-4 sample normalized by the maximum intensity; Raman spectra of other samples normalized by the intensity of the peak at around 549 cm -1 .

Fig. 8a and 10 (
Fig. 8a and b show the light transmittance spectra in the range of 275 to 1000 nm of PG and the heat-treated PG samples, respectively.The insets of the figures illustrate the change in transparency of the real samples.The light transmittance (%) spectra are derived from the optical absorption spectra through the equation:

Fig. 8 .
Fig. 8.Light transmittance spectra of PG (a) and the heat-treated PG samples (b), respectively.Insets: the optical photographs of the samples (maximum size: about 1 cm).

Table 2
Density (ρ) of PG and the heat-treated PG samples, respectively.