Structural and optical properties in Tm3+/Tm3+–Yb3+ doped NaLuF4 glass-ceramics

Int J Appl Glass Sci. 2021;12:485–496. wileyonlinelibrary.com/journal/ijag 485 Int J Appl Glass Sci. 2021;00:1–12. | 1 wileyonlinelibrary.com/journal/ijag Received: 20 February 2021 | Revised: 24 May 2021 | Accepted: 9 June 2021 DOI: 10.1111/ijag.16322 SPECIAL ISSUE ARTICLE Structural and optical properties in Tm3+/Tm3+– Yb3+ doped NaLuF4 glassceramics Jose J.Velázquez1 | RolindesBalda2,3 | JoaquinFernández4 | GiulioGorni5,6 | MercedesSedano5 | AliciaDurán5 | DušanGalusek1,7 | Maria J.Pascual5 This is an open access article under the terms of the Creat ive Commo ns Attri butio nNonCo mmerc ialNoDerivs License, which permits use and distribution in any medium, provided the original work is properly cited, the use is noncommercial and no modifications or adaptations are made. © 2021 The Authors. International Journal of Applied Glass Science published by American Ceramics Society and Wiley Periodicals LLC 1FunGlass, Alexander Dubček University of Trenčín, Trenčín, Slovakia 2Applied Physic Department I, Superior School of Engineering, Basque Country University, Bilbao, Spain 3Materials Physics Center, CSICUPV/ EHU, San Sebastian, Spain 4Donostia International Physics Center, San Sebastian, Spain 5Ceramic and Glass Institute (ICVCSIC), Madrid, Spain 6Cells, ALBA Synchrotron, Cerdanyola del Vallès, Spain 7Joint Glass Centre of the IIC SAS, TnUAD, and FChPT STU, Trenčín, Slovakia Correspondence Jose J. Velázquez, FunGlass, Alexander Dubček University of Trenčín, Študentská 2, 911 50 Trenčín, Slovakia. Email: [email protected] Funding information Euskal Herriko Unibertsitatea, Grant/Award Number: GIU17/014; European Regional Development Fund, Grant/Award Number: 313011R453 and 313011W442; Ministerio de Economía y Competitividad, Grant/ Award Number: MAT201787035C21P/- 2P; the European Union´s Horizon 2020 research and innovation programme, Grant/Award Number: 739566; Gobierno del Pais Vasco, Eusko Jaurlaritza, Grant/Award Number: PIBA201824 Abstract Transparent NaLuF4glassceramics (GCs) doped with Tm3+ and Tm3+/Yb3+ have been prepared by meltingquenching followed by thermal treatment at temperatures near the glass transition temperature. The crystallization process has been studied using Xray diffraction (XRD) and highresolution transmission electron microscopy (HRTEM). NaLuF4 nanocrystals (NCs) ranging 9– 30nm in size are the only crystalline phase, the crystal size increasing with the dopant concentration. Energy dispersive Xray (EDX) measurements confirm the Tm3+ and Yb3+ incorporation in the NCs. Optical characterization included the analysis of upconversion (UC) as well as the Nearinfrared (NIR) luminescence. NIR emission spectra of Tm3+ and Yb3+ in codoped samples confirmed an efficient energy transfer between both ions. No UC emissions are observed in Tm3+ singledoped glass and GCs. Yb3+ incorporation favors the Tm3+- Tm3+ UC processes resulting in Tm3+ blue, yellowishred and NIR UC emissions after excitation at 975nm. Blue UC emission is also observed in the codoped samples after Tm3+ excitation at 791nm. These effects were more evident for the GCs compared to the base glasses, confirming the RE ions incorporation in the NCs. As a result, these GCs can be used to tune the UC emission from NIR to blue by selective excitation. KEYWORDS crystallization, glass ceramics, glass manufacturing, optical properties, photo luminescence, processing, secondary VELÁZQUEZ et al.486 2 | LIŠKA ET AL. niobyl(3+) cation (NbO 3+ ), which enters deformed polyhedra (octahedra, tetragonal pyramids) with a coordinationcovalent bond to oxygen. 16,17 2 | METHOD 2 . 1 | Thermodynamic model of Shakhmatkin and Vedishcheva Especially in the field of oxide glasses, the thermodynamic model of Shakhmatkin and Vedishcheva has been successfully applied in previous years. 19 The SVTDM model considers glasses and melts to be an ideal solution formed by salt products of equilibrium chemical reactions of simple starting substances, e.g. oxides, halides, etc. These saltlike products have the identical stoichiometry as the crystalline compounds that exist in the phase diagram of the system under consideration. The model does not use any adjustable parameters. Only the molar Gibbs energies of pure crystalline phases and the analytical composition of the considered system are used as input parameters. The minimization of the system ' s Gibbs energy has to be performed with respect to the molar amount of each system species constrained by the overall system composition to reach the equilibrium system composition. 18,19 The presentday databases of thermodynamic properties comprising the molar Gibbs energies of various species (e.g. the FACT database 20,21 ) enable the routine construction of the Shakhmatkin and Vedishcheva model for many significant multicomponent glass systems. However, in many cases, mainly of nonsilicate multicomponent glass systems, the needed thermodynamic data are not found in any contemporary thermodynamic database. In our previous work, 6,22 we proposed a method for obtaining estimates of missing thermodynamic parameters by reproducing the structural information. 2 . 2 | Decomposition of Raman spectra based on glass composition from TD model The method of numerical analysis of Raman spectra was suggested by Malfait and his collaborators. 2325 The basic assumption of this method is that the Raman spectra are the sum of the partial Raman spectra (generated by the individual structural elements) multiplied by the abundance of these structural elements. The series of Raman spectra obtained for a series of glasses with different compositions span a linear vector space with a dimension given by a number of independent structural elements (i.e. structural elements that independently change their relative abundance) with different partial Raman spectra (PRS). Each measured spectrum is recorded with any scale, i.e. it is known except for the size of a multiplication factor. 2 . 3 | Multivariate curve resolution The multivariate curve resolution (MCR) method 19,2628 decomposes the set of experimental Raman spectra in the spectra of pseudo pure components (called loadings) and relative abundances of these components (called scores). It is essential to emphasize that MCR does not use any data of the system composition. The MCR result must be compared with the result of Malfait ' s spectra decomposition based on the TD model. 3 | EXPERIMENTAL PART Raman spectra were measured at room temperature using a confocal microscope (LabRam HR, Horiba Jobin– Yvon) with backscattering geometry and 532nm excitation line of Nd:YAG laser. For correction of the temperature dependent population of phonon levels, the intensities of Raman spectra were reduced using the Gammon– Shuker relation 29 : where 𝜈𝜈0 and 𝜈𝜈 are the frequency of excitation light and the Raman shift respectively. T is the thermodynamic temperature, k is the Boltzmann constant, and I ( ω ) is the measured Raman intensity. More details can be found in. 16,17 Raman spectra were multiplied by a constant, thus the maximum value of spectral intensity equals to one. 4 | RESULTS AND DISCUSSION On the basis of experimental structural data 16,17 ten system components were considered in the ZnONb 2 O 5 - P 2 O 5 thermodynamic model of Shakhmatkin and Vedishcheva – ZnO (Z), Nb 2 O 5 (Nb) , P 2 O 5 (P), ZnP 4 O 11 (ZP2), Zn(PO 3 ) 2 (ZP), Zn 2 P 2 O 7 (Z2P), Zn 3 (PO 4 ) 2 (Z3P), NbO(PO 3 ) 3 (1/2NbP3), (NbO) 4 (P 2 O 7 ) 3 (Nb2P3), and NbOPO 4 (1/2NbP). The equilibrium molar amounts of SVTDM system components recalculated for xNb 2 O 5 ·50ZnO·(50 - x)P 2 O 5 , ( x=0 , 1 , 3 , 5 , 7 , 10 , 12 ) glasses from the results of structural analysis published in 16,17 are summarized in Table 1 together with the structural groupings represented by the Q n units (i.e. tetrahedron PO 4 with n bridging oxygen atoms) of considered system components. The molar amounts reported in Table 1 correspond to little bit changed glass composition when compared with the prescribed one, i.e. xNb 2 O 5 ·50ZnO·(50 - x)P2O5 , ( x=0 , 1 , 3 , 5 , 7 , 10 , 12 ). This actual glass composition is given in Table 2 together with the glass transition temperature. The molar Gibbs energies of Z, Nb, P, and Z3P were taken from the FACT database. 20,21 The unknown molar Gibbs ( 1 ) I red (𝜈𝜈)=(𝜈𝜈0−𝜈𝜈)−4𝜈𝜈[1−exp. ( −h𝜈𝜈∕kT)]I(𝜈𝜈) 2 | VELÁZQUEZ et al. 1 | INTRODUCTION The search for new and more developed luminescent systems is still a challenge due to the wide range of optical applications that requires higher efficiencies, such as plasma display panels; white lightemitting diodes; fluorescent lamps, optical biosensors or biological markers, and laser cooling of solids.1– 6 Between these systems, rareearth (RE)- doped nanostructured oxyfluoride glassceramics (OxGCs) are very interesting materials due to the combination of the low phononenergy (300400cm−1) fluoride nanocrystals (NCs), with the mechanical, thermal, and chemical properties of oxide glasses.7,8 Moreover, the controlled crystallization mechanism allows obtaining NCs with sizes up to 50nm in which the RE ions are incorporated, dismissing the Rayleigh scattering and maintaining the transparency.9 Glassceramics based on LaF3, YF3, and RLnF4 (R=K, Na)10– 13 phases have been described as efficient luminescent materials. In particular, sodium lanthanide tetrafluorides, with the general formula NaLnF4, (with Ln=Gd, Y, La, or Lu) have been reported to be ideal hosts for RE3+ ions that can act as emitting centers and have been widely studied.14– 17 NaLuF4 exhibits two possible crystalline phases, cubic (α) and hexagonal (β), and it is an exceptional converter host for RE3+ ions.18,19 Nevertheless, the number of publications regarding NaLuF4 GCs obtained by meltingquenching is relatively scarce mainly due to the difficulty to obtain the desired crystalline phase with the proper degree of crystallization.20,21 GCs from the base composition 70SiO2– 8.5B2O3– 9.5Na2CO3– 6NaF6LuF3 (mol%) were developed by Chen et al.22 In this work, visible UC luminescence was obtained after excitation at 980nm of Yb3+ ions and subsequent energy transfer to Er3+ or Tm3+. The UC luminescence intensity of the GCs increase around 10000 and 2000 times, respectively, compared to the corresponding glass. Kück et al.23 described the optical properties of the Pr3+- doped pure Na7Lu13F46 crystalline phase and concluded that that system is suitable for use in cascadeemitting phosphors for Xe discharge lamps. From the different RE ions, Tm3+ ions exhibit interesting characteristics that in conjunction with the Yb3+ ions make them suitable candidates for laser operation at 0.8 and 1.5μm.24– 26 The Tm3+ ions have three metastable excited levels, 1D2, 1G4, and 3H4 that allows the emission in the blue, red, and infrared regions after pumping. Another characteristic is that the emission from the 3H4level to the ground state coincides with the window of the silica fibers near 0.8 μm. Moreover, many different energy transfer channels are possible between Yb3+ – Tm3+ due to the existence of significant resonance energy levels.27 In our previous works,28– 30 sodium lutetium fluoride glassceramics doped with ErF3 and codoped with ErF3YbF3 were studied in detail from a structural and optical perspective and for their suitable use in optical fibers. Transparent GCs based on a cubic solid solution NaxLu2x−1F7x−3 were successfully synthesized and the incorporation of RE3+ ions was confirmed by XRD and energy dispersive Xray spectroscopy (EDX).28 In the present work, the crystallization mechanism, and structural and optical properties of Tm3+- doped and Tm3+– Yb3+ codoped GCs are reported from the same base glass composition, 70SiO2– 5Al2O3– 2AlF3– 2Na2O– 18NaF– 3Lu2O3 (mol%). Particular attention is paid to the effect of codoping and RE3+- ion concentration on the crystallization kinetics and UC and NIR emissions. 2 | EXPERIMENTAL PROCEDURE Glasses with the composition 70SiO2– 5Al2O3– 2AlF3– 2Na2O– 18NaF– 3Lu2O3 (mol%) doped with xTmF3, x=0.1, 0.5 and codoped with yYbF3, y=0, 1, and 2 (mol%) have been prepared by the meltingquenching method. This composition was based on a previous study of the research group.30 The raw materials on reagent grade were SiO2sand (SaintGobain, 99.6%), Al2O3 (Panreac), Na2CO3 (Panreac, 99.5%), NaF (Panreac, 99.95%), Lu2O3 (Alfa Aesar, 99.9%), AlF3 (Alfa Aesar, 99.9%), TmF3 (Aldrich, 99.99%), and YbF3 (Alfa Aesar, 99.999%). The raw materials were mixed in a Turbula mixer for 3h. The batches of 100g. were calcined at 1250°C for 0.5h and melted at 1650°C for 1h in a Pt crucible with 125cm3 in volume covered with a Pt cap and then quenched on a brass mold. The melting process was repeated twice to improve the glass homogeneity. The glasses were annealed at 580°C for 30min to remove residual stresses. From now on, the obtained glass samples will be labelled as G0.1Tm, G0.5Tm, G0.5Tm1Yb, and G0.5Tm2Yb, for 0.1, 0.5 Tm3+ and 1, and 2 Yb3+ concentrations (in mol%), respectively; corresponding glassceramics will be denoted as GC. Glass specimens were heattreated at 600°C for 20h, using a heating rate of 10°C/ min to get the corresponding GCs. To perform the optical characterization, polished glass sheets (1cm x 1cm and thickness of 2mm) heattreated at 600°C during 20h were selected. One polished sheet of base glass of each composition and with the same shape was also used for comparison with the corresponding glassceramics. The thermal properties of the obtained glasses (glass transition temperature (Tg), softening point (Td), and thermal expansion coefficient (α)) were determined by VELÁZQUEZ et al.487 2 | LIŠKA ET AL. niobyl(3+) cation (NbO 3+ ), which enters deformed polyhedra (octahedra, tetragonal pyramids) with a coordinationcovalent bond to oxygen. 16,17 2 | METHOD 2 . 1 | Thermodynamic model of Shakhmatkin and Vedishcheva Especially in the field of oxide glasses, the thermodynamic model of Shakhmatkin and Vedishcheva has been successfully applied in previous years. 19 The SVTDM model considers glasses and melts to be an ideal solution formed by salt products of equilibrium chemical reactions of simple starting substances, e.g. oxides, halides, etc. These saltlike products have the identical stoichiometry as the crystalline compounds that exist in the phase diagram of the system under consideration. The model does not use any adjustable parameters. Only the molar Gibbs energies of pure crystalline phases and the analytical composition of the considered system are used as input parameters. The minimization of the system ' s Gibbs energy has to be performed with respect to the molar amount of each system species constrained by the overall system composition to reach the equilibrium system composition. 18,19 The presentday databases of thermodynamic properties comprising the molar Gibbs energies of various species (e.g. the FACT database 20,21 ) enable the routine construction of the Shakhmatkin and Vedishcheva model for many significant multicomponent glass systems. However, in many cases, mainly of nonsilicate multicomponent glass systems, the needed thermodynamic data are not found in any contemporary thermodynamic database. In our previous work, 6,22 we proposed a method for obtaining estimates of missing thermodynamic parameters by reproducing the structural information. 2 . 2 | Decomposition of Raman spectra based on glass composition from TD model The method of numerical analysis of Raman spectra was suggested by Malfait and his collaborators. 2325 The basic assumption of this method is that the Raman spectra are the sum of the partial Raman spectra (generated by the individual structural elements) multiplied by the abundance of these structural elements. The series of Raman spectra obtained for a series of glasses with different compositions span a linear vector space with a dimension given by a number of independent structural elements (i.e. structural elements that independently change their relative abundance) with different partial Raman spectra (PRS). Each measured spectrum is recorded with any scale, i.e. it is known except for the size of a multiplication factor. 2 . 3 | Multivariate curve resolution The multivariate curve resolution (MCR) method 19,2628 decomposes the set of experimental Raman spectra in the spectra of pseudo pure components (called loadings) and relative abundances of these components (called scores). It is essential to emphasize that MCR does not use any data of the system composition. The MCR result must be compared with the result of Malfait ' s spectra decomposition based on the TD model. 3 | EXPERIMENTAL PART Raman spectra were measured at room temperature using a confocal microscope (LabRam HR, Horiba Jobin– Yvon) with backscattering geometry and 532nm excitation line of Nd:YAG laser. For correction of the temperature dependent population of phonon levels, the intensities of Raman spectra were reduced using the Gammon– Shuker relation 29 : where 𝜈𝜈0 and 𝜈𝜈 are the frequency of excitation light and the Raman shift respectively. T is the thermodynamic temperature, k is the Boltzmann constant, and I ( ω ) is the measured Raman intensity. More details can be found in. 16,17 Raman spectra were multiplied by a constant, thus the maximum value of spectral intensity equals to one. 4 | RESULTS AND DISCUSSION On the basis of experimental structural data 16,17 ten system components were considered in the ZnONb 2 O 5 - P 2 O 5 thermodynamic model of Shakhmatkin and Vedishcheva – ZnO (Z), Nb 2 O 5 (Nb) , P 2 O 5 (P), ZnP 4 O 11 (ZP2), Zn(PO 3 ) 2 (ZP), Zn 2 P 2 O 7 (Z2P), Zn 3 (PO 4 ) 2 (Z3P), NbO(PO 3 ) 3 (1/2NbP3), (NbO) 4 (P 2 O 7 ) 3 (Nb2P3), and NbOPO 4 (1/2NbP). The equilibrium molar amounts of SVTDM system components recalculated for xNb 2 O 5 ·50ZnO·(50 - x)P 2 O 5 , ( x=0 , 1 , 3 , 5 , 7 , 10 , 12 ) glasses from the results of structural analysis published in 16,17 are summarized in Table 1 together with the structural groupings represented by the Q n units (i.e. tetrahedron PO 4 with n bridging oxygen atoms) of considered system components. The molar amounts reported in Table 1 correspond to little bit changed glass composition when compared with the prescribed one, i.e. xNb 2 O 5 ·50ZnO·(50 - x)P2O5 , ( x=0 , 1 , 3 , 5 , 7 , 10 , 12 ). This actual glass composition is given in Table 2 together with the glass transition temperature. The molar Gibbs energies of Z, Nb, P, and Z3P were taken from the FACT database. 20,21 The unknown molar Gibbs ( 1 ) I red (𝜈𝜈)=(𝜈𝜈0−𝜈𝜈)−4𝜈𝜈[1−exp. ( −h𝜈𝜈∕kT)]I(𝜈𝜈) | 3 VELÁZQUEZ et al. dilatometry. The dilatometric analysis was carried out using a Netzsch Gerätebau dilatometer, model 402 PC/1 with a heating rate of 5°C /min in the air; the estimated error of Tg is ±2°C. The thermal expansion coefficient was determined in the range 100°C– 450°C. The chemical composition of the glass samples, focussing on the Fluorine content, was analyzed by Xray fluorescence spectroscopy (XRF) using a PANalytical spectrometer. All oxides were determined employing the melting method with Li2B4O7, whereas elemental fluorine analysis was performed on pressed pellets of powdered glass (63μm, 0.3g) in order to avoid fluorine volatilization. Sieved glassceramic powders with the particle size <63μm were used for both XRD and TEM analysis. XRD measurements have been performed with an Xray diffractometer D8 ADVANCE (Bruker) equipped with a Lynx Eye detector. The patterns were collected with monochromatic CuKα1radiation (λ=1.54056Å). Diffractograms were collected in the range in the 10≤2θ≤70° using a step size of 0.02 ° and 1s. the acquisition time for each step. This technique was used to achieve two objectives; determination of the crystalline phase and estimation of the crystal size. For crystalline phases determination, the EVADiffractPLUS software was used meanwhile mean crystal sizes were resolved from the Scherrer equation with the corresponding errors.31 where θ is the angle of the diffraction maximum, β its full width at half maximum (FWHM), βi the instrumental broadening and λ is the wavelength. The θ and β parameters were obtained by fitting the most intense peaks to pseudoVoigt functions. Highresolution transmission electron microscopy (HRTEM), including scanning transmission microscopyhigh angle annular dark field (STEMHAADF) and Xray energy dispersive spectroscopy (EDXS), were recorded on a JEOL 2100 field emission gun transmission electron microscope operating at 200 kV and providing a point resolution of 0.19 nm. The TEM was equipped with an EDXS energy dispersive Xray spectrometer (INCA xsight, Oxford Instruments). EDX analysis was performed in STEM mode, with a probe size of ca. 1nm. Samples were prepared by dispersing the fine powder in ethanol with ultrasonic agitation; a droplet of the suspension was put on a copper holey carbon grid. Conventional transmission spectra were performed with a Cary 5spectrophotometer. The steadystate emission measurements were made with a Tisapphire ring laser (0.4cm−1linewidth) as exciting light. The fluorescence was analyzed with a 0.25 monochromator, and the signal was detected by an extended IR Hamamatsu H10330A75 photomultiplier and finally amplified by a standard lockin technique. Visible emission was detected by a Hamamatsu R636 photomultiplier. All measurements were performed at room temperature. Accuracy of emission measurements is between 3% and 5% depending on the experimental conditions. 3 | RESULTS AND DISCUSSION 3.1 | Thermal, XRD, and HRTEM analysis Transparent glasses were obtained after melting for the different dopant concentrations. As expected, the corresponding XRD patterns (not shown) did not reveal the presence of any crystalline phase. Thermal properties were characterized by dilatometric analysis and results are shown in Table 1. Tg does not show any constant increasing tendency with the dopant content, but the doped and codoped samples present Tg values ranging from 566°C to 589°C. At the same time, the thermal expansion coefficient (α) has the same value as in the undoped glass, 7.9°C−1, decreasing for the codoped glasses down to 7.3°C−1 with the increasing amount of Yb3+ ions. Similar behaviour was found by the authors for the same composition but (1) ∅= 0.94 ⋅ 𝜆 cos𝜃 √ 𝛽2−𝛽2 i , TABLE 1 Fluorine loss (%), Glass transition temperature (Tg), softening point temperature (Td) and thermal expansion coefficient (α) of undoped and doped glasses. Mean NCs size and thermal expansion coefficient (α) of corresponding GCs Glass F loss (%) α 10−6 (°C−1) ±0.5 Tg (°C) ±2 Td (°C) ±5 GCs NCs size (nm) ±1 α 10−6 (°C−1) ±0.5 Undoped 48 7.9 579 663 Undoped 9 8.6 G0.1Tm 42 7.9 583 664 GC0.1Tm 12 8.2 G0.5Tm 45 7.9 566 648 GC0.5Tm 15 8.6 G0.5Tm−1Yb 42 7.6 589 685 GC0.5Tm−1Yb 27 7.9 G0.5Tm−2Yb 52 7.3 572 673 GC0.5Tm−2Yb 30 9.0 VELÁZQUEZ et al.488 2 | LIŠKA ET AL. niobyl(3+) cation (NbO 3+ ), which enters deformed polyhedra (octahedra, tetragonal pyramids) with a coordinationcovalent bond to oxygen. 16,17 2 | METHOD 2 . 1 | Thermodynamic model of Shakhmatkin and Vedishcheva Especially in the field of oxide glasses, the thermodynamic model of Shakhmatkin and Vedishcheva has been successfully applied in previous years. 19 The SVTDM model considers glasses and melts to be an ideal solution formed by salt products of equilibrium chemical reactions of simple starting substances, e.g. oxides, halides, etc. These saltlike products have the identical stoichiometry as the crystalline compounds that exist in the phase diagram of the system under consideration. The model does not use any adjustable parameters. Only the molar Gibbs energies of pure crystalline phases and the analytical composition of the considered system are used as input parameters. The minimization of the system ' s Gibbs energy has to be performed with respect to the molar amount of each system species constrained by the overall system composition to reach the equilibrium system composition. 18,19 The presentday databases of thermodynamic properties comprising the molar Gibbs energies of various species (e.g. the FACT database 20,21 ) enable the routine construction of the Shakhmatkin and Vedishcheva model for many significant multicomponent glass systems. However, in many cases, mainly of nonsilicate multicomponent glass systems, the needed thermodynamic data are not found in any contemporary thermodynamic database. In our previous work, 6,22 we proposed a method for obtaining estimates of missing thermodynamic parameters by reproducing the structural information. 2 . 2 | Decomposition of Raman spectra based on glass composition from TD model The method of numerical analysis of Raman spectra was suggested by Malfait and his collaborators. 2325 The basic assumption of this method is that the Raman spectra are the sum of the partial Raman spectra (generated by the individual structural elements) multiplied by the abundance of these structural elements. The series of Raman spectra obtained for a series of glasses with different compositions span a linear vector space with a dimension given by a number of independent structural elements (i.e. structural elements that independently change their relative abundance) with different partial Raman spectra (PRS). Each measured spectrum is recorded with any scale, i.e. it is known except for the size of a multiplication factor. 2 . 3 | Multivariate curve resolution The multivariate curve resolution (MCR) method 19,2628 decomposes the set of experimental Raman spectra in the spectra of pseudo pure components (called loadings) and relative abundances of these components (called scores). It is essential to emphasize that MCR does not use any data of the system composition. The MCR result must be compared with the result of Malfait ' s spectra decomposition based on the TD model. 3 | EXPERIMENTAL PART Raman spectra were measured at room temperature using a confocal microscope (LabRam HR, Horiba Jobin– Yvon) with backscattering geometry and 532nm excitation line of Nd:YAG laser. For correction of the temperature dependent population of phonon levels, the intensities of Raman spectra were reduced using the Gammon– Shuker relation 29 : where 𝜈𝜈0 and 𝜈𝜈 are the frequency of excitation light and the Raman shift respectively. T is the thermodynamic temperature, k is the Boltzmann constant, and I ( ω ) is the measured Raman intensity. More details can be found in. 16,17 Raman spectra were multiplied by a constant, thus the maximum value of spectral intensity equals to one. 4 | RESULTS AND DISCUSSION On the basis of experimental structural data 16,17 ten system components were considered in the ZnONb 2 O 5 - P 2 O 5 thermodynamic model of Shakhmatkin and Vedishcheva – ZnO (Z), Nb 2 O 5 (Nb) , P 2 O 5 (P), ZnP 4 O 11 (ZP2), Zn(PO 3 ) 2 (ZP), Zn 2 P 2 O 7 (Z2P), Zn 3 (PO 4 ) 2 (Z3P), NbO(PO 3 ) 3 (1/2NbP3), (NbO) 4 (P 2 O 7 ) 3 (Nb2P3), and NbOPO 4 (1/2NbP). The equilibrium molar amounts of SVTDM system components recalculated for xNb 2 O 5 ·50ZnO·(50 - x)P 2 O 5 , ( x=0 , 1 , 3 , 5 , 7 , 10 , 12 ) glasses from the results of structural analysis published in 16,17 are summarized in Table 1 together with the structural groupings represented by the Q n units (i.e. tetrahedron PO 4 with n bridging oxygen atoms) of considered system components. The molar amounts reported in Table 1 correspond to little bit changed glass composition when compared with the prescribed one, i.e. xNb 2 O 5 ·50ZnO·(50 - x)P2O5 , ( x=0 , 1 , 3 , 5 , 7 , 10 , 12 ). This actual glass composition is given in Table 2 together with the glass transition temperature. The molar Gibbs energies of Z, Nb, P, and Z3P were taken from the FACT database. 20,21 The unknown molar Gibbs ( 1 ) I red (𝜈𝜈)=(𝜈𝜈0−𝜈𝜈)−4𝜈𝜈[1−exp. ( −h𝜈𝜈∕kT)]I(𝜈𝜈) 4 | VELÁZQUEZ et al. doped and codoped with Er3+ and Yb3+ ions.28 Moreover, Sroda, Gorni and de PablosMartin et al. also observed a similar effect in different oxyfluoride glasses.16,32,33 The thermal expansion coefficient of GCs samples was also measured and the values are shown in Table 1. In general, all GCs show higher values than the glass samples but in the range of error. This behavior is opposite to that observed by Reben and Środa in PbF2NCs glass, where the addition of fluorine decreased the thermal expansion coefficient and weakened the structure of the glass.34 In our case, it seems that the amount of fluorine incorporation in the structure due to the addition of dopant is not sufficient to affect the structure of the glassy phase in the GCs. Table 1 also presents the estimated F loss of each glass obtained from the chemical analysis of the corresponding glasses. From these results, the fluorine loss varies between 42 and 52wt%. The chemical analysis of the other components (not shown) displays values roughly in agreement with the nominal ones. These values indicate that a high fluorine loss occurs during the melting process due to the high melting temperatures involved, at around 1650°C, as is expected in these compositions. Taking into account the effect of the dopant addition on the glass Tg and our previous results,28– 30,35 the heat treatment temperature and time were selected as 600°C during 20h. Figure 1A shows the XRD patterns of the corresponding GCs samples. The results for the different dopant concentrations are almost identical with the diffraction maxima that are assigned to sodium lutetium fluoride cubic solid solution with a general formula NaxLu2x1F7x3 as previously reported in.28,30 The peaks become narrower and more intense when increasing dopant content indicating that the NCs growth is favoured. Moreover, the magnified region correspondings to the (111) crystallographic plane, Figure 1B, shows a constant shift towards lower angles with the increasing content of RE3+ dopant ions, mainly due to incorporation of the larger Tm3+ (ionic radius, rion=0.88Å) and Yb3+ ions (rion=0.87Å) that substitute the Lu3+ (rion=0.85Å) increasing the unit cell volume associated to the cubic phase. Table 1 also shows the mean size of NCs calculated by using the Scherrer equation. The increasing dopant content results in increasing size, varying from 9 to 30nm. In this case, the dopant content plays a notable role in crystal growth as observed for other GCs.36,37 The increasing NCs size is also related to the close relationship between the crystallization activation energy and doping level. Crystallization activation energy decreases with the increasing doping level.38 Thus, samples with higher dopant concentration are more susceptible to crystallize. HRTEM together with EDX techniques were used to look deeply into the nanostructure and dopants distribution, see Figures 2 and 3. HRTEM micrographs of the glasses, Figure 2A,B, show the typical welldispersed phaseseparated amorphous droplets. These phaseseparated droplets are the precursors of the NaLuF4NCs that will form after adequate heat treatment28 observed in Figure 2C,D. Moreover, HRTEM micrographs of GC0.5Tm– 2Yb show a broad crystal distribution that fits well with a pseudoVoight function centered at 13nm (inset Figure 2C), confirming the results obtained by XRD diffraction. It should be noted that unlike the hexagonal LaF3 oxyfluoride NCs,33,39 cubic NaLuF4NCs form from each phaseseparated droplet, this is typical behavior of alkali lanthanide tetrafluoride NCs.13,40,41 The size of the NCs is thus similar to the size of the precursor phaseseparated FIGURE 1 (A) XRD patterns for undoped, xTm3+- doped (x=0.1, 0.5mol%) and 0.5Tm3+– yYb3+- codoped (y=1, 2mol%) GCs. (B)Magnified regions of (111) peak of the corresponding GCs VELÁZQUEZ et al.489 2 | LIŠKA ET AL. niobyl(3+) cation (NbO 3+ ), which enters deformed polyhedra (octahedra, tetragonal pyramids) with a coordinationcovalent bond to oxygen. 16,17 2 | METHOD 2 . 1 | Thermodynamic model of Shakhmatkin and Vedishcheva Especially in the field of oxide glasses, the thermodynamic model of Shakhmatkin and Vedishcheva has been successfully applied in previous years. 19 The SVTDM model considers glasses and melts to be an ideal solution formed by salt products of equilibrium chemical reactions of simple starting substances, e.g. oxides, halides, etc. These saltlike products have the identical stoichiometry as the crystalline compounds that exist in the phase diagram of the system under consideration. The model does not use any adjustable parameters. Only the molar Gibbs energies of pure crystalline phases and the analytical composition of the considered system are used as input parameters. The minimization of the system ' s Gibbs energy has to be performed with respect to the molar amount of each system species constrained by the overall system composition to reach the equilibrium system composition. 18,19 The presentday databases of thermodynamic properties comprising the molar Gibbs energies of various species (e.g. the FACT database 20,21 ) enable the routine construction of the Shakhmatkin and Vedishcheva model for many significant multicomponent glass systems. However, in many cases, mainly of nonsilicate multicomponent glass systems, the needed thermodynamic data are not found in any contemporary thermodynamic database. In our previous work, 6,22 we proposed a method for obtaining estimates of missing thermodynamic parameters by reproducing the structural information. 2 . 2 | Decomposition of Raman spectra based on glass composition from TD model The method of numerical analysis of Raman spectra was suggested by Malfait and his collaborators. 2325 The basic assumption of this method is that the Raman spectra are the sum of the partial Raman spectra (generated by the individual structural elements) multiplied by the abundance of these structural elements. The series of Raman spectra obtained for a series of glasses with different compositions span a linear vector space with a dimension given by a number of independent structural elements (i.e. structural elements that independently change their relative abundance) with different partial Raman spectra (PRS). Each measured spectrum is recorded with any scale, i.e. it is known except for the size of a multiplication factor. 2 . 3 | Multivariate curve resolution The multivariate curve resolution (MCR) method 19,2628 decomposes the set of experimental Raman spectra in the spectra of pseudo pure components (called loadings) and relative abundances of these components (called scores). It is essential to emphasize that MCR does not use any data of the system composition. The MCR result must be compared with the result of Malfait ' s spectra decomposition based on the TD model. 3 | EXPERIMENTAL PART Raman spectra were measured at room temperature using a confocal microscope (LabRam HR, Horiba Jobin– Yvon) with backscattering geometry and 532nm excitation line of Nd:YAG laser. For correction of the temperature dependent population of phonon levels, the intensities of Raman spectra were reduced using the Gammon– Shuker relation 29 : where 𝜈𝜈0 and 𝜈𝜈 are the frequency of excitation light and the Raman shift respectively. T is the thermodynamic temperature, k is the Boltzmann constant, and I ( ω ) is the measured Raman intensity. More details can be found in. 16,17 Raman spectra were multiplied by a constant, thus the maximum value of spectral intensity equals to one. 4 | RESULTS AND DISCUSSION On the basis of experimental structural data 16,17 ten system components were considered in the ZnONb 2 O 5 - P 2 O 5 thermodynamic model of Shakhmatkin and Vedishcheva – ZnO (Z), Nb 2 O 5 (Nb) , P 2 O 5 (P), ZnP 4 O 11 (ZP2), Zn(PO 3 ) 2 (ZP), Zn 2 P 2 O 7 (Z2P), Zn 3 (PO 4 ) 2 (Z3P), NbO(PO 3 ) 3 (1/2NbP3), (NbO) 4 (P 2 O 7 ) 3 (Nb2P3), and NbOPO 4 (1/2NbP). The equilibrium molar amounts of SVTDM system components recalculated for xNb 2 O 5 ·50ZnO·(50 - x)P 2 O 5 , ( x=0 , 1 , 3 , 5 , 7 , 10 , 12 ) glasses from the results of structural analysis published in 16,17 are summarized in Table 1 together with the structural groupings represented by the Q n units (i.e. tetrahedron PO 4 with n bridging oxygen atoms) of considered system components. The molar amounts reported in Table 1 correspond to little bit changed glass composition when compared with the prescribed one, i.e. xNb 2 O 5 ·50ZnO·(50 - x)P2O5 , ( x=0 , 1 , 3 , 5 , 7 , 10 , 12 ). This actual glass composition is given in Table 2 together with the glass transition temperature. The molar Gibbs energies of Z, Nb, P, and Z3P were taken from the FACT database. 20,21 The unknown molar Gibbs ( 1 ) I red (𝜈𝜈)=(𝜈𝜈0−𝜈𝜈)−4𝜈𝜈[1−exp. ( −h𝜈𝜈∕kT)]I(𝜈𝜈) | 5 VELÁZQUEZ et al. regions, and then they grow more with the adequate heat treatment in temperature and time.28 Moreover, the Fast Fourier Transformation (FFT) from HRTEM micrographs revealed crystalline structures with interplanar distances of 0.31nm, attributed to (111) planes of the sodium lutetium fluoride cubic phase. To further study, the elemental composition and the incorporation of the RE3+ ions in the NCs, EDX analysis in STEM mode was performed for the Tm3+- doped and Tm3+- Yb3+ codoped GCs (Figure 3). This indicates the nanocrystals are Naand Lurich, with the Tm3+ and Yb3+ ions concentrated in the nanocrystals, similarly to previously observed for different fluoride NCs and RE3+ dopants.14,16,42 The presence of the Tm3+ and Yb3+ ions in the NCs supports the results obtained by XRD patterns in which the diffraction maxima are shifted to lower angles, suggesting the incorporation of the RE3+ ions in the crystal structure of the NCs. 3.2 | Optical properties 3.2.1 | Transmission spectra The transmission spectra were obtained for all samples in the 300– 2000nm range. As an example, Figure 4shows the spectra as a function of wavelength for the glass and GC samples codoped with 0.5TmF31YbF3 (in mol%). The spectra show the bands corresponding to the transitions starting from the 3H6ground state to the different higher levels 1D2, 1G4, 3F2, 3F3, 3H4, 3H5, and 3F4 of Tm3+ together with the 2F7/2 → 2F5/2 absorption of Yb3+. As can be seen, the samples are transparent with a transmittance as high as 91% and 89% for the glass and glassceramic, respectively. After the heat treatment, the absorption edge is shifted to longer wavelengths in the GC sample. This redshift has been attributed to the scattering of shortwavelength light by the NCs present in the GC sample.43,44 The spectra obtained for the glass and glassceramic samples doped with 2mol% YbF3 are similar, except for the double Yb3+ band intensity. 3.2.2 | NIR luminescence The nearinfrared emission in the 900– 2200nm spectral range was obtained for the single doped samples at room temperature by exciting at 791nm in resonance with the 3H4 (Tm3+) level. The codoped samples were also excited at 975nm in resonance with the 2F7/2 → 2F5/2 absorption of Yb3+. This wavelength corresponds to the maximum absorption of Yb3+ ions in these matrices. The fluorescence FIGURE 2 (A) and (C) HRTEM micrograph of 0.5Tm3+ doped glass and GC treated at 600°C20h; size distribution of the NCs is shown as insets in (C). (B) Magnified image of phase separation droplets. (D) HRTEM micrograph of 0.5Tm3+- doped GCs with corresponding SAED image shown as an inset (A) (B) (C) (D) VELÁZQUEZ et al.490 2 | LIŠKA ET AL. niobyl(3+) cation (NbO 3+ ), which enters deformed polyhedra (octahedra, tetragonal pyramids) with a coordinationcovalent bond to oxygen. 16,17 2 | METHOD 2 . 1 | Thermodynamic model of Shakhmatkin and Vedishcheva Especially in the field of oxide glasses, the thermodynamic model of Shakhmatkin and Vedishcheva has been successfully applied in previous years. 19 The SVTDM model considers glasses and melts to be an ideal solution formed by salt products of equilibrium chemical reactions of simple starting substances, e.g. oxides, halides, etc. These saltlike products have the identical stoichiometry as the crystalline compounds that exist in the phase diagram of the system under consideration. The model does not use any adjustable parameters. Only the molar Gibbs energies of pure crystalline phases and the analytical composition of the considered system are used as input parameters. The minimization of the system ' s Gibbs energy has to be performed with respect to the molar amount of each system species constrained by the overall system composition to reach the equilibrium system composition. 18,19 The presentday databases of thermodynamic properties comprising the molar Gibbs energies of various species (e.g. the FACT database 20,21 ) enable the routine construction of the Shakhmatkin and Vedishcheva model for many significant multicomponent glass systems. However, in many cases, mainly of nonsilicate multicomponent glass systems, the needed thermodynamic data are not found in any contemporary thermodynamic database. In our previous work, 6,22 we proposed a method for obtaining estimates of missing thermodynamic parameters by reproducing the structural information. 2 . 2 | Decomposition of Raman spectra based on glass composition from TD model The method of numerical analysis of Raman spectra was suggested by Malfait and his collaborators. 2325 The basic assumption of this method is that the Raman spectra are the sum of the partial Raman spectra (generated by the individual structural elements) multiplied by the abundance of these structural elements. The series of Raman spectra obtained for a series of glasses with different compositions span a linear vector space with a dimension given by a number of independent structural elements (i.e. structural elements that independently change their relative abundance) with different partial Raman spectra (PRS). Each measured spectrum is recorded with any scale, i.e. it is known except for the size of a multiplication factor. 2 . 3 | Multivariate curve resolution The multivariate curve resolution (MCR) method 19,2628 decomposes the set of experimental Raman spectra in the spectra of pseudo pure components (called loadings) and relative abundances of these components (called scores). It is essential to emphasize that MCR does not use any data of the system composition. The MCR result must be compared with the result of Malfait ' s spectra decomposition based on the TD model. 3 | EXPERIMENTAL PART Raman spectra were measured at room temperature using a confocal microscope (LabRam HR, Horiba Jobin– Yvon) with backscattering geometry and 532nm excitation line of Nd:YAG laser. For correction of the temperature dependent population of phonon levels, the intensities of Raman spectra were reduced using the Gammon– Shuker relation 29 : where 𝜈𝜈0 and 𝜈𝜈 are the frequency of excitation light and the Raman shift respectively. T is the thermodynamic temperature, k is the Boltzmann constant, and I ( ω ) is the measured Raman intensity. More details can be found in. 16,17 Raman spectra were multiplied by a constant, thus the maximum value of spectral intensity equals to one. 4 | RESULTS AND DISCUSSION On the basis of experimental structural data 16,17 ten system components were considered in the ZnONb 2 O 5 - P 2 O 5 thermodynamic model of Shakhmatkin and Vedishcheva – ZnO (Z), Nb 2 O 5 (Nb) , P 2 O 5 (P), ZnP 4 O 11 (ZP2), Zn(PO 3 ) 2 (ZP), Zn 2 P 2 O 7 (Z2P), Zn 3 (PO 4 ) 2 (Z3P), NbO(PO 3 ) 3 (1/2NbP3), (NbO) 4 (P 2 O 7 ) 3 (Nb2P3), and NbOPO 4 (1/2NbP). The equilibrium molar amounts of SVTDM system components recalculated for xNb 2 O 5 ·50ZnO·(50 - x)P 2 O 5 , ( x=0 , 1 , 3 , 5 , 7 , 10 , 12 ) glasses from the results of structural analysis published in 16,17 are summarized in Table 1 together with the structural groupings represented by the Q n units (i.e. tetrahedron PO 4 with n bridging oxygen atoms) of considered system components. The molar amounts reported in Table 1 correspond to little bit changed glass composition when compared with the prescribed one, i.e. xNb 2 O 5 ·50ZnO·(50 - x)P2O5 , ( x=0 , 1 , 3 , 5 , 7 , 10 , 12 ). This actual glass composition is given in Table 2 together with the glass transition temperature. The molar Gibbs energies of Z, Nb, P, and Z3P were taken from the FACT database. 20,21 The unknown molar Gibbs ( 1 ) I red (𝜈𝜈)=(𝜈𝜈0−𝜈𝜈)−4𝜈𝜈[1−exp. ( −h𝜈𝜈∕kT)]I(𝜈𝜈) 6 | VELÁZQUEZ et al. spectra corresponding to the single doped samples doped with 0.1 and 0.5Tm3+ (in mol%) are shown in Figure 5. The spectra show a strong emission band centered around 1650nm which corresponds to the 3F4 → 3H6 transition together with a less intense emission band centered around 1450nm and corresponding to the 3H4 → 3F4 transition. The longwavelength tail of the 3F4 → 3H6 emission is not completely observed due to the upper limit of the detector. Similar spectra are obtained for the glass and glassceramic samples, but the ratio of the emission intensity of the transition 3H4 → 3F4 to that of the 3F4 → 3H6 decreases with increasing Tm3+ concentration as is shown in Figure 5. This reduction in the intensity of the 1450nm emission with concentration has been attributed to crossrelaxation between 3H4 → 3F4 and 3F4 → 3H6 transitions that enhance the latter.45,46 The emission spectra of the codoped samples obtained under the same experimental conditions by exciting at 791nm show, in addition to the Tm3+ emissions around 1450 and 1650 nm, the Yb3+ (2F5/2 → 2F7/2) emission around 1000nm. The presence of the Yb3+ emission after excitation of Tm3+ ions confirms the energy transfer from FIGURE 3 (A) and (C) STEM image of NaLuF4 NCs in GC0.5Tm3+ and GC0.5Tm3+– 2Yb3+. (B) and (D) EDX of a line scan crossing same aLuF4NCs in GC0.5Tm3+ and GC0.5Tm3+– 2Yb3+ (A) (B) (C) (D) VELÁZQUEZ et al.491 2 | LIŠKA ET AL. niobyl(3+) cation (NbO 3+ ), which enters deformed polyhedra (octahedra, tetragonal pyramids) with a coordinationcovalent bond to oxygen. 16,17 2 | METHOD 2 . 1 | Thermodynamic model of Shakhmatkin and Vedishcheva Especially in the field of oxide glasses, the thermodynamic model of Shakhmatkin and Vedishcheva has been successfully applied in previous years. 19 The SVTDM model considers glasses and melts to be an ideal solution formed by salt products of equilibrium chemical reactions of simple starting substances, e.g. oxides, halides, etc. These saltlike products have the identical stoichiometry as the crystalline compounds that exist in the phase diagram of the system under consideration. The model does not use any adjustable parameters. Only the molar Gibbs energies of pure crystalline phases and the analytical composition of the considered system are used as input parameters. The minimization of the system ' s Gibbs energy has to be performed with respect to the molar amount of each system species constrained by the overall system composition to reach the equilibrium system composition. 18,19 The presentday databases of thermodynamic properties comprising the molar Gibbs energies of various species (e.g. the FACT database 20,21 ) enable the routine construction of the Shakhmatkin and Vedishcheva model for many significant multicomponent glass systems. However, in many cases, mainly of nonsilicate multicomponent glass systems, the needed thermodynamic data are not found in any contemporary thermodynamic database. In our previous work, 6,22 we proposed a method for obtaining estimates of missing thermodynamic parameters by reproducing the structural information. 2 . 2 | Decomposition of Raman spectra based on glass composition from TD model The method of numerical analysis of Raman spectra was suggested by Malfait and his collaborators. 2325 The basic assumption of this method is that the Raman spectra are the sum of the partial Raman spectra (generated by the individual structural elements) multiplied by the abundance of these structural elements. The series of Raman spectra obtained for a series of glasses with different compositions span a linear vector space with a dimension given by a number of independent structural elements (i.e. structural elements that independently change their relative abundance) with different partial Raman spectra (PRS). Each measured spectrum is recorded with any scale, i.e. it is known except for the size of a multiplication factor. 2 . 3 | Multivariate curve resolution The multivariate curve resolution (MCR) method 19,2628 decomposes the set of experimental Raman spectra in the spectra of pseudo pure components (called loadings) and relative abundances of these components (called scores). It is essential to emphasize that MCR does not use any data of the system composition. The MCR result must be compared with the result of Malfait ' s spectra decomposition based on the TD model. 3 | EXPERIMENTAL PART Raman spectra were measured at room temperature using a confocal microscope (LabRam HR, Horiba Jobin– Yvon) with backscattering geometry and 532nm excitation line of Nd:YAG laser. For correction of the temperature dependent population of phonon levels, the intensities of Raman spectra were reduced using the Gammon– Shuker relation 29 : where 𝜈𝜈0 and 𝜈𝜈 are the frequency of excitation light and the Raman shift respectively. T is the thermodynamic temperature, k is the Boltzmann constant, and I ( ω ) is the measured Raman intensity. More details can be found in. 16,17 Raman spectra were multiplied by a constant, thus the maximum value of spectral intensity equals to one. 4 | RESULTS AND DISCUSSION On the basis of experimental structural data 16,17 ten system components were considered in the ZnONb 2 O 5 - P 2 O 5 thermodynamic model of Shakhmatkin and Vedishcheva – ZnO (Z), Nb 2 O 5 (Nb) , P 2 O 5 (P), ZnP 4 O 11 (ZP2), Zn(PO 3 ) 2 (ZP), Zn 2 P 2 O 7 (Z2P), Zn 3 (PO 4 ) 2 (Z3P), NbO(PO 3 ) 3 (1/2NbP3), (NbO) 4 (P 2 O 7 ) 3 (Nb2P3), and NbOPO 4 (1/2NbP). The equilibrium molar amounts of SVTDM system components recalculated for xNb 2 O 5 ·50ZnO·(50 - x)P 2 O 5 , ( x=0 , 1 , 3 , 5 , 7 , 10 , 12 ) glasses from the results of structural analysis published in 16,17 are summarized in Table 1 together with the structural groupings represented by the Q n units (i.e. tetrahedron PO 4 with n bridging oxygen atoms) of considered system components. The molar amounts reported in Table 1 correspond to little bit changed glass composition when compared with the prescribed one, i.e. xNb 2 O 5 ·50ZnO·(50 - x)P2O5 , ( x=0 , 1 , 3 , 5 , 7 , 10 , 12 ). This actual glass composition is given in Table 2 together with the glass transition temperature. The molar Gibbs energies of Z, Nb, P, and Z3P were taken from the FACT database. 20,21 The unknown molar Gibbs ( 1 ) I red (𝜈𝜈)=(𝜈𝜈0−𝜈𝜈)−4𝜈𝜈[1−exp. ( −h𝜈𝜈∕kT)]I(𝜈𝜈) | 7 VELÁZQUEZ et al. Tm3+ to Yb3+ ions. As an example, Figure 6 displays the emission spectra for the codoped samples with 0.5Tm1Yb. After excitation of level 3H4, the energy transfer to the 2F5/2state of Yb3+ ions occurs. Although the 3H4 and 3H5levels of Tm3+ are located at higher and lower energies respectively, a nonresonant phononassisted energy transfer (PAET) between both ions takes place.47,48 The NIR luminescence intensities of the Tm3+ emissions measured under 791nm (Tm3+) excitation, in the single and codoped samples with 1mol% of Yb3+ (Figures 5B and 6) are quite similar for the 3F4→3H6 transition and nearly independent of the presence of Yb3+. However, the intensity of the 3H4→3F4 transition slightly decreases with the addition of Yb3+ ions probably due to the Tm3+→Yb3+ energy transfer. Once the 3H4level is populated by the absorption of 791nm photons, energy is transferred to the 2F5state of Yb3+. Moreover, the presence of energy transfer from Yb3+ to Tm3+ is demonstrated by the observation of the 3F4 → 3H6Tm3+ emission after excitation at 975nm in resonance with the Yb3+ absorption at 975nm, as shown in Figure 7. This figure shows the spectra of the glass and glassceramic samples codoped with 1 and 2mol% of Yb3+. As can be seen, the emission intensity for the codoped samples with 2mol% of Yb3+ is around twice those of the samples with 1mol%. Moreover, compared to the emission obtained under Tm3+ excitation at 791nm, the intensity of the 3F4 → 3H6 increases around four times when exciting Yb3+ ions. This is in agreement with the high absorption crosssection of Yb3+ at 975nm if compared with the one of Tm3+ at 791nm. The 2F5/2 (Yb3+)→3H5(Tm3+) energy transfer followed by nonradiative processes populate the 3F4 emitting level. 3.2.3 | NIR to visible upconversion (UC) Upconversion (UC) emission from Tm3+ ions in the codoped samples has been observed at room temperature after 975nm excitation. The UC emission spectra show bands at 478, 651, 700, and 800nm corresponding to the 1G4 → 3H6, 1G4 → 3F4, 3F2,3 → 3H6, and 3H4 → 3H6 transitions of Tm3+ ions respectively. Figure 8shows that the increase in the UC intensity in the GCs with respect to those obtained in the glass. However, this increment is not as big as expected. Thus, the small increase in the UC emission intensity in the GCs, if compared with the untreated glass deserves a particular comment. It looks like that after the heat treatment the fraction of RE3+ ions in the amorphous and crystalline phases could change if compare with those of the precursor glass. In particular, the presence of Yb3+ ions FIGURE 4 Transmission spectra for glass and glassceramic (GC) samples codoped with 0.5mol% Tm3+and 1mol% Yb3+ FIGURE 5 Room temperature emission spectra of Tm3+ ions in the single doped glass and GC samples doped with (A) 0.1 and (B) 0.5mol% Tm 3+ VELÁZQUEZ et al.492 2 | LIŠKA ET AL. niobyl(3+) cation (NbO 3+ ), which enters deformed polyhedra (octahedra, tetragonal pyramids) with a coordinationcovalent bond to oxygen. 16,17 2 | METHOD 2 . 1 | Thermodynamic model of Shakhmatkin and Vedishcheva Especially in the field of oxide glasses, the thermodynamic model of Shakhmatkin and Vedishcheva has been successfully applied in previous years. 19 The SVTDM model considers glasses and melts to be an ideal solution formed by salt products of equilibrium chemical reactions of simple starting substances, e.g. oxides, halides, etc. These saltlike products have the identical stoichiometry as the crystalline compounds that exist in the phase diagram of the system under consideration. The model does not use any adjustable parameters. Only the molar Gibbs energies of pure crystalline phases and the analytical composition of the considered system are used as input parameters. The minimization of the system ' s Gibbs energy has to be performed with respect to the molar amount of each system species constrained by the overall system composition to reach the equilibrium system composition. 18,19 The presentday databases of thermodynamic properties comprising the molar Gibbs energies of various species (e.g. the FACT database 20,21 ) enable the routine construction of the Shakhmatkin and Vedishcheva model for many significant multicomponent glass systems. However, in many cases, mainly of nonsilicate multicomponent glass systems, the needed thermodynamic data are not found in any contemporary thermodynamic database. In our previous work, 6,22 we proposed a method for obtaining estimates of missing thermodynamic parameters by reproducing the structural information. 2 . 2 | Decomposition of Raman spectra based on glass composition from TD model The method of numerical analysis of Raman spectra was suggested by Malfait and his collaborators. 2325 The basic assumption of this method is that the Raman spectra are the sum of the partial Raman spectra (generated by the individual structural elements) multiplied by the abundance of these structural elements. The series of Raman spectra obtained for a series of glasses with different compositions span a linear vector space with a dimension given by a number of independent structural elements (i.e. structural elements that independently change their relative abundance) with different partial Raman spectra (PRS). Each measured spectrum is recorded with any scale, i.e. it is known except for the size of a multiplication factor. 2 . 3 | Multivariate curve resolution The multivariate curve resolution (MCR) method 19,2628 decomposes the set of experimental Raman spectra in the spectra of pseudo pure components (called loadings) and relative abundances of these components (called scores). It is essential to emphasize that MCR does not use any data of the system composition. The MCR result must be compared with the result of Malfait ' s spectra decomposition based on the TD model. 3 | EXPERIMENTAL PART Raman spectra were measured at room temperature using a confocal microscope (LabRam HR, Horiba Jobin– Yvon) with backscattering geometry and 532nm excitation line of Nd:YAG laser. For correction of the temperature dependent population of phonon levels, the intensities of Raman spectra were reduced using the Gammon– Shuker relation 29 : where 𝜈𝜈0 and 𝜈𝜈 are the frequency of excitation light and the Raman shift respectively. T is the thermodynamic temperature, k is the Boltzmann constant, and I ( ω ) is the measured Raman intensity. More details can be found in. 16,17 Raman spectra were multiplied by a constant, thus the maximum value of spectral intensity equals to one. 4 | RESULTS AND DISCUSSION On the basis of experimental structural data 16,17 ten system components were considered in the ZnONb 2 O 5 - P 2 O 5 thermodynamic model of Shakhmatkin and Vedishcheva – ZnO (Z), Nb 2 O 5 (Nb) , P 2 O 5 (P), ZnP 4 O 11 (ZP2), Zn(PO 3 ) 2 (ZP), Zn 2 P 2 O 7 (Z2P), Zn 3 (PO 4 ) 2 (Z3P), NbO(PO 3 ) 3 (1/2NbP3), (NbO) 4 (P 2 O 7 ) 3 (Nb2P3), and NbOPO 4 (1/2NbP). The equilibrium molar amounts of SVTDM system components recalculated for xNb 2 O 5 ·50ZnO·(50 - x)P 2 O 5 , ( x=0 , 1 , 3 , 5 , 7 , 10 , 12 ) glasses from the results of structural analysis published in 16,17 are summarized in Table 1 together with the structural groupings represented by the Q n units (i.e. tetrahedron PO 4 with n bridging oxygen atoms) of considered system components. The molar amounts reported in Table 1 correspond to little bit changed glass composition when compared with the prescribed one, i.e. xNb 2 O 5 ·50ZnO·(50 - x)P2O5 , ( x=0 , 1 , 3 , 5 , 7 , 10 , 12 ). This actual glass composition is given in Table 2 together with the glass transition temperature. The molar Gibbs energies of Z, Nb, P, and Z3P were taken from the FACT database. 20,21 The unknown molar Gibbs ( 1 ) I red (𝜈𝜈)=(𝜈𝜈0−𝜈𝜈)−4𝜈𝜈[1−exp. ( −h𝜈𝜈∕kT)]I(𝜈𝜈) 8 | VELÁZQUEZ et al. competing with Tm3+ ions in a finite crystalline volume would reduce the density of the UC centers and, therefore, the UC efficiency. Moreover, increasing the Yb3+ content up to 2mol% does not lead to an increase in the total Tm3+ UC luminescence. Moreover, the overall intensity is lower. This effect could be attributed to energy migration among Yb3+ ions due to the higher concentration, to energy back transfer from Tm3+ to Yb3+ and/or to energy transfer to quencher centers.49 The UC processes involved in the population of the relevant emitting states of Tm3+ ion are depicted in Figure 9. Firstly, the Yb3+ ion at the ground state was excited to the 2F5/2state by absorbing one 975nm photon and then transferred the energy to Tm3+ ions in the ground state following the mechanism 2F5/2 → 2F7/2 (Yb3+):3H6 → 3H5 (Tm3+) (ET1). The 3H5level decays mainly nonradiatively populating the 3F4 (Tm3+) state. Then, a second energy FIGURE 6 Room temperature emission spectra of Tm3+ and Yb3+ ions in the codoped glass and GC samples codoped with 0.5Tm3+- 1Yb3+ obtained under excitation at 791nm FIGURE 7 Room temperature emission spectra of Tm3+ ions in the glass and GC samples codoped with (A) 0.5Tm3+- 1Yb3+ and (B) 0.5Tm3+- 2Yb3+ obtained under excitation at 975nm FIGURE 8 Room temperature UC emission spectra obtained under 975nm excitation for glass (blue line) and GC (red line) samples codoped with 1 (A) and 2 (B) Yb3+ VELÁZQUEZ et al.493 2 | LIŠKA ET AL. niobyl(3+) cation (NbO 3+ ), which enters deformed polyhedra (octahedra, tetragonal pyramids) with a coordinationcovalent bond to oxygen. 16,17 2 | METHOD 2 . 1 | Thermodynamic model of Shakhmatkin and Vedishcheva Especially in the field of oxide glasses, the thermodynamic model of Shakhmatkin and Vedishcheva has been successfully applied in previous years. 19 The SVTDM model considers glasses and melts to be an ideal solution formed by salt products of equilibrium chemical reactions of simple starting substances, e.g. oxides, halides, etc. These saltlike products have the identical stoichiometry as the crystalline compounds that exist in the phase diagram of the system under consideration. The model does not use any adjustable parameters. Only the molar Gibbs energies of pure crystalline phases and the analytical composition of the considered system are used as input parameters. The minimization of the system ' s Gibbs energy has to be performed with respect to the molar amount of each system species constrained by the overall system composition to reach the equilibrium system composition. 18,19 The presentday databases of thermodynamic properties comprising the molar Gibbs energies of various species (e.g. the FACT database 20,21 ) enable the routine construction of the Shakhmatkin and Vedishcheva model for many significant multicomponent glass systems. However, in many cases, mainly of nonsilicate multicomponent glass systems, the needed thermodynamic data are not found in any contemporary thermodynamic database. In our previous work, 6,22 we proposed a method for obtaining estimates of missing thermodynamic parameters by reproducing the structural information. 2 . 2 | Decomposition of Raman spectra based on glass composition from TD model The method of numerical analysis of Raman spectra was suggested by Malfait and his collaborators. 2325 The basic assumption of this method is that the Raman spectra are the sum of the partial Raman spectra (generated by the individual structural elements) multiplied by the abundance of these structural elements. The series of Raman spectra obtained for a series of glasses with different compositions span a linear vector space with a dimension given by a number of independent structural elements (i.e. structural elements that independently change their relative abundance) with different partial Raman spectra (PRS). Each measured spectrum is recorded with any scale, i.e. it is known except for the size of a multiplication factor. 2 . 3 | Multivariate curve resolution The multivariate curve resolution (MCR) method 19,2628 decomposes the set of experimental Raman spectra in the spectra of pseudo pure components (called loadings) and relative abundances of these components (called scores). It is essential to emphasize that MCR does not use any data of the system composition. The MCR result must be compared with the result of Malfait ' s spectra decomposition based on the TD model. 3 | EXPERIMENTAL PART Raman spectra were measured at room temperature using a confocal microscope (LabRam HR, Horiba Jobin– Yvon) with backscattering geometry and 532nm excitation line of Nd:YAG laser. For correction of the temperature dependent population of phonon levels, the intensities of Raman spectra were reduced using the Gammon– Shuker relation 29 : where 𝜈𝜈0 and 𝜈𝜈 are the frequency of excitation light and the Raman shift respectively. T is the thermodynamic temperature, k is the Boltzmann constant, and I ( ω ) is the measured Raman intensity. More details can be found in. 16,17 Raman spectra were multiplied by a constant, thus the maximum value of spectral intensity equals to one. 4 | RESULTS AND DISCUSSION On the basis of experimental structural data 16,17 ten system components were considered in the ZnONb 2 O 5 - P 2 O 5 thermodynamic model of Shakhmatkin and Vedishcheva – ZnO (Z), Nb 2 O 5 (Nb) , P 2 O 5 (P), ZnP 4 O 11 (ZP2), Zn(PO 3 ) 2 (ZP), Zn 2 P 2 O 7 (Z2P), Zn 3 (PO 4 ) 2 (Z3P), NbO(PO 3 ) 3 (1/2NbP3), (NbO) 4 (P 2 O 7 ) 3 (Nb2P3), and NbOPO 4 (1/2NbP). The equilibrium molar amounts of SVTDM system components recalculated for xNb 2 O 5 ·50ZnO·(50 - x)P 2 O 5 , ( x=0 , 1 , 3 , 5 , 7 , 10 , 12 ) glasses from the results of structural analysis published in 16,17 are summarized in Table 1 together with the structural groupings represented by the Q n units (i.e. tetrahedron PO 4 with n bridging oxygen atoms) of considered system components. The molar amounts reported in Table 1 correspond to little bit changed glass composition when compared with the prescribed one, i.e. xNb 2 O 5 ·50ZnO·(50 - x)P2O5 , ( x=0 , 1 , 3 , 5 , 7 , 10 , 12 ). This actual glass composition is given in Table 2 together with the glass transition temperature. The molar Gibbs energies of Z, Nb, P, and Z3P were taken from the FACT database. 20,21 The unknown molar Gibbs ( 1 ) I red (𝜈𝜈)=(𝜈𝜈0−𝜈𝜈)−4𝜈𝜈[1−exp. ( −h𝜈𝜈∕kT)]I(𝜈𝜈) | 9 VELÁZQUEZ et al. transfer process involving Yb3+ and Tm3+ ions can occur through the mechanism 2F5/2 → 2F7/2 (Yb3+):3F4 → 3F2,3 (Tm3+) (ET2). Subsequent multiphonon relaxation from level 3F2,3 populates the 3H4state and a third energy transfer upconversion process, 2F5/2 → 2F7/2 (Yb3+):3H4 → 1G4 (Tm 3+ ) (ET3) populates the 1 G4state. 46 To obtain information about the processes involved in the UC emission after 975nm excitation, the UC emission spectra have been obtained at different pump power densities. The dependence of the UC emission on the pump power density is related to the number of photons (n) involved in the process and gives information about the UC mechanisms according to the relation Figure 10shows the logarithmic plot of the UC blue and NIR (3H4→3H6) emission intensities of the codoped GC sample with 0.5TmF3−1YbF3 (in mol%) as a function of the pump power laser density. As can be seen for the blue emission, the slope is 2.08 which indicates that two photons are required to excite the electrons to the 1G4state. The slope in the case of the 3H4 → 3H6 transition is 1.58. However, it is clear from the energy level diagram of Tm3+ ions (Figure 9) that two photons are needed to populate this level and three photons are needed to populate the 1G4 level. UC to the 3H4 level occurs via two successive energy transfers from Yb3+ ions in the 2F5/2 excited state. An excited Yb3+ ion in the 2F5/2state transfers its energy nonresonantly to a Tm3+ ion in the ground state 3H6, exciting it to the 3H5state from where the 3F4 level is populated (ET1 in Figure 9). A second Yb3+ ion transfers its energy to the Tm3+ ion promoting it from the 3F4 to the 3F2,3 excited state (ET2 in Figure 9) and then decays nonradiatively to the3H4level. A third energy transfer step from an excited Yb3+ ion can excite the Tm3+ ions to the 1G4level (ET3 in Figure 9). The observed slopes in Figure 10 indicate the presence of a saturation effect which reduces the experimental power dependence. The obtained power dependences represent the lower limit of the number of photons involved in the UC mechanisms.50,51 This behavior, previously observed in other systems, has been attributed to the competition between the decay rate of the intermediate states and the UC rates. When UC dominates over linear decay for the depletion of the intermediate excited states, the slope of the luminescence from the upper state n is almost linear.52 No UC emissions were observed in the single doped samples excited at 791 and 975 nm. However, UC emission of Tm3+ ions is also observed in the codoped glass and glassceramics samples with 1 and 2mol% Yb3+ under 791nm excitation (see Figure 11). This is probably due to the reduction of the interionic distances between the Tm3+ ions in the NCs favored by the presence of Yb3+ ions, increasing the probability of the Tm3+- Tm3+ energy transfer. The UC mechanism of Tm 3+ - Yb 3+ codoped samples is as follows. After (2) Iem ∝ ( P pump)n. FIGURE 9 Energy level diagram of Yb3+ and Tm3+ ions together with the different energy transfer processes (ET) responsible for the upconverted emissions after excitation at 975nm (Yb3+) FIGURE 10 Logarithmic plot of the pump power dependence of the blue (1G4 → 3H6) and NIR (3H4 → 3H6) integrated emission intensities for the GC sample codoped with 0.5Tm3+- 1Yb3+ Slope: 2.08 Slope: 1.58 100. 01 ,000.0 PowerDensity (W/cm 2 ) 0 0 1 10 Integrated Intensity(arb. units)