Synthesis, structural characterization and ionic conductivity of mixed alkali titanium phosphate glasses

Glasses with formula Na3-xLixCaTi(PO4)3 [10(3-x) mol. % Na2O 10x mol. % Li2O 20 mol. % CaO 20 mol. % TiO2 30 mol. % P2O5] (0 ≤ x ≤ 3) were prepared by standard melt-quenching technique, and their structural and physical properties were characterized by thermal analysis, density measurements, Raman, and impedance spectroscopy. When Na is gradually replaced by Li, molar volume, glass transition temperature (Tg) and ionic conductivity values decrease, pass through a minimum around the composition x = 1.5, then increase, while density values increase, pass through a maximum, then decrease. The non-linear variation of these physical properties is a result of the classical mixed alkali effect. Powder X-ray diffraction shows that crystallization of the glasses leads to the formation of a Nasicon phase for the compositions x = 0 and x = 0.5, and to a mixture of phases for the other compositions. Raman spectroscopy study shows that the glass structure contains P2O7 and PO4 groups, and short -Ti-O-Ti-O-Tichains, formed by TiO6 octahedra linked to each other through corners. These chains are linked by phosphate tetrahedra to form -O-Ti-O-P-Olinkages.


Introduction
Metal phosphates in both crystalline and glassy forms are extensively studied for their numerous properties and applications in various fields such as energy, health, environment and catalysis [1][2][3][4] . Among these phosphates, those belonging to the Nasicon family (Nasicon: sodium super ionic conductors) have attracted much attention due to their diverse properties [5][6][7][8] . These compounds, developed first for their remarkable ionic conductivity 9,10 , exhibit other characteristic properties, such as for example, low thermal expansion 11,12 . Also, the wide variety of possible formulations allowed the development of new materials with more specific properties such as magnetism 13 , luminescence 14,15 or energy storage 16 . Some compounds of the Nasicon family can also be prepared in the glassy form; they are called Nasiglasses (Na super ionic glass) [17][18][19][20][21] . We previously reported the structural and electrical properties of Na3CaTi(PO4)3 which exists in both crystalline and glassy forms 20,22 . The structure of the crystalline compound, which belongs to the Nasicon family, consists of a three-dimensional framework built of PO4 tetrahedra sharing corners with AO6 octahedra (A= Ca, Ti); Na + ions occupy the interstitial sites 22 . The structure of the glassy form contains P2O7 4and PO4 3groups, and short -Ti-O-Ti-O-Ti-chains, formed by TiO6 octahedra linked by corners 20 . Substitution of sodium for lithium in Na3CaTi(PO4)3 led us obtaining new mixed alkali glasses: Na3-xLixCaTi(PO4)3 (0 ≤ x ≤ 3) 23 . The present work reports their synthesis, structural characterization and electrical properties.

Sample preparation
Na3-xLixCaTi(PO4)3 [10(3-x) mol. % Na2O -10x mol. % Li2O -20 mol. % CaO -20 mol. % TiO2 -30 mol. % P2O5] (0 ≤ x ≤ 3) glasses were prepared from Na2CO3, CaCO3, and Li2CO3 carbonates, TiO2 oxide, and (NH4)H2PO4 ammonium dihydrogen phosphate as starting materials (with purity not less than 99.8 %), by the melt quenching method, according to the following reaction scheme: (3-x) Na2CO3 + x Li2CO3 + 2 CaCO3 + 2 TiO2 + 6 (NH4)H2PO4 2 Na3-xLixCaTi(PO4)3 + 6 NH3 + 5 CO2 + 9 H2O The reagent amounts required to obtain 10 g of glass were mixed in a mortar and transferred to an alumina boat. The powder mixture was then heated, with a rate of 10 °C/min, respectively to 200 °C, 400 °C and 600 °C for 6 hours at each step. After grinding, the powder was transferred to a Pt crucible and heated (10 °C/min) to 800 °C (2 h), 950 °C (2 h) and finally to 1050 °C (20 min). The molten glass obtained was quenched in the air in a graphite mold to create bulk samples for conductivity measurements. Under these conditions, all glasses are colorless and homogeneous, without bubbles. Bulk samples were next annealed for 4 h at a temperature 20 °C below the glass transition temperature (Tg), followed by slow cooling (1 °C/min) to room temperature, in order to remove residual stress. Bulk samples were cut as round disks of 10 mm and 20 mm in diameter and 2 mm thick.

Chemical analysis
Lithium and sodium concentrations for the vitreous composition x = 0.5 were determined by flame atomic emission spectrometry (FAES) using a lab-built instrument. A 5.3 cm slot burner head, a 25 mm diameter, 75 mm focal length fused silica lens, and a handheld Ocean Optics USB 4000 spectrometer (Ocean Optics, Dunedin, FL, USA) were used in the instrumental setup. The flame composition was controlled by different regulators set at 12.5 L/min air and 1.0 L/min acetylene. The integration time was 50 ms and standard reference solutions used in this measurement were: 20 ppb, 50 ppb, 100 ppb, 200 ppb Li and Na standards concentrations. 100 mg of the corresponding glass (with composition x = 0.5) was diluted in 10 ml of concentrated HNO3 to make a 100 ml solution. The weight percentages obtained for Na (12.13 %) and Li (0.803 %) are close to the theoretical values (Na: 13.25 % ; Li: 0.80 %).

Density measurements
Density measurements were performed on blocks of glasses, using Archimedes' method, on an analytical balance (± 0.0001 g) with an attached density kit; diethyl phthalate was used as the immersion liquid. The density (ρ) was obtained from the following equation: ρ = [ma/(ma-ml)] ρl, where ma is the mass of the sample in air, ml is the mass of the sample fully immersed in the liquid, and ρl is the density, at room temperature, of the liquid used (diethyl phthalate). The error in the density measurements is within ± 0.03 g/cm 3 . The molar volume (Vm = M/ρ) of the glasses was calculated from the measured density (ρ) and molecular weight (M). M was calculated using the following formula:

Powder X-ray diffraction
Powder X-ray diffraction (PXRD) data were collected at room temperature with a Bruker D2 Phaser XRD system, with CuKα radiation and a Lynxeye 1D Detector, in the 2θ range of 10-60 ° at an increment of 0.01 °.

Thermal analysis
Differential scanning calorimetry (DSC, SDT Q600) and differential thermal analysis (DTA, DTG-60H) were used to determine the glass transition, crystallization and melting temperatures (Tg, Tc, Tm) of the glasses. Glass samples (~20 to 40 mg) were transferred to platinum crucibles, and the DTA curves were recorded in an air atmosphere, with a heating rate of 10 °C/min and accuracy of ± 5 °C. The DSC curves were recorded in a flow of argon with a rate of 100 ml/min and a temperature ramp of 5 °C/min.

Raman spectroscopy
Raman spectra were recorded on a confocal micro-Raman Labram (Horiba/Jobin-Yvon) spectrophotometer with a backscattering mode at room temperature in the range 150-1800 cm -1 . The excitation source was a 532 nm continuous laser. A holographic Notch filter was used to reject the Rayleigh diffusion. The backscattered light was collected through a 100x objective and selectively transmitted toward a cooled CCD detector.

Ionic conductivity
Bulk samples (20 mm and 10 mm in diameter and 2 mm thick) were polished using sand-paper and polishing cloths to optical transparency. The samples were sputtered with ~50 nm thick ion-blocking gold electrodes on both sides of the sample. Silver wires were then attached to the gold electrodes with silver conductive ink. Ionic conductivity values were determined with electrochemical impedance spectroscopy measurements using a Gamry Instruments Series G750 potentiostat with four probes in the potentiostatic mode. The applied voltage and ac perturbations were 0 V and 1 V, respectively, over a frequency range of 10 Hz to 300 KHz and a temperature range of 300 K to 673 K.

Results and Discussion
The vitreous state of the samples was inspected by PXRD and DSC/DTA. The absence of any sharp peak in the diffraction pattern (   Table 1. The density values vary in the range between 2.32 and 2.85 g/cm 3 . The substitution of sodium for lithium-ion is accompanied with a non-linearity in density variation. The density passes through a minimum for the composition x = 1.5 as well as the molar volume passes through a maximum (Fig. 2). This non-linear behavior has already been observed in other glasses 24-27 .  Figure 3 shows DSC/DTA curves of Na3-xLixCaTi(PO4)3 (0 ≤ x ≤ 3) glasses. All glasses exhibit an endothermic change due to glass transition (Tg), followed by an exothermic change due to crystallization (Tc), then an endothermic change due to the melting (Tm) of the crystalline phases. Table 1 shows the values of the characteristic temperatures (Tg, Tc, Tm). Tg was determined as the inflexion point of the first endothermic peak, Tc as the maximum of the exothermic peak, and Tm as the minimum of the second endothermic peak. Some glass compositions show more than one value of Tc and Tm because their crystallization leads to a mixture of phases, and each phase has its crystallization and melting temperatures. We gave in Table 1 only the values of the first peak. The drift observed, between Tg and Tc, in DTA thermograms of some glass compositions, is probably due to the relaxation phenomenon of glasses 28 . Fig. 4 shows the variation of Tg versus lithium content x (0 ≤ x ≤ 3). As it can be seen in this Figure

Ionic conductivity
Ionic conductivity measurements were completed for all Na3-xLixCaTi(PO4)3 (0 ≤ x ≤ 3) glass compositions, however, for x = 2.5 and x = 3, the resistance at room temperature was so high that it was impossible to measure it. The Nyquist plots, at different temperatures, are shown in Fig. 7, for two glass compositions (x = 0; 2). They exhibit complete semicircles at a higher frequency, a clear local minimum, and then the impedance increases at low frequency. At 300°C it is clear that the low frequency impedance continuously increases, indicating the samples are pure ion conductors. Similar impedance behavior was observed for x = 0.5; 1 and 1.5. The local minimum of the semicircle corresponds to the bulk resistance (R) and shifts to lower Z' values with increasing temperature, indicating that the dc conductivity is a thermally activated process. Ionic conductivity was calculated using the formula σ = e/(R.S), where e and S are respectively the thickness and the surface area of the glass disc, and R its resistance. Ionic conductivity variation follows the Arrhenius law: σ = (A/T) exp(-Ea/kT) (σ: conductivity, A: pre-exponential factor, k: Boltzmann constant, T: temperature, Ea: activation energy). Values of the ionic conductivity at 300 °C are given in Table 1. The log(σT) vs 1/T plots are displayed in Fig. 8.
Ionic conductivity variation of these glasses, versus alkali ion content, shows a non-linear evolution (Fig. 9), as observed above for density, molar volume and glass transition temperature. The same behavior of conductivity was observed in other vitreous systems and was attributed to the mixed alkali ions effect (MAE) 26,27,[37][38][39] . The effect of Li + ions in Na3-xLixCaTi(PO4)3 (0 ≤ x ≤ 3) glasses can be observed as a decrease in the conductivity giving rise to a minimum at the composition x = 1. For low lithium content (Li2O/Na2O ≤ 1; x ≤ 1.5), the conductivity is governed by Na + ions, but their mobility is hampered by Li + ions. In contrary, for low sodium content, the conductivity is governed by Li + ions, but their mobility is hampered by Na + ions. This means that the two types of alkali ions (Na + and Li + ) have distinctly different conduction pathways, and implies that Li + ions tend to block the pathways for the Na + ions and vice versa. We suppose that this is the main reason for the MAE. Our hypothesis was based on the dynamic structure model (DSM) 40,41 , which gives a theoretical explanation of this effect. By definition, the MAE occurs when two types of alkali ions are introduced into a glassy network. The progressive substitution of an alkali ion for another, while total alkali content in the glass being constant, is often accompanied by non-linear variations in physical properties related to the alkali ion movement and structure properties 42,43 . The DSM suggests that the two alkali ions can create their proper distinct local environments in the glass. Consequently, two preferred diffusion paths for each type of ion are formed which are dependent on the ionic composition and provide a possible explanation for the MAE.

Conclusion
Structural and physical properties of Na3-xLixCaTi(PO4)3 (0 ≤ x ≤ 3) glasses have been studied. When Na + is gradually replaced by Li + , density, molar volume, glass transition temperature and ionic conductivity vary in a non-linear manner, as a result of the mixed alkali ions effect. Crystallization of the glasses leads to the formation of a Nasicon phase for the compositions x = 0 and x = 0.5, and to a mixture of phases for the other compositions. Raman spectroscopy study shows that the glasses contain P2O7, PO4 groups and short -Ti-O-Ti-O-Ti-chains, probably linked by distorted PO4 tetrahedra. One can conclude that the glass structure is based on -Ti-O-Pand -Ti-O-Ti-O-linkages.