Vibrational spectroscopic and bond valence study of structure and bonding in Al₂O₃-containing AgI–AgPO₃ glasses

Abstract

We present a detailed investigation of the effects of synthesis conditions on glasses xAgI–(1 − x)AgPO₃ with 0 ≤ x ≤ 0.4. Raman and infrared spectroscopy of glasses synthesized in platinum crucibles showed that their metaphosphate structure remains largely unaffected by increasing the amount of AgI. This result contradicts recent findings on similar glasses synthesized by melting in alumina crucibles, which were found to exhibit a strongly changing phosphate structure with addition of AgI. The glass transition temperature and ionic conduction properties of the latter glasses also show strong deviation from the analogous properties of glasses melted in platinum crucibles, as in this work. To reveal the origin of such effects, glasses with the nominal composition 0.4AgI–0.6AgPO₃ were synthesized in alumina crucibles and their structures were compared with glasses prepared in platinum crucibles with composition yAl₂O₃–(1 − y)[0.4AgI–0.6AgPO₃] and y = 0.0, 0.04 and 0.07. Vibrational spectroscopy showed that melting in alumina crucibles, especially in the presence of P₂O₅, leads to doping the AgI–AgPO₃ glass with Al₂O₃ leached from the crucible. The bond valence approach was employed to assist assigning the new Raman bands that develop when Al₂O₃ modifies the metaphosphate structure, with an exponential relationship describing the dependence of the P–O stretching frequency on bond strength. Combination of this relationship with bond valence principles allowed us to associate the formation of aluminum-triphosphate species Al_2/3Ag₃P₃O₁₀ with the Raman trends observed for AgI–AgPO₃ glasses doped with Al₂O₃ either intentionally or unintentionally, i.e. when melted in alumina crucibles for long times.

---

1. Introduction

Ion conducting glasses attract much attention for fundamental studies of their ion transport mechanism and for technological applications in solid state electrochemical devices.^1–25 Some of the best glassy electrolytes, known as superionic conductive glasses, result from doping silver metaphosphate glass (AgPO₃) with silver halide salts; this process leads to room temperature ionic conductivity as high as 10⁻² S cm⁻¹.^26–32 This conductivity value is about five orders of magnitude higher than the conductivity of AgPO₃ base glass and analogous to that of molten salts, e.g. KNO₃.³ Over the years, research efforts have been focused on understanding the origins of the remarkably enhanced ionic conductivity in glasses doped with metal halide salts, in order to control the short- and/or medium-range order structures that favor ionic conduction in doped glasses.

Early studies in the field suggested that the doping salt forms its own conduction pathways^1,3,6 or microdomains/clusters^26,33,34 which percolate through the base oxide glass and facilitate ion transport. Later investigations showed that the incorporated doping salts expand the glass matrix, and this creates additional free volume^30,35 or pathway volume³⁶ available for ion transport and, thus, leads to higher ionic conductivity as a result of the increased charge carrier mobility in the expanded glass matrix. In a different approach the emphasis was placed on the local coordination of Ag ions in AgI-based glasses, where the evolution of mixed I/O environment around Ag ions and the establishment of O–Ag–I conduction sites were suggested to be the main factors controlling the mobility of the Ag ion carriers.³⁷ On the other hand, a correlation between ionic conductivity and the Ag–I distance was demonstrated in families of AgI-doped glasses, and glasses having longer Ag–I distance were found to exhibit lower activation energy and higher ionic conductivity.³⁸ In a unified treatment for ion transport in glassy and polymer electrolytes, it was proposed that very similar forces are involved in both ion migration and structural relaxations processes in such highly ion conducting materials, e.g. AgI-doped glasses.¹⁸ A recent analysis of conductivity data for glasses AgI–AgPO₃ showed that the mobility of Ag ions does not vary significantly with AgI content; it is rather the increase in the effective Ag ion carrier concentration that causes the enhancement in ionic conductivity.³⁹

Despite their differences, all models cited above highlight the crucial role of AgI in enhancing ionic conductivity. In addition to this role of AgI, other studies emphasized the importance of the synthesis method on structure and properties of glasses xAgI–(1 − x)AgPO₃. Thus, it was found^40,41 that when the AgPO₃ glass is prepared under dry ambient synthesis environments it exhibits enhanced direct current ionic conductivity (σ_dc) by almost two orders of magnitude and remarkably elevated glass transition temperature (T_g) by as much as 95 °C, relative to AgPO₃ glasses developed in common laboratory ambient environments.^26–32 This impressive increase in σ_dc and T_g was attributed to the absence of water traces when the handling of starting materials is performed under dry ambient conditions.^40,41 Similarly prepared glasses xAgI–(1 − x)AgPO₃ were found to display three distinct regimes of conductivity defined by thresholds at x_c1 = 0.09 and x_c2 = 0.38, which correlate well with the three elastic phases determined from calorimetric measurements.^41–43 Thus, it was suggested that the reversibility window 0.09 < x < 0.38 is suppressed or completely diminished in the presence of water traces because of the depolymerization of phosphate chains and the reduction in network connectivity. The Raman spectra of dry AgI–AgPO₃ glasses showed a strong dependence on the AgI content; this was interpreted in terms of changes in glass structure from chain-like to ring-like upon increasing AgI content^41,43 and found to contradict earlier Raman results for glasses synthesized under common laboratory ambient conditions which showed little or no change in the phosphate structure as AgI is added to AgPO₃.^31,44,45 It is noted that AgI-induced changes in the topology of the phosphate network, from long chains to large and small rings,^41,43 were considered to be at the origin of the excess constraints in AgI–AgPO₃ glasses.⁴⁶

The fundamental question whether the method of preparation affects structure and properties of AgI-containing phosphate glasses continues to attract interest, in view also of new applications of such materials in the field of photonics.⁴⁷ In this context, we note a recent investigation focusing on glass preparation by high- and low-temperature routes, i.e. melt-quenching versus mechano-synthesis by ball-milling.⁴⁸ This study revealed differences in thermal properties (e.g., T_g) and Raman spectra of AgI–AgPO₃ glasses prepared by the two different synthesis methods, and attributed them to different energy transfer mechanisms involved in the two synthesis techniques.

In a previous brief report,⁴⁹ we studied the effect of synthesis methods on structure and properties of the AgPO₃ base glass by varying the atmosphere of handling the starting materials, i.e. dry versus normal laboratory ambient conditions, starting materials, melting time and temperature, and type of crucible used for melting (i.e., Pt versus alumina). A key finding was that melting in alumina crucibles, which is a practice used widely for the synthesis of superionic glasses, gives AgPO₃ glasses being unintentionally contaminated with up to ca. 3 mol% Al₂O₃. This Al₂O₃-doping was found to cause enhanced phosphate network connectivity and drastically increased σ_dc and T_g relative to Al₂O₃-free AgPO₃ glasses prepared in Pt crucibles. In view of these findings for AgPO₃, we have studied and report here on the structure and properties of glasses xAgI–(1 − x)AgPO₃ with 0 ≤ x ≤ 0.4. The synthesized glasses were investigated by combining Raman and infrared spectroscopic techniques as they provide complementary structural information, while glass transition temperature and ionic conductivity were also measured as a function of AgI content and compared with literature data. To investigate the role of crucible on the AgI–AgPO₃ system, we studied glasses 0.4AgI–0.6AgPO₃ melted in alumina and Pt crucible, and compared their vibrational spectra with those recorded for glasses yAl₂O₃–(1 − y)[0.4AgI–0.6AgPO₃] prepared in Pt crucibles, with y = 0.0, 0.04 and 0.07. A key finding was that synthesis in alumina crucibles leads to unintentional doping of the 0.4AgI–0.6AgPO₃ glass with aluminum oxide, a process causing the increase of the O/P ratio and, consequently, the progressive change of the phosphate structure as its modification exceeds the metaphosphate composition (O/P = 3.0). Application of the bond valence approach allowed to propose the formation of Al-triphosphate, Al_2/3Ag₃P₃O₁₀, as a main structure formed when the metaphosphate glass is doped with Al₂O₃ either intentionally or unintentionally when using alumina crucibles and long melting times. These results are discussed in relation to previous studies on AgI–AgPO₃ glasses, and contribute to answering open questions in the field of ion conducting glasses.

2. Experimental

2.1 Glass synthesis

As noted in the Introduction, our earlier study on the AgPO₃ base glass showed that melting in alumina crucibles leads to contamination with Al₂O₃.⁴⁹ To avoid this problem, Pt crucibles were employed in this work to synthesize glasses in the series xAgI–(1 − x)AgPO₃, 0 ≤ x ≤ 0.4, using as starting materials dry NH₄H₂PO₄ (99.995%), AgNO₃ (99.995%) and AgI (99.9%). As for the case of AgPO₃ glass, we performed all weighting and mixing manipulations of the starting materials in a glove bag purged with dry N₂ gas, i.e. under dry ambient conditions (ca. 3–5% relative humidity). Each batch was then transferred quickly to an electrical furnace held at 150 °C and slowly calcinated up to 300 °C to remove the volatile products which result from the decomposition of the reactants. The temperature of the furnace was then increased to 700–900 °C depending on composition; the melts were repeatedly stirred and kept at the melting temperature for two hours. For convenience, the synthesis parameters of the studied glasses are summarized in Table 1 with their nominal compositions. Glasses in this series were obtained in the form of ca. 0.5–1.0 mm-thick discs with ca. 10–15 mm of diameter by quenching the melts in stainless steel pre-forms. This procedure gives smooth glass surfaces and suitable specimen sizes for infrared reflectance and electrical conductivity measurements when the AgI content is up to about x = 0.4. All glass specimens were stored under vacuum, with no additional heat or surface treatment to avoid exposing the samples to lab ambient for long periods of time.

Table 1 Synthesis parameters for glasses xAgI–(1 − x)AgPO₃ and values determined in this work for glass transition temperature, T_g, dc ionic conductivity at 293 K, σ_dc, activation energy for ionic conductivity, E_a, infrared absorption at 2815 cm⁻¹, α_max, and water content, C_H2O. Density data, ρ, are from ref. 39 and are used to estimate the water content in glass in wt%, W_H2O. The starting materials for glass synthesis glasses were NH₄H₂PO₄, AgNO₃ and AgI, which were handled under dry ambient conditions

Glass	x	Crucible	Synthesis conditions	Tg (°C) ± 1	log σdc (S cm^−1) ± 0.03	Ea (eV) ± 0.01	αmax (cm^−1) ± 0.1	CH2O (mol L^−1) ± 0.01	ρ (g cm^−3)	WH2O (wt%) ± 0.005
A1	0	Pt	700 °C, 2 h	193	−6.23	0.53	31.0	0.09	4.35	0.037
A2	0.1	Pt	800 °C, 2 h	183	−5.50	0.49	19.2	0.06	4.56	0.022
A3	0.2	Pt	900 °C, 2 h	171	−4.77	0.42	17.7	0.05	4.77	0.019
A4	0.3	Pt	900 °C, 2 h	141	−4.11	0.39	13.7	0.04	4.98	0.014
A5	0.4	Pt	900 °C, 2 h	120	−3.45	0.36	12.3	0.04	5.20	0.012

---

A second series of glasses was prepared with composition yAl₂O₃–(1 − y)[0.4AgI–0.6AgPO₃] as summarized in Table 2; the first glass (B1) being identical to glass A5 of Table 1. Two additional glasses with starting composition also 0.4AgI–0.6AgPO₃ were prepared by quenching melts from alumina crucibles using appropriate batches of NH₄H₂PO₄, AgNO₃ and AgI (glass B2) and P₂O₅, AgNO₃ and AgI (glass B3). For glasses B4 (y = 0.04) and B5 (y = 0.07) we used Pt crucibles, the same starting materials NH₄H₂PO₄, AgNO₃, AgI and Al₂O₃, as well as the same temperature and melting time (Table 2). Glasses in this series were prepared also under dry ambient conditions for the weighting and mixing manipulations of the starting materials.

Table 2 Synthesis parameters for glasses yAl₂O₃–(1 − y)[0.4AgI–0.6AgPO₃]. Starting materials were handled under dry ambient conditions

Glass	y	Crucible	Starting materials	Synthesis conditions
B1	0	Pt	NH4H2PO4, AgNO3, AgI	900 °C, 2 h
B2	0	Alumina	NH4H2PO4, AgNO3, AgI	900 °C, 1.5 h
B3	0	Alumina	P2O5, AgNO3, AgI	700 °C, 0.5 h
B4	0.04	Pt	NH4H2PO4, AgNO3, AgI, Al2O3	1100 °C, 0.5 h
B5	0.07	Pt	NH4H2PO4, AgNO3, AgI, Al2O3	1100 °C, 0.5 h

---

To assist Raman and infrared assignments, we prepared silver-pyrophosphate (Ag₄P₂O₇) by melting appropriate amounts of dried NH₄H₂PO₄ and AgNO₃ in Pt crucible. It was found that melt quenching results in polycrystalline Ag₄P₂O₇, while glass formation using the same quenching technique required addition of at least 20 mol% AgI (i.e., glass 0.2AgI–0.8Ag₄P₂O₇). Commercially available silver-orthophosphate (Ag₃PO₄) in polycrystalline form was considered also for Raman and infrared measurements.

2.2 Spectroscopic techniques

Raman spectra were measured in vacuum at the backscattering geometry (Renishaw inVia Raman Microscope), using the 514.5 nm line of an Ar ion laser for excitation. All spectra were recorded at room temperature with 2 cm⁻¹ resolution. Examination of glass specimens under the Raman microscope at the end measurements revealed no signs of laser-induced modifications at the employed irradiation conditions, i.e. 0.1 mW μm⁻² at the sample. The Raman spectra reported here are in the form of temperature-reduced intensity, I_red, calculated from the expression I_red = I(ν)/[n(ν) + 1] where I(ν) is the measured Raman intensity and n(ν) is the Bose–Einstein factor, n(ν) = 1/[exp(hcν/k_BT) − 1], where ν is the Raman shift in cm⁻¹, c is the speed of light, T is the temperature in K, and k_B and h are the Boltzmann and Planck constants, respectively.

Infrared (IR) spectra were measured on a vacuum Fourier transform spectrometer (Bruker, Vertex 80v), in quasi-specular reflectance and in transmittance modes. For each spectrum, 200 scans were collected at room temperature with 2 cm⁻¹ resolution and averaged for evaluation. Reflectance spectra were measured separately in the far- and mid-IR range and then merged to form a continuous spectrum in the range 30–7000 cm⁻¹. Analysis of such reflectance spectra by Kramers–Kronig transformation yielded the absorption coefficient spectra, α(ν), from the expression α(ν) = 4πνk(ν) where ν is the infrared frequency in cm⁻¹ and k(ν) is the imaginary part of the complex refractive index.¹⁴

2.3 Electrical and thermal properties

Impedance spectroscopy was employed to determine the ionic conductivity of glasses xAgI–(1 − x)AgPO₃. Circular gold electrodes were evaporated on the parallel surfaces of each disc-shaped glass specimen. The sample was then placed in a sample holder and isothermal conductivity curves were obtained for temperatures 20–160 °C in the frequency range 10 MHz to 5 Hz using an HP4192A impedance analyzer. Sufficient time was allowed between measurements to ensure thermal stability and constant readings. The direct current ionic conductivity, σ_dc, was deduced at each temperature from the low-frequency plateau of the corresponding curve, and the activation energy for ionic conduction, E_a, was obtained from the Arrhenius dependence of ionic conductivity vs. inverse temperature using the expression E_a = −k_B[dln σ_dc/d(1/T)], where k_B is the Boltzmann constant and T is the temperature in K.

The glass transition temperature, T_g, was determined by Differential Scanning Calorimetry (DSC, TA Q200 instrument) from the first heating circle at a scan rate of 10 °C min⁻¹. Glasses for DSC measurements were sealed in Al pans and measured under dry N₂. At least two different samples were run for each glass composition. T_g was determined from the inflection point of the DSC curve and was found reproducible within ±1 °C.

3. Results and discussion

3.1 Glasses xAgI–(1 − x)AgPO₃

All glasses in the series AgI–AgPO₃ were synthesized under identical conditions, apart from some variations in melting temperature (Table 1). Therefore, changes in glass structure and properties should be associated with the amount of AgI added to the AgPO₃ base glass.

3.1.1 On the structure of glasses xAgI–(1 − x)AgPO₃. Reduced Raman scattering and IR absorption coefficient spectra shown in Fig. 1a and b, respectively, manifest clearly their complementary nature due to differences in selection rules. The spectra are presented normalized on the strongest band to facilitate comparison; Raman spectra are normalized on the 1140 cm⁻¹ band and IR spectra on the band at 1235 cm⁻¹.

Fig. 1 (a) Temperature reduced Raman, and (b) infrared absorption coefficient spectra of glasses xAgI–(1 − x)AgPO₃. For comparison, Raman spectra were normalized on the strongest band at ca. 1140 cm⁻¹ and infrared spectra on the 1235 cm⁻¹ band.

As discussed earlier for glass AgPO₃,⁴⁹ the picture emerging from the consideration of its Raman and infrared spectra is that based mainly on a polymer-like metaphosphate chain structure which is formed by linked phosphate tetrahedra having two bridging (Ø) and two terminal (O) oxygen atoms, (PØ₂O₂)⁻. This structure is consistent with earlier studies of AgPO₃ glass and is widely accepted in the literature.^{31,40,44,45,50–60} Using the phosphate Qⁱ terminology where i is the number of bridging oxygen atoms per phosphate tetrahedron, the (PØ₂O₂)⁻ tetrahedron is termed Q² and has the PO₃⁻ stoichiometry. The presence of Q² units explains key features of the Raman and infrared spectra in Fig. 1a and b for AgPO₃ glass, x = 0. The strongest Raman band at 1144 cm⁻¹ originates from the symmetric stretching of the terminal PO₂⁻ groups in Q² units, ν_s(PO₂⁻), while the band at 676 cm⁻¹ is due to the symmetric stretching of P–O–P bridges within the phosphate backbone, ν_s(P–O–P). Likewise, the strongest infrared band at 1235 cm⁻¹ is due to the asymmetric stretching of PO₂⁻, ν_as(PO₂⁻), while the corresponding mode of P–O–P in chains of Q², ν_as(P–O–P), gives the band at 895 cm⁻¹. Besides these bands, the infrared spectrum shows at high-frequencies additional features at ca. 990 and 1058 cm⁻¹. Based on previous works,^51,59 the 990 cm⁻¹ band is attributed to ν_as(P–O–P) for Q² units in metaphosphate rings and the one at 1058 cm⁻¹ to the asymmetric stretching of PO₃²⁻ groups, ν_as(PO₃²⁻), which terminate the metaphosphate chains. For convenience, the assignments of infrared and Raman bands are summarized in Table 3.

Table 3 Frequencies and assignments of Raman and infrared bands measured in this work for metaphosphate glasses xAgI–(1 − x)AgPO₃, pyrophosphate glass 0.2AgI–0.8Ag₄P₂O₇, and orthophosphate crystal Ag₃PO₄

Raman (cm^−1)	Infrared (cm^−1)	Assignment	References
a Band intensity: vs = very strong, s = strong, m = medium, w = weak, vw = very weak, sh = shoulder.b Vibration mode: ν = stretching, δ = bending, νs, ν1 = symmetric stretching, νas, ν3 = asymmetric stretching, ν2 = symmetric bending, ν4 = asymmetric bending.c Q^i: a phosphate tetrahedron with i bridging and 4-i terminal oxygen atoms.
Metaphosphate glasses xAgI–(1 − x)AgPO3
1220–1225 (vw)^a	1235 (vs)	νas(PO2^−)^b, Q^2^c	21, 31, 43 and 49–60
1144–1138 (vs)	1140 (sh, vw)	νs(PO2^−), Q^2	21, 43–45 and 49–60
	1090 (sh, w)	νas(PO3^2−), Q^1	21, 59 and 60
	1058 (s)	νas(PO3^2−), end chains	21, 31, 43 and 49–59
1000 (vw)	—	νs(PO3^2−), Q^1	21, 53, 56 and 57
	990 (s)	νas(P–O–P), Q^2 in rings	51, 59 and 60
	895 (s)	νas(P–O–P), Q^2 in chains	21, 31, 43 and 49–60
680 (s), 710, 778 (sh)	710 (m), 778 (m)	νs(P–O–P), Q^2	21, 43 and 49–60
—	475 (s), 518 (s), 585 (m)	δ(PO2^−) + δ(PO3^2−)	21 and 61
—	128–122	ν(Ag–O), ν(Ag–O/I)	14, 21, 51 and 55
120 (m-s)	—	ν(Ag–I)	44 and 45
45 (m-s)	—	Boson mode	44 and 45
Pyrophosphate glass 0.2AgI–0.8Ag4P2O7
1075 (m), 1130 (sh)	1085 (vs)	νas(PO3^2−), Q^1	21, 59, 61 and 66
1002 (vs)	1000 (vw)	νs(PO3^2−), Q^1	21, 55, 57–59 and 61
912 (w)	—	ν1(PO4^3−), Q^0	21, 61 and 67
—	903 (s)	νas(P–O–P), Q^1	21, 57–59, 61 and 66
710 (m)	705 (m)	νs(P–O–P), Q^1	21, 55, 57–59 and 61
	527 (s), 555 (s), 590 (vw)	δas(PO3^2−), Q^1 species	61 and 66
345 (w)	415 (vw), 475 (w), 510 (s)	δs(PO3^2−), Q^1 species	61 and 66
	207 (s)	δ(P–O–P), Q^1 species	61
	140 (s)	ν(Ag–O)	14, 51 and 55
48 (m)	—	Boson mode	44 and 45
Orthophosphate crystal Ag3PO4
955 (vw), 1015 (vw)	960 (vs)	ν3(PO4^3−), Q^0	21, 59, 61 and 67
910 (vs)	—	ν1(PO4^3−), Q^0	21, 57, 58, 61 and 67
550 (vw)	552 (s)	ν4(PO4^3−), Q^0	21, 59, 61 and 67
405 (w)	—	ν2(PO4^3−), Q^0	21, 61 and 67
	198 (m)	ν(Ag–O)	14, 51 and 55

---

Addition of AgI to AgPO₃ glass affects clearly the low-frequency Raman spectra with bands developing progressively at ca. 45 and 120 cm⁻¹ (Fig. 1a). Similar features were observed in earlier studies and were associated with the boson band and Ag–I stretching, respectively.^44,45 However, the higher-frequency range (>250 cm⁻¹) of the Raman spectra remains largely unaffected by AgI addition except for a small shift of ν_s(PO₂⁻) from 1144 cm⁻¹ for x = 0 to 1138 cm⁻¹ for x = 0.4 and the development of very weak Raman activity at ca. 900–1050 cm⁻¹ with a peak discernible at ca. 1000 cm⁻¹ for x = 0.4. The present results are in agreement with earlier Raman studies on xAgI–(1 − x)AgPO₃ glasses^44,45 but differ substantially from those of recent reports^41,43 that show large changes in the Raman spectra; these were manifested mainly by the progressive loss of intensity of the ν_s(PO₂⁻) band at 1140 cm⁻¹ and its nearly disappearance for x ≥ 0.4 as well as by the growth of a new band at 1008 cm⁻¹ which dominates eventually the Raman spectrum for x ≥ 0.4. The origin of such pronounced differences in the Raman spectra of AgI–AgPO₃ glasses will be examined in Section 3.2.

Evidence for the incorporation of AgI in the glass structure is provided also by infrared absorption below ca. 250 cm⁻¹ where the vibrations of Ag ions against their sites are active. For the AgPO₃ glass a broad band is measured at 128 cm⁻¹ (Fig. 1b, x = 0) and can be attributed to Ag–O vibrations in oxide sites,^14,51,55 while doping of AgPO₃ with AgI causes the downshift of this band to 122 cm⁻¹ for x = 0.4. We note that crystalline β-AgI (wurtzite) and α-AgI (superionic phase) show their main infrared absorption at ca. 110 cm⁻¹ due to Ag–I stretching in tetrahedral iodide site hosting the silver ion.^62,63 Therefore, the downshift of the 128 cm⁻¹ band is consistent with the formation of mixed I/O coordination environments for Ag ions upon increasing AgI content. This is in line with the proposed formation of O–Ag–I conduction sites and their key role in controlling the mobility of Ag ions.³⁷

The higher-frequency infrared spectrum is affected slightly by AgI addition (Fig. 1b). This is manifested in several ways including the decrease of relative intensity at 990 cm⁻¹, the parallel increase in intensity at 895 and 1058 cm⁻¹ and the gradual growth of a shoulder at ca. 1090 cm⁻¹. These spectral changes are systematic and should mirror variations in the metaphosphate structure. They can be explored on the basis of assignments in Table 3 and include mainly a partial destruction of metaphosphate rings (990 cm⁻¹ band), the opening of which would give metaborate chains (bands at 895 and 1058 cm⁻¹). To search for the origin of such changes we consider first the formation of AgPO₃ glass, which can be described by the following reaction driven by thermal decomposition of the AgNO₃ and NH₄H₂PO₄ starting materials:^64,65

AgNO₃ + NH₄H₂PO₄ → AgPO₃ + (3/2)H₂O↑ + NH₃↑ + NO₂↑ + (1/4)O₂↑ (1)

In the presence of AgI, part of the evolved water in reaction (1) may lead to the opening of metaphosphate rings as follows:

The above reaction is consistent with the observed variations in relative intensity of the infrared bands at 990, 895 and 1058 cm⁻¹. Water may induce also the formation of shorter metaphosphate chains and pyro-phosphate species as described below:

(n + 2)Ag⁺[PO_3.5–(PO₃)_n−2–PO_3.5]⁻⁽ⁿ⁺²⁾ + 2AgI + H₂O → nAg⁺[PO_3.5–(PO₃)_n−4–PO_3.5]⁻ⁿ + 4Ag⁺P₂O₇⁴⁻ + 2HI↑ (3)

Evidence for the formation of isolated pyro-phosphate species, (P₂O₇)⁴⁻ or (PO₃–O–PO₃)⁴⁻ = 2Q¹, is provided by the shoulder developing in the IR at ca. 1090 cm⁻¹ which can be attributed to the asymmetric stretch of pyro-phosphate terminal P–O bonds,^21,59,60 denoted ν_as(PO₃²⁻) in Table 3. Indeed, the IR spectrum of the pyro-phosphate glass 0.2AgI–0.8Ag₄P₂O₇ in Fig. 2 shows its strongest band at ca. 1085 cm⁻¹. Formation of a minor amount of (P₂O₇)⁴⁻ by reaction (3) is also consistent with the observed weak Raman feature at ca. 1000 cm⁻¹ (Fig. 1a, x = 0.4), with the strongest Raman signal of (P₂O₇)⁴⁻ being measured at 1002 cm⁻¹, ν_s(PO₃²⁻) (Fig. 2 and Table 3).

Fig. 2 (a) Temperature reduced Raman, and (b) infrared absorption coefficient spectra of pyrophosphates in glassy, 0.2AgI–0.8Ag₄P₂O₇, and polycrystalline, Ag₄P₂O₇, forms.

In summary, incorporation of AgI in glasses AgI–AgPO₃ leaves the metaphosphate structure practically unaffected. Nevertheless, some small but systematic spectral changes were observed to develop in the Raman and infrared spectra with AgI addition and these could be associated with the opening of metaphosphate rings and the creation of shorter metaphosphate chains and minor amount of isolated pyro-phosphate units. It was proposed that such structural changes are induced by water which is evolved in situ during thermal decomposition of NH₄H₂PO₄ used for AgPO₃ glass synthesis (reactions (1–3)). In any case, the minor changes in the Raman and infrared spectra detected in this study cannot explain the strong influence of AgI on the phosphate structure reported recently for AgI–AgPO₃ glasses.^41,43

3.1.2 Properties of glasses xAgI–(1 − x)AgPO₃. Addition of AgI to AgPO₃ decreases drastically the glass transition temperature (T_g) and enhances remarkably the direct current ionic conductivity (σ_dc) as seen in Table 1. It is clear that the increase of ionic conductivity is due to the corresponding reduction in activation energy (E_a) with AgI content. The evolution of properties observed here as a function of AgI content is compared in Fig. 3 with literature data for T_g,^{26,29,30,43,68} σ_dc,^{26,29–31,43,69} and E_a.^{26,29–31,43,68,69} Starting with T_g (Fig. 3a) we note that values reported in ref. 41 and 43 are considerably larger than those found in earlier studies and in the present work. A large difference in T_g is noted already for the AgPO₃ base glass (x = 0) and it spans the range from ΔT_g = 94 °C in comparison to ref. 29 to ΔT_g = 61 °C in comparison to this work. Novita et al.^40–43 attributed such large differences in T_g, as well as in ionic conduction properties (Fig. 3b and c), to large differences in residual water content. It was argued that their method for glass preparation gives dry glasses in contrast to other methods used in literature which produce glasses with, presumably, higher water content.

Fig. 3 Comparison of results obtained in this work with literature data for (a) glass transition temperature, T_g, (b) room temperature dc ionic conductivity, σ_dc, and (c) activation energy, E_a, measured as a function of AgI content for glasses xAgI–(1 − x)AgPO₃.

Because residual water would certainly influence glass properties, it is of interest to investigate the presence of water in glasses synthesized in the present work. For this purpose we used IR spectroscopy which is a very sensitive technique to determine water in glass. Following the approach employed for AgPO₃ glass,⁴⁹ we have measured IR transmittance spectra, T(ν), and present in Fig. 4 the corresponding absorption coefficient spectra, α(ν), as a function of AgI content. The α(ν) spectra were calculated using the expression:

α(ν) = −ln[T(ν)/T(5000)]/d (4)

where d is the sample thickness and the transmittance T(ν) is normalized to the transmittance at 5000 cm⁻¹ to account for reflection losses.⁷⁰ The spectra in Fig. 4 show that it is nearly impossible to prepare absolutely dry AgI–AgPO₃ glasses with the glass preparation conditions employed in this study, as manifested by the presence of the ca. 2815 cm⁻¹ band which is due to O–H stretching of water molecules.^70–72 Nevertheless, the values of the maximum absorption at 2815 cm⁻¹, α_max, in Table 1 show that addition of AgI to AgPO₃ reduces the tendency of glass to retain water.

Fig. 4 Infrared absorption spectra in the region of O–H stretching for glasses prepared in this work in the system xAgI–(1 − x)AgPO₃.

The experimental α_max values can be used to evaluate the water content in glass, C_H2O = α_max/(2.303ε_H2O), but this requires knowledge of the molar extinction coefficient (ε_H2O) for the band at 2815 cm⁻¹. For the AgPO₃ glass we employed previously⁴⁹ the value ε_H2O = 180 L mol⁻¹ cm⁻¹ derived from the conversion factor of about 100 ppm OH/cm⁻¹ reported for phosphate glasses,⁷⁰ while in an earlier study⁷³ of calcium-metaphosphate glasses the value ε_H2O = 120 L mol⁻¹ cm⁻¹ was used for the absorption band at about 3000 cm⁻¹. This suggests that an average value of ε_H2O = 150 L mol⁻¹ cm⁻¹ could be used in the present study. We note also that ε_H2O data have been collected, for boro-silicate glasses mainly, to evaluate water content in glass by IR spectroscopy.⁷⁴ The derived ε_H2O vs. frequency curve indicates the value ε_H2O = 155 L mol⁻¹ cm⁻¹ at 2815 cm⁻¹ in good agreement with the average value noted above. Therefore, we use the value ε_H2O = 150 L mol⁻¹ cm⁻¹ to estimate the water content in AgI–AgPO₃ glasses assuming that the molar extinction coefficient is independent of AgI content. The estimated water content is given in Table 1 and shows a decrease from 0.09 to 0.04 moles H₂O per liter of glass from x = 0 to x = 0.4.

The density data reported in ref. 39 allow for the estimation of water content also in wt% (Table 1), to facilitate comparison with NMR results of Mustarelli et al.⁷⁵ on the effect of water on the thermal properties of AgPO₃ glass. It was found that the water content of glasses prepared in Pt crucibles using NH₄H₂PO₄ and AgNO₃ (as in the present study) decreases by increasing temperature of melt and melting time, leading at the same time to an increase of T_g. The driest AgPO₃ glass was melted at 700 °C and it was found to be NMR-water-free with T_g = 187 °C, while a glass with water content of 0.41 wt% had T_g = 172 °C.⁷⁵ While there are differences in sensitivity and way of calibration when NMR and IR data are used to evaluate water content in glass, we note that our AgPO₃ base glass with IR-evaluated water content of 0.04 wt% has glass transition higher by ΔT_g = 6 °C than the NMR-water-free glass of ref. 75. On these grounds we conclude that the synthesis method employed here gives glasses with very low water content even for the base glass, and the water content decreases further when AgI is added to AgPO₃. This trend should be due mainly to the use of less amount of NH₄H₂PO₄ when synthesizing glasses with increased AgI content, resulting in less water being evolved in situ during thermal decomposition of the NH₄H₂PO₄ starting material (reaction (1)). Therefore, the decrease of T_g by 73 °C upon increasing the amount of AgI from x = 0 to x = 0.4 (Table 1) cannot be related to traces of water but it should reflect a strong plasticizing effect of AgI which, upon incorporation in the glass structure, loosens the overall connectivity of the metaphosphate backbone.^3,7

To understand the origin of differences in T_g and conduction properties between earlier reports and the Novita et al.^40–43 data for AgPO₃ base glass (see Fig. 3), we synthesized in our previous study⁴⁹ AgPO₃ glasses using P₂O₅ and Ag₃PO₄ as starting materials like in ref. 40–43. When Pt crucible was used for melting at 900 °C for 2 h, the resulted glass had practically identical Raman and IR spectra, as well as T_g and conductivity values, with the glass prepared from NH₄H₂PO₄ and AgNO₃ in Pt crucible under milder conditions (melting at 700 °C for 2 h). However, melting P₂O₅ and Ag₃PO₄ in alumina crucible at 700 °C for 22 h, similarly to Novita et al.,^40–43 resulted in AgPO₃ glass with T_g = 255 °C and improved electrical conductivity. This was attributed to the unintentional contamination of AgPO₃ with up to about 3 mol% Al₂O₃, as was verified by comparison with spectra and properties of glasses xAl₂O₃–(1 − x)AgPO₃ (x = 0.01, 0.03, 0.05) prepared from NH₄H₂PO₄, AgNO₃ and Al₂O₃ in Pt crucibles.⁴⁹

Based on these findings, we return now to AgI–AgPO₃ glasses and summarize in Table 4 the synthesis protocols employed by authors whose T_g and conductivity data are compared in Fig. 3. While a variety of crucibles was used for melting, what really differentiates the synthesis method of ref. 43 from the others is the combination of alumina crucible with corrosive P₂O₅ and very long melting time (i.e., overnight). This synthesis protocol and the previous discussion for the AgPO₃ base glass, suggest that the use of alumina crucible with P₂O₅ and long melting times may cause unintentional contamination of AgI–AgPO₃ glasses with Al₂O₃ extracted from the crucible during melting. This possibility is examined in the following section.

Table 4 Summary of synthesis protocols for glasses xAgI–(1 − x)AgPO₃ with properties shown in Fig. 3 as a function of AgI content

Crucible	Starting materials	Melting temperature	Melting time	References
Alumina	AgPO3 + AgI	530–600 °C	2 h	26
Not specified	NH4H2PO4 + AgNO3 + AgI	527 °C	Several h	29
Pyrex	NH4H2PO4 + AgNO3 + AgI	500 °C	Not specified	30
Pt	NH4H2PO4 + AgNO3 + AgI	700 °C	30 min	31
Not specified	Ag3PO4 + P2O5 + AgI	900 °C	Not specified	41
Alumina	Ag3PO4 + P2O5 + AgI	700 °C	Overnight	43
Porcelain	NH4H2PO4 + AgNO3 + AgI	Not specified	15 min	68
Not specified	AgPO3 + AgI	600–700 °C	1–2 h	69
Pt	NH4H2PO4 + AgNO3 + AgI	700–900 °C	2 h	This work

---

3.2 Glasses yAl₂O₃–(1 − y)[0.4AgI–0.6AgPO₃]

Glasses with composition yAl₂O₃–(1 − y)[0.4AgI–0.6AgPO₃] were prepared following the synthesis protocols outlined in Table 2 and their Raman and infrared spectra are presented and discussed below.

3.2.1 Vibrational spectra of glasses yAl₂O₃–(1 − y)[0.4AgI–0.6AgPO₃]. The Raman and IR spectra of glasses yAl₂O₃–(1 − y)[0.4AgI–0.6AgPO₃] are reported in Fig. 5a and b, respectively. Starting with the Raman spectra of the binary glasses B1 and B2, because they have the same nominal composition 0.4AgI–0.6AgPO₃ and were prepared from the same starting materials, we find that the use of alumina crucible results in well-defined spectral changes for glass B2. These are manifested at high frequencies by the broadening and downshift of the ν_s(PO₂⁻) band to 1135 cm⁻¹ and the enhancement of features at 958, 1002, 1023, and 1082 cm⁻¹, while the 713 cm⁻¹ shoulder gains relative intensity in the region of ν_s(P–O–P). All these changes become more pronounced for glass B3, for the preparation of which alumina crucible was combined with the use of P₂O₅. The similarities of the Raman features of the ‘binary’ B3 glass with those of the ternary B4 glass with 4 mol% Al₂O₃ are striking and suggest similar structural characteristics, which are definitely different from those of the binary metaphosphate glass B1. Based on the Raman profiles, we suggest that glass B3 is unintentionally contaminated with Al₂O₃ leached from the crucible and that it contains more than 4 mol% Al₂O₃, while glass B2 should be contaminated with a lower amount of Al₂O₃. This difference between B3 and B2 should be attributed to the fact that P₂O₅ being the anhydride of the strong phosphoric acid is corrosive and reacts effectively with the basic Al₂O₃. On the other hand, NH₄H₂PO₄ is a salt of the dihydrogen phosphate ion, H₂PO₄⁻, which is the conjugate base of phosphoric acid and, as a weaker acid, is relatively less reactive to Al₂O₃. For the ternary glass B5, with 7 mol% Al₂O₃, the features at 958, 1023, 1082 and 1120 cm⁻¹ have almost disappeared while the 1002 cm⁻¹ band dominates the Raman spectrum and weaker features appear at 713 and 914 cm⁻¹. Comparison with the Raman spectrum of pyrophosphate species in Fig. 2a suggests that glass B5 should contain mainly isolated pyrophosphate species (P₂O₇)⁴⁻ with characteristic bands at 1002 cm⁻¹, ν_s(PO₃²⁻), and 713 cm⁻¹, ν_s(P–O–P), a smaller amount of other phosphate species giving the 914 cm⁻¹ Raman band, and no metaphosphate-type units.

Fig. 5 (a) Temperature reduced Raman, and (b) infrared absorption coefficient spectra of glasses yAl₂O₃–(1 − y)[0.4AgI–0.6AgPO₃]. For glass preparation conditions see Table 2.

A continuous evolution of glass structure is suggested also by the corresponding infrared spectra in Fig. 5b. Starting with the metaphosphate band of B1 at 1235 cm⁻¹, ν_as(PO₂⁻), we note its progressive downshift and reduction in intensity from B1 to B4 and its disappearance for B5. In agreement with Raman spectrum, the IR spectrum of B5 shows the dominant presence of isolated pyrophosphate species with ν_as(PO₃²⁻) at 1085 cm⁻¹, ν_s(PO₃²⁻) at ca. 1000 cm⁻¹, ν_s(P–O–P) at 712 cm⁻¹ and ν_as(P–O–P) at 900 cm⁻¹. The IR shoulder at ca. 958 cm⁻¹ is indicative of a minor amount of a different type of phosphate species, which should exist also in glasses B2 to B4.

We note at this point that the evolution of the Raman and IR spectra in Fig. 5 is very similar to that exhibited by the Raman^41,43 and IR⁴³ spectra of glasses xAgI–(1 − x)AgPO₃ which were prepared in alumina crucibles using P₂O₅ and long melting times. Based on our findings in Fig. 5, we suggest that the structure and properties of glasses prepared and studied in ref. 41 and 43 deviate from those of the pristine xAgI–(1 − x)AgPO₃ metaphosphate series and should rather correspond to alumino-phosphate glasses with composition changing progressively with increasing x from metaphosphate toward pyrophosphate; this being probably the result of Al₂O₃ leaching from the alumina crucible at prolong melting. Thus, deviations from the metaphosphate composition and structure should be the main reason for the observed differences in Fig. 3 between physical properties of ref. 41 and 43 and those of earlier studies and this work.

3.2.2 Modifying role of Al₂O₃ in glasses yAl₂O₃–(1 − y)[0.4AgI–0.6AgPO₃]. To explain the effect of Al₂O₃ on the phosphate structure of glasses yAl₂O₃–(1 − y)[0.4AgI–0.6AgPO₃] which were prepared here in Pt crucibles with y = 0 (B1), y = 0.04 (B4) and y = 0.07 (B5), we consider their nominal O/P ratio assuming that Al₂O₃ acts as typical glass modifier. Then the O/P ratio increases from 3.0 (B1) to 3.21 (B4) and to 3.38 (B5). Recalling that the O/P value is 3.0 and 3.5 for metaphosphate and pyrophosphate compositions respectively, it is clear that the nominal composition of B4 and B5 is between those of meta- and pyrophosphate with glass B5 approaching the pyrophosphate composition. Thus, the phosphate network of B4 and B5 is over-modified with respect to B1; the same should hold for glasses B2 and B3 judging from their Raman and IR spectra. The effect of Al₂O₃ incorporation in the structure of Ag-metaphosphate glasses appears similar to that of In₂O₃ in Zn-phosphate glasses of composition (50 − x)ZnO–xIn₂O₃–50P₂O₅.⁷⁶ As shown by Koudelka et al.⁷⁶ replacement of ZnO by In₂O₃ leads to the progressive depolymerization of metaphosphate chains due to the increase in the O/P ratio. Besides the same effect of Al₂O₃, one could not exclude here the possibility of some additional phosphate modification by reactions similar to reaction (3) which may lead to some losses of iodine by conversion of AgI to Ag₂O and HI.

Having considered the modifying role of Al₂O₃, which is introduced either unintentionally (B2, B3) or intentionally (B4, B5) in the phosphate structure, we explore next the origin of the Raman bands at 914, 958, 1023, 1082 and 1120 cm⁻¹ in Fig. 5a to understand the nature of phosphate species formed by Al₂O₃-induced modification of the phosphate structure of B1. In this context, we note that the evolution of the Raman spectra in Fig. 5a is very similar to that found for Ag-phosphate glasses Ag₂O–nP₂O₅ with 0.66 ≤ n ≤ 1.⁷⁷ Besides some differences in peak frequencies, there is good resemblance of the Raman spectra of glasses B4 and B3 with those of Ag₂O–nP₂O₅ having n = 0.75 (O/P = 3.17) and n = 0.66 (O/P = 3.26) respectively.

As the Ag₂O content increases above the metaphosphate composition in glasses Ag₂O–nP₂O₅ (i.e., n < 1) new Raman bands develop at 964, 1008, 1090 and 1120 cm⁻¹.⁷⁷ The bands at 1008 and 1090 cm⁻¹ were attributed to P–O stretch in end groups of metaphosphate chains, the band at 1120 cm⁻¹ to ν_s(PO₂⁻) in metaphosphate chains, and the 964 cm⁻¹ band to the totally symmetric stretch of orthophosphate PO₄³⁻ units (Q⁰).⁷⁷ In regards to this 964 cm⁻¹ band, the closest band in Fig. 5a is the one at 958 cm⁻¹. Since the latter band is not present in the Raman spectrum of B5 which has a more modified phosphate network, the 958 cm⁻¹ band should arise from phosphate species other than Q⁰. Interestingly, a weak band also at 958 cm⁻¹ develops in the IR spectra of Fig. 5b and exists as a shoulder for glass B5. This observation shows clearly that the Raman and IR bands at 958 cm⁻¹ could not originate from the same phosphate species.

Considering the previous attribution of the 964 cm⁻¹ band to PO₄³⁻ units,⁷⁷ we present in Fig. 6 the Raman and IR spectra of polycrystalline silver-orthophosphate (Ag₃PO₄). The totally symmetric stretch of PO₄³⁻ units, ν₁(PO₄³⁻), gives rise to the strong Raman band at 910 cm⁻¹ while the asymmetric stretch, ν₃(PO₄³⁻), has very strong IR activity at 960 cm⁻¹ (see Table 3). Therefore, the presence of orthophosphate PO₄³⁻ units in glasses developed in this study would be signaled by the 914 cm⁻¹ Raman band and the IR feature at 958 cm⁻¹.

Fig. 6 (a) Raman, and (b) infrared absorption coefficient spectra of polycrystalline silver-orthophosphate, Ag₃PO₄.

3.2.3 P–O bonding in glasses yAl₂O₃–(1 − y)[0.4AgI–0.6AgPO₃]: a bond valence approach. The origin of the remaining Raman bands at 958, 1023, 1082 and 1120 cm⁻¹ can be sought in terms of the bond valence (BV) approach introduced by Brow and coworkers to explain the decrease in the Raman stretching frequency of terminal P–O bonds, ν(P–O), upon increasing ZnO content in glasses xZnO–(1 − x)P₂O₅ with 0.50 ≤ x ≤ 0.71.⁵⁷ In their approach, ν(P–O) was correlated with the P–O bond strength expressed in valence units (vu). To obtain the P–O bond strength in a phosphate tetrahedron the valence of phosphorus (+5) is distributed along the four P–O bonds, with the restriction that the sum of the vu values assigned to the bonds connected to an oxygen atom must equal two in order to satisfy the valence requirements of oxygen. Along these lines, the strength of the P–O terminal bond decreases from 1.50 vu in metaphosphate (Q²) to 1.33 vu in pyrophosphate (Q¹) and to 1.25 vu in orthophosphate (Q⁰) structures, and ν(P–O) correlates linearly with the strength of the terminal P–O bond, s(P–O).⁵⁷ As discussed by Brow and co-workers, the resulting valence unit values for P–O bonds are not expected to be quantitatively accurate. However, they give useful relative bond strengths and thus relative Raman stretching frequencies of the corresponding bonds, and this allows probing the evolution of structure and bonding in phosphates systems.

It is of interest to examine whether the vibrational response of neutral Q³ tetrahedra fits also in the above BV approach, considering that Q³ constitutes the structure building unit of v-P₂O₅ and remains a basic structural unit in ultraphosphate glasses with metal oxide content below 50 mol%.⁵⁸ For the Q³ tetrahedron, it is straightforward to realize that each of the three bridging P–O(P) bonds has strength of 1.0 vu while the terminal P=O bond has strength of 2.0 vu. The corresponding Raman stretching frequencies of these bonds in v-P₂O₅ are ν(P–O–P) = 635 cm⁻¹ and ν(P=O) = 1380 cm⁻¹ according to ref. 78. The ν(P–O) versus s(P–O) data for Q², Q¹, and Q⁰ units in Ag-phosphate glasses studied in this work are shown in Fig. 7 together with data for v-P₂O₅. Data are included also for Zn-phosphate⁵⁷ and Li-phosphate⁷⁹ glasses for which characteristic Raman bands have been identified for Q², Q¹, and Q⁰ units.

Fig. 7 Correlation between the Raman frequency of phosphorus–oxygen stretch, ν(P–O) in cm⁻¹, and the bond strength, s(P–O) in valence units, for Li-, Zn- and Ag-phosphate glasses. The lines are exponential fits to the data according to eqn (5).

As shown in Fig. 7 the symmetric Raman stretching frequency ν(P–O) is found to correlate well with the strength s(P–O) of the bond, with the v-P₂O₅ data forming nicely the lower and upper limits in the correlations of the three phosphate glass systems. The correlation between ν(P–O) (in cm⁻¹) and s(P–O) (in vu) can be described by the exponential equation:

ν(P–O) = a − b exp[−ks(P–O)] (5)

The derived fitting parameters of eqn (5) in the three M-phosphate systems are in the order M = Ag, Li, Zn: a = 1627.8, 1531.4, 1496.1 cm⁻¹, b = 4049.3, 5450.6, 6741.0 cm⁻¹, and k = 1.40, 1.80, 2.05 vu⁻¹ (R² = 0.998, 0.998, 0.996). It is noted that an exponential expression was found to describe also the correlation between B–N stretching frequency and bond length in boron–nitrogen materials including boron–oxynitride amorphous thin films.⁸⁰

Having established eqn (5) for metal-phosphate glasses, we return now to bands at 958, 1023, 1082 and 1120 cm⁻¹ measured in the Raman spectra of glasses yAl₂O₃–(1 − y)[0.4AgI–0.6AgPO₃] (Fig. 5). Since eqn (5) was derived for symmetric stretching vibrations, we examine first the polarization characteristics of the bands in question before applying eqn (5). Polarized Raman spectra in the VV and VH backscattering geometries for glasses with y = 0.04 and 0.07 are shown in Fig. 8. It is clear that the bands at 958, 1023, 1082 and 1120 cm⁻¹ are polarized and thus they correspond to totally symmetric modes, as is the case with all other bands already assigned to symmetric stretching vibrations, i.e. ν_s(PO₂⁻) at ca. 1140 cm⁻¹, ν_s(PO₃²⁻) at ca. 1000 cm⁻¹, and ν_s(P–O–P) in the region 630–780 cm⁻¹.

Fig. 8 Polarized Raman spectra of glasses yAl₂O₃–(1 − y)[0.4AgI–0.6AgPO₃] for y = 0.04 (a), and y = 0.07 (b). The Raman spectra were measured in the backscattering geometry with polarizations VV and VH, where the first letter indicates the polarization of the exciting laser beam and the second the polarization of the collected scattered light.

Application of eqn (5) for Ag-phosphate glasses, with s = ln[b/(a − ν)]/k, gives for the Raman bands at 1120 and 1082 cm⁻¹ terminal P–O bond strengths of 1.48 and 1.43 vu, which are just below the nominal 1.50 vu strength in the PO₂⁻ unit of metaphosphate tetrahedra, Q². This suggests that the 1120 and 1082 cm⁻¹ Raman bands can be attributed to ν_s(PO₂⁻) in shorter (fragmented) metaphosphate chains caused by Al₂O₃-induced modification. The large bandwidth of the 1120–1082 cm⁻¹ envelop is indicative of a distribution of P–O bond strengths and chain lengths, as was the case for modified zinc-phosphate glasses.⁵⁷ Therefore, the broadening and downshift of the ν_s(PO₂⁻) mode from 1138 to 1120 cm⁻¹ and to 1082 cm⁻¹ in Fig. 5a suggests a progressive weakening of the terminal P–O bond in PO₂⁻ units as Al₂O₃ is introduced in the phosphate structure. This point of view is consistent with earlier NMR studies on sodium-phosphate glasses, where the shift of the Q²–³¹P peak to less negative values was correlated with a decreasing bonds order of terminal P–O bonds upon increasing Na₂O content.⁸¹

The Raman features at 1023 cm⁻¹ and 958 cm⁻¹, which show similar dependence on Al₂O₃ content, correspond to terminal P–O bond strengths of 1.36 and 1.29 vu respectively. These values are above and below the nominal P–O strength of 1.33 vu in PO₃²⁻ units of isolated pyrophosphate species (P₂O₇)⁴⁻, for which eqn (5) gives a frequency value of 1002 cm⁻¹ for the bond strength of 1.33 vu. Therefore, while the Raman band at 1002 cm⁻¹ is the clear signature of isolated pyrophosphate species, the bands at 1023 and 958 cm⁻¹ could signal the presence of PO₃²⁻ end groups in short phosphate fragments, just before they convert to isolated (P₂O₇)⁴⁻ species at higher Al₂O₃ contents.

To examine this scenario, we explore possible structures in glasses yAl₂O₃–(1 − y)[0.4AgI–0.6AgPO₃] which would be compatible with the Raman features at 1023 and 958 cm⁻¹. For simplicity, we consider the composition y = 0.0625 which is between y = 0.04 and 0.07 and, when ignoring AgI, gives the stoichiometry Al₂O₃–9AgPO₃. A structural unit corresponding to this stoichiometry is shown in Fig. 9; this unit resulting as the product of Al₂O₃-induced opening of three-member metaphosphate rings. A similar scheme has been proposed by Brow et al.⁸² for Al₂O₃-modified NaPO₃ glass with 10 mol% Al₂O₃. Evidence for the gradual destruction of three-member metaphosphate rings, (P₃O₉)³⁻, is provided by the intensity reduction from B1 to B5 of the ca. 778 cm⁻¹ IR band (Fig. 5b). According to Rulmont et al.⁵⁹ an IR band in the region 765–780 cm⁻¹ is characteristic of ν_s(P–O–P) in cyclotriphosphates, (P₃O₉)³⁻. For reasons of stoichiometry, aluminum should be six-fold coordinated to oxygen, Al(6), in the proposed aluminum-triphosphate Al_2/3Ag₃P₃O₁₀ structure (Fig. 9). This is in line with NMR spectroscopy on Al₂O₃–NaPO₃ glasses which showed that octahedral Al sites are the dominant aluminate sites in glasses with Al₂O₃ content lower than 12.5 mol%.⁸²

Fig. 9 Schematic presentation of a three-member metaphosphate ring, 3Ag⁺(P₃O₉)³⁻, which opens by interaction with Al₂O₃ to form aluminum-triphosphate, Al_2/3Ag₃P₃O₁₀, where aluminium is six-fold coordinated to oxygen, Al(6). The assignment of P–O bond strengths a–d (in vu) in the Al_2/3Ag₃P₃O₁₀ structure was based on the combination of the bond valence approach with experimental Raman frequencies in Fig. 5 and eqn (5).

We apply now the BV approach to the Al_2/3Ag₃P₃O₁₀ structure and assign P–O bond strengths in vu values as shown schematically in Fig. 9. Starting with the four P–O(P) bridging bonds, each bond should have strength of a = 1.0 vu to satisfy the oxygen valence. This leaves strength of b = 1.5 vu for each P–O terminal bond in the PO₂⁻ unit in order for the phosphorus atom in Q² to have a sum of 5.0 vu. To satisfy the valence of the two terminal oxygen atoms in PO₂⁻, the charge-balancing Ag⁺ ion contributes 1/2 = 0.5 vu to each Ag–O bond. We consider next the two PO₃²⁻ end groups, where each Al(6) contributes 3/6 = 0.5 vu to the Al(6)–O(P) bond. The question now is how to distribute the remaining phosphorus valence of 4.0 vu in the PO₃²⁻ end groups among three bonds, i.e. two P–O(Al) and one P–O(Ag) bond. To tackle this question we use the experimental Raman frequencies of 1023 and 958 cm⁻¹ and make the assumption that the 958 cm⁻¹ band corresponds to the symmetric stretching of the two terminal P–O(Ag) bonds. Then eqn (5) gives strength of c = 1.29 vu for the P–O(Ag) bond, and this requires a contribution of 0.71 vu from Ag to the Ag–O(P) bond to satisfy the oxygen valence. For charge neutrality, the remaining valence on Ag (i.e., 0.29 vu) should be equally distributed to the oxygen atoms in the two Al(6)–O(P) bonds. Therefore, the strength of each P–O(Al) bond should be equal to d = 2–0.5–(0.29/2) = 1.355 vu. This gives for phosphorus in the PO₃²⁻ end groups a sum of valences equal to 2 × 1.355 + 1.29 + 1.0 = 5.0 vu as expected. To check the compatibility of the estimated P–O(Al) bond strength of d = 1.355 vu with experiment, we use eqn (5) to calculate the Raman frequency for the corresponding P–O(Al) symmetric stretch. The result is 1020 cm⁻¹, in very good agreement with the experimental Raman value of 1023 cm⁻¹. For convenience, the final assignments of Raman and infrared bands for glasses yAl₂O₃–(1 − y)[0.4AgI–0.6AgPO₃] are summarized in Table 5.

Table 5 Frequencies and assignments of Raman and infrared bands for the Al₂O₃-containing glasses yAl₂O₃–(1 − y)[0.4AgI–0.6AgPO₃] studied in this work

Raman (cm^−1)	Infrared (cm^−1)	Assignment
a Vibration mode: ν = stretching, δ = bending, νs, ν1 = symmetric stretching, νas, ν3 = asymmetric stretching, ν2 = symmetric bending, ν4 = asymmetric bending.b Q^i: a phosphate tetrahedron with i bridging and 4-i terminal oxygen atoms.
1180–1260	1235	νas(PO2^−)^a, Q^2^b in meta-phosphate chains/rings
	1225–1202	νas(PO2^−), Q^2, in shorter metaphosphate chains
1138		νs(PO2^−), Q^2, in meta-phosphate chains/rings
1120, 1082		νs(PO2^−), Q^2, in shorter metaphosphate chains
1023		νs(P–O(Al)), in Al2/3Ag3P3O10
1002	1000	νs(PO3^2−), Q^1
958		νs(P–O(Ag)), in Al2/3Ag3P3O10
	958	ν3(PO4^3−), Q^0
914		ν1(PO4^3−), Q^0
	900	νas(P–O–P), Q^1
713	712	νs(P–O–P), Q^1
	530, 555, 590	δas(PO3^2−), Q^1
350	410, 470, 505	δs(PO3^2−), Q^1
	170	ν(Ag–O) + δ(P–O–P), Q^1
120		ν(Ag–I)
45		Boson mode

---

We note that assignment of the 1023 cm⁻¹ band to the symmetric stretch of the two P–O(Ag) bonds in the PO₃²⁻ end groups, and application of the approach described above would give strength of 1.32 vu for the P–O(Al) bonds. According to eqn (5) the corresponding Raman band would be at 990 cm⁻¹, but this does not agree with the experimental spectra of Fig. 5a.

In conclusion, the Al_2/3Ag₃P₃O₁₀ structure of Fig. 9, with its assigned bond valences, is consistent with the measured Raman spectra. Participation of Al in P–O–Al(6) bonding causes the splitting of the P–O stretching mode in PO₃²⁻ groups into two modes; one at 958 cm⁻¹ due to P–O(Ag) stretching and the other at 1023 cm⁻¹ for the P–O(Al) stretching. Therefore, combination of Raman spectra with the bond valence approach provides solid evidence for the incorporation of Al in the structure of glasses B2–B5 synthesized according to the protocols in Table 2.

4. Conclusions

Glasses xAgI–(1 − x)AgPO₃, 0 ≤ x ≤ 0.4, were synthesized using AgI, AgNO₃ and NH₄H₂PO₄ starting materials, handled under dry ambient conditions and melted in Pt crucibles. Glass transition temperature, ionic conductivity, Raman, IR reflectance and transmission measurements were performed to evaluate properties, structure and residual water as a function of AgI content.

Raman and IR reflectance spectroscopy showed that AgI additions to the AgPO₃ glass leave the metaphosphate structure largely unaffected, in line with earlier Raman studies on similar glasses prepared under ambient laboratory conditions.^31,44,45 Some small changes in the Raman and IR spectra were observed in this work to develop with AgI content and were attributed to the opening of metaphosphate rings and the creation of shorter metaphospate chains, due to interaction with water evolving in situ from the thermal decomposition of NH₄H₂PO₄ used in glass synthesis. Quantification of the water content by IR transmission spectroscopy showed that the method employed here gives glasses with very low water content which decreases further as AgI is added to AgPO₃, i.e. from 0.09 to 0.04 moles H₂O per liter of glass from x = 0 to x = 0.4. Such minor changes detected in the Raman and IR spectra of this study cannot explain the strong influence of AgI on the phosphate structure observed earlier^41,43 for AgI–AgPO₃ glasses prepared by combining alumina crucibles with corrosive P₂O₅ and long melting times.

To explore the origin of such pronounced differences, we synthesized glasses in the system yAl₂O₃–(1 − y)[0.4AgI–0.6AgPO₃] with y = 0.0, 0.04 and 0.07 using Pt and alumina crucibles, different starting materials (including P₂O₅) and different melting temperatures and times. Raman and IR spectroscopy showed that glasses prepared in alumina crucibles with y = 0.0 exhibit spectral trends which are very close to those of glasses prepared in Pt crucibles with y = 0.04 and 0.07. These results demonstrate clearly that prolong melting in alumina crucibles, especially in the presence of P₂O₅, leads to doping the 0.4AgI–0.6AgPO₃ glass with Al₂O₃ leached from the crucible, a process causing the increase of the O/P ratio and, thus, the progressive change of composition, structure and properties from metaphosphate toward pyrophosphate. These findings explain the observed deviations in glass transition temperature and ionic conduction properties of glasses AgI–AgPO₃ prepared in alumina crucibles under corrosive conditions^41,43 from those reported in this work and in previous studies on glasses prepared under milder conditions.

To understand the nature of alumino-phosphate species formed by Al₂O₃-induced modification of the metaphosphate structure and to assist in assigning key Raman features, we employed the bond valence approach introduced in Raman studies of phosphate glasses.⁵⁷ By extending this approach to Raman data from v-P₂O₅, it was found that an exponential relationship describes the dependence of the P–O symmetric stretching frequency on P–O bond strength in Li-, Zn- and Ag-phosphate glasses. Application of this relationship, in combination with bond valence principles, allowed the proposition of the Al-triphosphate species Al_2/3Ag₃P₃O₁₀ as the main species formed when the glass 0.4AgI–0.6AgPO₃ is doped intentionally (y = 0.04 and 0.07) or unintentionally (alumina crucible) with Al₂O₃.

Acknowledgements

Partial support by the Greek General Secretariat for Research and Technology and the European Commission, through the European Social Fund for Regional Development NSRF 2007–2013 action ‘Development of Research Centres – KPHΠIΣ’, project 447963 ‘New Multifunctional Nanostructured Materials and Devices – POLYNANO’, is gratefully acknowledged. I. K. acknowledges also support from the NHRF/EC grant HPMD-CT-2000-00033.

References

1. T. Minami, J. Non-Cryst. Solids, 1985, 73, 273–284.
2. H. L. Tuller and M. W. Basroum, J. Non-Cryst. Solids, 1985, 73, 331–350.
3. M. D. Ingram, Phys. Chem. Glasses, 1987, 28, 215–234.
4. J. L. Souquet, Solid State Ionics, 1988, 28–30, 693–703.
5. W.-K. Lee, J. F. Liu and A. S. Nowick, Phys. Rev. Lett., 1991, 67, 1559–1561.
6. M. Tatsumisago, Y. Shinkuma and T. Minami, Nature, 1991, 354, 217–218.
7. C. A. Angell, Annu. Rev. Phys. Chem., 1992, 43, 693–717.
8. K. Funke, Prog. Solid State Chem., 1993, 22, 111–195.
9. E. I. Kamitsos, J. A. Kapoutsis, G. D. Chryssikos, G. Taillades, A. Pradel and M. Ribes, J. Solid State Chem., 1994, 112, 255–261.
10. M. Tatsumisago, N. Torata, T. Saito and T. Minami, J. Non-Cryst. Solids, 1996, 196, 193–198; M. Tatsumisago, T. Saito and T. Minami, J. Non-Cryst. Solids, 2001, 293, 10–18.
11. J. W. Zwanziger, K. K. Olsen and S. L. Tagg, Phys. Rev. B: Condens. Matter Mater. Phys., 1993, 47, 14618–14621; K. K. Olsen, J. W. Zwanziger, P. Hartmann and C. Jager, J. Non-Cryst. Solids, 1997, 222, 199–205.
12. J. Kincs and S. W. Martin, Phys. Rev. Lett., 1996, 76, 70–73.
13. R. Lewandowska, K. Krasowski, R. Bacewicz and J. E. Garbarczyk, Solid State Ionics, 1999, 119, 229–234; J. E. Garbarczyk, T. K. Pietrzak, M. Wasiucionek, A. Kaleta, A. Dorau and J. L. Nowinski, Solid State Ionics, 2015, 272, 53–59.
14. C. P. Varsamis, E. I. Kamitsos, N. Machida and T. Minami, J. Phys. Chem. B, 1997, 101, 3734–3741; C. P. Varsamis, E. I. Kamitsos and G. D. Chryssikos, Phys. Rev. B: Condens. Matter Mater. Phys., 1999, 60, 3885–3898.
15. H. Jain, S. Krishnaswami and O. Kanert, J. Non-Cryst. Solids, 2002, 307, 1017–1023.
16. K. Funke, R. D. Banhatti and C. Cramer, Phys. Chem. Chem. Phys., 2005, 7, 157–165.
17. P. Vinatier and Y. Hamon, Applications of Ion Transport in Disordered Solids, in Charge Transport in Disordered Solids, ed. S. Baranovski, Wiley and Sons, 2006, pp. 403–433.
18. M. D. Ingram, C. T. Imrie, J. Ledru and J. M. Hutchinson, J. Phys. Chem. B, 2008, 112, 859–866; M. D. Ingram and C. T. Imrie, Solid State Ionics, 2011, 196, 9–17.
19. S. Bhattachaya and A. Ghosh, J. Phys. Chem. C, 2010, 114, 5745–5750; B. Deb and A. Ghosh, J. Phys. Chem. C, 2011, 115, 14141–14147.
20. F. Behrends and H. Eckert, J. Phys. Chem. C, 2011, 115, 17175–17183; J. Tsuchida, J. Schneider, M. T. Rinke and H. Eckert, J. Phys. Chem. C, 2011, 115, 21927–21941.
21. C. P. E. Varsamis, E. I. Kamitsos, T. Minami and N. Machida, J. Phys. Chem. C, 2012, 116, 11671–11681.
22. A. Sanson, A. C. Armellini, R. Grissenti and P. Fornasini, J. Phys. Chem. C, 2013, 117, 6081–6087.
23. R. Christensen, G. Olson and S. W. Martin, J. Phys. Chem. B, 2013, 117, 2169–2179.
24. M. Tatsumisago and A. Hayashi, Int. J. Appl. Glass Sci., 2014, 5, 226–235; T. Hakari, M. Nagao, A. Hayashi and M. Tatsumisago, Solid State Ionics, 2014, 262, 147–150.
25. K. Funke, R. D. Banhatti, P. Grabowski, J. Nowinski, W. Wrobel, R. Dinnebier and O. Magdysyuk, Solid State Ionics, 2015, 271, 2–9.
26. J. P. Malugani, A. Wasniewski, M. Doreau, G. Robert and A. A. Rikabi, Mater. Res. Bull., 1978, 13, 427–433.
27. S. W. Martin and A. Schiraldi, J. Phys. Chem., 1985, 89, 2070–2076; S. W. Martin, J. Am. Ceram. Soc., 1991, 74, 1767–1783; S. W. Martin, Eur. J. Solid State Inorg. Chem., 1991, 28, 163–205.
28. B. Roling, M. D. Ingram, M. Lange and K. Funke, Phys. Rev. B: Condens. Matter Mater. Phys., 1997, 56, 13619–13622.
29. M. B. M. Mangion and G. P. Johari, Phys. Chem. Glasses, 1988, 29, 225–234.
30. D. L. Sidebottom, Phys. Rev. B: Condens. Matter Mater. Phys., 2000, 61, 14507–14516.
31. S. Bhattacharya and A. Ghosh, Phys. Rev. B: Condens. Matter Mater. Phys., 2004, 70, 177703; S. Bhattacharya, D. Dutta and A. Ghosh, Phys. Rev. B: Condens. Matter Mater. Phys., 2006, 73, 104201.
32. H. Takahashi, H. Nakanii, T. Sakuma and Y. Onoda, Solid State Ionics, 2008, 179, 2137–2141; H. Takahashi, H. Nakanii and T. Sakuma, Solid State Ionics, 2005, 176, 1067–1072.
33. G. Carini, M. Cutroni, A. Fontana, G. Mariotto and F. Rocca, Phys. Rev. B: Condens. Matter Mater. Phys., 1984, 29, 3567–3572.
34. M. Tachez, R. Mercier, J. P. Malugani and A. J. Dianoux, Solid State Ionics, 1986, 20, 93–98; C. Rousselot, M. Tachez, J. P. Malugani, R. Mercier and P. Chieux, Solid State Ionics, 1991, 44, 151–158.
35. J. D. Wicks, L. Borjesson, G. Bushnell-Wye, W. S. Howells and R. L. McGreevy, Phys. Rev. Lett., 1995, 74, 726–729; J. Swenson and L. Borjesson, Phys. Rev. Lett., 1996, 77, 3569–3572.
36. S. Adams and J. Swenson, Phys. Rev. Lett., 2000, 84, 4144–4147; A. Hall, J. Swenson, S. Adams and C. Meneghini, Phys. Rev. Lett., 2008, 101, 195901.
37. P. Mustarelli, C. Tomasi and A. Magistris, J. Phys. Chem. B, 2005, 109, 17417–17421.
38. A. Sanson, F. Rocca, C. Armellini, G. Dalba, P. Fornasini and R. Grisenti, Phys. Rev. Lett., 2008, 101, 155901.
39. A. C. M. Rodrigues, M. L. F. Nascimento, C. B. Bragatto and J. L. Souquet, J. Chem. Phys., 2011, 135, 234504.
40. D. I. Novita and P. Boolchand, Phys. Rev. B: Condens. Matter Mater. Phys., 2007, 76, 184205.
41. D. I. Novita, P. Boolchand, M. Malki and M. Micoulaut, Phys. Rev. Lett., 2007, 98, 195501.
42. M. Micoulaut, M. Malki, D. I. Novita and P. Boolchand, Phys. Rev. B: Condens. Matter Mater. Phys., 2009, 80, 184205.
43. D. I. Novita, P. Boolchand, M. Malki and M. Micoulaut, J. Phys.: Condens. Matter, 2009, 21, 205106.
44. J. P. Malugani and R. Mercier, Solid State Ionics, 1984, 13, 293–299.
45. A. Fontana, F. Rossi, C. Armellini, G. D'Angelo, G. Tripodo and A. Bartolotta, Philos. Mag. B, 1999, 79, 2073–2080.
46. B. P. Rodrigues and L. Wondraczek, J. Chem. Phys., 2013, 138, 244507.
47. I. Konidakis, G. Zito and S. Pissadakis, Opt. Lett., 2012, 37, 2499–2501; I. Konidakis, G. Zito and S. Pissadakis, Opt. Lett., 2014, 39, 3374–3377; I. Konidakis and S. Pissadakis, Materials, 2014, 7, 5735–5745.
48. P. Grabowski, J. L. Nowinski and K. Kwatek, J. Non-Cryst. Solids, 2015, 415, 51–56.
49. I. Konidakis, C. P. E. Varsamis and E. I. Kamitsos, J. Non-Cryst. Solids, 2011, 357, 2684–2689.
50. R. F. Bartholomew, J. Non-Cryst. Solids, 1972, 7, 221–235.
51. E. I. Kamitsos, J. A. Kapoutsis, G. D. Chryssikos, J. M. Hutchinson, A. J. Pappin, M. D. Ingram and J. A. Duffy, Phys. Chem. Glasses, 1995, 36, 141–149.
52. G. B. Rouse, P. J. Miller and W. M. Risen, J. Non-Cryst. Solids, 1978, 28, 193–207.
53. G. J. Exarhos, P. J. Miller and W. M. Risen, J. Chem. Phys., 1974, 60, 4145–4155.
54. B. N. Nelson and G. J. Exarhos, J. Chem. Phys., 1979, 71, 2739–2747.
55. G. J. Exarhos, Vibrational Studies of Glass Structure and Localized Interactions, in Structure and Bonding in Non-Crystalline Solids, ed. G. E. Walrafen and A. G. Revesz, Plenum, New York, 1986, pp. 203–217.
56. A. Matic and L. Borjesson, Philos. Mag. B, 1998, 77, 357–362.
57. R. K. Brow, D. R. Tallant, S. T. Myers and C. C. Phifer, J. Non-Cryst. Solids, 1995, 191, 45–55.
58. R. K. Brow, J. Non-Cryst. Solids, 2000, 263 & 264, 1–28.
59. A. Rulmont, R. Cahay, M. Liegeois-Duyckaerts and P. Tarte, Eur. J. Solid State Inorg. Chem., 1991, 28, 207–219.
60. L. L. Velli, C. P. E. Varsamis, E. I. Kamitsos, D. Moncke and D. Ehrt, Phys. Chem. Glasses, 2005, 46, 178–181.
61. S. D. Ross, Inorganic Infrared and Raman Spectra, Mc Graw Hill, London, 1972, ch. 7 and 10.
62. G. L. Bottger and A. L. Geddes, J. Chem. Phys., 1967, 46, 3000–3004.
63. G. Burns, F. H. Dacol and M. W. Shafer, Phys. Rev. B: Solid State, 1977, 16, 1416–1425.
64. C. Weiping and Z. Lide, J. Phys.: Condens. Matter, 1997, 9, 7257–7267; R.-C. Wang, Y.-S. Gao and S.-J. Chen, Nanotechnology, 2009, 20, 375605.
65. K. Wakasa and M. Yamaki, J. Mater. Sci.: Mater. Med., 1992, 3, 141–144.
66. E. Steger and B. Käβner, Spectrochim. Acta, Part A, 1968, 24, 447–456.
67. M. Mroczkowska, J. L. Nowinski, G. Z. Zukowska, A. Mroczkowska, J. E. Garbarczyk, M. Wasiucionek and S. Gierlotka, J. Power Sources, 2007, 173, 729–733.
68. A. Schiraldi, E. Pezzati and P. Baldini, Phys. Chem. Glasses, 1986, 27, 190–193.
69. V. Clement, D. Ravaine and C. Deportes, Solid State Ionics, 1988, 28–30, 1572–1578.
70. P. R. Ehrmann, K. Carlson, J. H. Campbell, C. A. Click and R. K. Brow, J. Non-Cryst. Solids, 2004, 349, 105–114.
71. H. Ebendorff-Heidepriem, W. Seeber and D. Ehrt, J. Non-Cryst. Solids, 1993, 163, 74–80.
72. A. M. Efimov and V. G. Pogareva, J. Non-Cryst. Solids, 2000, 275, 189–198.
73. J. C. Arriagada, W. Burckhardt and A. Feltz, J. Non-Cryst. Solids, 1987, 91, 375–385.
74. Y. Abe and D. E. Clark, J. Mater. Sci. Lett., 1990, 9, 244–245.
75. P. Mustarelli, C. Tomasi, A. Magistris and S. Scotti, J. Non-Cryst. Solids, 1999, 163, 97–103.
76. L. Koudelka, A. Racicky, P. Mosner, I. Rösslerova, L. Montagne and B. Revel, J. Mater. Sci., 2014, 49, 6967–6974.
77. M. Scagliotti, M. Villa and G. Chiodelli, J. Non-Cryst. Solids, 1987, 93, 350–360.
78. R. K. Brow, D. R. Tallant, J. J. Hudgens, S. W. Martin and A. D. Irwin, J. Non-Cryst. Solids, 1994, 177, 221–228.
79. M. Tatsumisago, Y. Kowada and T. Minami, Phys. Chem. Glasses, 1988, 29, 63–66.
80. M. Dussauze, E. I. Kamitsos, P. Johansson, A. Matic, C. P. E. Varsamis, D. Cavagnat, P. Vinatier and Y. Hamon, J. Phys. Chem. C, 2013, 117, 7202–7213.
81. R. K. Brow, R. J. Kirkpatrick and G. L. Turner, J. Non-Cryst. Solids, 1990, 116, 39–45.
82. R. K. Brow, R. J. Kirkpatrick and G. L. Turner, J. Am. Ceram. Soc., 1990, 73, 2293–2300.

---

Footnotes

† Present address: Foundation for Research and Technology-Hellas (FORTH), Institute of Electronic Structure and Laser (IESL), P.O. Box 1385, 71110, Heraklion, Crete, Greece.

‡ Present address: Piraeus University of Applied Sciences (TEI of Piraeus), Department of Electrical Engineering, 250 Thivon & P. Ralli Str. 12244 Egaleo-Athens, Greece.

---