Raman Spectroscopy and Statistical Analysis of the Silicate Species and Group Connectivity in Cesium Silicate Glass Forming System
The Raman spectra of %Cs2O-(100 − x)%SiO2 (, 22, 27, 33, and 37 mol%) glasses and melts were measured in the temperature range of 293 to 1553 K. The concentrations of the species were calculated as a function of the composition and temperature based on the deconvolution analysis of the spectra. It was found that a dynamic equilibrium among structural units in the melts with mol% can be described by disproportionation reaction . The enthalpy of this reaction was found to be equal to 32 ± 6, 43 ± 8, 56 ± 10, and 52 ± 9 for , 27, 33, and 37 mol%, respectively. The nonideal entropy of mixing (Δ) depends on the melt temperature and increases almost linearly with increasing temperature. The , , and distributions with ranging from 0 to 55 mol% were modeled using experimental data for the concentrations of the units.
Knowledge of the structure of amorphous materials is fundamental for understanding of their physical and chemical properties. Structure-properties correlations of the silicate glasses and melts are of major importance in geochemical and technological fields. Therefore, an adequate understanding of the composition and temperature dependencies of glass and melt properties requires detailed information on their structure. Numerous studies of silicate glasses have demonstrated that silicon-oxygen tetrahedra with various numbers of bridging oxygen atoms are the fundamental structural blocks of these glasses. The silicon-oxygen tetrahedra are called units, where –4 is number of bridging oxygen atoms per SiO4 tetrahedron. Schematic 2-dimensional representation of the species is shown in Figure 1(a). units form a continuous random network of glasses or silicate anions in melts via the Si-O-Si bridging linkages. The random network of SiO2 glass consists of SiO4 tetrahedra interconnected via their oxygen apexes, where all oxygen atoms are bridging ( units). The addition of alkali oxides to SiO2 leads to the breaking of the Si-O-Si linkages and the formation of terminal oxygen atoms, each of which belongs only to one silicon atom, that is, the formation of species with in the glass structure. Schematic representation of the formation of species with nonbridging oxygen atoms () is shown in Figure 1(b). MAS-NMR studies have shown that the concentrations of various types of units depend on the concentration of the modifier oxide and the type of alkali cation [1, 2]. Namely, the equilibriumis shifted to the right with increasing cationic power of the metal cation (, where is the valency and is the ionic radius, Li+ > Na+ > K+). The Raman spectroscopy studies of rubidium and cesium silicate glasses [3, 4] allow the assertion that the above-mentioned regularity is valid for all alkali cations.
As known, equilibrium (1) () is shifted to the right with increasing temperature in the case of the sodium and potassium silicate melts [5–12]. However, no significant changes in the local structure of lithium silicate melts were found [5, 12]. Moreover, thermodynamic calculations of the distribution in the Li2O-SiO2 system have shown that equilibrium (1) is shifted to the left with increasing temperature, that is, towards an increase in the concentration of units . To the best of our knowledge, no data on the distribution and its changes with temperature in cesium silicate melts can be found in the literature. Therefore, the aim of this paper is a structural study of Cs2O-SiO2 glasses and melts by high-temperature Raman spectroscopy.
High-temperature Raman spectroscopy is a useful method for investigating not only the glass structure but also the melt structure over a wide temperature range [5–15]. As is known, the high-frequency region (800–1200 cm−1) of the spectra of alkali silicate glasses and melts is characterized by a series of Raman bands originating from the stretching vibrations of various units. It is assumed that the intensities of these bands are proportionate to the concentrations of species [3, 5, 7, 8, 10–15]. Therefore, these bands can be used for the quantitative description of the structural changes of silicate glasses and melts depending on both the composition and temperature.
Although knowledge of the concentrations of species is important for describing the structure of silicate glasses and melts, it does not provide a clear idea of their structure, because the distribution provides no information about the interconnection between units. Currently, only a limited number of publications address this problem, and all of them focus on the Na2O-SiO2 system [6, 16–20]. Therefore, this paper studies this problem for Cs2O-SiO2 glasses and melts.
2. Material and Methods
2.1. Sample Preparation
Glass samples of the compositions %Cs2O-(100 − x)%SiO2 (, 22, 27, 33, and 37 mol%) were prepared by the conventional melting and quenching method from reagent-grade SiO2 and Cs2CO3. The initial reagents were dried (120°C, for 2 h), weighed, and mixed in the required ratios. The batch (5 g) was melted in a platinum crucible at –1100°C for 2–5 h (depending on the melt composition) to obtain a bubble-free melt. The obtained melt was cooled directly in a platinum crucible in air at room temperature and then used for Raman measurements at various temperatures. To avoid glass hydration, Raman scattering measurements were started immediately after the sample preparation. It should be noted that due to the high viscosity of the 17%Cs2O-83%SiO2 melt, we failed to prepare a bubble-free glassy sample. Thus this sample was only studied in the glassy state.
2.2. Raman Experiments
Raman scattering measurements were performed on a specially designed high-temperature apparatus, based on a DFS-24 double monochromator. To record the Raman spectra at different temperatures, the small platinum crucible was placed into a vertical compact electrical furnace. Thus, all spectra were recorded in 180° geometry. The operating temperature range was 20–1300°C and was controlled within ±1°C. The second-harmonic of an LTI-701 solid-state pulsed laser (nm, mW) operated at a modulation frequency of 8.7 kHz was used as the excitation source. The pulse duration of the acoustooptic switch was 2 μs. An uncooled FEU-79 photomultiplier was used to detect the Raman signal. A gated photon counting system was applied to minimize the thermal radiation signal. The spectral width of the slit was 6 cm−1 in all measurements. A detailed description of the experimental setup and recording conditions of the high-temperature Raman spectra can be found in [21–23].
To compare the spectra obtained at various temperatures, they were reduced to obtain the temperature- and frequency-dependent scattering intensity:where and are the observed and reduced Raman intensities, respectively, and are the Raman shift and wavenumber of the excitation source, respectively, and , , , and represent Planck’s constant, Boltzmann’s constant, speed of light, and temperature, respectively.
2.3. Simulation of the , , and Distributions
As it is known, Zachariasen’s rules  for glass formation are focused only on a local configuration of cation-oxygen polyhedra and their connectivity to each other (via corners, not edges or faces). Based on these rules it is possible to assume that the most important point in a modeling of a local structure of silicate glasses is a coordination number of glass-forming cations rather than a network dimensionality and topology as a whole. Any regular or random network where each node has four linkages with the nearest nodes and each bond joins only two nodes reproduces completely the distribution in pure SiO2 (all structural units are ). Formally, such network can be expressed in form of the table consisting of five lines and columns, when is a number of nodes in the system. Top line contains the serial number of node and other four lines contain the serial numbers of nodes which joined to the given in the top line. Thus, each column describes linkages between five nodes. Each node in this table was interpreted as a silicon atom and had four linkages with the nearest nodes. The bond between two nodes was interpreted as a Si-O-Si (hereafter we will use the abbreviation ) bridging bond (some examples of bridging bonds are shown in Figure 1(c)). To calculate the distribution, it is necessary to break the preassigned number of bonds (the number of the bonds that must be broken is determined by the glass composition), and then the obtained configuration must be analyzed. In this approach, each node with unbroken bonds and each broken bond were interpreted as a unit and a Si-O−+M M+−O-Si nonbridging bond, respectively. Three parameters, , , and , were introduced into the modeling procedure to achieve the best agreement between the experimental concentrations and the calculated distribution. These parameters do not depend on the glass composition but do depend on temperature, and in addition to the concentration of various types of nodes, they determine the probability of the linkage break (Figure 2 demonstrates a role of these parameters in the computational algorithm). Taking into account the random character of the linkage choice and the table of linkages, 50 configurations were generated for each composition at a fixed temperature. The required concentrations of units were obtained by averaging of all 50 configurations of the table. The table was also analyzed to study the (bridging bonds) and (group connectivity) distributions based on the distribution data (e.g., schematic 2-dimensional representations of the and groups are shown in Figure 1(c)). Here, is the number of bridging oxygen atoms, and , , , and indicate the type of connected units. The modeling was performed in a composition range of 0 to 55 mol% Cs2O content at two fixed temperatures (293 and 1223 K). A more detailed description of the modeling procedure can be found in [25, 26].
3.1. Raman Spectra
3.1.1. Glass Spectra
The Raman spectra of glasses with the composition %Cs2O-(100 − x)%SiO2 (, 22, 27, 33, and 37 mol%) are shown in Figure 3, in which the symbolic designations near the spectra (left side) indicate the Cs2O content. The Raman spectra of glasses with relatively low Cs2O contents (, 22, and 27 mol%) exhibit two bands with peak intensities at 510 and 598 cm−1, one weak wide band at approximately 785 cm−1, and a line at 1100 cm−1, having the highest intensity in each spectrum, with a shoulder at 1150 cm−1. The peak position at 510 cm−1 is gradually shifted toward higher frequencies and the 598 cm−1 line is slightly shifted toward lower frequencies with increasing Cs2O content in the given composition range. In addition, the intensity of these bands increases, whereas the intensity of the weak line at 785 cm−1 decreases with the addition of cesium oxide. For the high-frequency envelope, the shape of this contour is most strongly affected by an increase in the modifier oxide content, which causes a gradual resolution deterioration of the 1100 and 1150 cm−1 bands and a drastic decrease in intensity for the high-frequency shoulder accompanied by an increase in the intensity of the 1100 cm−1 Raman band at mol%. The 1100 cm−1 line becomes more symmetric, and its width decreases. A new band at approximately 930 cm−1 distinguishes the 33Cs and 37Cs spectra from the mol% spectra. The intensity of this band increases with increasing . It also should be noted that the intensity of the 510–530 cm−1 band is greater than that of the 598 cm−1 band at mol%, and at mol%.
3.1.2. High-Temperature Raman Spectra
The Raman spectra of the 22%Cs2O-78%SiO2 and 27%Cs2O-73%SiO2 samples measured at different temperatures are shown in Figures 4(a) and 4(b). As seen in Figure 4(a), the peak intensity at 530 cm−1 decreases slightly and its position is gradually shifted toward lower frequencies with increasing temperature. The peak intensity at 598 cm−1, in contrast to that of 530 cm−1 band increases with increasing temperature and remains at essentially the same frequency. No significant changes are observed in the peak intensity or the shape of the high-frequency envelope (1000–1200 cm−1) at relatively low temperatures (up to ~898–1003 K). Only one symmetric wide line with maximum near 1100 cm−1 is observed in the high-frequency region at higher temperatures, and the intensity of this line decreases with increasing temperature. A new weak band at 930 cm−1 that appeared in the Raman spectra measured at temperatures above 1003 K is another peculiarity of the high-temperature spectra. All of the above-mentioned features of the changes in the 22Cs spectra with temperature are observed in the 27Cs spectra as well.
The Raman spectra of glasses and melts with modifier oxide contents of 33 and 37 mol% are shown in Figures 5(a) and 5(b), respectively. In contrast to the previous spectra, the line near 920–930 cm−1 is observed at all temperatures, and its intensity increases with temperature. This line is most clearly observed in the 37Cs spectra. The peak intensity at 1100 cm−1 obviously decreases with increasing temperature, and its position shifts slightly toward lower frequencies. In addition, the width of this band increases. The intensities of the low-frequency bands (520–530 and 598 cm−1) depend weakly on temperature, but their width significantly increases with temperature. Finally, the formation of another new Raman line in the melt spectra with a maximum near 420 cm−1 should be noted (see Figure 5(b)). This line is observed in the previous spectra only as an unresolved low-frequency shoulder near the 530 cm−1 band.
3.2. Modeling Results
The results of the modeling of the and distributions in the Cs2O-SiO2 system at two different temperatures (293 K, solid lines; 1223 K, dotted lines) are shown in Figures 6(a) and 6(b), respectively. In addition to our experimental data the literature data [1, 3, 27] on the concentrations of units in the Cs2O-SiO2 glasses and melts are also shown in Figure 6(a). As seen in this figure, the modeled curves well describe the experimental data for both glasses and melts over a wide compositions range. Temperature changes have no significant effect on the concentration of units in glasses and melts with relatively low Cs2O contents (up to ~15 mol%) as well as in a narrow composition interval near 40 mol%. In turn, most considerable changes in the distribution are observed for samples with disilicate and metasilicate compositions. Here, the concentrations of the dominant type of structural units ( and for disilicate and metasilicate, resp.) decrease with increasing temperature and are accompanied by an increase in the concentration of other structural units with differing by ±1. In the aggregate, this phenomenon causes the curves describing the dependences of the and units on the composition to be slightly wider for the melts than for the glasses, but the positions of their maxima do not change with temperature.
As seen in Figure 6(b), only bridging bonds with or 1 may be observed in the system among all possibilities (, , , , , , , , , ) of bridging bonds. The maxima of the curves decrease with increasing and are approximately the same for the bonds with the same values. The maxima of the and curves depend weakly on temperature (approximately 1%), whereas the positions of these maxima are shifted toward higher Cs2O concentrations. In addition, an increase in the width of these curves is also observed. The width and position of the peak of the and curves behave similarly, but their maxima decrease more dramatically with increasing temperature (approximately 20%). The curves with are shown in the inset to Figure 6(b). As observed, the dependence of the concentration of the given bridges on temperature is opposite to that typical for bridging bonds with . In this case, a significant increase in the and concentrations with increasing temperature is observed. Nevertheless, the concentration of these Si-O-Si bonds is low in both glass and melt and is less than 1% and 4% at 293 and 1223 K, respectively.
The modeling results of the concentrations of the and groups as a function of composition at different temperatures (293 and 1223 K) are shown in Figures 7(a)–7(d). The concentrations were calculated in two ways: relative to the concentration of the appropriate type of structural unit (Figures 7(a) and 7(c) for and , resp.),and relative to the total concentration of units (Figures 7(b) and 7(d) for and , resp.),In these equations, and are the amounts of different types of groups and structural units, respectively.
As seen in Figure 7(a), the gradual increase in concentration of the modifier oxide leads to the following transformations of the groupings: . In addition, , , and groups are also formed in the glass structure. As seen in the inset to Figure 7(a), however, their concentrations are lower than 7%. Both the coexistence regions and the variety of the groups are higher in melts than in glasses. The , , and groups supplement the list of groups at high temperature (1223 K). Their maximum concentration exceeds 7% in the melt structure and is low in the glasses. The , , , , and groups may also appear in the melt structure but their concentration is less than 7% (see Figure 7(a)). The shape of the , , and curves depends slightly on temperature, whereas the , , and curves are subject to dramatic changes. An increase in the width of the curves and a shift in the position of their maxima toward large are typical for all of them. The ratio between the maxima of these curves at 293 K isand that at 1223 K is
The region of existence of the groups is somewhat broader than that of and extends to mol% (see Figure 7(c)). The transformation of groups, depending on , can be represented by the sequence . In addition, some amount of groups (less than 3% of the total amount of units) and groups (less than 4%) can exist in the glass structure with disilicate and metasilicate compositions, respectively (see the inset in Figure 7(c)). The region of coexistence of the groups increases by ~2.5% and reaches 55% as the temperature increases up to 1223 K. It is also accompanied by an increase in the width of the curves, a shift of the maxima of these curves toward large values, and the leveling of their maxima. The concentration of the dominant type of structural group (at a given composition) decreases and the fraction of the groups that are untypical for glass increases with increasing temperature. As before, no significant changes are observed in the distribution at mol%.
The and distributions calculated according to (3) and (4) are shown in Figures 7(a)–7(d). Figures 8(a) and 8(c) represent the and distributions relative to the and contents, respectively, and Figures 8(b) and 8(d) represent the concentrations of the and groups relative to the total amount of units, respectively. As seen in Figure 8(a), an increase in concentration in the modifier oxide leads to the transformation of the groups in the following sequence: . It should be noted that the concentration of the and groups relative to units in the glass structure is low at mol% (less than 1.5%; see Figure 8(b)), although their fraction relative to the total amount of groups exceeds 40%. Moreover, the groups are only formed in the glass structure in an amount exceeding 1% at mol%, as follows from Figure 8(b). As before, an increase in the width of the curves and a shift in the position of their maxima toward large values are observed with increasing temperature. The maxima change such that the difference between maxima decreases with increasing temperature.
The distribution has the simplest form, which is evidently related to the low variety of such groups. (A change in from 4 to 1 gives only four types of groups.) The concentration of the groups is negligible in the glass structure. Therefore, a consequence of the transformations of the groups looks similar to . The data presented in Figures 8(c) and 8(d) support the assumption that the changes in the curves will be similar to those described above.
4.1. Raman Spectra and Structure of the %Cs2O-(100 − x)SiO2 (x = 17, 22, 27, 33, and 37 mol%) Glasses
It is rational to divide the overall frequency range into low-frequency (400–700 cm−1) and high-frequency (800–1200 cm−1) intervals to analyze the obtained Raman spectra. The low-frequency interval is related to the stretching and some of the bending vibrations of Si-O-Si linkages. Two narrow lines (490 cm−1 (D1) and 602 cm−1 (D2)) along with a broad intense line (~450 cm−1) are observed in the Raman spectrum of g-SiO2 (e.g., ). It is accepted that the D1 and D2 lines can be related to the symmetric oxygen breathing vibration of three- (D2) and four-membered (D1) siloxane rings, consisting of SiO4 tetrahedra [28–32]. The investigation of the Raman spectra of alkali silicate glasses with high SiO2 content  has shown that the bands near 490 and 602 cm−1 are gradually shifted toward higher and lower frequencies, respectively, with increasing Cs2O content. Thus, it can be assumed that the Raman bands near 510–530 and 598 cm−1 in our spectra have the same origin as the D1 and D2 lines, respectively. The change in the intensity of these lines as a function of glass composition (see Figure 3) shows that the increase in modifier oxide concentrations leads to changes in the statistical distribution of -membered rings, wherein the concentration of three-membered rings gradually increases. This observation is in accordance with the results published in , where an increase in concentrations of three-membered rings with increasing Cs2O was shown based on NMR data. The concentration of the four-membered rings changes weakly at 17% < < 22% and decreases at higher Cs2O content.
The Raman bands originating from the symmetric stretching vibration of the Si-O− terminal groups of various () units are located in the high-frequency range of spectra of alkali silicate glasses :(i)the band at 1050–1100 cm−1 is due to the symmetric stretching vibrations of the terminal oxygen atoms of tetrahedra with one nonbridging oxygen (NBO) atom, that is, units;(ii)the band at 920–950 cm−1 results from the Si-O− stretching of tetrahedra with two NBO ( units);(iii)the Raman band near 900 cm−1 is attributed to the stretching vibration of the units ( tetrahedra with three NBO);(iv)finally, the line at 850 cm−1 is related to the symmetric stretching mode of anions.
As seen in Figure 3, only the 1050–1100 and 930 cm−1 bands are observed in the Raman spectra of studied glasses. It should be noted that the 1050–1100 cm−1 band exists in all spectra, whereas the 930 cm−1 band is only observed in the spectra of glasses with relatively high Cs2O contents (33Cs and 37Cs). In addition, a high-frequency shoulder with maximum at approximately 1150 cm−1 is observed in the Raman spectra of the 17Cs, 22Cs, and 27Cs samples. Although alkali silicate glasses have been studied for a long time, a review of the literature revealed that the origin of this shoulder is still controversial. In one series of publications [12, 13, 35], the 1150 cm−1 line was attributed to the Si-O stretching vibration in fully polymerized structural species, that is, the vibrations of units. However, based on a study of the Raman spectra of alkali silicate glasses with various compositions Matson et al.  have suggested that this line may be assigned to the vibrations of the units, which are structurally and vibrationally distinguished from those of the units, producing the 1050–1100 cm−1 band. They argued that the 1150 cm−1 shoulder has significantly greater intensity than could reasonably be assigned to residual g-SiO2 spectral features. In addition, they found no correlation between the intensity of this band and other bands (e.g., 450 cm−1) characteristic of the g-SiO2 spectrum. Based on these conclusions, the 1150 cm−1 shoulder was attributed to units, which have slightly stronger (shorter) Si-O− bond than the one producing the 1100 cm−1 line .
Matson et al.’s assumption concerning the origin of the 1150 cm−1 shoulder was confirmed later on by You with co-authors . The correlation between the Raman shift and connecting topology of adjacent units was found based on the quantum chemical calculation of the characteristic frequencies of species. In other words, it was demonstrated that the Raman shift of the symmetric stretching vibration of units decreases as the number of bridging oxygen atoms of the nearest-neighbor species adjacent to the given unit decreases. For example, the Raman shift of the group is higher than that of the group. In our opinion, the conclusions in  are strong evidence of Matson’s assumption. Thus, we will rely on Matson’s interpretation of the origin of 1150 cm−1 band in our paper.
The qualitative examination of the Raman spectra of Cs2O-SiO2 glasses (Figure 3) confirms that the structure of glasses with a Cs2O content below 33 mol% consists of and units (the existence of units is obvious and requires no evidence, although the 1150 cm−1 shoulder indirectly proves the presence of such structural units) and that the species are present, at the least, in the form of and groups. The Raman band at 930 cm−1 shows that units are formed in the 33Cs and 37Cs glasses. The presence of units in the structure of disilicate glasses is a result of satisfying the charge balance (NBO/Si = 1). Regarding the 37Cs glass, the presence of species can be identified only by the quantitative analysis of the corresponding spectrum.
The high-frequency envelope (800–1300 cm−1) of the registered Raman spectra was simulated as a superposition of the Gaussian lines to estimate the concentrations. The number of Gaussian lines was sufficient to reproduce the original spectra with a correlation factor of ≥0.98. The interpretation of the Raman bands described above was also taken into account. In addition, some results published in  were also taken into account: species with equal give more than one band and peak position of individual band depends on the structural position of units. For example, wavenumbers of NBO symmetric stretching vibration of species are located in the range of 1050 to 1150 cm−1, whereas units give a set of the individual peaks in the range of 930 to 1050 cm−1 . Several examples of the deconvolution results of the Raman signal of the studied samples in the high-frequency region are shown in Figure 9. Four Gaussian lines were sufficient to reproduce the low-temperature (293 K) Raman spectra of the 17Cs and 22Cs glasses, whereas five lines were needed to simulate the 27Cs, 33Cs, and 37Cs spectra. The H3 and H4 bands were attributed to the species. Because the 17Cs and 22Cs glasses consist of and units, it is possible to assume that only four types of structural groups (, , , and ) can exist in structure of these glasses. Considering the dependence of the Raman shift of groups on the , , and indexes established in , it was assumed that the and groups are the main contributors to the intensity of the H4 band and that the vibrations of the and groups are the main contributors to the intensity of the H3 band. This qualitatively agrees with the simulation results of the distribution represented in Figures 7(c) and 7(d). The H2 band is most likely due to the stretching vibrations of the Si-O-Si linkages [3, 31]. The origin of the H1 line is unclear. It is possible that this line is a result of the assumption of the Gaussian shape of the elementary bands in the spectra of glasses with relatively low Cs2O concentrations (17Cs and 22Cs); that is, this line is an error in the choice of the type of elementary bands. The relative area of the H1 band is the same for the 17Cs and 22Cs spectra (0.02) and its intensity increases with further increases in the Cs2O content. The H2 line behaves similarly. An increase in intensity of both H1 and H2 lines begins from the appearance of a new H5 line in the deconvolution of the Raman spectra. The H5 line indicates the formation of species in the structure of the samples. According to , it is possible to assume that the vibrations of the groups, connected with one or two units, for example, and groups, also contribute to the intensity of the H2 band at higher concentrations of the modifier oxide. Thus, the 1060 cm−1 line was designated as in the deconvolution of the 27Cs, 33Cs, and 37Cs spectra. In turn, the line can be attributed to the vibrations of the , , and groups according to the distribution represented in Figures 8(a) and 8(b). Finally, the H5 line was ascribed to groups.
The localized nature of the silicon-oxygen stretching motions of silicate units, containing SiO4 tetrahedra with one, two, three, or four nonbridging oxygen atoms [34, 36] allows us to use the relative integral intensities of the Gaussian components to calculate the concentrations.
If three types (, , ) of species coexist in a structure simultaneously, then their concentrations (, , ) can be obtained from the following system of equations:The coefficient proportionality, , was chosen to achieve a best accordance with data published in other papers [3, 4]. Furthermore, if there is reason to believe that units are absent in the glass structure (as in the 17Cs and 22Cs glasses), then the final equation does not make sense and the  and  concentrations can be calculated analytically from the first two equations without any experimental data. Considering the complicated nature of the and bands, two scenarios were calculated. In the first variant, the and values were equal to the areas of the and components, respectively. The integral intensity of the and bands was reduced on the and values in the second scenario. Here, and are the average values of the integral intensities of the H1 and H2 bands, respectively, measured from the deconvolution results of the high-frequency range of the Raman spectra of low-alkali glasses (17Cs and 22Cs). The peak positions, relative areas of the partial bands, and the  concentrations calculated according to system (4) are summarized in Table 1. The peak positions and FWHM values were established within ±5 cm−1. As seen in the table, the first calculation variant yields slightly higher  and  concentrations and a somewhat lower  value. The calculation results of the second scenario yields the opposite trend. Accounting for the and values produces higher  values and somewhat lower concentrations of and units. The greatest difference between the calculation results is observed for the 27Cs glass and is 3% for the  concentration.
4.2. High-Temperature Raman Spectra and Structure of the Cs2O-SiO2 Glasses and Melts
The Raman spectra of the 22Cs sample measured in the temperature range of 293 to 1553 K are shown in Figure 4(a). As seen from this figure, the change in temperature results in changes in the spectra in both low- and high-frequency ranges. According to the above-mentioned structural interpretation of the Raman bands, the significantly greater intensity of the 598 cm−1 band and significantly lower intensity of the 530 cm−1 band in the melt spectra in comparison with the glass spectra indicate a considerable influence of temperature on the distribution of -membered rings. These data support the assumption that the fraction of 4-membered rings decreases and fraction of 3-membered rings increases with increasing temperature. In turn, the changes in the shape of the high-frequency envelope and the appearance of a weak Raman signal at 930 cm−1 in the melt spectra (this band is absent in the glass spectrum) point to a structural transformation in the local structure of the sample. It can be argued at a qualitative level that the list of structural units for glasses and melts will differ. The local structure of the glassy sample includes only two structural units, and , whereas that of melts contains significant amounts of units (930 cm−1 line). The same conclusions may be drawn from the 27Cs spectra (Figure 4(b)). Changes in the ratio and the gradual increase in the intensity of the 920–930 cm−1 band also occur. Such obvious changes in the low-frequency range are not observed in the Raman spectra of samples with higher Cs2O contents (see Figures 5(a) and 5(b)). In this case, it is difficult to derive well-defined conclusions about dependence of the distribution of the -membered rings on temperature. At the same time, an increase in the intensity of the 920–930 cm−1 band and a decrease in the intensity of the 1090–1100 cm−1 band are observed with increasing temperature, as before. Thus, an increase in temperature leads to a decrease in the concentration of the units and an increase in the fraction of the species in all studied samples.
The high-frequency range of the Raman spectra measured at different temperatures was simulated as a superposition of the Gaussian lines to study the influence of temperature on the concentrations of species (see Figure 9). The parameters of the partial bands obtained from the modeling of glass spectra were used in the deconvolution of the spectra measured at different temperatures. Thus, the band designation and origin correspond to those accepted in the previous section. It was found that the low-temperature spectra of the 22Cs samples are well reproduced by the same set of partial bands as the glass spectra. However, the low-temperature set of partial bands is insufficient for modeling of the high-temperature spectra, and a new H5 component appears in deconvolution of these spectra. One more H6 line appears in the modeling of the spectra of the sample with the highest Cs2O content (37Cs). Both H5 and H6 bands were assigned to the units. The H6 line is more likely due to groups according to Figures 8(a) and 8(b).
The dependences calculated according to system (7) are summarized in Table 1 (an additional item, , was added to the denominator of the last equation of system (7) in the calculation of the local structure of the 37Cs sample). According to the obtained data, the local structure of the studied glasses does not change under a moderate increase in temperature. Further increases in temperature lead to a decrease in the concentration of species and an increase in concentrations of and units. These changes can be explained by the shift of the equilibrium,to the right with increasing temperature.
The temperature of the beginning of the shift of equilibrium (8) to the right depends on the sample composition and most likely corresponds to the glass-transition () temperature. The dynamic equilibrium (8) is “frozen” at temperatures below .
The  data can be used to determine the enthalpy of the reaction (8). The equilibrium constant of the disproportional reaction (7) expressed using the concentrations of the units is defined asIn turn, the enthalpy of equilibrium (8) is calculated from the Van’t Hoff equation:
Assuming that is independent of temperature above , it is possible to calculate the enthalpy values using the slope of the versus () line from the high-temperature experimental data. The data are shown in Figure 10. Thus, the values for 22Cs, 27Cs, 33Cs, and 37Cs are obtained as 32 ± 6, 43 ± 8, 56 ± 10, and 52 ± 9 kJ/mol, respectively. These results show that value depends on the melt composition and is highest at 33 mol% Cs2O. A similar trend has been observed for the sodium silicate system . However, one should be advised and understand that there are a number of other reasons for decreasing of with increasing SiO2 content: choice of the individual bands to modeling of poorly resolved high-frequency spectral envelope, Gaussian shape of individual peaks, an increase in experimental error at determination of the integral intensity of the weak bands ascribed to the units, and so forth. Thus, we can assert unambiguously that is constant for the melts with close to 33 mol% (). Based on this conclusion one can see that there is a quite clear tendency for increase in with increasing alkali cation radius: is approximately equal to 0 [7, 37], 20 [11, 22, 37, 38], 30 [10, 39], and 50 kJ/mol (this work) for lithium, sodium, potassium, and cesium silicate melts, respectively.
Maehara et al.  have shown that  data can be used to calculate the nonideal entropy of mixing () for the silicate glasses and melts:where , is the Avogadro constant, and is Boltzmann’s constant. As follows from Figure 6(a), the change in temperature does not significantly change the in glasses and melts with high SiO2 contents (mol%). A similar situation would be typical for glasses with lower SiO2 contents, but only at relatively low temperatures (less than ). As seen in Table 1, the local structure of the 22Cs, 27Cs, 33Cs, and 37Cs samples significantly changes at higher temperatures. Hence, considerable changes in values are expected in this case. The values as a function of temperature for the above-mentioned samples calculated by (11) are shown in Figure 11. As one can see, the entropy increases almost linearly with increasing temperature in the studied temperature range for all samples. The entropy change depends on the melt composition, the entropy increasing with modifier oxide content up to 33 mol% and then beginning to decrease.
The structure of the Cs2O-(100 − x)SiO2 glasses and melts was studied by high-temperature Raman spectroscopy. It was found that the concentration of species gradually decreases with increasing modifier oxide content. In turn, the fraction of units increases, reaches a maximum at mol%, and then starts to decrease. The species are observed in the glass structure at mol%. Their concentration increases with increasing Cs2O content. The concentrations of and units are higher in the melt structure than in the corresponding glasses. The increase in the concentration of these structural units is explained by the shift of equilibrium (8) to the right with increasing temperature. The enthalpy of equilibrium (8) depends on the melt composition and was found to be equal to 32 ± 6, 43 ± 8, 56 ± 10, and 52 ± 9 kJ/mol for 22Cs, 27Cs, 33Cs, and 37Cs, respectively. The nonideal entropy of mixing, , depends on the melt composition and increases linearly with increasing temperature at . The value also depends on the melt composition, increasing with the Cs2O content up to 33 mol% and then beginning to decrease.
The  experimental data were used to model the distribution in Cs2O-SiO2 glasses and melts. The developed approach allows us to describe the experimental data over a wide composition range for both glasses and melts. The configurations of the random linkages generated during the modeling were analyzed for the identification of and distributions. The results support the assumption that temperature changes weakly influence the and distributions at relatively low Cs2O contents (less than mol%). At higher Cs2O contents, bridges with are most sensitive to temperature. The direction of the change (increase/decrease) in concentration of the bridging bonds between one-type structural units depends on the glass (melt) composition, except for bridges, the concentration which always increases with increasing temperature at mol%. As for the groups, it was found that increasing temperature widens the variety of coexisting groups in the melt. The greatest change in the distribution of and groups is expected in melts with mol%, whereas the and distributions are more prone to changes in the melts with mol%.
Conflict of Interests
The authors declare that there is no conflict of interests regarding the publication of this paper.
Partial support by the RFBR (Project no. 14-08-00323_a) is gratefully acknowledged.
V. N. Bykov, A. A. Osipov, and V. N. Anfilogov, “Structural study of rubidium and caesium silicate glasses by Raman spectroscopy,” Physics and Chemistry of Glasses, vol. 41, no. 1, pp. 10–11, 2000.View at: Google Scholar
B. O. Mysen and J. D. Frantz, “Raman spectroscopy of silicate melts at magmatic temperatures: Na2O-SiO2, K2O-SiO2 and Li2O-SiO2 binary composition in the temperature range 25-1475 C,” Chemical Geology, vol. 96, no. 3-4, pp. 321–332, 1992.View at: Google Scholar
B. O. Mysen and J. D. Frantz, “Silicate melts at magmatic temperatures: in-situ structure determination to 1651°C and effect of temperature and bulk composition on the mixing behavior of structural units,” Contributions to Mineralogy and Petrology, vol. 117, no. 1, pp. 1–14, 1994.View at: Publisher Site | Google Scholar
L. Olivier, X. Yuan, A. N. Cormack, and C. Jäger, “Combined 29Si double quantum NMR and MD simulation studies of network connectivities of binary Na2O·SiO2 glasses: new prospects and problems,” Journal of Non-Crystalline Solids, vol. 293–295, no. 1, pp. 53–66, 2001.View at: Publisher Site | Google Scholar
V. N. Bykov, A. A. Osipov, and V. N. Anfilogov, “High-temperature device for registration of Raman spectra of melts,” Rasplavy, no. 4, pp. 28–31, 1997 (Russian).View at: Google Scholar
V. N. Anfilogov, V. N. Bykov, and A. A. Osipov, Silicate Melts, Nauka, Moscow, Russia, 2005.
A. A. Osipov and L. M. Osipova, “Structure of lithium borate glasses and melts: investigation by high temperature Raman spectroscopy,” Physics and Chemistry of Glasses: European Journal of Glass Science and Technology Part B, vol. 50, no. 6, pp. 343–354, 2009.View at: Google Scholar
A. A. Osipov and L. M. Osipova, “ distribution in silicates: alkali silicate glasses and melts,” Advanced Materials Research, vol. 560-561, pp. 254–258, 2012.View at: Google Scholar
W. J. Malfait, W. E. Halter, Y. Morizet, B. H. Meier, and R. Verel, “Structural control on bulk melt properties: single and double quantum 29Si NMR spectroscopy on alkali-silicate glasses,” Geochimica et Cosmochimica Acta, vol. 71, no. 24, pp. 6002–6018, 2007.View at: Publisher Site | Google Scholar
S. K. Sharma, T. F. Cooney, Z. Wang, and S. van der Laan, “Raman band assignments of silicate and germanate glasses using high-pressure and high-temperature spectral data,” Journal of Raman Spectroscopy, vol. 28, no. 9, pp. 697–709, 1997.View at: Google Scholar
P. McMillan, “Structural studies of silicate glasses and melts—applications and limitations of Raman spectroscopy,” American Mineralogist, vol. 69, no. 7-8, pp. 622–644, 1984.View at: Google Scholar
V. N. Bykov, A. A. Osipov, and V. I. Anfilogov, “Raman spectroscopy of melts and glasses in Na2O-SiO2 system,” Rasplavy, no. 6, pp. 86–91, 1998 (Russian).View at: Google Scholar
V. N. Bykov, O. N. Koroleva, and A. A. Osipov, “Structure of K2O-SiO2 melts: Raman spectroscopic data and thermodynamic simulation results,” Rasplavy, no. 3, pp. 50–59, 2008 (Russian).View at: Google Scholar