Introduction

For the glass industry, data of a large variety of glass properties are required to guarantee an error-free production process and the applicability within the field of interest. For the production of reinforcement fibres, such as those for polymer matrix composites, the knowledge of the liquidus temperature T Liq and the temperature T 3, which is attributed to a melt viscosity of 103 dPa s are quite essential, because at this temperature, the glass melt is drawn to fibres [1,2,3]. In an industrial melt, crystallisation before and during the drawing needs to be avoided, because it affects the quality of the fibre or may even disrupt the drawing process. T Liq is defined as the highest temperature at which crystals are thermodynamically stable in contact to the glass melt and hence both phases are in equilibrium [1, 4, 5]. Below this temperature, crystals might be precipitated, and in any case will not dissolve [1, 6]. Hence, T Liq is a key data point for an effective fabrication of the glass products, in case of temperature fluctuations below T Liq which might induce devitrification. Since T Liq delivers a clear indication of the minimum possible melting and forming temperature, the energy costs can then be estimated as well [4].

The melting and fining conditions in glass furnaces are strongly affected by the viscosity η [7, 8]. The processing temperature (denoted as T 3), is commonly recognised as fibre-forming temperature [2]. Also the temperatures throughout the glass cooling schedule are predetermined by the viscosity [9].

The difference between T 3 and T Liq is hereby of special importance: ΔT (=T 3 − T Liq) is often used as a first approximation of the tendency towards crystallisation during glass fibre drawing. Commercial experiences in the reinforcement fibre industry show that the forming temperature T 3 should be at least 50 K higher than T Liq. If ΔT is chosen too low, the production might be disrupted due to devitrification, if chosen too high, the energy costs are raised unnecessarily, which is of course unprofitable for the commercial glass production [10].

The glass-forming ability of a melt is closely related to the viscosity at T liq—the liquidus viscosity (η liq) of the system: a low T liq generally correlates with a high viscosity at the liquidus temperature, resulting in a large kinetic barrier to the atomic rearrangements required for crystallisation [11, 12]. Good glass-forming compositions are thus found close to eutectic compositions [11]. It is therefore important to have accurate knowledge of phase equilibrium diagrams for the glass-forming systems of commercial interest.

Particularly in multicomponent systems, small composition changes can cause dramatic changes in the liquidus temperature [4, 6]. Veit et al. [5] reported that compositional changes within the CaO–MgO–Al2O3–SiO2—glass melt even smaller than 3 wt% may result in shifts of T Liq by more than 20 K, causing a huge effect on ΔT as well.

For the present work, glass compositions in the quaternary system CaO–MgO–Al2O3–SiO2 have been prepared and characterised. The region of interest was chosen within a range attributed to the lowest liquidus temperatures according to Figs. 3.318 and 3.319 of Ref. [13]. At these compositions, T Liq is smaller than 1300 °C and hence comparatively low. These glass compositions are favourable because relatively low melting temperatures were the goal of this study which would enable an energy and environmentally advantageous production [14].

The invariant composition with an T Liq of 1130 °C has, according to Koch et al., a wt% composition of 22.0 CaO, 3.9 MgO 11.6 Al2O3, 62.5 SiO2 (which is in mol% 23.88 CaO, 5.89 MgO, 6.92 Al2O3, 63.31 SiO2). Since T Liq might vary strongly if the composition is changed slightly, a composition range next to this invariant point was chosen: 23.9–28.6 mol% CaO, 2.7–5.0 mol% MgO, 7.6–9.6 mol% Al2O3 and 59–63.4 mol% SiO2. This is highly advantageous, because commercially produced glasses occasionally show slight compositional changes and thus choosing a composition at the eutectic point might save heating energy, but it simultaneously increases the risks of disturbances and interruption during the production process: if the variation of T Liq is too high in case of small composition changes and exceeds the drawing temperature, crystallisation might occur. Therefore, not only the chosen composition itself is of interest, but also the properties in the composition range around it. This is why it was of interest for this study, whether small compositional changes have a stronger influence on T 3 or T Liq, especially regarding the overall impact on ΔT. Also the effect of the mixed alkaline earth (CaO and MgO) on the viscosity should be determined.

Altogether 28 glasses near the eutectic composition with the lowest liquidus temperature of 1130 °C have been melted and characterised [15]. The exact chemical composition as well as the liquidus temperature was determined and the temperature at a viscosity of 103 dPa s was derived from the viscosity curves for each glass composition. Furthermore, these properties were compared with those of glasses without MgO from Ref. [16].

Experimental procedure

The chemical compositions of the 28 glasses are summarised in Table 1. The used raw materials were pure quartz of grain size C, CaCO3 and Al(OH)3 and 4MgCO3·Mg(OH)2·5 H2O (all Merck AG). Each oxide concentration does not fluctuate by more than ±3 mol% from that of the centre glass mol% composition 61.2 SiO2, 26.3 CaO, 8.6 Al2O3 and 3.8 MgO.

Table 1 Chemical compositions according to XRF-values and the nominal composition of glasses in mol%
Full size table

The glasses 26, 27 and 28 had identical compositions as the glasses 9, 15 and 3, respectively, to check the reproducibility. The glass melt of 300 g was prepared from pure raw materials and then mixed by manual shaking.

A portion of the weighed raw materials was filled into a platinum crucible and heated up in a furnace with MoSi2 heating elements until the major part of carbon dioxide was evaporated. Then, the next portion of the raw materials was added and the temperature was increased to 1590 °C and held for another 2 h.

To achieve comparatively rapid cooling, the melt was cast in 10 l of cold water and subsequently dried at 120 °C to improve the glass homogeneity. Afterwards, the glass was melted at a temperature of 1600 °C again, kept for 3 h, and finally cast on heat resistant steel (AL7) moulds spray coated with boron nitride and preheated to 600 °C. The crucible with the residual, adherent glass was reheated and cast into the second mould. After casting, the glass was transferred to a muffle furnace preheated to a temperature in the range from 775 to 805 °C, depending on the glass composition, slightly above the glass transition temperature, T g. This cooling procedure was supplied to all glasses, and hence a direct comparison of densities and T g is facilitated although the cooling rate might have a slight effect on the glass properties [17].

The compositions of all samples were determined using X-ray fluorescence (XRF) with a PANalytical MagixPro (4 kW Rh Tube, Kassel, Germany).

The glass transition temperatures T g were determined via dilatometry in the temperature range from 200 to 900 °C using a heating rate of 5 K min−1 (dilatometer DIL 402 PC, NETZSCH Gerätebau GmbH, Germany). Cylindrical samples with length of around 20 mm and a diameter of 8 mm were drilled out of the first cast glass bulk.

The viscosity was determined via rotation-viscometry (Bähr VIS 403). The data were analysed by fitting the data points with polynomials using the software Origin 8.0 (OriginLab, Northampton, MA, USA) in order to obtain T 3. In this test, the torque required to maintain a preset rotational frequency of a cylindrical spindle inserted in a platinum crucible (containing exactly 13.8 cm3 of glass) was measured as a function of the temperature. Two runs at 250 and 10 rotations per minute were performed for each composition. From this torque, the corresponding viscosity was calculated.

The viscometer was calibrated using a DGG standard glass and was also tested with Advantex®, a registered trademark of Owens Corning which was used under license. The temperature accuracy of the viscometer was given with ±0.1 K. The measured T 3 of Advantex® was between 1170 and 1177 °C in those tests and corresponds very well with the theoretical value of 1173 °C. In some more experiments with Advantex®, it was found that a change in volume of up to 0.5 cm3 did not affect the T 3—determination by more than 1 K, indicating this systematic error is minimal. Since the investigated glasses are slowly crystallising glasses, a shift of viscosities due to small crystals within the melt can be ruled out. Overall, a maximum error of ±6 K is expected.

For the determination of T Liq (following the ASTM standard procedure C829-81 [18]), crushed glass with a grain size smaller than 850 µm was filled into a platinum boat 250 mm in length and 10 mm in width. The platinum boat was then put into an Orton tubular furnace with 8 built-in thermo-couples for temperature surveillance and remelted at 1500 °C for 1 h to reduce gas bubbles which might affect the interpretation of results. Afterwards, the temperature was reduced to the range of interest and held for 24 h to induce crystallisation within the glass melt. Subsequently, the position of the crystals was studied using an optical microscope. The position in the boat and its corresponding annealing temperature was determined, where the crystals are found—going from high to low temperatures and is then attributed to the temperature which corresponds to T Liq.

To ascertain the correctness and precision of the gradient furnace, a Standard Reference Glass (NIST 1416) from the National Institute of Glasses and Technology for the calibration of gradient furnaces was tested. This glass has a theoretical T Liq of 1147 ± 4 °C. The experimental results of 1145 and 1142 °C correlate well with the theoretical value, demonstrating the right temperature regime within the used gradient furnace. Then, around 50 trials with glasses of a similar composition to the tested glasses and a theoretical T Liq of 1220 °C were performed in the gradient furnace to test the repeatability of the measurement method. The values varied between 1214 and 1223 °C with a mean value of 1218.7 °C and a standard deviation of 2.7 K for this glass trial. Under the assumption that the values of the results are normally distributed, the measurement uncertainty is ±7 K at 3 SIGMA.

Results and discussion

Chemical composition, XRF

In Table 1, the nominal and the glass compositions determined by X-ray fluorescence (XRF) are summarised. The XRF-values added up to at least 99.74 wt% (indicating a very low amount of tramps) and have been normalised to 100 mol% for better comparison. In the following, the normalised XRF—data will be used for further correlations with the thermal properties.

In all samples, a shift of 0.4–1.1 mol% of the calcium oxide values from the intended CaO concentration was noticed. The higher CaO concentration mostly caused a decrease in the SiO2 and MgO concentrations, whereas Al2O3 showed always a similar deviation of around 0.1 mol%.

The glasses 26, 27 and 28 had the same chemical compositions as the glasses 9, 15 and 3, respectively, to check the reproducibility. The glasses 9 and 26 were very similar and are thus comparable with each other. But glass set 27 and 15 and glass duo 3 and 28 show shifts within the MgO and SiO2 concentrations of 0.8 and 0.2 mol%, respectively. Hence, the comparability is lower.

In Fig. 1, the analytically measured compositions (XRF) are illustrated in a 3D quaternary diagram, which is displayed from different angles for visualisation and better understanding.

Figure 1
figure 1

Temperature T 3 at a viscosity of 103 dPa s as a function of the composition (mol%) in a 3D phase diagram, with view from side (a), front (b) and top (c)

Full size image

Glass transition temperatures T g

The glass transition temperatures, T g, of the glasses are listed in Table 2. The transition temperatures of these 28 glasses vary between 754 and 783 ± 3 °C. It is well known that the relaxation times of glasses are only a few minutes at T g [19, 20]. Also, the residual stresses in the glasses are gradually eliminated with increasing annealing time or temperature. Since all T g values are lower than 800 °C, the right cooling temperature of the cooling furnace has been chosen and the glasses are relaxed.

Table 2 Measured glass transition temperatures (T g), processing temperatures (T 3)—and liquidus temperatures (T liq) of the studied glasses
Full size table

The T g values of the glasses 9 and 26 with almost identical compositions are 762 and 758 °C, respectively. Due to the limitations of the T g-calculation with the tangent method in the applied analysis program, an error of ±3 K has to be taken into account. Hence, the results can be treated as identical values. Also the samples with identical compositions, 27 (765 °C) and 28 (771 °C), show only small deviations of 3 and 2 K from the corresponding original glasses 15 (762 °C) and 3 (773 °C), respectively. This can easily be explained by the analytical error of the tangent method when using the evaluation software.

In Fig. 3, the obtained T g values are illustrated as function of the number of bridging oxygen per tetrahedron.

All glasses are peralkaline compositions, i.e. they have a ratio (MgO + CaO)/Al2O3 larger than 1. Hence, all tested samples contain network modifier concentrations which are high enough to compensate the negative charge of the [AlO4]− tetrahedra, and therefore aluminium oxide is built into the network as a network former, i.e. in fourfold coordination. When all [AlO4]− tetrahedra are charge compensated by the network modifiers Ca2+-ions or Mg2+-ions, any excess Ca2+ and Mg2+ form non-bridging oxygen (NBO) sites and hence decrease the number of Si–O–Si bridges and therefore the number of bridging oxygen per tetrahedron (BO/T) [21]. This means, the higher the concentration of network modifier is, the larger is the number of non-bridging oxygen sites and the lower is the number of bridging oxygen per tetrahedron (BO/T) and the more pronounced is the weakening of the glass structure. Obviously higher concentrations of network modifiers decrease the transition temperature. Also in the comparison with glasses from Ref. [16], where glasses with similar compositions but without MgO were characterised, the slope is clearly visible. It is interesting to notice that the distribution of the T g—values in the present paper is obviously broader than in the glasses which do not contain MgO.

The temperature T 3 at which the viscosity is 103 dPa s

In Table 2, the results for T 3 are summarised. It is obvious, that the comparatively small changes in the chemical composition are able to cause a huge variation of the temperature at which the viscosity is 103 dPa s and thus also of the working temperatures. This is a possible threat for ensuring a smooth process. The data show variations between 1237 °C (glass 19) to 1327 °C (glass 22), i.e. by 90 K, even though glasses with only small changes in the composition were studied. All viscosity curves had similar slopes showing that all glasses in this composition range have similar (temperature dependent) activation energies.

In Fig. 1, the T 3-temperatures of the glasses are presented as a function of the determined XRF-composition in a ternary phase diagram for better visualisation.

The colour code for the temperatures varies from the lowest value of T 3 for glass 19 with 1237 °C (violet) and the highest glass fibre-forming temperature at a viscosity of 103 dPa s for composition 22 with 1327 °C (dark red). This colouration is a guide for the eye.

Influence of the components

From the top view in Fig. 1, the effect of the calcium oxide concentration on the viscosity of the glass melt is apparent. The higher the CaO concentration in the glass melt is, the lower is the temperature, attributed to a melt viscosity of 103 dPa s: the highest T 3-temperature, that of glass 22 (dark red-bottom of picture) with 1327 °C had a CaO concentration of only 23.5 mol%, whereas the glass 7 (violet—top of picture) with the lowest T 3 had a CaO concentration of 27.9 mol%. Moreover, the colouring between those two extrema is following the rainbow colouring of the legend with rising CaO concentration.

Furthermore, T 3 increases with higher SiO2 concentrations. This is also clearly visible in Fig. 2. An increasing Al2O3 concentration does not seem to affect T 3 significantly. Those results are correlating with Ref. [16] where ternary CaO–Al2O3–SiO2 glasses are described, i.e. without MgO and have been described in the literature as well [17].

Figure 2
figure 2

T 3 as function of oxide concentrations

Full size image

As shown in Fig. 3, there is a clear correlation between the BO/T and the T 3 (and also T g). Inserting more network modifiers such as CaO and MgO into the glass structure leads to an increase in non-bridging oxygens bonds within the network. Hence, the glass with the lowest concentration of the network formers Al2O3 and SiO2 (glass 7) has the lowest BO/T and thus the lowest network connectivity. This explains the decrease in T 3 as well: the lower the network connectivity, the smaller is the rigidity, i.e. the resistance to movement of the clusters in the melt and in consequence the lower is the viscosity. Therefore, the temperature at the viscosity of 103 dPa s, which is of interest in this study, is lower for glasses with high concentration of alkaline earths.

Figure 3
figure 3

T 3 and T g as a function of the number of bridging oxygen per tetrahedron

Full size image

This is proved by the comparison of the 5 glass duos 15 and 27, 17 and 22, 2 and 14, 4 and 14, and 6 and 13, with almost identical Al2O3 and CaO concentrations: T 3 is significantly decreasing if SiO2 is substituted by MgO.

Furthermore, if comparing glasses 4 and 12 with almost identical MgO and Al2O3 concentrations, it is visible, that the substitution of SiO2 by CaO results in decreasing temperatures at a viscosity of 103 dPa s, just as expected. Other examples which show similar behaviour are the glass duos 16 & 8 and 11 & 13. Additionally, T 3 values from Ref. [16] without any MgO were compared to similar glass compositions from this glass series. In Table 3 and Fig. 4, the results are displayed and show a linear trend. If 5 mol% of SiO2 are exchanged by MgO, T 3 decreases by 60–80 K. A substitution of 3.2 mol% MgO by SiO2—hence, more network modifiers are in the system—increases T 3 by 66 K.

Table 3 Change in the temperature T 3 due to substitution of SiO2 by MgO
Full size table
Figure 4
figure 4

Shift of T 3 as function of the of substitution of SiO2 by MgO

Full size image

However, when interpreting Fig. 1—the front view—and also Fig. 2d the influence of the MgO concentration on the T 3 does not show a clear trend: it does not seem to be affected by increasing MgO concentrations. This is because the temperature of the CMAS-glasses depends more strongly on the CaO- than on the MgO concentration. The glasses 7, 8, 16 and 22, which have similar MgO concentrations of around 2.5 mol% showed increasing T 3-temperatures which are all between 1238 and 1327 °C (see Fig. 1 Front view/Table 2). This can easily be correlated with the increase in the CaO concentration (see Fig. 1, top view), which increases from 23.5 to 27.9 mol%. The direct comparison of different glass sets with similar concentrations of silica and alumina (1 and 7 and 12) shows that the substitution of CaO by MgO increases T 3. This is not observed in the glass sets 12 & 23 and 5 & 25; however, the temperature shifts of only 6 and 9 K, respectively, are within the error of the method and therefore not significant.

For the set 1, 7 and 12, an interesting effect was noticed if one network modifier is exchanged by another. For the glass duo 7 and 1, with increasing MgO concentration from 2.5 to 3.2 mol%, the temperature increases by 31 K, whereas from glass 7–12, even though another 1.4 mol% are added, the temperature changes only by 4 K. Also for the glass set 12 and 23, where the MgO—concentration is increased from 4.0 to 4.6 mol%, the measured temperature change is only 6 K, and for glass set 25 and 5, where the MgO concentration is increased from 3.6 to 5.4 mol%, the temperature change is only 9 K. Although in both sets, the temperature change is contradicting to the trend, it can still be concluded that the effect of the alkaline earths gets less pronounced if their concentrations are higher. It has already been described by, e.g. Vogel that the viscosity changes are less significantly if more alkaline network modifiers are added. In the glass compositions studied in this paper, this effect is also observed, if there is already a network modifier incorporated into the network (CaO) and a second one is added (MgO).

It was expected that the incorporation of MgO as substitute of the larger CaO would have a decreasing stimulus on the viscosity. Instead, as mentioned before, the T 3 values increase, the more CaO is exchanged even though the number of bridging oxygen is identical for each set. This is due to the more covalent character of the Mg–O bonds in comparison with the more ionic Ca–O bonds.

The substitution of one network former atom against the other (Al2O3 vs SiO2) does not have any impact on T 3: the glass sets 13 & 25 and 23 & 28, with similar CaO and MgO concentrations (as network modifiers) show both temperature shifts of only 3 and 9 K, which can be considered as identical within the error margin of the analysis. The substitution of 1 mol% SiO2 against 1 mol% Al2O3 slightly increases the number of bridging oxygen within the network, because 2 aluminium tetrahedra are incorporated if only one silicon tetrahedron is taken out, but this obviously does not noticeably affect T 3.

Liquidus temperature T liq

The liquidus temperatures determined via a tubular gradient furnace are listed in Table 2. With 20 tests of the same glass composition, the error distribution of the gradient furnace was determined to be 7 K. Liquidus temperatures of 1165, 1164 and 1169 °C were measured for the similar compositions 9, 19 and 26, respectively. In analogy, the glasses 3 and 28, also very comparable in composition, had values of 1138 and 1142 °C.

Within the investigated composition region, the temperatures vary between 1138 °C for glass 3 and a maximum value of 1192 °C for composition 7. All determined values were lower than expected in comparison with the liquidus temperatures given by the figures of Ref. [13], where only a temperature lower than 1300 °C was reported. Even though a liquidus temperature of 1130 °C for a composition next to the investigated composition range was reported by Koch et al. [15], higher values were expected because of the fact that not the eutectic but only a neighbouring range was examined.

The glasses with magnesium oxide concentration have a lower T Liq range than the glasses from Ref. [16], which had liquidus temperatures varying between 1157 and 1214 °C.

ΔT: the difference between T 3 and T Liq

Obviously, the composition changes of 6 mol% SiO2, 4.3 mol% CaO, 2.0 mol% Al2O3, and 4.0 mol% MgO had a greater impact on the forming temperature at the viscosity of 103 dPa s than on the crystallisation of the glasses. In comparison with T 3, the T Liq—data changed only by 53 K, whereas the temperature at a viscosity of 103 dPa s showed a difference of 90 K between the lowest and the highest measured values. Close to the eutectic range, liquidus temperatures usually drop significantly even with small composition changes. Hence, it was expected that T Liq would show a more pronounced variation, whereas the viscosity and T 3 should not be influenced as strongly with those small variations in the compositions. From this observation, it can be concluded that the viscosity of the glasses is the decisive factor for ΔT.

In Fig. 5, the difference ΔT between the processing (T 3) and the liquidus temperature (T Liq) as a function of composition is displayed. It is visible that the difference between the two temperatures T 3 and T Liq varies between 65 and 160 K. Thus, all glasses would meet the requirement of ΔT = 50 K and be suitable for the glass fibre-forming processes. To avoid the risk of interruptions due to crystallisations while forming the glass melt, a temperature difference of ≥70 K is more advantageous, which then excludes glass 18 (dark red) with a ΔT of 65 K.

Figure 5
figure 5

ΔT between T 3 and T Liq as a function of the composition (mol%) in a 3D phase diagram, with view from side (a), front (b) and top (c)

Full size image

But high values of ΔT, which guarantee process stability, are not the only parameter to consider. More energy is needed to produce the final product if the processing temperature of the glass melt is increased. Hence, a ΔT value, which is too high, is not favourable either. Nevertheless, all compositions with a ΔT higher than 70 K could probably be adjusted easily. With minor additions of alkali oxides such as lithium oxide, sodium oxide or potassium oxide, T 3 could be lowered without changing T liq considerably [22]. However, glass compositions with a ΔT higher than 130 K should be excluded, because then T 3 is too high and the energy costs increase. Minor additions of alkali oxides may not decrease T 3 sufficiently to get a suitable ΔT. For the glasses of this study, T 3 was the decisive factor for ΔT and is increased above 1300 °C, even though T Liq is lower than 1160 °C. Hence, all glass composition with low calcium oxide composition would be a bad choice, because of their high working temperatures at a viscosity of 103 dPa s which leads to the favourable high ΔT but increase the energy costs ineffectively because the production requirements are already met with ΔT = 50 K, which means they do not need to be significantly higher.

The more calcium oxide the compositions contain, the smaller is ΔT. This effect is achieved, because T 3 decreases to temperatures of around 1240 °C, whereas the liquidus temperature is increased to at least 1160 °C (glass 7) and up to 1192 °C (glass 18).

Small composition changes generally cannot be ruled out in the commercial glass production. This is why a composition range with the smallest risk to failure needs to be chosen. The best processing range for the glass melts should be within the green zone in Fig. 5 Top view, because firstly the liquidus temperatures do not change significantly (around 15 K) and are secondly also the lowest of the whole studied composition range. The working temperature is between 1270 and 1280 °C, leading to a ΔT of 110–120 K.

As mentioned above, the working temperature can be decreased with small additions of alkali oxides. Addition of 1 wt% Li2O already decreased the working temperature of a glass with a similar composition by around 50 K [22].

In summary, even though the glasses 7 and 16 might have lower ΔT temperatures but the working temperature of the glasses 23, 25 and 28 can be adjusted to the same region as glasses 7 and 16. Hence, the energy needed to draw those fibres would be similar, but the difference to the crystallisation temperature would be higher because of the lower liquidus temperatures of the glasses 23, 25 and 28 and therefore more advantageous for a smooth production process.

Since the magnesium concentration within the glass compositions does not have any noticeable impact on ΔT (see front view in Fig. 5), it can be concluded, that for the choosing of the composition, the decisive factor is the calcium concentration, followed by the silica-concentration (top view in Fig. 5). The MgO concentration might therefore be the component to be changed easily—maybe in exchange for the alkali oxides. It is concluded that the composition of the glasses 23, 25 or 28 would be the most suitable choice as reinforcement glass fibre composition: they have sufficiently high ΔT and T 3 is comparatively small or can be adjusted.

Obviously, the production of commercial reinforcement glass fibres does not use laboratory raw materials but rather technical raw materials which contain tramp oxides, such as Fe2O3, TiO2 or alkali oxides. It can be expected, that they also affect T 3 and T Liq but it is not possible to predict reliable values for the influence of these tramps. Of course, the inclusion of network modifiers such as Fe2O3 or alkalies (which are also often tramps in technical raw materials) would definitely decrease T 3 for same the reasons explained earlier.

But the extent of the decrease cannot be predicted without exactly knowing in which raw material the tramps are included. If for instance the raw material clay is used for the glass melting which includes at least trace quantities of alkalies, the reduction of T 3 would be much higher, since the alkali oxides are then basically substituted for the network formers SiO2 and Al2O3. But if the alkalies are coming into the melt through the raw material lime (which is a technical raw material for CaO) the effect is probably not as pronounced.

It is assumed that the crystallisation is inhibited the more components a glass melt contains, but it is not clear, whether the liquidus temperature is affected, or just the crystallisation rate. For all these reasons, it is not possible to estimate the variation of ΔT when technical raw materials are used instead of pure raw materials.

Conclusion

The design of new technologically important glasses requires a thorough knowledge on the effect of composition on key properties of the product. Altogether, 28 glasses with compositions between 23.9 and 28.6 mol% CaO, 2.7 and 5.0 mol% MgO, 7.6 and 9.6 mol% Al2O3, and 59 and 63.4 mol% SiO2 were melted from pure raw materials. The composition of each glass sample was determined via X-ray fluorescence to ensure a correct comparison of the compositions and their influence on the thermal property values. The glass transition temperature, T g, the forming temperature at a viscosity of 103 dPa s T 3 and the liquidus temperature T Liq of the glasses were measured. T g ranged from 754 to 783 ± 3 °C; T 3 varied between 1237 and 1327 ± 6 °C and T liq between 1138 and 1192 °C ±7 K. It was found that the small composition changes had a larger effect on the viscosity than on the crystallisation temperature.

The network modifier calcium oxide had the major effect on viscosity changes, overlapping the influence of magnesium oxide. Substitution of SiO2 by Al2O3 did not result in significant changes in viscosity. All glass compositions met the safety requirements of a minimum ΔT = 50 K and are therefore suitable as reinforcement fibres although to ensure energy saving processes neither T Liq nor ΔT should not be too high.