In this study, glasses along the join x K2O - (1-x) SiO2 were
studied with modifier fractions up to x=0.35. Up to x=25% the glasses were found to
be reasonably resistant to moisture if kept under dry conditions.|
Fig. 1: Wide-angle x-ray scattering patterns of several glasses along the join x K2O - (1-x) SiO2. The respective molar fractions of K2O are indicated.
The First Sharp Diffraction Peak (FSDP) moves towards larger momentum transfer values as the K2O content increases. At the same time the width of the peak appears to decrease. The position of the FSDP corresponds to the inverse of a characteristic length of the structure. In silicate glasses, the characteristic length is commonly interpreted as the correlation length between voids formed in association with the free electron pairs of oxygen. The correlation length varies from 400pm (5% sample) to 320pm (25% sample).
Fig. 2: Diffractogram of the x=4.99% sample. The two overlapping lines required to fit
the measured data are indicated by broken lines.
In the case of the 5% K2O sample, the line broadening trend is so predominant that no symmetric line shape (above theangle-dependent decaying background) such as a single Gaussian can fit the data. This suggests that the broadening is caused by the superposition of two FSDPs from two simultaneously present amorphous phases with different void correlation length.
Fig. 3: WAXS peak positions (top) and line widths (bottom) as a function of the K2O
fraction. The open symbols refer two the second component found in the x=4.99% sample.
Errors attached to the fits (Gaussian plus linear background) are smaller than the symbols.
The decrease of the average distance between voids with increasing modifier content contradicts the idea that the voids are stuffed by added modifier, as has been suggested on the basis of simulations. The reason for this apparent discrepancy is that the modifier ions (K+) come with oxygen counter-ions, which in turn increases the amount of voids and hence reduces the correlation length. However, the creation of non-bridging oxygens (NBOs) due to modifier addition causes the network to become less strained. Larger voids can relax to form smaller ones and the local order is increased - hence the FSDP narrows.
Fitting NMR spectra with several overlapping Gaussians reveals the fractions of Si atoms in
SiO4 tetrahedra with varying number of NBOs (Qn species with n brigding
Fig. 4: 29Si NMR spectra of the x K2O - (1-x) SiO2 glasses. The broken lines indicate the change in the chemical shift for the three Qn species when the K2O fraction increases.
In the 5% sample, only Q4 and Q3 species are present, while in the 35% sample, only Q2 and Q3 are found. In all intermediate glasses, all three species have a noticeable population. The chemical shift of each individual peak moves towards less negative values (i.e. deshielding) as the K2O fraction increases. All three species are displaced to an approximately equal extent, indicating that the modifier causes the overall network to relax rather than resulting only in local relaxation.
Fig. 5: Fractions of the different Qn species as function of the K2O fraction,
as obtained from multiple-Gaussian fits to the spectra. Full circles denote Q4, open
circles Q3, crosses Q2.
As the modifier content increases, the fraction of the Q3 component grows at the expense of Q4. There is, however, a discontinuity at 17% K2. For this glass, the Q3 component is considerably more predominant than in its neighbours. The Q3 fraction obtained from the analysis of NMR spectra is based on the assumption that all Q3 species are in similar environments, i.e. that a Gaussian distribution of electron densities around Si atoms in Q3 species exists.
Fig. 6: Dependence of the fraction of non-bridging oxygens on the K2O fraction.
The line represents values calculated on the basis of the overall chemical composition,
the circles correspond to the evaluation of the distribution of Qn species as found by NMR.
The NBO fraction can be computed independently from the Qn distribution and from the chemical composition of the sample (assuming one NBO per monovalent modifier). The high Q3 fraction in the 17% sample causes a discrepancy between the two values of the NBO fraction. This indicates that a single Q3 fit component is not accurate, and that the Q3 distribution is wider than a Gaussian line suggests. This points to the coexistence of two amorphous phases with different Qn distribution.