Matrix-controlled channel diffusion of sodium in amorphous silica

Abstract

To find the origin of the diffusion channels observed in sodium-silicate glasses, we have performed classical molecular dynamics simulations of Na2O-4SiO2 during which the mass of the Si and O atoms has been multiplied by a tuning coefficient. We observe that the channels disappear and that the diffusive motion of the sodium atoms vanishes if this coefficient is larger than a threshold value. Above this threshold the vibrational states of the matrix are not compatible with those of the sodium ions. We interpret hence the decrease of the diffusion by the absence of resonance conditions. PACS numbers: 61.43. Fs, 61.43.Bn, 66.30.Hs, 63.50.+x The mechanism of ionic transport in amorphous materials [1,2] at the atomic scale is not completely elucidated so far and therefore it is the subject of numerous experimental [3,4] and numerical [5][6][7][8][9] studies. Among the most studied systems are the sodium-silicate glasses since they contain the essential ingredients, namely the amorphous matrix (silica) and the mobile ions (sodium), as a first step in the simulation of more complex glasses of higher practical interest.In previous studies [10-12], we have shown by means of classical molecular dynamics simulations of Na 2 O-4SiO 2 (NS4), that the sodium atoms diffuse through a well connected network of pockets (which represents only a limited fraction of the entire available space) that we have called "channels" to be coherent with the literature. The existence of the channels, which are not due to microsegregation effects [6,7], has been confirmed by the existence of a pre-peak in the partial Na-Na structure factor at a wave-vector q = 0.95Å −1 [13]. This pre-peak has also been observed experimentally [14] and numerically [15] in another study. We have also shown that the location of the channels is strongly correlated to the positions of the non-bridging oxygens [13] and Horbach et al. have shown that the sodium dynamics should be related to that of the underlying silica network [15]. This suggests that the origin of the channels could be related to the dynamical properties of the matrix. To check this idea we present in this letter classical molecular dynamics simulations on a series of "toy" systems in which the atomic masses of both the oxygen and silicon atoms have been systematically changed after artificially multiplying their experimental values by a common factor µ varying from 0.5 to infinity, the usual NS4 system being recovered for µ = 1.By studying both the mean square displacement of the sodium atoms and the characteristics of the channels, we find a change in the sodium diffusion properties when the parameter µ is increased above a value of about 30. Above this threshold, the sodium diffusion decreases and the channels can no more be clearly defined. Guided by the concomitant change in the short time characteristics of the velocity autocorrelation function, we have calculated the vibrational density of states (VDOS) for both the sodium atoms and the atoms of the matrix for various values of µ. We obse...

Formation of an Alkali Diffusion Barrier and Stain‐Resistant Glass Surface

“…The fact that a small concentration of boron, for example 1.5 atomic percent diffused into pre‐existing glass networks, is adequate to stop alkali diffusion suggests that a relatively small number of sites, or channels, contribute to alkali diffusion as reported in previous publications, alkali ions have preferential coordination sites involving nonbridging oxygens, and their frequency of motion depends on the population of these sites . As shown in this work by IRRS and XPS investigations, boron diffusion into the surface significantly reduces nonbridging oxygen concentration in the vicinity of the surface.…”

Sodium diffusion in an artificially frozen silica glass

“…The interactions between the particles are given by a modified version of the so-called 'BKS' potential [19,20] which is able to reproduce the structure as well as the dynamics of several sodium-silicate systems [10][11][12][13]21] (for more details see [11]). In this study we start from two standard NS4 samples previously generated and relaxed during 1.4 ns (10 6 steps) [17] at 8 different temperatures: 4000, 3050, 2500, 2325, 1900, 1675, 1025 and 0 K. We have shown in previous studies [10][11][12][13] that the properties of these samples are in good agreement with the available experimental data. Then, after these 1.4 ns of relaxation time, we have frozen all the movements of the silicon and oxygen atoms (including of course the vibrations) at these 8 'freezing' temperatures T fr .…”

“…Then, after these 1.4 ns of relaxation time, we have frozen all the movements of the silicon and oxygen atoms (including of course the vibrations) at these 8 'freezing' temperatures T fr . In our previous work [17] we have shown that this implies that the sodium diffusion is impossible in these samples. In order to permit the sodium diffusion we have performed three additional simulations in which we have given a supplemental thermal energy to the sodium atoms in order to reach the diffusive regime after an additional 1.4 ns.…”

“…This pre-peak has also been observed experimentally [14] and numerically [15,16] in other studies. Since the sodium dynamics appeared to be related to that of the underlying silica network [15] we have recently performed classical molecular dynamics simulations of a series of 'toy' systems in which the atomic masses of both the oxygen and silicon atoms have been systematically changed after artificially multiplying their experimental values by a common factor l varying from 0.5 to infinity [17]. Contrarily to what is observed in a standard NS4 system (l = 1), these simulations have shown that in a frozen silica matrix (l = 1) the diffusion of the sodium atoms is impossible.…”

A revised view on the mixed-alkali effect in alkali borate glasses

“…The simulations by Hornbach et al [39] and Sunyer et al [40] show that the location of channels strongly correlates to the position of negative network units. The authors suggest that the channels could be related to the dynamic properties of the matrix [41] and thus a matrix-controlled diffusion mechanism is anticipated [12]. In our approach the correlation between ionic jumps and network relaxation are not taken into account, however, the correlations between jumps of neighbouring alkali ions are of great importance (see below).…”