Local structures and Al/Si ordering in lanthanum aluminosilicate glasses explored by advanced 27Al NMR experiments and molecular dynamics simulations

Abstract

The structures of 15 La-Al-Si-O glasses, whose compositions span 11-28 mol% La(2)O(3), 11-30 mol% Al(2)O(3), and 45-78 mol% SiO(2), are explored over both short and intermediate length-scales by using a combination of solid-state (27)Al magic-angle spinning (MAS) nuclear magnetic resonance (NMR) spectroscopy and molecular dynamics (MD) simulations. MAS NMR reveals Al speciations dominated by AlO(4) groups, with minor but significant fractions of AlO(5) (5-10%) and AlO(6) (≲3%) polyhedra present in all La(2)O(3)-Al(2)O(3)-SiO(2) glasses; the amounts of Al([5]) and Al([6]) coordinations increase for decreasing molar fraction of Si. The MD simulations reproduce this compositional trend, with the fractional populations of AlO(p) groups (p = 4, 5, 6) according well with the experimental results. The modeled La speciations mainly involve LaO(6) and LaO(7) polyhedra, giving a range of average La(3+) coordination numbers between 6.0 and 6.6; the latter increases slightly for decreasing Si content of the sample. Besides the expected bridging and non-bridging O species, minor contributions of oxygen triclusters (≤9%) and free O(2-) ions (≤4%) are observed in all MD data. The glass structures exhibit a pronounced Al/Si disorder; the MD simulations reveal essentially random SiO(4)-SiO(4), SiO(4)-AlO(p) and AlO(p)-AlO(q) (p, q = 4, 5, 6) associations, including significant amounts of AlO(4)-AlO(4) contacts, regardless of the n(Al)/n(Si) molar ratio of the glass. The strong violation of Al([4])-Al([4]) avoidance is verified by 2D (27)Al NMR experimentation that correlates double-quantum and single-quantum coherences, here applied for the first time to aluminosilicate glasses, and evidencing AlO(p)-AlO(q) connectivities dominated by AlO(4)-AlO(4) and AlO(4)-AlO(5) pairs. The potential bearings from distinct fictive temperatures of the experimental and modeled glass structures are discussed.

27Al NMR Studies of Aluminosilicate Glasses

“…Although the structural fragment of Fig. 18 stems from a MD-generated model, we observe a very good agreement between the predictions from MD simulations and NMR experiments in the Al and O speciations: all of features (1)-(3) are fully corroborated by experiments [74,76,89,104,116] (whereas the presence of O [0] and O [3] species in AS glasses still awaits direct experimental proof ).…”

“…At odds with the conventional structural model of AS glasses, the results suggested a pronounced network disorder (discussed further in Section 11), as reflected both in strong violations of Loewenstein's rule and in Al [4] -NBO contacts associated with Q Al n<4 tetrahedra. Subsequent advanced MAS NMR experimentation [69,76] utilizing through-space 27 Al- 27 Al and 17 O- 27 Al dipolar interactions has unambiguously confirmed these preliminary findings (Section 11).…”

“…The program incorporates restricted parameter ranges (as opposed to fixed parameters) in the numerical fitting: this is useful for precluding physically unreasonable results and ensuring consistency in the handling of a large series of AS glasses, for instance for monitoring dependencies of the NMR parameters on a glass-composition variable [76,89,116]. Surprisingly, among the free deconvolution software available to date, only EGdeconv appears to offer options for restricted parameter fitting.…”

“…S > 1=2 cross-polarization (CP) [82,176,177], where ν nut CT for the S-spin is selected first, as its choice has the overall strongest bearings on the magnetization-transfer efficiency. (ii) To establish optimum conditions for homonuclear recoupling of spin-1/2 nuclei [178,179] or halfinteger spins [76,[171][172][173][180][181][182], as well as for (iii) spin-1/2 applications at fast MAS (>35 kHz), e.g., for identifying CP conditions and locating suitable proton decoupling powers [178].…”

“…Following previous developments of efficient protocols for spin dynamics calculations [243,244] and orientational averaging/spectra interpolation [245][246][247], we employ a program developed in our laboratory for deconvoluting 1D MAS NMR spectra [76,89,116] (courtesy of K. Okhotnikov and B. Stevensson), which bears some similarities with the EGdeconv package [118]. The program incorporates restricted parameter ranges (as opposed to fixed parameters) in the numerical fitting: this is useful for precluding physically unreasonable results and ensuring consistency in the handling of a large series of AS glasses, for instance for monitoring dependencies of the NMR parameters on a glass-composition variable [76,89,116].…”