Unravelling the effect of nano-heterogeneity on the viscosity of silicate melts: Implications for glass manufacturing and volcanic eruptions

Highlights

• Materials characterization at the nanoscale is essential for reliable viscosity data.

• Nanocrystal formation increases viscosity and glass transition temperature of melts.

• The effect is observable in Fe₃O₄- and TiO₂-bearing nanocrystallized silicate melts.

• Raman features correlate with viscosity increase and nanostructural changes.

• A remarkable decoupling between calorimetric and viscometric results was identified.

Abstract

Traditionally depicted as homogeneous materials, manmade and volcanic glasses are often subjected to nanostructuration during demixing and crystallization of their parental melt. While the controlled formation of nanocrystals has been exploited for decades in the industrial production of glass-ceramics to obtain materials with superior properties, nano-heterogeneities have been recently recognized in volcanic products erupted explosively. Here, a multi-pronged approach is adopted to highlight how the emergence of nano-heterogeneities exerts a previously unexplored influence on the viscosity of multicomponent silicate melts. Calorimetry and viscometry provide evidence of how the formation of TiO₂- and FeO-bearing nanocrystals increases the effective viscosity of suspensions. Raman spectroscopy and transmission electron microscopy allow the correlation between the increase in viscosity and the degree of sample modification. This study has implications for glass and glass-ceramics manufacturing and for the development of better numerical models of volcanic eruptions, which are key for the operational assessment of volcanic risk.

Introduction

The glassy state is a nonequilibrium and hybrid state of matter that features properties of both solids and liquids [1], [2], [3], [4]. Glasses appear solid on a human timescale, but they inexorably relax toward the equilibrium and ultimately crystallize in the limit of infinitely long time [4]. Due to the almost unlimited applications that glasses have accomplished in our society, one can argue that the world has entered the “Glass age” [5]. In fact, the modern technological diversification of glass production includes, but is not restricted to, the well-established floating of flat glass [6] and extrusion of glass tubes [7], sintering of glass powders [7], drawing of glass fibers [8], and the newly prototyped glass 3D printing [9,10].

The melt viscosity plays a key role in all these manufacturing processes, especially during stages such as fining, molding and annealing. A comprehensive understanding of glass chemistry and physics underpins the development of better models to predict the viscosity of glass-forming melts over an extended range of temperature (T) and chemical composition (X). Thus, the research seeking to find a description of the viscosity of glass-forming melts as a function of T dates back to 1920s, when the seminal studies of Vogel [11], Fulcher [12] and Tammann et al. [13] were published.

Over the past 60 years [14,15], our understanding of melt viscosity has allowed us to control the process of crystallization [16] that plays a crucial role in fine-tuning the production of glass-ceramics. These materials are composed by a residual glassy phase hosting a variable volume fraction of functional crystals [17], which have a typical size varying from a few nanometers to several microns. Due to their peculiar nature, glass-ceramics are characterized by superior physical properties such as high mechanical strength and low thermal expansion; they are essential for optical, electronic, catalytical and medical applications [18]. Viscosity has been indeed recognized as a key parameter of glass ceramization protocols in the industry [19], [20], [21], [22]. On the other hand, understanding the influence of undesired demixing and crystallization on melt viscosity currently represents an important topic in the development of glasses for radioactive waste immobilization. The formation of spinel and the presence of insoluble residues during the melting process have been shown in fact to strongly affect the discharge efficiency of the high-throughput glass melters designed for these purposes [23,24].

The effect of so-called microlites (i.e. micrometric crystals) on the viscosity of melts is also of importance in Earth Sciences, and especially in the dynamics of magmas feeding volcanic eruptions. At depth, magma consists of a multicomponent aluminosilicate melt containing crystals and dissolved volatiles. The eruptive style is primarily controlled by the degree of phase separation between the carrying phase (melt + crystals) and the bubbles, which form due to the exsolution of volatiles from the melt. Effusive eruptions tend to have a low-viscosity carrying phase that allows the exsolved gas to flow through it and separate, leading to the emission of gas and lava. In contrast, explosive eruptions occur when there is little phase separation due to the high viscosity of the carrying phase; the exsolved gas cannot escape and so the growth in exsolved gas volume directly drives the rapid expansion needed to generate a volcanic explosion characterized by magma fragmentation [25,26]. At eruptive temperature, the viscosity of the magma can vary by several orders of magnitude [27] depending on both the continuous evolution of the melt composition (i.e. chemical effect) and the formation of microlites (i.e. physical effect) induced by the decompression [28] during its ascent.

Although nanocrystals represent a key component of glass-ceramics and have been increasingly recognized [29], [30], [31], [32], [33], [34], [35], [36], [37], [38] in volcanic and experimental products, the effect of their formation on the viscosity of initially homogeneous glass-forming melts has received little attention to date [[19], [20], [21], [22],39,40]. Here, we employ differential scanning calorimetry (DSC), viscometry, Raman spectroscopy and scanning transmission electron microscopy (STEM) to isolate and distinguish the compositional and textural effects of nanocrystallization on the viscosity of systems of interest for materials and Earth sciences. We show that subtle nanostructural changes greatly increase the overall viscosity of Ti-doped lithium magnesium aluminosilicate melts. The compositional similarity of this model system with several mass-produced glasses and glass-ceramics makes this study central to our understanding of the effect of demixing and nanocrystallization on the flow property of industrial melts. Afterwards, we demonstrate that our approach can be generalized and used to monitor and quantify the viscosity increase of volcanic melts due to the formation of iron-bearing nanocrystals. With this aim, we examine the product of the most common form of volcanism on Earth, such as basalt from Mt. Etna (Italy). Basalts are low-viscosity magmas and basaltic volcanoes are well-known to primarily manifest gentle effusion of lava. However, effusive eruptions can be suddenly interrupted by violent explosions releasing a powerful quantity of ash and gas in the atmosphere, which impact local populations [41], disrupt air traffic [42] and have the potential to drop the global temperature [43].

In view of suspension rheology concepts, one must note that all technical and natural processes shown here result in heterogeneous liquids. At its simplest, two phases would be present such as a single crystal type is suspended in a viscous matrix. However, if one considers multiple transformations at the nano-, meso‑, and microscale, including liquid demixing and gas exsolution, multi-phase suspensions/emulsions can be formed during these processes. To address the overall effect on shear viscosity, we refer to the relative viscosity η_rel [15]:

$$lo{g}_{10}{\eta }_{rel}=lo{g}_{10}\eta -lo{g}_{10}{\eta }_0=lo{g}_{10}\frac{\eta }{{\eta }_0}$$

where η is the effective shear viscosity of the heterogeneous liquid (suspension, emulsion) and η₀ is the shear viscosity of the crystal-, bubble-, and droplet-free homogeneous liquid. In what follows the definition displayed in Eq. (1) is used to distinguish between the viscosity specific of a simple homogeneous melt or of the carrier phase in a complex mixture (labelled by η₀), and the overall effective viscosity of a bi(multi)phase heterogeneous melt (labelled by η). However, one has to stress that only for suspensions containing an inert filler η₀ is a constant under isothermal conditions, i.e. η₀ is independent of the filler volume fraction. For so-called reactive fillers, η₀ can increase with the filler fraction, if the reaction depletes network modifiers in the liquid matrix. Vice versa, a decrease of η₀ with the filler volume fraction is expected for reactions leading to a depletion of network formers in the melt. In turn, η₀ will be also dependent on the volume fraction of the precipitate for melts that are prone to demix and to crystallize non-polymorphically. In these cases, it would be beneficial to further normalize viscosity to η₀ of the initially quenched melt.