Magma degassing can trigger crystal growth by increasing the magma liquidus temperature. As crystallization greatly increases magma viscosity, this process can strongly influence eruptive dynamics. We use a microscope and heated stage to obtain the first direct observations of degassing-driven crystal growth in natural basaltic melts at magmatic temperatures. Samples from Mount Etna, Italy (0.39 wt% H2O), and Kilauea volcano, Hawaii (0.18 wt% H2O) were heated in air at 1 bar, and held isothermally for 0.5–17 h between 1190 °C and 1270 °C, before cooling to solidus temperatures. On heating, bubble growth at >900 °C indicated volatile exsolution. In the hydrous Etna sample, isothermal conditions produced numerous new plagioclase crystals that grew to ≤160 μm at maximum rates of 5.2–18 × 10−6 cm s−1. Growth rates and crystal morphologies (tabular to spherulitic) depended on dwell temperature. Growth slowed dramatically after 20 min as equilibrium was approached. In the H2O-poor Kilauea sample, few new crystals appeared; they grew at maximum rates of 1.7–6.5 × 10−6 cm s−1. On cooling, crystal nucleation and growth were strongly influenced by preexisting crystal textures, highlighting the importance of studying natural samples. Our results document rapid crystal growth triggered by melt devolatilization when the H2O content of the glass is sufficiently high. Resultant swift, substantial changes in magma texture are a key control on lava rheology at Etna and elsewhere.


The volatile contents and crystal textures of magmas are critical in determining their rheology and eruptive behavior (e.g., Gonnermann and Manga, 2007). Volatile exsolution increases the melt liquidus temperature (Tliq), so degassing-induced undercooling can induce crystallization without cooling (Sparks and Pinkerton, 1978; Cashman, 1992; Hort, 1998) or even during heating (Blundy et al., 2006). Degassing-driven crystallization triggers profound rheological changes in ascending intermediate and silicic magmas (Swanson et al., 1989; Sparks et al., 2000), and may trigger highly explosive basaltic volcanism (e.g., Sable et al., 2006).

To assess whether degassing-driven crystallization is significant on the timescales of magma ascent or lava flow advance, crystal growth rates (G) are needed. Textural analysis of natural samples allows estimates of mean G (Gmean) using crystal size distribution (CSD) theory (Crisp et al., 1994). However, maximum G (Gmax) can be orders of magnitude higher than Gmean, as systems are perturbed from, and then return to, equilibrium (e.g., Cashman, 1993; Hort, 1998). Experimental studies have examined degassing-induced crystallization in silicic (e.g., Hammer and Rutherford, 2002; Couch et al., 2003) and intermediate-composition (e.g., Szramek et al., 2006) magmas. However, the high temperatures of basaltic melts pose experimental challenges, and so the kinetics of degassing-induced crystallization of mafic melts have not been well constrained.

Gmax can be estimated from quenching experiments of short duration (<30 min). However, such estimates for plagioclase in basalts are highly variable (10−5 to 10−9 cm s−1; Simakin and Salova, 2004; Orlando et al., 2008) and depend on experimental conditions. The best way of determining G is by direct observation at high temperature using a heated microscope stage (Kirkpatrick et al., 1976, 1979), as used recently by Schiavi et al. (2009, 2010) to analyze cooling-driven crystallization of anhydrous, synthetic basaltic trachyandesites. Here we present the first direct observations of degassing-driven crystallization in variably crystalline and variably degassed natural samples at magmatic temperatures.


We studied a snow-quenched porphyritic K-trachybasalt from Mount Etna, Italy (46 vol% crystalline), collected from a small pahoehoe flow during the A.D. 2008–2009 eruption, and an air-cooled basalt from Kilauea, Hawaii (10 vol% crystalline), collected from a blue glassy pahoehoe flow in April 2009. Double-polished wafers, ≤50 μm thick and ≤0.8 mm across, were placed in a Linkam TS1500 stage on a Zeiss Axioscope. Experiments involved heating, an isothermal phase, and cooling to subsolidus temperatures (Table 1). Time-lapse imaging was used to monitor textural changes and measure crystal growth rates. Sample water contents (Etna 0.39 wt%, Kilauea 0.18 wt%) were measured using thermogravimetric analysis–differential scanning calorimetry (TGA-DSC). Exothermic peaks during degassing at >1000 °C indicated ∼20%–30% and <5% crystallization in Etna and Kilauea samples, respectively. See the GSA Data Repository1 for details of sample composition, TGA-DSC methods and results, and heated microscopy experiments.



In all experiments, samples swiftly became opaque at ∼650 °C due to oxide nucleation within glass. Bubble growth occurred in the melt after crossing the glass transition temperature (Tg) at ≥900 °C. In the Etna wafers, bubbles were seen near preexisting phenocrysts, where a thin wedge of melt remained translucent. In the Kilauea wafers, bubble growth began before oxide nucleation rendered the sample fully opaque, and therefore bubble formation could be clearly observed (Table 2). Melting of crystal phases was seen from ∼1090 °C (Etna) and 1135 °C (Kilauea). Samples regained translucency as the oxides melted. Bubble coalescence permitted gas escape within a few minutes of isothermal conditions being reached. Above 1170 °C, plagioclase crystals began to melt. Clinopyroxene and olivine crystals completely melted by 1230 °C, or within a few minutes when held at lower temperatures.

Isothermal Conditions

Isothermal temperatures (1190–1270 °C) were selected to provide subliquidus conditions (liquid + spinel ± plagioclase), but where little crystallization would occur without magma degassing. In all experiments on Etna samples, but only some on Kilauea samples, preexisting plagioclase crystals ceased to melt when the isothermal temperature was reached, and instead began to grow. Fe-Ti oxides were resorbed very slowly in all experiments, at rates of ∼10−8 cm s−1.

All isothermal experiments on Etna samples showed new plagioclase crystals (≥1.5 μm wide and ≥10 μm long) after ∼2 min (Figs. 1A and 1B). Nucleation occurred on preexisting crystals, bubbles, and in apparently crystal-free melt; morphologies varied with temperature. When the sample was held at ≤1230 °C, pale cloudy regions propagated into darker melt; we interpret these as spherulitic growth (Fig. 1C). Well-defined acicular crystals were rare at the lowest temperature (1190 °C), but were far more abundant at 1250 °C (Fig. 1B). At the highest temperature (1270 °C), rare new plagioclase crystals were tabular, and gradual darkening of melt regions probably indicated growth of small (<1.5 μm) oxides.

In contrast, isothermal experiments in Kilauea samples showed only sparse (and acicular) or no new crystal growth at temperatures of 1200–1250 °C. Additionally, new crystals were observable only where they grew against a background of opaques, due to poor contrast with the melt (Fig. 1D).

Crystal Growth

Time-lapse images were used to measure plagioclase G along length and width axes (Fig. 2), with 4–15 crystals measured per experiment (Table 1). Growth was most rapid in the long axis direction and in the first ∼15 min; thereafter plagioclase growth slowed (Fig. 2A), apparently due to impingement. Where acicular morphologies dominated, impingement left an open framework of randomly oriented crystals, and further growth occurred by widening (Fig. 2B).

Gmax was calculated from the steepest portion of each length-time or width-time curve. In the Etna sample, Gmax (sustained for 3–12 min; Fig. 2) increased with increasing temperature from 5.2 × 10−6 to 1.8 × 10−5 cm s−1 for length (Fig. 3) and 3.7 × 10−7 to 2.7 × 10−6 cm s−1 for width. Although there is considerable scatter in the data (Fig. 3), we found no dependence of G on crystal size or distance to nearest neighbor. Preexisting plagioclase crystals with initial lengths of 47–382 μm grew at 1.9–5.7 × 10−7 cm s−1, which may indicate a negative size dependence of G over a greater size range than represented in the new crystal population. New crystals grew in only one Kilauea experiment, with a Gmax of 1.7–6.5 × 10−6 cm s−1 for length.


In the Etna samples, new plagioclase crystals appeared after ∼20–30 °C of cooling from the isothermal temperature. Star-shaped opaque clinopyroxenes appeared at <1200 °C. In the Kilauea samples, paired plagioclase-pyroxene laths appeared after 20–50 °C of cooling (Fig. 1E), and star-shaped pyroxenes appeared by ∼1150 °C; crystallization was spatially heterogeneous. G increased rapidly 40–60 °C below the appearance temperature, producing dendritic crystals (Fig. 1F), and growth was limited by impingement. Gaps were filled by new crystals down to ∼970 °C, and no further growth was observed <900 °C. If samples were subjected to a second heating cycle, neither degassing (bubble growth and bursting) nor crystallization were observed.


Our results provide direct observations of isothermal crystal growth at magmatic temperatures in natural basaltic samples and confirm that degassing can trigger crystallization in basaltic melts.

Degassing-Driven Crystallization

Phase equilibria for the Etna sample were calculated using MELTS (Ghiorso and Sack, 1995; Asimow and Ghiorso, 1998) and predict a Tliq of ∼1260 °C for the oxidized, degassed samples. Spinel is the predicted liquidus phase, and plagioclase appears at 1210 °C. At 1190 °C, a crystallinity of 14% is predicted. The crystallinity was >14% at the start of all our isothermal dwells; hence, to achieve equilibrium, crystals should have melted. Instead, plagioclase growth was pervasive up to 1250 °C, perhaps indicating MELTS calibration problems for highly alkaline melts and low pressures (Simakin and Salova, 2004). A degassing-driven liquidus shift, which undercools the melt, would favor high-temperature crystallization. The degassing-induced undercooling, ΔTdg (°C), can be estimated from water loss, H2O (wt%), using the expression: 

(Médard and Grove, 2008). Complete degassing of Etna and Kilauea samples (0.39 wt% and 0.18 wt% H2O, respectively) would provide ΔTdg of 15 °C and 7 °C, respectively. The total effective undercooling, ΔTtot, is ΔTdg plus the undercooling due to cooling, ΔTcool, (Tliq minus the sample temperature). ΔTcool ranges from ∼10 °C (at 1250 °C) to ∼80 °C (at 1190 °C) (Fig. 3). Plagioclase morphologies vary with undercooling, ranging from tabular at low ΔTtot to spherulitic at high ΔTtot (Figs. 1 and 3), in agreement with previous experimentation (e.g., Lofgren, 1974; Hammer and Rutherford, 2002). The switch from lengthening to widening with time reflects decreasing ΔTtot as crystallization pushes the melt composition closer to equilibrium (e.g., Hammer et al., 1999; Hammer and Rutherford, 2002).

Etna and Kilauea samples experience similar ΔTtot, but undergo contrasting amounts of crystallization. Nucleation cannot be a limiting factor as nucleation sites are abundant (vesicles, spinel crystals) and both samples rapidly nucleate plagioclase on cooling. Instead we propose that a threshold degree of degassing (between 0.18 and 0.39 wt%) is required to drive crystallization, a concept supported by previous studies of silicic magmas (Hoblitt and Harmon, 1993; Wright et al., 2007). Although calculating ΔTtot during eruptions is difficult, crystal textures in Etna and Kilauea lavas suggest ΔTtot <65 °C, similar to the experiments. The more crystal- and water-rich nature of Etna lava likely reflects greater degassing-driven crystallization (e.g., Sparks and Pinkerton, 1978) associated with a higher initial water content (<3.7 wt% for Etna [Collins et al., 2009] versus <1 wt% for Kilauea [Edmonds et al., 2009]).

Controls on Crystallization

Our measured plagioclase Gmax values are compared with previous results in Table 3. G is commonly assumed to depend on either temperature or ΔTtot. Diffusion-controlled kinetics should cause an exponential increase of G with temperature (e.g., Orlando et al., 2008), consistent with the Arrhenius temperature-diffusivity relation. Our results support this relationship (Fig. 3). More complex relationships between ΔTtot and G have been proposed in some experimental (Kirkpatrick et al. 1979; Schiavi et al., 2009, 2010; Table 3) and theoretical (Kirkpatrick, 1975) studies. We note, however, that the experimental data of Schiavi et al. (2010) combine growth for new and preexisting crystals. Our G values for preexisting crystals are 1–2 orders of magnitude lower than those for new crystals, suggesting complex size-GT relationships.

Extrapolation of the best fit curve for our Etna data coincides well with the G values of Schiavi et al. (2009) (Fig. 3). We also note that our Gmax is similar to that measured by Simakin and Salova (2004) in water-saturated basalts at lower temperature (Table 3). Although G appears to depend on temperature when considering only fully degassed samples, melt viscosity is probably the key control as it controls the rate of diffusion.


Our results allow estimation of the time scale of crystallization and rheological stiffening in conduits and lava flows. In Etna lavas, 30–500 μm crystals are inferred to grow during degassing on ascent (CSD analysis; Armienti et al., 1994). Gmax at ascent temperatures (∼1080–1150 °C) is ∼2–4 × 10−6 cm s−1, from the curve in Figure 3 (inset). This gives growth times of 0.2–7 h for this crystal population, consistent with estimates of ascent times at Etna (e.g., Andronico et al., 2005). However, Gmax cannot be sustained for hours, and perhaps growth rates are faster in more volatile-rich melt at greater depth. On eruption, Etna lavas contain ≤50 vol% crystals (Pinkerton and Sparks, 1978), formed during storage and ascent. Our experiments suggest ∼20–30 vol% of this may form due to degassing on ascent. Relative viscosity increases only ∼10× as crystallinity rises from 0 to 30 vol%, but by up to four orders of magnitude over the interval from 30 to 60 vol% (Pinkerton and Stevenson, 1992). Determining the extent of degassing-driven crystallization during ascent is therefore critical for quantifying the profound rheological changes that influence eruptive behavior and hazards. Measurement of Gmax represents a crucial step toward this goal.

This work was funded by Natural Environment Research Council grant NE/I016414/1. Tuffen was supported by a Royal Society University Research Fellowship. We thank two anonymous reviewers and the editor for constructive comments that greatly improved the manuscript.

1GSA Data Repository item 2013059, details of thermogravimetric analysis–differential scanning calorimetry and heated stage microscopy methods, heated stage experimental setup, and two movies of crystal growth in the Mount Etna sample, is available online at www.geosociety.org/pubs/ft2013.htm, or on request from editing@geosociety.org or Documents Secretary, GSA, P.O. Box 9140, Boulder, CO 80301, USA.