Introduction

Explosive-eruption dynamics are very complex and depend on a large number of parameters, such as the characteristics of the volcanic system (e.g., geometry of the conduit), the chemical and rheological characteristics of the magma involved, and the dynamics of both magma rise and fragmentation (Verhoogen 1951; Sparks 1978). As a result, textural features (e.g., bubble size and concentration) of the products of explosive eruptions provide important insights into eruption dynamics (Cheng and Lemlich 1983; Houghton and Wilson 1989; Cashman and Mangan 1994; Klug and Cashman 1996; Sahagian and Proussevitch 1998; Blower et al. 2001, 2002; Gaonac’h et al. 2005; Toramaru 2006).

Experimental and numerical studies have shown that fragmentation in silica-rich systems occurs after a rapid decompression that leads to a non-equilibrium continuous nucleation process (Blower et al. 2001, 2002). The resulting products are characterized by a high vesicle number per unit volume. In particular, textural characteristics of rhyolitic products can be considered as representative of the magma conditions at fragmentation. The content of vesicles, expressed as vesicle number density (N V, number of vesicles per unit volume) is controlled by magma properties, such as composition, temperature, viscosity, diffusivity, and interfacial energy, and it strongly depends on the decompression rate (Toramaru 2006).

In this work, we present a detailed textural analysis of pumices lapilli produced by the May 6, 2008, Chaitén climactic phase (Layer β, 6 May; Alfano et al. 2011) with the main objective of determining the associated explosive dynamics.

The May 2008 Chaitén eruption

The May 2008 Chaitén eruption (Fig. 1a) developed from an initial highly explosive phase, which lasted approximately 2 weeks, to a second, less explosive phase characterized by dome extrusion, explosions and collapses, and the generation of pyroclastic density currents (Alfano et al. 2011). The few seismic data available indicate that the eruption was preceded by a very short period of precursory seismic signals (2 days; Lara 2009) followed by a series of several explosive events with different intensity and duration. During the initial phase, the eruptive products were dispersed to the east, over a wide area up to 600 km away from the vent, producing a tephra deposit of about 0.5 km3 bulk volume. The explosive phase reached a climax on May 6 with an associated 19-km-high column and a bulk volume of about 0.1–0.2 km3 dispersed towards the NE (Fig. 1b; Alfano et al. 2011). The pyroclastic products erupted during the climactic phase produced a lapilli-rich layer (Layer β; Alfano et al. 2011) characterized by a thickness exceeding 20 cm within 10 km from the vent (Fig. 1c). The material erupted during this phase is composed mainly of lithic fragments (∼80 wt. %) produced by the disruption of the old dome, and a minor fraction of juvenile fragments represented by vesicular pumices (∼10 wt.%) and fresh and not-altered obsidian fragments (∼10 wt.%) (Alfano et al. 2011). The climax of the eruption was followed by a second phase started with the beginning of the dome extrusion. The shifting of activity from explosive to effusive occurred within the same explosive phase, as shown by effusion of an obsidian flow simultaneous with the explosive events are reported (Castro et al. 2012).

Fig. 1
figure 1

a Location of the Chaitèn volcano, Chile. b Isopach map of the tephra fallout produced from the May 6, 2008, climactic explosion (in centimeters; modified after Alfano et al. 2011). The pumice clasts analyzed were collected about 5 km from the vent along the dispersal axis (black circle in figure). c Picture of the deposit where the clasts were collected (courtesy of Laura J. Connor, University of South Florida, USA). The deposit in this location consists, from bottom to top, of a basal layer of brownish ash with lithic lapilli at the very base produced by the opening explosive event of May 1–2, 2008, (Layer α); the lapilli layer produced during the climactic explosion of May 6, 2008 (Layer β); the complex sequence of tephra layers ranging from fine ash to lapilli which represents the activity after May 6, 2008, (layers χ−ν); a reworked layer of ash covers the entire sequence (Alfano et al. 2011). d Picture showing a selection of the pumice lapilli studied in this work

Full size image

The pumices (Fig. 1d) are aphiric (<1 vol% crystals) and rhyolitic (Castro and Dingwell 2009; Alfano et al. 2011). The crystal population is composed by microphenocrysts (0.5–1.0 mm) of plagioclase, Fe-Ti oxides, orthopyroxene, and biotite, with associated rare microlites of plagioclase and biotite. The volatile fraction, mainly water, is estimated to be in the range 1.3–2.3 wt.% (Castro and Dingwell 2009). In contrast, obsidian fragments show higher degrees of crystallinity (∼2–5 vol% crystals) with plagioclase crystals up to 2–3 mm diameter and a lower water content in the range 0.5–1 wt.% (Castro and Dingwell 2009). Analysis of the microlite composition and decompression experiments carried out by Castro and Dingwell (2009) have shown that the dynamics of the eruption are characterized by a rapid rise of a water-saturated rhyolitic magma from from depths of >5 km, with an estimated average velocity of about 0.5 m/s and a short magma ascent time. However, the viscosity (∼106–108 Pa s) is estimated to be at least one order of magnitude too low to produce a magma autobrecciation as a result of shear during its rise in the conduit (Castro and Dingwell 2009).

Methods

Density and porosity measurements

A density and porosity study was carried out on a population of 100 pumice lapilli (P1 to P100) of the climactic phase (i.e., Layer β, May 6; Alfano et al. 2011) collected about 5–6 km north-east of the crater. Pumice clasts are all 2–6 cm diameter (Fig. 1d) small enough, 8.5 ± 4.9 cm3, to have cooled rapidly with little post-fragmentation vesicle expansion (Thomas and Sparks 1992; Tait et al. 1998).

The density distribution of the pumice clasts was determined using a hydrostatic balance. In order to include all the superficial vesicles in the measurement, the clasts were wrapped using parafilm (Houghton and Wilson 1989). Results were converted into bulk porosity (ratio between the volume of all the vesicles and the volume of the pumice including the vesicles) based on the average solid density measured on powdered pumice clasts using a helium pycnometer at the University of Geneva (Quantachrome ULTRAPYC 1200e). Then, 50 clasts were selected, taking care to cover the entire range of density, and a characterization of the porosity was carried out. Density measurements using the helium pycnometer were carried out on unwrapped clasts and converted using the average solid density of the powder in order to obtain the value of the closed porosity (ratio between the volume of the vesicles not connected with the surface and the volume of the pumice clasts including all the vesicles). Open porosity (ratio between the volume of the vesicles connected with the surface and the volume of the pumice clasts including all the vesicles) was obtained by the difference between bulk and closed porosity and represents all the interconnected vesicles also connected with the surface. Relative values of open and closed porosity were also calculated as ratios between the volume of open and closed vesicles and the total volume of vesicle in each clast.

In addition, density measurements on ten juvenile obsidian clasts coeval with the pumice clasts were carried out using the helium pycnometer and represent a non-vesicular endmember of the products of the explosive event.

Textural analysis of pumice lapilli

Textures were studied of seven pumice lapilli selected from representative density classes of the pumice population (i.e., most frequent and endmember density classes; Table 1). Thin sections cut at random orientations were made from four samples, and two oriented thin sections, orthogonal and parallel to vesicle elongation, were taken from three samples (in order to represent oriented structures present in the pumice clasts). For each of the ten thin sections, a set of 17 images was acquired at four different magnifications. An image of the entire thin section was taken using a Nikon Super Coolscan 4000 (resolution 157.5 pixels/mm). Fourteen scanning electron microscope (SEM) backscatter images of parts of the thin section were taken using the JEOL JSM7001F at the University of Geneva (resolutions 267, 1,070, and 2,670 pixels/mm) following the nesting strategy described by Shea et al. (2010b). Two additional images (resolution 1,000 × 1,000 pixels) were extracted from the section image in order to analyze vesicles down to 0.5 mm equivalent diameter to cover the entire range of vesicle size. Images were prepared for analysis using Adobe Photoshop CS3, rebuilding manually vesicle walls and producing binary images that were processed using JMicrovision (www.jmicrovision.com). The image analysis was carried out in order to study the morphology of the vesicles, the vesicle wall thickness, and the 2D vesicle size distribution (VSD).

Table 1 Summary of data for clast density, vesicularity, and textural features for low (italic) and high (bold) density samples
Full size table

Vesicle morphology was studied based on the frequency distribution of the aspect ratio (AR; ratio between width and length of a vesicle) and the solidity factor (SF; ratio between the area of a vesicle and the area of the convex hull of the vesicle, which is the line of shortest distance which connects the maximum projections on a particle outline). Aspect ratio describes the elongation of the vesicle, varying between extremely elongate (<0.2), very elongate (0.2–0.4), moderately elongate (0.4–0.6), slightly elongate (0.6–0.8), and not elongate (0.8–1.0) (Blott and Pye 2008). The solidity factor describes the roughness of the outline of the vesicle. Smooth vesicles have a convex outline, with few or no concavities, so the projected area of the particle will be almost equal to the area of the convex hull and the resulting SF will be close to 1. As roughness increases, the outline will be characterized by a larger number of concavities, and the associated SF will be reduced (Blott and Pye 2008).

Vesicle wall thickness was measured superimposing four grids of parallel lines oriented in four different directions (0°, 45°, 90°, and 135°) to the thin section images at high magnification (2,670 pixels/mm). The distribution was obtained by deleting the areas occupied by vesicles from the grid and measuring the length of the remaining segments.

VSD was studied based on the determination of the vesicle number per unit area (N A, mm−2) for each thin section and a geometric size class distribution with constant ratio 10−0.1 (Sahagian and Proussevitch 1998; Shea et al. 2010b). N A distributions were calculated for each of the 17 images of each section, and a total N A distribution was determined by convolution of the data obtained from each single image. N A distributions were converted to number of vesicles per unit volume (N V, mm−3) by dividing N A for the central value of diameter of each size class (Cheng and Lemlich 1983). The vesicle volume fraction of the sample was calculated by multiplying N V for each bin class with the volume of the corresponding equivalent sphere. Resulting values were corrected for glass content using factor equal to the ratio between the measured and calculated porosity.

Results

Density of pumice lapilli

Pumice lapilli show a unimodal density distribution (400–1,300 kg/m3; main mode at 700 kg/m3; Fig. 2a) with a high-density tail, ranging between 1,000 and 1,300 kg/m3 (i.e., ∼10 % of the distribution; Fig. 2a). Bulk porosity, calculated based on helium pycnometer bulk density of 2,242 ± 14 kg/m3, ranges between 43 % and 80 %. No clear correlation between pumice volume and porosity was found. Porosity analyses were carried out on 38 pumice lapilli with densities less than 800 kg/m3, representing the most frequent density classes, and on 12 pumice clasts from the high-density tail. Bulk and open porosity decrease accordingly, with an average value of the open porosity equal to 53 ± 9 %. In contrast, closed porosity remains roughly constant, with an average value of 15 ± 4 % (Fig. 2b). Relative open porosity varies between 70 % and 86 % of the bulk porosity. In contrast, closed porosity varies between 14 % and 30 % of the bulk porosity showing a slightly increase for the high-density tail pumice clasts (27 ± 11 %) with respect to the pumice clasts with density less than 800 kg/m3 (21 ± 5 %) (Fig. 2b).

Fig. 2
figure 2

a Density distribution of the 100 pumice lapilli analyzed in this work. b Porosity characterization of the pumice clasts showing the variation of the closed and open porosity in relation with the density and of the bulk porosity

Full size image

Obsidian lapilli are characterized by helium pycnometer density of 2,267 ± 33 kg/m3, very close to solids density of the powdered pumice clasts (2,242 ± 14 kg/m3), which indicates that these products can be considered as not vesicular.

Description of the thin sections

A qualitative and quantitative characterization of the vesicles was carried out on pumice lapilli of both endmembers and modal density classes (Fig. 2a). Analyses on sections with generic orientation have been carried out on samples P48, P70, P38, and P26; analyses on orientated sections have been carried out on samples P02, P25, and P39.

The analyzed pumice lapilli are characterized by highly stretched vesicles and an almost crystal-free glass groundmass (Fig. 3a). No evidence of post-fragmentation expansion or breadcrusting was observed. Vesicles tend to be larger in clasts with lower density, and coalescence is greater in low-density clasts (cf., P48, P02, P70, and P38, Fig. 3a). Vesicle morphologies are highly irregular, and no particular differences are observed in sections oriented perpendicular and parallel to vesicle elongation (cf., P02, P25, and P39, Fig. 3a). Vesicles in low-density clasts show regular shapes with regular convex outlines (cf., P02, P25, and P39, Fig. 3a). High-density pumice clasts are characterized by stretched vesicles occasionally presenting indented walls (cf., P25 and P39, Fig. 3a). Aspect ratios show unimodal distribution for all the analyzed sections except for the P25 parallel section (Fig. 3b). Vesicles generally show a high degree of elongation, with median values of AR ranging between 0.4 and 0.6. A slightly higher degree of elongation is found for samples P70 and P38 (cf., Table 1). Sections orthogonal to the direction of vesicle elongation show AR values slightly higher than sections parallel to it (difference < 10 %). Vesicles have various shapes, with SF varying widely from 0.3 to 1. Outlines of vesicles are more irregular for the most elongate vesicles, with SF (Table 1) varying from high values (about 0.9 = regular outline), for slightly and not elongate vesicles, and decreasing progressively as elongation increases. This behavior is particularly evident in samples P25 and P39, where vesicles with high elongation can be very irregular (SF < 0.7).

Fig. 3
figure 3

a Selection of SEM images for some analyzed clasts with increasing density from top to bottom. Vesicles are in black, glass walls in white. Width of the images is indicated for each magnification (millimeters). Clasts number is indicated on the right. b Frequency distribution of aspect ratio (AR) of the vesicles of all the low- and high-density pumice clasts analyzed in this work

Full size image

In order to investigate the distribution of vesicle walls, samples were divided in two groups by density: low-density samples (blue symbols; P48, P02, P38, and P70; density between 441 and 671 kg/m3) and high-density samples (red symbols; P25, P39, and P26; density between 859 and 1,271 kg/m3) (Fig. 3b). Median thickness of vesicle walls are in the range 4–8 and 10–15 μm for low- and high-density samples, respectively (Fig. 4), corresponding to a 2D glass fraction of 0.3–0.5 and 0.6–0.7.

Fig. 4
figure 4

Thickness distribution of vesicle walls based on the measured length of the segments connecting adjacent vesicles

Full size image

Vesicle size distribution (VSD)

Despite vesicles being texturally different in terms of their morphology and wall thickness, only slight variations in the VSD are observed. N A is similar for all samples, equal to 1.3 ± 0.5 × 103 mm−2 for the low-density samples to 1.0 ± 0.2 × 103 mm−2 for the high-density samples. Average vesicle number per unit area is 1.2 ± 0.4 × 103 mm−2. As a result, N V gives similar values for the two density classes of samples (9.9 ± 4.4 × 104 mm−3 for low-density pumice clasts and to 8.7 ± 2.0 × 104 mm−3 for high-density pumice clasts). These values correspond to different ranges of vesicle volume fraction (VVF) equal to 0.50–0.63 and 0.32–0.44 for low- and high-density pumice clasts, respectively (cf., Table 1).

Distribution of vesicle sizes are described plotting volume fractions (corrected for the melt) with vesicle sizes expressed as diameters of equivalent spheres (Fig. 5). Vesicle size are distributed unimodally with mode of 0.05–0.08 mm, with the only exception of the orthogonal section of P02 and the parallel section of P25 that show mode at 0.08–0.13 mm. Observed minimum and maximum sizes of the vesicles are ∼0.01 and ∼3 mm, respectively. Vesicles with equivalent diameter larger than 1 mm are mostly present in the clasts with the lowest density (P48 and P02). VSD is similar for all clasts and do not vary with the orientation of the sections nor with respect to vesicle elongation.

Fig. 5
figure 5

VSD expressed in volume fraction corrected for the melt for all the low and high density pumice clasts analyzed in this study. The average vesicle volume fraction is reported for each single plot

Full size image

Volume fraction of the analyzed clasts was normalized to the average VSD to better compare associated textures and show similar distributions (Fig. 6a). The most significant exceptions are represented by the content in vesicles <0.03 mm for the sample P70 and the content in large vesicles (>0.3 mm) for the low-density samples (P48 and P02). Nonetheless, all samples show similar values in the range 0.05–0.13 mm, in agreement with the modes of the volume fraction distributions. Given the small differences between samples, an average N corrV distribution inferred for the bulk magma was calculated, and the converted cumulative number density (N corrV  > d; cubic millimeter) was plotted versus particle diameter (Fig. 6b). The distribution is characterized by two populations of vesicles, both of which follow power-law trends with different slope. For small vesicles (d < 16 μm), N corrV  > d is characterized by a power-law trend with exponent (E 1) equal to 1.1. For larger vesicles (d > 16 μm), N corrV  > d is characterized by a power-law trend with exponent (E 2) equal to 3.6. However, power-law exponents E 1 and E 2 for individual samples vary respectively in the ranges 1.0–1.7 and 3.5–4.2 (cf., Table 1). As a result, the average number of vesicles per unit volume corrected for the melt (N corrV ) inferred for this explosive phase is 1.3 ± 0.5 × 105 mm−3.

Fig. 6
figure 6

a VSD expressed in volume fraction normalized for the average volume fraction (volume fraction/average volume fraction) for all the analyzed samples showing the differences in vesicle populations. Symbols as in Fig. 5. b Average cumulative N corrV distribution. Standard deviation is indicated for each point

Full size image

Figure 7 shows the relation between mass eruption rate (MER, kilograms per second; Wilson and Walker 1987) and the N V for the Chaitén explosion of May 6, 2008, and other studied eruptions of basaltic, rhyolitic, and phonolitic magmas. Basaltic eruptions seem to show a trend between MER and N V, whereas rhyolitic and phonolitic eruptions do not. Values of the eruptive parameters of all the eruptions are collected in Table 2.

Fig. 7
figure 7

Log–log plot of the mass eruption rate (MER) versus the vesicle number density corrected for the melt (N corrV ). References as in Table 2

Full size image
Table 2 Summary table of main eruptive parameters of studied eruptions
Full size table

Decompression rate

The decompression rate that characterized the Chaitén climactic phase of May 6, 2008, was calculated using the model of Toramaru (2006) defined for homogeneous nucleation of the vesicles. A rhyolitic magma (74 SiO2 wt.%; Alfano et al. 2011) rich in volatiles (2–5 H2O wt.%) and a temperature interval in the range 775–850 °C were considered (Castro and Dingwell 2009). Only vesicles with diameter <0.01 mm (Shea et al. 2011) have been considered in the calculation, assuming they correspond to the last nucleation event before the fragmentation (N fV  = 7.3 ± 3.0 × 104 mm−3). The resulting decompression rate is on order of 8.1 ± 2.9 MPa s−1, corresponding to an exit velocity in the range 330–790 m/s (assuming a mixture composed of 50 % vesiculated magma and 50 % not-vesiculated magma with density 1,490 ± 100 kg/m3). If the total N corrV is considered, the value of the estimated decompression rate increases to 12.1 ± 4.3 MPa s−1, corresponding to an exit velocity in the range 490–1,180 m/s.

Discussion

Pumice lapilli erupted during the climactic phase of May 6, 2008, of Chaitén volcano are characterized by a unimodal VSD with a predominance of small vesicles with modal diameter between 0.05 and 0.08 mm (Fig. 5) that does not vary significantly with clast density. Unimodal distribution and the high frequency of small vesicles suggest that vesiculation occurred over a very short time and relatively late during magma ascent (Klug et al. 2002), as also suggested by the absence of microcrystals and by the relatively low viscosity (∼106–108 Pa s−1) of the rhyolitic melt (Castro and Dingwell 2009). Given that there was not enough time for vesicles to expand, only a small number of vesicles with diameter >1 mm was observed, and they are only present in the clasts with low density (P48 and P02; cf, Figs. 3a and 5).

The lack of large vesicles in the dense pumice clasts may be also enhanced by processes of collapse that produced the irregular vesicles characterized by lower values of SF observed in sections P25 and P39 (cf., Table 1 and Fig. 4a). Open porosity shows high values suggesting that coalescence may have played a role in the evolution of magma porosity. However, it is also possible that fractured vesicle walls might have contributed to reach these high values. Open porosity increases along with the bulk porosity, as the probability of vesicles to coalesce producing a complex network of interconnected vesicles increases with the number and the volume of the vesicles. This may have favored the degassing process that produced the collapsed vesicles observed in samples P25 and P39. Pumice lapilli of the high-density tail show how an increase in relative closed porosity corresponds to a decrease in open porosity through collapse processes.

Cumulative N V plots produce power-law trends that are usually interpreted as the result of vesicle nucleation under non-equilibrium conditions, which is characteristic of explosive eruptions, especially of silica-rich magmas (Mangan and Cashman 1996; Blower et al. 2001, 2002). In fact, similar trends are reported for the products of the 1875 eruption of Askja volcano, which show VSD characterized by two power-law trends with the branch representing intermediate and coarse vesicle size characterized by exponents in the range 2.3–5.1 (Carey et al. 2009). In addition, the 1980 Mt. St. Helens eruption shows a power-law trend with exponent of about 3.4 (Rust and Cashman 2011), and the 1912 eruption of Novarupta shows a power law-trend with exponent of 3.9 (Adams et al. 2006a, b).

The relation between the evolution of the volatile fraction present within a melt and the explosivity of the eruption is not totally understood. Considering the values of MER and N V (cf. Table 2 and Fig. 7), Chaitén volcano’s May 6 explosion is characterized by values close to those estimated for layers 2 and 5 of Cotopaxi, Etna 122 BC, and units B, D, and F of Towada volcano. Differences of at least one order of magnitude in the values of MER and N V are observed when comparing Chaitén eruption with the other cases reported in Table 2. Houghton et al. (2010) assert that there is a positive correlation between MER and N V. According to Rust and Cashman (2011), this correlation cannot be generalized to all cases, and it is limited to those cases in which vesiculation occurs in near-equilibrium condition. Basaltic eruptions show an increase of the vesicle number density with the mass eruption rate associated with a shift in the eruptive style from Hawaiian/Strombolian (Kilauea Iki, 1959; Stromboli, 2002; Villarica, 2004) to sub-Plinian (Izu-Oshima, 1986) and Plinian eruptions (Etna 122 bc, Fontana Lapilli and Quilotoa 800 bc, Tarawera, 1886). Vesicle number density of phonolithic eruptions does not show a clear correlation with eruption style (e.g., Vesuvius 512 ad and Vesuvius 79 ad), but more phonolitic eruptions should be studied to confirm this observation. More complex is the behavior of andesitic/rhyolitic eruptions. Sub-Plinian andesitic/rhyolitic eruptions (Chaitén May 6, 2008, and Towada volcano Units B, D, and G) follow the trend of basaltic Plinian/sub-Plinian eruptions, with the exception of the Unit B of the 1875 Askja eruption, which shows N V values more similar to Plinian andesitic/rhyolitic eruptions. Andesitic/rhyolitic Plinian eruptions partially follow the trend of basaltic Plinian/sub-Plinian eruptions (Cotopaxi layers 1, 2, and 5, Novarupta episode II, and Towada Units A, C, E, and F) and partially the trend of phonolitic eruptions (Askja 1875 Units C and D, Mt. St. Helens May 18, 1980, Novarupta 1912 episode III, and Taupo 1.8 ka). It is important to notice that the MER range is similar for most Plinian/sub-Plinian eruptions independently on the composition (i.e., mostly 106–109 kg/s), while the largest N V values are shown by both Plinian andesitic/rhyolitic eruptions and Phonolitic eruptions. The lowest MER and N V values are shown by Strombolian and Hawaiian eruptions. We can conclude that Plinian and sub-Plinian eruptions are difficult to distinguish only based on N V and MER values (with N V mostly >104 mm−3 with the exception of Tarawera 1886) but are very different from Strombolian and Hawaiian eruptions that are characterized by MER < 105 kg/s and N V < 5 × 104 mm−3. Regardless of the general trends shown by Fig. 7, the relation between N V and MER is complex, especially when we consider the high degree of uncertainty in the determination of both parameters. The high uncertainty in N V results from determination obtained typically by the statistical 2D analysis of pumice samples, with the assumption that a small number of pumice clasts can be considered a representative of the whole magma. In addition, MER can be currently estimated only within a factor of 10 due to the large uncertainties associated with both the existing expressions that relate plume height and MER and the current strategies used to determine the erupted mass and plume height (Mastin et al. 2009).

The textural features found in Chaitén products suggest an eruption driven by a violent decompression of the magmatic system that triggered the homogeneous vesiculation of a water super-saturated magma. Considering the large content of lithic fragments (∼80 %; Alfano et al. 2011), we think that this sub-Plinian event was generated by the disruption of the old obsidian dome and the consequent opening of an ∼800-m radius vent (Smithsonian Institution 2008). According to the calculated N V, for a temperature interval of 775–850 °C and a water content of 2–5 wt.%, the decompression rate estimated to have produced such a sub-Plinian explosion is about 10 MPa s−1 (Toramaru 2006). This value of decompression rate agrees with values calculated for past eruptions and presented by Toramaru (2006) using the decompression rate meter for homogeneous nucleation. As an example, the Chaitén explosion of May 6 shows values of column height, composition, and N V (cf., Table 2) similar to the sub-Plinian episodes of the historical eruptions of Towada caldera and Izu-Oshima (Toramaru 1990, 2006; Blower et al. 2002), whose decompression rate is estimated to have been in the range between 6.3–91.0 MPa s−1, for Towada caldera, and 4.9 MPa s−1, for Izu-Oshima (cf., Table 2). This behavior is related to the different mechanism controlling vesiculation. In basaltic magmas, where the viscosity is low, vesiculation is controlled by diffusion and coalescence, with the result that same decompression rates produce lower N V than in rhyolitic melts, where vesiculation is controlled by the high viscosity (Toramaru 1995; Klug and Cashman 1996). In contrast, the values of N V and decompression rate estimated for the 1980 eruption of Mt. St. Helens are one order of magnitude higher than those estimated for the Chaitén explosion. In this particular case, the sector-collapse that triggered the eruption (Holasek and Self 1995) caused a high decompression rate.

Crystallinity can play a role in eruption dynamics. As an example, the decompression rate values calculated for the 79 ad eruption of Vesuvius (Shea et al. 2011) assuming an heterogeneous nucleation of the vesicles gives values in the range 0.4–6.2 MPa s−1, in the same order as those for the Chaitén eruption. The 79 ad Plinian eruption of Vesuvius (cf., Table 2: Gurioli et al. 2005; Shea et al. 2010a) is characterized by N V of one to two orders of magnitude higher than the N V calculated for the pumice clasts of Chaitén. This aspect shows the important effect that the absence of microcrystals in Chaitén melt had on the eruption dynamics, as higher decompression rate are required to trigger vesiculation.

The presence of 10 % of non-vesicular, juvenile obsidian fragments within the tephra deposit and the simultaneous obsidian effusion documented by Castro et al. (2012) suggest that the pumice lapilli are the relict of a volatile-rich batch of magma, which was involved in the very beginning of the eruption. The abrupt difference in density and porosity of pumice and obsidian clasts and the lack of intermediate varieties indicate a dynamic of eruption involving volatile-rich magma and so was able to vesiculate and produce the sub-Plinian phase and a volatile-poor magma. Castro et al. (2012) explain the relation between the volatile-rich and the volatile-poor magma as the result of a degassing through a magma fracturing process induced by shear stress. The collapse morphology observed in the vesicles of the high-density samples indicates clearly that a degassing process was developing during the magma rise. In addition, all the samples, regardless of density have highly elongated vesicles, with medium AR in the range 0.4–0.6 (cf., Fig. 3b), indicating that a high shear stress was acting on the magma body, extending its influence to the middle of the conduit, as a result of both the high ascent velocity of the magma (0.5 m/s; Castro and Dingwell 2009) and its rise through a narrow dike (Wicks et al. 2011). However, the low values of viscosity (Castro and Dingwell 2009) and the degassing process acting on the magma body (Castro et al. 2012) indicate that conditions did not favor magma autobrecciation. Consequently, there must have been other factors acting on the system that triggered of the explosive phase. Fast ascent, enhanced by the low viscosity of the magma, and the absence of microcrystals did not allow for a significant vesicle nucleation and growth, reducing the efficiency of the shear-induced degassing. As a result, large portions of the magma body could reach shallow crustal levels highly supersaturated in water. In this situation, the preexisting dome had a critical role acting as a plug obstructing the volcanic conduit and causing pressure to increase. When failure of the dome occurred, the magma decompressed rapidly, triggering the nucleation of bubbles and the consequent sub-Plinian eruption that produced the Layer β deposit.

Conclusions

  • The juvenile products, in the size range 2–6 cm, of the climactic sub-Plinian explosion of May 6, 2008, of Chaitén volcano (Layer β) are characterized by a density range 400–1300 kg m−3 (bulk vesicularity ranging between 54 vol.% and 81 vol.%) with vesicle diameters <4 mm, irregular vesicle morphologies, some vesicle collapse structures, median vesicle walls thickness varying between 4 and 15 μm, and unimodal VSD with modal values in the range 0.05–0.08 mm and a total N V of 1.3 ± 0.5 × 105 mm−3.

  • The open porosity decreases with bulk porosity and represents the main fraction of the vesicles in the clasts (78 ± 8 %). This high degree of interconnection between vesicles favored degassing processes that produced morphologies of vesicle collapse and a higher fraction of closed porosity observed in the pumice clasts with high density (21 ± 5 % for pumice clasts with density >800 kg m−3; 27 ± 11 % for pumice clasts with density <800 kg m−3).

  • Unimodal VSD and the power-law trend of the cumulative N V plots indicate that the magma was not in equilibrium with the volatile fraction and produced a rapid and continuous homogeneous nucleation that occurred in the later phases of the magma rise through the conduit. The rapid rise and the absence of microcrystals delayed the magma degassing that started at shallow levels.

  • Fragmentation was triggered by the nucleation of vesicles due to a sudden decrease of pressure estimated to be about 10 MPa s−1, produced by the failure of the preexisting obsidian dome during magma rise. After a highly explosive phase, the activity shifted to the effusive phase that involved volatile-poor magma, which started to erupt simultaneously with the explosive activity. Relicts of this magma batch are included in the tephra deposit as non-vesicular obsidian clasts.