Fourier transform infrared spectroscopy (FTIR)

For samples indicated in table 6.1, Fourier transform infrared spectroscopy (FTIR) was used to measure the water content in double polished wafers, 50-120 µm thick.

Clusters of crystals were avoided during these measurements, however small isolated crystals were unavoidable: the water contents from the raw data ΦH2O,measured have been re-calculated to account for the fraction of small isolated crystals φ_isolated, giving the true water content of the glass: Φ_H2O,glass

ΦH2O,glass = Φ_H2O,measured

1− φ^∗p(φisolated), (6.4)

and then for the bulk crystal fraction, giving the water fraction of the bulk sample ΦH2O,bulk:

ΦH2O,bulk = ΦH2O,glass(1− φ^∗p), (6.5)

so that the reported values are directly comparable to the TGA measurements.

The FTIR used for these measurements was the Thermo Nicolet IR interferom-eter at Lancaster University with a Continuum Analytical microscope, KBr beam-splitter, MCT-A detector and 100 µm square aperture. 128 spectra were collected at 4 cm⁻¹ spectral resolution within the mid-IR range of 4000-1000 cm⁻¹, with a 12-point linear baseline correction used to quantify absorption peaks at 3550 cm⁻¹ (-OHT) and 1630 cm⁻¹ (H2Om). Sample thickness was determined using reflectance fringes (von Aulock et al., 2014). Sources of error include uncertainty in sample thickness, absorption coefficients and peak quantification (von Aulock et al., 2014) and typically combine to create uncertainty of 10%.

6.4 Results

6.4.1 Bubble populations

Bubble fractions calculated using the vesicularity method presented in section 6.3.1 for each of the fully water quenched samples are plotted against down flow dis-tance from the vent, in figure 6.7. These measurements show clearly the significant reduction in bubble fraction with distance.

In addition, the bubble shapes change with distance, as can be seen in the SEM images in figure 6.5. The most proximal sample contains bubbles that appear to be spheroidal. With increasing distance, the bubble shapes become more distorted and large, non-spheroidal bubbles become more common. In all cases, the smallest

6.4. Results

0 5 10 15 20 25

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1

distance from vent (km)

bubble fraction

Figure 6.7: Plot of total bubble fraction using vesicularity method against distance from vent. Site 3, sample sp07 is shown in dark blue; site 4, sample sp11 is shown in pale blue; and site 6, sample s07 is shown in red. The error bars indicate the standard deviation from 13-15 measurements for each site.

bubbles appear to be closer to spherical; the limit above which bubbles lost their sphericity was a radius of approximately 0.1 mm.

6.4.2 Crystal populations

All samples contain phenocrysts and microphenocrysts of plagioclase feldspar and pyroxene. These are often found in clusters of significantly intergrown crystals (glomerocrysts), and are also found in apparent isolation. The higher resolution SEM images used to create the montages in figure 6.5 were used to measure the size and aspect ratio of crystals. The most numerous type of crystal in any sample was always isolated plagioclase, followed by clustered plagioclase, clustered pyroxene, and isolated pyroxene (one example image is shown in figure 6.8).

The observed crystal populations and textures change both with distance from the vent and with type of quench. The crystal population that most closely repre-sents the crystal cargo of flowing lava is likely to be captured by the most rapidly quenched samples: for site 1, this is naturally cooled sample sp01; for sites 3, 4 and 6, these are totally water quenched samples sp07, sp11, and s07 (figure 6.5;

see figure 6.2 for sample locations). As a result of being naturally cooled rather

6.4. Results

isolated plagioclase

isolated pyroxene

clustered pyroxene

clustered plagioclase

length

width glomerocryst

Figure 6.8: Examples of clustered (glomerocryst) and isolated plagioclase and pyroxene crystals, from the fully water quenched site 6 sample. The image is approximately 1 mm wide.

than quenched, the site 1 sample contains relatively little clean glass, and it can be difficult to confidently distinguish microlites that grew during cooling from micro-lites and microphenocrysts carried by the lava while it flowed. The totally water quenched samples from the remaining sites contain clean glass with no sign of mi-crolite growth, and can be used to understand how the crystal cargo changed as the lava flowed away from the vent.

From the SEM montage images (figure 6.5), it is clear that the crystal fraction (φ^∗_p) increases with distance from the vent. This increase was quantified by measur-ing the area, length, and width of individual crystals in the higher resolution images (e.g. figure 6.8) in ImageJ: crystal fractions based on area measurements are plotted in figure 6.9.

In addition to the bulk increase in crystal fraction, there are more subtle changes in the characteristics of the crystal population with distance from the vent. Since crystal size measurements (i.e. length and width, as shown in figure 6.8) were made using two-dimensional thin sections, it is necessary to account for the true three-dimensional shapes and sizes of the crystals. CSDCorrections (Higgins, 2000) is a piece of software designed to calculate the three-dimensional crystal population that could have produced an observed two-dimensional crystal population. This software is commonly used in volcanological applications (e.g. Higgins and Meilleur,

6.4. Results

0 5 10 15 20 25

0 0.05

0.1 0.15 0.2 0.25

distance from vent (km)

crystal fraction

Figure 6.9: Plot of total crystal area fraction against distance from vent. Site 3, sample sp07 is shown in dark blue; site 4, sample sp11 is shown in pale blue; and site 6, sample s07 is shown in red. Error bars show standard deviations based on repeated measurements of an area of 5.7 mm² (equivalent to 8 high resolution SEM images, which is visually a roughly representative area for each sample). The large standard deviation for site 3 is due to the heterogeneties created by the clustered crystals; in other samples, this effect is reduced due to the larger fraction of non-clustered crystals.

2009; Vona et al., 2013; McClinton et al., 2014), and the crystal size distributions resulting from inputting observed two-dimensional size distributions in these lavas are shown in figure 6.10. Both the shift to larger crystal sizes and the increase in crystal fraction with increasing distance from vent are clear. This trend is seen in all four groups of crystals.

In order to calculate the effect of crystals on rheology, two parameters are needed:

the volume fraction and the maximum packing fraction, which in turn is affected by aspect ratio and particle roughness. The volume fraction will be the same whether crystals are dealt with in isolation or as part of clusters; the aspect ratio and rough-ness are not. The glomerocrysts tend to be more equidimensional than isolated crystals, but have a higher rugosity. The former factor will tend to increase the maximum packing fraction, while the latter will tend to reduce it, compared to the smooth, more elongate isolated crystals.

Dealing with a polydisperse population of crystals in general is never trivial, and typically results in a complicated effective medium method, in which the smallest

6.4. Results

Figure 6.10: Number of crystals per unit bubble-free volume (population density) in each crystal length interval, as calculated from CSDCorrections. a) isolated plagioclase, b) clustered plagioclase, c) isolated pyroxene, and d) clustered pyroxene. Site 3, sample sp07 is shown in dark blue; site 4, sample sp11 is shown in pale blue; and site 6, sample s07 is shown in red.

crystals are considered first and additional populations of increasingly large crystals are then added. This level of complexity is probably unnecessary here: the crystal sizes vary by less than an order of magnitude, which is typically quoted as the minimum size difference needed for a continuous medium method to be justified (Farris, 1968). As a result, instead of considering the effects of the different crystal populations independently, the crystals that compose the glomerocrysts have been treated as individuals, rather than clusters. Although this is a simplification, it does seem justified here, as discussed above.