Experimental solidification of an andesitic melt by cooling

Chemical Geology 283 (2011) 261–273

Contents lists available at ScienceDirect

Chemical Geology j o u r n a l h o m e p a g e : w w w. e l s ev i e r. c o m / l o c a t e / c h e m g e o

Research paper

Experimental solidification of an andesitic melt by cooling Gianluca Iezzi a,b,⁎, Silvio Mollo b, Guglielmo Torresi a,c, Guido Ventura b, Andrea Cavallo b, Piergiorgio Scarlato b a b c

Dipartimento DIGAT, Università G. d'Annunzio, Via Dei Vestini 30, I-66013 Chieti, Italy Istituto Nazionale di Geofisica e Vulcanologia, Via di Vigna Murata 605, 00143 Roma, Italy Institut für Mineralogie, Universität Hannover, Hannover, Germany

a r t i c l e

i n f o

Article history: Received 22 October 2010 Received in revised form 24 January 2011 Accepted 28 January 2011 Edited: D.B. Dingwell Keywords: Andesitic melt Experimental solidification Nucleation Crystal coarsening Disequilibrium phase Glass-forming ability (GFA)

a b s t r a c t Solidification experiments at (a) five different cooling rates (25, 12.5, 3, 0.5 and 0.125 °C/min) between 1300 and 800 °C, and (b) variable quenching temperatures (1100, 1000, 900 and 800 °C) at a fixed cooling rate of 0.5 °C/min were performed on an andesitic melt (SiO2 = 58.52 wt.% and Na2O + K2O = 4.43 wt.%) at air conditions from high superheating temperature. The results show that simultaneous and duplicated experiments with Pt-wire or Pt-capsule produce identical run-products. Preferential nucleation on Ptcontainers or bubbles is lacking. Plagioclase and Fe–Ti oxide crystals nucleate firstly from the melt. Clinopyroxene crystals form only at lower cooling rates (0.5 and 0.125 °C/min) and quenching temperatures (900 and 800 °C). At higher cooling rates (25, 12.5 and 3 °C/min) and quenching temperature (1100 °C), plagioclase and Fe–Ti oxide crystals are embedded in a glassy matrix; by contrast, at lower cooling rates (0.5 and 0.125 °C/min) and below 1100 °C they form an intergrowth texture. The crystallization of plagioclase and Fe–Ti oxide starts homogeneously and then proceeds by heterogeneous nucleation. The crystal size distribution (CSD) analysis of plagioclase shows that crystal coarsening increases with decreasing cooling rate and quenching temperature. At the same time, the average growth rate of plagioclases decreases from 2.1 × 10− 6 cm/s (25 °C/min) to 5.7 × 10− 8 cm/s (0.125 °C/min) and crystals tend to be more equant in habit. Plagioclases and Fe–Ti oxides depart from their equilibrium compositions with increasing cooling rate; plagioclases shift from labradorite–andesine to anorthite–bytownite. Therefore, kinetic effects due to cooling significantly change the plagioclase composition with remarkable petrological implications for the solidification of andesitic lavas and dikes. The glass-forming ability (GFA) of the andesitic melt has been also quantified in a critical cooling rate (Rc) of ~ 37 °C/min. This value is higher than those measured for latitic (Rc ~ 1 °C/min) and trachytic (Rc b 0.125 °C/min) liquids demonstrating that little changes of melt composition are able to significantly shift the initial nucleation behavior of magmas and the following solidification paths. © 2011 Elsevier B.V. All rights reserved.

1. Introduction The transition from a silicate melt to a fully solidified magmatic rock is an important phase transformation occurring on the Earth. The melt to rock transition involves vitrification and/or crystallization, two processes related to the melt composition and to temperature/ pressure variation (Dowty, 1980; Lofgren, 1980; Kirkpatrick, 1981; Cashman, 1991; Lasaga, 1997; Hammer, 2008). Dynamic crystallization experiments carried out to investigate the nucleation behavior of silicate melts mostly concentrated on peridotitic and basaltic liquids (Conte et al., 2006; Hammer, 2006; Pupier et al., 2007; Schiavi et al., 2009 and references therein). Conversely, few data are available for intermediate and evolved compositions (Swanson, 1977; Naney and Swanson, 1980; Couch, ⁎ Corresponding author at: Dipartimento DIGAT, Università G. d'Annunzio, Via Dei Vestini 30, I-66013 Chieti, Italy. Tel.: +39 0871 3556147; fax: +39 0871 3556047. E-mail address: [email protected] (G. Iezzi). 0009-2541/$ – see front matter © 2011 Elsevier B.V. All rights reserved. doi:10.1016/j.chemgeo.2011.01.024

2003; Hammer, 2004; Iezzi et al., 2008); crystallization data for andesitic melts are instead completely lacking (Iezzi et al., 2009). Importantly, recent data on latitic and trachytic liquids demonstrated that small compositional differences have important effects on the nucleation behavior of silicate melts (Iezzi et al., 2008). The aim of this study is to investigate the crystallization behavior of an andesitic melt under dynamic cooling conditions. Experiments were performed under variable cooling rates and final quenching temperatures at the oxygen fugacity of air, which is the appropriate fugacity for magmas at shallowest crustal levels (e.g. dikes; Burgisser and Scaillet, 2007) and emplacing lava flows or domes (Burkhard, 2005a, 2005b). The results allow us to (i) investigate the nucleation and growth of plagioclase and Fe–Ti oxide in the andesitic melts, (ii) obtain information on the ability of such melts to crystallize, (iii) shed light on the textural and compositional (disequilibrium) features observed in the outer portions of aphyric and degassed andesitic lavas and dikes, and (iv) constrain physical models for the emplacement of andesitic magmas.

262

G. Iezzi et al. / Chemical Geology 283 (2011) 261–273

Fig. 1. Experimental solidification conditions (Tables 1a and 1b). The star symbol represents the experiment quenched at 1300 °C. The straight lines are the five different cooling rates used (Table 1a). The open circles correspond to the solidification experiments quenched at 1100, 1000 and 900 °C (Table 1b).

2. Starting material, experimental and analytical methods A natural andesitic rock from the Island of Panarea (Aeolian Islands, Italy) was finely ground and ~ 10 g of powder was loaded in a Pt crucible and melted two times in air at 1400 °C for 200 min. The obtained starting material was then analyzed by X-ray powder diffraction (XRPD) and transmission optic microscope observations, which did not reveal the presence of any crystalline phases. A Deltech DT-31 vertical rapid-quench furnace was used for the preparation of the starting glass and for all the high-T solidification experiments; the temperature was monitored by two thermocouples (Pt87Rh13-Pt, type R) close to (~5 mm) the experimental charges. The samples were placed at the center of the furnace into the 2 cm3 uniform hot-zone having a thermal gradient b5 °C. The heating rate used for each experiment is reported in Fig. 1. At the beginning of cooling, the starting glassy material was held at 1400 °C for 40 min (Fig. 1) which is 234 °C above the liquidus temperature (1166 °C) indicated by MELTS (Ghiorso and Sack, 1995). The high degree of superheating was chosen

to favor the dissolution of submicrometric crystals, equilibrate and homogenize the melt and remove as much as possible gas bubbles. Successively, temperature was lowered to 1300 °C in 5 min. Below 1300 °C, the following conditions were applied (Tables 1a and 1b; Fig. 1): (a) five different cooling rates (25, 12.5, 3, 0.5 and 0.125 °C/min) between 1300 and 800 °C (final quenching temperature) and (b) a fixed cooling rate of 0.5 °C/min with variable quenching temperatures (1100, 1000, 900 and 800 °C). Set (a) of experiments was designed to investigate the role of cooling rate (Table 1a), whereas set (b) was performed to follow the crystallization path of the andesitic melt (Table 1b and Fig. 1). A further experiment was run at 1400 °C (dwell time of 40 min), cooled in 5 min to 1300 °C, and then quenched. The analysis of this run-product by electronic microscopy and by XRPD did not reveal crystals; only few bubbles were observed (≤ 1 vol.%). Thus, we can assume that all the kinetic experiments performed below 1300 °C, sets (a) and (b), initiated from a homogeneous and crystal-free andesitic melt. The composition of this melt has been considered the initial one. In wt.%, the melt composition is: SiO2 = 58.52(± 0.7), TiO2 = 0.59(± 0.06), Al2O3 = 17.24(± 0.49), Fe2O3 = 7.67(± 0.35), MnO = 0.18(± 0.04), MgO = 4.14(± 0.17), CaO = 7.73(± 0.33), Na2O = 2.46(± 0.07), K2O = 1.97(± 0.01), P2O5 = 0.15(± 0.02). Experiments were conducted using both Pt-wires (loops 1 mm in diameter) and Pt-capsules (internal diameter of 3 mm). Runs were carried out simultaneously or separately by applying identical T and t conditions (Tables 1a and 1b). This was necessary to evaluate possible effects of the sample container on the nucleation behavior, and to verify the reproducibility of the observed solidification path (Tables 1a and 1b). Each experimental charge contained ~ 100 mg of the starting glass. Quenching was achieved by dropping the sample into a water bath at room temperature with a quenching rate N 300 °C s− 1. Recovered run-products were mounted in epoxy and then ground and polished to expose both their internal and external parts, specially the contact between the Pt-container and silicate material. They were analyzed with an Electronic Probe Micro Analysis (EPMA) and a Field Emission Gun-Scanning Electron Microscopy (SEM) both installed at Istituto Nazionale di Geofisica e Vulcanologia, Roma (Italy). EPMA is a Jeol-JXA8200 combined EDS-WDS (five spectrometers with twelve crystals) and SEM is a Jeol-JSM6500F equipped with an EDS detector.

Table 1a Solidification experiments under variable cooling rates (before quenching) and fixed thermal range. The number reported in parentheses refers to the standard deviation. Thermal range (°C)

1300-800-quenching

Run label

BC-DPP-0.125

BW-DPP-0.125

AC-DPP-0.5

AW-DPP-3

BC-DPP-3

BW-DPP-3

Sample container

Pt-capsule

Pt-wire

Pt-capsule

Pt-wire

Pt-capsule

Pt-wire

Cooling rate (°C/min–°C/h) Time duration (minute–hour) Crystal content (vol.%)

0.125–7.5 4000–66.7 80 (3)

Run label

AW-DPP-12.5

BC-DPP-12.5

BW-DPP-12.5

AW-DPP-25

BC-DPP-25

BW-DPP-25

Sample container

Pt-wire

Pt-capsule

Pt-wire

Pt-wire

Pt-capsule

Pt-wire

Cooling rate (°C/min–°C/h) Time duration (minute–hour) Crystal content (vol.%)

12.5–750 40–0.6 20 (3)

0.5–30 1000–16.7 78 (3)

3–180 167–2.8 28 (3)

25–1500 20–0.3 10 (2)

Table 1b Solidification experiments under variable quenching temperatures and fixed cooling rate to 0.5 °C/min. The number reported in parentheses refers to the standard deviation. Run label

DC-DPP-0.5

DW-DPP-0.5

Sample container

Pt-capsule

Pt-wire

Thermal range (°C) Time duration (minute–hour) Crystal content (vol.%)

1300–1100 400–6.67 13 (1)

EC-DPP-0.5

FC-DPP-0.5

FW-DPP-0.5

Pt-capsule

Pt-capsule

Pt-wire

1300–1000 600–10 72 (3)

1300–900 800–13.34 73 (3)

G. Iezzi et al. / Chemical Geology 283 (2011) 261–273

263

EPMA allowed us to determine the chemical composition of the runproducts over an area of more than some μm2. The analyses were performed using an accelerating voltage of 15 kV and an electric current of 10 nA. For glasses, a slightly defocused electronic beam was used with a counting time of 5 s on background and 15 s on peak. For crystals, the beam size was 1–2 μm with a counting time of 20 and 10 s on peaks and background, respectively. The following standards were used: jadeite (Si and Na), corundum (Al), forsterite (Mg), andradite (Fe), rutile (Ti), orthoclase (K), barite (Ba), celestine (S), fluorite (F), apatite (P and Cl) and spessartine (Mn). Sodium and potassium were analyzed first to accurately determine their amounts. The crystal analyses were used only if compositions are 0.96 b Ca + Na + K b 1.04 and 3.96 b Si + Al b 4.04 a.p.f.u. (calculated on 8 oxygens) for plagioclases, and 2.97 b Fe + Mg + Al + Ti b 3.03 a.p.f.u. (calculated on 4 oxygens) for oxides. SEM images were collected using the back-scattered electron (BSE) mode at 80 to 10,000 magnifications. The accuracy of EDS chemical analysis has been checked on the same chemical standards used for EPMA; differences in SiO2, Al2O3, Fe2O3, MgO, CaO, Na2O and K2O contents, obtained by EPMA-WDS and SEM-EDS were less than 10% relative. Image analysis was performed using the commercial software Image Pro Plus. Two examples of the procedure adopted to obtain phase segmentations and relative binary images of single phases (glass and plagioclase) are reported in Fig. 2 for crystal-poor and crystal-rich run-products. Plagioclase crystals frequently impinged on each other limiting the identification of a single crystal; following Pupier et al. (2007), white digital lines were drawn between impinged plagioclases allowing their identification. In the final binary images, plagioclase crystals were automatically counted to obtain their shortest and longest axes, corresponding to the best-fitting ellipses of the bi-dimensional crystal sections (Higgins 2006). The crystal size distribution (CSD) data for plagioclase populations were calculated with the program CSD-Corrections 1.38, that also accounts for stereological corrections, using the 2D longest dimensions of plagioclases (Higgins 2000, 2006); the crystal habit of plagioclase were fixed to 1:3:7 (intermediate between the shapes of large and small plagioclase crystals, see below); each CSD plot was constructed considering about 1000–1500 crystals, discarding those with lengths b 3 μm.

temperatures (1000, 900 and 800 °C), the crystal content varies from ~ 70 to ~ 80 vol.% (Fig. 3b). Crystals of plagioclase are intergrown in all the run-products (Fig. 3); crystal impingement, agglomeration and coalescence are also observed, mostly at lower cooling rates (0.5 and 0.125 °C/min) (Fig. 3a). The size of plagioclase varies with cooling rates: (i) at higher cooling rates (25 and 12.5 °C/min) (Fig. 3a) and quenching temperature (1100 °C) (Fig. 3b), crystals reach a maximum length of ~ 0.2 mm; (ii) at lower cooling rates (0.5 and 0.125 °C/min) (Fig. 3a) and quenching temperatures (1000, 900 and 800 °C) (Fig. 3b), the maximum crystal length is ~ 0.5 mm; (iii) at 3 °C/min (Fig. 3a), the value is ~ 0.35 mm. Larger crystals (length N 20 μm) show equant (2:3) or prismatic (2:5) shapes, whereas smaller crystals (length b 20 μm) are acicular (1:10); the latter shape frequently occurs in pools of glass surrounded by larger crystals, especially at higher cooling rates (Fig. 3a) and quenching temperature of 1100 °C (Fig. 3b). Taking in account these observations, we were forced to fix the plagioclase crystal shape to 1:3:7 (see before) for all the CSD plots; however, checks performed by using other crystal shape (1:2:4 and 1:1:10) did not significantly affect the CSD results. Some plagioclases with size of hundreds of μm show irregular and, to a lesser extent, smoothed and rounded contours (Fig. 3). Tiny crystals frequently mantle the larger ones. Additionally, they do not dispose isotropically around the larger crystals but arrange along preferential directions, i.e. roughly normal or parallel to the edges of large plagioclases. Such spatial arrangements produce crystal intergrowths with regular geometries (Fig. 3). Plagioclase develops nearly continuous crystal networks at lower cooling rates (0.5 and 0.125 °C/min) (Fig. 3a) and temperatures (1000, 900 and 800 °C) (Fig. 3b). Micrometric Fe–Ti oxides have acicular shapes, whereas larger crystals (maximum size of 300 μm) show equant habits (Fig. 3). Their crystal size increases as the cooling rate and quenching temperature decrease. Generally, larger crystals result from the attachment of smaller and equant ones (Fig. 3b). Clinopyroxenes show always acicular shapes and dendritic textures. Their size is of the order of few microns. Notably, they invariably occur in pools of glass surrounded by large plagioclase crystals (Fig. 3).

3. Results

The anorthite (An) content of plagioclase decreases as the cooling rate and quenching temperature decrease (Table 2 and Fig. 4). At 25 and 12.5 °C/min, crystal compositions vary from An60 to An90 and cluster in the interval of An80–90. At 0.5 and 0.125 °C/min, they are in the range of An55–75. At the intermediate cooling rate of 3 °C/min, crystal compositions cover those observed at both higher and lower cooling rates forming two clustered ranges of An85–90 and An60–70. At the fixed cooling rate of 0.5 °C/min, plagioclase compositions are in the range of An65–85, An60–80 and An55–80 for quenching temperatures of 1100, 1000 and 900 °C, respectively (Fig. 4). The oxide compositions are plotted in Fig. 5. They are mainly close to the magnetite and/or Ti-magnetite end-members. However, at higher cooling rates (25 and 12.5 °C) and quenching temperature (1100 °C) some crystals are solid solutions with the spinel s.s. endmember. Clinopyroxenes were not analyzable by EPMA-WDS because of their small size. However, SEM-EDS analyses reveal compositions ranging from pigeonite to augite. Glass compositions show large variations as a function of the glass proximity (or not) to crystalline phases (Tables 3a and 3b; Fig. 6). Glasses from run-products obtained at lower cooling rates (0.5 and 0.125 °C/min) and quenching temperatures (900 and 800 °C) are always embedded between large plagioclase crystals. At cooling rates of 0.5 and 0.125 °C/min and quenching temperature of 800 °C, glasses are characterized by a MgO content lower than that of the starting

Run-products show the following general features: (i) duplicated experiments performed in both Pt-capsule and Pt-wire (Tables 1a and 1b) have the same textural and compositional features, (ii) iron and alkali loss have never been detected, (iii) preferential crystallization on Pt-walls of the sample container or near bubble surfaces have not been observed (Fig. 3). Notably, bubbles are frequently reported in crystallization studies performed at atmospheric pressure (Pupier et al. 2007); they nucleate by exsolution of low water amounts retained in nominally anhydrous silicate glasses in response to successive crystallization events (Castro et al., 2008). 3.1. Textural features Textural features of run-products are displayed in Fig. 3. Glasses, plagioclases, and Fe–Ti oxides are always present, whereas clinopyroxenes appear only at lower cooling rates (0.5 and 0.125 °C/min) (Fig. 3a) and quenching temperatures (900 and 800 °C) (Fig. 3b). Plagioclase is the most abundant crystalline phase. The crystal content of run-products varies as follows (Tables 1a and 1b): (i) run-products solidified at 25, 12.5, and 3 °C/min have a crystal content ≤ 28 vol.% (Fig. 3a), (ii) at 0.5 and 0.125 °C/min, the crystal content reaches a maximum value of ~ 80 vol.% (Fig. 3a); (iii) run-products quenched at 1100 °C have ~ 13 vol.% of crystals (Fig. 3b); and (iv) at lower

3.2. Phase compositions

264

G. Iezzi et al. / Chemical Geology 283 (2011) 261–273

G. Iezzi et al. / Chemical Geology 283 (2011) 261–273

265

andesitic composition, reflecting the crystallization of the clinopyroxene at temperatures b 900 °C (Tables 3a and 3b; Fig. 6). The progressive increase of the crystal content with decreasing cooling rate and quenching temperature produces glasses characterized by lower CaO and Al2O3 and higher SiO2 and K2O amounts (Tables 3a and 3b; Fig. 7).

1997). On the basis of the above considerations, our results indicate that φ is very close to 1 for bubble and Pt substrata for this andesitic melt.

4. Discussion

Textures of plagioclase from our experiments are similar to those described for basic melts solidified under experimental (Lofgren, 1980, 1983; Dunbar et al., 1993, 1995; Burkhard, 2005a; Hammer, 2006; Walton and Herd, 2007 and references therein) and natural (Phillpotts et al., 1998) conditions. At higher cooling rates (≥ 3 °C/min) and quenching temperature of 1100 °C, the initial nucleation event and successive growth of plagioclase (and Fe–Ti oxide) are more easily observable because the crystallization is at an incipient stage (Figs. 3 and 7a,b,c,d). The crystallization sequence can be tracked as follows: (i) after the homogeneous onset of nucleation, an additional single crystal nucleate and growth along the most energetically favored crystallographic direction [100] (Kostov and Kostov, 1999; Deer et al., 2001), (ii) the next one heterogeneously grow normally to the pre-existing crystal forming a L-shaped crystal network (Fig. 7a), (iii) U-, rectangular- and sieve-shaped plagioclase networks (Fig. 7a,b,c,d) result from successive and repeated heterogeneous growth events at step (ii); (iv) large plagioclases grew by agglomeration mechanisms or by single crystal growth (Fig. 7c,d); and (v) finally, smaller and elongated plagioclase crystals preferentially nucleate and, in part, agglomerate on the rims of larger ones (Fig. 7c,d). At the lower cooling rates (0.5 and 0.125 °C/min) and quenching temperatures (1000, 900 and 800 °C), the amount of crystals is high (Tables 1a and 1b). Plagioclases show evident agglomeration or coalescence textures, as well as heterogeneous nucleation features. The more elongated crystals agglomerate preferentially parallel to the [100] crystallographic direction (the longest dimension) forming continuous and self-impinging large crystal networks (Fig. 7e,f). A successive crystallization of tiny Fe–Ti oxides, clinopyroxenes, and plagioclases occurs within the residual melt among the larger plagioclases. Tiny crystals show dendritic textures and form crystal chains disposed along crystallographic directions (epitaxial growth) bridging the edges of larger plagioclases (Figs. 3 and 7e,f). This heterogeneous growth mechanism has been recently demonstrated to be very common for silicate melts solidified in laboratory or for lavas (Hammer et al., 2010). The plagioclase crystallization process has been quantified through CSD analysis (Marsh, 1998; Higgins, 2000 and 2006; Zieg and Marsh, 2002). CSD has been limited to plagioclase because of the low number of measurable Fe–Ti oxides and small size of clinopyroxenes. CSDs of plagioclase crystallized under variable cooling rates and quenching temperatures are reported in Fig. 8. For a crystal length b 10 μm, the CSDs overlap with the exception of the curve at 3 °C/min (Fig. 8). CSDs have different slopes at lengths between 10 and 40 μm (Fig. 8). At 3 °C/min, the CSD curve displays three different almost log-linear slopes and a downturn at lengths b 10 μm (Fig. 8). A similar evolution is also shown by the CSD curve for the run-product quenched at 1000 °C. At cooling rates of 0.5 and 0.125 °C/min and quenching temperature of 900 °C, CSDs show slopes with a high value for lengths b 10 μm and a low value for lengths N 10 μm (Fig. 8). In light of the CSD theory and as evidenced by plagioclase textures (Figs. 3a,b and 7), our results indicate that as both the cooling rate and quenching temperature decrease, the crystallization process progressively changes from a continuous growth of a single nucleus to a “crystal growth by coarsening”. As observed by CSD studies on natural volcanic rocks (e.g., Higgins, 2006; Hammer, 2008), the crystallization

4.1. Initial nucleation behavior It has been frequently reported that variable superliquidus heat treatments and long experimental time durations above the liquidus temperature reduce the nucleation rate and increase the incubation time of the first crystallizing phase (Tsuchiyama, 1983; Pupier et al., 2007; Hammer, 2008). These effects can be caused by (i) the kinetic rearrangement of the melt structural configuration (short-range order) during cooling as a function of superliquidus temperature (Hammer, 2008) and (ii) the presence of microscopic gas bubbles or solid-particles persisting above the liquidus temperature (De Benedetti, 1995; Davies and Ihinger, 1998; Pupier et al., 2007). The latter explanation seems more convincing because relaxation kinetics of the melt is extremely rapid in time (from milli- to micro-seconds) at superliquidus temperatures (Richet, 2002; Webb, 2005; Dingwell, 2006). Accordingly, high and/or prolonged superliquidus heat treatments affect successive nucleation events by reducing the amount of bubbles and possible solid particles. Basaltic and depolymerised melts nucleate more vigorously and in a shorter time with respect to more evolved and polymerised silicate liquids. The former are also more inclined to nucleate heterogeneously on different foreign substrates, as demonstrated by the high number of crystals on the boundary of sample holders or around bubbles (Berkebile and Dowty, 1982; Tsuchiyama, 1983; Davis and Ihinger, 1998; Pupier et al., 2007). Our experiments show no evidence for preferential nucleation processes, because the crystals are homogeneously distributed in the inner, intermediate, and outer portions of sample holders. Additionally, textures of the run-products are the same for experiments duplicated under identical conditions or performed by using different sample holders (Fig. 3). The random occurrence of isolated crystal-rich patches surrounded by glass (Fig. 3) at the higher cooling rates (25, 12.5, and 3 °C/min) and quenching temperature (1100 °C) indicate that crystals start to nucleate homogeneously. Thus, the first nucleation event inside the andesitic melt is homogeneous and the solidification path (s) can be entirely ascribed to the intrinsic crystallization properties of the melt and to the applied cooling conditions (Tables 1a and 1b; Fig. 1). This initial homogeneous nucleation behavior is similar to that observed for anhydrous latitic and trachytic melts (Iezzi et al. 2008). It also extends results from previous experimental studies on water saturated silicic compositions (SiO2 N 70 wt.%) indicating the lack of preferential crystallization on gas bubbles or sample containers (Swanson, 1977; Fenn, 1977; Naney and Swanson, 1980; Couch, 2003; Martel and Schmidt, 2003). The work of critical cluster formation (Wc) in crystallizing systems is expressed by (Schmelzer, 2003; Fokin et al., 2006): het

Wc

hom

= Wc

×φ

ð1Þ

where the superscripts hom and het refer to homogeneous and heterogeneous, and φ is a parameter ranging from 0 to 1. This parameter accounts for the substratum chemistry, the composition of the melt, and the crystal-chemical similarity between the substratum (bubble and Pt) and the heterogeneous nuclei (Lofgren 1983; Lasaga

4.2. Crystallization of plagioclase

Fig. 2. Examples of image analysis performed on highly (left column) and poorly crystalline (right column) run-products. From top to bottom: untreated BSE-SEM images, gray level variations for glass (blue) and plagioclase (red), phase segmentations after tuning contrast and brightness, binary images for plagioclase and glass. Light gray levels correspond to clinopyroxene and Fe–Ti oxide.

266

G. Iezzi et al. / Chemical Geology 283 (2011) 261–273

G. Iezzi et al. / Chemical Geology 283 (2011) 261–273

267

Table 2 Representative An-rich composition of plagioclase. Thermal range (°C)

1300-800

1300-800

1300-800

1300-800

1300-800

1300-1100

1300-1000

1300-900

Cooling rate (°C/min)

25

12.5

3

0.5

0.125

0.5

0.5

0.5

SiO2 (wt.%) Al2O3 (wt.%) Fe2O3 (wt.%) CaO (wt.%) Na2O (wt.%) K2O (wt.%) Total (wt.%) An (mol.%) Ab (mol.%) Or (mol.%)

48.15 32.64 0.62 16.45 2.02 0.14 99.62 82.80 16.4 0.80

49.64 31.75 0.73 15.33 2.57 0.26 99.84 79.30 19.30 1.40

51.38 30.96 0.58 13.50 2.64 0.27 99.75 71.70 26.60 1.70

51.15 29.64 0.48 13.45 3.56 0.55 99.58 65.30 32.60 2.10

56.65 26.67 0.39 12.73 3.32 0.74 100.26 63.20 33.20 3.50

47.86 33.21 0.49 15.66 2.33 0.14 99.49 76.10 23.10 0.60

51.81 30.26 0.37 13.84 3.16 0.25 99.87 69.80 28.70 1.50

51.70 29.75 0.51 13.10 3.20 0.60 99.32 67.70 30.40 1.80

Fig. 4. Variation in anorthite content of plagioclase as a function of experimental conditions. The lines are guide for eye and pass through the representative composition of plagioclase (see Table 2). The star symbols correspond to the equilibrium composition of plagioclase (see text).

Fig. 5. Oxide compositions (wt.%) of all experiments. The two ellipses enclosed the oxide crystals with a significant amount of Al and Mg found only in solidification experiments at 25 and 12.5 °C/min between 1300 and 800 °C and at 0.5 °C/min between 1300 and 1000 °C.

is dominated by nucleation at the micrometric crystal size, whereas at larger sizes, crystal growth processes prevail. Our CSD data show that at conditions of cooling rate of 3 °C/min and quenching temperature of 1000 °C (Fig. 8), the CSD peaks at ~ 10 μm; this supports Ostwald ripening for these two solidification conditions. The increasing population density with decreasing cooling rate at crystal lengths N 10 μm, indicates that “crystal growth by coarsening” is operating (Fig. 8). Such a feature is consistent with ex-situ (Pupier et al., 2007) and in-situ (Schiavi et al., 2009) experimental observations on basaltic melts and with the textural observations on natural rocks (Higgins, 1998, 2002; Higgins and Roberge, 2003). The crystal growth by coarsening also agrees with a successive heterogeneous nucleation of plagioclase after an initial homogeneous nucleation event. The rate of coarsening decreases with increasing cooling rate or quenching temperature (Fig. 8), as it is also indicated by plagioclase textures (Fig. 3). The maximum crystal growth rate (Gmax) calculated as the ratio between the largest crystal size and time, shows that the growth rate decreases as the experimental time increases. At 25, 3, and 0.125 °C/ min, the largest crystal sizes are ~ 200, ~ 350, and ~ 500 μm, respectively. Assuming that the crystal growth starts at 1160 °C, i.e. the liquidus temperature computed by MELTS (Ghiorso and Sack, 1995), the computed crystal growth rates are 2.1 × 10− 5 (25 °C/min), 4.9 × 10− 6 (3 °C/min), and 3 × 10− 7 cm/s (0.125 °C/min). These

values resemble those calculated for plagioclase crystals grown in laboratory (Lasaga, 1997; Pupier et al., 2007) and/or in natural andesitic rocks (Stewart and Fowler, 2001). According to Zieg and Marsh (2002), the average growth rate (G) of a batch closed system can be calculated as: G = 1=m × t

ð2Þ

where m is the slope of CSD and t is the time of cooling. For the estimate of m, we used the slopes of CSD that approach as much as possible to a linear trend (Fig. 9). The calculated G values decrease from 2.4 × 10− 6 to 5.7 × 10− 7 to 6.4 × 10− 8 cm/s with decreasing cooling rate from 25 to 3 to 0.125 °C/min, respectively. Consistently, the coarsening effect becomes more important at lower cooling rates. Finally, the larger plagioclases result from preferential agglomeration normal to the [100] crystallographic direction being the larger plagioclases more equant than the smaller ones (see before). 4.3. Disequilibrium chemistry of plagioclase An important outcome from our cooling experiments concerns the composition of plagioclase. Crystals are enriched in anorthite (Fig. 4) relative to the plagioclase composition (An55) computed at the thermodynamic equilibrium at 800 °C by MELTS (Ghiorso and Sack,

Fig. 3. BSE-SEM images of the solidified DPP run-products obtained at five different cooling rates (25, 12.5, 3, 0.5 and 0.125 °C/min) between 1300 and 800 °C (a), and variable quenching temperatures (1100, 1000, 900 and 800 °C) at the fixed cooling rate of 0.5 °C/min (b). The black bars are equal to 100 μm, whereas the gray bars are equal to 10 μm.

268

G. Iezzi et al. / Chemical Geology 283 (2011) 261–273

Table 3a Average chemical composition of glass from variable cooling rates. The number reported in parentheses refers to the standard deviation. The iron is calculated all as ferric. Cooling rate

–

0.125 °C/min

0.5 °C/min

3 °C/min

Run label

Starting material

BC-DPP-0.125

AC-DPP-0.5

AW-DPP-3

BC-DPP-3

BW-DPP-3

Number of point analyses

19

13

14

15

16

15

SiO2 (wt.%) TiO2 (wt.%) Al2O3 (wt.%) Fe2O3 (wt.%) MnO (wt.%) MgO (wt.%) CaO (wt.%) Na2O (wt.%) K2O (wt.%) P2O5 (wt.%) Total (wt.%)

58.52(0.70) 0.59(0.06) 17.24(0.49) 7.67(0.35) 0.18(0.04) 4.14(0.17) 7.73(0.33) 2.46(0.07) 1.97(0.10) 0.15(0.02) 100.64(0.66)

69.24(2.38) 0.94(0.07) 15.20(1.20) 2.60(0.40) 0.07(0.05) 0.83(0.28) 4.59(0.79) 2.56(0.25) 3.55(0.63) 0.21(0.07) 99.79(0.76)

72.69(1.75) 0.24(0.03) 12.86(0.79) 2.73(0.73) 0.09(0.02) 1.40(0.35) 2.02(0.48) 2.78(0.03) 4.73(0.24) 0.04(0.03) 99.62(0.25)

59.44(0.73) 0.61(0.07) 16.15(0.07) 7.34(0.33) 0.14(0.04) 3.43(0.23) 6.99(0.27) 2.75(0.08) 2.21(0.12) 0.21(0.10) 99.36(0.47)

60.23(1.42) 0.65(0.11) 15.99(0.47) 7.48(0.84) 0.16(0.04) 4.01(0.70) 6.82(0.62) 2.26(0.22) 2.42(0.43) 0.22(0.15) 100.24(0.43)

59.83(1.17) 0.72(0.10) 15.32(0.60) 8.31(0.81) 0.17(0.04) 4.27(0.51) 7.08(0.39) 2.24(0.15) 2.41(0.35) 0.17(0.04) 100.52(0.53)

Cooling rate

12.5 °C/min

Run label

AW-DPP-12.5

BW-DPP-12.5

BC-DPP-12.5

AW-DPP-25

BC-DPP-25

BW-DPP-25

Number of point analyses

13

11

12

15

15

15

58.01(0.82) 0.65(0.06) 15.43(0.88) 8.37(1.02) 0.18(0.04) 4.61(0.88) 7.38(0.65) 2.46(0.22) 2.05(0.22) 0.19(0.08) 99.37(0.48)

58.26(1.08) 0.59(0.05) 15.35(1.06) 8.13(0.79) 0.17(0.04) 4.61(0.80) 7.24(0.82) 2.53(0.24) 2.10(0.27) 0.19(0.13) 99.19(0.48)

59.65(1.78) 0.60(0.17) 16.22(0.65) 7.64(1.21) 0.15(0.04) 3.87(0.42) 6.77(0.42) 2.66(0.15) 2.09(0.20) 0.20(0.08) 99.86(0.38)

SiO2 (wt.%) TiO2 (wt.%) Al2O3 (wt.%) Fe2O3 (wt.%) MnO (wt.%) MgO (wt.%) CaO (wt.%) Na2O (wt.%) K2O (wt.%) P2O5 (wt.%) Total (wt.%)

25 °C/min

59.38(1.18) 0.64(0.11) 16.19(0.66) 7.51(0.50) 0.16(0.04) 3.89(0.40) 6.85(0.54) 2.57(0.13) 2.07(0.20) 0.18(0.04) 99.43(0.29)

59.49(0.83) 0.60(0.10) 16.09(0.56) 7.63(0.76) 0.16(0.04) 3.98(0.37) 6.79(0.32) 2.61(0.16) 2.04(0.13) 0.16(0.03) 99.56(0.41)

58.62(1.89) 0.59(0.16) 15.58(0.98) 7.81(1.18) 0.16(0.03) 4.21(1.15) 7.39(0.73) 2.86(0.33) 1.82(0.32) 0.21(0.05) 99.27(0.51)

Table 3b Average chemical composition of glass solidified with a cooling rate of 0.5 °C/min and variable quenching temperatures. The number reported in parentheses refers to the standard deviation. The iron is calculated all as ferric. Thermal range

1300–1100 °C

1300–1100 °C

1300–1000 °C

1300–900 °C

1300–900 °C

1300–800 °C

Run label

DW-DPP-0.5

DC-DPP-0.5

EC-DPP-0.5

FC-DPP-0.5

FW-DPP-0.5

AC-DPP-0.5

Number of point analyses

10

8

8

13

11

14

SiO2 (wt.%) TiO2 (wt.%) Al2O3 (wt.%) Fe2O3 (wt.%) MnO (wt.%) MgO (wt.%) CaO (wt.%) Na2O (wt.%) K2O (wt.%) P2O5 (wt.%) Total (wt.%)

60.21(0.72) 0.60(0.05) 16.20(0.31) 7.16(0.28) 0.16(0.04) 3.85(0.32) 6.52(0.32) 2.86(0.07) 2.25(0.12) 0.19(0.04) 100.02(0.59)

60.14(0.60) 0.54(0.07) 16.46(0.24) 6.74(0.33) 0.15(0.03) 3.74(0.17) 6.43(0.07) 2.71(0.10) 2.38(0.06) 0.19(0.03) 99.46(0.45)

61.16(1.92) 0.72(0.17) 14.08(1.70) 7.32(1.17) 0.20(0.04) 4.96(1.09) 5.55(0.45) 2.33(0.25) 2.81(0.40) 0.18(0.05) 99.32(0.82)

62.02(0.99) 0.80(0.19) 13.39(2.47) 7.05(1.83) 0.24(0.01) 5.29(0.41) 5.57(0.38) 1.95(0.23) 2.64(0.15) 0.24(0.07) 99.18(0.49)

62.56(0.96) 0.64(0.09) 12.76(0.99) 8.58(0.82) 0.24(0.05) 5.63(0.46) 4.42(0.21) 2.38(0.13) 2.81(0.15) 0.23(0.06) 100.24(0.40)

72.69(1.75) 0.24(0.03) 12.86(0.79) 2.73(0.73) 0.09(0.02) 1.40(0.35) 2.02(0.48) 2.78(0.03) 4.73(0.24) 0.04(0.03) 99.62(0.25)

1995). The average crystal compositions progressively depart from that of equilibrium as the cooling rate increases or the quenching temperature decreases (Fig. 4). However, plagioclases analyzed at 3 °C/min show two compositionally distinct populations (An85 and An65) that resemble those measured at higher and lower cooling rates. This feature highlights that An-rich crystals are the first to nucleate from the melt. As a consequence of their formation, Na and Si concentrations increase into the melt leading to the formation of relatively An-poor plagioclases. The kinetic effects of cooling on crystal compositions have been scarcely experimentally investigated (Mollo et al., 2010 and 2011 and references therein). In general, metastable nucleating crystals have a lower energetic barrier to overcome and a relative shorter induction time when compared to the thermodynamically stable phases (Kirkpa-

trick, 1983; Lasaga, 1997). The kinetic control on the formation of metastable crystals is predicted qualitatively by the Ostwald step rule (Zhang, 2008). In silicate crystals and liquids, the energetic barrier of nucleation roughly scales with the number of IVSi–O and to a lesser extent of IVAl–O bonds, i.e. degree of polymerization and average bond strength. As a consequence, Si–O rich crystals can appear delayed with respect to the Si–O poor ones, also for the same NBO/T (Naney and Swanson, 1980; Kirkpatrick, 1983; Sunagawa, 1992; Iezzi et al., 2008; Mollo et al., 2011). Since An-rich plagioclases contain less Si, they are favored to nucleate although all feldspars have tetrahedral framework structures. Metastable nucleation and/or heterogeneous (disequilibrium) composition are facilitated with increasing cooling rate according to a progressive and differential reduction of ionic mobility as the glass transition region is approached (Roskosz et al., 2005, 2006; Villeneuve et

Fig. 6. Chemical composition of glasses obtained at five different cooling rates (25, 12.5, 3, 0.5 and 0.125 °C/min) between 1300 and 800 °C (left panel), and variable quenching temperatures (1100, 1000, 900 and 800 °C) at the fixed cooling rate of 0.5 °C/min (right panel). Star symbols refer to the starting melt composition.

G. Iezzi et al. / Chemical Geology 283 (2011) 261–273

269

270

G. Iezzi et al. / Chemical Geology 283 (2011) 261–273

Fig. 7. BSE-SEM images (a, b, c and d) of run-products with a low crystal content, i.e. DW-DPP-0.5, AW-DPP-25, BC-DPP-25 and AW-DPP-12.5 (scale bars are 10 μm). BSE-SEM images (e and f) of run-products with a high crystal content, i.e. BW-DPP-0.125 and FW-DPP-0.5 (scale bars are 100 μm).

al., 2008). Moreover, depolymerized melts favor a more efficient chemical transport of cations and, consequently, a shorter time for crystals to re-equilibrate with the melt (Dingwell, 2006). Notably, by comparing the crystallization behavior of andesitic (this study) and latitic melts (Iezzi et al., 2008) cooled at the same conditions (i.e., 0.5 and 0.125 °C/min) we observe that (i) the plagioclase composition is approaching that of equilibrium at 800 °C for the andesite, whereas this do not occur for the latite, and that (ii) the clinopyroxene form at the end of crystallization from the andesitic melt, in contrast it oversteps the plagioclase formation in the latite. Therefore, the chemical disequilibrium and metastable nucleation mostly affect silicic melts rather than mafic ones (Naney and Swanson, 1980; Kirkpatrick, 1983; Sharp et al., 1996; Iezzi et al., 2008). It is worth nothing that the increasing An-content in plagioclase have been addressed to the effect of cooling rate by previous experimental (Kirkpatrick, 1983; Iezzi et al., 2008; Mollo et al., 2011) and natural (Chistyakova and Latypov, 2009) studies. Additionally, our experiments evidence that also An-poor plagioclases are characterized by disequi-

librium textures. They crystallize from residual pools of silicic melts surrounded by An-rich crystal networks. An-poor plagioclases have low abundances, mostly at higher cooling rates. Textural observations also indicate that they are partially dissolved if not surrounded by a further heterogeneous nucleation of An-rich plagioclases (Fig. 10). The effect of cooling rate produces disequilibrium compositions also for Fe–Ti oxides (Fig. 5). Some oxide crystals with significant amount of Al and Mg contents are observed for the two higher cooling rates or at 0.5 °C/min quenched at 1100 °C. This compositional variation due to rapid cooling rates has been observed for basalts solidified in laboratory (Villeneuve et al., 2008) and in natural conditions (Zhou et al., 2000). 4.4. Glass forming ability of the andesitic melt The glass-forming ability (GFA) represents the ability of a silicate melt to persist in a metastable liquid state (or not) at sub-liquidus conditions during cooling (Fokin et al., 2003, 2006; Fan et al., 2007).

G. Iezzi et al. / Chemical Geology 283 (2011) 261–273

Fig. 8. CSD curves of plagioclase from experiments performed at variable cooling rate (top panel) and quenching temperature (bottom panel). The two inserts highlight the CSD plots between 0 and 90 μm.

The GFA is measured through the critical cooling rate (Rc) which is the minimum value at which a liquid can be frozen to a solid glass without forming crystals (Fan et al., 2007; Iezzi et al., 2009). The dependence of the GFA on melt composition is shown in Fig. 11, where the crystal contents of andesitic (this study), latitic, and trachytic (Iezzi et al., 2008) melts are plotted as a function of the cooling rate and quenching temperature. The andesite crystallizes in a significant shorter time and lower undercooling degree relative to latite and trachyte. This compositional feature is better represented by the

Fig. 9. CSD of plagioclase crystals and regression lines for the log-linear part of the CSD plots from experiments performed at 25, 3 and 0.125 °C/min. The crystal length ranges and the regression equations are also reported in brackets. The slope (m in Eq. (2)) of the fits are used to calculate the average crystal growth rate (G).

271

Fig. 10. BSE-SEM images of BW-DPP-12.5 (top panel) and DW-DPP-0.5 (bottom panel) run-products, cooled at 12.5 and 0.5 °C/min and quenched at 800 and 1100 °C, respectively. They show different textural and chemical features of coexisting plagioclases with An-rich and An-poor compositions. Scale bars are 100 μm (top panel) and 10 μm (bottom panel).

reduced glass transition parameter (Trg) which is the ratio between the glass transition temperature (Tg) at a viscosity of 1012 Pa s and the melting temperature (Tm) (Fokin et al., 2003; Fan et al., 2007; Iezzi et al., 2009). Notably, glass-forming liquids (silicates, metals, polymers, etc.) are significantly affected by variation in Trg and fragility (m); this latter is a measure of the sensitivity of a melt viscosity to deviate from an Arrhenian behavior (Giordano and Dingwell, 2003). On one hand, a lower Trg implies higher nucleation rate (I), a lower incubation time (τ), and a higher crystal growth (G) at the maximum of nucleation (Fokin et al., 2003, 2005). On the second hand, a lower Trg and higher m determine higher critical cooling rate and lower GFA (Fan et al., 2007). These relationships constrain the higher facility, intensity, and rapidity of mafic melts to crystallize relative to silicic ones (Iezzi et al., 2009). In Fig. 11, the linear fit between the crystal content and cooling rate of 25, 12.5, and 3 °C/min indicates that the Rc is ~ 37 °C/min for the andesite. Iezzi et al. (2008) found that the Rc is b 0.125 and ~ 1 °C/ min for the trachyte and latite, respectively. These significant changes, induced by a restricted bulk composition variation, agree with the logarithmic relationship between Rc and Trg (Iezzi et al., 2009). This demonstrates that little changes in chemical composition are able to significantly shift the initial nucleation behavior of silicate liquids and the following solidification paths. 4.5. Volcanological and petrological implications Tsuchiyama (1983) demonstrated that textural and compositional features of natural melts are comparable with those reproduced in

272

G. Iezzi et al. / Chemical Geology 283 (2011) 261–273

andesite (this study) latite (Iezzi et al., 2008) trachyte (Iezzi et al., 2008) 100

100 40

90 80

crystal content (vol. %)

90

30

37 C/min

˚

80

20

70 60

10

50

0

70 60 0

10

20

30

50

40

40

40

30

30

20

20

10

10

0

0 0

5

10

15

20

cooling rate (˚C/min)

25

800

900

1000

1100

1200

1300

solidification range (˚C)

Fig. 11. Plots of cooling rate vs crystal content (left panel) and quenching temperature vs crystal content (right panel). Data for latite and trachyte are from Iezzi et al. (2008). The insert (left panel) shows the linear fit (dotted line) of data from cooling experiments at 3, 12.5 and 25 °C/min; the critical cooling rate calculated for the andesite is ~ 37 °C/min.

laboratory at high superheating conditions. However, the spatial relevance of cooling rates used in laboratory must be considered when comparing experimental and natural systems. Several studies (Dunbar et al., 1995; Neri, 1998; Xu and Zhang, 2002; Burkhard, 2005a; Harris et al., 2005) evidenced that cooling rates decrease from tens to fractions of °C/min moving from the outermost (few decimetres) portions of dikes and lavas towards their innermost (few meters) parts. Therefore, cooling conditions used in our experiments are relevant for cooling conditions of natural silicate melts with thickness of several meters. Textural and chemical features of our experiments are similar to those observed in aphyric and thin lava flows (Tamura et al., 2003; Mattioli et al., 2006) characterized by (i) glass and crystal compositions showing a large chemical variability, (ii) Al-, Mg-rich oxides, and (iii) An-rich plagioclases coexisting with silicic residual melts. Chistyakova and Latypov (2009) studied mafic nearly aphyric dikes and found that centimeter- to meter-sized dikes show, on average, An-rich plagioclases at the outer margins, where the cooling rate is higher; conversely, An-poor crystals occur in the inner and central parts of dikes, where slower cooling rates develop, in agreement with our experiments. The andesitic lavas and dikes sampled in the Tonga arc and reported in Hekinian et al. (2008) represent the most pertinent comparison with our experimental conclusions, since these rocks were emplaced in sub-marine environments where they rapidly cooled. These andesitic rocks have thickness variable from dm to few meters. The plagioclase crystals in the groundmass of these lavas have compositions An65–85, which perfectly overlap the chemical variability observed in Fig. 4 at variable cooling rates. It is worth noting that disequilibrium features and time– temperature–spatial scales of our experiments indicate that extreme care should be adopted during the sampling of andesitic rocks for petrological and geochemical investigations. The occurrence of a wide range of textural and chemical features in the collected samples could be interpreted as a result of open magmatic systems (e.g. magma mixing) instead of crystallization under high cooling conditions. Therefore, the analysis of different portions of andesitic lavas or dikes is of fundamental importance to discriminate crystallization processes occurred in magmatic reservoirs from those developed at subaerial or submarine conditions. Finally, the crystal content of our run-products shows a sudden increase from 13 to 72 vol.% with decreasing cooling rate from 3 to

0.5 °C/min, respectively (Tables 1a and 1b; Fig. 11). This implies that the viscosity of an andesitic magma will be strongly controlled by its crystal content for cooling rates b 3 °C/min. Furthermore, the crystal growth rates calculated from our experiments indicate that large crystals of plagioclase may not form only in intratelluric conditions. The impingement between these crystals will result in the attainment of the critical crystal fraction (Kerr and Lister, 1991; Costa, 2005) at temperatures between 1100 and 1000 °C (Tables 1a and 1b; Fig. 3b). Therefore, the rheology of andesitic magmas is largely controlled by the cooling experienced during their emplacement (see also Del Gaudio et al., 2010). Acknowledgements We are very grateful to the anonymous reviewers and C. Martel for their revisions and suggestions; D. Dingwell is warmly acknowledged for his editorial handling. We are also very grateful to Prof. D. Dolfi for allowing us to use experimental petrology facilities at the Università degli Studi Roma Tre (Roma). This study has been supported by several institutions and programs: (a) Università G. d'Annunzio — “Fondi Ateneo” to G. Iezzi, (b) Dipartimento della Protezione Civile — INGV in the frame of the 2004–2006 project to G. Iezzi, and (c) the FIRB — MIUR “Sviluppo Nuove Tecnologie per la Protezione e Difesa del Territorio dai Rischi Naturali” to P. Scarlato, A. Cavallo, and S. Mollo. References Berkebile, C.A., Dowty, E., 1982. Nucleation in laboratory charges of basaltic composition. American Mineralogist 67, 886–899. Burgisser, A., Scaillet, B., 2007. Redox evolution of a degassing magma rising to the surface. Nature 445, 194–197. Burkhard, D.J.M., 2005a. Nucleation and growth rates of pyroxene, plagioclase, and Fe– Ti oxides in basalt under atmospheric conditions. European Journal of Mineralogy 17, 675–685. Burkhard, D.J.M., 2005b. Crystallization and oxidation during emplacement of lava lobes: In: Manga, M., Ventura, G. (Eds.), Geological Society of America, Special Paper, 396, pp. 67–80. Cashman, K.V., 1991. Textural constraints on the kinetics of crystallization of igneous rocks: In: Nicholls, J., Russell, J.K. (Eds.), Reviews in Mineralogy, 24, pp. 259–309. Castro, J.M., Beck, P., Tuffen, H., Nichols, A.R.L., Dingwell, D.B., Martin, M.C., 2008. Timescales of spherulite crystallization in obsidian inferred from water concentration profiles. American Mineralogist 93, 1816–1822. Chistyakova, S., Latypov, R., 2009. On the development of internal chemical zonation in small mafic dikes. Geological Magazine 147, 1–12.

G. Iezzi et al. / Chemical Geology 283 (2011) 261–273 Conte, A., Perinelli, C., Trigila, R., 2006. Cooling kinetics experiments on different Stromboli lavas: effects on crystal morphologies and phase composition. Journal of Volcanology and Geothermal Research 155, 179–200. Costa, A., 2005. Viscosity of high crystal content melts: dependence on solid fraction. Geophysical Research Letters 32 (22), 1–5. Couch, S., 2003. Experimental investigation of crystallization kinetics in a haplogranite system. American Mineralogist 10, 1471–1485. Davis, M.J., Ihinger, P.D., 1998. Heterogeneous crystal nucleation on bubbles in silicate melts. American Mineralogist 83, 1008–1015. De Benedetti, P.G., 1995. Metastable Liquids, Concepts and Principles. Princeton University Press, Princeton, NY. Deer, W.A., Howie, R.A., Zussman, J., 2001. Framework Silicates: FeldsparsSecond Edition. The Geological Society, London. Del Gaudio, P., Mollo, S., Ventura, G., Iezzi, G., Taddeucci, J., Cavallo, A., 2010. Cooling rate-induced differentiation in anhydrous and hydrous basalts at 500 MPa: implications for the storage and transport of magmas in dikes. Chemical Geology 270, 164–178. Dingwell, D.B., 2006. Transport properties of magmas: diffusion and rheology. Elements 2, 281–286. Dowty, E., 1980. In: Hargraves, R.B. (Ed.), Crystal Growth and Nucleation Theory and the Numerical Simulation of Igneous Crystallisation. Princeton University Press, pp. 419–485. Dunbar, N.W., Riciputi, L.R., Jacobs, G.K., Naney, M.T., Christie, W., 1993. Generation of rhyolitic melt in an artificial magama: implications for fractional crystallization processes in natural magmas. Journal of Volcanology and Geothermal Research 57, 157–166. Dunbar, N.W., Jacobs, G.K., Naney, M.T., 1995. Crystallization processes in an artificial magma: variations in crystal shape, growth rate and composition with melt cooling history. Contribution to Mineralogy and Petrology 120, 412–425. Fan, G.J., Choo, H., Liaw, P.K., 2007. A new criterion for the glass-forming ability of liquidus. Journal of Non-Crystalline Solids 353, 102–107. Fenn, P.M., 1977. The nucleation and growth of alkali feldspars from hydrous melts. The Canadian Mineralogist 15, 135–161. Fokin, V.M., Zanotto, E.D., Schmelzer, J.W.P., 2003. Homogeneous nucleation versus glass transition temperature of silicate glasses. Journal of Non-Crystalline Solids 321, 52–65. Fokin, V.M., Nascimento, M.L.F., Zanotto, E.D., 2005. Correlation between maximum crystal growth rate and glass transition temperature of silicate glasses. Journal of Non-Crystalline Solids 351, 789–794. Fokin, V.M., Zanotto, E.D., Yuritsyn, N.S., Schmelzer, J.W.P., 2006. Homogeneous crystal nucleation in silicate glasses: a 40 years perspective. Journal of Non-Crystalline Solids 352, 2681–2714. Ghiorso, M.S., Sack, R.O., 1995. Chemical mass-transfer in magmatic processes 4. A revised and internally consistent thermodynamic model for the interpolation and extrapolation of liquidus-solid equilbria in magmatic systems at elevated temperatures and pressures. Contribution to Mineralogy and Petrology 119, 197–212. Giordano, D., Dingwell, D.B., 2003. The kinetic fragility of natural silicate melts. Journal of Physics: Condensed Matter 15, S945–S954. Hammer, J.E., 2004. Crystal nucleation in hydrous rhyolite: experimental data applied to classical theory. American Mineralogist 89, 1673–1679. Hammer, J.E., 2006. Influence of fO2 and cooling rate on the kinetics and energetics of Fe-rich basalt crystallization. Earth and Planetary Science Letters 248, 618–637. Hammer, J.E., 2008. Experimental studies of the kinetics and energetics of magma crystallization: In: Putirka, K.D., Tepley, F.J. (Eds.), Reviews in Mineralogy, 69, pp. 9–59. Hammer, J.E., Sharp, T.G., Wessel, P., 2010. Heterogeneous nucleation and epitaxial crystal growth of magmatic minerals. Geology 38, 367–370. Harris, A.J.L., Bailey, J., Calvari, S., Dehn, J., 2005. Heat loss measured at a lava channel and its implications for down-channel cooling and rheology: In: Manga, M., Ventura, G. (Eds.), Geological Society of America, Special Paper, 396, pp. 125–146. Hekinian, R., Mühe, R., Worthington, T.J., Stoffers, P., 2008. Geology of a submarine volcanic caldera in the Tonga Arc: dive results. Journal of Volcanology and Geothermal Research 176, 571–582. Higgins, M.D., 1998. Origin of anorthosite by textural coarsening: quantitative measurements of a natural sequence of textural development. Journal of Petrology 39, 1307–1323. Higgins, M.D., 2000. Measurement of crystal size distributions. American Mineralogist 85, 1105–1116. Higgins, M.D., 2002. A crystal size-distribution study of the Kiglapait layered mafic intrusion, Labrador, Canada: evidence for textural coarsening. Contribution to Mineralogy and Petrology 144, 314–330. Higgins, M.D., 2006. Quantitative Textural Measurements in Igneous and Metamorphic Petrology. Cambridge University Press, Cambridge. Higgins, M.D., Roberge, J., 2003. Crystal size distribution of plagioclase and amphibole from Soufrière Hills Volcano, Montserrat: evidence for dynamic crystallizationtextural coarsening cycles. Journal of Petrology 44, 1401–1411. Iezzi, G., Mollo, S., Ventura, G., Cavallo, A., Romano, C., 2008. Experimental solidification of anhydrous latitic and trachytic melts at different cooling rates: the role of nucleation kinetics. Chemical Geology 253, 91–101. Iezzi, G., Mollo, S., Ventura, G., 2009. Solidification behaviour of natural silicate melts and volcanological implications. In: Lewis, N., Moretti, A. (Eds.), Volcanoes: Formation, Eruptions and Modelling. Nova publishers, New York, pp. 127–151.

273

Kerr, R.C., Lister, J.R., 1991. The effects of shape on crystal settling and on the rheology of magmas. Journal of Geology 99, 457–467. Kirkpatrick, R.J., 1981. Kinetics of crystallization of igneous rocks: In: Lasaga, A.C., Kirkpatrick, R.J. (Eds.), Reviews in Mineralogy, 8, pp. 321–395. Kirkpatrick, R.J., 1983. Theory of nucleation in silicate melts. American Mineralogist 68, 66–77. Kostov, I., Kostov, R.I., 1999. Crystal Habits of Minerals. Bulgarian Academic Monographs, Sophia. Lasaga, A.C., 1997. Kinetic Theory in the Earth Sciences. Princeton University Press, Princeton, NY. Lofgren, G., 1980. In: Hargraves, R.B. (Ed.), Experimental Studies on the Dynamic Crystallisation of Silicate Melts. Princeton University Press, Princeton, pp. 487–565. Lofgren, G., 1983. Effect of heterogeneous nucleation on basaltic textures: a dynamic crystallization study. Journal of Petrology 24, 229–255. Marsh, B.D., 1998. On the interpretation of crystal size distributions in magmatic systems. Journal of Petrology 39, 553–599. Martel, C., Schmidt, B., 2003. Decompression experiments as an insight into ascent rates of silicic magmas. Contribution to Mineralogy and Petrology 144, 397–415. Mattioli, M., Renzulli, A., Menna, M., Holm, P.M., 2006. Rapid ascent and contamination of magmas through the thick crust of the CVZ (Andes, Ollague region): evidence from a nearly aphyric high-K andesite with skeletal olivines. Journal of Volcanology and Geothermal Research 158, 87–105. Mollo, S., Del Gaudio, P., Ventura, G., Iezzi, G., Scarlato, P., 2010. Dependence of clinopyroxene composition on cooling rate in basaltic magmas: implications for thermobarometry. Lithos 118, 302–312. Mollo, S., Putirka, K., Iezzi, G., Del Gaudio, P., Scarlato, P., 2011. Plagioclase-melt (dis) equilibrium due to cooling dynamics: implications for thermometry, barometry and hygrometry. Lithos. doi:10.1016/j.lithos.2011.02.008. Naney, M.T., Swanson, S.E., 1980. The effect of Fe and Mg on crystallization in granitic systems. American Mineralogist 65, 639–653. Neri, A., 1998. A local heat transfer analysis of lava cooling in the atmosphere: application to thermal diffusion-dominated lava flows. Journal of Volcanology and Geothermal Research 81, 215–243. Phillpotts, A.R., Shi, J., Brustman, C., 1998. Role of plagioclase crystal chains in the differentiation of partly crystallised magma. Nature 395, 343–346. Pupier, E., Duchene, S., Toplis, M.J., 2007. Experimental quantification of plagioclase crystal size distribution during cooling of a basaltic liquid. Contribution to Mineralogy and Petrology 155, 555–570. Richet, P., 2002. Enthalpy, volume and structural relaxation in glass-forming silicate melts. Journal of Thermal Analysis and Calorimetry 69, 739–750. Roskosz, M., Toplis, M.J., Besson, P., Richet, P., 2005. Nucleation mechanisms: a crystal chemical investigation of phases forming in highly supercooled aluminosilicate liquids. Journal of Non-Crystalline Solids 351, 1266–1282. Roskosz, M., Toplis, M.J., Richet, P., 2006. Kinetic vs. thermodynamic control of nucleation and growth in molten silicates. J Non-Cryst Solids 352, 180–184. Schiavi, F., Walte, N., Keppler, H., 2009. First in-situ observation of the crystallization processes in a basaltic–andesitic melt with the moissanite cell. Geology 37, 963–966. Schmelzer, J.W.P., 2003. Kinetic and thermodynamic theories of nucleation. Materials Physics and Mechanics 6, 21–33. Sharp, T.G., Stevenson, R.J., Dingwell, D.B., 1996. Microlites and “nanolites” in rhyolitic glass: microstructural and chemical characterisation. Bulletin of Volcanology 57, 631–640. Stewart, M.L., Fowler, A.D., 2001. The nature and occurrence of discrete zoning in plagioclase from recently erupted andesitic volcanic rocks, Montserrat. Journal of Volcanology and Geothermal Research 106, 243–253. Sunagawa, I., 1992. In situ investigation of nucleation, growth, and dissolution of silicate crystals at high temperatures. Annual Review of Earth and Planetary Sciences 20, 113–142. Swanson, S.E., 1977. Relation of nucleation and crystal-growth rate to the development of granitic textures. American Mineralogist 62, 966–977. Tamura, Y., Yuhara, M., Ishii, T., Irino, N., Shukuno, H., 2003. Andesites and dacites from Daisen Volcano, Japan: partial-to-total remelting of an andesite magma body. Journal of Petrology 12, 2243–2260. Tsuchiyama, A., 1983. Crystallization kinetics in the system CaMgSi2O6–CaAl2Si2O8: the delay in nucleation of diopside and anorthite. American Mineralogist 68, 687–698. Villeneuve, N., Neuville, D.R., Boivin, P., Bachèlery, P., Richet, P., 2008. Magma crystallization and viscosity: a study of molten basalts from the Piton de la Fournaise volcano (La Réunion island). Chemical Geology 256, 241–250. Walton, E.L., Herd, C.D.K., 2007. Dynamic crystallization of shock melts in Allan hills 77005: implications for melt pocket formation in Martian meteorites. Geochimica et Cosmochimica Acta 71, 5267–5285. Webb, S.L., 2005. Silicate melts at extreme conditions: In: Miletich, R. (Ed.), EMU Notes in Mineralogy, 7, pp. 64–95. Xu, Z., Zhang, Y., 2002. Quench rates in water, air and liquid nitrogen and inference of temperature in volcanic eruption columns. Earth and Planetary Science Letters 200, 315–330. Zhang, Y., 2008. Geochemical Kinetic. Princeton University Press, Princeton, NY. Zhou, W., Van der Voo, R., Peacor, D.R., Zhang, Y., 2000. Variable Ti-content and grain size of titanomagnetite as a function of cooling rate in very young MORB. Earth and Planetary Science Letters 179, 9–20. Zieg, M.J., Marsh, B.D., 2002. Crystal size distribution and scaling laws in the quantification of igneous textures. Journal of Petrology 43, 85–101.