A total of eight isothermal melting experiments were conducted to bracket phase appearance/disappearance temperatures at 1 atm. (see insert, Fig. 6; Table 2). The liquidus is bracketed at 1184 ± 2ºC by the appearance of olivine in the temperature interval between the two highest temperature runs. Similarly, plagioclase appears at 1177ºC with an uncertainty of ± 5ºC, and augite appears at 1167 ± 5ºC. Thus, the melt under equilibrium conditions is multiply saturated with olivine, plagioclase, and augite at 17ºC below the liquidus. Pigeonite is the last phase to appear at 1113 ± 12ºC. Fe-Ti oxides are present in low abundance in all runs quenched below 1050ºC and are found in a number of higher temperature experiments. The solidus is not directly constrained by these experiments, but is estimated at 1076ºC, based on a linear extrapolation of temperature vs. melt fraction (Table 2).

The results of 23 controlled cooling experiments are presented in Table 3. In every case, as discussed above, the sample was initially conditioned for ~1 hr at ~10ºC above the liquidus and then cooled at a specified rate to the quench temperature. As has been shown in previous studies of dynamic crystallization (Lofgren, 1980), the resulting textures are extremely sensitive to the experimental parameters, such as whether the sample remained subliquidus or was superheated before cooling, the dwell temperature and time, and the quench temperature. The lack of phenocrysts in the quench zones of Unit 1 suggests that the lava was probably moderately superheated on eruption. This inference, in part, lead to our choice of a dwell temperature slightly above the liquidus. The dwell time was chosen to ensure complete fusion and chemical homogenization of the sample and for convenience. By selecting these experimental parameters, however, we make no a priori judgment regarding the conditions that governed the solidification of Unit 1. Rather, the insights gained from the experiments give merit to the interpretations advanced for development of the textures of the lavas.

Figure 6 shows the results of the controlled cooling experiments as a function of cooling rate and temperature of quenching. The solid curves delimit the first appearance of a particular phase on cooling, (e.g., olivine-in). These phase-appearance curves can be compared with the equilibrium crystallization sequence in Figure 6. The most striking result is the dramatic suppression of olivine crystallization at moderate to high cooling rates. For example, at 10ºC/hr olivine first appears 30ºC below its equilibrium appearance temperature, while at 100ºC/hr olivine crystallization is suppressed by more than 190ºC, and perhaps entirely. In contrast, the delay in the plagioclase nucleation is suppressed by only 7º-10ºC relative to the equilibrium appearance temperature at a cooling rate of 10ºC/hr, increasing to ~30ºC at 100ºC/hr, and is ~80ºC at 1000ºC/hr. Suppression of augite crystallization is slightly greater than found for plagioclase at moderate cooling rates, but is similar at the highest cooling rates examined (Fig. 6). Pigeonite, which appears at ~1113ºC under equilibrium conditions, is not observed in any of the controlled cooling run products. From the visually estimated modes (Table 3), glass was only a minor constituent below a quench temperature of ~1100ºC at a cooling rate of 10ºC/hr and below ~1050º-1030ºC at higher cooling rates.

The experiments show two broadly classified types of textural changes as both cooling rate and quench temperature are varied. The first involves an increase in the temperature range of crystallization with increasing cooling rate (Fig. 6), while the second shows a decrease in crystal size with increasing cooling rate. While both of these cooling-related textural characteristics have been reported for other basalt cooling-rate experiments (e.g., Walker et al., 1976, 1978; Grove and Walker, 1977; Grove and Raudsepp, 1978; Grove and Beaty, 1980), our results differ from these studies in the dominant role played by plagioclase, which appears as the liquidus phase for the cooling-rate experiments (although olivine is the sole liquidus phase in the equilibrium experiments, as noted above).

Changes in the overall textural characteristics of the experimental charges are best seen in the whole-charge images shown in Figure 7. The images are arranged from lowest cooling rate (10ºC/hr) to highest cooling rate (1000ºC/hr) from left to right, and from highest quench temperature (1150ºC) to lowest quench temperature (1000ºC) from top to bottom. As shown in Figure 6, crystallization is delayed (i.e., occurs at a temperature lower than the equilibrium appearance temperature) in all of the controlled cooling experiments. However, the kinetic delay (t), or time lag in passing from the equilibrium appearance temperature to the actual temperature of appearance, varies among the phases and with cooling rate. For example, t for plagioclase, estimated from Figure 6, is 290 s for a cooling rate of 1000ºC/hr, 1080 s for a cooling rate of 100ºC/hr, 1580 s for a cooling rate of 50ºC/hr, and 3600 s for a cooling rate of 10ºC/hr. For olivine, t is ~12,000 s at 50ºC/hr and ~9650 s at 10ºC/hr. The results for plagioclase are in good agreement with estimates of t made by Nabelek et al. (1978) for a high alumina basalt composition. These authors interpreted the increase in kinetic delay with decreasing cooling rate as reflecting a decrease in the driving force for plagioclase nucleation as the degree of undercooling is reduced. The t-cooling rate dependence for olivine implies the opposite effect.

At cooling rates >10ºC/hr, plagioclase is the first phase to crystallize. Plagioclase microlites form initially as isolated crystals, sometimes, but not always, located around the margin of the experimental charge (e.g., Fig. 7B, Fig. 7H). Early plagioclase crystals typically appear acicular in two dimensions (Fig. 8), suggesting that their three-dimensional form is actually that of thin tablets (Higgins, 1994). This interpretation is supported by the occasional appearance of equant to slightly or wholly skeletal tablets in thin section (Fig. 8). Early appearance of very thin tablet-shaped plagioclase crystals is common in more silicic magmas (e.g., Muncill and Lasaga, 1988; Hammer et al., 1999) and is also reported in other experiments on basaltic compositions, where plagioclase is a near-liquidus phase (Grove, 1978; Lofgren, 1980). It appears that this morphology is preferred for nucleation, although energetic constraints lead to more equant forms as crystallization progresses (Hammer et al., 1999).

Further cooling produces a second plagioclase texture, that of dendritic crystallization fronts that nucleate at a few points and radiate across the experimental charge (e.g., Fig. 7F). Dendritic crystal growth is dominated by plagioclase, which forms plumose crystals (Fig. 9A). At low to moderate cooling rates, plagioclase and pyroxene crystallize together (Fig. 9B) to form an intrafasciculate (bundled) texture (e.g., Grove, 1978; Lofgren et al., 1978; Walker et al., 1978). At very high cooling rates, pyroxene may be the dendrite-forming phase, although the crystallization front appears spherulitic and individual phases are impossible to distinguish (Fig. 9C). As shown above for the initiation of crystallization, time scales for solidification (i.e., crystallization in excess of ~90 vol%) also increase with decreasing rates of cooling. For example, dendritic crystal growth results in solidification in ~720 s at cooling rates of 1000ºC/hr, in >3600 s and <5400 s for cooling at 100ºC/hr, in >7200 s and <10,800 s in the 50ºC/hr experiments, and in >18,000 s and <36,000 s in the 10ºC/hr experiments. Thus, both the solidification rate and the minimum time scale required for glass formation are dependent on the cooling rate (e.g., Uhlmann et al., 1979; 1981).

Both the size of solitary plagioclase laths and the coarseness of dendritic intergrowths are found to be strongly affected by the rate of cooling (e.g., Walker et al., 1976; 1978; Grove and Walker, 1977; Grove and Bence, 1977; Grove, 1978; Lofgren et al., 1979; Schiffman and Lofgren, 1982; Cashman, 1993). Individual plagioclase widths were measured as the average of the minimum dimension (measured between swallowtail extensions that form perpendicular to [010]; e.g., Grove and Walker, 1977). At a single cooling rate, plagioclase size increases with time (Fig. 10), a trend best exemplified by the 50ºC/hr sequence, where a linear fit to the data yields a plagioclase growth rate of 1.6 × 10^-7 cm/s. Similarly, Grove and Raudsepp (1978) estimated growth rates for plagioclase of 1-10 × 10^-7 cm/s for cooling rates of 1.7º-150ºC/hr. An exception to this trend lies in the 10ºC/hr cooling sequence, where the experiment with the lowest quench temperature (longest crystallization time) has a more finely crystalline texture than experiments quenched at 1054ºC and 1102ºC, for reasons possibly related to small differences in melt preconditioning before cooling (Lofgren et al., 1978; Nabelek et al, 1978; Walker et al., 1978).

Plagioclase dendrite coarseness varies both within and between experimental samples (Fig. 7, Fig. 11). Within individual samples, progressive advance of dendritic growth fronts results in a gradation in dendrite size, with early-formed dendrites in a given sample consistently coarser than later-formed dendrites in the same sample (e.g., Fig. 11C). Between samples, maximum dendrite thickness increases with decreasing rate of cooling (Fig. 11). A semi-quantitative analysis of dendrite coarsening as a function of cooling rate is provided by comparing the maximum width (earliest dendrite formation) of individual dendrites among the run products (Grove and Walker, 1977; Grove, 1978; Schiffman and Lofgren, 1982). Figure 12 shows the results of such measurements made on BSE images for experiments quenched at 1000ºC. The linear log-log relationship between plagioclase width and cooling rate for these experiments is similar to that reported by Grove and Walker (1977) for lunar basalt.

Only samples cooled at the slower (10º-50ºC/hr) rates showed development of textures other than dendritic. Experiment #12 (10ºC/hr quenched at 1054ºC) is the only charge with an intergranular texture (Fig. 7C). Here, crystals of both augite and plagioclase are well-formed and similar in size, with the number of plagioclase microlites per unit area ~10⁴/mm², comparable to plagioclase number densities of rapidly cooled (10º-20ºC/hr) Hawaiian aa flows (Crisp et al., 1994; Katz, 1997). Such nucleation site densities are characteristic of the textures generated by flow of lava through open channels. Experiment #41 (50ºC/hr, quenched at 1050ºC) is the only charge exhibiting textures approaching those of the intersertal flows of Unit 1, where radiating intergrowths of plagioclase and pyroxene form between solitary laths and are separated by ~10%-20% matrix glass (e.g., compare Fig. 13 with Fig. 5A).