One of the best-known outcomes of controlled cooling experiments on basalts of lunar compositions is the suppression of the temperature of plagioclase appearance with increasing rates of cooling (e.g., Walker et al., 1976; Grove and Walker, 1977; Grove and Raudsepp, 1978; Grove and Beaty, 1980), a measure of the kinetic delay (incubation time) required for plagioclase nucleation. While suppression of phase appearance is a general phenomenon for all phases, the cooling rate effect on plagioclase was found to be most dramatic, particularly in circumstances where the equilibrium temperature of plagioclase appearance was well below that of the liquidus. However, even in this situation, plagioclase suppression is not as extreme when the temperature at which cooling is initiated is below the liquidus, even if this temperature is still well in excess of that of plagioclase appearance (Bianco and Taylor, 1977). When plagioclase is a liquidus or near-liquidus phase, extreme plagioclase suppression is not seen until cooling rates are very high (600ºC/hr) (Grove, 1978), and early plagioclase crystallization is in the form of solitary crystal laths (Grove, 1978; Nabelek et al., 1978; Grove and Bence, 1977; Schiffman and Lofgren, 1982). The results of our experiments fit nicely into this framework.

As noted above, under equilibrium conditions Unit 1 liquid is multiply saturated with olivine, pyroxene, and plagioclase within 17ºC of the liquidus. Moreover, plagioclase nucleates easily relative to pyroxene, and forms initially as tabular (acicular in thin section) laths. Soon after plagioclase nucleation, crystallization proceeds rapidly through the propagation of plagioclase + pyroxene dendrites from a small number of nucleation sites. Where discernible, pyroxene appears to nucleate on the plagioclase (with the possible exception of the very high cooling-rate experiments, where the reverse may be true, e.g., Fig. 11D). Progressive dendritic crystallization in a single experimental sample results in inhomogeneous textures, with later-formed dendrites being more finely crystalline. These intrafasciculate textures are common features of crystallization from multiply saturated liquids that lack abundant pre-existing nucleation sites (Grove, 1978; Lofgren et al., 1978; Walker et al., 1978).

One important feature of our experiments that runs counter to previous controlled cooling experiments, and important in the context of interpreting the cooling history of Unit 1, is the apparent difficulty of olivine nucleation at cooling rates >10ºC/hr. In Grove's (1978) dynamic crystallization study of Luna 24 ferrobasalt, the suppression of crystallization of olivine was modest (25º-60ºC) at cooling rates of 10º-500ºC/hr and was only slightly higher for plagioclase. Grove (1978) initiated cooling after superheating the sample at 15ºC above the liquidus for 20 to 40 min, not substantially different from the run conditions used in the present study. However, for Luna 24, plagioclase was the equilibrium liquidus phase, while olivine appeared ~15ºC below the liquidus. In the work of Grove and Raudsepp (1978) on quartz normative lunar basalt (QNB 15597), olivine and pyroxene both appear early in the equilibrium crystallization sequence and show little delay in nucleation up to cooling rates of 150ºC/hr, while plagioclase appearing late in the crystallization sequence required strong undercooling (150ºC) before nucleation.

Although it is difficult to compare experiments on lunar compositions directly with the present work, despite similarities in experimental method, the differences in the relative suppression of olivine and plagioclase crystallization are significant. Of possible importance are compositional differences between starting materials. Although the SiO₂ and Al₂O₃ contents are similar in the lunar and terrestrial samples (SiO₂ = 47.5-48.5 wt%; Al₂O₃ = 12.6-13.95 wt%), both QNB 17797 and Luna 24 are fairly evolved compositions with Mg# (atomic ratio Mg/[Mg+Fe_total]) of 0.22-0.36, and are characteristically poor in alkalis (Na₂0 + K₂0 < 0.36 wt%). Unit 1 has a Mg# of 0.51 and the sum Na₂0 + K₂0 = 2.37. In terms of normative composition, Unit 1 is composed of 48% feldspar component (or + ab + an) (Table 1), compared to <36% for the lunar compositions. These compositional differences are clearly responsible for the differences in the equilibrium phase relations, in particular the early appearance of plagioclase shown for Unit 1. We also speculate that they may have important consequences for near-liquidus melt structure. On the basis of these compositional considerations, it is expected that near liquidus melt for Unit 1 is more fully polymerized than either of the lunar compositions and, therefore, that the activation barrier for nucleation of tectosilicates (plagioclase) will be lower than for nucleation of nesosilicates (olivine). The reverse situation would presumably be true for the lunar compositions.

Although this explanation is conjectural and worthy of further study, the experimental observation that olivine crystallization is strongly suppressed at high cooling rates in terrestrial tholeiitic basalt, while plagioclase is not, has implications for the petrographic interpretation of Unit 1 lavas. For example, the microphenocryst assemblage for Unit 1 is dominated by plagioclase and augite, in subequal proportions, while the mesostasis comprises 10%-40% of the rock. Olivine was reported only in trace (<1%) amounts in 6 of the 11 thin sections described by the Shipboard Scientific Party (Duncan, Larsen, Allan, et al., 1996). However, based on equilibrium modes (Fig. 14) at melt fractions 0.4, olivine should comprise greater than 12% of the rock. It is reassuring that this expectation, based on equilibrium phase relations, is not born out by the mineralogy of Unit 1 (e.g., paucity of olivine). Rather the mineralogy is consistent with the textural evidence that the lavas solidified at moderately high cooling rates.

Two other important features of Unit 1 lavas can be addressed using the cooling rate experiments described above. First is confirmation of the origin of the finely crystalline bands so prevalent in Unit 1, and second is the origin of the intersertal (mesostasis-rich) textures evident throughout much of the flow. Both of these features, together with the exceptional flow thickness, are unusual relative to textural features described in other subaerial basalt cores from the Northeast Atlantic (Roberts, Schnitker, et al., 1984; Eldholm, Thiede, Taylor, et al., 1987; Larsen, Saunders, Cliff, et al., 1994), from the Columbia River Basalt Group (Ho and Cashman, 1997), or from Hawaii (Katz, 1997).

Bands identified as finely crystalline, moderately vesicular flow tops by the Shipboard Scientific Party (Duncan, Larsen, Allan, et al., 1996) have intrafasciculate (radiating intergrowths of plagioclase and pyroxene) textures that are qualitatively similar to final textures produced in the rapid (>100ºC/hr) cooling rate experiments. Quantitatively, both plagioclase and pyroxene dendrites are substantially coarser in the natural samples than in the experiments. The qualitative similarities between the textures leads us to conclude that these bands are indeed the result of high cooling rates expected from the initial cooling of pahoehoe lobes (Kesthelyi and Denlinger, 1996).

Somewhat puzzling, however, is the apparent absence of the glassy selvage that characterizes Hawaiian pahoehoe flows. There are two possible (and mutually consistent) explanations for this. The first comes from evidence provided above for the ease of crystallization (difficulty of glass formation) in the Site 989 Unit 1 composition. Data from Figure 6 are replotted in Figure 15 to show the time and temperature corresponding to various degrees of crystallinity. Two bracketing curves are drawn in Figure 15, one delineating the boundary of visible plagioclase appearance, and one for the boundary of solidification (Vc, volume of crystals 0.9 of the total volume V). We have estimated the location of the nose of the curve based on a projection of the data, together with the observation that temperature of the nose is commonly ~0.77 that of the liquidus (Uhlmann et al., 1979, 1981). These data show that cooling times of ~10 s (cooling rates of ~ 20º-30ºC/s) are required for glass formation in this composition. While these rates are rapid, they lie within the range of critical cooling rates (1º-100ºC/hr) determined for glass formation in lunar basalts (Uhlmann et al., 1981).

Kesthelyi and Denlinger (1996) show that glassy rinds on the order of 1-3 mm in thickness form at the surface of pahoehoe lobes at critical (for glass formation) cooling rates of 10º-30ºC/s. As cooling rates are limited by the rate of radiative cooling from the flow surface, if substantially higher cooling rates are required for glass formation (e.g., 100ºC/s), then no parts of the flow should experience rapid enough cooling for glass formation, and all parts will solidify (Kesthelyi and Denlinger, 1996). Thus, we anticipate that the flow top glass on individual Unit 1 flows was likely to have been thin (a few mm) and easily abraded. Additionally, the diffuse upper boundaries of the finely crystalline zones may reflect rapid successive emplacement of flow lobes. This would have the effect of both resorbing or reheating and recrystallizing flow top glass, and, perhaps, of coarsening individual dendrites just below the glassy flow surface.

Finally, as discussed earlier, the internal textures of the individual flow units are intersertal, a classification based on the presence of 10%-40% mesostasis (Duncan, Larsen, Allan, et al., 1996). Although the mesostasis is now composed of clay minerals, it is likely that it was glassy before hydrothermal alteration. Based on this observation, it is concluded that the time scales of cooling were less than those required for complete solidification of the flow. Figure 16 gives a histogram of individual flow thicknesses (based on data provided in Duncan, Larsen, Allan, et al., 1996) showing that >70% of the individual flows are 1 m thick. Based on the cooling rate measurements of Hon et al. (1994), and assuming that the lower solidification front of an individual flow lobe propagates at a rate ~50% of that of the upper solidification front, we estimate that the time for cooling of individual flows, ~1 m thick, would be on the order of 10⁴ s. This is somewhat longer than the maximum time for partial solidification (10³-10⁴ s) shown in Figure 15.

Based on the experimental data provided above, we conclude that the quenched (intrafasciculate) textures characteristic of the 989 Unit 1 flow tops, the diffuse upper flow boundaries, and the intersertal (mesostasis-rich) nature of the individual Unit 1 flows are consistent with the original interpretation of Unit 1 as a compound (multiple) flow composed of numerous individual flows that erupted and chilled rapidly and finally cooled to ambient conditions as a single unit. Three possible origins for such a compound flow are a series of overflows from an active lava channel, a near vent shield, or a compound breakout from a distal lava tube. We feel that either of the last two interpretations are possible, with the third perhaps most likely.

Open lava channels typically produce flows with aa morphologies and finely crystalline textures. The observed intersertal textures are distinctly different from those typical of aa channel lavas in Hawaii (K.V. Cashman, unpubl. data) and of inferred aa lavas from cores drilled at Site 990 (e.g., Bucker et al., Chap. 5, this volume), thus making the first option unlikely. It would also be unusual for a lava channel overflow deposit to reach the extreme thickness of Unit 1 (i.e., >69 m). Finally, pipe vesicles, such as those located at the flow base, are not seen in aa flows. The presence of pipe vesicles at the flow base suggests emplacement on a near-horizontal slope, perhaps consistent with either a lava shield or a tube breakout at a break in slope. The low vesicularity (<15%) (Duncan, Larsen, Allan, et al., 1996) argues against accumulation near the vent, as Hawaiian near-vent lavas have vesicularities closer to 70%, whether tube-fed (Cashman et al., 1994) or channelized (Lipman and Banks, 1987). Thus, we are left with the conclusion that this unit may represent multiple break-outs from a distal tube. For such a circumstance to lead to a thickness of 69 m would seem to require that the breakout occur at a break in slope, perhaps to form a lava fan delta (such as that formed at the base of Hilina Pali on Kilauea Volcano). This interpretation still allows the original hypothesis that the progressive decrease in individual flow thickness (Fig. 4) results from an exponentially waning eruption rate.

More generally, we note that Unit 1 is unique among flows observed in other basalt cores. Thick flows are rare in Hawaiian drill cores, with a maximum flow thickness of <35 m recorded for a single inflated pahoehoe flow in the distal Hawaiian Scientific Drilling Project core (Katz, 1997). Thick flows are also uncommon in Northeast Atlantic hardrock core, with only one aa flow (of an evolved basaltic composition) identified in excess of 50 m in thickness (Larsen, Saunders, Cliff, et al., 1994). This is also true for flows recovered at Site 642 on the Vøring margin (Leg 104). Thick flows are common in the Columbia River Basalt Group where lava ponded in the Pasco Basin or was channelized through the ancestral Columbia River Gorge, but these flows are usually simple in form, with one or two identifiable flow units (e.g., Reidel and Hooper, 1989). Unusually thick flows (50-60 m) also occur in the onshore flood basalt succession of East Greenland, but, as for the Columbia River Basalt Group, these flows are interpreted as simple flows of aa, although some may be of the inflation type (Hon et al., 1994). Thus the uniqueness of the Unit 1 flow may reflect unusual pre-eruption geometries required for its formation.