Crystal Clusters, Feldspar Glomerocrysts, and Magma Envelopes in the Atascosa Lookout Lava Flow, Southern Arizona, USA: Recorders of Magmatic Events

Abstract

The Atascosa Lookout trachyandesite lava flow is the youngest and most compositionally primitive unit in the middle Tertiary Atascosa–Tumacacori–Cerro Colorado volcanic complex (ATCC). The flow hosts a variety of objects of contrasting origin, including (1) clusters of plagioclase ± chromian diopside, magnesian augite, quartz, hornblende, and orthopyroxene; (2) amoeboid-shaped quartz-bearing enclaves; (3) plagioclase crystals with a concentric interior zone of small melt inclusions (dusty plagioclase); (4) plagioclase crystals with cores filled with large melt inclusions (honeycomb plagioclase); (5) plagioclase glomerocrysts. The groundmass of the trachyandesitic flow is trachydacite. Some crystal clusters, enclaves, and plagioclase glomerocrysts are surrounded by diffuse envelopes of trachydacite higher in K and Mg and lower in Si than the trachydacitic groundmass of the flow. This envelope material is interpreted as foreign magma that engulfed these objects as it invaded their host magma. Both the crystal clusters and plagioclase glomerocrysts may be the remains of cumulate crystal layers, disrupted by influxes of magma into their reservoirs. Crystals in the lava flow originated in at least three distinct magmas and their hybrids. The groundmass of the lava flow preserves evidence for repeated infusion of envelope magma into the system. These influxes fueled the invasion of crystal clusters, plagioclase glomerocrysts, enclaves, and swirls of the envelope magma into the groundmass of the Atascosa Lookout lava flow. Despite the compositional and textural variety apparent in the lava flow, the magmas involved in its development may have been genetically closely related. The collection of features in the lava flow resulted from the development of compositional layers in the magma, accumulation of crystal-rich horizons, disturbance of the system by repeated magma influx, and minor crustal assimilation.

INTRODUCTION

Phenocryst assemblages and magmatic enclaves in igneous rocks preserve information about the evolution of magmatic systems. The presence of a diversity of phenocryst types, as well as textural features such as rimmed or embayed phenocrysts, has been interpreted to represent evidence for interaction of compositionally and thermally distinct magmas during the history of magmatic systems (Gerlach & Grove, 1982; Davidson et al., 1990; Stamatelopoulou-Seymour et al., 1990; Kawamoto, 1992; Davidson & Tepley, 1997; Clynne, 1999). The present study evaluates the petrogenetic significance of multiphase clusters of crystals, droplet-shaped quartz-bearing enclaves, two texturally contrasting populations of melt inclusion-bearing plagioclase, and plagioclase feldspar glomerocrysts in the Atascosa Lookout lava flow, a middle Tertiary trachyandesite from the Atascosa Mountains of southern Arizona. Detailed textural, compositional, and isotopic investigation of these features indicates that interaction of isotopically similar magmas, possibly differing from one another only in having undergone contrasting amounts of crustal assimilation and fractional crystallization, can produce an array of textures and compositions that might otherwise be attributed to hybridism between magmas of strongly contrasting origin.

Recent refinements of models for the evolution of magmatic systems have focused on the development of compositional zonation resulting in layered magma chambers (e.g. Hildreth, 1979; Michael, 1983; Baker & McBirney, 1985), the dynamics of withdrawal of magma from zoned systems (e.g. Spera et al., 1986; Trial et al., 1992), the compositional effects of eruptions on zoned magmas (Byrd & Nash, 1993), the physical and compositional consequences of influxes of mafic magma near the base of magma chambers (e.g. Sparks et al., 1977; Eichelberger, 1980; Dorais et al., 1990; Pallister et al., 1992; Cioni et al., 1995; Wiebe, 1996), and the effects of vesiculation during rise of magma batches (e.g. Eichelberger, 1980, 1995; Koyaguchi & Blake, 1989). This study seeks to use textural and mineral compositional data, along with isotopic ratios of crystals and of chilled magmas, to examine a very heterogeneous lava flow, which is a record of a part of the lifespan of a shallow, intermediate to silicic, magma system. On the basis of its stratigraphic position, its composition relative to those of preceding volcanic deposits, and the range of mineral varieties and their textural relationships, the Atascosa Lookout lava flow is suggested to represent magma that occupied the lower reaches of a magma system. The flow provides an opportunity to integrate concepts of magma chamber zoning, influx of magma into zoned chambers, the response of minerals to disequilibrium conditions, and the relative timing of magmatic events.

GEOLOGIC BACKGROUND

Tectonic and structural setting

The Atascosa, Tumacacori, and Cerro Colorado Mountains of southern Arizona (Fig. 1) host a thick succession of middle Tertiary andesitic to rhyolitic rocks ranging in age from ∼27·7 Ma to ∼23·3 Ma (based on ⁴⁰Ar/³⁹Ar geochronology from this study). The Atascosa–Tumacacori–Cerro Colorado volcanic complex (ATCC) is one of several middle Tertiary intermediate to silicic volcanic centers in southern Arizona, New Mexico and Colorado, related to the transition from convergent orogenesis to back-arc extension (Christiansen & Lipman, 1972; Elston, 1984; Gans et al., 1989; Leeman & Fitton, 1989; Lipman & Glazner, 1991; Bove et al., 1995). Christiansen & Lipman (1972) and Leeman & Fitton (1989) correlated this transition with ridge–trench collision and development of the San Andreas transform boundary. Back-arc extension throughout the western USA was associated with the development of metamorphic core complexes and the explosive eruption of silicic to intermediate magma from several caldera complexes (e.g. Elston, 1984; Ratte’ et al., 1984; Bryan, 1995). In southern Arizona, extension resulted in metamorphic core complex development, intrusion of ‘anorogenic’ granites and subvolcanic intrusions in the Baboquivari Mountains west of the ATCC (Fig. 1) (Goodwin & Haxel, 1990), and in the generation of the ATCC, Chiricahua, and Peloncillo explosive volcanic complexes (Bryan, 1995). ATCC volcanism began around 28 My ago, just before the detachment phase of metamorphic core complex development in the Baboquivari Mountains (Goodwin & Haxel, 1990).

The ATCC is composed of slightly alkaline, silica-saturated rocks, typical of felsic volcanism associated with continental extension. Nd isotopic analyses from this study (the focus of a future paper) indicate that the initial ϵ_Nd of the Atascosa Lookout lava flow is approximately −1·5. DePaolo et al. (1992) compiled ϵ_Nd data from rocks of crustal and mantle derivation in the southwestern USA. In the Baboquivari Mountains, due west of the Atascosa Mountains, rock that they used to represent a crustal component has an ϵ_Nd of −12; rock representing the mantle component from the Baboquivari Mountains has an ϵ_Nd of 1. In view of these regional values, the low negative ϵ_Nd of the Atascosa Lookout lava flow is consistent with a strong contribution of mantle material to the flow.

Volcanic stratigraphy

All of the middle Tertiary rocks in the Atascosa, Tumacacori, and Cerro Colorado Mountains are part of the Montana Peak or Atascosa Formations, named and described by Webb & Coryell (1954) in the Atascosa Mountains. Both formations are heterogeneous in terms of rock types and mineral assemblages. The Montana Peak Formation has been mapped in areas adjacent to the Atascosa and Tumacacori Mountains and correlated with rocks of the ATCC (Nelson, 1963; Knight, 1970; Riggs & Busby-Spera, 1990). The Montana Peak Formation directly overlies the shaly Oro Blanco member of the sedimentary Bisbee Formation. It consists of andesitic and dacitic ignimbrites and lava flows. Its uppermost deposit is a distinctive red andesitic agglomerate that underlies the Atascosa Formation throughout the Atascosa Mountains. The Atascosa Formation overlies the Montana Peak Formation, and consists of a complex succession of rhyolitic to rhyodacitic ignimbrites and lava flows (Fig. 2). The Atascosa Lookout lava flow, the youngest preserved unit in the Atascosa Mountains and Tumacacori Mountains, is the focus of the present study. It is a trachyandesite, and has the most primitive composition in the Atascosa Formation succession, with the exception of basalt flows that cap the Cerro Colorado Mountains, and may be significantly younger than the rest of the rocks of the complex.

The Atascosa and Tumacacori Mountains are essentially contiguous. The Cerro Colorado Mountains are separated from the northern tip of the Tumacacori Mountains by a plain, ∼30 km wide, formed by Basin and Range faulting (Fig. 1). The boundary between the Atascosa and Tumacacori ranges consists of a structurally complex, narrow, deep topographic valley. The Atascosa Mountains, on the south side of the valley, are dominated by ignimbrites totaling >300 m in thickness, but consisting of several mineralogically and texturally distinctive units, rather than a single thick caldera-fill tuff (Fig. 3). The rocks of the Tumacacori Mountains, on the north side of the valley, consist mostly of voluminous lava domes (Seaman et al., 1995), overlain by at least eight ignimbrites, which were deposited over 250–500 ky following the extrusion of the lava. Despite the dominance of ignimbrites in the Atascosa Mountains and lava flows in the Tumacacori Mountains, a case cannot be made at present for the presence of classic caldera features such as thick caldera-fill ignimbrite, caldera walls, and moat deposits. Absence of these features suggests that the accumulation of volcanic strata in the complex resulted from episodic eruptions that did not lead to a single catastrophic collapse of the roof of a large magma chamber.

Geochronology

⁴⁰Ar/³⁹Ar dating of hornblende and sanidine from several ignimbrites and lava flows constrains the timing of eruptions in the three mountain ranges. Samples were prepared and isotopic ratios measured in the New Mexico Geochronology Research Laboratory under the supervision of William McIntosh and Matthew Heisler, using methods and instruments described by Gregory & McIntosh (1996). All errors are reported at the 1σ level.

Volcanism began by about 26·00 ± 0·11 Ma (laser fusion, sanidine) in the Cerro Colorado Mountains, slightly later in the Tumacacori Mountains (25·51 ± 0·26 Ma; isochron, hornblende), and persisted for at least 2·1 My. The youngest middle Tertiary rock in the complex, with the possible exceptions of basalt flows that cap the Cerro Colorado range, is the Atascosa Lookout lava flow, the subject of this study (23·88 ± 0·09 Ma; isochron, based on 20 hornblende crystals).

THE ATASCOSA LOOKOUT LAVA FLOW

Whole-rock analyses of the Atascosa Lookout lava flow indicate that the bulk composition of the flow is trachyandesite. However, the aphyric groundmass of the flow is trachydacite, on the basis of averages of broad beam (10 μm) microprobe analyses (Table 1, Fig. 4). Abundant crystal clusters, plagioclase glomerocrysts, enclaves, and plagioclase and hornblende phenocrysts account for ∼20 vol. % of the lava flow, and drive the whole-rock composition to lower overall silica content. Envelopes of aphyric chilled magma, contrasting in color and composition with the groundmass, are referred to in this study as ‘magma envelopes’. They typically surround crystal clusters and in some instances surround plagioclase glomerocrysts and enclaves. The following sections present descriptions of constituents of the Atascosa Lookout lava flow.

DESCRIPTION OF THE ROCKS

The trachydacitic groundmass

The trachydacitic groundmass of the Atascosa Lookout lava flow is brown in plane-polarized light. It consists of micron-sized plagioclase crystals and brown glass. Plagioclase and hornblende are the only abundant crystals in the lava flow (a total of ∼20 vol. % of flow), occurring both as independent phenocrysts and in crystal clusters. Clinopyroxene and quartz phenocrysts are rare in the groundmass, but both are common in crystal clusters. Five to 10 small (to 0·05 mm) groundmass clinopyroxene crystals per thin section are typical; more than one quartz phenocryst (to 1 mm) in a thin section would be unusual. Quartz phenocrysts are embayed and have a thin rim of very fine clinopyroxene. Plagioclase crystals occur in a variety of sizes. The smallest (∼0·05–0·1 mm long) are elongate and typically show albite twinning. Larger plagioclase crystals are blocky (0·1–3·0 mm long) and display well-developed, generally plateau-style zoning [terminology of Blundy & Shimizu (1991)]. Hornblende crystals occur in a range of sizes (0·1–2·0 mm long), are green to brown to red–brown, and typically have black rims with lower Si and Mg concentration and higher Al and Fe concentration than the hornblende inside the rims. These rims are discussed in more detail in a later section.

With close inspection it is possible to identify indistinct zones slightly grayer in color, swirled through the groundmass (Fig. 5). The grayer zones host up to 15 vol. % very small (to 15 μm) square hematite crystals. These zones are similar in appearance and mineral assemblage to the magma envelopes. The swirled zones are too small for statistically significant quantitative analyses, but X-ray compositional mapping indicates that they have high K, high Mg, and low Si relative to the composition of the lava flow groundmass. These characteristics also distinguish the magma envelopes that most commonly surround crystal clusters, and suggest that the swirled glass and the magma envelopes represent the same mingled magma.

Crystal clusters

Clusters of crystals are homogeneously distributed through the Atascosa Lookout flow. Two or three crystal clusters typically occur in a thin section (Fig. 6a). They contain from approximately five to 100 quartz, clinopyroxene, orthopyroxene, plagioclase, and/or hornblende crystals (Fig. 6b). The clusters range from <1 mm to ∼1 cm in diameter. The crystals in the clusters are ∼0·5–5 mm in their longest dimension. Virtually all of the clusters contain plagioclase crystals, which generally have high-calcium cores with low-amplitude oscillatory zoning and low-calcium rims (Table 2). Other minerals typically found in the clusters are hornblende, diopside, chromian diopside (to 0·8 wt % Cr₂O₃), magnesian augite, orthopyroxene (En_66–69), and quartz. Any combination of these minerals can be found together in a single crystal cluster. The combination of chromian diopside or orthopyroxene and quartz in the same crystal cluster is common. Only the quartz crystals are irregularly shaped (Fig. 6c).

Textural relations of margins and interiors of crystal clusters indicate that the clusters are not xenolith fragments. The boundaries between clusters and the surrounding trachydacitic groundmass are not smooth. Crystals from the clusters protrude out into the trachydacitic groundmass (Fig. 6b), and groundmass of the host is present between crystals in some clusters (Fig. 6d). These characteristics demonstrate that the crystal clusters are not rock fragments, but rather were loose aggregates of crystals when they entered the host magma.

Magmatic enclaves

Droplet-shaped, quartz-bearing enclaves, with grains considerably coarser than that of the groundmass, occur throughout the flow. The enclaves are lighter in color than the trachydacite groundmass, and have smooth to lobate margins (Fig. 7). They range from 0·5 to 2·0 cm in diameter. Some inclusions have discernible, but not abundant (to 10 vol. %) glass; in most inclusions, crystals are identifiable at even the highest magnification. Plagioclase occurs in fine-grained (∼0·2 mm), somewhat radiating sprays. Quartz crystals are equally fine grained, irregularly shaped, and are interstitial to the plagioclase crystals. Sparse quartz phenocrysts in the enclaves are highly anhedral, with embayed margins. Ilmenite needles occur throughout the enclaves, and are about the same length as the plagioclase laths. Some enclaves are surrounded by envelopes of the same gray–brown aphyric trachydacitic, blocky oxide-rich glass (magma envelopes) that surrounds many crystal clusters. Enclaves consist of plagioclase (∼45%), quartz (∼25%), ilmenite (∼15%), and vesicles (∼15%). They are trachyandesitic to dacitic (Table 1, Fig. 4). Although they contain quartz, and the groundmass of the flow does not, the enclaves contain 3–5 wt % less SiO₂ and ∼3 wt % less total alkalis than the groundmass of the Atascosa Lookout lava flow (Table 1).

The coarse-grained texture relative to the host, and the composition of these enclaves (only slightly less silicic than their host), make it at first seem difficult to interpret these objects as chilled magmatic enclaves because chilled magmatic enclaves are typically very fine-grained relative to their host, and their composition is considerably less silicic. The contrast in liquidus temperatures between hotter, less silicic magma droplets, and cooler, more silicic host magma, leads to rapid crystallization of magmatic enclaves. All of the enclaves observed in the Atascosa Lookout lava flow are amoeboidal and droplet shaped. The shape of these bodies is the only evidence that they represent crystallized magma droplets, but in the absence of any angular enclaves, it is difficult evidence to refute. The small difference in temperature between the magma droplets and the host would account for the coarser than typical grain size of these enclaves; rather than chilling and rapidly growing very small crystals, slower cooling would have allowed for the growth of larger crystals. The droplets might have been significantly larger when they were incorporated into the host, and might now represent partially assimilated remains of larger droplets. Alternatively, rather than magma droplets, the enclaves might have been mushy segregations of crystals when they were incorporated into the host, although it seems unlikely that a crystal segregation body would take on the smooth-edged droplet shape of the enclaves. Given the constraints imposed by the shape of these bodies on their origin, and their slightly less silicic composition than their host, the enclaves are interpreted in this work as magmatic droplets that crystallized in the present host.

Plagioclase phenocrysts and plagioclase glomerocrysts

Five textural types of plagioclase crystals are present in the Atascosa Lookout lava flow. The first type of plagioclase, honeycomb plagioclase phenocrysts, has central zones with large melt inclusions (Fig. 8a). The second type, dusty plagioclase phenocrysts, contains fine melt inclusions that can occur in the interior or near the rim of the crystals (Fig. 8b). The third type, clear plagioclase phenocrysts, has no inclusions. The fourth textural type of plagioclase, plagioclase laths, occurs in the groundmass; these crystals are ∼0·02–0·1 mm long. Finally, plagioclase glomerocrysts are aggregates of a few to several plagioclase phenocrysts (Fig. 8c). They are referred to as ‘plagioclase glomerocrysts’ throughout this paper, following Hogan (1993), to distinguish them from the multiphase crystal clusters. The plagioclase crystals that constitute the glomerocrysts are generally honeycomb plagioclase. Glomerocrysts have lobate to serrate internal interlocking grain boundaries with no quenched melt along crystal boundaries. Zoning and twinning in plagioclase are typically truncated by other plagioclase crystals in the glomerocryst. Large round melt inclusions are commonly present in glomerocryst interiors, and a thin, continuous, inclusion-free plagioclase rim surrounds each glomerocryst.

Magma envelopes

Most crystal clusters, and some enclaves and plagioclase glomerocrysts, are rimmed by a 0·5–1·0 mm envelope of fine-grained, wispy, gray–brown material (Fig. 9a). This envelope material contains abundant fine, blocky hematite pseudomorphs after magnetite (Table 3). The envelopes will be referred to as ‘magma envelopes’ throughout this paper; the material composing the envelopes will be referred to as ‘envelope magma’ or ‘envelope glass’, and that composing the groundmass of the lava flow as ‘groundmass magma’ or ‘groundmass glass’ (now devitrified). Compositional mapping of the magma envelopes indicates that they are more potassic, more magnesian, and less silicic than the groundmass of the lava flow (Fig. 9b–d). Quantitative microprobe analyses of envelopes and of the groundmass of the lava flow corroborate differences in K, Mg, and Si concentrations between the two materials (Fig. 4, Table 1).

Textural summary

In summary, any model for the origin of the Atascosa Lookout flow must include an explanation of the following features:

1. a diverse collection of crystals, crystal clusters, and chilled magmas are present in the Atascosa Lookout lava flow.

2. Crystal clusters are not rock fragments; crystal margins, rather than broken rock edges, protrude into the surrounding trachydacitic groundmass (Fig. 6b).

3. Mineral phases in crystal clusters are texturally distinct. Quartz crystals are irregularly shaped (Fig. 6c). Plagioclase, orthopyroxene, and clinopyroxene crystals are fresh and euhedral.

4. Crystal clusters are commonly surrounded by an envelope of trachydacite higher in K and Mg, and lower in Si than the trachydacitic groundmass (Fig. 9); enclaves are less typically surrounded by a magma envelope.

5. Wispy patches of oxide-rich, high-K, high-Mg, low-Si trachydacite also occur throughout the trachydacitic groundmass (Fig. 5).

6. Fine-grained, droplet-shaped trachyandesitic to dacitic enclaves are aphyric with lobate margins against the surrounding trachydacite groundmass (Fig. 7), indicating that they formed by liquid–liquid mixing.

7. Plagioclase glomerocrysts are aggregates of multiple interlocking crystals. The interiors of some of the glomerocrysts are inclusion filled (Fig. 8c), but the interiors of other glomerocrysts contain remnant, inclusion-free cores of individual crystals. The rims of the glomerocrysts are unembayed. The boundary between the interior inclusion-filled zones and the rims of the glomerocrysts are identical in shape to the margin of the complex, but corners are rounded and indistinct.

8. Honeycomb plagioclase crystals have cores dominated by large melt inclusions, and thin, continuous rims; dusty plagioclase crystals have a concentric zone of fine melt inclusions separating inclusion-free cores and outer zones.

9. Fe–Ti oxide minerals in the envelope magma and in the swirled patches of gray groundmass are square, suggesting that these oxides crystallized as magnetite.

MINERAL COMPOSITIONS

Microanalysis of minerals was carried out using the Cameca SX50 electron microprobe in the Department of Geosciences at the University of Massachusetts, under the direction of Michael Jercinovic. This microprobe is equipped with four wavelength-dispersive spectrometers for quantitative analysis and a PGT-IMIX energy-dispersive spectrometer for rapid qualitative analysis. For instrument control and implementation of quantitative analyses, the Cameca SX50 operating system was run in the Solaris environment on a Sun Sparc-20 computer. Corrections for differential groundmass effects were made on-line using the PAP procedure (Pouchou & Pichoir, 1984a, 1984b). Mineral analyses of grids of points in the largely crystalline enclaves were carried out using a focused 15 kV, 15 nA beam with counting times per element ranging from 20 to 40 s. Groundmass and magma envelope analyses were carried out on grids of points at 15 kV and 5 nA, with a 20–40 s count time. Areas for grid analyses were chosen on the basis of absence of crystals. However, the groundmass of the Atascosa Lookout lava flow is devitrified, and micron-sized plagioclase crystals are an abundant and unavoidable component of the groundmass. Inevitably, averaging of several hundred groundmass points (Table 1) includes analyses of plagioclase microlites, which are a crystalline groundmass component.

Plagioclase feldspar

The five textural types of plagioclase crystals place constraints on interpretations of crystallization and disequilibrium epochs in the Atascosa Lookout lava flow. Compositional characteristics of each feldspar type (Fig. 10), and implications for magma evolution, are presented in this section.

Honeycomb plagioclase

Honeycomb plagioclase crystals contain a lower-calcium interior zone (∼An_40–45) with large melt inclusions, mantled by a higher-calcium (∼An_50–60), euhedral rim, which itself grades to lower calcium concentration (∼An₄₅) (Fig. 11a and b). Thin magma envelopes surround some honeycomb plagioclase crystals, and indicate that crystals were in contact with envelope magma before incorporation into the present groundmass magma.

Dusty plagioclase

Dusty plagioclase crystals are compositionally complex. Some dusty plagioclase crystals contain an inclusion-free, higher-calcium interior zone (∼An₅₉) that grades to a lower-calcium zone (∼An₃₂) rimmed by a melt inclusion laden ‘dusty’ zone (∼An_40–50), finally mantled by a rim that grades from higher to lower calcium concentration (∼An_60–30) (Fig. 11b). Other dusty plagioclase crystals have no inclusion-free interior, just a core filled with fine melt inclusions. Some dusty plagioclase crystals have a very thin high-calcium outermost rim. Magma envelopes have not been observed in association with dusty plagioclase crystals.

Clear plagioclase

Plagioclase phenocrysts without melt inclusions are also common in the Atascosa Lookout lava flow. They are euhedral, with flat zoning profiles (∼An₅₅) from their cores to near their rims. Their thin outermost rims typically show increased An concentration. Because clear plagioclase phenocrysts host no melt inclusions, are strongly euhedral, and are not surrounded by magma envelopes, they are interpreted to have grown in the present trachydacitic groundmass magma after the events that led to the development of dusty and honeycomb plagioclase crystals.

In addition to occurring as phenocrysts, clear plagioclase crystals are abundant in the crystal clusters. These crystals generally have more anorthitic cores, which show fine oscillatory zoning across a small compositional range (An_55–60), and a thin less anorthitic rim (∼An₄₅), suggesting a more complex history than that of the clear plagioclase phenocrysts.

Plagioclase laths

Plagioclase laths in the groundmass of the Atascosa Lookout lava flow (An_48–52) are generally comparable in composition with the outer zones of clear plagioclase phenocrysts. They are interpreted to have grown later than the clear phenocrysts, as the groundmass lava cooled. Calcium-richer rims that occur on some of the laths may have grown from decompression of the magma during its ascent to the surface.

Plagioclase glomerocrysts

Plagioclase crystals in glomerocrysts typically have the honeycomb texture, indicating that the component crystals of the glomerocrysts were resorbed either during or before the event(s) that sutured the component crystals together to make the glomerocrysts. Component crystals of the glomerocrysts have cores of composition similar to that of the cores of honeycomb plagioclase (An_65–55). Glomerocrysts have thin, optically continuous high-calcium (∼An₆₀) rims that surround the entire glomerocryst.

Summary: plagioclase crystals

Honeycomb plagioclase crystals have a lower-calcium interior zone, filled with large melt inclusions. This zone is surrounded by inclusion-free plagioclase of similar composition. All honeycomb plagioclase crystals have a higher-calcium inclusionless rim. Dusty plagioclase crystals have a higher-calcium interior zone that grades to a lower-calcium zone, surrounded by a zone of fine melt inclusions. The crystals grade to lower calcium concentrations toward their rims. Contrast in compositional zoning of the honeycomb and dusty plagioclase crystals suggests that honeycomb plagioclase crystals nucleated and completed their early stages of growth in less calcic magma than that in which dusty plagioclase crystals developed. Both varieties of plagioclase underwent resorption, as indicated by melt inclusions.

Clear plagioclase phenocrysts show little zoning and no resorption. They are interpreted to have grown after the resorption episode(s) recorded in the honeycomb and dusty plagioclase crystals, or in a different magma altogether. In contrast, clear plagioclase crystals in crystal clusters have higher-calcium cores that grade to lower-calcium concentration, with an overprint of weak oscillatory zoning, and thin low-calcium rims. These crystals are interpreted to have grown in more calcic magma than that in which the clear plagioclase phenocrysts grew. Plagioclase laths have composition similar to that of the outer zones of the clear plagioclase phenocrysts, and are interpreted to have grown synchronously with the outermost rims on the plagioclase phenocrysts.

Plagioclase glomerocrysts are composed of honeycomb plagioclase crystals with zoning similar to that described above. They are interpreted to have formed by some process that caused the suturing of honeycomb plagioclase crystals contemporaneously with, or after, the resorption episode that generated the honeycomb texture.

Pyroxene

Crystal clusters can contain either orthopyroxene (En_66–69) or clinopyroxene, or both (Table 4). Orthopyroxene crystals are blocky, large (to 0·3 mm), and show no zoning. They are much less ubiquitous than clinopyroxene crystals in clusters, and occur in 5–10% of all clusters. Clinopyroxene crystals are most abundant in crystal clusters, but sparse, small (to 0·2 mm) clinopyroxene crystals occur throughout the trachydacite groundmass. Clinopyroxene also occurs as fine-grained crystals rimming rare, strongly embayed quartz phenocrysts. In crystal clusters, clinopyroxene is magnesian augite, diopside, or chromian diopside (Table 4, Fig. 12). The most notable characteristic is the high concentration of chromium (to 0·8 wt % Cr₂O₃) in some crystals. Clinopyroxene crystals with high concentrations of Cr₂O₃ are bright grass green in plane-polarized light, compared with the darker, duller green of the chromium-poor crystals. Chromian clinopyroxene is richer in MgO (MgO 15·5 to >17·0 wt %) compared with Cr-poor clinopyroxene (MgO 14·5–15·0 wt %). The cores of diopside and chromian diopside in crystal clusters are typically slightly richer in Al₂O₃ than rims, and slight oscillatory zoning in Al₂O₃ concentration occurs in the core region. Sector zoning has not been observed in any clinopyroxene crystals in the lava flow. Orthopyroxene and chromian clinopyroxene crystals are the most magnesian, compositionally primitive crystals in the lava flow, and have been identified only in the crystal clusters.

Compositions of small clinopyroxene crystals in the groundmass of the Atascosa Lookout lava flow, and of clinopyroxene that rims quartz phenocrysts, contrast with those of clinopyroxene in the crystal clusters. Small clinopyroxene crystals in the groundmass of the flow contain less SiO₂ than crystals in clusters. However, they span a significant compositional range (Table 4). Small clinopyroxene crystals that rim embayed quartz phenocrysts (Fig. 13) have much higher SiO₂ and lower Al₂O₃ concentrations than either groundmass clinopyroxene or cluster clinopyroxene crystals (Table 4).

Hornblende

Hornblende is the only abundant ferromagnesian phase in the groundmass of the Atascosa Lookout flow. Most of the hornblende crystals are dull olive green to green–brown, and invariably have rounded margins and thick black rims. Hornblende crystals in crystal clusters and in the groundmass are indistinguishable in appearance and composition (Table 5). They show no discernible compositional zoning. They are transitional between tschermakitic and hastingsitic hornblende, on the basis of the classification of Giret et al. (1980).

Fe–Ti oxide minerals

Electron microprobe analyses of Fe–Ti oxides, and their recalculation and allocation of Fe²⁺ and Fe³⁺ indicate that Fe–Ti oxides are members of the hematite–ilmenite series and are predominantly hematite (Table 3). Small (<0·05 mm) oxide minerals in the magma envelope glass and in the gray wispy patches that occur throughout the groundmass are distinctive in terms of their blocky, cubic shapes. The crystal form of these oxides suggests an origin as magnetite, and later oxidation to hematite. The prevalence of the cubic crystal form of oxides in the envelope and in the patches in the groundmass may indicate that the envelope magma, from which these oxides crystallized, was less oxidized than either the Atascosa Lookout lava flow magma or the enclave magma.

STRONTIUM ISOTOPIC DATA

Strontium isotopic ratios were measured in several components of the Atascosa Lookout lava flow, including plagioclase glomerocrysts, chromian clinopyroxene from crystal clusters, trachydacitic groundmass, trachyandesitic to dacitic enclaves, and magma envelopes. All Sr isotopic analyses were performed in the Isotope Geochemistry Laboratory in the Department of Earth and Space Sciences at the University of California at Los Angeles (UCLA), under the supervision of Jon Davidson and Peter Holden. A microdrill was used to extract clean rim material from single plagioclase glomerocrysts, and to sample enclaves, magma envelopes, and groundmass of the lava flow. Clinopyroxene crystals were separated from their groundmass as whole crystals. Sr concentrations of glomerocrysts and clinopyroxene crystals were determined by isotope dilution at UCLA. The Sr concentration of the groundmass was determined by X-ray fluorescence spectrometry at the University of Massachusetts. Sr concentrations of enclaves were determined by electron microprobe by averaging concentrations determined with long count times (200 s per analysis) and high current (100 nA). Rb concentrations of all materials were determined by laser ablation inductively coupled mass spectrometry under the direction of Susan Keydel and Dula Amarasiriwardena in the Department of Natural Sciences at Hampshire College. Results of these analyses are shown in Table 6, and initial Sr isotopic ratios (at 23·8 Ma) are plotted against Sr concentration in Fig. 14.

The initial Sr ratios of the groundmass of the lava flow are higher than those of enclaves, clinopyroxene crystals, plagioclase glomerocrysts, or magma envelope material. Cores of plagioclase glomerocrysts have higher initial Sr ratios than their high-calcium rims, and the core isotopic composition of one of the glomerocrysts is virtually identical to that of the enclaves. It should be pointed out that cores of plagioclase glomerocrysts do not necessarily correspond to the cores of the honeycomb plagioclase crystals that make up the glomerocrysts. Hence Sr isotopic ratios of cores of the three glomerocrysts analyzed cover a range of values, probably reflecting analysis of different parts of the component crystals. Glomerocryst rims have initial Sr ratios similar to those of magma envelopes. Initial Sr ratios of clinopyroxene crystals are slightly lower, but similar to those of the magma envelopes.

The similarity in isotopic ratios of clinopyroxene in crystal clusters, plagioclase in glomerocrysts, and groundmass, enclaves, and envelopes is permissive of a broad genetic relationship between all of these materials, consistent with their coexistence in a single lava flow. If the magmas represented by the groundmass, the enclaves, and the magma envelopes are all products of partial melting of a single parent composition, differences in their initial ratios are likely to be a result of assimilation of crust. Reasonable candidates for crustal assimilant in south–central Arizona are Jurassic and Proterozoic plutonic rocks. Present-day ⁸⁷Sr/⁸⁶Sr of samples of these rocks are approximately 0·71278 and 0·78546, respectively (Farmer & DePaolo, 1984). Jurassic basement contains approximately 199 ppm Rb and 218 ppm Sr, and has ⁸⁷Rb/⁸⁶Sr = 1·665; Proterozoic basement contains approximately 221 ppm Rb and 160 ppm Sr, and has ⁸⁷Rb/⁸⁶Sr = 3·983 (Farmer & DePaolo, 1984). Assimilation of even the Jurassic country rock at 23·8 Ma (⁸⁷Sr/⁸⁶Sr_23·8Ma ∼ 0·71222) could have increased the strontium isotopic ratio of the envelope magma. A simple mass balance calculation indicates that if the analyzed Jurassic country rock were the assimilant, assimilation of country rock with envelope magma in a country rock–envelope magma ratio of 0·028:0·972 would be sufficient to raise the strontium isotopic ratio of the envelope magma to the average initial ratio of the enclaves (0·707814). Assimilation of Jurassic country rock with enclave magma, in a country rock–envelope magma ratio of 0·016:0·984, could raise the initial ratio of the enclaves to that of the groundmass of the Atascosa Lookout lava flow. Assimilation of Proterozoic country rock (⁸⁷Sr/⁸⁶Sr_23·8Ma ∼ 0·78411) in a country rock–envelope magma ratio of 0·002:0·998 would raise the initial ratio of the envelope magma to that of the enclaves. Assimilation of Proterozoic country rock with enclave magma in a country rock–enclave magma ratio of only 0·001:0·999 would raise the initial ratio of the enclave magma to that of the groundmass. Hence, very little assimilation of material representing reasonable country rock composition would be necessary to slightly diversify the strontium isotopic ratios of the magmas represented in the Atascosa Lookout lava flow.

GEOTHERMOMETRY AND GEOBAROMETRY

The variety of textural relationships and, in particular, evidence for textural disequilibrium between minerals in the Atascosa Lookout lava flow complicate the use of geothermometry and geobarometry. Because hornblende and plagioclase are ubiquitous in the Atascosa Lookout lava flow, and because euhedral crystals of each can be analyzed, the hornblende–plagioclase geothermometer (Spear, 1980, 1981; Nabelek & Lindsley, 1985; Blundy & Holland, 1990; Holland & Blundy, 1994) was used here, and was applied to unresorbed groundmass hornblende–plagioclase pairs, as well as to hornblende and plagioclase crystals in crystal clusters. Cation abundances and Fe²⁺/Fe³⁺ ratios were determined following the method of Robinson et al. (1982).

For any choice of cluster and groundmass crystals, phenocrysts in the groundmass produce temperatures lower than those calculated from cluster crystals, although absolute uncertainties would certainly overlap. The average temperature calculated for hornblende–plagioclase pairs from the groundmass of the Atascosa Lookout lava flow is 801°C (n = 12), with a range from 795 to 811°C. The average temperature calculated for hornblende–plagioclase pairs from the crystal clusters is 838°C (n = 19). The range of temperatures calculated from cluster crystal pairs is larger (823–862°C).

DISCUSSION

First-order observation of the array of minerals in the Atascosa Lookout lava flow suggests that some of the minerals (i.e. chromian clinopyroxene, orthopyroxene, and quartz) are not compatible with an origin in the present groundmass of the lava flow. Evidence for the presence of more than one magma in the flow, and contrasting initial strontium isotopic ratios reinforce the impression that some components of the flow are exotic to the present groundmass.

Plagioclase crystals, which exhibit a variety of textures and occur in various subenvironments in the lava flow, are particularly essential in interpreting the history of the lava flow. The first part of this discussion focuses on plagioclase feldspar crystals, then their relationships to other phases represented in the lava flow are considered. The Atascosa Lookout lava flow hosts two types of plagioclase crystals (honeycomb and dusty) that contain melt inclusions. Compositional zoning, as well as differences in the size and position in the crystal of the inclusions, indicate that these types of crystals had different histories. Several workers (Kuno, 1950; Dungan & Rhodes, 1978; Gerlach & Grove, 1982; Nelson & Montana, 1992) have suggested that honeycomb plagioclase forms as a result of partial dissolution. Others have suggested that it is a skeletal texture, formed as a result of rapid growth during undercooling of a magma (Kuo & Kirkpatrick, 1982; Anderson 1984; Kawamoto, 1992). In the Atascosa Lookout lava flow, the higher An rims on honeycomb plagioclase make it difficult to suggest an origin for these crystals by rapid cooling in a thermally perturbed, supposedly cooler and less anorthitic magma. Rather, partial dissolution may have occurred, before An-rich rim growth, as a result of (1) influx of hotter, more anorthitic magma, (2) ascent of the magma to lower pressure, or (3) influx of H₂O into the magma chamber, stabilizing more anorthitic plagioclase. Although the other two origins cannot be ruled out, physical evidence exists, in the form of magma envelopes and swirled envelope magma in the groundmass, for an influx of hotter, less silicic magma into the groundmass magma of the Atascosa Lookout lava flow.

Plagioclase glomerocrysts are aggregates of honeycomb plagioclase crystals. Their origin provides some constraint on the history of the flow. Hogan (1993) presented a model for the formation of plagioclase glomerocrysts involving partial resorption of closely spaced crystals. He suggested that resorption along boundaries of the grains produces an intergranular melt, which crystallizes either slowly to produce overgrowths on existing crystals, or quickly to preserve a chilled melt boundary between subgrains in the crystal. Following Hogan (1993), plagioclase glomerocrysts in the Atascosa Lookout lava flow are interpreted to have developed by resorption of closely spaced plagioclase crystals and development of boundary zone plagioclase melts. Subsequent cooling and crystallization sutured the crystals together into glomerocrysts. The boundary melts apparently cooled slowly enough to produce plagioclase overgrowths because no intercrystalline glass has been observed in the glomerocrysts. The cores of the component plagioclase crystals in glomerocrysts contain less calcium than the rims, similar to the cores of honeycomb plagioclase crystals that are not part of glomerocrysts. Cores of the glomerocrysts also have higher initial ⁸⁷Sr/⁸⁶Sr than their thin continuous higher-calcium rims. Both plagioclase glomerocrysts and honeycomb plagioclase crystals have a thin, continuous outer high-calcium plagioclase rim that in some cases has an outermost selvage of more sodic plagioclase (Fig. 11a). Calcium zoning and isotopic variation in these crystals are consistent with the origin of honeycomb plagioclase crystals in a relatively less calcic, higher ⁸⁷Sr/⁸⁶Sr magma than that in which their rims crystallized.

In contrast to the honeycomb plagioclase crystals, the dusty plagioclase crystals have high-calcium cores that grade to lower calcium concentrations inside a zone of fine melt inclusions (Fig. 11b). Dusty plagioclase has been interpreted as partially melted relatively Ab-rich plagioclase (Kuno, 1936, 1950; Larsen et al., 1938; MacDonald & Katsura, 1965; Sakuyama, 1979; Tsuchiyama, 1985; Bloomfield & Arculus, 1989; Kawamoto, 1992). This interpretation is consistent with compositional zoning in dusty plagioclase crystals in the Atascosa Lookout lava flow. The higher-calcium innermost core of dusty plagioclase crystals indicates that these crystals originated in a relatively calcic magma, compared with the magma in which the honeycomb plagioclase crystals originated. The abrupt transition to more sodic plagioclase, still in the core of the dusty plagioclase, is consistent with crystallization from a less calcic magma. The subsequent development of the dusty texture seems to suggest that after the crystallization of less calcic plagioclase, the magma was perturbed by influx of calcic magma, depressurization, or dehydration.

The composition of clear plagioclase crystals (An_50–55) is similar to the composition of outermost zones of dusty plagioclase crystals (Fig. 10), perhaps indicating that the rims of dusty plagioclase crystals crystallized under the same conditions as the clear plagioclase. However, clear plagioclase crystals can be large (to 2 mm) and the rims on dusty feldspar crystals are only to 0·3 mm thick, indicating that growth of matrix crystals outpaced precipitation of plagioclase onto existing plagioclase crystals. The thin higher-calcium outermost rinds on matrix plagioclase crystals may be a result of magma cooling followed by decompression associated with rise of the magma to the eruption site.

The high calcium concentrations of plagioclase crystals from crystal clusters are consistent with an origin in a magma more calcic than the present groundmass. Hybridization with a less calcic magma could account for the zoning to lower calcium composition that is typical of these crystals. Clinopyroxene crystals (some rich in chromium), and orthopyroxene crystals, coexist with plagioclase crystals in crystal clusters. Compositions of these crystals are also consistent with their origin in more calcium-rich, primitive magma than that represented by the present groundmass of the lava flow. In contrast, the quartz that occurs with pyroxene and calcic plagioclase in the crystal clusters, as well as hornblende, is not likely to have crystallized from the same magma that gave rise to those minerals. The quartz and hornblende crystals may be phenocrysts that resided in a more silicic magma that mixed with the native magma of the plagioclase and pyroxene crystals.

Compared with clinopyroxene crystals in clusters, clinopyroxene crystals in the groundmass of the Atascosa Lookout lava flow contain less SiO₂ (Table 4). Although silicon concentration tends to decrease in pyroxene with decreasing temperature at constant pressure (Thompson, 1974), silicon activity in the magma also affects silicon concentration in pyroxene (Deer et al., 1978). Hornblende–plagioclase geothermometry indicates that hornblende and plagioclase in the matrix of the Atascosa Lookout lava flow crystallized at lower temperature than the same minerals in crystal clusters, and implies that groundmass clinopyroxene crystallized at a lower temperature than clinopyroxene in crystal clusters. Lower crystallization temperature would be expected to favor less silicon in groundmass clinopyroxene rather than the reverse. This may suggest that Si activity was the controlling factor on pyroxene composition, and that the magma from which the groundmass clinopyroxene crystallized (presumably represented by the present lava groundmass) had higher Si activity than that from which the clinopyroxene crystals in the crystal clusters grew. Clinopyroxene crystals that rim large, irregularly shaped quartz phenocrysts (Fig. 13) have still higher SiO₂ concentrations than the groundmass clinopyroxene crystals. They probably grew by reaction of the quartz with surrounding magma.

Hornblende is ubiquitous in the Atascosa Lookout lava flow. It occurs as a phenocryst phase in the groundmass of the flow, and is abundant in crystal clusters. In all settings, hornblende crystals are surrounded by thick opaque rims. Rutherford & Hill (1993) examined hornblende crystals with rims consisting of small plagioclase, pyroxene, and Fe–Ti oxides from rocks of Mount St Helens. They suggested that the rims resulted from reaction of the hornblende crystal with the melt, with the reaction triggered by (1) decrease in water concentration in the magma, (2) influx of higher-temperature or less hydrous magma, or (3) fluxing of the magma with CO₂-rich fluid. The black rims surrounding hornblende crystals in the Atascosa Lookout lava flow are finer grained than those examined by Rutherford & Hill. Electron microprobe analysis of the dark rims yields compositions similar to those of crystal interiors, but lower in Si, Mg, Ca, and K, and higher in Al and Fe than fresh hornblende. These characteristics, except for depletion in Ca, are consistent with the reaction suggested by Rutherford & Hill (1993), in which amphibole reacts with melt, enriching the melt in Si, K, and Na, and depleting it in Al, Fe, and Ca. The ubiquity of the rims on hornblende crystals, both in clusters and in the groundmass of the lava flow, indicates that at least some hornblende crystals formed in cooler, more silicic magma before hybridization. Rim growth would be consistent with dehydration of the magma associated with ascent to the surface, or with influx of hotter magma into the chamber.

A MAGMA CHAMBER MODEL

Textural and compositional characteristics of crystal clusters, single plagioclase crystals, plagioclase glomerocrysts, and enclaves can be integrated into a model of a dynamic magma system that accumulated the residue of a variety of magmatic processes. The model presented here is only one of several possible based on the data available from the flow, but it illustrates the constraints imposed on evolution of the system by data derived from the flow.

Calculations of the density of magmas represented by the groundmass, the enclaves, and the magma envelopes, using the method of Bottinga et al. (1982), indicate that the envelope magma had about the same density as the groundmass magma, and that the enclave magma was the densest of the three magmas represented in the lava flow (Table 1). One possible model to account for the textures and compositions of features of the Atascosa Lookout lava flow would involve at least two influxes of envelope magma into the bottom of a magma chamber initially consisting of a layer of denser enclave magma overlain by a layer of groundmass magma (Fig. 15a). Plagioclase crystals destined to become honeycomb and glomerocryst plagioclase could have originated in the less calcic enclave magma; those destined to become dusty plagioclase could have originated in the more calcic envelope magma that invaded the chamber. Hybridization of these two magmas (Fig. 15b) would have had the following effects on the crystals in the less calcic enclave magma: (1) resorption of single plagioclase crystals to produce honeycomb texture; (2) resorption and suturing together of plagioclase crystals in cumulate layers to produce plagioclase glomerocrysts. In the more calcic invading envelope magma, hybridization would cause calcic plagioclase crystals destined to become dusty plagioclase to develop the abrupt drop in calcium concentration shown in their cores.

Quartz and hornblende crystals (from the enclave magma, which contains quartz in the groundmass and as phenocrysts), and clinopyronene and orthopyroxene crystals (from the envelope magma) may have accumulated in heterogeneous crystal mats in the hybrid magma. Further influx of envelope magma (Fig. 15c) may have disrupted these mats and partially melted the crystal accumulations, and, with cooling, clusters of crystals may have been sutured together with partial melts of their component crystals. This influx could also have caused the partial dissolution of the high-calcium cored plagioclase crystals that originated in the envelope magma, leading to the formation of dusty plagioclase texture. Finally, influx of the buoyant envelope magma may have penetrated the hybrid layer, allowing mingling of the envelope and groundmass magmas, producing the swirled texture preserved in the present lava flow groundmass, and propelling objects from the hybrid layer into the groundmass magma (Fig. 15c). Those objects are commonly surrounded by envelopes of the invading buoyant magma.

The most primitive crystals in the crystal clusters, the chromian clinopyroxene and the orthopyroxene, seem unlikely crystallization products of even the least silicic magma represented in the Atascosa Lookout lava flow (the envelope magma). However, the isotopic similarity between the envelope magma and chromian clinopyroxene crystals supports the origin of the clinopyroxene in a magma isotopically similar to the envelope magma. Fractional crystallization may have driven the evolution of the ancestral envelope magma well beyond its composition at the time of chromian clinopyroxene crystallization. Hence, the orthopyroxene and the chromian clinopyroxene crystals may have formed in a hybrid magma with at least one end-member considerably more mafic than any magma now represented in the lava flow, although the envelope magma may be the liquid remnant of that magma.

This model is consistent with quartz and at least some of the hornblende crystals forming in a different, more silicic magma than the one in which the chromian clinopyroxene and orthopyroxene crystals formed. It also accounts for the contrasting zoning profiles of honeycomb and dusty plagioclase, as well as the occurrence of magma envelopes around some objects. It is consistent with density contrasts between the three magmas, and with the isotopic data.

SUMMARY AND CONCLUSIONS

The Atascosa Lookout lava flow is a sample of a deep portion of a long-lived, predominantly felsic magmatic system. The flow preserves evidence of crystal accumulation, possible crustal assimilation, and more than one influx of hotter magma. Plagioclase glomerocrysts and crystal clusters are evidence for development of cumulate layering. Magma envelopes are interpreted to be remnants of the invading magma in which the dusty plagioclase crystals began to crystallize. Enclaves are evidence of the preservation of silicic, viscous, crystal-rich magma in a host only slightly more silicic.

Despite the variety of crystal textures and types of magma represented in the Atascosa Lookout lava flow, the isotopic similarity of crystals in clusters, crystals in glomerocrysts, groundmass glass, and enclave glass, suggests that the crystals formed in similar magmas, and that the magmas are partial melts of similar source rocks. The contrast in crystallization temperature between hornblende–plagioclase pairs in crystal clusters and the same phases in the groundmass (∼50°C) may support the hypothesis that crystals in the clusters, perhaps remnants of cumulate layered crystals, crystallized before the last hybridization of their host with a cooler magma.

Crystal clusters, enclaves, and plagioclase glomerocrysts document at least a part of the history of a complex magma system that gave rise to several explosive and effusive eruptions. The interpretation of components of even this single lava flow has proven to be difficult and involves a variety of analytical techniques. The model offered is undoubtedly only one of many permitted by the data available. Textural and compositional variety similar to that observed in the Atascosa Lookout lava flow are somewhat common in intermediate rocks, and may indicate that multiple hybridization events between closely related magmas are a common factor in the evolution of intermediate magmas. The remote relationship between intermediate to silicic magmatic rocks produced in shallow crustal settings and their lower-crustal and mantle ancestry underscores the importance of integrating mineral compositions, glass compositions, textural relationships, and isotopic data in the interpretation of their origin and evolution.

I thank Michael Clynne, Jon Davidson, Ron Vernon, Dave Wark and Mike Williams for thorough and helpful reviews of this manuscript. I am grateful to John Hogan, Chris Knowlton, Jim McLelland, John Reid, Frank Ramos, LeeAnn Srogi, Dave Wark and Mike Williams for discussions, and to Dula Amarasiriwardena, Jon Davidson, Kurt Hollocher, Peter Holden, Mike Jercinovic, Sue Keydel, Bill McIntosh and Dave Wark for opening their laboratories to this work. Mike Vollinger’s careful sample preparation is gratefully acknowledged. National Science Foundation Grant EAR-9406576 partially supported this research.

REFERENCES