The Atka volcanic center, the largest modern, central Aleutian magmatic complex, covers 360 km(2) with an estimated eruptive volume of similar to200 km(3). The oldest exposed rocks, the Old Harbor Series, are a suite of basaltic andesite to dacite flows, pyroclastic rocks and lahars of presumed Tertiary age. The youngest phase of activity on Atka probably began 1-2 Ma ago near the center of the present volcanic field, and consists of basaltic and basaltic andesite flows which produced a large shield volcano, Atka Volcano. Subsequent to this activity, satellite vents formed along the margin of this structure. Collapse of the ancestral Atka Volcano produced a caldera 5 km in diameter and was accompanied by the eruption of a large dacitic flow (Big Pink). After caldera formation, four major volcanic cones (Korovin, Mount Kliuchef, Sarichef and Konia) grew around the margins of Atka caldera. This study focuses on the Atka Volcano, Korovin and Mount Kliuchef volcanic centers. Atka lavas range from basalt through dacite, with all major volcanic vents erupting lavas spanning similar compositional ranges. They are porphyritic with phenocrysts of plagioclase (25-36%), olivine (0.4-3.6%) or orthopyroxene (2.8%), clinopyroxene (1-12%) and a single Fe-Ti oxide (0-2%). Hydrous phases are lacking even in the most evolved lavas. Two generations of plagioclase occur in the Korovin suite but not in lavas from Mount Kliuchef. Korovin plagioclase also displays resorption surfaces and inclusion zones which are either absent in the phenocrysts of Mount Kliuchef lavas or very rare. Similarly, clinopyroxene in Korovin has reaction rims and abundant glass inclusions which are absent in the Mount Kliuchef lavas. Lavas erupted in the Atka volcanic field vary from 47 to 66% SiO2, although the complex is dominated volumetrically by basaltic lavas. On major-element Harker diagrams, mafic and intermediate Atka Volcano lavas form linear trends but the most siliceous lavas do not fall along these trends. By contrast, Mount Kliuchef lavas are compositionally bimodal whereas Korovin samples range from 51 to 62 wt% silica with intermediate lavas clustering around 55% SiO2. Incompatible trace-element abundances of Korovin lavas do not increase with silica content but those of Atka Volcano and Mount Kliuchef do. On incompatible-incompatible element diagrams, the three suites behave very differently. Differences also exist between the REE characteristics of the three Atka suites. The petrographic and geochemical differences between the three suites suggest formation by markedly different crustal magmatic processes. The characteristics of the Atka Volcano suite suggest formation of the basalt to dacite suite by crystal fractionation of an anhydrous phenocryst assemblage. Incompatible-incompatible element trends suggest the mafic Mount Kliuchef lavas were also produced by crystal fractionation, a conclusion supported by quantitative mass balance calculations. The difference in slope between the Atka Volcano and Mount Kliuchef suites indicates, however, that there were significant differences in these fractionation schemes. Modeling as well as geochemical characteristics suggest that the Mount Kliuchef dacites are not the product of crystal fractionation. They may represent (1) crustal melts, (2) complete melts of siliceous lava produced during earlier magmatic stages, or (3) residual liquids from the Atka Volcano magmatic system. In contrast to these suites, the Korovin lavas could not have been formed by crystal fractionation. Rather the andesites are probably the result of mixing b The Atka data suggest: (1) large volcanic centers may be supplied by a series of non-communicating, crustal magma chambers; (2) local factors may be important in determining liquid-lines-of-descent; and (3) very different magmatic processes may occur in adjacent magma chambers. If the complexity of the Atka suite is a common feature of arc magmatic plumbing systems, reconstructing differentiation processes requires analytical sample sets which are well constrained spatially and temporally.
