Phosphorus and aluminum zoning in olivine : contrasting behavior of two nominally incompatible trace elements

Abstract

Phosphorus zoning in olivine is receiving considerable attention for its capacity to preserve key information about rates and mechanisms of crystal growth. Its concentration can vary significantly over sub-micron spatial scales and form intricate, snowflake-like patterns that are generally attributed to fast crystal growth. Ostensibly similar aluminum enrichment patterns have also been observed, suggesting comparable incorporation and partitioning behavior for both elements. We perform 1-atm crystallization experiments on a primitive Kīlauea basalt to examine the formation of P and Al zoning as a function of undercooling − ΔT (− ΔT = Tliquidus − Tcrystallization) during olivine growth. After 24 h spent at Tinitial = 1290 °C (10 °C above olivine stability), charges are rapidly cooled to final temperatures Tfinal = 1220–1270 °C, corresponding to undercoolings − ΔT = 10–60 °C (with Tliquidus = 1280 °C). Compositional X-ray maps of experimental olivine reveal that only a small undercooling (≤ 25 °C) is required to produce the fine-scale enrichments in P and Al associated with skeletal growth. Concentration profiles indicate that despite qualitatively similar enrichment patterns in olivine, P and Al have contrasting apparent crystal/melt mass distribution coefficients of KPol/melt = 0.01‒1 and KPol/melt = 0.002‒0.006. Phosphorus can be enriched by a factor > 40-fold in the same crystal, whereas Al enrichment never exceed factors of 2. Glass in the vicinity of synthetic and natural olivine is usually enriched in Al, but, within analytical uncertainty, not in P. Thus, we find no direct evidence for a compositional boundary layer enriched in P that would suffice to produce P enrichments in natural and synthetic olivine. Numerical models combining growth and diffusion resolve the conditions at which Al-rich boundary layers produce the observed enrichment patterns in olivine. In contrast, the same models fail to reproduce the observed P enrichments, consistent with our observation that P-rich boundary layers are insignificant. If instead, P olivine/melt partitioning is made to depend on growth rate, models adequately reproduce our observations of 40-fold enrichment without boundary layer formation. We surmise that near-partitionless behavior (KPol/melt close to 1) of P is related to the olivine lattice being perhaps less stiff in accommodating P during rapid crystallization, and/or to enhanced formation of vacancy defects during fast growth. Our results confirm that P is a robust marker of initial rapid growth, but reveal that the undercooling necessary to induce these enrichments is not particularly large. The near-ubiquitous process of magma mixing under volcanoes, for instance, is likely sufficient to induce low-to-moderate degrees of undercooling required for skeletal growth.