Zircon is widely used to simulate melt generation, migration and evolution within the crust and mantle. The achievable performance of melt modelling generally depends on the availability of reliable trace element partition coefficients (D). However, a large range of D_REE values for zircon from natural samples and experimental studies has been reported, with values spanning up to 3 orders of magnitude. Unfortunately, a gap of knowledge on this variability is evident. In this study we model the crystallization processes of common REE-bearing minerals from granitic melts and show that the measured zircon D_REE would be elevated if there is crystallization of REE-enriched minerals subsequent to zircon. Nevertheless, compared to zircon D_REE values measured from experimental studies, this mechanism appears to have a less significant influence on those from natural granite samples since the quantity of crystallized REE-enriched minerals is very low in natural magmatic systems and/or most of them crystallize prior to zircon. Combined with recently published studies, this work supports that analysis of natural zircon/host groundmass pairs provides more robust D_REE values applicable to natural systems than those measured from experimental studies, which can be used to constrain the provenance of detrital zircons.

锆石被广泛用于模拟地壳和地幔内部的熔体生成，迁移和演化。熔体建模的可实现性能通常取决于可靠的痕量元素分配系数（D）的可用性。然而，据报道，来自天然样品和实验研究的锆石的D _REE值范围很大，范围高达3个数量级。不幸的是，关于这种可变性的知识差距是显而易见的。在这项研究中，我们模拟了花岗岩熔体中常见的含REE矿物的结晶过程，并表明，如果锆石之后富含REE的矿物发生结晶，则测得的锆石D _REE将会升高。不过，与锆石D _REE相比根据实验研究得出的数值，这种机理似乎对天然花岗岩样品的影响较小，因为在天然岩浆系统中，富含稀土元素的结晶矿物的数量非常少，并且/或者其中大多数都在锆石之前结晶。与最新发表的研究相结合，这项工作支持对天然锆石/宿主地物对的分析提供了比从实验研究中测得的值更可靠的适用于自然系统的D _REE值，可用于限制碎屑锆石的出处。