We faced this question in gold exploration in the Abitibi Mobile Belt. The subduction of pre-existing rocks led to gold-enriched volcanic intrusions/extrusions that were either Na-rich or K-rich. The higher grade gold was associated with the K-rich magmas. Presumably the sediments or other rocks(?) that were subducted were rich in K, but we had no answer.

Thank you Michael,

Yes maybe the subduction of K-rich sediments is more important than what we thought.

The composition of subduction-related magma (but also of magma generated by a subduction-modified mantle in its geological past) strongly depends on the composition of the thin veneer of sediments recycled with the bulk of the slab (altered oceanic crust + lithospheric mantle).

The composition of GLOSS (Global Subducting Sediments) shows the same inter-elemental fractionation recorded in magmas generated in present-day subduction settings. This means that what is transferred to the mantle depths is the main responsible for the final composition of subduction-related melts. Of course the composition of the recycled sediments depend on the composition of the eroded continental crust, but some general features can be outlined. One of these considers the fact that K is strongly concentrated into the upper crust (average 2.87 wt%) compared to the lithospheric mantle (0.05 wt%).

The typical fingerprints of subducting sediments (and of upper continental crust) are strong positive peaks at K and Pb in primitive mantle-normalized diagrams, the same features observed for typical subduction-related magmas.

The higher incompatibility of K vs. Na is another important factor, but it is not the main one. If so, you should expect always potassic/ultrapotassic compositions also in within-plate settings, which is very rare.

Thank you Michele for your comprehensive answer.

Kamchatka is the typical subduction zone with three volcanic belts subparallel to the trench from volcanic front to back arc setting. The high-K rocks exist in all three belts, including volcanic front (Duggen et al., 2007; Nishizawa et al., 2017). These rocks are quite rare in all belts, but they exist.

In addition, it was shown by numerous research, that Kamchatka subduction zone almost free of subducted sediments (e.g. Kersting & Arculus, 1995; Turner et al., 1998) and crust assimilation.

Unfortunately, I cannot say that I know the reason for that for the moment. During 1975-1976 Tolbachik eruption both middle-K and high-K series were erupted. And during Tolbachik massive history (Late-Pleistocene-recent time) both series were overlap.

Our major element simple calculations on the Tolbachik middle-K and high-K rocks show that it is possible to explain both series due to different crystallization conditions - water-rich melts produce middle-K trend and water-poor melts produce high-K trend (Churikova et al., 2015a). So probably different P-T-H2O-fO2 conditions can be responsible for the origin of high-K rocks in subduction setting.

Thanks Tatiana for answering my question.

When it comes to the factors that influence the alkaline rocks (either sodic and potassic) geochemistry several items such as contribution of sediments, pressure under which melting occurs, the composition of the source are suggested. Although there are lots of papers on the alkaline rocks but surprisingly just a few papers deal with the reasons that for example why partial melting of mantle generates an potassic magma and not a sodic one or vice versa. One of the references which I found suggests that regardless of the metasomatism agent, partial melting of the phlogopite bearing mantle sources produces the potassic and ultrapottasic melts while the sodic melts are formed through melting of the amphibole-bearing mantle sources (Rosenthal et al 2009). However I think there are examples which contradict this statement...

In many arcs, K-contents of the magmas increases with distance from the trench meaning also depth of the subducted slab. High K is often explained by (a) lower degrees of partial melting and/or (b) breakdown of phlogopite. Regarding (a): Na contents seems to be buffered by Na in pyroxene (for example in sodic pyroxene of eclogite) whereas all or most K goes into the melt.

Differences in subducted sediment composition may also play a role role: clay-rich vs. clay poor.

Finally: most if not all subducted related magmas are enriched in LILE and in a way, K is also a LILE (large and lithophile), and like other LILE it does not fit into te crystal structures of mantle minerals. Furthermore, as soon as water transport is involved (dehydration of the slab!), LILE's are relatively mobile and end up in the melts.

Dear Ulrich, thank you so much, I really liked your answer!

Dickinson (1975, Geology, v. 3, no. 2, p. 53-56) showed a scatter plot of K20 concentrations (at 57.5% SiO2) versus distance from the trench at the top of the inclined seismic zone in several intra-oceanic and continental-margin magmatic arcs. K20 (57.5) increases with increasing distance from the trench in both oceanic and continental-margin arcs. At every distance, however, K20 is higher for continental-margin arcs than for intra-oceanic arcs.

The general increase in K20 vs distance in both oceanic and continental-margin arcs may be related to progressively lower degrees of partial melting with increasing depth, or to the breakdown of phlogopite in the slab (as suggested in Ulrich Knittel's answer). It could also be related to partial melting within or above a slab that is progressively water-depleted with increasing depth (as suggested in Tatiana Churokova's answer).

The higher K2O of continental-margin arcs (relative to intra-oceanic arcs) may be related to subduction of sediments eroded from the continent, as suggested in Michele Lustrino's answer. In addition, continent-margin-arc magmas may assimilate some K from the continental crust through which they rise.

We suggest that high K2O is signature of the involvement of high CO2-CO fluid in the generation or modification of subduction magmas. Indeed such fluid dissolve Au well so association of K2O rich magmas and Au deposits is casual.

In my group we have some unpublished data that can prove this connection directly. Hope that they will be publically demonstrated soon.