Our study began on a very specific date—30 April 2019—and, like many discoveries, it started almost by accident. At the end of an exciting day in the field, a few unexpected observations set us on a path we hadn’t planned. It was the second-to-last day of our spring expedition in a remote area of eastern Nepal. The route we had followed—through rural landscapes and forests of blooming rhododendrons—lay far from the usual trekking trails. For several days we had been walking across phyllites, relatively soft rocks that rarely form good outcrops and tend to weather into dusty paths. So the excitement was real when, that afternoon, the first exposures of more competent rocks appeared. What we found was a thick sequence of calc-silicate gneisses derived from marly sedimentary protoliths. The outcrops extended almost continuously for a couple of kilometres, and because the dirt road cutting through them had only recently been opened, the rocks were exceptionally fresh—quite unusual in a subtropical climate where weathering is typically pervasive.
At the time, we were already working on calc-silicate rocks, which are relatively common in the Himalaya. Yet the structures we observed that day immediately stood out. They closely resembled those of migmatites—crustal rocks that have undergone partial melting—hosting very coarse-grained leucosome pods and veins, i.e., portions crystallized from melt. Such features are typically seen in rocks derived from pelitic or felsic compositions, but we had never observed, nor found described in the literature, similar structures in rocks of marly composition. In our field notebook, we noted that these features pointed to in situ partial melting of the meta-marls: in other words, the melt was internally derived, rather than being introduced from outside.
From the field to the microscope: the “the balls-bearing rock”
Back in the lab, we prepared four thin sections from the collected samples and examined them under the microscope. What we saw was striking. The leucosomes were extremely heterogeneous in both mineral assemblage and composition. Alongside the coarse-grained minerals already visible in the field—clinopyroxene, garnet, and titanite—the matrix revealed something unusual, earning our samples the nickname “the balls-bearing rock.”
The leucosomes consist largely of potassium feldspar (silicate domain), within which rounded aggregates—up to a centimetre in size—are embedded. These “balls” are rich in scapolite and represent domains of (silico-)carbonate composition. Similar domains, but much finer-grained, also occupy distinct interstitial positions within the leucosomes: they consist of K-feldspar for the silicate domain, and of thin films of calcite and/or scapolite for the (silico-)carbonate one. Both domain types preserve the sequential structure typical of rocks crystallized from a melt.
What the microstructures revealed
These microstructural features pointed to the presence of two immiscible liquids: one silicatic and the other (silico-)carbonatic, which crystallized in separate domains within the leucosomes. Particularly intriguing were the thin interstitial films composed of calcite ± scapolite. We interpret these as the final crystallization products of the residual (silico-)carbonatic liquid. Because these domains contain more than 60% calcite by volume, they can be classified as carbonatites, following the definition proposed by Mitchell (2005), who considers a rock to be a carbonatite if it contains more than 30% primary igneous carbonate, regardless of its silica content. Carbonatites are relatively rare magmatic rocks, often strongly enriched in rare earth elements, which makes them especially important from an economic and resource perspective. For this reason, the identification of carbonatitic melts in our “ball-bearing rock” was particularly intriguing—but it also demanded further confirmation.
To test whether the silicatic and (silico-)carbonatic liquids were indeed conjugate phases, we calculated their partition coefficients for major elements. The results were clear: K₂O is strongly partitioned into the silicate liquid, whereas Na₂O—and especially CaO—preferentially enter the carbonate liquid. This explains why Ca- and Na-rich minerals (calcite and scapolite) and K-rich minerals (K-feldspar) crystallized in distinct interstitial domains within the leucosomes. Similarly, MnO, MgO, and FeO are preferentially concentrated in the carbonate liquid, while Al₂O₃ and SiO₂ are enriched in the silicate one. These patterns closely match experimental data obtained for hydrous, silica-rich, potassium-bearing systems, thus confirming our immiscibility hypothesis.
What’s new: a new perspective on carbonatite genesis
Since the 1960s, carbonatites have largely been interpreted as rocks crystallized from mantle-derived magmas formed by low degrees of partial melting of carbonated peridotites. Our results show that this is not the whole story. Partial melting of marly metasediments in orogenic settings can also generate carbonatitic liquids. This crustal model broadens the framework for carbonatite formation, highlighting the role of anatexis of carbonate-bearing sediments in the continental crust. It shows that carbonatitic melts can form without any contribution from mantle-derived magmatism—no ultramafic precursors, no deep mantle melting required. While our findings do not challenge the idea that most carbonatites originate in the mantle, they do question the widespread assumption that crustal anatexis cannot produce true carbonatites. In doing so, they offer a possible explanation for carbonatites found in collisional settings that lack a clear mantle signature.
What’s next
One complication arising from our data is that our (silico-)carbonate liquids are relatively enriched in SiO₂ and depleted in alkalis compared with those produced in experimental immiscibility studies. We believe this reflects the starting compositions used in most experiments, which are typically silica-undersaturated and alkali-rich (especially in Na₂O), unlike the average composition of marly sediments. However, to test this hypothesis, new experiments specifically designed to replicate the composition of marly sediments will be needed. Only then can we fully assess how immiscibility behaves in such systems.
Take home message
One of the most exciting aspects of this work is the reminder that careful petrographic observations still matter. By paying close attention to the microstructures of metamorphic rocks—even using relatively simple and traditional methods—we can uncover entirely new processes and rethink long-standing geological models.