Komatiites and sedimentary rocks sampled during the International Continental Drilling Program (BARB1-2–3-4–5) in the Barberton greenstone belt, South Africa, were analyzed for 146Sm-142Nd systematics. Resolved negative l142Nd values (down to 7.7 ± 2.8) were identified in komatiites from the 3.48 Ga Komati Formation and this signature correlates with low Hf/Sm ratios measured in these samples. The negative l142Nd point to a source with subchondritic Sm/Nd ratio which formed during the Hadean. No analytically resolvable 142Nd anomalies were measured in crustal detritus-rich, Si-rich, Ca-Fe-rich sediments and cherts from the Buck Reef (3.42 Ga) and the Fig Tree Group (3.23–3.28 Ga). Our new measurements are incorporated into a larger set of 147Sm-143Nd and 176Lu-176Hf data to better understand the 142,143Nd-176Hf isotope signatures in the mantle source at the time of komatiite crystallization. Our calculations show that the 142,143Nd-176Hf isotope signatures and Hf/Sm ratios cannot be produced by recycling into the komatiite source of detrital sediments like those sampled in the Barberton area. Only cherts have the required trace element characteristics – low Hf/Sm, radiogenic e176Hf –but the trace element concentrations in the cherts are so low that unrealistic amounts of chert would need to be added. We propose a four-stage model for the formation of these rocks. Negative l142Nd and low Hf/Sm ratios developed during the crystallization of a deep magma ocean soon after Earth accretion. The material that ultimately became the source of komatiites was a residual liquid produced by 50% crystallization leaving a bridgmanite/ferropericlase/Ca-perovskite cumulate. The Lu-Hf and Sm-Nd isotope systems were decoupled at this stage. After extinction of 146Sm around 4 Ga, parent/daughter ratios fractionated during a melt extraction event. With this model we explain the positive e176Hf and slightly negative e143Nd in these samples. The 3.55 Ga Schapenburg komatiites in another part of the Barberton belt share similar chemical signatures, supporting our model of fractionation in a deep magma ocean early in Earth’s history.