Abstract Diamonds have been brought to the Earth’s surface from at least 2.82 Ga onward by igneous and tectonic processes, and they have been redistributed since then by sedimentary processes into secondary diamond deposits. None of the known tectonically emplaced diamond deposits are economically viable, and only two types of igneous rock, kimberlite and lamproite, sometimes carry diamonds and can occasionally be economic to mine. Where diamonds are present in kimberlite and lamproite, the concentrations are –3 ppm, acquired by random sampling of diamond source rocks in the subcontinental lithosphere. Diamond-forming processes in the lithosphere were episodic since ~3.57 Ga, and all primary diamond deposits show evidence of two or more diamond generations. The earliest diamond-forming episode at ~3.4 ± 0.2 Ga appears to have been a worldwide metasomatic event triggered by CO2-rich, probably subduction-derived fluids that produced diamonds associated with garnet harzburgite. Further diamond populations have formed in association with craton accretion, subduction, slab melting, magmatic modifications of the lithospheric mantle, obduction tectonics, and metasomatic infiltration. In the process, diamonds formed in association with metasomatized harzburgite were supplemented predominantly by metasomatic diamond growth in eclogite, with occasional significant contributions from grospyditic, lherzolitic, and websteritic sources as well as sublithospheric ultradeep sources, notably majorite. Diamond-bearing igneous bodies exploit preexisting zones of weakness in the crust. They probably traverse most of the distance from the mantle to the surface as thin dikes or dike swarms, with nearurface expressions dictated by multiple intrusions, their volatile content, the presence or absence of cap rocks, local structures, and the ambient stress field, interaction with ground water , and degrees of preservation from erosion. Average diamond values per carat for a given diamond occurrence vary by approximately three orders of magnitude (US$1-$1,000/carat). A multistage model for the formation of diamond deposits is presented for the Kaapvaal craton that takes into account the tectonic history of the craton as well as the complexities observed within diamond populations of its various primary diamond deposits. Although the details of this model are craton specific, the general features of the model are applicable to other cratons.

Abstract Natural diamond occurrences are reviewed to show that kimberlites and lamproites, though our principal source rocks for economic diamonds, are not primary deposits in the sense that diamond originated within them. The igneous hosts simply provided the transport medium for diamonds that formed within harzburgitic and eclogitic source rocks in the subcontinental lithosphere. A distinction is made between diamond source rocks and “primary” diamond deposits defined as kimberlites, lamproites, or any other igneous rock type originating deep enough to have sampled the diamondiferous source rocks. “Primary” deposits are then analysed in terms of sources, transport media, and depositional sites. Although sizes and grades among “primary” deposits vary greatly, “giant” deposits that are governed by different geological processes than smaller deposits cannot be identified. However, the enormous dilution in diamond grade between source rocks and “primary” deposits suggests that concentrations of eclogitic source rocks in the upper mantle may represent truly “giant” deposits that would dwarf even the largest and richest “primary” deposits.