Miocene to Recent volcanism in northwestern Arabia produced the largest intraplate volcanic field on the Arabian plate (Harrat Ash Shaam, Jordan). The chemically and isotopically diverse volcanic field comprises mafic alkali basalts and basanites. The magmas underwent limited fractional crystallization of ol ± cpx ± plag and rare samples have assimilated up to 20% of Late Proterozoic crust en route to the surface. However, there are subtle Sr–Nd–Pb isotopic variations (⁸⁷Sr/⁸⁶Sr = 0·70305–0·70377, ¹⁴³Nd/¹⁴⁴Nd = 0·51297–0·51285, ²⁰⁶Pb/²⁰⁴Pb = 18·8–19·2), which exhibit marked correlations with major elements, incompatible trace element ratios and abundances in relatively primitive basalts (MgO >8·5 wt %), and cannot be explained by fractional crystallization and crustal contamination alone. Instead, the data require polybaric melting of heterogeneous sources. Semi-quantitative melt modelling suggests that this heterogeneity is the result of small degree melts (2–5%) from spinel- and garnet-facies mantle, inferred to be shallow Arabian lithosphere, that mixed with smaller degree melts (<1%) from a predominantly deep garnet-bearing asthenospheric(?) source with ocean island basalt characteristics. The latter may be a ubiquitous part of the asthenosphere but is preferentially tapped at small degrees of partial melting. Volcanism in Jordan appears to be the result of melting lithospheric mantle in response to lithospheric extension. With time, thinning of the lithosphere allowed progressively deeper mantle (asthenosphere?) to be activated and melts from this to mix with the shallower lithospheric mantle melts. Although Jordanian intraplate volcanism is isotopically similar to examples of Late Cenozoic volcanism throughout the Arabian peninsula (Israel, Saudi Arabia), subtle chemical and isotopic differences between Yemen and Jordan intraplate volcanism suggest that the Afar plume has not been channelled northwestwards beneath the Arabian plate and played no role in producing the northern Saudi Arabian and Jordan intraplate volcanic fields.

We present trace element and Sr–Nd–Hf–Pb isotope compositions for clinopyroxenes from anhydrous spinel peridotite and garnet ± spinel pyroxenite xenoliths of Pan-African lithospheric mantle from Jordan, including the first high-precision double-spike Pb isotope measurements of mantle clinopyroxene. Clinopyroxenes from the peridotites are variably Th–U–LILE–LREE enriched and display prominent negative Nb, Zr and Ti anomalies. MREE–HREE abundances can generally be modelled as partial melting residues of spinel lherzolite with primitive-mantle-like composition after extraction of 5–10% melt, whereas the enrichments in Th–U–LILE–LREE require a Pan-African or later metasomatic event. The large range of Nd, Sr, Pb and Hf isotope ratios in both peridotites and pyroxenites (e.g. εNd ∼ 1·4–17·5; ²⁰⁶Pb/²⁰⁴Pb ∼ 17·2–20·4; εHf ∼ 0·6–164·6) encompasses compositions more radiogenic than mid-ocean ridge basalt (MORB), and Pb isotopes cover almost the entire range of oceanic basalt values. εHf values are some of the highest ever recorded in mantle samples and are decoupled from εNd in the same samples. Marked correlations between Sr–Nd–Pb isotopes, LILE–LREE enrichments and HFSE depletion suggest that the metasomatizing agent was a carbonatitic-rich melt and isotopic data suggest that metasomatism may have been related to Pan-African subduction. The metasomatic melt permeated depleted upper mantle (<16 kbar) during Pan-African subduction at ∼ 600–900 Ma, and the variably metasomatized material was then incorporated into the Arabian lithospheric mantle. There is no evidence for recent metasomatism (1·2 Ga) pre-metasomatic Lu–Hf signatures of the depleted upper mantle that was the protolith of the Arabian lithospheric mantle. The ‘resistance’ of the Lu–Hf isotopic system to later metasomatic events resulted in the development of extremely heterogeneous Hf isotopic signatures over time that are decoupled from other isotopic systems. No mantle sample in this study exactly matches the chemical and isotopic signature of the source of Jordanian intraplate basalts. However, the xenolith compositions are broadly similar to those of the source of Arabian intraplate basalts, suggesting that the numerous Cenozoic intraplate volcanic fields throughout Arabia may be the product of melting upper mantle wedge material fertilized during Pan-African subduction and incorporated into the Arabian lithospheric mantle. We propose a model whereby the proto-Arabian lithospheric mantle underwent a major melting event in early Proterozoic–late Archean times (at the earliest at ∼1·2 Ga). Island-arc volcanism and major crust formation occurred during the Pan-African orogeny, which liberated fluids and possibly small-degree melts that migrated through the mantle creating zones of enrichment for certain elements depending upon their compatibility. Immobile elements, such as Nb, were concentrated near the base of the mantle wedge providing the source of the Nb-rich Jordanian volcanic rocks. More mobile elements, such as LILE and LREE, were transported up through the mantle and fertilized the shallow mantle source of the Jordanian xenoliths. Following subduction, the mantle wedge became fossilized and preserved distinct enriched and depleted zones. Lithospheric rifting in the Miocene triggered partial melting of spinel-facies mantle in the lower lithosphere, which mixed with deeper asthenospheric garnet-facies melts as rifting evolved. These melts entrained segments of variably carbonatite-metasomatized shallow lithospheric mantle en route to the surface.