The Paleocene flood basalts of the Faroe Islands form a central part of the North Atlantic Igneous Province, but have proven difficult to date because of very low-grade burial metamorphism in the chabazite-thomsonite to the laumontite zeolite zone. We present 17 replicated K/Ar and 8 Ar/Ar whole-rock analyses of basalts from the >3 km thick lower basalt formation, the age of which has been debated for years. Samples are from the massive core of thick, exposed flows, and two boreholes (Vestmanna-1 and Lopra-1). Six samples are drill cuttings. Extensive microprobe work and mass balance calculations show that roughly 60% of the potassium of the dated basalts resides in plagioclase, interstitial cryptocrystalline rhyolite and smectitic clay, the rest mainly forming thin rims of alkali feldspar on plagioclase. Six basalts fulfil the following criteria: (1) they are almost homogeneous in K and Ar (ages on different splits vary by <4 Ma); (2) the only low-temperature phase present is smectite (saponite±minor interstratified chlorite-smectite); and (3) max. c. 6% of total K occurs in smectite. This smectite replaced metastable interstitial glass during early burial and has a trivial effect on measured ages. The six basalts give mean K/Ar whole-rock ages of 56.5 ± 1.3 to 58.9 ± 1.3 Ma (1σ), which are interpreted as igneous ages consistent with mapped palaeomagnetic reversals and unpublished Ar/Ar dates. They suggest that the oldest drilled lavas erupted at 58.8 ± 0.5 Ma (1σ) in the later part of magnetochron C26r accumulating at >2 km Ma⁻¹, and that the volcanism came to a slow end at 56.4 ± 0.5 Ma in the beginning of chron C24r. The Lower Basalt Formation is overlain by 10m of coal-bearing sediments and 2 km syn-breakup lavas, deposited in early C24r (>55 Ma).

The remaining 11 basalts are either inhomogeneous, carry > 10% of the total K in clay, have C/S > saponite or contain traces of zeolites, secondary quartz, dioctahedral smectite or celadonite as probable indicators of prolonged alteration, and they give low or variable K/Ar ages. The Ar/Ar analyses include five of these poor samples plus three of the first group. They give a plateau age of 55.7 ± 0.9 Ma (1σ) for two exposed flows assigned to chron C25n, but of 60–63 Ma for six drilled lavas assigned to C26r. We argue that the later Ar/Ar ages are too high due to ³⁹Ar recoil loss out of the sample (0–25%) or relocation during irradiation and should be ignored.